Farmed saltwater crocodiles A genetic improvement program

A report for the Rural Industries Research and Development Corporation

by Sally Isberg, Peter Thomson, Frank Nicholas, Stuart Barker and Chris Moran

October 2004 RIRDC Publication No 04/147 RIRDC Project No US-109A

ii

© 2004 Rural Industries Research and Development Corporation. All rights reserved. ISBN 1 74151 055 4 ISSN 1440-6845 Farmed saltwater crocodiles–A genetic improvement program Publication No. 04/147 Project No. US-109A The views expressed and the conclusions reached in this publication are those of the author and not necessarily those of persons consulted. RIRDC shall not be responsible in any way whatsoever to any person who relies in whole or in part on the contents of this report. This publication is copyright. However, RIRDC encourages wide dissemination of its research, providing the Corporation is clearly acknowledged. For any other enquiries concerning reproduction, contact the Publications Manager on phone 02 6272 3186. Researcher Contact Details Associate Professor Chris Moran Centre for Advanced Technologies in Animal Genetics and Reproduction (ReproGen) Faculty of Veterinary Science, Sydney, NSW, 2006 Phone: (02) 9351 3553 Fax: (02) 9351 2114 Email: ChrisM@vetsci.usyd.edu.au

Sally Isberg Centre for Advanced Technologies in Animal Genetics and Reproduction (ReproGen) Faculty of Veterinary Science, Sydney, NSW, 2006 Phone: (02) 9351 3553 Fax: (02) 9351 2114 Email: sisb7192@mail.usyd.edu.au

In submitting this report, the researcher has agreed to RIRDC publishing this material in its edited form. RIRDC Contact Details Rural Industries Research and Development Corporation Level 1, AMA House 42 Macquarie Street BARTON ACT 2600 PO Box 4776 KINGSTON ACT 2604 Phone: 02 6272 4819 Fax: 02 6272 5877 Email: rirdc@rirdc.gov.au. Website: http://www.rirdc.gov.au Published in October 2004 Printed on environmentally friendly paper by Canprint

iii

Foreword Production from farmed crocodile for the skin trade is an emerging industry that began in the mid-1980s. Research to improve the efficiency of crocodile production has so far concentrated on management -related practices, such as nutrition and housing, with a notable exception of research in the area of genetics. For the Australian crocodile industry to continue to develop, genetic improvement programs should be incorporated into farm management practices to simultaneously improve the efficiency of crocodile production. Currently, it takes approximately three years to raise a juvenile crocodile to slaughter size (1.6 – 1.9 metres in total length). By implementing a selection program based on reproductive performance, juvenile growth rates and juvenile survival rates, breeding from superior parents will increase the productivity of commercial crocodile farms. The major benefits to the industry will be decreasing overhead costs by growing animals to marketable size in a quicker time, increasing profitability by offsetting some of the production costs per animal and increasing the number of animals obtained from the farm each year. The major aim of this project was to create a practical genetic improvement program for immediate adoption by the Australian crocodile industry, to be called CrocPLAN. First, this involved defining crocodile breeding objectives in consultation with industry members. Second, data analyses were completed to obtain estimates of relevant genetic and phenotypic parameters (heritability, repeatability, and correlations). Finally, crocodile breeding values (CBVs) were estimated and incorporated into a crocodile economic selection index ($CESI) using the relative economic weights (estimated by fourth year economics student, Ms Emily Gray) for each breeding objective. Recommendations and current limitations for implementation of CrocPLAN are discussed. Furthermore, the assessment of a parentage testing kit using microsatellite markers provides a means for pedigree verification and a method of expanding the genetic improvement program beyond unitised breeding pens. The results presented in this study are from data collected at Janamba Croc Farm, Middle Point, Northern Territory, Australia. The value of these data and the assistance of Janamba Croc Farm should be recognised by the Australian crocodile industry. This project was funded from RIRDC Core Funds which are provided by the Australian Government. This report, an addition to RIRDC’s diverse range of over 1000 research publications, forms part of our New Animal Products R&D program, which aims to accelerate the development of viable new animal industries. Most of our publications are available for viewing, downloading or purchasing online through our website: • downloads at www.rirdc.gov.au/fullreports/index.html • purchases at www.rirdc.gov.au/eshop Simon Hearn Managing Director Rural Industries Research and Development Corporation

iv

Acknowledgments The outcomes reported in this project arose from a collaborative effort between RIRDC, the University of Sydney and Janamba Croc Farm. In particular, without the meticulous data collection of Mr Stuart Barker, manager of Janamba Croc Farm, this study would not have eventuated. Although, Stuart is co-author of this study, we would like to ensure that due acknowledgement is given to him because without foresight such as his, new initiatives do not occur particularly for emerging industries. Associate Professor Chris Moran and Professor Frank Nicholas not only supported this project academically, but used their own consultancy funds to finance a workshop on genetic improvement of crocodiles in Darwin (February, 2002). Thanks must also go to the team at Janamba Croc Farm- Jim (Collin) Young, Wayne Gurney, John Nash, Tracy Saunders and Lindsay Cleary, who have helped over the years in the collection of data, subsequently used in these analyses. Also, the knowledge they distilled about the animals greatly assisted in Sally’s understanding for modelling purposes. Thanks to Professor Grahame Webb and Mr Charlie Manolis for their help, discussions on everything crocodilian, and most importantly, access to their extensive library resource. Thanks also to staff at Wildlife Management International Pty. Ltd., in particular Dr Adam Britton, Boyd Simpson, Jacob Bar-Lev, David Ottway and Brenton Whelan, for their advice, support, and welcome scientific discussion. A special thanks to Professor John James for providing invaluable advice on the unspoken rules of genetic analysis and the functioning of genetic improvement programs. Your wealth of information is unequalled. A special mention to Dr Yizhou Chen, Zung Doan, and Jaime Gongora for guiding Sally through laboratory procedures. The help of the Biometry Section (Kyle Kieffer, Kath Bartimote, and Prof. Mick O’Neill) and David Liu for computing and statistical software was greatly appreciated. Thanks for the support of everyone in the Animal Genetics postgraduate lab, in particular Denbigh Simond for suggesting the usage of CBVs (introduced in Chapter 9). Also, thanks to Ms Emily Gray for estimating the economic values required for this study. With regard to Chapter 8, thanks to Dr Adam Stow and Dr Kyall Zenger for advice on how to proceed with the microsatellite work, and Dr Nancy FitzSimmons and Dr Jake Gratten for their advice and guidance on designing the tissue biopsy punch used to collect the breeder tissue samples. Extended thanks to Mr John Olsen and Mr Björn Isberg for deciphering the schematics, manufacturing and donating the tissue biopsy tool. We gratefully acknowledge the financial assistance of the Rural Industry Research and Development Corporation (RIRDC) and Dr Peter McInnes (RIRDC Research Manager, New Animal Products). Also, Sally would like to thank the University of Sydney and the Faculty of Agriculture, Food and Natural Resources for her scholarship, with an additional acknowledgement to the Faculty of Veterinary Science for providing the resources necessary to complete this PhD. Sally Isberg Chris Moran

v

Abbreviations To minimise repetition, abbreviations were used for the traits analysed in this study. These are defined below, although are also described within the text.

Reproductive traits

ClSize total number of eggs collected in a clutch. NoViable Number of viable eggs in a clutch. NoHatch Number of live, healthy hatchlings in a clutch. HatchR NoHatch as a proportion of ClSize. AvSVL average snout-vent length of hatchlings in a clutch. HDays number of days between hatching date and the 1st of January in that particular year to

indicate time of hatch. Nesting whether the female nested or not in a particular year. Production traits

HHL hatchling head length (in millimetres) HSVL hatchling snout-vent length (in millimetres) HTL hatchling total length (in millimetres) InvHL inventory head length (in millimetres) InvSVL inventory snout-vent length (in millimetres) InvTL inventory total length (in millimetres) CullHL slaughter head length (in millimetres) CullTL slaughter total length (in millimetres) CullBwidth slaughter belly width (in millimetres) Quality traits

SR number of scale rows General

SE Standard error SD Standard deviation CBV Crocodile breeding values $CESI Crocodile economic selection index

vi

Contents Foreword............................................................................................................................................................... iii

Acknowledgments ................................................................................................................................................ iv

Abbreviations ........................................................................................................................................................ v

Executive summary............................................................................................................................................. xii

1. Introduction....................................................................................................................................................... 1

1.1 General introduction ..................................................................................................................................... 1

1.2 The Australian crocodile industry................................................................................................................. 1 1.2.1 The product ........................................................................................................................................... 1 1.2.2 Skin grading system .............................................................................................................................. 2 1.2.3 Export markets ...................................................................................................................................... 3

1.3 General introduction to genetic improvement programs .............................................................................. 3

1.4 Defining breeding objectives ........................................................................................................................ 4

1.5 Selection criteria ........................................................................................................................................... 4

1.6 Estimating breeding values........................................................................................................................... 5

1.7 Relative economic weights ........................................................................................................................... 5 1.7.1 An example of economic values in the Australian pig industry ............................................................ 6

1.8 Molecular genetics: uses in animal breeding ................................................................................................ 7 1.8.1 Microsatellites and parentage determination......................................................................................... 7 1.8.2 Marker assisted selection (MAS) .......................................................................................................... 7

1.9 Objectives ..................................................................................................................................................... 8

2. Resource description and analytical methodolgy ........................................................................................... 9

2.1 Overview of Janamba Croc Farm ................................................................................................................. 9 2.1.1 Breeding stock..................................................................................................................................... 10 2.1.2 Egg collection and incubation ............................................................................................................. 12 2.1.3 Clutch identification marking system.................................................................................................. 13 2.1.4 Rearing hatchlings and juveniles......................................................................................................... 14 2.1.5 Stocking densities and grading............................................................................................................ 16 2.1.6 Feeding................................................................................................................................................ 17 2.1.7 Feeding routine and cleaning............................................................................................................... 17 2.1.8 Disease management ........................................................................................................................... 17 2.1.9 Slaughtering and processing................................................................................................................ 18 2.1.10 Skin grading ...................................................................................................................................... 19

2.2 Structure of family groups .......................................................................................................................... 20

2.3 Traits and observations ............................................................................................................................... 22 2.3.1 Reproductive traits .............................................................................................................................. 22 2.3.2 Production traits .................................................................................................................................. 22 2.3.3 Survival traits ...................................................................................................................................... 24 2.3.4 Quality trait – Number of scale rows................................................................................................... 24 2.3.5 Explanatory variables .......................................................................................................................... 25

2.4 Statistical analysis....................................................................................................................................... 25 2.4.1 The linear mixed model....................................................................................................................... 27 2.4.2 Types of models .................................................................................................................................. 28 2.4.3 Heritability and repeatability estimates ............................................................................................... 29

vii

3. Industry perspectives on breeding objectives for saltwater crocodile genetic improvement programs.. 31

3.1 Abstract....................................................................................................................................................... 31

3.2 Introduction ................................................................................................................................................ 31

3.3 Materials and methods................................................................................................................................ 31

3.4 Results and discussion ................................................................................................................................ 32 3.4.1 Reproductive output ............................................................................................................................ 32 3.4.2 Survival ............................................................................................................................................... 33 3.4.3 Food conversion efficiency ................................................................................................................. 33 3.4.4 Age at Slaughter .................................................................................................................................. 34 3.4.5 Skin grade............................................................................................................................................ 34 3.4.6 Skin quality - a trait of future importance?.......................................................................................... 34

3.5 Implications ................................................................................................................................................ 35

4. Quantitative analysis of reproduction traits ................................................................................................. 36

4.1 Abstract....................................................................................................................................................... 36

4.2 Introduction ................................................................................................................................................ 36

4.3 Materials and Methods ............................................................................................................................... 37 4.3.1 Animals and data collection ................................................................................................................ 37 4.3.2 Statistical methods............................................................................................................................... 38

4.4 Results ........................................................................................................................................................ 40 4.4.1 Univariate models ............................................................................................................................... 40 4.4.2 Multi-trait model and parameter estimates.......................................................................................... 41 4.4.3 Binomial model and repeatability ....................................................................................................... 42

4.5 Discussion................................................................................................................................................... 42 4.5.1 Accounting for female age using proxies ............................................................................................ 42 4.5.2 Relationships between the “egg” traits................................................................................................ 43 4.5.3 Relationship between hatchling size and egg traits ............................................................................. 43 4.5.4 Relationship between time of nesting and other traits......................................................................... 43 4.5.5 Nesting success ................................................................................................................................... 44

4.6 Implications ................................................................................................................................................ 44

5. Quantitative analysis of age at slaughter ...................................................................................................... 45

5.1 Abstract....................................................................................................................................................... 45

5.2 Introduction ................................................................................................................................................ 45

5.3 Methods and materials ................................................................................................................................ 45 5.3.1 Experimental animals .......................................................................................................................... 45

Inventory age (days) ......................................................................................................................................... 48 5.3.2 Rearing environments.......................................................................................................................... 48 5.3.3 Statistical methods............................................................................................................................... 49 5.3.4 Univariate modelling........................................................................................................................... 49 5.3.5 Multivariate modelling ........................................................................................................................ 50 5.3.6 Genetic parameter estimates................................................................................................................ 50

5.4 Results and discussion ................................................................................................................................ 50 5.4.1 Hatchling traits .................................................................................................................................... 50 5.4.2 Inventory traits .................................................................................................................................... 51 5.4.3 Age at slaughter................................................................................................................................... 52 5.4.4 Multivariate modelling and genetic parameter estimates .................................................................... 52

5.5 Implications ................................................................................................................................................ 53

viii

6. Quantitative Analysis of Juvenile Survival ................................................................................................... 54

6.1 Abstract....................................................................................................................................................... 54

6.2 Introduction ................................................................................................................................................ 54

6.3 Methods and materials ................................................................................................................................ 54 6.3.1 Experimental animals .......................................................................................................................... 54 6.3.2 Statistical methods............................................................................................................................... 55 6.3.3 Genetic parameter estimates................................................................................................................ 55

6.4 Results and discussion ................................................................................................................................ 55 6.4.1 Risk factors.......................................................................................................................................... 57

6.5 Implications ................................................................................................................................................ 57

7. Quantitative analysis of scale row number ................................................................................................... 58

7.1 Abstract....................................................................................................................................................... 58

7.2 Introduction ................................................................................................................................................ 58

7.3 Methods and materials ................................................................................................................................ 59 7.3.1 Experimental animals .......................................................................................................................... 59 7.3.2 Methodology for collection of scale row data ..................................................................................... 59 7.3.3 Comparing counting methods.............................................................................................................. 61 7.3.4 Full-sib heritability estimate................................................................................................................ 62 7.3.5 Correlation with hatchling morphometric measurements................................................................... 62 7.3.6 Animal model...................................................................................................................................... 62 7.3.7 Combining the data-sets ...................................................................................................................... 62

7.4 Results and discussion ................................................................................................................................ 63 7.4.1 Comparing methods ............................................................................................................................ 63 7.4.2 Correlation with hatchling traits .......................................................................................................... 63 7.4.3 Heritability estimates........................................................................................................................... 64 7.4.4 Confounding effects of incubation temperature .................................................................................. 64 7.4.5 Possible imprinting effects on scale row number ................................................................................ 66

7.5 Implications ................................................................................................................................................ 67

7.6 Special acknowledgements ......................................................................................................................... 67

8. Analysis of microsatellites and parentage testing in saltwater crocodiles.................................................. 68

8.1 Abstract....................................................................................................................................................... 68

8.2 Introduction ................................................................................................................................................ 68

8.3 Materials and methods................................................................................................................................ 69 8.3.1 Animals and Sampling ........................................................................................................................ 69 8.3.2 Experimental Protocol......................................................................................................................... 69 8.3.3 Microsatellite and Population Genetic Analysis.................................................................................. 69

8.4 Results ........................................................................................................................................................ 71 8.4.1 Confirmation of correct parentage assignment.................................................................................... 71 8.4.2 Simulating a situation where one or neither parent is known.............................................................. 71

8.5 Discussion and implications ....................................................................................................................... 72

8.6 Special acknowledgements ......................................................................................................................... 72

ix

9. Designing and implementing the genetic improvement program: CrocPLAN.......................................... 73

9.1 Abstract....................................................................................................................................................... 73

9.2 Crocodile economic selection index ($CESI)............................................................................................. 73 9.2.1 Estimated crocodile breeding values (CBVs) for the breeding objectives .......................................... 73 9.2.2 Relative economic values .................................................................................................................... 78 9.2.3 The crocodile economic selection index ($CESI) ............................................................................... 79

9.3 Case study: CrocPLAN implementation on Janamba Croc Farm............................................................... 80 9.3.1 Crocodile-specific industry issues to consider .................................................................................... 80 9.3.2 Response to selection .......................................................................................................................... 81 9.3.3 Juvenile selection ................................................................................................................................ 81 9.3.4 Across-herd selection .......................................................................................................................... 82 9.3.5 Applying CrocPLAN on a leading crocodile breeding farm ............................................................... 83

10. Conclusions and recommendations ............................................................................................................. 86

10.1 CrocPLAN................................................................................................................................................ 86

10.2 Areas of further research .......................................................................................................................... 87 10.2.1 Reproductive traits ............................................................................................................................ 87 10.2.2 Food conversion efficiency and growth rate ..................................................................................... 88 10.2.3 Skin grade.......................................................................................................................................... 88 10.2.4 Industry-wide CrocPLAN ................................................................................................................. 88

10.3 Conclusion ................................................................................................................................................ 88

Literature cited.................................................................................................................................................... 89

Appendix I. .......................................................................................................................................................... 94

Appendix II.......................................................................................................................................................... 96

x

List of Tables Table 1.1. A range of prices received for saltwater crocodile skins..............................................................3 Table 1.2. ..............Increment of improvement and economic values for various breeding objectives in the

Australian pig industry.................................................................................................................6 Table 2.1. Farm number of saltwater (C. porosus) and freshwater (C. johnstoni) crocodile numbers at

Janamba Croc Farm ...................................................................................................................10 Table 2.2. A summary of C. porosus breeding at Janamba Croc Farm from 1989 to 2002........................12 Table 2.3. Summary of male and female identification numbers used in this study, along with the year

when the animals were first put together. ..................................................................................21 Table 2.4. Summary of fixed effects and covariates used in the analysis of

the performance traits ................................................................................................................26 Table 3.1. Responses to industry survey.....................................................................................................32 Table 4.1. Crocodilian reproductive traits of economic importance ...........................................................38 Table 4.2. Summary statistics for the reproductive traits used to estimate variance components...............38 Table 4.3. Significant terms for the reproductive traits from univariate REML modelling........................40 Table 4.4. Regression coefficients for egg length (EL) and egg width (EW) produced from univariate

REML analyses..........................................................................................................................40 Table 4.5. Estimates of repeatability (bold on diagonal) and phenotypic correlations (below diagonal)

with their approximate standard errors in parenthesis below each estimate for saltwater crocodile reproductive traits.......................................................................................................41

Table 5.1. Crocodile morphological traits measured as possible selection criterion for age at slaughter ...46 Table 5.2. Summary statistics for the morphometric traits used to estimate

variance components..................................................................................................................46 Table 5.3. Description of terms used to model the traits of interest............................................................48 Table 5.4. Significant terms for the morphometric traits from univariate REML modelling .....................51 Table 5.5. Hatchling trait genetic (above diagonal; standard errors below) and phenotypic (below

diagonal) correlation estimates. .................................................................................................51 Table 5.6. Estimates of heritability (h2), phenotypic standard deviations (σP), genetic (rg; above diagonal)

and phenotypic correlations (rp; below diagonal) for hatchling and inventory morphometric traits and age at slaughter...........................................................................................................52

Table 6.1. Estimates (± SE) of year effects, their hazard ratios and the antilog of the 95% confidence interval using the Cox Proportional Hazards Model..................................................................57

Table 7.1. Summary statistics for the number of scale rows on the belly skin for 3206 crocodiles.............................................................................................................59

Table 7.2. Summary statistics for the comparison between scale row counting methods...........................61 Table 7.3. Simple linear regression coefficients (± SE) describing the difference between the number of

scales on the left and right of the midline using the same counting method (Simple and SimpleR), and the difference between the counting methods (WMI and SimpleR) ..................63

Table 7.4. Genetic (rg) and phenotypic (rp) correlations between number of scale rows and hatchling head length (HHL), snout-vent length (HSVL) and total length (HTL).............................................63

Table 7.5. Estimates of variance components and heritability for the number of scale rows for progeny incubated at 32oC constant temperature.....................................................................................64

Table 7.6. Separate regression equations for male (MSR) and female (FSR) scale row count with the average of their progeny (SR) for the years 1998 to 2001 .........................................................65

Table 8.1. Microsatellite loci trialled on 32 adult C. porosus for use in a parentage determination kit........................................................................................................70

Table 9.1. Raw data for the fate of juveniles produced from the pairs at Janamba Croc Farm...................77 Table 9.2. Relative economic values for crocodile breeding objectives. Abb. is the abbreviated term for

the respective breeding objective. ..............................................................................................79 Table 9.3. Various parameters required to predict the response to selection. .............................................85

xi

List of Figures Figure 2.1. Location of Janamba Croc Farm in relation to Darwin, Northern Territory, Australia................9 Figure 2.2. Schematic diagram of Janamba Croc Farm, Middle Point, NT, Australia.................................11 Figure 2.3. Scute marking system for crocodilians ......................................................................................13 Figure 2.4. One of the twenty-six fibreglass tubs.........................................................................................14 Figure 2.5. The tub and intermediate area....................................................................................................14 Figure 2.6. Environmentally-controlled nursery ..........................................................................................15 Figure 2.7. Small nursery pens.....................................................................................................................15 Figure 2.8. Large nursery pens.....................................................................................................................15 Figure 2.9. Two intermediate pen designs allow the animals to hide, and feel less threatened, during

cleaning and other procedures....................................................................................................16 Figure 2.10. Slaughter size crocodiles inside a channel pen ..........................................................................16 Figure 2.11. Diagram of a crocodile belly skin with the features used in measurement and grading, in which

the most valuable area is the belly pattern .................................................................................20 Figure 2.12. Measuring morphological traits on crocodiles ...........................................................................23 Figure 5.1. Histogram of belly widths available for analysis .......................................................................47 Figure 6.1. Kaplan-Meier estimated baseline survival function for crocodiles between hatch

and slaughter ..............................................................................................................................56 Figure 7.1. Belly region of a crocodile.........................................................................................................60 Figure 7.2. Examples of WMI method for counting scale rows...................................................................61 Figure 7.3. Percentage (%) male and average scale row number (squares with linear trendline; adapted

from Manolis et al., 2000) versus constant incubation temperature (oC) ...................................66 Figure 7.4. Systematic matings to test hypothesis that genetic imprinting is responsible for scale row

expression ..................................................................................................................................66 Figure 9.1. Pair CBVs (± SE) for the reproduction breeding objective, number of hatchlings produced per

clutch per year (NoHatch)..........................................................................................................74 Figure 9.2. Pair CBVs (± SE) for the breeding objective, age at slaughter (days) .......................................75 Figure 9.3. A) Log hazard pair estimates (± SE) of juvenile survival. B) Pair CBVs (± SE) for juvenile

survival at 1095 days (or three years) ........................................................................................76 Figure 9.4. Pair CBVs (± SE) for number of scale rows ..............................................................................78 Figure 9.5. Crocodile economic selection index ($CESI) values for each breeding pair at Janamba Croc

Farm used in this study ..............................................................................................................80 Figure 9.6. Schematic representation of a proposed selection program for a crocodile production system

(modified from Hicks et al., 1998).............................................................................................82

xii

Executive summary Background Profit from farmed crocodile is essentially a function of the returns and costs from the average lifetime productivity of the herd. The aim of a genetic improvement program is to improve the total economic value of the herd, and consequently maximise profit. To date, no research has been conducted to evaluate the potential of a genetic improvement program in the Australian crocodile industry. By implementing a selection program based on reproductive performance, juvenile growth rates and juvenile survival rates, the resultant superior breeding animals will increase the profitability of crocodile farms. The major benefits to the industry will be decreasing overhead costs by growing animals to marketable size in a quicker time, increasing profitability by offsetting some of the production costs per animal and increasing the number of animals obtained from the farm each year. The major aim of this project was to create a practical genetic improvement program for immediate adoption by the Australian crocodile industry, to be called CrocPLAN. Research Breeding objectives were defined for a crocodile genetic improvement program and a survey was distributed to industry members. Reproduction, production and survival data were obtained from Janamba Croc Farm, Northern Territory, Australia between 1994 and 2002. Analyses were conducted to obtain relevant genetic and phenotypic parameters (heritability, repeatability and correlation) to determine the relative importance of genetic effects on economically-important traits. These traits included the number of hatchlings produced per female per year, slaughter age, juvenile survival and the number of scale rows. Relative economic weights for each breeding objective were estimated by fourth year economics student, Ms Emily Gray. These were combined with estimated crocodile breeding values (CBVs) into a crocodile economic selection index ($CESI). This selection index ranks animals based on a single dollar value, allowing selection decisions to be made. Outcomes Key outcomes from this study included: Breeding objectives thought to be important for improving crocodile farm profitability were skin grade, number of hatchlings per female per year, survival rate, food conversion efficiency, and age to slaughter (growth rate). In the future, skin “quality” (scale row number and regularity, shape and thickness) traits may also be incorporated into a multi-trait selection index. A survey of industry members asked for priorities to be given to each objective. Some participants indicated they had already begun selecting candidates for traits including growth, survival and skin tensile strength, suggesting that a genetic improvement program will be readily adopted by the industry. For the reproduction breeding objective, number of hatchlings per female per year, repeatability was high 0.68. Repeatability estimates for the other possible selection criteria for this objective were also high ranging from 0.24 (hatchability) to 0.68 (initial clutch size and time of nesting). Phenotypic correlations between this objective and its criteria ranged from negligible (0.03) to high (0.86). Resultant CBVs ranged between -8.75 and 15.09 hatchlings, when expressed as a deviation with the average herd CBV centred at zero. This is a difference of 23.84 hatchlings per female per year. The breeding objective, age at slaughter, and its possible selection criteria (hatchling and inventory morphometric measurements) were analysed. Heritability estimates were all high ranging from 0.40 to 0.60. Additionally, the genetic (-0.81) and phenotypic (-0.82) correlations between slaughter age and inventory head length were high. The deviations from herd average CBV for age at slaughter ranged between -158 days and 144 days, a difference of 302 days. From the survey results, the breeding objective ranked the highest priority (4.9 of a possible 5) was juvenile survival. Analysis of the number of juveniles surviving to slaughter size produced a

xiii

heritability estimate of 0.15 for log survival. This analysis was conducted using an extension of Cox’s proportional hazard analysis to include a frailty (random effect) term. Using approximate CBVs estimated for juveniles surviving to three years of age, offspring from the highest-ranked pair have a 20.3% reduced risk of mortality compared to the herd average, whilst offspring from the lowest-ranked pair have a 23.5% greater risk of mortality. Although a premium price is not yet received for additional skin quality traits, such as number of scale rows and thickness, it was considered worthy of investigation to determine if there is any genetic variation for selection to occur in the future. Data for number of scale rows only was available. The heritability estimate using data from Janamba Croc Farm was 0.37, whilst the inclusion of additional data from Wildlife Management International Pty. Ltd (Darwin, Northern Territory) produced a heritability estimate of 0.44. The range of CBVs using the Janamba data only was between 0.92 rows and -1.20 rows, expressed as a deviation from the average offspring scale row count when centred at zero. A parentage testing kit was developed using eleven previously-published microsatellite markers. These markers detected a 5.6% pedigree error rate from 107 juvenile samples collected from 16 known-breeding pairs, which is consistent with pedigree error rates in other industries. If consistent, this level of pedigree error could have an impact on the accuracy of genetic parameter and breeding value estimation. In addition, these markers were shown to be adequately polymorphic to retrospectively determine parentage in situations where maternity and/or paternity may not be known, such as when multiple males and females breed in lagoons. In these situations, a 2% error in parentage assignment was predicted. The use of these microsatellite markers will broaden the scope of a breeding program allowing progeny to be tested from adults maintained in large breeding lagoons for selection as future breeding animals. Implications Combining the CBVs produced for each of the breeding objectives and the relative economic weights produced by Ms Emily Gray into a crocodile economic selection index ($CESI), a single dollar-value index was estimated for each breeding pair at Janamba Croc Farm. Each $CESI value was expressed as a dollar ($) deviation from the average herd profitability with offspring from the pair B16 having the largest CESI values (+$4,748), whereas offspring from the pair B01 have the lowest potential value (-$5,257). The response to selection was predicted to be a $324 increase in profit per annum per pair, assuming no genetic and phenotypic correlation amongst breeding objectives. This result reinforces the potential for implementing a genetic improvement program (CrocPLAN) on Australian crocodile farms. Future Research Before the industry invests a large amount of capital into a genetic improvement program, a cost-benefit analysis should be commissioned that takes account of the various production and breeding systems represented in the industry. In addition, an integrated performance recording system needs to be initiated to record pedigrees, to optimise individual identification (scute-cuts, tags, etc), and decide on the various breeding objectives and relevant selection criteria. It would be advantageous for an industry-wide service to be available similar to the service provided for the pig industry (PigBLUP at http://agbu.une.edu.au/PIGBLUP/index.html). Producers would provide pedigree, performance and reproductive records on particular animals. These would then be combined with data from other animals from that farm and other farms, and $CESI values for each individual would be calculated. This would facilitate across-herd trade of genetically superior (fertile) eggs, hatchlings, juveniles, and adults, and minimise inbreeding rates and maximise the effective population size. The data provided by Janamba Croc Farm has provided a wonderful foundation for estimating the quantitative parameters presented in this study. However, there were many gaps in the data where complete analyses could not be conducted. One particular limitation was the limited pedigree structure. Follow up analyses when a more complex pedigree structure becomes available will be greatly beneficial. The preferred model for quantitative analysis is the “animal” model, which takes

xiv

into account all the relationships between individuals within the data-set and ultimately produces an CBV for each animal. Also, calculation of the genetic and phenotypic correlations between breeding objectives will be essential to more accurately predict response to selection. Data must still be collected on feed intake and growth rate to enable quantitative genetic analysis of feed conversion efficiency. Also, skin grade was not investigated since appropriate data were not available. However, with the large number of skins not meeting first grade requirements a high priority should be placed on research to evaluate whether there is any genetic basis for skin damage. One possibility might be to conduct a behavioural study since anecdotal evidence suggests that some clutches are more “aggressive” than others, even at the time of hatching, implying a familial and possibly genetic basis for the differences in aggressive behavior and resultant skin damage. Continued industry enthusiasm for a genetic improvement program will provide an opportunity to address some of these current limitations. Conclusion Recommendations for CrocPLAN implementation are given using Janamba Croc Farm as a model. Selection of replacement breeder animals will occur when the animals reach slaughter size (approximately 2.8 years), after adjusting for the relevant fixed effects. Juveniles will be selected on their $CESI value using individual performance records integrated into an index with records from relatives. 10% of adult breeding pairs will be replaced each year, with selection intensities (i) of 2.88 and 2.05, respectively for males and females. Generation interval was estimated to be 13 years for both males and females using the assumptions that: sexual maturity will occur at eights years of age for both males and females; a replacement rate of 10% is used; and, the oldest six pairs are replaced each year (that is, each animal breeds for ten years after sexual maturity). A recommendation for maximising genetic variation is the industry-wide adoption of a CrocPLAN allowing across-herd selection of replacement animals. With a functional crocodile economic selection index, $CESI-values can be obtained at any time, allowing across-herd trade of genetically superior (fertile) eggs, hatchlings, juveniles, and adults. The occasional inclusion of wild-harvested animals may also provide additional variability. Publications Publications to date include: 1. Isberg, S.R., Y. Chen, S.G. Barker and C. Moran. 2004. Analysis of microsatellites and parentage

testing in saltwater crocodiles. J. Hered. Accepted. 2. Isberg, S and C. Moran. 2003. Increasing production efficiency: A genetic approach. Crocodile

Capers 8:4. Queensland Department of Primary Industries, Townsville, Qld. 3. Isberg, S and C. Moran. 2004. A parentage determination kit for saltwater crocodiles. Crocodile

Capers. Queensland Department of Primary Industries, Townsville, Qld. Submitted. 4. Isberg, S.R., F.W. Nicholas, P.C. Thomson, S.G. Barker, S.C. Manolis, and C. Moran. 2003.

Defining breeding objectives for saltwater crocodile genetic improvement programs. Proc. Assoc. Advmt. Anim. Breed. Genet. 15:166-169.

Material presented in this report will be submitted in a thesis to be submitted for the Degree of Philosophy, University of Sydney, Australia.

1

1. Introduction 1.1 General introduction

Production from farmed crocodile for the skin trade is an emerging industry. Recent Australian research to improve the efficiency of crocodile production has concentrated on husbandry-related practices. Examples of this include improving skin quality using single grow-out pens (MacNamara et al., 2003), optimising incubation temperature regimes to improve skin quality (Manolis et al., 2000) whilst not compromising post-hatching growth and survival (Webb and Cooper-Preston, 1989), and reducing feed costs using a processed pelletised food (Davis, 2001). To complement these studies, this study investigates the possibility of incorporating a genetic improvement program to simultaneously improve the efficiency of crocodile production. This is the first time that a genetic improvement program has been investigated in any of the Crocodylia. As such, no literature of genetic and phenotypic parameter estimates exist for review. Instead, this chapter outlines genetic improvement programs from the perspective of providing an emerging industry with the information required for developing and implementing an economic selection index. The main objective of genetic improvement programs is to maximise profit by selecting candidates for breeding that are genetically superior for traits that either lower production costs, or alternatively, increase returns through productive attributes sought by the market. In turn, this increases the overall value of the herd. To achieve this, a concise understanding of the whole farming system; the product market, including supply, demand, competitors and constraints; and genetic improvement program design, development and implementation are needed. These are all briefly outlined in the chapter below.

1.2 The Australian crocodile industry

The Australian crocodile industry is based on the production of saltwater crocodiles (Crocodylus porosus) for the international skin trade. For those animals whose skins meet the requirements of an export-quality, first-grade skin, 80% of the total product value is derived from the production of skin, with the remaining 20 per cent derived from the sale of meat (15%) and by-products (5%; backstrap, head/skull, feet). For the 2001-2002 financial year, the Australian crocodile industry’s total (skins and meat) export earnings were reported to be just over A$1.8 million (Australian Bureau of Statistics, 2002). At present, there are 14 commercial farms in Australia, with six in both the Northern Territory and Queensland, and two in Western Australia (Love and Langenkamp, 2003).

1.2.1 The product Saltwater crocodile skins are considered superior to other crocodilian skins. Saltwater crocodiles have a greater number of scale rows (average 31.2, range 27-37; Manolis et al., 2000) compared with other crocodilian species (Brazaitis, 1987). This adds to the aesthetic appeal of the finished product, with a more prominent scale pattern being displayed. Another superior quality of saltwater crocodile skins is the absence of osteoderms (bones in the dermal layer of the skin) in the belly skin. Osteoderms increase the risk of tearing during tanning (MacNamara et al., 2003) and produce a pitted, discoloured appearance in the finished skins (Thorbjarnarson, 1999). Skins are predominantly sold as raw belly skins, which includes the tail and throat. The aim is to maximise the uncut area of the belly skin, allowing one whole skin to be manufactured into the final product (Manolis et al., 2000). Saltwater crocodile skins are mainly used for the manufacture of handbags and shoes (MacNamara et al., 2003). Crocodiles are generally harvested on farms when they have a belly width ranging between 35 and 45cm (2-5 years of age; Treadwell et al., 1991). This is the industry-preferred belly width for manufacturing handbags to minimise the wastage of off-cuts during product

2

manufacture. There is occasional demand for smaller skins to be manufactured into small leather goods such as watchstraps, and also for larger skins (greater than 50cm) in response to fashion trends towards larger-sized handbags (MacNamara et al., 2003). Skins are sold on a ‘$ per centimetre’ belly-width basis in conjunction with a stringent, yet subjective, grading system (discussed below). First-grade skins are highly sought on the international market (traded in US$/cm). Contrary to this, there is a weaker demand for lower-grade skins and they attract a much lower price, usually being sold domestically (AU$/cm) or sold for tanning and re-imported for manufacture. The skins are sold as “green” skins which have been salted and chilled to preserve skin structure during storage, and to retard bacterial growth (Van Jaarsveldt, 1987). Meat is an important by-product. On average, slaughter animals yield 5-6kg of flesh. Retail cuts of meat are generally packaged as inner and outer tail fillet (40 per cent of the total meat), body (30 per cent), back fillet (15 per cent), satay (10 per cent) and legs (five per cent) (unpublished production records). The meat is sold predominantly as boneless, although bone-in is sometimes preferred.

1.2.2 Skin grading system Skins are subject to a stringent grading system and are classified most commonly as either first, second or third grade, although fourth grade and reject categories also exist. Skins are graded according to the number and severity of blemishes on the belly area. A first grade skin will have no blemishes, have four appendages and be well preserved. The presence of any bite marks, abrasions or knife holes results in an automatic downgrading of the skin. Second grade skins may have an imperfection such as a missing appendage, a hole within the belly region or scales lifting from the leather, whilst third grade skins have two or more blemishes (Manolis et al., 2000). Sources of damage are most likely to be caused by bite marks from fighting (caused by inappropriate stocking densities and large size disparities in pens), scratches from rough floor surfaces, but may also include bacterial and fungal infections such as brown spot and pox virus (Manolis et al., 2000). Inadequate preparation of skins can also cause downgrading, such as the blurring of scales due to inadequate salting (MacNamara et al., 2003) or “red heat” from inappropriate storage temperatures (Van Jaarsveldt, 1987). Strong demand and premium export prices are received only for first grade skins. In contrast, second and third grade skins are in competition with other crocodilian skins (particularly the large number of alligator and Caiman spp. skins available) and receive a large discount in their value. The prices received by Australian producers vary since producers individually negotiate with the tanneries. Table 1.1 illustrates the range of prices received by Australian producers. A second grade skin is worth about 50-75 per cent of a first grade skin, whilst a third grade skin only about 25 per cent (Davis and Peucker, 2001). The ascending pricing regime with increasing belly width (Table 1.1) demonstrates that there is a demand for skins larger than 45cm. However, the increased costs of production associated with producing a larger skin is cost-prohibitive, along with an increase in the risk of blemishes, and thus skins being downgraded. With the strong pricing differential between first and lower grade skins directly influencing farm income, an issue of large concern is the number of skins failing to meet first grade requirements (Manolis et al., 2000). MacNamara et al. (2003) reported that only 50 per cent of crocodiles currently being produced in Australia meet first grade requirements, and this is thought to be an overestimate.

3

1.2.3 Export markets The major export destinations for Australian saltwater crocodile skins are France, Italy, Japan and Singapore. Australia exports approximately 11,000 skins (inclusive of all skin grades) per annum (Davis and Peucker, 2001) and the Northern Territory is the largest supplier, producing 6500 skins per annum (MacNamara et al., 2003). However, the Australian industry is small, supplying only one per cent of the global market of crocodilian skins (DOTRS, 2001) and the largest concern of skin buyers is the insufficient supply of first grade saltwater crocodile skins (Manolis et al., 2000).

Table 1.1. A range of prices received for saltwater crocodile skins (Barker, pers. comm.). These prices are all shown in $US, although second and third grade skins are seldom exported and are traded on the domestic market in Australian dollars. (AU$1 = US$0.78; 11/01/04).

Prices received per centimetre (US$/cm)

Belly Width (cm) First Grade Second Grade Third Grade

18-24 4.50-5.00

25-34 5.50-6.00

35-39 7.50-8.50 3.50-4.50 1.50-2.00

40-45 8.50-10.00

46-50 10.00-11.00

1.3 General introduction to genetic improvement programs

Farm profit is essentially a function of the returns and costs from the average lifetime productivity of the herd. The aim of a genetic improvement program is to improve the total profitability of the herd, and consequently maximise farm profit. After identifying the traits of economic importance in the production system, and using an economic selection index, genetically-superior candidates can be selected to be the parents of the next generation (Smith, 1983). The selection of a candidate is based on the overall profitability of the animal, in dollar-terms. This is achieved using the weighted sum of a candidate’s breeding value for each economically-important trait, where the weights are the economic values (Bourdon, 2000). The economically-important traits are known as breeding objectives, and the economic selection index is defined as

H = v1BV1 + v2BV2 +...+ vmBVm [1.1] where H = aggregate breeding value of the animal for profitability,

vi = the economic value for the ith breeding objective (expressed as the increase in profitability per unit increase in the breeding objective),

BVi = the true breeding value for the ith breeding objective, and m = the total number of breeding objectives in the selection index.

Candidates are ranked based on an estimate of their aggregate breeding value. The rate of genetic improvement is maximal when selection is based upon H. However, H is usually never known exactly. Instead breeding values need to be estimated, with greater or lesser precision, using an index I, as follows (Nicholas, 2003)

4

Ii = EBVi = b1C1 + b2C 2 +...+ bnC n [1.2] where Ii = index value (best but imperfect estimate of BVi),

EBV i = the estimated breeding value for the ith breeding objective, bi = weighting factor maximising the correlation between Ii and BVi, Cn = clues (or selection criteria) to the candidate’s breeding value, and n = the number of clues available for estimating the breeding value.

The first step in designing a genetic improvement program is to determine appropriate breeding objectives. These breeding objectives are a set of performance traits that have economic importance and, therefore, influence profitability. The second step is to identify the clues, or selection criteria, that maximise the correlation between the true and estimated breeding values. These selection criteria must be observable in the population and relate to the particular breeding objective. Genetic and phenotypic parameters, such as heritability, genetic and phenotypic correlations, can then be estimated from the breeding objectives and selection criteria. Once appropriate selection criteria have been identified, breeding values can be estimated for each animal to be combined into the selection index. Finally economic values need to be estimated for each breeding objective. These are all discussed in the sections below.

1.4 Defining breeding objectives Defining the overall breeding goal is the most important step in designing a genetic improvement program (Ponzoni and Newman, 1989). The breeding goal of the producer is to increase the profitability of the enterprise using the economic selection index defined in equation 1.1. The breeding goal is the weighted sum of economically-important traits, known as breeding objectives. The traits of economic importance are obviously those that influence farm profitability, either by affecting net returns or production costs. For example, an Australian pig producer may want to decrease the time it takes to reach the desired market weight, decrease back-fat depth, and to increase the number of piglets born alive per litter. Since this is the first time that a genetic improvement study has been contemplated for the Australian crocodile industry, one of the most important aims of this study was to define the breeding objectives for improvement.

1.5 Selection criteria Since the true breeding value of an individual for a selection objective is almost always unknown, an estimate of the breeding value is required. This is achieved by utilising clues to an individual’s breeding value. This can be the phenotypic performance of the candidate for the breeding objective itself, or alternatively, the phenotypic performance for a trait that has a non-zero correlation with the breeding objective (James, 1982). Information from a candidate for selection, as well as information from its relatives, can be incorporated into the selection index (Nicholas, 2003). Some objectives are both expensive and difficult to measure, for example, food conversion ratio in pigs. Genetic studies have shown that there is a high, although variable, genetic correlation between food conversion ratio and average daily gain in pigs (average -0.53, range -1.24-0.34; references within Clutter and Brascamp, 1998). Therefore, measuring the selection criteria, average daily gain, provides a basis for improving the selection objective, food conversion ratio. Another use of selection criteria is when an objective can only be measured accurately on an animal post-slaughter. An example of this is the carcase cut, loin eye area (LEA). However, an animal can not be used as a breeding animal post-slaughter. Instead, back-fat thickness and average daily gain are selection criteria which can be measured on the live animal. Johnson et al. (1999) reported genetic correlations of -0.27 and 0.36 for back-fat thickness and average daily gain, respectively with the objective LEA. Measuring suitable selection criteria instead of

5

the actual breeding objective may also permit earlier selection of candidate breeding animals and thus reduction of the generation interval. If measures of lifetime performance are a selection objective, then any delay in selection until data on the objective becomes available will increase the generation interval. In addition, there is an increase in the maintenance cost of keeping these potential replacement breeders if selection decisions cannot be made early. For these reasons, there is considerable interest in utilising performance records at the earliest stages possible for indirect selection. In crocodiles, sexual maturity does not occur until at least eight years of age (Elsey et al., 1993) and thus there is a high maintenance cost associated with keeping a non-producing animal for that period of time. However, care must also be taken using early, indirect selection criteria so as not to compromise the efficiency or accuracy of improving lifetime productivity and profitability. To address these concerns, it is necessary to estimate the genetic and phenotypic variances of each potential objective and criterion, and all the genetic and phenotypic covariance between them. The method of choice for animal breeders to estimate these parameters is Restricted Maximum Likelihood (REML) (briefly discussed below). Knowledge of these parameters allows efficient genetic improvement program design (Harris and Newman, 1994), prediction of direct and correlated response to selection (Falconer and Mackay, 1996), and most importantly in the context of this study, the estimation of candidate breeding values (Robinson, 1991).

1.6 Estimating breeding values The most common type of statistical model used to predict breeding values based on phenotypic observations is the linear mixed model. Mixed models, based on Henderson’s method 3 (Henderson, 1953) enable adjustment for both fixed effects and random effects, jointly. Fixed effects are the effects attributable to a finite set of levels of a factor that occur in the data, or are pre-set by the operator. The effects can occur as either categorical or continuous variates, for example different year (categorical), herd (categorical) or age (covariate) effects. In contrast, random effects are ascribed to an infinite set of factor levels, of which only a random sample are deemed to occur in the data (Searle et al., 1992). In the context of animal breeding, examples of random effects are sire, dam, permanent environmental, maternal effects and, of course, random error. Solutions for these mixed models are most commonly achieved using Restricted Maximum Likelihood (REML), since REML has the ability to accommodate unbalanced and complex data sets, and missing values (Robinson, 1991). Variance components, used to estimate genetic parameters, are produced as part of the solutions to the mixed model equations. In addition, the prediction of a random effect for each animal is produced. This prediction is known as a Best Linear Unbiased Predictor or BLUP (Henderson, 1953; 1973). These BLUP values are the estimated breeding values (EBVs) required for inclusion in the economic selection index. Some advantages of estimating breeding values calculated using BLUP methodology are that they can accommodate complex data structure including non-random matings, environmental trends, sire groups or populations, herd differences in genetic merit, and selection and culling bias (Nicholas, 2003). Most importantly however, BLUP can include and correctly account for the records of relatives using a relationship matrix and therefore utilise all available data for estimating a candidate’s breeding value.

1.7 Relative economic weights In equation 1.1, economic values are required to indicate the relative importance of a marginal change in the respective breeding objective, as a dollar value. When combined into the economic selection index, the economic values act as weightings for the estimated breeding values to produce the best estimate of the aggregate (true) breeding value for each available candidate. The aggregate breeding values are then given as a single dollar value, on which candidates can be ranked for selection.

6

Economic weights have conventionally been calculated from profit functions using the simplified formula (Ponzoni and Gifford, 1990)

Profit = Income – Expenses [1.3] Therefore, before estimating economic values, all sources of income and expense need to be identified. The economic values are the partial derivatives, with respect to each selection objective, of the profit function in equation 1.3. The partial derivative allows the economic value to represent a change in profit resulting from a one unit change in the trait using genetic selection, whilst all other traits are held constant (Ponzoni and Gifford, 1990). Using the simple profit function assumes that the profitability of animals is a linear function of measurable production characteristics (Goddard, 1983). Alternative methods of modelling economic values are now becoming more widely used. These other methods include models such as the bioeconomic model (Cameron and Crump, 2001) and the generalised Cobb-Douglas function (Bright, 1991), and are based on the concept of profit maximisation using production economics theory. The models have the advantage of not assuming linearity, can include interactions between traits and are responsive to changes in the production system due to improvements in genetic and managerial changes.

1.7.1 An example of economic values in the Australian pig industry Economic selection indices are used in the major livestock industries. As an example, the Australian pig industry is a relevant model for the crocodile industry since the females have litters of offspring (multiple births) and the juveniles are slaughtered at particular weights to meet market-specified demands. The traits identified in the breeding goal and their relative economic values, as reported by Cameron and Crump (2001), are given in Table 1.2. One decision that needs to be made when estimating economic values is the basis on which the weights will be derived. To illustrate this, the economic values shown in Table 1.2 are given from two different bases: ‘per slaughter pig’ and ‘per farrowing’ basis. The differences in economic values produced using the different bases reflect litter size and offspring mortality rates. Based on the economic values shown in Table 1.2, the objective, the number of piglets born alive, contributes most significantly to the overall production system, followed by feed conversion ratio, carcase back-fat and dressing percentage.

Table 1.2 Increment of improvement and economic values for various breeding objectives in the Australian pig industry (Cameron and Crump, 2001). The economic values were derived on a ‘per slaughter pig’ and a ‘per farrowing’ basis for comparison.

Economic values in AU$

Breeding objective Increment per slaughter pig per farrowing

Average daily gain 10g/day 0.49 3.97

Carcase back-fat -1 mm -2.05 -16.55

Feed conversion ratio -0.1 kg/kg -2.11 -17.04

Number born alive 1 piglet 3.56 31.72

Dressing percentage 1% 1.39 11.25

7

1.8 Molecular genetics: uses in animal breeding Accurate selection of candidates using EBVs depends on correct pedigree information (Nicholas, 2003). As mentioned above, genetic improvement programs are based upon the selection amongst candidates using EBVs that have been estimated using information from the candidate and its relatives. Therefore, errors in assigned parentage may lead to the incorrect genetic evaluation of a candidate. These errors can result in the real genetic improvement being less than was expected (Visscher et al., 2002) and can hence compromise the success of the improvement program. Microsatellite markers can provide a way of identifying pedigree errors, or alternatively, identify the parents of individuals that would have previously been excluded from analysis. In addition, information from microsatellite markers can be used to investigate marker linkage and to create genetic linkage maps. These uses form the basis of continuing molecular genetic research with the ultimate goal of selecting animals using marker-assisted selection (MAS) (Hayes and Goddard, 2001).

1.8.1 Microsatellites and parentage determination Microsatellites are small, tandem repeat motifs (one to five base pairs) that occur widely across the genome. They have become the genetic markers of choice since they are highly polymorphic and amenable to automated analysis. The first microsatellite markers for crocodilians were developed using American alligators (Alligator mississippiensis; Glenn et al., 1996, 1998). Although many of these loci could be amplified in Crocodylus spp., they were not sufficiently variable for population genetic studies. To overcome this problem, FitzSimmons et al. (2001) developed 26 microsatellite markers from three crocodile species: the American crocodile (C. acutus), saltwater crocodile (C. porosus) and Johnston River crocodile (C. johnstoni). In addition to developing these primers, FitzSimmons et al. (2001) evaluated some of the markers on C. porosus, although not with the objective of developing a parentage determination kit. Some of these primers have since been used to identify purebred Siamese crocodiles (C. siamensis) for reintroduction into Vietnam (FitzSimmons et al., 2002).

1.8.2 Marker assisted selection (MAS) Selection of an individual, in the context outlined above, is based on the phenotypic performance of the candidate and its relatives. However, this selection is based on little or no knowledge of what is actually occurring at the DNA level (Georges, 2001). The phenotype measurable on an individual for a particular quantitative trait is the result of additive effects and interactions between numerous loci scattered throughout the genome as well as numerous identified and unidentified environmental factors (Nicholas, 2003). These loci responsible for quantitative variation are known as quantitative trait loci, or QTL (Geldermann, 1975). The major objective of QTL studies is to identify genes and markers for inclusion into genetic improvement programs using marker-assisted selection (MAS) (Khatkar et al., 2004). This technology has the potential to increase the rate of genetic gain more than is currently achieved using phenotypic data alone. In a simulation study, Meuwissen and Goddard (1996) showed that there was a 64% increase in genetic gain using both BLUP and MAS compared to BLUP alone. Additional benefits include pre-selection of younger candidates, which increases selection differentials and shortens the generation interval (Khatkar et al., 2004). There are three types of MAS available for selection of candidates, although they vary in their accuracy and utility (Georges, 2001). The first two methods involve markers that are linked to the QTL. These markers have no direct effect on the phenotypic performance of the trait. Using the first method, the marker is linked to the QTL but there is no information on linkage disequilibrium. Therefore, the association between the QTL and markers needs to be re-established between generations since the relationship only exists within families. Alternatively, a second method exists when the same association between the QTL and marker allele exists across the population and across generations. This association is founded upon linkage disequilibrium and is becoming the most widely preferred method of MAS

8

(Hayes and Goddard, 2003, 2004). The third, and most desirable, type of MAS is when the gene and the mutation causing the phenotypic difference are discovered. Direct selection for a particular mutation can then occur providing an accurate and reliable method of selection. In addition, discovering the mutation can provide fundamental information of the biology of the underlying trait (Georges, 2001).

1.9 Objectives In an attempt to create a multi-trait genetic improvement program ready for immediate industry adoption, the objectives of this study were: 1. Identify, by means of industry collaborations and a survey, breeding objectives of major economic importance within the Australian crocodile industry for inclusion into a genetic improvement program. 2. Obtain univariate and multivariate mixed model estimates of variance and covariance components for genetic effects, common environment effects, and their corresponding relevant genetic and phenotypic parameters for the following traits 1. Reproductive traits 1.1. Number of eggs laid 1.2. Number of fertile eggs 1.3. Number of eggs hatching to produce a viable offspring 1.4. Timing of reproductive activities 1.5. Regularity of breeding cycles 2. Age at slaughter 2.1. Hatchling size 2.2. Inventory (~ nine months) size 2.3. Age at slaughter 3. Survival, and 4. Number of scale rows. 3. To derive relevant economic values to weight each respective breeding objective and identify priorities in creating a multi-trait genetic improvement program. 4. To develop a microsatellite-based parentage determination kit to ensure the integrity of the pedigree records used to aid in identifying the correct animals for selection as future breeding animals. Also, a microsatellite-based parentage determination kit broadens the application of a selection program on crocodile farms by enabling animals housed in multiple mating environments, and their offspring, to potentially be included in improvement program.

9

2. Resource description and analytical methodolgy Janamba Croc Farm has provided data and animal resources essential for the research reported in this thesis. In the initial part of this chapter, the farm, its history and its operation will be described. The second part of this chapter describes the data available and statistical methodology used to obtain parameter estimates for crocodile performance traits.

2.1 Overview of Janamba Croc Farm The crocodile farm is located 70km southeast of Darwin, as shown in Figure 2.1. It was established in 1980-81. Initially it produced Johnston River (Crocodylus johnstoni) skins as the saltwater crocodile (C. porosus) management program had not been fully devised. Also, CITES restrictions at that time only allowed a limited number of C. porosus hatchlings and breeders to be sourced through the Northern Territory Government trial egg harvest and the Northern Territory Government problem crocodile program, respectively. With a change of ownership in 1985, much of the infrastructure was installed. Expansion of infrastructure continued when the farm changed hands again in September 1989 to the present owners, Deugro Pty Ltd. Changes in CITES regulations in late 1987 allowed Australia to trade C. porosus products on the world markets for the first time since bans were implemented in the late 1960s. At this time, the current focus on C. porosus production began. Currently there are approximately 10,000 animals held on the farm, of which around 220 are kept primarily as breeding animals. The increase in C. porosus and respective decrease in C. johnstoni numbers over the period of current ownership are shown in Table 2.1. Unfortunately animal numbers prior to this are unknown.

Figure 2.1. Location of Janamba Croc Farm in relation to Darwin, Northern Territory, Australia (modified from Browne, 2004).

10

Table 2.1. Farm number of saltwater (C. porosus) and freshwater (C. johnstoni) crocodile numbers at Janamba Croc Farm. The numbers prior to these dates are unknown.

C. porosus C. johnstoni January, 1990 2,500 7,000 January, 1993 4,500 7,100 January, 1995 6,500 3,400 January, 1997 8,500 150 July, 2000 10,000 15

The farm consists of 40 hectares of gently sloping land, backing onto Harrison’s dam and surrounding floodplains. This topography provides fairly good drainage. The layout of the farm is presented schematically in Figure 2.2.

2.1.1 Breeding stock The breeding stock at Janamba Croc Farm was mainly derived from wild-caught animals. Unfortunately, since there is no reliable method for estimating ages of wild crocodiles, especially of mature animals, nothing is known about their prior reproductive history and status. The only information recorded for the animals is their length at the time of arrival at Janamba.

2.1.1.1 Breeding infrastructure There are two types of breeding pens in which individual sires and dams can be distinguished:

• B pens- these pens have just one male and one female and produce about 25% of nests on the farm, and

• UB pens- most have just one male and one female although some do contain two females. The UB-pens produce about 30% of nests on Janamba Croc Farm.

Most of the animals have been together for a number of years and have produced multiple clutches. The breeding animals were maintained in two types of pens: the B- and UB-pens. The B-pens contain a single male and single female, whilst the UB-pens usually have a single male and 1-2 females. The B-pens are much smaller (144m2) in comparison to the UB-pens (600-800m2). The UB-pens are considered to better mimic the natural environment with deeper water (up to 5m) and adequate basking banks allowing the animals to regulate their own body temperature. In contrast, the B-pens have two small water ponds with one being slightly larger than the other and no more than one metre deep. Under tropical conditions, the animals in the B-pens do not have the same thermoregulatory behaviour patterns available as the UB-pen animals. The possible inability to thermoregulate as effectively could affect the developing embryos during the period of four to eight weeks between mating and oviposition (Lang, 1980). The second water pond in the B-pens provides a refuge for the females, which are sometimes subject to lethal or damaging male aggression. Two of the females included in this study were seriously injured by their male partners. Breeding crocodiles are fed once a week, generally receiving either one chicken each or equivalent in red meat.

The majority of breeding animals are maintained within lagoons, producing about 45% of nests. Although the female parents in the lagoons can usually be identified relatively easily due to their nest-protecting behaviour, male parents cannot. This inability to identify paternity among multiple males in the lagoons means that lagoon-bred progeny have to be excluded from analysis and genetic parameter estimation. A secondary aim of this research is to develop microsatellite markers to permit identifying paternity.

11

Figure 2.2. Schematic diagram of Janamba Croc Farm, Middle Point, Northern Territory, Australia

123

45

6

910

78

Office

House

water

pon

Mangoes

Mangoes

lunchroom

abattoir

workshop salting room meatroom

egg processingroom

freezerNative bush

17 18 19 20

14 15 16

11 12 13

9 10 5 6 7 8

1 3 2

4

Intermediates and Tubs

New Nursery

drainage channel

Native bush

Native bush

S2 S1 S8

S9

S4 S3 S7

S6 S5

UB4 UB3 UB2

UB5

UB1

UB6 UB7

UB12

UB14

UB13

UB9

UB8UB17

UB10 UB15

UB16

UB11 Lagoon2

Lagoon 3

Lagoon4

Lagoon 5

Lagoon 1

Roads

Bore 1

Bore 2

Display Pens

Channels

50 metres

N

1

2

20 16 18 14 12 10 17 19 15 11

8 4 6

7 9 5 3 13

B Pens

Spare pens

Native bush

12

2.1.1.2. Breeding history at Janamba The first captive-bred nest was collected in January of the 1987/88 season and detailed records have been kept for subsequent breeding seasons. Table 2.2 shows the trend in reproductive performance on the farm between the 1989/90 and 2001/02 breeding seasons.

2.1.1.3. Feeding the breeding stock The breeding crocodiles are fed only once a week, generally receiving one chicken each or equivalent in red meat (kangaroo, horse, beef or pig). There are also wild birds that are possibly caught and eaten by the animals, such as Radjah shelduck (Tadorna radjah), magpie geese (Anseranas semipalmata), pied heron (Ardea picata) and whistling kites (Haliastur (Milvus) sphenurus). Water pythons (Liasis fuscus) are also possibly eaten. The lagoons also sustain a limited supply of native, freshwater fish.

Table 2.2. A summary of C. porosus breeding at Janamba Croc Farm from 1989 to 2002.

Nesting season

Number of nests

Av. number of eggs per

clutch

Av. % of fertile eggs per clutch

Total number of live hatchlings

1989/90 11 NR NR 184 1993/94 43 48 87.2 1153 1994/95 50 44 86.1 1284 1995/96 51 44 88.3 1344 1996/97 54 45 90.8 1647 1997/98 52 45 92.4 1495 1998/99 63 47 90.9 1995 1999/00 62 48 94.9 2053 2000/01 69 48 87.4 2091 2001/02 63 46 86.5 1790

NR- denotes not recorded.

2.1.2 Egg collection and incubation C. porosus eggs are collected from the breeding pens between November and May, with approximately 75% collected in December/January. When the clutches are laid, they are collected as soon as possible to minimise embryo mortality. Daily monitoring is undertaken to observe any nests being constructed. Nests are carefully opened until the nest cavity (where all the eggs are laid) is exposed. Each egg is marked with a pencil along the dorsal midline, to indicate its “upright” position within the nest. The eggs are carefully transferred to a padded crate and transported to a designated egg-processing area. The eggs are washed with water, re-marked along the dorsal midline, and labelled with their clutch number. Infertile eggs are detected by “candling” with a small torch. Eggs that have been collected within the first 24 hours after oviposition are still translucent and infertile members of the clutch are recognised by the absence of sub-embryonic fluid. This fluid is detectable as a floating mass on top of the yolk when the eggs are moved from side to side. If the clutch has been laid more than 24 hours before collection, the beginning of the opaque band provides a more precise indication of a fertile egg. The fertile eggs are then placed into their own crate, marked with the estimated date of hatching (80 days after day of laying), and artificially incubated at 32oC in relative humidity between approximately 90-95%. Comprehensive records are maintained on all aspects of nesting including initial number of eggs in the clutch, number infertile or died before collection, number that died during incubation, number that hatched alive and the number that were culled due to deformities.

13

2.1.3 Clutch identification marking system Immediately after hatching, all animals are scute cut for clutch (not individual) identification. Scute cutting involves the removal of a unique sequence of the raised, triangular osteoderms that occur along the animal’s dorsal surface (Figure 2.3). This method, as described by Richardson et al. (2002), is one of the few permanent and practical identification methods available although it is not without problems. Errors can result from incorrect marking originally; reading numbers incorrectly either initially or during routine management procedures; incorrect addition of numbers; numbers changing over time due to the regrowth of scutes that were not cut correctly (for example, only half the scute was removed) or other animals modifying the scute sequence by the biting of scutes. Scutes are counted starting from the point of bifurcation of the single row of scutes into the double row, which continues anteriorly (Figure 2.3). At this point of bifurcation, the single row that extends posteriorly is designated as the hundreds row (100’s), the left hand row (looking from behind the animal) is the tens row (10’s) and the right hand row is the ones (1’s). All hatchlings bred from Janamba’s breeding stock are cut at the 60 scute to allow their distinction from animals sourced from other localities. The addition of the remaining numbers identifies the clutch of origin, which can then be traced back to the parental records. For example, consider an animal with the 60, 30, 5 and 6 scutes cut. The 60 indicates an animal bred on Janamba, whilst the sum of the 30, 5 and 6 indicate that it was the 41st clutch.

Figure 2.3. Scute marking system for crocodilians. The 60 (green) scute is cut on all Janamba-bred crocodiles. The year scutes (100T and 10S; yellow) allow the age of the animal to be determined. To save confusion, the 1, 10, 100, 9 and 90 scutes are never removed.

The 100T scute (10th scute on the left hand row) and the 10S scute (10th scute on the right hand row) are designated as year scutes. By using different combinations of these year scutes, the age of the animal can be determined. For example, in 2000 both the 100T and the 10S scutes were cut, in 2001 only the 100T, in 2002 only 10S and then in 2003 no year scutes were cut. The identification of year of hatching is important for management purposes such as keeping year-cohorts together and permitting the identification of animals failing to reach slaughter size within four years. These animals can then be removed from the production system and sold into niche markets for small skins or taxidermy.

14

The point of bifurcation is sometimes difficult to identify, especially immediately after hatching as the scutes can be stuck together or not fully separated. To ensure no confusion, the 1, 10 and 100 scutes are never cut so that the point of bifurcation can always be used as a reference point. For this same reason, the 9 and the 90 scutes are also preserved to allow easy identification of year scutes. In addition to the clutch-identifying scute cuts, those offspring from the B- and UB-pens that hatched in the 2000 and 2001 hatching seasons were also given individual scute cuts so their progress could be monitored through the production system. This was achieved using a different combination of cuts involving the 100’s row.

2.1.4 Rearing hatchlings and juveniles After emergence from the shell, some hatchlings still have a large, retained yolk sac. These animals were returned to the incubator as the warmth promotes faster digestion of the yolk sac so they can begin feeding. Prior to 1998, all hatchlings were placed in fibreglass tubs. There are twenty-six of these circular tubs arranged in two rows under a shade-cloth covered area (Figures 2.4 and 2.5). The total floor area of each tub is 4.5m2 with water covering approximately 30% of the area maintained at a depth of 10cm. During the day, half of the lid is opened to permit basking but is closed at night to retain heat. Bore water is allowed to flow constantly into the rear of each pen and ranges in temperature between 29oC and 31oC, depending on the time of year. A heat lamp is also provided during the night for additional warmth. The clutches that hatch first are placed into these tubs since the weather is still relatively warm at the end of the wet season. Stocking densities are maintained at 0.09m2/animal.

Figure 2.4. One of the twenty-six fibreglass tubs. Half the lid is open to permit basking, but will be closed at night to retain warmth. Note the pipe into the back of the tub (arrow) providing a constant flow of water.

Figure 2.5. The tub and intermediate area. The intermediate pens are on either side of the two rows of thirteen tubs. The entire area is covered by shade-cloth to reduce exposure to direct sunlight.

After construction of an environmentally controlled nursery in 1998 (Figure 2.6), animals that hatch early in the year are still placed in the tubs until they reach capacity. The nursery is reserved for animals that hatch in the cooler parts of the year. This nursery is divided into two sections: small and large pen areas. The small pen area contains 80 pens, whilst the large pen area contains 40 pens, divided by an internal wall. The small pens are 2m2 (2m × 1m) each (Figure 2.7), whilst each large pen is 6m2 (2m × 3m; Figure 2.8). Animals are maintained at 0.09m2/animal and 0.32m2/animal in the small and large pens, respectively. Light is provided primarily by fluorescent tubes turned on between approximately 0700 and 1630, with additional natural light from sky-lights in the ceiling. Black plastic hide-areas are provided over half of each pen to discourage piling. An advantage of the nursery is the ability to control temperature and, to a lesser extent, humidity. Doors and shutters on the edge of the building can be

15

opened or closed to promote ventilation or retain warmth, depending on external environmental conditions. Shallow water covers approximately 40% of the available floor area and can be maintained at preset levels using thermostat-controlled valves supplying heated water to pens. The water temperature is maintained between 30.5 and 31.5oC. Water is drained prior to cleaning each morning and not re-filled until the afternoon. The pens are also left without water over the weekend in an attempt to break disease cycles.

Figure 2.6. Environmentally-controlled nursery. Doors and shutters can be opened or closed to either ventilate or retain warmth. Inside, an internal division separates the small pen area from the large pen area.

Figure 2.7. Small nursery pens.

Figure 2.8. Large nursery pens.

During their first year, if the yearlings become too large for the tubs, they are transferred into intermediate pens. These are also the pens to which the majority of the hatchlings will be transferred in December, at an average age of nine months. The intermediate pens cover an area of 12.6m2 of which approximately 35% is dry concrete that slopes into water with a maximum depth of 20cm. There are two different intermediate pen designs as shown in Figure 2.9. The first layout includes a concrete division within the water to allow refuge for the animals (Figure 2.9A), whilst the other layout has no division (Figure 2.9B). In both cases, a hide-board is provided for added security. Animals are stocked to a density of 0.32m2/animal in each of these pens. Water constantly flows into the pens at the temperatures mentioned above for the tubs. The pens are exposed to ambient air temperatures and natural diurnal photoperiods although the whole area is covered by shade-cloth to reduce exposure to direct sunlight.

16

Figure 2.9. Two intermediate pen designs allow the animals to hide, and feel less threatened, during cleaning and other procedures. A) the pen is split into two water sections by a concrete subdivision. The animals tend to hide in the back subdivided area whilst the front section is drained for cleaning. B) no subdivision of water area is present in these pens, so the hideboard is very important to provide sanctuary for the animals.

Depending on size, at about 1-1.5 years old (or 0.6 - 0.9m in total length), the animals are relocated into channels for approximately 1-1.5 years until they reach slaughter size. The channels derive their name from the two “channels” of water contained within each pen (Figure 2.10). They are 8 × 25m (200m2) with stocking densities maintained at approximately 1m2/animal. One of the channels within each pen, termed the “flow” side, has water constantly flowing through it, again at temperatures between 29 and 31oC, depending on the time of year.

Figure 2.10. Slaughter size crocodiles inside a channel pen. Clearly visible are the two “channels” of water from which the pens derive their name.

2.1.5 Stocking densities and grading Stocking densities are extremely important, not only to minimise food competition but also to minimise fighting and, therefore, maintain higher skin grades. Minimising the size range of the animals within each pen is also imperative in maintaining an efficient production system. If there is a large size disparity between the animals within a pen, the larger, more dominant animals will inhibit the smaller members of their cohort from eating to full capacity, further stunting their growth and leading to possible mortality. A large size disparity between the animals also increases the probability and severity of damage to the skin

1m 1m

3.55m

0.2m

0.75m

1.2m

3.55m

3.55m 3.55m

A. B.

0.2m

1.2m

Hideboard

Scale 1m

Hide-board

Land Land Water WaterWater

17

due to bites, increasing the downgrading of skins and ultimately affecting the final product value. For this reason, animals are constantly graded between pens to even the distribution of size.

2.1.6 Feeding Hatchlings are initially fed five times per week until the youngest clutch is about three months old. Feeding is then reduced to four times a week. The intermediate and channel animals are fed three times a week during the warmer months. Channel feeds are decreased to two times a week for about three months a year during cooler weather. Feed sheets are used to monitor the amount of food that each pen receives. Hatchlings are fed a minced-meat mixture, containing 70% red meat (kangaroo, beef, horse) and 30% chicken heads, with a 1% by weight vitamin supplement (Petvite®, Igy Manufacturing Pty. Ltd., Australia) and 2% by weight dicalcium phosphate. Intermediate animals are fed a coarser mince, consisting also of 70% red meat and 30% chicken heads, but excluding the supplements. The channels containing younger animals receive a mash of equal proportions of red meat and chicken heads. As the channel animals become larger, they are gradually weaned onto chicken heads alone.

2.1.7 Feeding routine and cleaning The animals are fed late in the afternoon, as this is better suited to the management of the farm. Food is fed along the waterline within the pens. This helps simulate natural feeding behaviour of charging out of water to grab their prey on the bank. Particularly when encouraging hatchlings to begin feeding, it is important to break the food into small pieces. The chunks of meat get proportionally larger as the crocodiles grow. The following morning any excess food is removed from the pens and the pens are hosed and scrubbed. The nursery pens and tubs are scrubbed with Kwiksan (active constituent Benzalkonium Chloride; Campbell Brothers Ltd, Qld). Intermediate pens and channels are “medicated” with chlorine (using sodium hypochlorite; Pool Resources (SA) Pty Ltd, SA) twice weekly. The chlorine is used mainly to prevent algal growth, as slipping represents a potentially life threatening occupational health and safety risk whilst working in the pens.

2.1.8 Disease management The main disease problems on the farm occur in the first year within the nursery and tubs, although some outbreaks occur to a lesser extent in intermediate pens and channels. Outbreaks of fungal disease (mycotic dermatitis or superficial mycosis) represent a particular problem in the tubs. The most common isolate of the disease pathogen found on farms is Fusarium sp., although other fungi have been identified including Candida sp, Syncephalastrum sp, Aspergillus sp, Trichosporon sp, Penicillium sp, and Curvularia sp (Buenviaje et al., 1994, 1998). Buenviaje et al. (1994) concluded that because of the superficial nature of the infestation, it was not possible to confirm the primary fungus of infection. The typical gross characteristic is a grey, gelatinous appearance of affected skin. Lesions develop over a period of one week usually beginning on the feet, although as the disease progresses the lesions can be present at any location on the body but predominantly affecting the dorsal skin of the head (Buenviaje et al., 1998; personal observation). If detected early, mortality rarely occurs. Treatment is relatively easy, requiring the addition of formalin to a final concentration of 10% v/v in the tub or pen water. Formalin is administered during the day when the water flow is turned off. At night, the water flow is turned back on to flush away the formalin. Treatment is usually only required for one week. However, if the disease is not detected early, treatment must be continued for a longer period and high mortalities can result. Since 1998, there have been outbreaks of bacterial septicaemia in both the nursery and, to a smaller degree, in the tubs. These outbreaks have caused high mortality, particularly of those animals in good condition and showing relatively better growth rates. Gram-negative bacteria are most often involved, suggesting that endotoxaemia is an important aspect of pathogenesis, both in acute and chronic illness (Shotts, 1981; Ladds and Sims, 1990). Shotts (1981) suggested that similar cases of endotoxaemia in

18

alligators resulted as a secondary infection or opportunistic nature, induced by inappropriate husbandry particularly related to temperature, feeding or overcrowding. In contrast, Ladds and Sims (1990) suggested that as acute bacterial septicaemia can be diagnosed in crocodiles in good condition, the disease may be caused by a primary infection of the relevant pathogen(s). Husbandry practices definitely influence the immunological status, although under Janamba’s management regime, crocodiles appear to beat the risk of septicaemia until another stress factor, such as grading, is introduced. Ladds and Sims (1990) suggested that most animals that die are the best-growing animals, with poorer-growing animals rarely affected, since the superior-growing animals experience a relatively higher level of physiological and nutritional stress. Unfortunately most of the bacteria that cause the disease are ubiquitous and opportunistic. Hence, it is impossible to eradicate them. However, minimising their impact is achievable. The most common isolates occurring at Janamba include Edwardsiella tarda and Providencia (Proteus) rettgeri, although Salmonella spp, Pseudomonas spp and other Proteus spp have also been found (bacteriology reports, Berrimah Veterinary Laboratories, Darwin). The septicaemic onset has no definite signs, although the animals do appear lethargic in comparison to normal. As the septicaemia progresses, neurological symptoms can be observed, such as standing on land with the head elevated and swaying from side to side, and swimming in circles (Ladds et al., 1996; personal observations). Pathology reports often reveal oedema of the subcutis and other tissues (eg pharyngeal, perirenal), which is frequently associated with clear or bloodstained fluid in body cavities (pathology reports, Berrimah Veterinary Laboratories, Darwin; Ladds and Sims, 1990). The liver and kidneys also often appear pale. The stomach is usually empty. Treatment is dependent upon the bacteria involved and their susceptibility to various antibiotics. Antibiotics are administered orally within food, as other methods such as injections, etc. are too laborious and stressful on the animals, especially when there are large numbers of animals involved. Unfortunately, oral administration limits the choices of antibiotics available. The ones commonly used at Janamba are Trimidine (sulfadimidine (430mg/g), trimethoprim (86mg/g)) and Terramycin (oxytetracycline hydrochloride (50mg/g), neomycin sulphate (50mg/g)). Amoxocillin is an alternative antibiotic that Janamba has not yet required. With only a limited number of antibiotics available, care must be taken to minimise the development of antibiotic resistance. Hence, sensitivity tests after bacteriological isolation are used to identify the most effective choice of antibiotic. Since the onset of this disease problem, certain husbandry procedures have been introduced to attempt to reduce the incidence, especially within the nursery. Skylights have been installed and the shed ventilation improved. The shed shutters are opened and closed, as determined by the external weather conditions, throughout the day. Chlorine is also put in the drainage gutters twice a week to sterilise the places that cannot be manually cleaned. Every year, the nursery and tubs are thoroughly cleaned, and then kept empty for about four weeks to break any disease cycles that are present. Within the nursery, this includes thoroughly cleaning the inside walls, grates and gutters. In addition, a couple of days before grading or other management practices are undertaken, water temperatures are reduced within the nursery to overcome immunosuppression caused by high temperatures (Lang, 1987).

2.1.9 Slaughtering and processing In general, animals are slaughtered between 1.6m and 1.9m in total length, or two to three years of age. This corresponds to a belly width between 35 and 45cm. Prior to 2001, Janamba slaughtered and processed its own animals using an on-site abattoir. A change of management procedures has made the majority of the following descriptions obsolete, and confidentiality prohibits further description of current procedures. Animals were killed with a low-velocity, short 0.22-calibre bullet into the brain from the posterior end of the skull. This caused minimal disturbance to animals in a pen. They were then pithed (spinal cord and blood vessels severed behind the skull), washed with 20% hypochlorite and hung overnight in a chiller (5oC). The following morning, the animals were weighed, and their head length and total length measurements were recorded. Each skin was given a tag number, issued by Environment Australia (EA; previously Australian Nature Conservation Agency or ANCA) in accordance with CITES regulations.

19

This tag number must remain with the skin until it is processed, so that the animal can be identified in relation to its species, country of origin and the farm in which it was processed. After measurements, the animals are skinned and their flesh boned-out. The average amount of flesh derived per animal is 5kg, which is vacuum-packaged into various cuts and frozen for distribution. The main cuts of retail flesh are tail-fillet (inner and outer), back-fillet, legs, satay and body flesh, although carcase orders can be specified. Heads, feet, backstraps and tail tips were all frozen ready to be made into curio items to be sold to the tourist trade. The majority of skins produced are sold as “belly” skins in the Australian industry. The aim is to maximise the uncut area of the regularly arranged scales on the animals’ ventral surface. After removal, skins need to be cleaned on the inside by a water blaster (150kPa) to remove excess tissue. They are then cured with salt (NaCl). Curing aims to preserve the skin’s protein-fibre structure so that it can be restored to something like its original state before tanning is commenced (Van Jaarsveldt, 1987). However, the aim is not to completely dehydrate the skins as then they become brittle, can suffer from grain cracking (skin between the scales crack) and are difficult to rehydrate for tanning (David, 1987). Curing must also be sufficient to preserve the skin structure during storage and transit, which can sometimes be for considerable periods. Spoilage of skins by micro-organisms prior to tanning is a major concern as “scale-slip” can occur, downgrading the value of the skin. An advantage of using salt is that it retards most bacterial growth (Van Jaarsveldt, 1987). Halophilic bacteria are a major concern due to their salt tolerance. However, by keeping the skins in a cool environment, with adequate air circulation, bacterial spoilage is negligible. After salting, skins are rolled firstly by folding the four limbs and flank area inwards, and then tightly rolling starting from the tip of the jaw area to the end of the tail. This allows the valuable belly area to remain flat. The skins at Janamba were then stored in crates of fifteen to allow adequate air circulation in a chiller at 0oC. The following day the skins are resalted with clean, dry salt, as the old salt will already be at saturation. The skins are given a third, light resalt prior to shipping, usually within a month after processing.

2.1.10 Skin grading Skins are assessed by an objective measure of quality and also a quantitative measure of belly width. The market specifies the sizes required: generally between 35-45 cm is desired. The skin is measured between the third osteoderm down the flank from the forelegs on either side without stretching the skin (Figure 2.11). The quality of the skin is the most important factor, as it will ultimately affect the quality of the finished product. There is no standard way of assessing a skin, as the specifications change between the purchasers: a skin rejected by one purchaser may be accepted by another. However, basic guidelines are available. Basically, grading classifies skins into one of four grades: first, second, third or fourth. The main determinants influencing the grade classification of a skin are the number, location and severity of any scars, scratches and holes. A first-grade skin is clear of all damage within the belly region, although one or two small holes may be present outside the belly region. The skin should be well preserved with no flesh on the underside, contain four legs and have two rows of osteoderms from the lateral surface on both sides. A second grade skin meets all of the first grade requirements, except that it will have an imperfection such as a hole somewhere within the belly region, have scales lifting from the leather, or have one appendage missing. A third grade skin has two or more of the conditions that rejected it from the second grade classification. Fourth grade skins have no substantial area unblemished and are rejected.

20

Figure 2.11. Diagram of a crocodile belly skin with the features used in measurement and grading, in which the most valuable area is the belly pattern (from Van Jaarsveldt, 1987).

2.2 Structure of family groups As mentioned previously, there are two types of breeding pen environment at Janamba Croc Farm where the sires and dams can be correctly identified as the parents of particular clutches. It was from these pens, the B- and UB-pens, that the data were collected in family groups for inclusion in this study. The majority of these animals were housed in unitised (1 male, 1 female) breeding pens. Only two pens (UB04 and UB06) housed 1 male and 2 females. Records of half-sibling performance were available in these pens. In addition, since the animals were housed in pairs where both the male and female parents of a clutch could be identified, the effects of sire and dam could not be estimated independently but instead were collectively estimated as a pair effect. In total, there were 30 family groups used in this study. With the exception of one pair, all of these adult animals were wild-harvested. Thus, no information regarding their age or breeding history is known. However, these pairs had been housed at Janamba for a substantial number of years (Table 2.3). As a result, a considerable number of offspring records have been collected for use in this study. The exception was one pair that was captive-raised, in which the female was bred at Janamba Croc Farm (born in 1993), whilst the male was from a wild-harvested clutch of eggs raised in captivity (born May, 1989). All breeding animals were assumed unrelated for the purposes of this study.

21

Table 2.3. Summary of male and female identification numbers used in this study, along with the year when the animals were first put together. Data used in this study were collected between 1994 and 2002, with the last column (Nesting Data Since) indicating when data collection started for each pair.

Male ID Female ID Pair No. Together Since Nesting Data

Since B01M B01F 1 1991 1994 B02M B02F 2 1991 1994 B03M B03F 3 1991 1995 B04M B04F 4 1993 1995 B05M B05F 5 1996 1997 B06M B06F 6 1998 2000 B07M B07F 7 1991 1994 B09M B09F 8 1997 1999 B10M B10F 9 1987 1994 B11M B11F 10 1994 1995 B12M B12F 11 1999 2001 B13M B13F 12 1987 1994 B14M B14F 13 1991 1995 B15M B15F 14 1987 1994 B16M B16F 15 1992 1994 B19M B19F 16 1994 1996 B20M B20F 17 1996 2000 UB01M UB01F 18 1991 1997 UB03M UB03F 19 1997 1998

UB04-1F 20 1991 1995 UB04M UB04-2M 21 1991 1994

UB05M UB05F 22 1990 1995 UB06-1F 23 1991 1997 UB06M UB06-2F 24 1995 1999

UB07M UB07F 25 1993 1996 UB08M UB08F 26 1998 1998 UB09M UB09F 27 1992 1996 UB10M UB10F 28 1991 1994 UB13M UB13F 29 2001 2001 UB16M UB16F 30 1994 1995

22

2.3 Traits and observations The traits analysed in this study were reproductive, production, survival and skin quality traits. The first part of this section briefly discusses each of these traits that were available for analysis. It should be noted that all data were excluded from an animal if there was any ambiguity with scute cuts, parent identity, or uncertainty when cross-checked against other records. The second section outlines the explanatory variables that were available for inclusion in the models.

2.3.1 Reproductive traits Seven reproductive traits were recorded from the females, namely

- initial clutch size (ClSize): total number of eggs collected in a clutch, - number of viable eggs (NoViable): ClSize minus the infertile eggs and those that died before

collection, - number of live, healthy hatchlings (NoHatch): NoViable minus those that died during

incubation or were euthanased due to abnormalities (lethargy, kinked spine, external yolk sac, external organs, etc),

- hatchability (HatchR): NoHatch as a proportion of ClSize, - average snout-vent length (AvSVL)1: random sample of 10 hatchling snout-vent lengths

(~30%) from each clutch, to indicate hatchling size, - hatch days (HDays): number of days between hatching date and the 1st of January in that

particular year, and - nesting (Nesting): whether the female nested or not in a particular year (a binary trait where 0

= no and 1 = yes) The data available had been collected between 1994 and 2002. The binary trait Nesting was analysed separately from all the measurement-variable traits. Further details of the characteristics of the data will be given in Chapter 5. The lack of pedigree structure of the breeding animals, combined with only one generation for analysis, prohibited a genetic study of the reproductive traits. However, estimates of repeatability and phenotypic correlations were derived, providing a benchmark for future studies. This is discussed further below.

2.3.2 Production traits Morphological measurements were taken on juveniles at three stages through the production system. The major trait of interest was the age of the animal at slaughter, or expressing it as a selection objective, decreasing the time to reach slaughter size. However, it may be advantageous to use earlier morphometric records (at hatching and/or inventory) as an aid to selection at a later stage. Further details of the characteristics of these data are given in Chapter 6.

1 In the context of the reproductive traits, AvSVL is an index of reproductive output.

23

2.3.2.1 Description of morphological traits A description of each of the measured traits, as defined by Webb and Messel (1978), is given below. A more thorough description of the stages of measurement are given in the following sections. (1) Head length (HL, mm): (Figure 2.12A) Tip of snout to median posterior platform of supraoccipital

bone. Animals measured at hatching and during the yearly inventory were measured with digital callipers to the nearest 0.01 mm. Animals measured in the abattoirs were measured to the nearest 5 mm with a flexible tape measure.

(2) Snout-vent length (SVL, mm): (Figure 2.12B) Tip of snout to anterior proximity of cloaca. Only animals at hatching and at the yearly inventories were measured for this trait. A clear, plastic ruler is placed at the tip of the snout and the measurement was taken to the nearest 1mm.

(3) Total length (TL. mm): (Figure 2.12C) Tip of snout to tip of tail. The snout was butted up against an inlaid ruler and a measurement to the nearest 1 mm was taken to the tip of the tail. Animals in the abattoir were measured to the nearest 10 mm with a flexible tape measure.

(4) Bodyweight (Bwt, kg). Measured to the nearest 0.1 kg using a Salter spring balance (25kg ± 100g). (5) Bellywidth (CullBwidth, mm). Measured between the third osteoderm down the flank from the

forelegs, on either side without stretching, using a metal ruler, rounding down to the lowest centimetre.

Figure 2.12. Measuring morphological traits on crocodiles. A) Measuring head length (HL) using digital callipers. B) Measuring snout-vent length (SVL) using a clear, plastic ruler. C) Measuring total length (TL) against a ruler.

2.3.2.2 Hatchling traits Prior to 2000, a random sample of 10 animals (~30%) from each newly-hatched clutch was measured for snout-to-vent length. However, for the breeding seasons between 2000 and 2002, every individual was measured for hatchling head length (HHL), snout-vent length (HSVL) and total length (HTL). A total of 1607 hatchling head length measurements were available for analysis.

A

B

C

24

2.3.2.3 Yearling inventory traits An annual inventory is conducted at Janamba Croc Farm in either December or early January (average age 265.1 days; range 163-341 days). Inventory head length (InvHL) measurements have always been made on all juveniles. These measurements have been made available from 1994 to 2002 (n = 4859). In addition, 6-8% of animals are also measured for SVL (InvSVL) and TL (InvTL) in each year, although all juveniles from the 2001 cohort were measured for all three traits. Overall, 20% of the juveniles had observations on all three traits.

2.3.2.4 Slaughter traits As mentioned previously, crocodiles are generally slaughtered between 1.6 and 1.9 metres to obtain a belly width between 35 and 45 cm. However, the decision to select an animal for slaughter is based upon visual assessment. Therefore, there is substantial variation in the resultant belly-widths. In addition, management decisions to slaughter poor-growing animals further contribute to the range of belly-widths obtained. Although the major trait of interest at slaughter is the individual’s age, other morphological measurements have also been collected including head length (CullHL) and total length (CullTL). Later, belly width (CullBwidth) is measured during skin grading. Data for these traits were available from 1994 to 2001 cohorts.

2.3.3 Survival traits To conduct a survival analysis, the fate of each individual within a clutch needs to be known. There are three alternatives: the animal died, was slaughtered or still remains in the production system. Mortality was recorded on a daily basis during routine farm procedures. These data included the date of death and a description of the crocodile’s condition upon death and/or a possible cause of death. Abattoir records provided data for those animals slaughtered for processing. Additionally, those animals that had neither died nor been slaughtered, that is were still alive in the production system, had dummy records created for them. These dummy records were then censored at the last date of data collection (December 2, 2003). A binary censoring status variable was then constructed for these data. Those animals that died were coded 1. Alternatively, those animals that were slaughtered or had dummy records created were coded 0, since they did not die as a result of genetic or environmental reasons. Data were available for clutches hatched between 1994 and 2002. A total of 5035 records were available for analysis consisting of 1302 mortality records, 2151 slaughter records and 1582 dummy records.

2.3.4 Quality trait – Number of scale rows These data were collected from two different facilities using two different techniques. There were 1739 offspring records collected at Janamba Croc Farm from 30 pairs (average number of offspring per pair = 58; range 23-119) in 2001 and 2002. A second set of data was kindly provided by Wildlife Management International (WMI). It was originally collected for the study by Manolis et al. (2000). Data collection continued after the completion of the Manolis et al. (2000) study, and so data from 1999 and 2002 inclusive were available. Only offspring incubated at 32oC and with parental records were included in these analyses. There were 1467 WMI individuals available for analysis from 19 known family groups (19 sires; 18 dams) with an average of 75 offspring per family (range 23-114). A subsequent total of 3206 crocodiles were available for analysis from the two facilities. A more thorough description of the scale row counting methods is given in Chapter 8.

25

2.3.5 Explanatory variables Outlined below are the explanatory variables available to model the traits listed above. Mixed-model analysis was used extensively in these studies, and involves the separation of explanatory variables into fixed or random effects (more detail of mixed-model methodology is given in the next section). The explanatory variables described below were included into the fixed-effects part of the mixed model. Table 2.4 indicates the fixed effects and covariates available for inclusion with each performance trait. More specific descriptions of the explanatory variables can be found for each performance trait (reproductive, production, survival and quality) in the relevant chapters. 1. Year: This was the year in which a particular clutch hatched. For the reproductive, productive and

survival traits, nine years of data were available. The quality trait, number of scale rows, was collected at two facilities. There were two years of data collected at Janamba Croc Farm and four years collected from Wildlife Management International.

2. Pen: This effect was fitted to account for any variation between the two breeder-pen environments (B-pens or UB-pens) and, therefore, had two levels.

3. EnvI: The effect of initial rearing environment on the juvenile animals. This effect had two levels: tubs and nursery, and was only available for the inventory analysis of traits.

4. Gender: Records of the gender of an individual were only available for those animals that reached slaughter size (CullAge). Gender was determined in the abattoir by manual palpation for the presence of a penis. These were coded 1 for males and 2 for females.

5. Collected (Coll): This fixed effect was used only for the reproductive traits, with the exception of the traits ClSize, HDays, and Nesting, to describe the number of days between when the eggs were laid (oviposition) and when they were collected for incubation (ten discrete levels). The levels, or number of days between oviposition and collection, were estimated using the band markings described by Richardson, et al. (2002).

5. Egg dimensions: An average of a random sample of 10 egg lengths (EL) and 10 egg widths (EW) from each clutch. In the context of the reproductive traits, previous biological studies have indicated that female size, and age, is strongly correlated with various egg dimensions (Thorbjarnarson, 1996). It was hoped that these EL and EW measurements may act as proxies for female age since the age of females in this study was unknown.

6. InvAge: The age of the animal at inventory was used during the analysis of InvHL, InvSVL and InvTL. Using the clutch-specific scute cuts, the individual’s age was determined as the number of days between the date of hatch and the date of inventory.

7. HDays: Besides being a reproductive trait in a slightly different context, this explanatory variable was used during the analysis of the productive traits to adjust for the climatic differences that occur as hatching proceeds later into the year. The variable is the difference between the date of hatch and the 1st of January in that year.

8. SOI(m): This was a specific explanatory variable created for the reproductive trait, Nesting. SOI(m) is the southern oscillation index for month m in that year with m being either June, July, August, September, average of June and July or average of July and August. The data was obtained from the Australian Commonwealth Bureau of Meteorology website (http://www.bom.gov.au/climate/current/soihtm1.shtml). This variable was included since previous studies have found strong correlations between the SOI and the probability of females nesting in particular years.

2.4 Statistical analysis Since the breeding pairs were kept in unitised pens, separate sire and dam effects could not be estimated. Instead, the term “Pair” was used to estimate the genetic variance and covariance components for the reproductive, production, survival and quality traits. The models were derived from the adaptation of the following general linear model.

26

SOI

HD

ays

InvA

ge

EW

EL

Col

lect

ed

Gen

der

EnvI

Pen

Fixe

d ef

fect

Yea

r

Mod

el ty

pe

Line

ar

Line

ar

Line

ar

Line

ar

Line

ar

Line

ar

Bin

ary

Line

ar

Line

ar

Line

ar

Cox

’s p

ropo

rtion

al

haza

rd

Line

ar

Trai

t

ClS

ize

NoV

iabl

e

NoH

atch

Hat

chR

AvS

VL

HD

ays

Nes

ting

Hat

chlin

g

Inve

ntor

y

Cul

lAge

Surv

ival

Num

ber o

f sc

ale

row

s

Tab

le 2

.4. S

umm

ary

of f

ixed

eff

ects

and

cov

aria

tes

used

in th

e an

alys

is o

f th

e pe

rfor

man

ce tr

aits

. The

tick

()

indi

cate

s th

at th

e ef

fect

was

in

corp

orat

ed in

to th

e un

ivar

iate

mod

el f

or th

e tra

it, a

lthou

gh d

oes

not i

ndic

ate

sign

ifica

nce,

whi

lst t

he c

ross

()

indi

cate

s th

at th

e ex

plan

ator

y va

riate

was

eith

er n

ot a

vaila

ble

for i

nclu

sion

or n

ot a

ppro

pria

te.

Reproductive Production Survival Quality

27

2.4.1 The linear mixed model The general statistical linear mixed model, in matrix notation, used in these analyses is εZuXβy ++= [2.1] where y is a vector of n observable random variables (traits of interest), β is a vector for all fixed effects, u is a vector of random effects, X and Z are known design or incidence matrices for the fixed and random effects, respectively, and ε is the vector of random residual error terms. The expectation of y was assumed to be

E⎡ ⎤ ⎡ ⎤⎢ ⎥ ⎢ ⎥=⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦

y Xβu 0ε 0

whilst the variance of the random effects, var(u), and residuals, var(ε), is var(u) = G and var(ε) = R, respectively, where G and R are known non-singular matrices (apart from variance and covariance components); and cov(u, ε’) = 0 (Henderson and Quaas, 1976; Robinson 1991). To simultaneously generate solutions for the fixed effects (Best Linear Unbiased Estimates; BLUE) and random effects (Best Linear Unbiased Predictor; BLUP), Henderson (1976) presented the following mixed-model equations

⎥⎦

⎤⎢⎣

⎡=⎥

⎦

⎤⎢⎣

⎡⎥⎦

⎤⎢⎣

⎡

+ yRZ'yRX'

uβ

GZRZ'XRZ'ZRX'XRX'

1-

-1

1-1-1-

-1-1

ˆ

ˆ

which need to be solved iteratively for β and u , since R and G depend on variance and covariance parameter estimates. REML is used to estimate the (co)variance parameters (Patterson and Thompson, 1971).

2.4.1.1 Fixed effects The fixed effects were outlined in Section 2.3.5, and Table 2.4 indicated the fixed effects and covariates available for inclusion in the vector β for each performance trait. To evaluate fixed effects, Wald test-statistics were used, which are a generalisation of a z-test for testing multiple parameters. The test statistic has an asymptotic chi-squared probability distribution (Dobson, 1990). A 5% significance level was chosen to evaluate explanatory variables by backward elimination.

2.4.1.2 Random effects As mentioned previously, the main genetic effect of interest in these analyses is Pair. In addition, previous studies on crocodilians have noted strong clutch effects influencing traits such as growth rate, survival and time to reach slaughter size. To account for these clutch effects, an interaction term between Pair and Year was used. In conjunction with the Pair random term that accounted for the genetic effects, the interaction term adjusted for pair effects nested within year, alternatively known as a common-environment effect. In the future, as pedigree structures become more advanced in the crocodile industry, separate sire and dam effects will be able to be estimated. Until then and for the purposes of this study, the vector u in model 2.1 consists of pair and clutch random effects. To extend model 2.1, the mixed model with these two random effects is: y = Xβ + Z1p + Z2c + ε [2.2] where y is the vector of observations, β is the vector of fixed effects, p is the vector of pair genetic effects, c is the vector of common environment effects (Clutch), and ε is the vector of residual effects. It was assumed that these random effects were uncorrelated, i.e. cov(p,c) = 0 and var(c) = 2

cσI , where 2cσ

is the common environment variance. This gives:

28

⎥⎦

⎤⎢⎣

⎡=

cp

u , and

2

2p

c

⎡ ⎤σ= ⎢ ⎥σ⎣ ⎦

P 0G

0 I

where I and P are the identity matrix and matrix of genetic relationships between animals, respectively, and 2

pσ is the genetic variance.

2.4.2 Types of models As an extension of model 2.2, two types of general analyses were performed. Firstly, all traits were analysed using a univariate analysis. For the measurement-variable traits, a linear univariate mixed-model approach was used. However, a binomial model was used for the binary trait, Nesting, whilst Cox’s proportional hazards model was used for the survival trait. Second, the measurement-variable traits and their significant univariate explanatory variables were combined into a multivariate analysis. The univariate and multivariate linear, and binomial mixed models were fitted using GenStat 6 (version 6.1, 2002) and ASReml (version 1.10, 2003) software, whilst Survival Kit V3.12 (Ducrocq and Sölkner, 1994, 1998) was used to fit Cox's proportional hazard models. The significance of each random term was evaluated using a likelihood ratio test (Dobson, 1991). Likelihood was produced for the linear models using restricted (or residual) maximum likelihood as described by Patterson and Thompson (1971). For the binomial trait, an approximate linear likelihood was produced using the method of Gilmour et al. (1985) and Schall (1991), whilst a partial likelihood function was used for Cox’s proportional hazard models as described by Cox (1972) and Ducrocq and Casella (1996). To test the significance of the random effects, the difference between -2 × log likelihood of the full model (includes random term) and reduced model (excludes random term) is compared to a chi-square distribution with 1 degree of freedom. A 5% significance level was chosen for evaluating all random effects.

2.4.2.1 Linear model analyses Univariate models for all linear traits (reproductive, production and quality) were initially evaluated in GenStat 6. A 5% significance level was chosen to evaluate explanatory variables by backward elimination. The resultant univariate models were then combined into a multivariate analysis using ASReml to evaluate intraclass correlations between traits. The following form of multivariate model was fitted

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡+

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡+

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡=

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡

3

2

1

3

2

1

3

2

1

3

2

1

3

2

1

3

2

1

εεε

uuu

Z000Z000Z

βββ

X000X000X

yyy

where y1, y2 and y3 are the vectors of phenotypic observations of length n1, n2 and n3, representative of the number of records for each trait; X1, X2 and X3 are incidence matrices of 0s and 1s relating the records and fixed effects; β1, β2 and β3 are the unknown quantities of each fixed effect including the underlying means for each trait; Z1, Z2 and Z3 are incidence matrices relating the records and random effects; u1, u2 and u3 are vectors of the unknown additive-genetic effects; and ε1, ε2 and ε3 are the vectors of the unknown random errors.

29

The variance structure of the random elements in the model had the following form

⎥⎥⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢⎢⎢

⎣

⎡

=

⎥⎥⎥⎥⎥⎥⎥

⎦

⎤

⎢⎢⎢⎢⎢⎢⎢

⎣

⎡

23

223

213

232

22

212

231

221

21

23

223

213

232

22

212

231

221

21

var

ε

ε

ε

ε

ε

ε

ε

ε

ε

g

g

g

g

g

g

g

g

g

σσσ

σσσ

σσσ

σσσ

σσσ

σσσ

III

000

III

000

III

000

000

GGG

000

GGG

000

GGG

εεεuuu

3

2

1

3

2

1

where G is the numerator relationship matrix; 21gσ , 2

2gσ , and 23gσ are the genetic (co)variance

components for the traits being considered; and 2ε1σ , 2

ε2σ , and 2ε3σ are the residual (co)variance

components of the traits. It was the variance components produced from the multivariate analysis that were used to derive the quantitative parameter estimates reported in this study.

2.4.2.2 Binomial model analysis The only binomial trait analysed during this study was whether a female nested or not in a particular year, Nesting. The responses were No/Yes and coded as 0/1. The model was again initially evaluated in GenStat 6, but ASReml was used to obtain the final variance component estimates. The generalised linear mixed model form is

log ' '1

ii i

i

pp

⎛ ⎞= +⎜ ⎟−⎝ ⎠

x β z u

where pi is the probability of an event occurring, in this case of a particular female nesting in a particular year.

2.4.2.3 Cox’s Proportional Hazards analysis Cox’s proportional hazards model was considered the most appropriate model for analysing the survival data. Mixed-effect models in survival analyses are known as frailty models, and Cox’s proportional hazard model makes no assumptions of the baseline hazard function. Survival kit V3.12 (Ducrocq and Sölkner, 1994; 1998) was used to evaluate the following frailty model:

0log ( ) log ( ) ' 'ij ij ijh t h t= + +x β z u where hij(t) is the hazard function for the jth individual from the ith pair from the lth clutch at time t, and h0(t) is the unspecified baseline hazard function. This type of analysis allows for deaths to occur over a continuous time scale and allows records to be censored if the individual has not died before the study period has ended (Southey et al., 2001).

2.4.3 Heritability and repeatability estimates Estimating parameters for the measurement-variable traits varied due to certain data constraints in each data-set. A brief definition of the parameters estimated is given below, although a more appropriate description is given in the respective chapters for the reproductive, production and quality traits. For this chapter and subsequently, 2

Pairσ is the pair variance component, 2Clutchσ is the clutch (common

environment) variance component, and 2εσ is the residual, or error, variance component.

Measurement-variable reproductive traits- Only repeatability could be estimated for these traits due to the

lack of pedigree structure and/or generational data. Repeatability was estimated as

Repeatability = Rep = 22Pair

2Pair

σσσ

ε+

30

Production and quality traits- All heritabilities (production, survival and quality characters) were derived using full-sibling correlations. Falconer and Mackay (1996) defined heritability to be t/r where t is the correlation between the relatives and r is the relationship. In the case of full-sibling data, r is ½, so the heritability estimate using full-sibling data is

h2 = 2 × full-sib correlation Using the variance components from the fitted mixed model,

HeritabilityFS = h2FS = 22

Clutch2Pair

2Pair

σσσσ

2ε++

×

Approximate standard errors (SE)- Standard errors for both the heritability and repeatability estimates

were derived using the following equation (Van Vleck, 2000)

SE(h2) = SE(Rep) = ( )[ ]

)1)((11)1)(1(22 2

22

−−−+−−

FFNktktN

where )1(/1

2 −⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛−= ∑

=

FNnNkF

ii , ni is the number of individuals within each family

group, ∑=

=F

iinN

1= total number of observations in the trait, F = number of families and for

heritability t = h2FS/2, whereas for repeatability, t = Rep (the repeatability estimate).

Phenotypic correlations- Again there were slight differences in the estimation of phenotypic correlations

between each data set. For the reproductive traits, say Y1 and Y2, the phenotypic correlation )( pρ was obtained as follows (Searle, 1961)

))((

),cov()Pair,Paircov(22

Pair22

Pair

2121

2211 εε σ+σσ+σ

εε+=ρ p

where )Pair,Paircov( 21 and ),cov( 21 εε is the covariance between traits 1 and 2 for the pair and residual terms, respectively. In addition, clutch (co)variance components were estimated for the production and quality data sets

Genetic correlations- These were estimated only for the production and quality traits. For two traits Y1 and

Y2, the genetic correlation )( gρ was obtained as follows (Searle, 1961)

)(

)Pair,Paircov(2Pair

2Pair

21

21σ×σ

=ρg

Binary reproductive trait- Repeatability was estimated using an extension from Lynch and Walsh (1998; Thomson, pers. comm.)

Repeatability = ( ) 3/π22Pair

2Pair

φ+σσ

where φ is the dispersion parameter included to manage any non-binomial variation. There are no readily available standard errors associated with binomial traits.

Survival- Heritability of juvenile survival was estimated on the log-hazard scale using a full-sib correlation as derived from the half-sib heritability estimate defined by Ducrocq and Casella (1996)

62Clutch

2Pair

2Pair2

logt 22hπ+σ+σ

σ×=

Approximate standard errors were derived as defined for heritability and repeatability above from Van Vleck (2000).

31

3. Industry perspectives on breeding objectives for saltwater crocodile genetic improvement programs

3.1 Abstract Genetic improvement of crocodiles to improve production efficiency is a novel concept. To quantify the level of industry support for such an initiative, a survey of the Australian crocodile industry was conducted. Forty questionnaires were sent to industry members (from farm managers to product manufacturers), which asked participants to assess potential breeding objectives on a scale of one (least important) to five (most important). The potential objectives included reproductive traits (number of hatchlings per female, nesting frequency, timing of nesting, and age of sexual maturity), survival rate, food conversion efficiency, and age at slaughter. The main source of income for crocodile farms is the sale of skins, with meat as the major by-product. Skins are sold according to an industry standard grading system, which currently does not include skin quality traits. In the future, skin “quality” (scale row number and regularity, shape and thickness) could also become important and was therefore included in the list of potential objectives. Nine responses were received and provide an indication of the priority that should be given to each objective. Interestingly, some participants indicated they had already begun selecting candidates for traits including growth, survival and skin tensile strength. Overall, the survey provided encouraging support for the adoption of an improvement program.

3.2 Introduction Crocodile production for the skin trade is an emerging industry in Australia. To date, research in the industry has concentrated on improving husbandry techniques, such as nutrition and housing, to improve production efficiency. No research has yet been conducted into the possibility of genetic improvement in captive breeding populations. Isberg et al. (2003) suggested breeding objectives for implementing an improvement program on crocodile farms. Those included were reproductive traits, juvenile survival, food conversion efficiency and age at slaughter. One major issue currently facing the Australian crocodile industry is the lack of skins meeting the requirements of first grade, blemish-free skins (Manolis et al., 2000; MacNamara et al., 2003). Therefore, skin grade was also included as an objective which has direct impact on farm income. In addition, skin quality traits were included in the list of objectives although economic incentives are not yet offered for these traits. This chapter presents the results of a survey of members of the Australian crocodile industry, designed to identify priorities for increasing production efficiency on Australian crocodile farms.

3.3 Materials and methods

The questionnaire was distributed in November, 2002 to all relevant industry members, including farm managers (of both commercial and tourist farms), tanners, government officials and researchers. As the Australian crocodile industry is small, only forty questionnaires were sent out. The questionnaire was designed as a “closed” survey that asked participants to rank (on a scale of 1 to 5) a provisional list of breeding objectives and to suggest other potential objectives. The traits provisionally included in the breeding goal were those described by Isberg et al. (2003) and were classified into three groups: 1) reproductive performance (number of hatchlings per female, frequency of nesting, time of nesting and age of sexual maturity), 2) production (survival, food conversion efficiency and age at slaughter), and 3)

32

quality (skin grade, number of scale rows and skin thickness). The list and format of the closed questions is shown in Table 3.1.

3.4 Results and discussion

There were twelve respondents to the survey from the forty distributed, with nine completing the form: two being commercial skin/meat production facilities alone, one solely a tourist facility, five combined tourist and production facilities and one a crocodile researcher. These respondents included seven of the 14 commercial farms in Australia reported by MacNamara et al. (2003). While seven is obviously a small number of respondents, a response of 50% to a survey is regarded as quite adequate (Hager et al., 2002). Overall, responses were positive for a breeding program to increase efficiency. Table 3.1 presents the collective responses as a percentage of respondents that gave a particular score to a breeding objective (importance percentage). An average ranking preference of respondents indicated by the mean score is also presented in Table 3.1. This ranking provides an indication of subjective priority for the objectives.

Table 3.1. Responses to industry survey. Participants were asked to score, on a scale of 1 (least important) to 5 (most important), each objective. Figures under “importance” are the percentage of responses in each category. Mean score gives a priority ranking to each objective.

Importance (%) Mean

n 1 2 3 4 5 Score

Reproductive traits

No. of hatchlings/female 8 0 0 12.5 37.5 50 4.4

Nesting frequency 9 11.1 0 11.1 33.3 44.4 4.0

Timing of nesting 9 22.2 11.1 33.3 11.1 22.2 3.0

Earlier sexual maturity 7 14.3 28.6 28.6 14.3 14.3 2.9

Production traits

Survival 9 0 0 0 11.1 88.9 4.9

Food conversion efficiency 8 0 0 12.5 25 62.5 4.5

Age at slaughter 8 0 0 25 12.5 62.5 4.4

Quality Issues

Improve skin grade 8 12.5 0 0 12.5 75 4.4

Number of scale rows 7 14.3 14.3 14.3 28.6 28.6 3.4

Skin thickness 8 12.5 12.5 37.5 0 37.5 3.4

3.4.1 Reproductive output The maintenance of breeding stock is a substantial cost in crocodile production. Pen construction is a major cost item, particularly where breeding pairs are maintained. Females can produce between 25 and 30 hatchlings per year (Treadwell et al., 1991). The actual clutch size is greater than the number of live hatchlings produced, due to egg infertility and embryonic mortality. If higher hatch rates could be achieved, this would allow a reduction in breeder overheads. From the four provisional objectives for reproductive traits, the most important was number of hatchlings per female (mean score 4.4, Table 3.1). The frequency of a female nesting was the next priority (4.0),

33

followed by timing of nesting (3.0) and then sexual maturity (2.9). These are discussed individually below: i) Selecting for females that produce more hatchlings per year is desirable. The majority of breeding animals on farms were harvested as sexually-mature adults from the wild (Webb, 1989). This creates a problem when trying to estimate genetic parameters as there is no generational data. Instead, only repeated measures on an individual are available. In addition, since the animals are wild-caught, their ages are unknown and there is no reliable technique for estimating age. Age is an important variable when modelling reproductive traits and it has been well documented in the crocodilians that as female size increases (which is related to age), so does the number, size and viability of eggs produced (references within Ferguson, 1985). ii) Nesting frequency is analogous to mating interval in other livestock industries. Saltwater crocodiles in northern Australia breed only once per year, primarily during the wet season between late-October and April (Richardson et al., 2002), when they lay only one clutch. Anecdotal evidence indicates that some animals will miss breeding for one season every third or fourth year. No biological studies have investigated this phenomenon, although factors such as age and climatic conditions may be involved. iii) The timing of nesting during the breeding season influences when the hatchlings will emerge. As indicated above, crocodiles nest over a six-month period (Richardson et al., 2002). Under artificial incubation conditions (32oC; 90-95% relative humidity) the average incubation period is 80 days, which corresponds to a hatching period from February to June (Richardson et al., 2002). In tropical Australia, this means that the hatchlings emerge in either the hot, humid wet season (before April), or the cool, dry season (April-August) (Webb et al., 1987). Eggs that are laid earlier in the nesting season produce hatchlings in the wet season where warm temperatures promote eating, growth and better hatchling survival (Webb et al., 1978). In contrast, temperatures experienced by animals that hatch during the dry season may discourage the onset of feeding and reduce survival rates. Many farms use environmentally controlled rearing conditions particularly for juveniles that hatch later in the season. If hatching was over a shorter period, farm management costs could be reduced. iv) Selecting females that reach sexual maturity earlier was considered relatively unimportant by the survey participants. With the majority of breeding stock on farms being wild-harvested, only a few animals have been raised to sexual maturity in captivity. At present, this trait is relatively unimportant in the breeding program and it would be extremely difficult to investigate due to the lack of data to estimate the genetic parameters. However, it may be worth investigating the variation between maturing females at a future time.

3.4.2 Survival Webb (1989) recommended that achievable survival rates should be at least 95% in the first year (between hatching and one year) and at least 95% between one year old and slaughter (average 3.5 years; range 2-5 years; Treadwell et al., 1991). In reality, very few farms achieve such high survival rates, and there is large variation in this trait. This variation is affected by management regimes, but genetic effects may also impact on this rate. The survey results indicate that producers place a strong emphasis on this trait (mean score 4.9, Table 3.1; the highest of all mean scores).

3.4.3 Food conversion efficiency It has been estimated that 42-45% of the operating costs of crocodile production is accounted for by feed (Treadwell et al., 1991). Consistent with this estimate, the survey participants gave a mean score of 4.5 to this trait (Table 3.1). Treadwell et al. (1991) also reported that a crocodile harvested at 1.5m would have consumed, on average, a total of 120kg of food. Feed costs themselves are extremely variable between farms, mainly due to variation in the location and source of protein. Although food conversion efficiency has been estimated in biological and production-based studies (Garnett and Murray, 1986; Manolis et al., 1989; Webb et al., 1991), the genetic basis has not been analysed. No data for food conversion have

34

currently been collected with a genetic analysis in mind. This is an important trait that requires evaluation in the future.

3.4.4 Age at slaughter The majority of captive-raised juveniles reach harvest size within 3.5 years (Treadwell et al., 1991). Variation in this trait has been reported to occur due to clutch-specific effects (for example Manolis et al., 1989). Treadwell et al. (1991) determined that a decrease in the average age at slaughter from 3.5 to 2.5 years would increase the internal rates of return by 250%, using a small breeding farm model. Industry participants ranked this trait highly, with a mean score of 4.4.

3.4.5 Skin grade Skin grade mainly reflects physical damage to the skin (bites and abrasions). The largest concern to buyers of crocodile skins is the high proportion of skins failing the requirements for 1st grade classification (Manolis et al., 2000). Typically, only around 30% meet this grade (Isberg et al., 2003), although MacNamara et al. (2003) reported up to 50%. It is essential to improve the average grade to meet market demand. The mean score from the survey results for improving skin grade was 4.4. Although this is a relatively high score, it was interesting that this trait was not universally ranked at the maximum score of 5, given its perceived importance. A first grade skin is worth between US$7.50-10.00 per cm belly width (35 to 45cm), whilst a second grade skin is worth about 50-75 percent of first grade skins and third grade about 25 percent (Davis and Peucker, 2001). However, the issue of improving skin grade is complex. Since the industry is still developing, much of the downgrading of skins is likely to be caused by inappropriate management techniques, such as inappropriate stocking densities and pen designs. Crocodile behavior, especially fighting, also affects skin grade. Anecdotal evidence suggests that some clutches are more “aggressive” than others, even at the time of hatching, implying a familial and possibly genetic basis for the differences in aggressive behavior.

3.4.6 Skin quality - a trait of future importance? Saltwater crocodile skins, with their relatively small scales, evenly distributed belly-scale pattern, and lack of bone deposits in the belly scales, attract a premium price in comparison to other crocodilian skins. Skin quality, reflecting inherent properties of the hide such as shape and thickness, is assessed subjectively and is not rewarded under the current marketing system. Manolis et al. (2000) conducted a survey of persons involved in all aspects of the crocodile skin industry, in an effort to understand the issues that face producers of saltwater crocodiles to enhance their product. Four “quality” concerns that could be influenced by genetic selection were raised: skin shape, skin thickness, regularity of scale pattern and number of scale rows between the vent and the neck. Wild-harvested crocodiles are considered by buyers to have a superior skin shape and skin thickness compared with their captive counterparts (Manolis et al., 2000). Skins from the former are narrower per unit length and have less wastage during product manufacture (Manolis et al., 2000). By comparison, captive-raised crocodiles can be obese and show signs of stretching between the ventral scales. Extensive experience in other animal industries has shown that obesity is very amenable to selective improvement. Captive crocodiles grow faster than those in the wild, which may affect the thickness of the skin. At present, a premium price is not offered for skins with superior scale pattern regularity or number of scale rows. Manolis et al. (2000) investigated the inheritance of the number of scale rows but did not report any genetic parameter estimates. It is likely that scale row number and regularity could be considered as future selection objectives if a market reward is introduced for these traits. Despite the lack of economic incentive at present, the average score of survey participants was 3.4 for both skin thickness and number of scale rows, indicating that producers are aware that these could

35

become traits of importance in the future. One producer indicated that they are already selecting hatchlings on the basis of higher tensile strength skins.

3.5 Implications

The responses from the majority of participants were encouraging. Most are supportive of a genetic selection program. Interestingly, some participants reported they had already begun selection for various traits, although genetic analyses of these traits have not been formally conducted. Of the survey participants, two farms are already selecting juveniles on the basis of growth rates, two on skin quality (number of scale rows and skin thickness) and one on disease resistance. From the information received, it appears that most farms are already in a position to begin selecting juvenile replacements and the efficiency of this selection will be improved by availability of genetic parameter estimates. A few producers expressed concerns that improving efficiency through genetic selection is premature and that research should remain focused on husbandry-related issues. However, other livestock industries have shown that maximum improvement of productivity can best be achieved by simultaneous improvement of genetics and husbandry.

36

4. Quantitative analysis of reproduction traits 4.1 Abstract

Repeatability and phenotypic correlations were estimated for saltwater crocodile reproductive traits. No pedigree information was available to estimate heritability or genetic correlations, since the majority of breeder animals on farms are wild-caught. Also, the age of the female breeders could not be accounted for, so egg size measurements were used as proxies. The reproductive traits investigated were clutch size (total number of eggs laid), number of viable eggs, number of eggs that produced a live, healthy hatchling, hatchability, average snout-vent length of the hatchlings and time of nesting. A second data set was also created comprising binary data of whether or not the female nested. Repeatability estimates ranged from 0.24 to 0.68 for the measurable traits, with phenotypic correlations ranging from -0.15 to 0.86. Repeatability for whether a female nested or not was 0.58 on the underlying scale. Correlations could not be estimated between the measurement and binary traits due to confounding. These estimates are the first published for crocodilian reproduction traits.

4.2 Introduction

The Australian crocodile industry relies heavily on the wild-harvesting of eggs for subsequent grow-out of juveniles for the skin market. However, some farms also maintain a population of captive breeding animals with resultant offspring also being placed into the production system. Whilst there have been numerous biological studies conducted to investigate the reproductive characteristics of the saltwater crocodile, the purpose of this study was to obtain phenotypic parameter estimates for various reproductive traits for inclusion in a multi-trait genetic improvement program. Since heritabilities cannot be estimated due to the lack of a pedigree structure, repeatabilities have been estimated in order to set an upper bound for the heritabilities of these traits. Genetic correlations cannot be evaluated with the available data set. Results from the industry survey presented in Chapter 3 indicated four main selection objectives for crocodile reproductive efficiency, namely breeder output (number of live, healthy hatchlings per female per year), annual nesting of females, earlier nesting within a breeding season, and age of sexual maturity. Selecting for breeding animals that produce more hatchlings per year is an obvious way of increasing efficiency on a ‘per breeder’ basis since the overhead costs of breeder maintenance is reduced proportionally. There is anecdotal evidence that some females miss nesting every third or fourth year. Time of nesting within a year is important since the time when the offspring hatch is important in relation to seasonal effects on subsequent juvenile growth. Age of sexual maturity is currently seen as a low priority since only a few crocodiles have been raised in captivity for this purpose. Under natural and farmed conditions, saltwater crocodiles in northern Australia nest primarily during the wet season between late-October and April (Richardson et al., 2002). A female lays only one clutch of eggs in a nesting season. Webb and Cooper-Preston (1989) reported the average clutch size (± SE) from 416 wild nests to be 49.98 ± 0.56 eggs (range 2-78 eggs). They also described the elliptical, hard-shelled eggs as having average dimensions (average of clutch means) of: length 7.84 ± 0.02cm (n = 385; range of clutch means 6.57-8.95cm), width 4.90 ± 0.01cm (n = 383; range 4.18-5.50cm) and weight 109.19 ± 0.79g (n = 357; range 65.4-147.0g). The eggs are laid during a single laying event, normally at night or early in the morning, into a previously constructed mound nest. Not all eggs laid in a clutch will produce a viable hatchling. Some eggs will be infertile, whilst others will die during incubation. Whilst in the wild high embryo mortality caused by flooding, predation or overheating are reported (Richardson et al., 2002), these are greatly reduced in captivity although not negligible. Using artificial incubation techniques (32oC; 90-100% relative humidity), time to hatch is

37

approximately 80 days, which corresponds to a hatching period between February-June (Richardson et al., 2002).

4.3 Materials and methods

4.3.1 Animals and data collection Nesting records were collected from Janamba Croc Farm (Northern Territory, Australia) for the consecutive breeding seasons between 1994 and 2002. Data were collected from 30 pairs kept mostly in unitised (1 male, 1 female) breeding pens, although some pens consisted of 1 male and 2 females. With the exception of one pair, all breeding animals were wild-harvested. This had two implications. Firstly, the age structure of the breeding population could not be ascertained and therefore, statistical adjustments for age could not be made. Secondly, only a few captive-born juveniles have been raised to become replacement breeders and as such, no familial data are available for the traits between generations. Therefore, this data set contains only repeated measures on individuals for the reproductive traits of interest. All adults are assumed unrelated. A thorough description of breeding pens is given in Chapter 2 (Section 2.1). Briefly, the breeding animals were maintained in two types of pens: the B- and UB-pens. The B-pens contain a single male and single female, whilst the UB-pens usually have a single male and 1-2 females. Breeding crocodiles are fed once a week, generally receiving either one chicken each or equivalent in red meat. The standard practice on the farm is to collect the eggs within 24 hours of oviposition to minimise embryo mortality. A thorough description of egg collection and incubation is given in Chapter 2 (Section 2.1.2). Briefly, nests are carefully opened and each egg is marked along the dorsal midline to indicate its “upright” position so that it can be incubated in the same orientation (Ferguson, 1985). Infertile eggs are detected by “candling” with a small torch. Eggs that have been collected within the first 24 hours after laying are still translucent and infertile members of the clutch are recognised by the absence of sub-embryonic fluid. If the clutch has been laid more than 24 hours before collection, the beginning of the opaque band provides a more precise indication of a fertile egg (described by Richardson, et al., 2002) and also of the correct “upright” orientation. Fertile eggs are incubated at 32oC in a relative humidity between approximately 90-95%. Table 4.1 presents a description of the traits recorded between 1994 and 2002. Descriptive statistics for each trait are given in Table 4.2. A second data set was constructed for the trait Nesting. This data set consisted of binary data indicating whether a female nested in a particular year or not.

38

Table 4.1 Crocodilian reproductive traits of economic importance. Abb. is the abbreviated trait name used in the description of the models.

Abb. Description Clutch Size (eggs) ClSize Total number of eggs collected in a clutch Number Viable (eggs)

NoViable ClSize minus the infertile eggs and those that died before collection

Number Hatch (hatchlings)

NoHatch NoViable minus the those that died during incubation or were euthanased due to abnormalities (lethargy, kinked spine, external yolk sac, external organs, etc)

Hatchability HatchR NoHatch as a proportion of ClSize Average Snout-Vent Length (mm)

AvSVL Average of a random sampling of 10 hatchling snout-vent lengths (~30%) from each clutch. The animals were inverted and the tip of the snout placed at the beginning of a clear, plastic ruler and the measurement was taken to the nearest 1mm.

Hatch Days (days)

HDays Number of days between hatching date and the 1st of January in that particular year.

Nesting Nesting Whether female nested or not in a particular year. (0 = no; 1 = yes)

Table 4.2. Summary statistics for the reproductive traits used to estimate variance components.

N Mean ± SD Range

ClSize (eggs) 190 43.36 ± 9.37 23-68

NoViable (eggs) 190 39.47 ± 10.75 3-68

NoHatch (hatchlings) 190 31.25 ± 11.44 0-61

HatchR 190 0.72 ± 0.28 0-1.00

AvSVL (mm) 147 141.7 ± 4.73 126-152

HDays (days) 186 85.05 ± 29.19 25-175

Nesting (0/1) 229 0.83A 0-1 A Frequency of nesting successes.

4.3.2 Statistical methods The objective of this study was to estimate phenotypic parameters including repeatabilities (upper bounds for heritability) for reproductive traits in saltwater crocodiles. Using Restricted Maximum Likelihood (REML), appropriate univariate models for each trait were established before combining the traits into a multi-trait analysis for (co)variance component estimation. Repeatability for whether a female nested or not each year (Nesting) was estimated using a generalised linear mixed model (GLMM).

39

4.3.2.1 Linear models Univariate REML analyses were conducted in GenStat 6, version 6.1 (2002) for all traits, except frequency of nesting. The following was the initial model evaluated for all traits

Yijklm = µ + βLELjk + βWEWjk + Yeark +Penl + Collm + Pairj + εijklm where Yijklm is an observation on either ClSize, NoViable, NoHatch, HatchR, AvSVL or HDays; µ = the overall mean; ELjk and EWjk are the average egg lengths and egg widths of a random sample of 10 eggs for the clutch from the jth pair in the kth year; βL and βW are the regression coefficients associated with EL and EW respectively; Yeark is the fixed effect of the kth year (k = 1994,…,2002); Penl is the fixed effect for the breeder environments, i.e. B-pens or UB-pens (l = 1,2); Collm (Collected) is the number of days between oviposition and collection (m = 1,...,10) (all traits except ClSize and HDays); Pairj is the random effect of pair (assumed N(0,σ2

Pair)); and εijklm is the random residual effect (assumed N(0,σ2ε)). A 5%

significance level was chosen to evaluate explanatory variables by backward elimination. The subsequent univariate models were then combined for multi-trait analysis using ASReml (version 1.10, 2003). Again, fixed effects were omitted from the multivariate model if P>0.05 using backward elimination. Repeatability is defined as

Rep = 22Pair

2Pair

εσ+σσ

where 2Pairσ and 2

εσ are the estimated pair and residual variance components, respectively. The phenotypic correlations between two traits were estimated using standard equations (Searle, 1961).

Phenotypic standard deviations (σP) were estimated using 22Pair εσ+σ as defined by Gregory et al.

(1995). Note that the heritability of a trait will always be less than or equal to the repeatability (Falconer and Mackay, 1996).

4.3.2.2 Binomial model The trait Nesting was modelled using a GLMM in ASReml. The trait was binomial since the animal either nested (1) in a particular year or did not (0), and was modelled as

j(m)kmlk

jklm

jklm Pair SOIPenYearp1

plog +β+++µ=⎟

⎟⎠

⎞⎜⎜⎝

⎛

−

where pjklm is the probability of the female nesting, and µ, Yeark and Penl are fixed effects as described above. Previous studies by Webb et al. (1990) and McClure and Mayer (2001) have reported that the southern oscillation index (SOI) of certain months prior to the breeding season in a particular year affect the degree and timing of reproductive effort. For this study (m)

kSOI is the southern oscillation index for month m in the kth year with m being either June, July, August, September, average of June and July or average of July and August. These SOI measures were fitted in separate models to ascertain significance. So mβ is the regression coefficient of the corresponding SOI measure. SOI data were obtained for each year from the Australian Bureau of Meteorology website (http://www.bom.gov.au/climate/current/soihtm1.shtml; accessed September 9, 2003). Separate models were used to evaluate the significance of each month’s SOI within year. As in the linear model analysis, Pairj was specified as a random effect (assumed N(0,σ2

Pair)). This trait could not be included in the multivariate model due to confounding with other traits. That is, when an animal was coded as not having nested in a particular year (0), the other traits of interest were considered as missing values.

40

Repeatability for nesting probability was calculated on the unobserved scale using

Rep = ( ) 3/π22Pair

2Pair

φ+σσ

where 2Pairσ is the pair variance component and φ is the dispersion parameter included to manage any

non-binomial variation.

4.4 Results

4.4.1 Univariate models The terms found to be significant for each reproductive trait from univariate modelling are shown in Table 4.3 (those that have the superscript M remained significant in the multivariate model also). The random genetic effect, Pair, was highly significant in all models.

Table 4.3. Significant terms for the reproductive traits from univariate REML modelling. Levels of significance are shown as 0.001 < P ≤ 0.05 (**) and P < 0.001 (***). A dash (-) indicates that the term was not significant (P > 0.05), whilst NA specifies terms that were not applicable to that trait and were therefore not included in the initial model. Terms significant with the multi-trait REML model indicated with M. Nesting was not included in the multi-trait analysis due to confounding with the other traits.

Fixed terms Random Response EL EW Pen Year Coll Pair ClSize *** *** ** - NA ***M NoViable ** *** *** - - ***M NoHatch ** ** *** - *** ***M HatchR - - - - *** *** M AvSVL ***M - - ***M - ***M HDays NA NA - - NA ***M Nesting NA NA - *** NA ***

As mentioned before, with the exception of one pair, adult age is unknown. This resulted in the inability to adjust for age and/or parity during the modelling process. Crocodiles exhibit non-determinant growth, that is continue to grow throughout their life, although slowing as the animal gets older (Thorbjarnarson, 1996). Therefore, female size is related to female age. Crocodilian reproductive biology studies have shown clear allometric trends between female size, clutch size and egg dimensions (Webb et al., 1983; Thorbjarnarson, 1996). Utilising this information, egg dimensions (egg length and egg width) were used as proxies for adjusting for female age effects in the reproductive traits under investigation. With the exception of HatchR and HDays, egg length (EL) was a significant predictor for all models. For the traits ClSize, NoViable and NoHatch, egg width (EW) was also significant in conjunction with EL. Regression coefficients for EL and EW predicted from the univariate analyses of these traits are presented in Table 4.4.

Table 4.4. Regression coefficients for egg length (EL) and egg width (EW) produced from univariate REML analyses. Egg width was not significant in the AvSVL univariate model, indicated by NA.

Response Egg Length (± SE) Egg Width (± SE) ClSize -1.41 (0.29) 2.01 (0.41)

NoViable -1.09 (0.36) 2.37 (0.52) NoHatch -1.03 (0.38) 1.81 (0.57) AvSVL 1.00 (0.11) NA

41

Breeder pen type also had a significant effect on the traits of ClSize, NoViable and NoHatch. Compared to the B-pens, the UB-pens produced clutches with an average of 4.97 more eggs per clutch. Further, an additional 7.93 eggs per clutch were viable and an average of 8.36 additional hatchlings per clutch were produced. These results appear to be a cumulative across the reproductive characters. Possible explanations could include reduced exposure to male aggression due to a larger pen size, and the ability for crocodiles in the UB-pens to thermoregulate and maintain optimal in utero conditions for the embryos both prior and during gestation. The day of collection (Coll) after oviposition greatly affected the embryo survival rate and the overall number of live, healthy hatchlings (NoHatch). The majority of eggs were collected on day 0 (n = 142 nests) with 20, 10, 3, 1, 1 nests being collected on days 1, 2, 3, 5 and 10, respectively. The day of collection was not recorded for 13 clutches. It should also be noted that these days of collection are estimated from the bands that form around the eggs as incubation proceeds (Richardson et al., 2002). However, since they were collected relatively early into incubation, they were considered fairly accurate.

4.4.2 Multi-trait model and parameter estimates A multivariate analysis was undertaken with ASReml using the explanatory variables identified as being significant for each trait in their respective univariate models. Note that using this multivariate formulation, ASReml allows a separate set of explanatory variables to be specified for each trait. Those explanatory variables that remained significant at the 5% level are indicated by the superscript M in Table 4.3. Repeatability and phenotypic correlation estimates are shown in Table 4.5. Repeatability estimates are in bold on the diagonal, whilst phenotypic correlations are below the diagonal. The repeatability estimates were high and ranged from 0.24 (HatchR) to 0.68 (ClSize and HDays). Phenotypic correlations ranged from negligible (0.03: NoViable and HDays) to high (0.86: ClSize and NoViable) between the various traits.

Table 4.5. Estimates of repeatability (bold on diagonal) and phenotypic correlations (below diagonal) with their approximate standard errors in parenthesis below each estimate for saltwater crocodile reproductive traits. A indicates repeatability on the underlying scale. Phenotypic standard deviations (σP) are also presented for the measurement-variable traits.

ClSize NoViable NoHatch HatchR AvSVL HDays Nesting

ClSize 0.68

(0.02)

NoViable 0.86

(0.01) 0.55

(0.03)

NoHatch 0.54

(0.03) 0.73

(0.02) 0.34

(0.02)

HatchR -0.08 (0.04)

0.27 (0.04)

0.78 (0.02)

0.24 (0.02)

AvSVL -0.15 (0.04)

-0.06 (0.05)

0.15 (0.05)

0.30 (0.05)

0.57 (0.03)

HDays -0.13 (0.03)

-0.04 (0.03)

0.03 (0.04)

0.13 (0.04)

0.10 (0.04)

0.68 (0.02)

Nesting - - - - - 0.58A

σP 9.36 10.87 11.56 0.22 4.29 28.39 -

42

4.4.3 Binomial model and repeatability The binary trait Nesting, had significant year and pair effects. The southern oscillation index (SOI) values for the months of interest were all non-significant. Since it was not appropriate to include this trait in the multivariate model, the univariate GLMM variance component estimates were used to determine repeatability. The pair residual component was 2.49 with a dispersion parameter of 0.56, indicating less variation than expected under a binomial model using a logit-link function. Repeatability on the underlying scale for a female nesting was calculated to be 0.58.

4.5 Discussion Repeatability estimates set upper limits to heritability (Falconer and Mackay, 1996). The estimates of repeatability for reproductive characters in the saltwater crocodile were high compared to estimates reported in other livestock species (pigs: Serenius et al., 2003, Kerr and Cameron, 1995; rabbits: Ayyat et al., 1995), although the phenotypic correlations were similar to those reported in the other industries. In addition, the phenotypic correlations and non-genetic factors of crocodile reproduction traits presented in this study concur with the results of published biological studies.

4.5.1 Accounting for female age using proxies The ages of only one female and male used in this study were known. Both were captive-raised, with the female being bred at Janamba Croc Farm, whilst the male was from a wild-harvested clutch of eggs. The pair first nested when the female was six and the male ten years of age. The first clutch this pair produced (in 1998) consisted of 35 small eggs, in which 25 were infertile and the other 10 died during incubation. For the succeeding three years, clutch size varied increasing to 50, 52 and 46 in the years 1999, 2000 and 2001, respectively, and all are larger than the average clutch size reported for this entire data-set (43.36; Table 4.2). In the same years 1999-2000, the number of hatchlings produced was 14, 29 and 27, respectively, which is slightly below the farm average (31.25; Table 4.2) although expected to gradually increase with maturity. Unfortunately, the inability to account for female age could have compromised the estimates presented herein. The relationships between female size (related to age) and different aspects of reproductive output have been well documented (Thorbjarnarson, 1996). A clutch characteristic demonstrated by Deeming and Ferguson (1990) in the American alligator (A. mississippiensis) was more uniform mean egg widths between clutches than mean egg lengths, possibly indicating an oviducal limitation on egg width that does not affect length. In another study, Thorbjarnarson (1994) reported in Caiman crocodilus an inverse relationship between egg length and clutch size, suggesting limitations on linear egg placement within the oviduct of smaller animals (Ford and Seigel, 1989). That is, as the female continues to grow older, her pelvic canal diameter increases, determining the resulting egg and clutch sizes. Similar trends have also been shown in the poultry (Bell, 2002) and aquaculture (Guo-Sheng et al., 2002) industries. Using this information, it was hoped that egg length and width measurements would account for some clutch variation caused by female age. As can be seen from the regression coefficients in Table 4.4, as egg length increases by 1mm, the reproductive characters (ClSize, NoViable and NoHatch) decrease, similar to the trend shown in C. crocodilus (Thorbjarnarson, 1994). Contrary to this, as egg width increases by 1 mm, the reproductive characters increase. This was expected in accordance with the results presented by Deeming and Ferguson (1990). That is, if egg width is restricted by oviducal limitations, which in turn are restricted by female size (and age), then her ability to produce more eggs of a larger size will increase as the female grows (and ages).

43

4.5.2 Relationships between the “egg” traits The estimates of phenotypic correlations between clutch size, the number of viable eggs and the number of hatchlings produced were positive and high (Table 4.5). In contrast, the phenotypic correlations between hatchability and ClSize, NoViable and NoHatch were -0.08, 0.27 and 0.78, respectively. The inability to determine genetic correlations from the information provided by this data set questions the validity of using NoHatch and/or HatchR as the trait quantifying the number of offspring produced from each clutch. That is, the high, positive phenotypic correlation estimates obtained between NoHatch and ClSize could just be a numerical consequence of an increased clutch size resulting in more embryos surviving. However, using the trait HatchR instead of NoHatch creates modelling complexities, particularly with interpretation of results. That is, since the traits ClSize and NoViable have a large range of values (as shown in Table 4.2), from the perspective of a farm enterprise, a 50% hatch rate from a clutch of 50 eggs is better than 100% from 10 eggs. Many of these issues can not be further dealt with until a pedigree structure is available to estimate the relevant genetic parameters.

4.5.3 Relationship between hatchling size and egg traits In general, larger crocodilian eggs produce larger offspring (Ferguson, 1985). The regression coefficient of egg length from the univariate analysis on average snout-vent length was 1.00 mm/mm EL (SE 0.11mm; Table 4.4), which concurs with Ferguson (1985). The phenotypic correlations between AvSVL and the egg traits (ClSize, NoViable, NoHatch and HatchR) were not consistent. The correlation with ClSize was -0.15, indicating that as clutch size increases, the hatchling size decreases. This appears to be valid since as egg length increases, clutch size decreases whilst hatchling size increases. AvSVL has almost a zero correlation with NoViable (-0.06), whilst there is a small, positive relationship with number that hatch (0.15). There was a moderate correlation (0.30) between average snout-vent length and hatchability.

4.5.4 Relationship between time of nesting and other traits For time of nesting (as indicated by when the juveniles hatched, or HDays), phenotypic correlations with the other traits were all non-signficant. A small, negative correlation estimate was found with clutch size (-0.13), whilst the same magnitude estimate of opposite sign was found using hatchability (0.13) which may be indicative of the large effect clutch size had in deriving the trait hatchability. A small, positive correlation was found with AvSVL (0.10), whilst the correlation of HDays with NoViable and NoHatch were almost zero (-0.04 and 0.03, respectively). The repeatability estimated for time of nesting was 0.68, indicating that females have a tendency to lay at approximately the same time each year. This concurs with the result reported by McClure and Mayer (2001) whose study was similar in design and duration (nine years) to this study. Timing of nesting has been reported to be affected by social dominance hierarchies in American alligators, whereby the larger, more dominant females mate and nest first (Ferguson and Joanen, 1983; Ferguson, 1985). However, this study, and that of McClure and Mayer (2001), does not support this explanation since the crocodiles were in unitised pens where such social interactions do not exist. Contrary to the study of wild populations conducted by Webb et al. (1983), this study showed that there is a slight, negative relationship between clutch size and time of nesting under farm conditions. This again confirms the results of McClure and Mayer (2001). However, McClure and Mayer (2001) also reported that higher hatch rates were achieved with earlier nesting females. In contrast, a non-significant relationship between NoHatch and time of nesting was found in this study (phenotypic correlation 0.03 (SE 0.04); Table 4.5). Although non-significant, the clutches that are laid later produce slightly larger hatchlings as indicated by the average snout-vent length phenotypic correlation of 0.10 (SE 0.04; Table 4.5). However, in terms of improving production efficiency, the size of the offspring at hatchling has no real relevance to post-hatching growth rates (see Chapter 5).

44

4.5.5 Nesting success McClure and Mayer (2001) and Webb et al. (1990) reported significant relationships between whether a female nested in a particular year and the southern oscillation index (SOI) for a particular month. SOI is a measure of cyclic pattern of air pressure movement across the Pacific Ocean and is used to predict the onset of the wet season in Northern Australia (Webb et al., 1990). It is recognised that the degree and timing of nesting within the wet-season is closely associated with the climatic conditions prevailing late in the dry-season. Low water levels between August and October, combined with high temperature conditions, are generally associated with a reduced nesting effort. In contrast, high water levels and cool conditions result in maximal nesting activity. More specifically, late dry-season rains and intense periods of rain throughout the wet season are the environmental stimuli required for courtship and mating (Webb and Manolis, 1989). Due to this interrelationship between water level, rain and temperature, reproductive output in the wild can be predicted as much as 7-8 months prior to the nesting season (Webb et al., 1987). Webb et al. (1990) reported a significant relationship between the degree of nesting and the average SOI for June and July, whilst McClure and Mayer (2001) reported that August was a more appropriate predictor for their data. The SOIs used in this study (June to September, and the averages for June-July and July-September) were all non-significant. In contrast, the fixed effect, Year, was highly significant in the model. An explanation could be that Year is a term that encompasses not only air pressures and water heights (SOI), but also air temperature, relative humidity and other climatic factors. The random factor, Pair, accounted for a considerable amount of variation that McClure and Mayer (2001) did not attempt to account for in their modelling procedure (ANOVA).

4.6 Implications Only repeatability and phenotypic correlations could be estimated from the available data. Repeatabilities set upper limits for heritabilities. Notably all of the repeatabilities estimated in this study were very high. The conclusions that can be drawn from these estimates in relation to a genetic improvement program must be conservative, since heritability estimates for reproductive traits are usually substantially lower than these estimates of repeatability. However, these results provide a starting point for designing selection for future breeding animals.

45

5. Quantitative analysis of age at slaughter

5.1 Abstract Crocodile morphometric (head, snout-vent and total length) measurements were recorded at three stages during the production chain: hatching, inventory (average age (± SE) is 265.1 ± 0.4 days) and slaughter (average age is 1037.8 ± 0.4 days). Crocodile skins are used for the manufacture of exclusive leather products, with the most common-sized skin sold having 35cm to 45 cm in belly width. One of the breeding objectives for inclusion into a multi-trait genetic improvement program for saltwater crocodiles is the time taken for a juvenile to reach this size, or age at slaughter. A multivariate REML analysis provided (co)variance components for estimating the first published genetic parameter estimates for these traits. Heritability (± SE) estimates for the traits hatchling snout-vent length, inventory head length and age at slaughter were 0.60 (0.15), 0.59 (0.12) and 0.40 (0.10), respectively. There were strong negative genetic (-0.81 ± 0.08) and phenotypic (-0.82 ± 0.02) correlations between age at slaughter and inventory head length.

5.2 Introduction The majority of juvenile saltwater crocodiles in Australia reach harvest size in captivity within 3.5 years, although slaughter age has been reported to range between two and five years (Treadwell et al., 1991). Much of this variation has been attributed to large clutch effects (Manolis et al., 1989; MacNamara et al., 2003). However, no research has been conducted to quantify these clutch effects, or partition the variation in the time to reach slaughter size into genetic and other components. Decreasing the time to slaughter has obvious benefits in reducing costs. Design of a selection program with this long-term objective requires the accurate estimation of relevant genetic and non-genetic (co)variance components.

5.3 Methods and materials

5.3.1 Experimental animals Data were collected from juvenile crocodiles bred and raised at Janamba Croc Farm (Northern Territory, Australia) between 1994 and 2002. Progeny records from 30 pairs kept mostly in unitised (1 male, 1 female) breeding pens, although some pens consisted of 1 male and 2 females, were available. Data were collected at three stages: hatching, inventory and slaughter. These different stages are described below. A description of the morphological traits measured are given in Table 5.1, whilst descriptive statistics for all traits are given in Table 5.2. Hatching. Eggs were collected and incubated at 32oC as described in Chapter 2 (Section 2.1.2). After hatching, the crocodiles were marked for clutch identification by cutting scutes in a unique sequence. Scutes are vertical, triangular osteoderms on the dorsal midline of the posterior tail that bifurcate into two rows of more laterally flattened scales in about the middle third of the tail and continue cranially (Richardson et al., 2002). In addition to the clutch-identification scute cuts, the clutches in the 2001 and 2002 cohorts were also given individual scute cuts to allow individual animals to be tracked through the production system so that subsequent performance records could be attributed to these animals. Hatchling head length (HHL), snout-vent length (HSVL) and total length (HTL) were measured on every individual hatchling in the breeding seasons between 2000 and 2002.

46

Table 5.1. Crocodile morphological traits measured as possible selection criterion for age at slaughter. Head, snout-vent and total lengths were first described by Webb and Messel (1978), whilst the definition for belly width was described by Van Jaarsveldt (1987). Abb. is the abbreviated trait name used in the description of the models.

Abb. Description Head length HL Tip of snout to median posterior platform of supraoccipital bone.

For animals measured at hatching (HHL) and at inventory (InvHL), these were measured with digital calipers to the nearest 0.01 mm. At slaughter (CullHL), the crocodiles were measured to the nearest 5 mm with a flexible tape measure.

Snout-vent length

SVL Tip of snout to anterior proximity of cloaca. Only animals at hatching (HSVL) and at inventory (InvSVL) were measured. The animals were inverted and the tip of the snout placed at the beginning of a clear, plastic ruler and the measurement was taken to the nearest 1mm.

Total length TL Tip of snout to tip of tail. At hatching (HTL), the crocodiles were inverted and the tip of the snout placed at the beginning of a clear, plastic ruler and the measurement was taken to the nearest 1mm. At inventory (InvTL), the snout was butted up against an inlaid ruler and a measurement to the nearest 1 mm taken to the tip of the tail. Slaughter animals (CullTL) were measured to the nearest 10 mm with a flexible tape measure.

Bellywidth Bwidth Measured between the third osteoderm down the flank from the forelegs, on either side without stretching, using a metal ruler to the nearest 10 mm. Only measured on slaughter animals.

Table 5.2. Summary statistics for the morphometric traits used to estimate variance components.

n Mean ± SD Range Hatchling traits

HHL (mm) 1607 44.06 ± 1.60 37.49-47.97 HSVL (mm) 1599 143.4 ± 5.20 118-158 HTL (mm) 1604 305.4 ± 10.81 245-334

Inventory traitsInvHL (mm) 4859 90.83 ± 14.64 49.48-142.16 InvSVL (mm) 970 303.5 ± 58.55 152-502 InvTL (mm) 970 642.5 ± 124.58 319-1045 InvAge (days) 4859 265.1 ± 30.67 163-341

Slaughter traits CullHL (mm) 2133 219.9 ± 61.89 140-260 CullTL (mm) 2107 1682 ± 110.17 1240-2010 CullBwidth (mm) 1349 37.63 ± 2.94 28-47 CullAge (days) 2151 1037.8 ± 197.57 634-1859

Inventory. Hatchlings were initially placed in either fibreglass tubs or small pens in an environmentally-controlled nursery. As the animals get larger, they may be transferred into larger nursery pens, intermediate pens, remain in their initial environment or be transferred between the initial environments. In either December of that year or early January of the next year, head length (InvHL) measurements are made on all juveniles. These measurements were made available from 1994 to 2002. In addition, 6-8% of animals are also measured for SVL (InvSVL) and TL (InvTL) in each year, although all juveniles from the 2001 cohort were measured for all three traits. Overall, 20% of the juveniles had observations on all three traits.

47

Slaughter. After inventory, the juveniles were moved to the intermediate pens and later transferred to the channel (or grow-out) pens. Crocodiles are generally slaughtered between 1.6-1.9 metres in total length to obtain a belly width between 35 and 45 cm. However, the decision to select an animal for slaughter is based upon visual assessment. Therefore, there is substantial variation in the resultant belly-widths as shown in Figure 5.1 outside the desired 35 to 45cm belly width range. In addition, management decisions to slaughter poor-growing animals further contribute to the range of belly-widths obtained (Figure 5.1). At slaughter, head length (CullHL) and total length (CullTL) measurements were taken. Later, belly width (CullBwidth) was measured during skin grading. Data for these traits were available from 1994 to 2001.

Figure 5.1. Histogram of belly widths available for analysis.

Gender was also determined for each animal at this stage, as described in Table 5.3. From the data available there were 1946 male animals and 159 females. This ratio was not unexpected since the animals were incubated at 32oC which predominantly produces males (Richardson et al., 2002) in saltwater crocodiles.

48

Table 5.3. Description of terms used to model the traits of interest. Abb. is the abbreviated name used in the model descriptions.

Abb. Description Fixed effects

Year Year Year in which the animal hatched. For the hatchling, inventory and slaughter traits, the number of years were 3 (2000-2002), 9 (1994-2002) and 8 (1994-2001), respectively.

Initial rearing environment (Tubs/Nursery)

EnvI This refers to whether the hatchlings were initially placed in the tubs (coded as 1) or environmentally-controlled nursery (coded as 2). This terms attempts to encapsulate any environmental differences between observed measurements.

Gender (Male/Female)

Gender Gender was determined at slaughter by manual palpation and feeling for the presence of a penis, as described by Webb and Messel (1978).

Egg Length (mm)

EL Average of a random sampling of 10 egg length measurements (~30%) from each clutch as described in Chapter 4. Measurements were taken with digital callipers to 0.01mm.

Egg Width (mm)

EW Average of a random sampling of 10 egg width measurements (~30%) from each clutch as described in Chapter 4. Measurements were taken with digital callipers to 0.01mm.

Hatch Days (days)

HDays Number of days between hatching date and the 1st of January in that particular year.

Inventory Age (days)

InvAge This explanatory variable was only required for the inventory traits to describe the age of the crocodiles.

Inventory Age2 (days)

InvAge2 Square of InvAge.

Random effects

Pair Pair Random effect of the parents, assumed N(0, �2Pair)

Clutch Clutch Common environment effects modelled as an interaction between pair and year, assumed N(0, �2

Clutch)

5.3.2 Rearing environments A thorough description of the juvenile rearing environments is given in Chapter 2 (Section 2.1.4), although a brief description of each environment is presented below. Fibreglass tubs. Each of the fibreglass tubs covers an area of 4.5m2, in which approximately 30% was covered by water maintained at a depth of 10cm. Stocking densities were maintained at 0.09m2/animal. The main disadvantage with these tubs is that they were exposed to ambient air temperatures. During the day, half the lid was opened to permit basking but it was always closed at night to retain heat. Bore water constantly flowed through the tubs at temperatures between 29oC to 31oC, depending on the time of year. A heat lamp was also provided during the night. The main disadvantage with these tubs is that they were exposed to ambient air temperatures. Environmentally-controlled nursery. This nursery is divided into two sections: small and large pen areas. The small pen area contains 80 pens, whilst the large pen area contains 40 pens, divided by an internal wall. The small pens were 2m2 (2m × 1m) each, whilst each large pen was 6m2 (2m × 3m). All animals are initially placed in the small pens (0.09m2/animal) until they become too large in which time the large pens (0.32m2/animal) are recruited. Light was provided artificially by fluorescent tubes between approximately 0700 and 1630 An advantage of the nursery was the ability to control temperature and, to a lesser extent, humidity. Doors and shutters could be opened or closed to promote ventilation or retain

49

warmth. Shallow water (approximately 40% of the floor area) could be maintained a constant, preset levels between 30.5 and 31.5oC. Intermediate pens. These concrete pens cover an area of 12.6m2 and animals were stocked to a density of 0.32m2/animal in each pen. Water constantly flowed into the pens at the temperatures mentioned above for the tubs. The pens were exposed to ambient air temperatures and natural diurnal photoperiods although the whole area was covered by shade-cloth to reduce exposure to direct sunlight. Channel (or grow-out) pens. At about 0.8-0.9m in length, the animals were relocated into channels until they reached slaughter size. The channels derived their name from the two “channels” of water contained within each pen. They are 8m × 25m (200m2) with 200 animals placed into each (stocking density of 1m2/crocodile). One of the channels within each pen, termed the “flow” side, has water constantly flowing through it at the temperatures described above.

5.3.3 Statistical methods Full-sibling data were available for all measurement stages (i.e. hatching, inventory and slaughter). This provided an opportunity to estimate heritability, and genetic and phenotypic correlations using restricted maximum likelihood (REML). Since this was the first time that genetic and non-genetic components of crocodilian morphometric traits have been analysed, it was necessary to model the traits using univariate methods before combining them in a multivariate model to obtain (co)variance components. Univariate modelling were conducted in GenStat 6 (version 6.1, 2002). Descriptions of the explanatory variates used are given in Table 5.3. A 5% significance level was chosen to evaluate explanatory variables by backward elimination. Final multivariate modelling was conducted in ASReml (version 1.10, 2003).

5.3.4 Univariate modelling Hatchling traits. The following was the initial model evaluated for all hatchling traits

Yijk = µ + βLELjk + βWEWjk + Yeark + Pairj + Clutchjk + εijk where Yijk is an observation on either hatchling head length (HHL), snout-vent length (HSVL) or total length (HTL); µ = the overall mean; ELjk and EWjk are the average of a random sample of 10 egg lengths and 10 eggs widths, respectively, for the clutch from the jth pair in the kth year; βL and βW are the regression coefficients associated with EL and EW respectively; Yeark is the fixed effect of the kth year (k = 2000, 2001 or 2002); Pairj is the random effect of pair (assumed N(0,σ2

Pair)); Clutchjk is the common environment (random) effect of a clutch produced by the jth pair in the kth year (assumed N(0,σ2

Clutch)); and εijk is the random residual effect (assumed N(0,σ2

ε)). Inventory traits. The following was the initial model evaluated for all inventory traits

Yijkl = µ + βAInvAgeijk + βA2InvAge2ijk + βHDHDaysjk + Yeark + EnvIl + Pairj + Clutchjk + εijkl

where Yijkl is an observation on either inventory head length (InvHL), snout-vent length (InvSVL) or total length (InvTL); µ = the overall mean; InvAgeijk and InvAge2

ijk are the age and age-square of the ith individual from the jth pair in the kth year at inventory, HDaysjk is the number of days between hatching date and the 1st of January in that particular year for an individual from the jth pair in the kth year; βA, βA2

and βHD are the regression coefficients associated with InvAge, InvAge2 and HDays respectively; Yeark is the fixed effect of the kth year (k = 1994,…,2001); EnvIl is the effect of the initial rearing environment (l = 1 or 2); Pairj is the random effect of pair (assumed N(0,σ2

Pair)); Clutchjk is the common environment (random) effect of a clutch produced by the jth pair in the kth year (assumed N(0,σ2

Clutch)); and εijkl is the random residual effect (assumed N(0,σ2

ε)).

50

Age at slaughter. The following was the initial model evaluated for the actual breeding objective, age at slaughter (CullAge)

CullAgeijklm = µ + βHDHDaysjk + Yeark + EnvIl + Genderm + Pairj + Clutchjk + εijklm where µ = the overall mean; HDaysjk is the number of days between hatching date and the 1st of January in that particular year for an individual from the jth pair in the kth year; βHD is the regression coefficient associated with HDays; Yeark is the fixed effect of the kth year (k = 1994,…,2002); EnvIl is effect of the initial rearing environment (l = 1 or 2); Genderm is effect of the individual’s gender (m = 1 or 2); Pairj is the random effect of pair (assumed N(0,σ2

Pair)); Clutchjk is the common environment (random) effect of a clutch produced by the jth pair in the kth year (assumed N(0,σ2

Clutch)); and εijklm is the random residual effect (assumed N(0,σ2

ε)).

5.3.5 Multivariate modelling Problems were encountered when all traits were combined for a single multivariate analysis due to collinearity and data limitations. Instead, a multivariate analysis for the traits within each “stage” (hatchling and inventory) was conducted. One trait from each stage was then selected based on the number of observations available and biological validity, and incorporated into an overall multivariate analysis. Multivariate modelling was conducted in ASReml (version 1.10, 2003). Again, fixed effects were omitted from the multivariate model if P>0.05 using backward elimination.

5.3.6 Genetic parameter estimates Heritabilities were estimated using a full-sib correlation as follows

22Clutch

2Pair

2Pair2 2h

εσ+σ+σ

σ×=

using the REML estimates of the variance components, 2Pairσ , 2

Clutchσ and 2εσ . Phenotypic standard

deviations (σP) were estimated by 2ε

2Clutch

2Pair σ+σ+σ as described by Gregory et al. (1995), whilst

genetic (rg) and phenotypic (rp) correlations between the traits were estimated using the standard methods (Searle, 1961).

5.4 Results and discussion The terms found to be significant for each trait from the univariate models are shown in Table 5.4. The random genetic (Pair) and common environment (Clutch) effects were highly significant in all models. Traits (bold) that were used in the final multivariate analysis are indicated in Table 5.4, and those terms that have the superscript M remained significant in the multivariate model.

5.4.1 Hatchling traits Egg length (EL) had a significant effect on all hatchling traits (HHL, HSVL and HTL), whilst year was only significant for hatchling snout-vent length. Egg width (EW) was non-significant for all hatchling traits.

51

Table 5.4. Significant terms for the morphometric traits from univariate REML modelling. Levels of significance are shown as 0.001 < P ≤ 0.05 (**) and P < 0.001 (***). A dash (-) indicates that the term was not significant (P > 0.05) at the 5% level, whilst NA specifies terms that were not applicable or not available to that trait and were therefore not included in the initial model. Traits in bold were those used in the final multivariate REML model and the terms that remained significant are indicated with M.

Fixed terms

Random Year

EL EW EnvI HDays InvAge InvAge2 Gender Pair Clutch

HHL - *** - NA NA NA NA NA *** *** HSVL ***M *** - NA NA NA NA NA ***M ***M

HTL - *** - NA NA NA NA NA *** ***

InvHL ***M NA NA ***M - ***M *** NA ***M ***M

InvSVL *** NA NA *** - *** - NA *** ***

InvTL *** NA NA *** - *** - NA *** ***

CullAge ***M NA NA - - NA NA ***M ***M ***M

A multivariate analysis of the three hatchling traits revealed high genetic and phenotypic correlation estimates, as shown in Table 5.5. These high correlation values were not unexpected since these traits are not independent from each other, and as a result of these strong autocorrelations, it was decided that only one hatchling trait should be included in the final multivariate analysis. The industry standard measurement for size of hatchlings is snout-vent length, as the head is slightly convex upon hatching from the egg (Manolis, pers. comm.) and this interferes with accurate head length measurement. Webb et al. (1983) demonstrated this during an investigation of C. porosus nesting, in which snout-vent length upon hatching displayed higher correlation coefficients with various egg measurements (length, width and weight) and hatchling body weight compared to those of head length. Therefore, snout-vent length was the only hatchling morphometric measurement to be included in the multivariate analysis.

Table 5.5. Hatchling trait genetic (above diagonal; standard errors below) and phenotypic (below diagonal) correlation estimates.

HHL HSVL HTL HHL

0.97

(0.04) 0.82

(0.09) HSVL 0.79

(0.02) 0.88

(0.06) HTL 0.75

(0.03) 0.88

(0.01)

5.4.2 Inventory traits For all inventory traits (InvHL, InvSVL, InvTL), the terms Year, EnvI and InvAge were all highly significant. For InvHL only, an age-squared term (InvAge2) was also required. Due to collinearity, a multivariate model using all three inventory traits would not converge. Therefore, it was decided that only one trait could be included in the overall multivariate model. Since a study conducted by Webb and Messel (1978) on the morphological relationships of wild C. porosus, head length has become the standard industry measurement for evaluating juvenile lengths, as it is the most

52

accurate, economical and stress-free measurement to perform on the animals. For this reason, head length was the measurement preference at inventory, and consequently there are a greater number of measurements for inclusion into a multivariate analysis.

5.4.3 Age at slaughter The average age of juveniles reaching slaughter size at Janamba Croc Farm is 1037.8 days (SE 4.26) (or 2.84 years (SE 0.01); Table 5.2). Gender and year effects were highly significant in determining the CullAge of the crocodiles. Only 8% of the juveniles reaching slaughter size were female. The bias in sex ratio was caused by incubating at 32oC, which produces mostly males (86% males; Lang and Andrews, 1994). The difference in survival between males and females has not been investigated. After adjusting for the effects of year, pair and clutch, the difference between male and female slaughter age was only 47.24 days (SE 13.97).

5.4.4 Multivariate modelling and genetic parameter estimates After specifying the univariate models described above, the three traits (HSVL, InvHL and CullAge) were combined into a multivariate model using ASReml. Note that using this multivariate formulation, ASReml allows a separate set of explanatory variables to be specified for each trait. Those explanatory variables that remained significant at the 5% level are indicated by the superscript M in Table 5.4. Genetic parameter estimates are given in Table 5.6. Heritabilities and phenotypic standard deviations. The heritability estimates for the three traits are all high (Table 5.6). The heritability estimate for the breeding objective, slaughter age, is 0.40 (SE 0.10). The heritability estimates for the possible selection criterion, HSVL (0.60; SE 0.15) and InvHL (0.59; SE 0.12), are also highly heritable. In addition, there was also considerable variation present within each set of data (i.e. HSVL, InvHL and CullAge) as indicated by the phenotypic standard deviations. Since this data consisted of measurements taken from full-sibs, the estimates of heritability are most likely upwardly biased because of confounding between the genetic variance and common environment (or dominance) effects (Falconer and Mackay, 1996). When more complex pedigree data becomes available, re-estimation of these parameters should be undertaken. These estimates represent the first reported for such production traits in crocodilians.

Table 5.6. Estimates of heritability (h2), phenotypic standard deviations (σP), genetic (rg; above diagonal) and phenotypic correlations (rp; below diagonal) for hatchling and inventory morphometric traits and age at slaughter. Standard errors are in parentheses.

HSVL InvHL CullAge

h2 0.60 (0.15)

0.59 (0.12)

0.40 (0.10)

σP 5.12 12.97 180.42

HSVL -0.04 (0.22)

0.22 (0.22)

InvHL 0.12 (0.07)

-0.81 (0.08)

CullAge 0.01 (0.07)

-0.82 (0.02)

53

Genetic and phenotypic correlations. The rg and rp estimates, along with their approximate standard errors, are given in Table 5.6. There is a very strong relationship between InvHL and CullAge as identified by both the rg (-0.81; SE 0.08) and rp (-0.82; SE 0.02) values. The rg between HSVL and InvHL shows a negligible relationship despite the rp, indicating a weak relationship. There appeared to be a strong, positive rg between HSVL and CullAge, but the standard error of the same magnitude indicates substantial data limitations.

5.5 Implications These results represent the first genetic parameter estimates of crocodile morphometric traits. The breeding objective is to decrease the age of juvenile crocodiles at harvest size. The high heritability (0.40) estimated for this trait indicates great potential to improve this trait, especially in view of the large phenotypic standard deviation of 180.42. A possible selection criterion for age at slaughter is head length at inventory, indicated by strong, negative genetic and phenotypic correlations. The hatchling trait (snout-vent length) showed no significant correlation with either the inventory trait or age at slaughter, although some data limitations were indicated by large standard error approximations.

54

6. Quantitative analysis of juvenile survival

6.1 Abstract Mortality records of 1302 juvenile crocodiles were available for analysis. Crocodiles that were slaughtered during the study were treated as censored (n = 2151). Additionally, those animals that had neither died nor been slaughtered, i.e. were still alive in the production system, had dummy records created for them (n = 1582). These dummy records were then censored at the last date of data collection. There were a total of 3733 censored records. The data were all full-sib records from 30 parental pairs from Janamba Croc Farm (Northern Territory, Australia), collected over nine consecutive years. Data were analysed using an extension of Cox’s proportional hazards model to include frailty (random) terms to account for genetic effects. Heritability of log survival time for juvenile crocodile survival was 0.15 (SE 0.04). This estimate is the first genetic estimate of crocodile survival and is a fundamental element in the development of a genetic improvement program.

6.2 Introduction In Chapter 4, it was reported that the average number of live, healthy hatchings per female per year is 31.25 (SE 0.83). Of critical importance is the survival of these hatchlings through to production size of 1.6 to 1.9m. The first year (between hatching and one year) is the most critical and Webb (1989) recommended that producers should aim for survival rates of at least 95%. After the first year, the risk of mortality decreases and a 95% survival rate is the aim from one year old to slaughter at about 3.5 years on average (Webb, 1989). In reality, few farms achieve such high survival rates, and there is large variation in this trait. This variation is affected by management regimes, but genetic effects may also impact on survival rates. The objective of this study was to identify effects influencing juvenile crocodile mortality and to estimate relevant genetic parameters.

6.3 Methods and materials

6.3.1 Experimental animals Data were collected at Janamba Croc Farm (Northern Territory, Australia) between 1994 and 2002. Progeny records from 29 pairs kept mostly in unitised (1 male, 1 female) breeding pens, although some pens consisted of 1 male and 2 females, were available. All eggs were collected usually within 24 hours of oviposition and incubated at 32oC as described in Chapter 2 (Section 2.1.2). This excludes the possibility of incubation temperature influencing survival rates as described by Webb and Cooper-Preston (1989). After hatching, the crocodiles were marked for clutch identification by cutting scutes in a unique sequence. Scutes are vertical, triangular osteoderms on the dorsal midline of the posterior tail that bifurcate into two rows of more laterally flattened scales in about the middle third of the tail and continue cranially (Richardson et al., 2002). Mortality data were collected on a daily basis during routine feeding and cleaning procedures. The dead animal’s clutch of origin and thus parentage were determined from its scute cuts. The date of death and a description of the crocodile’s condition upon death and/or a possible cause of death were recorded. Subsequently, the age of the animal and its year of hatch were determined and assigned to the juvenile’s record. In addition, to adjust for differences in the time of year when an animal hatched, a covariate hatch days (HDays) was calculated based on the number of days between the date of hatch and the 1st of January in that particular year. A total of 1302 mortality records were used in this study and coded as 1 to indicate premature death prior to slaughter. Juvenile crocodiles are generally slaughtered at between 1.6m and 1.9m to obtain a belly width between 35cm and 45 cm. Slaughtered crocodiles were treated as censored and coded 0. Slaughtering an animal is

55

a management decision so the record cannot be considered a death. Censoring a record allows the record to be included in the study, but considers these animals to have left the study before a natural mortality could be observed. There were 2151 slaughter records incorporated into these data. The average age of juveniles at slaughter is 2.84 ± 0.01 years (Table 5.2, Chapter 5). Therefore, the clutches that hatched in 1999 to 2002 had a considerable number of missing records since these animals were still in the production system. Again to overcome any bias introduced from not including these animals, another set of censored records was created. 1582 dummy records were created and censored (coded 0) to account for the study period ending before mortality could be observed. These animals were censored at the last date of data collection (December 2, 2003) for these animals. A total of 3733 (74.14%) records were right censored with an average censoring time of 1017.5 days.

6.3.2 Statistical methods Cox’s proportional hazards model was considered the most appropriate model for analysing these data. Survival kit V3.12 (Ducrocq and Sölkner, 1994; 1998) was used to evaluate the following model

ln[hijk(t)] = ln[h0(t)] + (βHDHDaysjk + Yeark + Pairj + Clutchjk) where hijk(t) is the hazard function for the ith individual from the jth pair in the kth year at time t, h0(t) is the unspecified baseline hazard function, HDaysjk is the number of days between hatching date and the 1st of January in that particular year for an individual from the jth pair in the kth year; βHD is the regression coefficient for HDays; Yeark is the fixed effect of the kth year (k = 1994,…,2002); Pairj is the random effect of pair (assumed N(0,σ2

Pair)); and Clutchjk is the common environment (random) effect of a clutch produced by the jth pair in the kth year (assumed N(0,σ2

Clutch)). A 5% significance level was chosen to evaluate explanatory variables by backward elimination.

6.3.3 Genetic parameter estimates Heritability of log-survival time was estimated from the full-sib correlation as follows (Ducrocq and Casella, 1996)

62Clutch

2Pair

2Pair2

logt 22hπ+σ+σ

σ×=

using the estimates of the variance components, 2Pairσ and 2

Clutchσ .

6.4 Results and discussion The Kaplan-Meier estimate of the baseline survival function for crocodiles between hatch and slaughter is shown in Figure 6.1. This plot shows the probability of a crocodile surviving to any given day, and it demonstrates a high morality rate over the first approximately 400 days. In fact, the probability of a crocodile surviving to day 400 is only 56% in this study. After day 400, the mortality rate reduces, confirming the observation and recommendation of Webb (1989) that mortality rates decrease after the first year.

56

Figure 6.1. Kaplan-Meier estimated baseline survival function for crocodiles between hatch and slaughter. The vertical line indicates one year of age (365 days). The large declining steps apparent after day 1600 (4.4 years) occur due to the limited amount of data around those time points since the animals are slaughtered at an average age of 3.5 years.

All terms in the survival model were significant with the exception of hatch days (P>0.05). Of the nine years of data analysed in this study, 1998 had the greatest number of recorded deaths (n = 259) and was used as the base to compare the other years. Table 6.1 shows the regression coefficients estimates (± SE), hazard ratios (exponential coefficients) and the antilog of the 95% confidence interval for the year effects expressed as deviations from the 1998 base year. 1998 was the first year that the hatchlings were placed into an environmentally-controlled nursery. The high number of deaths in 1998 (n = 259) was a result of optimising management routines, for example ventilation and stocking densities. Compared to 1998, hatchlings produced in the other years had lower hazards of mortality, varying between 48% (in 1994) and 4% (in 2001). In other words, hatchlings born in 1994 had a 48% risk of mortality compared to those hatchlings born in 1998, whilst hatchlings in the 2001 cohort had a 4% risk of mortality. Heritability on the log survival time scale for juvenile crocodile survival was estimated to be 0.15 (SE 0.04). This heritability estimate of survival is comparable to other livestock species. For example, Southey et al. (2001) estimated the heritability of lamb survival to be between 0.12 and 0.20 at different rearing stages, using Cox’s proportional hazards model. For laying hens, Ducrocq et al. (2000) reported heritability estimates of 0.48 for the rearing period and 0.19 for the production period, using Cox’s model and Weibull’s model, respectively. Using Gibbs sampling, van Arendonk et al. (1996) reported a heritability estimate of 0.11 for piglet survival, whilst in a review of piglet survival studies, Knol et al. (2002) reported that heritability estimates were on average 0.04.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 200 400 600 800 1000 1200 1400 1600 1800

Survival time from hatch to slaughter (days)

Estim

ated

sur

vivo

r fun

ctio

n

365 days

57

Table 6.1. Estimates (± SE) of year effects, their hazard ratios and the antilog of the 95% confidence interval using the Cox Proportional Hazards Model. The hazard ratio is the antilog of the estimate and represents the chance of mortality. All are expressed as deviations from 1998 since 1998 had the greatest number of uncensored deaths. n is the number of recorded deaths in each year.

95% Confidence Interval Year Estimate ± SE Hazard Ratio Lower Upper n 1994 -0.74 ± 0.49 0.48 0.18 1.24 34 1995 -1.20 ± 0.29 0.30 0.17 0.53 120 1996 -1.34 ± 0.27 0.26 0.15 0.44 147 1997 -0.89 ± 0.25 0.41 0.25 0.67 242 1998 - 1.00 - - 259 1999 -1.64 ± 0.25 0.19 0.12 0.31 197 2000 -1.56 ± 0.25 0.21 0.13 0.34 209 2001 -3.27 ± 0.30 0.04 0.02 0.07 29 2002 -1.85 ± 0.28 0.16 0.09 0.27 65

6.4.1 Risk factors There were insufficient details provided with these data to investigate some of the risk factors that cause crocodile mortality. Possible risk factors for future inclusion would be runting and disease status such as bacterial septicemia and fungal dermatitis. Runting is a condition of hatchling crocodiles whereby they fail to grow in comparison to the rest of their cohort and generally appear emaciated (anorexic). Runts constitute a large proportion of juvenile mortalities on Australian crocodile farms. Peucker and Mayer (1995) proposed that the condition is inherited and appears to be clutch related. Buenviaje et al. (1994) suggested that runting was a failure to adapt to a particular rearing or management environment. Fungal dermatitis and bacterial septicemia are caused by opportunistic microorganisms and also represent a large cause of mortality. Death due to bacterial septicemia is of particular concern since it appears to mainly affect the best growing animals, with poorly growing animals rarely dying. These faster growing animals are believed to experience a relatively higher level of physiological and nutritional stress (Ladds and Sims, 1990), rendering them more prone to infection. Since the Australian crocodile industry is still in the early stages of development, no selection for disease resistance has begun. It is recommended for future studies, that specific causes of mortality be recorded so that targeted risk analyses can be undertaken to enhance the selection process. Specific variables for consideration could include runting, fungus, disease (e.g. bacterial septicaemia), hatchling deformities (such as external yolk sac, kinked spine, etc.), or management, for example.

6.5 Implications Using a full-sib survival analysis, heritability on the log scale was estimated to be 0.15 (SE 0.04). This estimate was consistent with heritability estimates for survival in other livestock industries. There were significant year effects, although the underlying causes of these year differences were not amenable to investigation. The base year, 1998, was used for comparison since it was the year with the greatest number of uncensored deaths, due to optimising management regimes. It is suggested that more comprehensive details on cause of death be kept so that risk analyses can be performed in the near future. This would also allow selection for disease-resistance to begin. This study is the first to estimate and report heritability for crocodile survival for incorporation into a genetic improvement program.

58

7. Quantitative analysis of scale row number

7.1 Abstract A total of 3206 scale row records, comprising 1739 full-sibling records from 30 families from Janamba Croc Farm (NT, Australia) and 1467 parent-offspring records from 19 families from Wildlife Management International, Pty. Ltd. (NT, Australia), collected at each facility using a different method, were analysed using ASReml. The full-sibling heritability estimate for the Janamba data was 0.37 (SE 0.03). The animal model estimate of heritability for the WMI data, also based predominantly on full-sibling data, was 0.42 (SE 0.04). The counts from three counting methods were evaluated by regression analysis on 100 individuals and were found to be highly correlated. Using the regression relationship, the WMI data were transformed and pooled with the Janamba data to give an animal model heritability estimate of 0.42 (SE 0.04). A multi-trait analysis revealed negligible correlations (both phenotypic and genetic) between hatchling size traits and the number of scale rows. There is ample genetic variation to incorporate this trait into a genetic improvement program for farmed saltwater crocodiles.

7.2 Introduction The Australian crocodile industry is based on the production of high quality skins from the saltwater crocodile (Crocodylus porosus) for the international skin trade. For those animals whose skins meet the requirements of an export-quality, blemish-free skin, 80% of the total product value is derived from the skin, with the remaining 20% from the sale of meat (15%) and by-products (5%; backstrap, head/skull, feet). In 1998, the estimated total value of Australian crocodile production was A$5 million per annum, with skins contributing A$4 million of this total (Stubbs, 1998). The major export destinations for Australian skins are France, Japan, Singapore and Italy (MacNamara et al., 2003). The Australian industry is small in relation to world trade (MacNamara et al., 2003) contributing less than 1% of the world trade in skins crocodilian skins (Stubbs, 1998). Saltwater crocodile skins, with their relatively small scales, evenly distributed belly-scale pattern, and lack of bone deposits (osteoderms) in the belly scales, attract a premium price in comparison to the skins of other crocodilian species. Skins are sold to the fashion market on a ‘$ per centimetre’ belly-width basis. The price received is also dependent upon a stringent grading system. Skin quality traits, reflecting inherent properties of the hide such as shape and thickness, are not rewarded under the current marketing system. Manolis et al. (2000) surveyed persons involved in all aspects of the crocodile skin industry in an effort to understand the issues that face producers of saltwater crocodiles to enhance their product. Four “quality” concerns that could be influenced by genetic selection were raised: skin shape, skin thickness, regularity of scale pattern and number of scale rows between the cloaca (vent) and the neck. The present study focuses on estimating the genetic parameters for the number of scale rows alone, since data for the other quality traits were not available. With more scale rows on the belly skin comprised of smaller scales, consumer appeal of the product increases. Manolis et al. (2000) reported that product manufacturers considered that skins with more belly scales would produce a higher quality fashion product, for which they would be willing to pay a premium price if sufficient numbers could be produced. In their study, Manolis et al. (2000) reported an average of 31.2 scale rows (range 27-37) for saltwater crocodiles, although fewer than 2% of animals had more than 35 scale rows. Manolis et al. (2000) reported a significant positive effect of incubation temperature (31, 32 and 33oC) on the average number of hatchling scale rows (33.3, 34.2 and 35.0, respectively). The standard incubation temperature for saltwater crocodiles on farms is 32oC, which maximises embryo survival, produces mostly males and increases post-hatching growth rates (Richardson et al., 2002). Manolis et al. (2000) also conducted a preliminary investigation of parental effects on variation in scale row number at 32oC by

59

regressing average clutch scale row counts against their maternal and paternal scale row counts separately. Data were analysed independently using 14 families only for both years. They reported non-significant maternal regressions, whereas the paternal regression was significant (P = 0.01) in one year but not in the next year (P = 0.05). Neither regression coefficients nor derived genetic parameter estimates were reported in the study. The purpose of the present study to estimate genetic parameters for the number of scale rows for possible inclusion into a multi-trait selection program using standard mixed model methodology, and to investigate the basis for the difference between paternal and maternal regression analyses.

7.3 Methods and Materials

7.3.1 Experimental animals A total of 1739 offspring records were collected at Janamba Croc Farm from 30 pairs (average number of offspring per pair = 58; range 23-119) in 2001 and 2002. The eggs were incubated at 32oC as described in Chapter 2 (Section 2.1.2). All scale counts were recorded using a simple method (described below) on the day of hatch. In addition, hatchling head length (HHL), hatchling snout-vent length (HSVL) and hatchling total length (HTL) were measured for all hatchlings, as described in Chapter 2 (Section 2.3.2.2). All measurements were made on the live animal. Scale row data from Wildlife Management International (WMI) were collected in 1999 for the study by Manolis et al. (2000). Data collection continued after the completion of the Manolis et al. (2000) study, and so data from 1999 and 2002 inclusive were kindly provided for this study. Only offspring incubated at 32oC, all of which had their ventral scale rows photocopied, and had parental records were included in these analyses. A total of 1467 individuals were available for analysis from 19 known family groups (19 sires; 18 dams) with an average of 75 offspring per family (range 23-114). All parents were wild caught and their incubation temperatures were unknown. Descriptive statistics for the number of scale rows from each facility are given in Table 7.1.

Table 7.1. Summary statistics for the number of scale rows on the belly skin for 3206 crocodiles.

Year n Mean ± SD Range 1st–3rd Quartile

Janamba 2001 921 30.74 ± 1.21 26-35 30-32

2002 818 30.46 ± 1.43 25-34 30-31

Overall 1739 30.61 ± 1.25 25-35 30-32

WMI 1999 105 32.24 ± 1.43 29-37 31-33

2000 448 32.23 ± 1.27 28-35 31-33

2001 363 32.52 ± 1.33 29-36 32-33

2002 501 32.66 ± 1.34 29-37 32-34

Overall 1467 32.44 ± 1.53 28-37 31-33

7.3.2 Methodology for collection of scale row data Data were collected at the two breeding facilities using different counting methods. An example of each counting method is shown in Figure 7.1 on a typical belly region of a crocodile. Regardless of method, the number of scale rows was counted along the midline of the animal (yolk scar). The count was inclusive of the first complete row anterior to the cloaca and the row before the longitudinally elongated collar scales of the lower neck region. A further description of each method is given below.

60

Figure 7.1. Belly region of a crocodile. Number of scale rows is counted between the cloaca and the row before the collar (longitudinally elongated neck scales). This individual (counting the asterisks) has 29 scale rows using the Simple method, and 30 using the WMI method (counting the plus signs).

Collar

Cloaca (vent)

Area for scale row count (inclusive)

****

**

*

***

*

*

*

*

***************

+

++++

++++++

++

+

+++

+

+

+

+

++

++

+

+

+++

61

7.3.2.1 Simple method. This method involves counting the number of scales on the left of the midline in a “count-them-as-they-appear” manner. That is, even if the row did not continue to the lateral extremities, it was still included in the count. 7.3.2.2 WMI method. This method involves counting the scale rows on the right hand side of the midline, but if there were small “split” scales, or greatly enlarged “fused” scales on the midline, making a determination as to whether they constituted one or two “real” ventral scale rows. Some examples are given in Figures 7.2a-d. If a row began on the midline and then split into two rows after one or two single scales, then it was counted as two rows (Figure 7.2a). However, if it split into two rows after three or more single scales, it was counted only as one row (Figure 7.2b). Similarly, if two rows began on the midline and continued for three or more scales before combining into one row, they were counted as two rows (Figure 7.2c). If two rows converged before three scales, this was counted only as one row (Figure 7.2d).

Figure 7.2. Examples of WMI method for counting scale rows. The heavy vertical line represents the midline. The first three scales dictate whether a row is counted as a “true” row or not. Figures 2a and 2c were counted as two rows, whereas Figures 2b and 2d were counted as one.

7.3.3 Comparing counting methods To compare the different counting methods, 100 individuals were counted using both the Simple method and the WMI method. In addition, the Simple method (which, as mentioned above, was normally counted on the left of the midline) was used to count the scales on the right of the midline for these individuals. This count was called SimpleR and allows a direct comparison between the number of scales on the left and right sides of the midline. A summary of scale row counts from these animals using the three different methods is in Table 7.2. Three regression analyses were conducted in GenStat 6 (version 6.1, 2002) to determine the linear relationships between the different counting methods.

Table 7.2. Summary statistics for the comparison between scale row counting methods.

Method Side of midline N Mean ± SD Range 1st–3rd Quartile

Simple Left 100 32.67 ± 1.6 29-37 31-34

SimpleR Right 100 33.53 ± 1.5 30-37 32-35

WMI Right 100 33.49 ± 1.5 30-37 32-35

˙ ˙ ˙ ˙˙ ˙

˙˙

˙˙

˙˙

˙˙ ˙ ˙

˙ ˙ ˙

˙˙ ˙ ˙ ˙

˙˙

c d

b a 2 rows 1 row

62

7.3.4 Full-sib heritability estimate A univariate analysis of scale row data collected from Janamba Croc Farm was conducted in ASReml (version 1.10, 2003; Gilmour et al. 2002). The initial model was

(J)ijk k j ijkSR Year Pair= µ + + + ε

where (J)ijkSR is the number of scale rows on the ith individual from the jth pair using the Simple method of

counting from Janamba Croc Farm only; µ = the overall mean; Yeark is the fixed effect of the kth year (k = 2001 and 2002); Pairj is the random effect of pair (assumed N(0,σ2

Pair)); and εijk is the random residual effect (assumed N(0,σ2

ε)). Using the above model, scale row heritability was estimated using a full-sib correlation as follows

22Pair

2Pair2 2h

εσ+σ

σ×=

using the REML estimates of the variance components, 2Pairσ and 2

εσ .

7.3.5 Correlation with hatchling morphometric measurements Hatchling morphometric traits (HHL, HSVL and HTL) were recorded in conjunction with the number of scale rows for the animals from Janamba Croc Farm. The above scale row model was then combined into a multi-trait analysis with the univariate models for HHL, HSVL and HTL as described in Chapter 5. Genetic (rg) and phenotypic (rp) correlations between the number of scale rows and the hatchling morphometric traits were estimated using the standard equations (Searle, 1961).

7.3.6 Animal model Although the WMI data were predominantly records of full-siblings, an animal model analysis was used in ASReml to incorporate the parental scale row records in the analysis and utilise all available pedigree information, as described by Mrode (2000)

ikik(WMI)ik AnimalYearSR ε+++µ=

where (WMI)ikSR is the number of scale rows on the ith individual using the WMI animals and counting

method; µ = the overall mean; Yeark is the fixed effect of the kth year (k = 1999,…,2002); Animali is the random effect of the ith individual (assumed N(0,σ2

Animal)); and εik is the random residual effect (assumed N(0,σ2

ε)). Heritability was then estimated as follows

22Animal

2Animal2h

εσ+σ

σ=

again using the REML estimates of the variance components.

7.3.7 Combining the data-sets The WMI data were transformed using the linear regression developed above (in section 7.3.3) and combined with the Janamba data for analysis using the univariate animal model in ASReml. The model included fixed effects for Year (as described above) and another generic term “Farm” (encoded 1=WMI; 2=Janamba) to account for inter-farm variation.

63

7.4 Results and Discussion

7.4.1 Comparing methods The descriptive statistics of the three different methods (Simple, SimpleR, WMI) for counting scale rows using the 100 individuals are presented in Table 7.2. There are highly significant linear relationships (Table 7.3) between the three counting methods and the regression equations can be used to transform data from one counting system to another. Besides increasing the number of scale rows on a belly skin, some product manufacturers also indicated that a more regular scale pattern was important (Manolis et al., 2000). Scale pattern regularity relates to the degree that a scale row divides into additional rows as it continues laterally (some examples in Figure 7.2). Although the Simple method is easier to conduct on the live animal, it provides no indication of scale pattern regularity. In contrast, the WMI method attempts to incorporate an index for scale pattern regularity in conjunction with counting the number of rows by only including those scale rows that form a more or less complete row. No study has yet been conducted to investigate the environmental and/or genetic effects on scale pattern regularity, or to recommend a measurement index for such a trait. It would be advantageous to conduct such a study before selection for scale row number commences to avoid any antagonistic effects between the two traits. In addition, photocopying the ventral aspect of each hatchling provides a valuable resource for future analysis should a scale pattern regularity index be created. Table 7.3. Simple linear regression coefficients (± SE) describing the difference between the number of scales on the left and right of the midline using the same counting method (Simple and SimpleR), and the difference between the counting methods (WMI and SimpleR). In addition, the equation to convert the WMI scale row counts to the Simple method for a combined analysis is also given. ** represents P < 0.01, whilst *** indicates P < 0.001. Correlation values are also given.

Explanatory variate Intercept Slope Correlation

Simple SimpleR 9.01 ± 2.52*** 0.71 ± 0.07*** 0.69

SimpleR WMI 7.25 ± 2.33** 0.78 ± 0.07*** 0.75

Simple WMI 6.75 ± 2.49** 0.77 ± 0.07*** 0.73

7.4.2 Correlation with hatchling traits All of the genetic correlations between the number of scale rows and the hatchling morphometric traits were estimated to be below 0.06 (SE 0.20, Table 7.4), whilst the phenotypic correlations were all below 0.05 (SE 0.05, Table 7.4). This indicates that the size of the hatchling has little, if any, effect on the number of scale rows on the belly skins of the crocodiles. Manolis et al. (2000) reported the temperature-sensitive period for scale row development to be before day 30 at both 30oC and 33oC. Therefore, the number of scale rows on an individual has been determined long before egg size could influence hatchling size (Webb et al., 1987).

Table 7.4. Genetic (rg) and phenotypic (rp) correlations between number of scale rows and hatchling head length (HHL), snout-vent length (HSVL) and total length (HTL). SE in parentheses.

(rg) (rp)

HHL 0.06 (0.20) 0.02 (0.05)

HSVL 0.06 (0.20) 0.05 (0.05)

HTL 0.03 (0.20) 0.03 (0.05)

64

7.4.3 Heritability estimates Heritability estimates and their approximate standard errors for all models are given in Table 7.5. The heritability estimate from the Janamba full-sibling analysis was 0.37 (SE 0.03). Utilising all available pedigree information from the WMI data, the heritability estimate was 0.42 (SE 0.04). These data were predominantly full-sibling records, although a few half-sibling records were also present. After transforming the WMI data set using the regression equation (Table 7.3), another animal model was fitted to the combined data. The resulting heritability estimate from all available data for this study was also 0.42 (SE 0.04).

Table 7.5. Estimates of variance components and heritability for the number of scale rows for progeny incubated at 32oC constant temperature. The Janamba data were analysed using a full-sibling correlation model, whilst the WMI data and the combined data were analysed using animal models. SE’s are in parentheses.

Janamba only WMI only Combined data

Model type Full-sibling Animal Animal

Variance component estimates (SE)

Pair/Animal 0.34 (0.10) 0.83 (0.22) 0.64 (0.13)

Residual 1.49 (0.05) 0.98 (0.12) 0.90 (0.07)

Heritability 0.37 (0.03) 0.42 (0.04) 0.42 (0.04)

These estimates are most likely upwardly biased since these data consisted of measurements taken from full-sibs and because of confounding between the genetic variance and common environment (or dominance) effects (Falconer and Mackay, 1996). When more complex pedigree data becomes available, re-estimation of these parameters should be undertaken.

7.4.4 Confounding effects of incubation temperature Manolis et al. (2000) found a significant, although variable, linear regression between paternal and average progeny scale row counts for the years included in their analyses (1998 and 1999). In contrast, they found no significant relationship found between maternal count and average progeny scale row counts in 1998 and 1999. The WMI data for 1998 and 1999 were re-analysed and included the additional data from 2000 and 2001. The results are presented in Table 7.6. The results in Table 7.6 confirm the observations of Manolis et al. (2000) that there is no significant regression of progeny scale row counts on maternal counts for any of the years, nor when the years were pooled. Variable results were obtained for the paternal regression analyses. The separate year analysis for 1999 and 2001 showed a significant regression of progeny count on paternal count, but not in 1998 or 2000. The significant regression of progeny count on paternal count, but not maternal count, was also supported when the data from all years was pooled. Conversion of these regression slopes to heritabilities gives h2 of 0.52 (SE 0.13) and -0.14 (SE 0.25) for paternal and maternal data respectively (Falconer and Mackay, 1996). The heritability estimated from the paternal regression is consistent with the estimates in Table 7.5, but the estimate from the maternal regression analysis is outside the permissible parameter space (is negative) and is clearly not significant from zero. Manolis et al. (2000) hypothesised that male parents contribute genetically to variation in progeny scale row number but female parents do not. However, a biologically mechanism for this is not immediately apparent. The effect of sex-linkage can be excluded since crocodilians exhibit temperature-dependent sex determination, and do not have sex chromosomes (Lang and Andrews, 1994). Maternal negative imprinting could provide a mechanism for the Manolis et al. (2000) hypothesis above. However, such

65

imprinting effects reported in the literature are restricted to single loci. The variation in scale row number is presumably polygenic. For imprinting to explain these observations, it would have to apply in the same direction to all loci, which is a priori unlikely.

Table 7.6. Separate regression equations for male (MSR) and female (FSR) scale row count with the average of their progeny (SR) for the years 1998 to 2001. Data over all years were pooled for the “combined” year analysis. R2 and P-values for the regression equations are also given. ‡ indicates differences between the results presented by Manolis et al. (2000) and herein, presumably caused by omissions between the data-sets.

Year Regression equation R2 P-value

1998 Male

Female

SR = 24.77 + 0.24(MSR)

SR = 33.37 – 0.04(FSR)

0.21‡

0.05‡

0.119‡

0.827‡

1999 Male

Female

SR = 18.91 + 0.42(MSR)

SR = 39.89 – 0.24(FSR)

0.43‡

0.16‡

0.008‡

0.139‡

2000 Male

Female

SR = 26.07 + 0.21(MSR)

SR = 38.49 – 0.19(FSR)

0.28

0.25

0.064

0.085

2001 Male

Female

SR = 16.46 + 0.51(MSR)

SR = 37.84 – 0.17(FSR)

0.59

0.06

<0.001

0.340

Combined Male

Female

SR = 24.16 + 0.26(MSR)

SR = 34.59 – 0.07(FSR)

0.24

0.02

0.034

0.595

A simpler explanation for the anomaly between progeny regressions on paternal and maternal counts is possible. Crocodiles exhibit temperature-dependent sex determination (Lang and Andrews, 1994). At constant incubation temperatures of 28oC, 29oC and 30oC, Lang and Andrews (1994) reported 100% females being produced, whilst at 31oC, 32oC and 33oC, the percentage of male hatchlings was found to be 16%, 86% and 17%, respectively (Figure 7.3). All progeny in the present study were incubated at 32oC and, as reported by Manolis et al. (2000), an intermediate scale row number was observed (Figure 7.3). However, all parents are wild caught. Thus, while it is likely that male parents were incubated during the critical phase of development at 32oC, female parents are much more likely to have been incubated at temperatures either side of 32oC. This introduces a substantial additional source of environmental variability into the maternal scale row counts, which could be as large as ±2 scale rows. Such a large environmental effect would be very likely to overwhelm the effect of genes affecting the maternal phenotypes. In this circumstance, the genetic contribution of the mother to her offspring is not going to be well predicted by the maternal phenotype, and the offspring-maternal regressions would be predicted to be zero or near to zero. There is an unresolvable confounding of genetic effects with environmental effects which will cancel out any maternal genetic signal. For the male parents, incubated most likely at 32oC, this environmental source of variation in scale row count will be reduced or absent when compared to progeny also incubated at 32oC. In this case, the genetic contribution of male parents to their progeny is less likely to be obscured and thus the significant positive regressions of progeny on male parents will be observable. To fully investigate these phenomena, standardising the incubation temperature of the parental generation is necessary. Analyses of full-sib records only are not affected by these problems, since all progeny were incubated at 32oC at both Janamba Croc Farm and WMI.

66

32

32.5

33

33.5

34

34.5

35

35.5

29.5 30 30.5 31 31.5 32 32.5 33 33.5

Incubation temperature (oC)

Aver

age

Num

ber o

f Sca

le R

ows

0

10

20

30

40

50

60

70

80

90

100

% m

ales

Average SR

%males

Figure 7.3. Percentage (%) male (indicated by curve; adapted from Lang and Andrews, 1994) and average scale row number (squares with linear trendline; adapted from Manolis et al., 2000) versus constant incubation temperature (oC).

7.4.5 Possible imprinting effects on scale row number Although a genetic explanation is probably not necessary for the difference between the paternal and maternal regression results, genetic imprinting will be explained here briefly. Genetic imprinting occurs when the expression of an allele is determined by the sex of the parent from which it was inherited (Nicholas, 2003). A good example in pigs is the paternally-determined expression of the insulin-like growth factor 2 gene, IGF2 (Van Laere et al., 2003). Segregation of IGF2 variants affects muscle growth, fat deposition and size of the heart in pigs. This QTL is one of many that contribute to variation in muscle growth (15-30%) and fat deposition (10-20%) in pigs (Van Laere et al., 2003). Van Laere et al. (2003) showed that the paternally-inherited variants were exclusively responsible for the effects of this locus on muscle mass and back-fat thickness. One method of testing the imprinting hypothesis on crocodile scale-row count would be to conduct systematic matings (illustrated in Figure 7.4). The four crosses required are A) a high scale-row male crossed with a high scale-row female, B) a high scale-row male crossed with a low-scale row female, C) a low scale-row male crossed with a high scale-row female, and D) a low scale-row male crossed with a high scale-row female. If genetic imprinting is implicated in variation in scale-row phenotype, the number of scale rows from matings A and B will be equal but greater than C and D, i.e. (A = B) > (C = D). However, if genetic imprinting is not responsible, then the number of progeny scale rows on B will be greater than A, and also D greater than C (i.e. (B > A) = (D > C)). This trial would have to be conducted under constant incubation temperatures (for both the parental and progeny generations).

Figure 7.4. Systematic matings to test hypothesis that genetic imprinting is responsible for scale row expression. The “high” and “low” indicate scale row number of the parental populations and the resultant progeny from the matings will be A, B, C, or D. It is the number of scale rows in the progeny that will determine whether the hypothesis is valid.

Male

high low

high A C

Fem

ale

low B D

67

7.5 Implications Heritability estimates for the number of scale rows (at 32oC) were high (0.37 and 0.42), although estimates derived predominantly from full-sibling data may be upwardly biased. Nevertheless, there is clearly substantial additive genetic variation for this trait which could be exploited in a genetic improvement program for farmed saltwater crocodiles. The negligible genetic correlation between the number of scale rows and the hatchling size traits indicates that selective improvement of scale row count can be achieved without adverse effects on hatchling size traits. Before commencing selection for scale row number, it would be advantageous to investigate the genetic effects, and create a measurement index, of scale pattern regularity to avoid any possible antagonistic selection effects.

7.6 Special acknowledgements Special thank you to Professor Grahame Webb and Mr Charlie Manolis from Wildlife Management International Pty. Ltd. for allowing me to use their scale row data presented in this chapter and further extend the usefulness of the genetic parameter estimates. Also, for providing comments on the manuscript.

68

8. Analysis of microsatellites and parentage testing in saltwater crocodiles1 8.1 Abstract Fifteen microsatellite loci were evaluated in farmed saltwater crocodiles for use in parentage testing. One marker (C391) could not be amplified. For the remaining fourteen, the number of alleles per locus ranged from 2 to 16, and the observed heterozygosities ranged from 0.219 to 0.875. The cumulative exclusion probability for all fourteen loci was 0.9988. The eleven loci that showed the greatest level of polymorphism were used for parentage testing with an exclusion probability of 0.9980. Using these eleven markers on 107 juveniles from 16 known-breeding pairs, a 5.6% pedigree error rate was detected. The usefulness of these markers was also evaluated for assigning parentage in situations where maternity and/or paternity may not be known. In these situations, a 2% error in parentage assignment was predicted. It is recommended that more microsatellite markers be used in these situations. The use of these microsatellite markers will broaden the scope of a breeding program, allowing progeny to be tested from adults maintained in large breeding lagoons, for selection as future breeding animals.

8.2 Introduction The Australian crocodile industry is a new and emerging industry based primarily upon the production of skins from the saltwater crocodile (Crocodylus porosus). For this industry to meet the demand for a high quality product, the adoption of a genetic improvement program is essential. The program will be based on the selection of candidates based on their own and their relatives’ phenotypic performances for selection criterion related to defined selection objectives (Isberg et al., 2003). Self evidently, the implementation of a successful breeding program will require correct pedigrees. Errors in assigned parentage decrease the accuracy of genetic evaluation of candidates, resulting in realised genetic improvement being less than expected (Visscher et al., 2002). Only a small number of adult crocodiles on Australian crocodile farms are kept in unitised breeding pens (1 male: 1 or 2 females). The majority of adults are maintained in breeding lagoons containing many males and females (Webb, 1989). In these situations, matings are neither observed nor recorded. It would be advantageous for industry, and for the success of a genetic improvement program, to include these animals and their offspring in a multi-trait selection program. A common method of marking a crocodile for identification is to cut scutes in a unique sequence. Scutes are vertical triangular osteoderms on the dorsal midline of the posterior tail that bifurcate into two rows of more laterally flattened scales in about the middle third of the tail and continue cranially (Richardson et al., 2002). In captive breeding, this method of marking is usually done on the day of hatch. This method, as described by Richardson et al. (2002), is one of the few permanent and practical identification methods available, although it is not without problems. Errors can result from incorrect cutting of the scute sequence at hatch, incorrect reading of the scute numbers later, or numbers changing over time due to the regrowth of incorrectly cut scutes. Other methods include toe tagging and microchips, which have their own disadvantages. If parentage errors are high, the success of a multi-trait selection program utilising records from relatives will be seriously compromised. Previously published microsatellite markers for crocodiles (FitzSimmons et al., 2001) have enabled a parentage determination kit to be assembled and evaluated in this study. The markers evaluated provide an indispensable complement to a selection program for farmed saltwater crocodiles, enabling confirmation of the pedigree records of juveniles from pens of known parents.

1 Accepted for publication as a research article in Journal of Heredity

69

Further, they widen the scope of the program by enabling the pedigrees of offspring produced from mass matings in lagoons to be deduced.

8.3 Materials and methods

8.3.1 Animals and sampling A total of 139 individuals were sampled from Janamba Croc Farm (Northern Territory, Australia). There were 16 known family groups consisting of 32 parents with an average of 6.7 offspring per family. Parents were long-term, known-breeding pairs housed in unitised pens. The parents were wild-caught and were assumed unrelated. Offspring from these pairs were collected as clutches of eggs, developed in an incubator and uniquely marked at hatching by scute cutting. Blood for DNA preparation was collected at various opportunities during the production system (either at the end of their first year or at slaughter). Blood samples were taken from eight adults and all of the juveniles using the occipital venous sinus technique described by Lloyd and Morris (1999), whilst tissue samples were taken from the remaining adults using a specifically-designed biopsy punch at any accessible location along the tail.

8.3.2 Experimental protocol DNA was extracted using standard phenol/chloroform protocols (Sambrook et al., 1989). Fifteen of the twenty-six microsatellites developed by FitzSimmons et al. (2001) were randomly chosen for this study (Table 8.1). For every microsatellite locus, the amplification reaction took place in a total volume of 15µL. PCR reagents included 1 unit of Taq DNA polymerase (various sources), 1X PCR buffer (Promega) and final concentrations of 0.1 mM dNTPs, 0.3 mM each of forward and reverse primer, 1.3-3.3 mM MgCl2 and approximately 50-100 ng of template DNA. Standard PCR conditions included a touchdown protocol with an initial denaturation at 95oC for 15 mins, followed by 3 cycles of 95oC for 40 sec, 63 oC for 1 min and 72 oC for 1 min 30 sec, followed by 5 cycles of 95oC for 40 sec, 61 oC for 1 min and 72 oC for 1 min 30 sec, followed by 35 cycles of 95oC for 40 sec, 59 oC for 1 min and 72 oC for 1 min 30 sec, finally being held at 72oC for 20 min. PCR products were run on a denaturing 6% polyacrylamide gel using an ABI 373 sequencer (Applied Biosystems, Inc.) and alleles were scored using GeneScan and Genotyper software (Applied Biosystems, Inc.).

8.3.3 Microsatellite and population genetic analysis Initially, only the parents were evaluated for all fifteen markers to determine the degree of polymorphism. Only those loci with two or more alleles were included in the population genetic analysis (11 loci; Table 8.1). Tests of linkage disequilibrium between loci were done using Arlequin 2.000 (Schneider et al., 2000). Allele frequencies for each locus were estimated using Cervus 2.0 (Marshall et al., 1998), as were the probabilities of exclusion. Cervus was used for parentage testing using a typing error rate of 0.01 with strict and relaxed confidence levels specified as 95% and 80%, respectively. Of the estimated 250 breeding animals on Janamba Croc Farm, only 32 parents were sampled (12.8% of the Janamba population) and each had all 11 loci typed. To identify the most likely parent, Cervus 2.0 uses a logarithm of the likelihood ratio (LOD score), as described by Marshall et al. (1998). After evaluating the likelihood of each possible combination of parents for a given offspring, the resulting LOD score produced for each possible parent is ranked and the parent with the highest LOD score is considered the most likely parent (Jones and Ardren, 2003). For the purposes of this study, there were three ways in which Cervus 2.0 was used to assign parentage. First, since the family groups used in this study came from unitised breeding pens where males and females are permanently paired, the markers were used to confirm correct data entry, scute cutting and animal identification, since mating errors were not possible. To simulate a situation where parentage is unknown, such as those in the breeding lagoons, two scenarios for parentage assignment were tested: (i) identifying

70

the male parent when the female is known, and (ii) identifying either parent with no prior knowledge of the other.

Table 8.1. Microsatellite loci trialled on 32 adult C. porosus for use in a parentage determination kit. For each locus, the number of alleles (k) and size range (Size) detected, effective number of alleles (NA), heterozygosity observed (HO) and expected (HE), and probability of exclusion was revealed.

Probability of Exclusion

Locus k Size NA HO HE

Neither parent

known

One parent

known

Cj127 16 353-415 6.54 0.813 0.861 0.542 0.705

Cj131 8 228-242 5.20 0.875 0.82 0.448 0.624

Cj101 6 345-367 3.29 0.625 0.707 0.281 0.45

CUD68 6 137-147 2.33 0.563 0.58 0.182 0.344

Cj16 6 156-187 2.46 0.719 0.603 0.199 0.369

Cj18 5 185-228 4.11 0.750 0.769 0.341 0.518

Cj105 4 365-371 1.92 0.563 0.488 0.116 0.224

Cp10 4 196-204 2.98 0.594 0.675 0.247 0.419

Cj119 4 178-188 2.86 0.594 0.66 0.229 0.393

Cj104 3 206-210 2.54 0.813 0.616 0.184 0.328

Cj122 3 375-387 1.25 0.219 0.201 0.202 0.095

Cj35A 2 159-161 1.88 0.375 0.476 0.110 0.179

CU4-121 A 2 178-180 1.56 0.469 0.365 0.064 0.147

CUJ-131 A 2 201-203 1.40 0.344 0.289 0.041 0.122

C391 B - - - - -

Average 5.07 2.88 0.594 0.579

Overall 0.9742 0.9988 A indicates that only the parents were evaluated due to PCR optimisation difficulties. B

indicates that marker did not amplify under any conditions tested

71

8.4 Results Of the fifteen loci evaluated, only eleven were genotyped on all available samples. C391 did not amplify using conditions recommended by FitzSimmons et al. (2002) or various other conditions tried. Cj35, CUJ-131 and CU4-121, whilst scored for the adult samples, were considered insufficiently polymorphic for inclusion into a parentage determination kit with each only revealing two alleles in the population sampled. Descriptions for each locus are in Table 8.1. The number of alleles detected ranged from two to 16 (average 5.1), whilst the effective number of alleles ranged between 1.3 and 6.5 (average 2.9). Genotype frequencies of the adult animals were within expectations of Hardy-Weinberg equilibrium at each locus (P > 0.05), with the exception of Cj104 (P = 0.032). However, this locus was kept in the analysis since minor deviations from Hardy-Weinberg equilibrium at few loci are unlikely to bias likelihood estimates considerably across all loci (Marshall et al., 1998). In addition, since only 32 adults (that were wild-caught from various locations) were sampled, these animals may not represent a sample from a natural population of the Australian saltwater crocodile. Tests of pairwise linkage disequilibrium, using a Markov chain, were conducted in Arlequin 2.000 (Schneider et al., 2000) and were significant for six from 120 tests, equivalent to the number expected by chance alone. Cj127 was the most informative locus with a probability of exclusion of 0.705, whilst the least informative was Cj122 (probability of exclusion = 0.10). Using all 14 loci, the combined probability of excluding the second parent when the first is known was 0.9988, whilst the probability of excluding the first parent when neither parent is known was 0.9742. After omitting the three loci that displayed only two alleles, the average probability of exclusion when one or neither parent is known is 0.9980 and 0.9678, respectively.

8.4.1 Confirmation of correct parentage assignment Using 11 markers, 5.6% of the juveniles were found to have genotypes incompatible with their recorded parentage. These individuals were re-tested to confirm they were not due to sample misidentification. In each case, the genotypes of at least four loci excluded these parents. Thus the animals must have been marked incorrectly when the scutes were cut, or the scutes were read incorrectly or the scute cut sequence had been modified by biting. Laboratory errors cannot be excluded but great care was taken with labelling and identification of samples. The multiple locus exclusion in each case means that mutation need not be considered as a reason for incompatible genotypes. These animals were omitted from analyses aimed at predicting parentage since the adults sampled came from unitised pens of known breeding pairs. No other adult crocodiles on the farm were sampled, so the true parents could not be identified.

8.4.2 Simulating a situation where one or neither parent is known

8.4.2.1 Female Parent is Known In some situations, the female parent of a clutch can be identified when collecting the eggs, due to maternal nest protection behaviour. In these situations where the mother is known, Cervus 2.0 (Marshall et al., 1998) provides the option of specifying one parent as known. All 16 males were then treated as potential fathers. For 99 of the juveniles used in this simulation, the known father was designated the most-probable father. However, for two of 101 juveniles, the most-probable father assigned was known to be incorrect. In these two cases, the true father, known to be located in the same pen as the mother, was the second most likely.

8.4.2.2. Neither Parent is Known In some cases, the mother cannot be identified whilst collecting the eggs from the nest. To simulate a situation where neither parent is known, the assignment of parentage was divided into two parts. First, the most likely parent was predicted from all possible parents available. For the animals used in this analysis,

72

100% were assigned correctly to the most likely parent. Secondly, the analysis was rerun with the most probable parent specified as known, whilst excluding all other potential parents of the same gender as the known parent. Again, all except for the two juveniles mentioned above, were assigned their correct other parent as the most-probable other parent. As mentioned previously, parentage allocation is achieved using exclusion probabilities, where a higher probability is given to a potential parent that is homozygous for the alleles present in the offspring across the greatest number of loci. Therefore, including extra microsatellites, even the three lowly informative markers, may have given the extra exclusion power required for 100% correct filiation. Alternatively, genotype information from siblings within the same clutch could be used to provide extra information about the true father’s genotype.

8.5 Discussion and implications Pedigree errors were shown to occur in 5.6% of cases tested in this study. This is within the ranges of pedigree errors revealed in other animal industries (8.7%-15.5% in sheep: Barnett et al., 1999; 2%-22% in cattle: references within Visscher et al., 2002) using microsatellite markers. The parentage errors detected in this study could have resulted from incorrect scute identification or less likely from sample misidentification, since the adults sampled were from known unitised breeding pens where mating errors were not possible. The eleven microsatellites evaluated in this study provide sufficient power (99.80%) to be useful in a parentage determination kit to confirm parentage of offspring from unitised breeding pens for selection as future breeding animals. Not all of the adults on the study farm were sampled. Thus, the true parents of these 5.6% of mis-assigned individuals could not be determined. The remaining adults on the study farm had unrestricted and unobserved choice of mates. These eleven microsatellites have shown adequate exclusionary power to determine parentage of these clutches retrospectively. This will provide the opportunity to include the progeny from animals maintained in uncontrolled lagoon breeding situations to be included in a multi-trait selection program. The possibility of multiple paternities in clutches from lagoons may require a greater probability of exclusion, which would be provided by using more loci.

8.6 Special acknowledgements Thanks to Wayne Gurney and John Nash for help in collecting DNA samples, and to Jaime Gongora, Zung Doan and Adam Stow for their guidance and invaluable technical advice. Thanks also to Nancy FitzSimmons, Jake Gratten and Tony English for advice on designing the tissue biopsy punch with special thanks to Mr John Olsen and Mr Björn Isberg for manufacture and donation of the tissue biopsy punch.

73

9. Designing and implementing the genetic improvement program: CrocPLAN 9.1 Abstract The previous chapters have aimed at estimating the genetic and phenotypic parameters for the breeding objectives and their relevant selection criteria. The estimated breeding values, or crocodile breeding values (CBVs), produced from the analyses in the previous chapters are presented in this chapter. Along with the relative economic values reported by Ms Emily Gray (Appendix II), the CBVs are incorporated into the crocodile economic selection index ($CESI). After ranking the breeding pairs from Janamba Croc Farm using $CESI, the pair B16 had the largest $CESI value (+$4,748), whilst B01 has the lowest value (-$5,257). In addition to these results, Janamba Croc Farm is used as a case study to recommend methods of future selection of candidates. The response to selection was predicted to be a $324 increase in profit per annum per breeding pair, assuming no genetic and phenotypic correlation amongst breeding objectives.

9.2 Crocodile economic selection index ($CESI)

9.2.1 Estimated crocodile breeding values (CBVs) for the breeding objectives In Chapters 4 to 7, heritability and repeatability estimates derived from the analyses were presented. For the reproductive and slaughter-age breeding objectives, these were obtained using a multivariate analysis with all possible selection criteria included. In contrast, the survival and scale-row variance component estimates were obtained using univariate analyses. In addition to these variance-component estimates, BLUP-EBVs were also calculated for each breeding pair. Separate sire and dam breeding values could not be estimated due to the structure of the data. Therefore, only pair EBVs were available for incorporation into the crocodile economic selection index ($CESI). Just as “ABVs” are the estimated breeding values for the Australian Dairy Industry, it is proposed that crocodile breeding values be termed CBVs. CBVs are reported separately for each breeding objective below.

9.2.1.1 Number of hatchlings produced The reproduction breeding objective investigated in Chapter 4 was the number of hatchlings produced per female per year (NoHatch). Possible selection criteria also included in the analysis were initial clutch size, number of viable eggs, hatchability, average hatchling snout-vent length and time of nesting. Only repeatability could be estimated since no pedigree structure was available. All repeatability estimates were high, ranging from 0.24 (hatchability) to 0.68 (initial clutch size and time of nesting). Phenotypic correlations between the traits ranged from negligible (0.03) to high (0.86). All traits were kept in the multivariate analysis for the estimation of CBVs (± SE) for NoHatch, shown in Figure 9.1.

74

-14

-10

-6

-2

2

6

10

14

18

UB

03U

B05

UB

16U

B04-1

UB

07B

06U

B04-2

UB

06-1B

09U

B13

UB

08U

B10

UB

06-2B

16B

15B

05B

14B

19B

04U

B01

B12

B02

B10

B07

B03

B13

UB

09B

11B

01B

20

Pair

No. H

atch

CBV

Figure 9.1. Pair CBVs (± SE) for the reproduction breeding objective, number of hatchlings produced per clutch per year (NoHatch). These CBVs were produced from the ASReml multivariate analysis described in Chapter 4.

The CBVs shown in Figure 9.1 range between -8.75 hatchlings (B20) and 15.09 hatchlings (UB03), a difference of 23.84 hatchlings per female per year. These CBVs are expressed as deviations from the average herd CBV centred at zero. Mating of a pair of offspring from parents both with a CBV of 15.09 like the pair UB03 will produce 15.09 hatchlings per annum greater than a pair whose parents have average breeding value, whilst progeny from B20 will have a reproductive performance of 8.75 hatchlings less than the herd average. One must be careful when interpreting these results. Unlike the other pairs in the analysis, the male and female in B20 are known-age, captive-raised crocodiles. The female was bred at Janamba Croc Farm, whilst the male was hatched from a wild-harvested egg. The pair first nested when the female was six and the male ten years of age. A more thorough description of this pair was given in Chapter 4 although to re-iterate, the first clutch this pair produced (in 1998) consisted of 35 small eggs, of which 25 were infertile and the embryos from the other 10 died during incubation. Since then the clutch size has increased to be slightly better than the farm average, although the number of hatchlings produced is still slightly below the farm average. Care must therefore be taken when comparing animals of young age, such as B20, to animals of presumed older age.

9.2.1.2 Age at slaughter The age when an animal is slaughtered directly influences the cost of production on crocodile farms. An investigation of this trait was presented in Chapter 5 and included the breeding objective, age at slaughter, and possible selection criteria including measurements taken at hatching and inventory (average age ~ 9 months). Heritability estimates were all high, being 0.40 and 0.60 for age at slaughter and hatchling snout-vent length, respectively. The genetic (-0.81) and phenotypic (-0.82) correlations between slaughter age and inventory head length were high, whereas the correlation estimates between hatchling snout-vent length and the other traits were either low and/or unreliable due to large standard errors. The resulting CBVs (± SE) for slaughter age are shown in Figure 9.2. The range of CBVs shown in Figure 9.2 is between -158 days (B16) and 144 days (B01), with a difference of 302 days. These results indicate that future offspring from pair B16 will be predicted to reach slaughter size 158 days before the herd average, whilst offspring from pair B01 take an additional 144 days to reach the marketable size.

75

-200

-160

-120

-80

-40

0

40

80

120

160

B16

B07

UB

08U

B07

B04

B12

B20

UB

01B

03U

B09

UB

06-2U

B16

UB

05U

B10

B02

UB

06-1U

B04-2

B14

UB

04-1U

B13

B19

B10

B11

B05

B06

UB

03B

13B

09B

15B

01

Pair

Age

at S

laug

hter

CBV

Figure 9.2. Pair CBVs (± SE) for the breeding objective, age at slaughter (days). These CBVs were produced from the ASReml multivariate analysis described in Chapter 5.

9.2.1.3 Juvenile survival The number of juveniles that reach slaughter size from each clutch directly influences farm income. This trait was investigated in Chapter 6 using a Cox’s proportional hazard analysis. The heritability estimate for log juvenile crocodile survival time was 0.15 (SE 0.04). The log hazard pair estimates are shown in Figure 9.3a. However, since the average age at slaughter is three years (Chapter 5), CBVs were calculated based on a juvenile surviving to 1095 days (or three years). These CBVs are presenting in Figure 9.3b. UB10 had the lowest log hazard estimate of -0.57 (antilog estimate (e-0.57) = 0.57), whilst B01 had the highest estimate of 0.74 (antilog estimate = 2.09). This means that a juvenile from a clutch produced by UB10 has a higher chance of surviving to slaughter than a juvenile produced by the pair B01. More specifically, if we denote S0(t) as the baseline survival function, that is the probability that an individual survives to age t, averaged across the population, then the survival function for offspring of UB10 will be [S0(t)]0.57 (increased survival) whereas those from B01 will have a survival function of [S0(t)]2.09 (reduced survival). So in general, the survival function for offspring of a particular pair will be [S0(t)]R, where R is the hazard ratio for a particular pair, being the antilog of the BLUP estimate on the log hazard scale. The baseline survival function, S0(t), is routinely available in survival analysis output, and has been shown in Chapter 6 (Figure 6.1). Since the hazard of mortality changes with time, it was decided that the most appropriate time to approximate breeding values was at day 1095 (or three years) since this was the average age at slaughter presented in Chapter 5. Juvenile survival CBVs are expressed as a percentage difference in survival to 1095 days, relative to the population average, and have been calculated as

{ }i

i

RR 10

i 00

[S (1095)]CBV 1 100 [S (1095)] 1 100S (1095)

−⎧ ⎫= − × = − ×⎨ ⎬

⎩ ⎭,

with approximate standard errors { } 100SE(BLUP)(1095)][S(1095)lnSR)SE(CBV iR

00ii ××−= where Ri and SE(BLUPi) are the hazard ratios (exponentiated hazard BLUP estimates) and standard error of the BLUP estimates, respectively.

76

From the CBVs (Figure 9.3b), offspring from pair UB10 have a 20.3% reduced risk of mortality compared to the herd average, whilst offspring from B01 have a 23.5% greater risk of dying before cull age. This difference between CBV estimates for UB10 and B01 were considered quite large, even with the adjustments for the large year effects being taken into account (Chapter 6). Therefore, various aspects of the raw data were investigated and are presented in Table 9.1. Without adjusting for year or clutch effects, the raw data in Table 9.1 confirm that some pairs have higher juvenile mortality rates than others. For UB10 and B01, which differ by the greatest amount in CBVs, there are similar differences also apparent in the raw data. From 164 offspring produced by pair B01 over seven clutches, 54.3 % went on to be recorded as deaths, with only 12.8% reaching slaughter size. In contrast, of the 205 offspring produced by UB10 over six clutches, only 9.3% died and 60.5% reached slaughter size during the period of observation. Therefore, the difference in CBV estimates between UB10 and B01 appear plausible. Given that this breeding objective was given the highest priority ranking in the industry survey discussed in Chapter 3, receiving a mean score of 4.9 from a possible five, these results imply that substantial genetic improvement can be made to meet this recognised industry need.

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

log

Haza

rd fo

r Juv

enile

Sur

viva

l

-35-30-25-20-15-10-505

101520253035

UB

10B

14B

07B

16U

B08

B02

UB

06-1U

B06-2

UB

07B

20B

05U

B16

UB

05U

B09

UB

03B

09B

12U

B01

B19

UB

04-2B

04B

15B

03B

10B

06U

B04-1

B13

B11

B01

Pair

Juve

nile

Sur

viva

l 109

5 da

y CB

V

Figure 9.3. A) Log hazard pair estimates (± SE) of juvenile survival produced using the Cox’s proportional hazards model described in Chapter 6. B) Pair CBVs (± SE) for juvenile survival at 1095 days (or three years). These log CBV estimates were produced using a Cox’s proportional hazards model described in Chapter 6.

A

B

77

Table 9.1. Raw data for the fate of juveniles produced from the pairs at Janamba Croc Farm. n is the number of clutches the pair has produced. ‘Total no. to farm’ is the sum of all offspring produced by the pair over n clutches. ‘No. of recorded mortalities’ and ‘No. of Juveniles slaughtered’ are the number of offspring from that pair that have been recorded as either died or slaughtered, respectively. The difference between the ‘Total no. to farm’ and the sum of ‘No. of recorded mortalities’ and ‘No. of Juveniles slaughtered’ are the animals still growing in the production system. These data have not been adjusted for year or clutch effects.

Pair n

Total No. to

Farm

No. of recorded

mortalities

No. of Juveniles

Slaughtered

% Juvenile

Mortality

% Juveniles

Slaughtered

B01 7 164 89 21 54.27 12.80

B02 7 194 28 102 14.43 52.58

B03 8 208 75 75 36.06 36.06

B04 6 167 41 83 24.55 49.70

B05 6 148 26 42 17.57 28.38

B06 1 30 21 3 70.00 10.00

B07 8 219 33 141 15.07 64.38

B09 3 115 21 15 18.26 13.04

B10 6 181 47 89 25.97 49.17

B11 7 171 81 62 47.37 36.26

B12 2 53 4 2 7.55 3.77

B13 5 118 49 27 41.53 22.88

B14 6 188 16 94 8.51 50.00

B15 9 287 93 74 32.40 25.78

B16 9 299 37 165 12.37 55.18

B19 7 177 50 54 28.25 30.51

B20 3 70 5 22 7.14 31.43

UB01 5 129 36 69 27.91 53.49

UB03 5 258 48 47 18.60 18.22

UB04-1 8 313 125 60 39.94 19.17

UB04-2 9 317 73 112 23.03 35.33

UB05 8 333 75 127 22.52 38.14

UB06-1 6 218 36 82 16.51 37.61

UB06-2 3 111 9 61 8.11 54.95

UB07 6 265 57 167 21.51 63.02

UB08 5 203 30 50 14.78 24.63

UB09 6 151 36 78 23.84 51.66

UB10 6 205 19 124 9.27 60.49

UB16 6 239 42 103 17.57 43.10

78

9.2.1.4 Number of scale rows The price received for an individual crocodile skin is currently determined by i) its width, and ii) its grade (determined by the number and severity of blemishes). Additional skin-quality traits have been identified by Manolis et al. (2000) that could, in the future, become important when marketing skins. Possible quality traits include: number of scale rows, regularity of scale pattern, and skin thickness. Currently, there is no premium received for any of these traits. However, in the future these traits may become important and were therefore considered worthy of investigation. Only data on number of scale rows were collected, with results of parameter estimates presented in Chapter 7. The heritability estimate using data from Janamba Croc Farm was 0.37. The CBVs (± SE) for number of scale rows are shown in Figure 9.4. The range of CBVs was between 0.92 rows (UB13) and -1.20 rows (B12).

-1.4-1.2

-1-0.8-0.6-0.4-0.2

00.20.40.60.8

11.2

UB

13U

B07

B06

B10

B05

B02

B19

UB

10B

04U

B16

B20

B13

UB

08U

B06-2

B09

B03

UB

06-1B

15U

B04-2

B07

B11

B01

UB

03B

14U

B04-1

UB

09U

B05

B16

UB

01B

12

Pair

Sca

le R

ow C

BV

Figure 9.4. Pair CBVs (± SE) for number of scale rows. These CBVs were produced from the ASReml univariate analysis described in Chapter 7 using only the Janamba scale row data.

9.2.2 Relative economic values Appendix II presents the full results from economic research performed by fourth year honours student, Ms Emily Gray, and these results are credited to her. The following economic values for unit improvements in the breeding objectives have been estimated: number of hatchlings produced per pair per year, age at slaughter and juvenile survival. An economic value could not be obtained for number of scale rows since there is currently no premium price received for a greater number of scale rows on a belly skin. An economic value for another breeding objective, food conversion efficiency, was also estimated. No phenotypic data were available for this breeding objective, so neither genetic parameter estimates or breeding value estimation has been completed for this trait. The relative economic values for each breeding objective estimated are presented in Table 9.2.

79

Table 9.2. Relative economic values for crocodile breeding objectives. Abb. is the abbreviated term for the respective breeding objective.

Breeding objective Abb. Increment

Economic value

(AU$)

Number of hatchlings per year

NoHatch Increase by one hatchling per year

41.95

Age at slaughter CullAge Increase time to reach slaughter size by one week

-25.68

Juvenile survival Surv Increase survival by 1% 52.37 Food Conversion Efficiency

FCE Increase food consumed by 1g/week

-4.67

9.2.3 The crocodile economic selection index ($CESI) In Chapter 1, an economic selection index was defined in equation 1.1 as

H = v1BV1 + v2BV2 +...+ vmBVm where H = aggregate breeding value of the animal for profitability, vi = the economic value for the ith breeding objective, BVi = the true breeding value for the ith breeding objective, and m = the total number of breeding objectives in the selection index. Remember that candidates are selected after ranking based on an estimate of their aggregate breeding value, H. The estimated crocodile breeding values (CBVs) for each breeding objective have been presented above in sections 9.2.1.1 to 9.2.1.4. Modifying equation 1.1 by substituting in the crocodile breeding objectives, the crocodile economic selection index ($CESI) appears as

$CESI = vNoHatchCBVNoHatch + vCullAgeCBVCullAge + vSurvCBVSurv + vSRCBVSR and after substituting the economic values from Table 9.2, the $CESI becomes

$CESI = 41.95(CBVNoHatch) - 25.68(CBVCullAge) + 52.37(CBVSurv) + 0(CBVSR) A $CESI value for each breeding pair was estimated and is shown graphically in Figure 9.5. Each pair is expressed as a dollar ($) deviation from the pair average. Pair B16 has the largest $CESI value (+$4,748), whilst B01 has the lowest value (-$5,257). Note that because covariances between the component CBVs could not be determined, it was not possible to determine standard errors for these $CESIs.

80

-$6,000

-$5,000

-$4,000

-$3,000

-$2,000

-$1,000

$0

$1,000

$2,000

$3,000

$4,000

$5,000

B16

B07

UB

08U

B07

B04

B12

UB

16U

B06

-2U

B10

B20

UB

01U

B05

UB

09U

B06

-1B

03B

02B

14U

B13

UB

04-2

UB

03U

B04

-1B

05B

19B

06B

10B

09B

11B

13B

15B

01

Pair

$CE

SI

Figure 9.5. Crocodile economic selection index ($CESI) values for each breeding pair at Janamba Croc Farm used in this study. The CESI value is expressed as a dollar ($) deviation from the herd average.

9.3 Case study: CrocPLAN implementation on Janamba Croc Farm So far, this Chapter has presented CBVs for current breeding pairs by evaluating their progeny’s performance (known as progeny testing). However, how should future breeding animals be selected? How should the selection program be implemented? How should generation interval be managed? These questions are addressed in Section 9.3.2. First, there are several industry-specific issues that must be considered before recommending implementation of a genetic improvement program for the Australian crocodile industry. These are outlined in Section 9.3.1 below.

9.3.1 Crocodile-specific industry issues to consider When implementing a genetic improvement program for crocodiles, the following considerations realities must be made: 1. A skewed sex ratio will affect number of female replacements available- The standard incubation temperature for developing embryos in the Australian crocodile industry is 32oC. The main reason for incubating at 32oC is to maximise embryo survival both pre- and post-hatching (Richardson et al., 2002). Lang and Andrews (1994) reported that 86% of embryos develop as males at 32oC. The disadvantage in terms of genetic improvement is that there will be few females among which to select replacements. Chapter 5 presented data from 2151 slaughter crocodiles collected over nine years. Only 8% or 172 were reported as females, equivalent to 19 females reaching surviving to slaughter size each year. 2. Female casualties- Male crocodiles not uncommonly kill their pen-mates. To ensure that there are enough replacement females, more breeding females may have to be selected, as a reserve, to cover possible high female mortality. 3. Age at sexual maturity- There is little literature on the age at which a captive-raised crocodile reaches sexual maturity. Elsey et al. (1993) reported that sexual maturity occurred at eight years of age for saltwater crocodiles. One pair at Janamba Croc Farm (B20), as mentioned previously in Chapter 4 and

81

section 9.2.1.1, first nested when the female was six and the male ten years of age. Whether the male was sexually mature before ten years of age but did not reproduce because the female was not yet ready, is unclear. For the purposes of estimating generation interval in this chapter, sexual maturity will be presumed to be at eight years of age for both males and females. 4. Juvenile quality- Once a female becomes sexually mature, the first few clutches she produces will be of lower quality than from older and more mature females, having lower initial clutch size, lower hatch rates, lower post-hatching growth and lower survival. To use pair B20 as an example (see Chapter 4), the first clutch this pair had in 1998 (female six years old; male ten) consisted of 35 eggs. Twenty-five of these were infertile and the other ten died pre-hatching. For the succeeding three years (2000 to 2002) the clutch size increased to 50, 52 and 46, whilst hatch rates increased to 28%, 56%, and 59%, respectively. Similar parity effects have also been shown to occur in gilts (Brüssow et al., 2002). 5. Identifying individuals- Janamba Croc Farm identifies all captive-bred hatchlings by a clutch-specific scute cut as described in Chapter 2 section 2.1.3. Hatchlings from the 2000 and 2001 cohorts were also marked for individual identification so that genetic and phenotypic correlations could be estimated. This allowed hatchlings to be tracked throughout the production system and, therefore, measurements at the various stages to be collated for an individual. Individual identification is important for eventual selection of individuals and will be required if a complex genetic improvement program is going to be successful.

9.3.2 Response to selection Breeding programs aim to increase the average performance of the herd. The success of a breeding program is assessed using the annual rate of genetic change, or response to selection, where the greater the rate of change, the more successful the breeding program. Therefore, when designing a genetic improvement program, obtaining the maximal rate of genetic change is the primary goal. The rate of genetic change is influenced by four main factors:

1) accuracy of selection, 2) selection intensity, 3) genetic variation, and 4) the generation interval.

as indicated by the formula

L

iσr

t∆ BVBVBV,BV

∧

= [9.1]

where t∆BV is the rate of genetic change per unit time (t), ∧

BVBV,r is the accuracy of selection, i is the

intensity of selection, BVσ is the standard deviation of breeding values in the population, and L is the generation interval (Bourdon, 2000).

9.3.3 Juvenile selection The data in this study were analysed using “Pair” as the random effect. This type of model is analogous to the traditional “Sire” model used in other livestock industries (Mrode, 2000), for example, pigs and dairy cattle. The “animal” model has since become the more widely used model mainly due to computational advances. The advantage of the animal model is that it incorporates a relationship matrix between all animals, allowing a breeding value to be predicted for every animal in the pedigree using all available data, even with missing data (Mrode, 2000). This allows the animals to be ranked by their breeding value, even if the individual is currently a juvenile and has no progeny of its own, using equation 9.1. In the future, with an increase of data collection across subsequent generations, an animal model could be employed for estimating CBVs for every crocodile (juvenile or adult) candidate for selection. Replacement breeder crocodiles will be selected when they reach slaughter size (average age 2.8 years; Chapter 5), after adjusting for the relevant fixed effects. When data become available for an “animal” model, records from the individual’s and relative’s records (full-sibs (both same and different clutches),

82

half-sibs, parents, grandparents) will be combined to estimate a CBV for each breeding objective for every candidate juvenile. These will be incorporated into the $CESI index using the relative economic weights (Appendix II). This will be used to rank the aggregate estimated breeding values of juveniles in dollar-terms and the desired number of replacements can be selected. An initial model of implementation is shown in Figure 9.6. Of course, inclusion of food conversion efficiency and skin-grade breeding objectives can be made when genetic parameter estimates become available. Also, skin quality breeding objectives could also be included, should these traits become economically important in the future.

Grandparents, G1 - Current breeding parents (wild-harvested, age unknown) Parents, G2 - Candidates are selected on the basis of their own, and their relatives’, growth, survival, and reproductive performance using a $CESI value. Individuals, G3 - Adequate information is now available to use an animal model. Candidates can now be selected using a $CESI value.

Figure 9.6. Schematic representation of a proposed selection program for a crocodile production system (modified from Hicks et al., 1998). There are three generations shown: G1 is the current breeding population that are wild-harvested and their ages unknown (indicated by the ?), G2 is the current juvenile population (offspring of G1) available as candidate breeding animals, whilst G3 will be the offspring of G2. Juveniles will be available for selection (G2 and G3) at an average age of 2.8 years. Selection for G2 will occur at time S1 based on their own and relatives’ growth, survival and reproductive performance using $CESI. Assuming that sexual maturity occurs at eight years of age, but given a few years for reproductive performance to stabilise, selection of G3 occurs at S2 using a $CESI index. Selection of the subsequent generations also occurs at 2.8 years.

9.3.4 Across-herd selection Figure 9.5 presents $CESI values for 30 breeding pairs from one commercial breeding farm, Janamba Croc Farm. The small effective population size (Ne) of this genetic resource creates limitations for a long-term genetic improvement program. Potential issues that could arise when attempting to implement a crocodile genetic improvement program are maximising the effective herd size and minimising inbreeding. The current effective population size at Janamba Croc Farm is 30 for known-pair matings. Therefore, the rate of inbreeding ( )

e2N1F =∆ is 0.017, and after 10 generations of selective breeding, the

inbreeding coefficient ( )10tF = will be 0.15 (Falconer and Mackay, 1996). The territorial nature of crocodiles, which can often result in mortality particularly of the females, greatly limits the stocking densities that can be used and therefore, the genetic resource available for selection. For optimal success of a genetic improvement program in the Australian crocodile industry, a long-term goal should be industry-wide adoption of a genetic evaluation program allowing the trade of genetically-superior stock between farms to increase the gene pool available for selection. Other livestock industries have designed their own nation-wide, and to a lesser extent international, breeding programs, allowing the trade of genetically superior stock between farms and thus increasing the gene pool for selection. Some industry examples are:

• Australian dairy industry- Australian Dairy Herd Improvement Scheme (ADHIS; http://www.adhis.com.au),

0 1 2 3 4 5 6 7 8 9 10

?-1 ?

S1

juvenile

Predicted age at sexual maturity

S2

0 1 2 3 4 5 6 7 8 9 10

juvenile

Average age at slaughter

S3

sub-adult

sub-adult

Time since start of breeding program

83

• Australian pig industry- PigBLUP (http://agbu.une.edu.au/PIGBLUP/index.html) • Australian lamb industry- LambPLAN (http://www.mla.com.au/lambplan), and • Australian beef industry, with clients worldwide- BREEDPLAN (http://breedplan.une.edu.au).

One advantage that these livestock industries have is the transfer of genetic stock using artificial reproductive technologies such as Artificial Insemination (AI) and Embryo Transfer (ET). These technologies are not available, nor foreseeable, for the crocodile industry. Therefore, trade of genetic superior stock will have to be as live crocodiles. With a functional crocodile economic selection index, $CESI-values can be obtained at any time, allowing trade in genetically superior (fertile) eggs, hatchlings, juveniles, and adults. Initially, genetic selection will be on a within-farm (or within-herd) operation for the Australian crocodile industry. Care will therefore be required when selecting replacement juveniles and mating pairs to maintain genetic diversity within the limited breeding population. As the industry begins to adopt CrocPLAN, more choices of mates will become available via trading of genetically superior stock.

9.3.5 Applying CrocPLAN on a leading crocodile breeding farm There is no clear defining method for creating and implementing a genetic improvement program on any farm in any industry. However, there are some clear guidelines that need to be considered. These are outlined in the context of a farm like Janamba Croc Farm.

9.3.5.1 Recommendations for pen allocation As discussed in section 9.3.5, the farm breeding system should be optimised to maintain maximal genetic diversity. Figure 2.2 shows a schematic plan of Janamba Croc Farm. The pens potentially available for inclusion in a selection program are the B- and UB-pens. Eventually it is hoped that the lagoon animals and their progeny will also be incorporated into the selection program. Currently no retrospective parentage testing of these lagoon-bred offspring occurs, although the possibility now exists with the microsatellite-based parentage testing kit presented in Chapter 8. At this stage, only parents and offspring from the unitised, known-breeding pairs will be considered in the selection program. There are 20 B-pens and 17 UB-pens. Each of the 20 B-pens can house only one male and one female crocodile (i.e. one pair), with a total of 40 crocodiles (20 males; 20 females). These pens are quite small and female deaths occur frequently (~1 per year) due to territorial behaviour. Of the 17 UB-pens, seven (UB01 to UB07) are much larger than the other ten (UB08 to UB17). In the larger UB-pens, it is recommended that a sex ratio of one male and three females be used, whilst in the smaller UB-pens, it is suggested that a sex ratio of one male to two females be used in each pen. The total number of animals that could then be housed in the UB-pens is 58 (17 males; 41 females). A total of sixty-one pairs from the B- (20 pairs) and UB-pens (41 pairs) would then be available for progeny testing. This would result in an effective population size of 92, with a rate of inbreeding (∆F) of 0.005, and an inbreeding coefficient of 0.053 after ten generations. These are substantially lower than the rate of inbreeding (∆F = 0.017) and inbreeding coefficient (F10 = 0.15) given for the 30 breeding pairs (section 9.3.5). The farm layout illustration in Figure 2.2 shows nine spare pens and ten display pens. It is proposed that these pens be used as “maturing pens”, or the pens where the replacement juveniles are kept until they reach sexual maturity. One to two juvenile animals could be housed in each of these pens, depending on their size and the size of the pen. Additional holding pens may also be required depending on the size and design of the pens, and the proposed stocking densities.

9.3.5.2 Choosing a replacement rate for breeding animals Assuming that all B- and UB-pens have been filled to capacity as per the recommendation in section 9.3.6.1, potentially 61 known-breeding pairs could be reproducing each year. Results from Chapter 4 showed that the nesting frequency of pairs was 83% (Table 4.2). Therefore, 51 clutches of a possible 61 will potentially be produced each year. The average number of hatchlings produced from each clutch was

84

shown to be 31.25 hatchlings (Table 4.2), resulting in approximately 1594 hatchlings entering the production system each year from these pairs alone. Assuming that the recommendation of Webb (1989) can be achieved, survival should be 95% in the first year, 1514 hatchlings will be available for inventory measurements. With Webb (1989) recommending survival rates of 95% for the remainder of the grow-out period, 1439 juveniles will be alive at slaughter size and available for selection. Presuming the percentage of females produced and surviving to slaughter size is consistent with that reported in Chapter 5 (8%), the number of female juveniles available for selection each year will be 115 whilst the remaining 1324 juveniles will be males. Let us assume arbitrarily that the oldest six pairs will be replaced each year, which is equivalent to 10% of the 61 breeding pairs. Therefore, the proportion of males and females selected is 0.5% ( )1324

6 and 5.2% ( )115

6 , respectively. This is equivalent to selection intensities (i) of 2.88 and 2.05 respectively for males and females. With twelve young breeder animals replacing twelve older animals each year, it will take ten years to turn over all breeder animals.

9.3.5.3 Generation interval In the case of crocodiles, juveniles will be selected at slaughter size, based on an economic index ($CESI). This will occur at an average of 2.8 years as shown in Table 5.2, after adjusting for the various effects that influence the age at slaughter (year, sex, pair, and clutch). It will then take a further 5.2 years for each animal to reach the assumed age of sexual maturity at eight years (Figure 9.6). Six months before the individual reaches (estimated) sexual maturity, it will be allocated into a pen with a prospective mate, for example directly after the last breeding season to replace the oldest existing pair. Generation interval is the time required to replace one generation with the next, or in a closed-population, the average age of parents when their offspring are born. This is difficult to determine at present because of the unknown age structure of the breeding population is a results of wild-harvesting the adults. However, if the recommendations of section 9.3.6.2 are adopted, it will take ten years to turn over all breeder animals, plus an additional 5.2 years until they reach sexual maturity. If all animals begin breeding at eight years of age and are replaced after ten years (when they are 18 years old), the average generation interval for both males and females is 13 years. The long generation interval is an intrinsic limiting factor for genetic improvement in the crocodile industry. A similar problem also exists in the forestry industry (Wei and Lindgren, 2001) but has not deterred breeding program implementation. In the above derivation of generation interval, it has been assumed that the generation interval is the same for both males and females. This may not be true. There is a scarcity of published information on the age of sexual maturity and the assumption of eight years for both males and females may be incorrect. This will need to be determined in the future. An additional breeding objective could be included into CrocPLAN to select for earlier breeding replacement breeder animals.

9.3.5.4 Selection response Further extending the model for response to selection (equation 9.1) to include differences between male and female selection, equation 9.1 becomes

( )fm

BVfBV,BVmBV,BVBVLL

σirir

t∆ ffmm

+

⎟⎠⎞

⎜⎝⎛ +

=∧∧

[9.2]

where mm BV,BV

r ∧ and ff BV,BV

r ∧ are the accuracy of selection for males and females, respectively; im and if

are the intensity of selection for males and females, respectively; BVσ is the standard deviation breeding values in the population; and Lm and Lf is the generation interval for males and females, respectively.

85

For the crocodile scenario presented in this study, the limited data structure prohibited the estimation of genetic and phenotypic correlations between breeding objectives. Therefore, assumptions of no genetic and phenotypic correlations (that is, rg =0 and rp = 0) between breeding objectives were used. Equation 9.2 was then modified to incorporate the available parameters as a weighted sum (economic values) of the breeding values and their accuracies, separately for males and females, as follows (James, pers. comm.)

L

)rσ(v)rσ(v)rσ(vi∆

23

2CBV

23

22

2CBV

22

21

2CBV

21

tBV

321++

= [9.3]

where i is the intensity of selection (that is 2.88 for males and 2.05 for females), vi is the relative economic values for the ith breeding objective,

2CBVi

σ is the variance of CBVs for the ith breeding objective, ri is the accuracy of selection of the ith breeding objective, and L is the generation interval (in this case, assumed 13 years for both males and females).

Accuracy for the ith breeding objective, ri, was determined by

2g

ii

iσ

SEP1r −= [9.4]

where SEPi is the average standard error of prediction of the CBVs for the ith breeding objective, and 2gi

σ is the genetic variance associated with the breeding objective. Using the 2

CBViσ and ri values for each breeding objective presented in Table 9.3, the response to

selection was predicted to be a $323.80 increase in profit per annum, assuming no genetic and phenotypic correlation amongst breeding objectives.

Table 9.3. Various parameters required to predict the response to selection.

Breeding objective 2CBVi

σ SEP 2gi

σ r

Number of hatchlings per year

36.89 3.15 45.75 0.78

Juvenile survival 117.15 10.09 119.90 0.15 Age at slaughter 5793.40 29.55 6542.56 0.87

86

10. Conclusions and recommendations 10.1 CrocPLAN Chapter 9 presents the crocodile economic selection index, or $CESI, which is the combination of economic values and the crocodile breeding values (CBVs) resulting from the analysis of each breeding objective. After ranking the breeding pairs from Janamba Croc Farm, there was a difference between the highest ranking pair (B16; +$4,748) and the lowest ranking pair (B01; -$5,257) of $10,006. The response to selection was predicted to be a $324 increase in profit per annum, assuming no genetic and phenotypic correlation amongst breeding objectives. This result reinforces the potential for implementing a genetic improvement program (CrocPLAN) on Australian crocodile farms. The aim of a genetic improvement program is to improve the total economic value of the herd, and consequently maximise farm profit. Identifying pairs, or preferably individuals, that are inhibiting profit maximisation, such as B01, and replacing them with genetically superior animals will improve farm profitability. The survey presented in chapter 3 stimulated some interesting comments from a few industry members. A few producers expressed concerns that improving efficiency through genetic selection is premature and that research should remain focused on husbandry-related issues (for example nutrition, housing). It is hoped that this study has provided strong evidence for the benefits of simultaneous improvement of genetics, along with improvements in husbandry. For those respondents who reported they had already begun selecting for various traits, the results presented in this study provide a formal basis for structuring selection decisions. Chapter 9 presented an example of how CrocPLAN can be implemented, and used Janamba Croc Farm as an example. Selection of replacement breeder animals will occur at slaughter size (approximately 2.8 years) after adjusting for the relevant fixed effects, based on a $CESI-value. Replacement of pairs will be equivalent to 10% of adult breeding animals per year, or the oldest six pairs. The proportion of male and female juvenile replacements is 0.5% and 5.2%, respectively, which is equivalent to selection intensities (i) of 2.88 and 2.05. Generation interval was estimated to be 13 years for both males and females using the assumption that sexual maturity occurs at eight years of age for both sexes, a replacement rate of 10% and that the animals will be maintained in the breeding system for ten years only after sexual maturity. Some issues that need to be carefully considered before implementing a genetic improvement program are herd size, inbreeding, across-herd selection and the introduction of new genetic stock. These issues are all related to the number of breeding animals. The $CESI values presented in this study are from 30 breeding pairs on one farming establishment, which is a limited resource in terms of a long-term genetic improvement program. For the most effective genetic improvement program to be obtained, industry-adoption of the program is recommended. This will allow the additional trade of genetically superior stock across herds, analogous to other industries such as in the Australian dairy industry, and maintain a large genetic pool in which selection decisions can be made. Utilising the parentage testing kit (Chapter 8) will also broaden the scope of a breeding program, allowing progeny to be tested from adults maintained in large breeding lagoons for selection as future breeding animals. Furthermore, the introduction of wild-harvested eggs and, and occasionally, adults into the selection program will add to the genetic diversity of captive breeding animals. Under strict sustainable management programs, wild-harvesting of eggs and adults is allowed in the Northern Territory and, to a lesser extent, Western Australia (Manolis et al., 2000). These eggs are then artificially incubated and the hatchlings placed onto the farms for grow-out. All hatchlings from a clutch have most-likely the one

87

mother. Again, using the parentage-testing kit (Chapter 8), the number of males that fathered the clutch1 can be identified and the relationship between hatchlings can be determined. These animals can then also be considered for selection when they reach slaughter size along with the captive-bred juveniles. The industry-wide adoption of CrocPLAN will require an integrated performance-recording system. This would require the recording of pedigrees, individual identification (scute-cuts, tags, etc) and, of course, the various breeding objectives and relevant selection criteria. However, before the industry invests a large amount of capital into a genetic improvement program, a cost-benefit analysis should be commissioned that takes account of the various production and breeding systems represented in the industry. However, with the large range in $CESI values produced in Chapter 9, it is expected that CrocPLAN will be highly cost effective just as similar programs are in other industries (beef cattle: Farquharson et al., 2002, 2003; dairy cattle: Norman et al., 2003; pigs: McPhee, 1977).

10.2 Areas of further research Other than the cost-benefit analysis recommended above to confirm the economic gains in productivity, there are some other areas that warrant further investigation. The data provided by Janamba Croc Farm have provided a valuable foundation for estimating the quantitative parameters presented in this study. However, there were many gaps in the data where complete analyses could not be conducted. One particular limitation was the limited pedigree structure. Follow-up analyses when a more complex pedigree structure becomes available will be greatly beneficial. The preferred model for quantitative analysis is the “animal” model, which takes into account all the relationships between individuals within the data-set and ultimately produces a $CESI value for each animal. Other areas of further research are outlined below individually.

10.2.1 Reproductive traits One of the limiting areas of crocodilian research is captive breeding, particularly in regards to age of sexual maturity and parity effects on clutch and offspring characteristics. The limitations of the reproductive data available for use in the present study have been mentioned numerous times throughout this study. However, the reproductive characteristics of crocodiles are very important traits that directly influence the number of juveniles that are placed into the production system. Even though the data provided for this study have been the most comprehensive available, the unknown age and relationship structure of the population made the estimation of heritability impossible for reproductive traits. Age and parity effect of males and females in the production system could only be examined for one pair (B20). Repeatabilities were estimated for the breeding objective, number of hatchlings per female per year, and other possible selection criteria. Another breeding objective that could be included in CrocPLAN eventually is age of sexual maturity, for the selection for younger maturing animals in an attempt to decrease the generation interval.

1 evidence of multiple paternity of clutches exists within the Crocodylia (Richardson et al., 2002).

88

10.2.2 Food conversion efficiency and growth rate Food contributes a significant cost (42-45%; Treadwell et al., 1991) to crocodile production. No data were available for quantitative genetic analysis in this study, although food conversion efficiency has been investigated in biological and production-based studies (Garnett and Murray, 1986; Manolis et al., 1989; Webb et al., 1991). It would be beneficial to conduct a study to investigate these traits simultaneously. Potentially, there could be a high genetic correlation between these traits as in pigs (Clutter and Brascamp, 1998), allowing the possibility of indirect selection on growth rate which is easy to record.

10.2.3 Skin grade Investigation into this breeding objective was not conducted since appropriate data were not available. However, with the large number of skins not meeting first grade requirements (Manolis et al., 2000; MacNamara et al., 2003), a high priority should be placed on research to evaluate whether there is any genetic basis for skin damage. One possibility might be to conduct a behavioural study since anecdotal evidence suggests that some clutches are more “aggressive” than others, even at the time of hatching, implying a familial and possibly genetic basis for the differences in aggressive behavior and resultant skin damage.

10.2.4 Industry-wide CrocPLAN It would be advantageous for an industry-wide service to be available similar to the service provided for the pig industry (PigBLUP at http://agbu.une.edu.au/PIGBLUP/index.html). Producers provide pedigree, performance and reproductive records on particular animals. These will then combined with other animals from that farm and other farms, and $CESI values for each individual will be calculated. This will allow across-herd selection and trade of genetically superior (fertile) eggs, hatchlings, juveniles, and adults. This in turn will minimise inbreeding rates and maximise the effective population size.

10.3 Conclusion The present study presents a wide range of quantitative genetic analyses for economically-important traits in the production of saltwater crocodiles in Australia. The general aim of the study was to investigate the possibility of implementing a genetic improvement program on Australian crocodile farms. Results have indicated that there is sufficient genetic variability present for a selection program to be successful. In addition, the estimation of economic values has allowed the immediate application of this program, given the availability of performance records. Furthermore, the availability of a parentage testing kit using microsatellite markers provides verification of pedigree and a method for expanding the genetic improvement program beyond unitised breeding pens.

89

Literature cited Allen, W.M., S. Berret, and J.D.J. Harding. 1970. Experimentally induced acute stress syndrome in

Piétrain pigs. Vet. Rec. 87:64-69. Australian Bureau of Statistics. 2002. Foreign Trade: Magnetic Tape Service, Cat. no. 5464.0, Canberra. Ayyat, M.S., I.F.M. Marai, and GH.A El-Sayiad. 1995. Genetic and non-genetic factors affecting milk

production and preweaning litter traits of New Zealand White does under Egyptian conditions. World Rabbit Science. 3:119-124.

Barnett N.L., I.W. Purvis, B. van Hest, and I.R. Franklin. 1999. The accuracy of current dam pedigree recording strategies employed by stud Merino breeders. Proceedings of the Association for the Advancement of Animal Breeding and Genetics. 13:373-376.

Bell, D.D. 2002. Egg production and egg weight standards for table-egg layers. Page 1079 in Commercial chicken meat and egg production. 5th ed. D.D. Bell, and W.D. Weaver, eds. Kluwer Academic Publishers, Massachusetts.

Bourdon, R.M. 2000. Understanding animal breeding. 2nd ed. Prentice-Hall, New Jersey. Brazaitis, P. 1987. The identification of crocodilian skins and products. Pages 373-386 in Wildlife

Management: Crocodiles and Alligators. Webb, G.J.W., S. C. Manolis, and P.J. Whitehead, eds. Surrey Beatty and Sons, Chipping Norton, Sydney.

Bright, G. 1991. Economic weights from profit equations: appraising their accuracy. Anim. Prod. 53:395-398.

Browne, A. (ed.) 2004. Page 426 in Explore Australia. 22nd edition. Explore Australia Publishing Pty. Ltd., Victoria.

Brüssow K.P., J. Ratky, H. Torner, I. Egerszegi, F. Schneider, L. Solti, and A. Tuchscherer. 2002. Follicular and oocyte development in gilts of different age. Acta Vet. Hung. 50:101–110.

Buenviaje, G.N., P.W. Ladds, and Y. Martin. 1998. Pathology of skin diseases in crocodiles. Aust. Vet. J. 76:357-363.

Buenviaje, G.N., P.W. Ladds, L. Melville, and S.C. Manolis. 1994. Disease-husbandry associations in farmed crocodiles in Queensland and the Northern Territory. Aust. Vet. J.. 71:165-173.

Cameron, N.D. and R.E. Crump. 2001. Economic values for the Australian pig industry. Page 146 in Manipulating pig production VIII. P.D. Cranwell, ed. Australian Pig Science Association, Werribee.

Clutter, A.C. and E.W. Brascamp. 1998. Genetics of performance traits. Pages 427-462 in The genetics of the pig. Rothschild, M.F. and A. Ruvinsky, eds. CAB International.

Cox, D. 1972. Regression models and life table (with discussion). J. R. Stat. Soc. 34B:187-220. David, D.N. 1987. Effects of Alligator skin storage techniques on “Red Heat”. Pages 393-95 in Wildlife

Management: Crocodiles and Alligators. Webb, G.J.W., S.C. Manolis, and P.J. Whitehead, eds. Surrey Beatty and Sons, Chipping Norton, Sydney.

Davis, B. and S. Peucker. 2001. Thinking of crocodile farming? Farming issues to consider. DPI note, Brisbane Available: http://www.dpi.gov.au/animals/4935. Accessed February 21, 2004.

Davis, B.M. 2001. Improved nutrition and management of farmed crocodiles- from hatching to harvest. Rural Industries Research and Development Corporation Publication No 01/123, Canberra. Available: http://www.rirdc.gov.au/reports/NAP/01-123.pdf. Accessed Oct. 8, 2003.

Deeming, D.C. and M.J.W. Ferguson. 1990. Methods for the determination of the physical characteristics of eggs of Alligator mississippiensis: A comparison with other crocodilian and avian eggs. Herp. J. 1:458-462.

Dobson, A. J. 1990. An Introduction to Generalized Linear Models. London, Chapman & Hall. DOTRS (Department of Transport and Regional Services). 2001. Crocodile farming. In Investing in

Northern Australia: Key Opportunities – Emerging Industry Sectors, Canberra. Available: http://www.dotars.gov.au/regional/northern_forum. Accessed February 11, 2004)

Ducrocq, V. and G. Casella. 1996. A Bayesian analysis of mixed survival model. Genet. Sel. Evol. 28:505-529.

Ducrocq, V. and J. Sölkner. 1994. ''The Survival Kit'', a FORTRAN package for the analysis of survival data. Proc. 5th World Cong. Genet. Appl. Livest. Prod., Ontario, Canada. XXII:51-52.

90

Ducrocq, V. and J. Sölkner. 1998. ''The Survival Kit -- V3.0'', a package for large analyses of survival data. Proc. 6th World Cong. Genet. Appl. Livest. Prod., Armidale, Australia. XXVII:447-448.

Ducrocq, V., B. Besbes, and M. Protais. 2000. Genetic improvement of laying hens viability using survival analysis. Genet. Sel. Evol. 32:23-40.

Elsey, R.M., T. Joanen. and L. McNease. 1993. Captive breeding of alligators and other crocodilians. In Proc. 2nd IUCN Crocodile Specialist Group Meeting, Darwin, Australia.

Falconer, D.S. and T.F.C. Mackay. 1996. Introduction to Quantitative Genetics. 4th ed. Longmand Chesire, London.

Farquharson, B., G.R. Griffith, S.A. Barwick, and R.G. Banks. 2002. Estimating returns from investment into beef cattle genetic technologies in Australia. Pages 188-190 in Update of research in progress at the Tamworth Agricultural Institute. B. Martin, ed. NSW Agriculture, Tamworth, Australia.

Farquharson, B., G. Griffith, S. Barwick, R. Banks, and B. Holmes. 2003. Estimating the returns from past investment into beef cattle genetic technologies in Australia. Economic Research Report No. 15. NSW Agriculture, Wagga Wagga, Australia.

Ferguson, M.W.J. 1985. Reproductive biology and embryology of the crocodilians. Pages 330-491 in Biology of the Reptilia. Gans, C., F.S. Billet, and P.F.A. Maderson, eds. John Wiley and Sons, New York.

Ferguson, M.W.J. and T. Joanen. 1983. Temperature-dependent sex determination in Alligator mississippiensis. J. Zool. (Lond.) 200:143-177.

FitzSimmons N.N., J.C. Buchan, P.V. Lam, G. Polet, T.T. Hung, N.Q. Thang, and J. Gratten. 2002. Identification of purebred Crocodylus siamensis for reintroduction in Vietnam. J. Exp. Zool. 294:373-381.

FitzSimmons N.N., S. Tanksley, M.R. Forstner, E.E. Louis, R. Daglish, J. Gratten, and S. Davis. 2001. Microsatellite markers for Crocodylus: new genetic tools for population genetics, mating system studies and forensics. Pages 51-57 in Crocodilian Biology and Evolution. G. Grigg, F. Seebacher, and C.E. Franklin, eds. Surrey Beatty and Sons, Chipping Norton, Australia.

Ford, N.B., and R.A. Seigel. 1989. Relationships among body size, clutch size, and egg size in three species of oviparous snakes. Herpetologica. 45:75-83.

Garnett, S.T., and R.M. Murray. 1986. Parameters affecting the growth of the estuarine crocodile, Crocodylus porosus, in captivity. Aust. J. Zool. 34:211-223.

Geldermann, H. 1975. Investigations on inheritance of quantitative characters in animals by gene markers. I. Methods. Theor. Appl. Genet. 46:319-330.

Georges, M. 2001. Recent progress in livestock genomics and potential impact on breeding programs. Theriogenology 56:15-21.

Gilmour, A.R., B.J. Gogel, B.R. Cullis, S.J. Welham and R. Thompson. 2002. ASReml User Guide Release 1.0, VSN International Ltd, UK

Gilmour, A.R., R.D. Anderson, and A.L. Rae. 1985. The analysis of binomial data by a generalised linear mixed model. Biometrika 72:593-599.

Glenn, T.C., H.C. Dessauer, and M.J. Braun. 1998. Characterization of microsatellite loci in American Alligators. Copeia 3:591-601.

Glenn, T.C., W. Stephan, H.C. Dessauer, and M.J. Braun. 1996. Allelic diversity in alligator microsatellite loci is negatively correlated with GC content of flanking sequences and evolutionary conservation of PCR amplifiability. Mol. Biol. Evol. 13:1151-1154.

Goddard, M.E. 1983. Selection indices for non-linear profit functions. Theor. Appl. Genet. 64:339-344. Gregory, K.E., L.V. Cundiff, and R.M. Koch. 1995. Genetic and phenotypic (co)variances for growth and

carcass traits of purebred and composite populations of beef cattle. J. Anim. Sci. 73:1920-1926.

Guo-Sheng, S., L.E. Liljedahl, and G.A.E. Gall. 2002. Genetic correlations between body weight at different ages and with reproductive traits in rainbow trout. Aquaculture 213:85-94.

Hager, M.A., S. Wilson, T.H. Pollak, and P.M. Rooney 2002. Response Rates for Mail Surveys of Nonprofit Organizations: A Review and Empirical Test. Proceedings of the Annual Meeting of the Academy of Management, Denver, Colorado.

Harris, D.L. and S. Newman. 1994. Breeding for profit: Synergism between genetic improvement and livestock production (A review). J. Anim. Sci. 74:2178-2200.

91

Harville, D.A. 1977. Maximum likelihood approaches to variance component estimation and to related problems. J. Am. Stat. Assoc. 72:320-340.

Hayes, B. and M.E. Goddard. 2001. The distribution of the effects of genes affecting quantitative traits in livestock. Genet. Sel. Evol. 33:209-229.

Hayes, B. and M.E. Goddard. 2003. Evaluation of marker assisted selection in pig enterprises. Livest. Prod. Sci. 81:197-211.

Hayes, B. and M.E. Goddard. 2004. Break-even cost of genotyping genetic mutations affecting economic traits in Australian pig enterprises. Livest. Prod. Sci. In Press, Corrected Proof, Available online 10 March 2004.

Henderson, C.R. 1953. Estimation of variance and covariance components. Biometrics 9:226-252. Henderson, C.R. 1973. Sire evaluation and genetic trends. Pages 10-41 in Proc. of the Animal Breeding

and Genetics Symposium in Honor of Dr. J.L. Lush, Champaign, IL: ASAS and ADSA. Henderson, C.R. and R.L. Quaas. 1976. Multiple traits evaluation using relatives’ records. J. Anim. Sci.

43:1188-1197. Hicks, C., W.M. Muir, and D.A. Stick. 1998. Selection index updating for maximizing rate of annual

genetic gain in laying hens. Poultry Science 77:1-7. Isberg, S.R., F.W. Nicholas, P.C. Thomson, S.G. Barker, S.C. Manolis, and C. Moran. 2003. Defining

breeding objectives for saltwater crocodile genetic improvement programs. Proceedings of the Association for the Advancement of Animal Breeding and Genetics 15:166-169.

James, J. W. 1982. Economic aspects of developing breeding objectives: general considerations. Pages 107-118 in Future developments in the genetic improvement of animals. Barker, J.S.F., K. Hammond. and A.E. McClintock, eds. Academic Press, Sydney, Australia.

Johnson, Z.B., J.J. Chewning, and R.A. Nugent. 1999. Genetic parameters for production traits and measures of residual feed intake in Large White swine. J. Anim. Sci. 77:1679-1685.

Jones A.G. and W.R. Ardren. 2003. Methods of parentage analysis in natural populations. Mol. Ecol. 12:2511-2523.

Kerr, J.C. and N.D. Cameron. 1995. Reproductive performance of pigs selected for components of efficient lean growth. Animal Science. 60:281-290.

Khatkar, M.S., P.C. Thomson., I. Tammen, and H.W. Raadsma. 2004. Quantitative trait loci mapping in dairy cattle: review and meta-analysis. Genet. Sel. Evol. 36:163-190.

Knol, E.F., J.I. Leenhouwers, and T. van der Lende. 2002. Genetic aspects of piglet survival. Livest. Prod. Sci. 78:47-55.

Ladds, P.W. and L.D. Sims. 1990. Diseases of captive crocodiles in Papua New Guinea. Aust. Vet. J. 67:323-330.

Ladds, P.W., J. Bradley, and R.G. Hirst. 1996. Providencia rettgeri meningitis in hatchling saltwater crocodiles (Crocodylus porosus). Aust. Vet. J. 74:397-8.

Lang, J.W. 1980. Reproductive behaviours of New Guinea and Saltwater crocodiles. SSAR Symp. Reprod. Biol. Conserv. Crocodiles, Milwaukee, Aug. 1980. (Abstr.)

Lang, J.W. 1987. Crocodilian thermal selection. Pages 307-317 in Wildlife Management: Crocodiles and Alligators. Webb, G.J.W., S.C. Manolis, and P.J. Whitehead, eds. Surrey Beatty and Sons, Chipping Norton, Sydney.

Lang, J.W. and H.V. Andrews. 1994. Temperature-dependent sex determination in crocodilians. J. Exp. Zool. 270:28-44.

Lloyd M. and P.J. Morris. 1999. Phlebotomy techniques in Crocodilians. Bulletin of the Association of Reptilian and Amphibian Veterinarians 9:12-14.

Love, G. and D. Langenkamp. 2003. Australian Aquaculture: Industry profiles for related species. ABARE eReport 03.8. Prepared for the Fisheries Resources Research Fund, Canberra, May.

MacNamara, K., P. Nicholas, D. Murphy, E. Riedel, B. Goulding, C. Horsburgh, T. Whiting, and B. Warfield 2003. Markets for skins and leather from the goat, emu, ostrich, crocodile and camel industries. Rural Industries Research and Development Corporation Publication No 02/142, Canberra. Available: http://www.rirdc.gov.au/reports/NAP/02-142.pdf. Accessed Sept. 5, 2003.

Manolis, S.C., G.J.W. Webb, S.G. Barker, and C. Lippai. 1989. Nutrition of crocodiles. Pages 291-312 in Proceedings of the Intensive Tropical Animal Production Seminar, Townsville, Queensland.

92

Manolis, S.C., G. Webb, and K. Richardson. 2000. Improving the quality of Australian crocodile skins. Rural Industries Research and Development Corporation Publication No 00/21, Canberra. Available: http://www.rirdc.gov.au/reports/NAP/00-21.pdf. Accessed Sept. 6, 2003.

Marshall T.C., J. Slate, L. Kruuk, and J.M. Pemberton. 1998. Statistical confidence for likelihood-based paternity inference in natural populations. Mol. Ecol. 7:639-655.

McClure, G. and R. Mayer. 2001. Factors influencing nesting of Crocodylus porosus on a Queensland farm. Page 256 in Crocodilian Biology and Evolution. G. Grigg, F. Seebacher, and C.E. Franklin, eds. Surrey Beatty and Sons, Chipping Norton, Australia.

McPhee, C.P. 1977. Economic returns from reducing backfat thickness in pigs by genetic and other means. Aust. J. Exp. Agric. Anim. Hus. 17:399-402.

Meuwissen, T.H.E., and M.E. Goddard. 1996. The use of marker haplotypes in animal breeding schemes. Genet. Sel. Evol. 28:161-176.

Mrode, R.A. 2000. Linear Models for the Prediction of Animal Breeding Values. CAB International, UK. Nicholas, F.W. 2003. Introduction to veterinary genetics. 2nd ed. Blackwell Publishing, Massachusetts. Norman, H.D., R.L. Powell, J.R. Wright, and C.G. Sattler. 2003. Timeliness and effectiveness of progeny

testing through artificial insemination. J. Dairy Sci. 86:1513-1525. Patterson, H.D. and R. Thompson. 1971. Recovery of inter-block information when block sizes are

unequal. Biometrika 58:545-554. Peucker, S. and R. Mayer. 1995. Runt observation studies. Crocodile Research Bulletin. 1:57-62. Ponzoni, R.W. and D.R. Gifford. 1990. Developing breeding objectives for Australian Cashmere Goats. J.

Anim. Breed. Genet. 107:351-370. Ponzoni, R.W. and S. Newman. 1989. Developing breeding objectives for Australian beef cattle

production. Anim. Prod. 49:35-47. Prayaga, K.C. and S.J. Eady. 2000. Rabbit farming for meat production in Australia: preliminary

estimation of economic values for production traits. Asian-Aust. J. Anim. Sci 13(Supp):357-59. Richardson K.C., G.J.W. Webb, and S.C. Manolis. 2002. Crocodiles: Inside Out. Surrey Beatty and Sons:

Chipping Norton, Australia. Robinson, G.K. 1991. That BLUP is a good thing: The estimation of random effects. Stat. Sci. 6:15-51. Sambrook J.E., E.F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual. Cold Spring

Harbor Laboratory Press: Cold Spring Harbor, NY. Schall, R. 1991. Estimation in generalized linear models with random effects. Biometrika 78:719-727. Schneider S., J. Kueffer, D. Roessli, and L. Excoffier. 2000. Arlequin Ver. 2.000: A software for

population genetic analysis. Genetics and Biometry Laboratory, University of Geneva, Switzerland. Searle, S.R. 1961. Phenotypic, genetic and environmental correlations. Biometrics. 17:474-480. Searle, S.R., G. Cassella, and G.E. McCulloch. 1992. Variance Components. John Wiley and Sons, Inc.,

New York. Serenius, T., M.-L. Sevón-Aimonen, and E.A. Mäntysaari. 2003. Effect of service sire and validity of

repeatability model in litter size and farrowing interval of Finnish Landrace and Large White populations. Livest. Prod. Sci. 81:213-222.

Shotts, E.B. 1981. Bacterial diseases of alligators: An overview. Pages 36-41 in Proc. 1st Annual Alligator Production Conference., University of Florida, Gainsville.

Smith, C. 1983. Effects of changes in economic weights on the efficiency of index selection. J. Anim. Sci. 56:1057-1064.

Southey, B.R., S.L. Rodriguez-Zas, and K.A. Leymaster. 2001. Survival analysis of lamb mortality in a terminal sire composite population. J. Anim. Sci. 79:2298-2306.

Stubbs, A. 1998. Information systems for new animal industries, RIRDC- Publ. No 98/139, Canberra. Available: http://www.rirdc.gov.au/reports/NAP/PTP-1A.doc. Accessed Sept. 6, 2003.

Thorbjarnarson, J.B. 1994. Reproductive ecology of the spectacled caiman (Caiman crocodilus) in the Venezuelan Llanos. Copeia 4:907-919.

Thorbjarnarson, J.B. 1996. Reproductive characteristics of the order Crocodylia. Herpetologica. 52:8-24. Thorbjarnarson, J. 1999. Crocodile Tears and Skins: International trade, economic constraints, and limits

to the sustainable use of crocodilians. Conserv. Biol. 13:465-470. Treadwell, R., L. McElvie, and G.B. Maguire. 1991. Pages 63-70 in Profitability of selected aquaculture

species. ABARE discussion paper 91.11, Canberra.

93

van Arendonk, J. A. M., C. van Rosmeulen, L.L.G. Janss, and E.F. Knol. 1996. Estimation of direct and maternal genetic (co)variances for survival within litters of piglets. Livest. Prod. Sci. 46:163-171.

Van Jaarsveldt, K.R. 1987. Flaying, curing and measuring crocodile skins. Pages 387-92 in Wildlife Management: Crocodiles and Alligators. G.J.W. Webb, P.J. Whitehead, and S.C. Manolis. eds. Surrey Beatty and Sons, Chipping Norton, Australia.

Van Laere, A.-S., M. Nguyen, M. Braunschweig, C. Nezer, C. Collette, L. Moreau, A.L. Archibald, C.S. Haley, N. Buys, M. Tally, G. Andersson, M. Georges and L. Andersson. 2003. A regulatory mutation in IGFs causes a major QTL effect on muscle growth in the pig. Nature. 425:832-836.

Van Vleck, L.D. 2000. Selection Index and Introduction to mixed model methods. CRC Press, Florida. Visscher P.M., J.A. Woolliams, D. Smith, and J.L. Williams. 2002. Estimation of pedigree errors in the

UK dairy population using microsatellite markers and the impact on selection. J. Dairy Sci. 85:2368-2375.

Webb, G. 1989. The crocodile as a production unit. Proceedings of the Intensive Tropical Animal Production Seminar, Townsville, Queensland.

Webb G.J.W., A.M. Beal, S.C. Manolis, and K.E. Dempsey. 1987. The effects of incubation temperature on sex determination and embryonic development rate in Crocodylus johnstoni and C. porosus. Pages 507-531 in Wildlife Management: Crocodiles and Alligators. Webb, G.J.W., P.J. Whitehead and S.C. Manolis, eds. Surrey Beatty and Sons, Chipping Norton, Australia.

Webb G.J.W. and H. Cooper-Preston. 1989. Effects of incubation temperature on crocodiles and the evolution of reptilian oviparity. Amer. Zool. 29:953-971.

Webb, G.J.W., G.J. Hollis, and S.C. Manolis. 1991. Feeding, growth, and food conversion rates of wild juvenile saltwater crocodiles (Crocodylus porosus). J. Herp. 25:462-473.

Webb, G. and C. Manolis. 1989. Crocodiles of Australia. Reed Books Pty Ltd, N.S.W. Webb, G.J.W., S.C. Manolis and H. Cooper-Preston. 1990. Crocodile management and research in the

Northern Territory 1988-1990. Pages 253-273 in Proc. 10th IUCN Crocodile Specialist Group Meeting, Gainesville, Florida.

Webb, G.J.W. and H. Messel. 1978. Morphometric analysis of Crocodylus porosus from the north coast of Arnhem Land, Northern Australia. Aust. J. Zool. 26:1-27.

Webb, G.J.W., H. Messel, J. Crawford, and M.J. Yerbury. 1978. Growth rates of Crocodylus porosus (Reptilia: Crocodilia) from Arnhem Land, Northern Australia. Aust. Wildl. Res. 5:385-399.

Webb, G.J.W., G.C. Sack, R. Buckworth, and S.C. Manolis. 1983. An Examination of Crocodylus porosus nests in two Northern Australian Freshwater Swamps with an analysis of embryo mortality. Aust. Wildl. Res. 10:571-605.

Webb, G.J.W., P.J. Whitehead, and S.C. Manolis. 1987. Crocodile management in the Northern Territory of Australia. Pages 107-124 in Wildlife Management: Crocodiles and Alligators. Webb, G.J.W., P.J. Whitehead and S.C. Manolis, eds. Surrey Beatty and Sons, Chipping Norton, Australia.

Wei, R.P-. and D. Lindgren. 2001. Optimum breeding generation interval considering buildup of relatedness. Can. J. For. Res. 31:722-729.

94

Appendix I.

Nes

ting

B

- - - - - -

2.49

465

0.

5595

77

HD

ays

-19.

9936

-6.5

7577

8.95

03

0.55

494

14.0

813

550.

448

25

5.72

8

-

AvS

VL

-7.1

6694

-3.0

6579

2.59

601

0.19

645

10.5

791

7.

8426

2

-2.1

8732

-

Hat

chR

-0.1

8314

6

8.16

E-02

0.36

7232

1.20

E-0

2

3.71

E-0

2

8.88

E-02

0.28

441

-

NoH

atch

38.3

674

49.4

974

45.7

498

87

.991

4

1.63

582

5.00

1

2.33

394

-

NoV

iabl

e

58.5

709

64.8

158

53

.248

8

42.2

881

0.56

6424

0.44

406

-6.0

3032

-

ClS

ize

59.2

382

28.2

802

28.6

688

20.0

656

2.12

E-02

1.02

315

-14.

5752

-

App

endi

x I:

Tab

le 1

Est

imat

es o

f gen

etic

and

resi

dual

var

ianc

e an

d co

varia

nce

com

pone

nts A

am

ong

the

repr

oduc

tive

traits

from

a m

ultiv

aria

te a

naly

sis.

Trai

t

ClS

ize

NoV

iabl

e

NoH

atch

Hat

chR

AvS

VL

HD

ays

Nes

ting

B

A G

enet

ic (f

irst l

ine)

and

resi

dual

(sec

ond

line)

var

ianc

es o

n di

agon

al (s

how

n in

bol

d), g

enet

ic c

ovar

ianc

es a

bove

the

diag

onal

, an

d re

sidu

al c

ovar

ianc

es b

elow

the

diag

onal

. B G

enet

ic a

nd re

sidu

al v

aria

nce

com

pone

nts o

n th

e un

derly

ing

scal

e of

a b

inom

ial d

istri

butio

n.

95

Appendix I: Table 2 Estimates of genetic and residual variance and covariance components A among the production traits from a multivariate analysis. Also, the (co)variance component estimates from the juvenile survival analysis using a Cox’s Proportional Hazards model.

Trait HSVL InvHL CullAge Survival

HSVL 7.81991

3.84535

14.5994

-0.81836

3.84611

50.6803

-16.2509 -

Pair

Clutch

Residual

InvHL

5.12924

49.5684

22.5846

95.9857

-463.978

-113.194 -

Pair

Clutch

Residual

CullAge

-28.9133 -1341.64

6542.56

1099.70

24907.5

-

Pair

Clutch

Residual

Survival - - -

0.18147

0.53289

Pair

Clutch A Genetic (first line; Pair), common environment (second line; Clutch) and residual (third line; Residual) variances on diagonal (shown in bold), genetic (first line) and common environment (second line) covariances above the diagonal, and residual covariances below the diagonal.

Appendix I: Table 3 Estimates of genetic and residual variance and covariance components A among the skin quality and hatchling measurement traits from a multivariate analysis.

Trait Scale Row HHL HSVL HTL

Scale Row 0.340777

1.48758

3.02E-02 8.87E-02 0.107076 Pair

Residual

HHL

2.51E-03

0.773147

1.41775

1.99512 4.20708 Pair

Residual

HSVL

0.236939 1.26878

5.83406

18.7983

12.3565 Pair

Residual

HTL

0.328489 9.7826 38.5544

30.1447

95.0239

Pair

Residual A Genetic (first line; Pair) and residual (second line; Residual) variances on diagonal (shown in bold), genetic covariances above the diagonal, and residual covariances below the diagonal.

96

Appendix II.

Estimating economic values for crocodile selection indices

E.M. Gray1, S.R. Isberg2, F.W. Nicholas2, C. Moran2, S.G. Barker3 and F. Ahmadi-Esfahani1*

1 Agricultural and Resource Economics, Faculty of Agricuture, Food and Natural Resources,

University of Sydney, NSW, 2006, Australia. 2 Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary Science, University of Sydney, NSW, 2006, Australia.

3 Janamba Croc Farm, PO Box 496, Humpty Doo, NT 0836 Australia. *Corresponding author: F.Ahmadi@agec.usyd.edu.au

Abstract For an economic selection index to operate, selection objectives need to be defined and their respective economic values (EVs) need to be estimated. This study seeks to estimate EVs for previously defined crocodile selection objectives including reproductive output (number of viable hatchlings), juvenile survival, decreasing the time taken to reach harvest size and decreasing food consumed. The method used for estimating these EVs is based on microeconomic theory using a Cobb-Douglas type production function, as opposed to the conventional method based on the profit function alone. The resultant EVs were $41.95 for each additional hatchling for reproductive output, $52.37 for a one per cent increase in juvenile survival, $25.68 for a reduction in the time to reach slaughter size by one week, and $4.67 for reducing food consumed by one gram per week. Sensitivity analyses are also conducted. There is a high demand for first-grade, blemish-free saltwater crocodile skins. These skins are traded in US$ per centimetre and the exchange rate greatly affected the resultant EVs obtained. In addition, the percentage of skins meeting the requirements of first-grade also greatly affected the resultant values. Various other scenarios are also tried, including interactions between the selection objectives. 1. Introduction The Australian crocodile industry is based on the production of saltwater crocodiles (Crocodylus porosus) for their skins. Meat is also produced, but is only considered a by-product. The industry has been estimated to be worth AU$5 million per annum (Stubbs, 1998) with AU$4 million deriving from skin sales. McNamara et al. (2003) reviewed the present status of the emerging Australian crocodile industry which began in the mid-1980s. Juvenile crocodiles are generally harvested with a belly width ranging between 35cm and 45cm, at an average age of 2.8 years1. Skins are sold on a ‘$ per centimetre’ belly-width basis in conjunction with a stringent, yet subjective, grading system. Skins are graded according to the number of blemishes on the belly area. A first grade skin will have no blemishes, four appendages and are well preserved. The presence of any bite marks, abrasions or knife holes results in an automatic downgrading of the skin. First-grade skins are highly sought on the international market. However, there is a weaker demand for lower-grade skins, which are in competition with other crocodilian skins, and receive a large discount in their value. As such, they attract a much lower price, usually being sold domestically or sold for tanning and re-imported for manufacture (Table 1).

1 See Chapter 5

97

Table 1: A range of prices received for saltwater crocodile skins (salted) Prices received per centimetre (US$/cm)

Belly Width (cm) First Grade Second Grade Third Grade

18-24 4.50-5.00

25-34 5.50-6.00

35-39 7.50-8.50 3.20-3.85 1.28-1.93

40-45 8.50-10.00

46-50 10.00-11.00

AU$1 = US$0.64 (20/08/03)

The prices received by Australian producers vary, as producers negotiate individually with the tanneries. Table 1 demonstrates a range of prices prices received by Australian producers. The ascending pricing regime with increasing belly width demonstrates that there is demand for larger skins in excess of 45cm, although increasing the growing out period increases production costs, the risk of blemishes and the subsequent downgrading of skins. The strictness of the skin grading system has significant repercussions, as indicated by the pricing differentials between first grade and lower grade skins in Table 1. This is representative of buyers preferring a first grade, blemish-free skin of lower “quality” (for example, American alligator) over a blemished saltwater crocodile skin. The largest concern of skin buyers is the undersupply of first-grade skins, with so many failing to meet the grading requirements (Manolis et al., 2000). Trade in crocodile skins can be divided into ‘classic’ skins, such as saltwater crocodile, versus others, such as caiman and alligator. Classic skins, such as those obtained from the saltwater crocodile, Nile crocodile (C. niloticus) and Siamese crocodile (C. siamensis) are distinguished by the absence of osteoderms in the belly scales. Classic skins are therefore valued more highly. Saltwater crocodile skins receive a premium price as their skins contain fewer osteoderms relative to the other “classic” skin species. The major export destinations for Australian saltwater crocodile skins are France, Italy, Japan and Singapore. Australia exports approximately 11,000 skins per annum (MacNamara, et al., 2003), supplying one per cent of the global market for crocodilian skins (DOTRS, 2001). Although individual producers negotiate with tanneries, they can be assumed to be price takers1. Isberg et al. (2003) defined selection objectives that affect the profitability of crocodile farms and divided them into two categories: those that influenced farm revenue, and those that influenced production costs (Table 2). This study was meant to estimate the relative change in profit per unit improvement in each breeding objective (Goddard, 1998), and thus report economic values (EVs) for each selection objective.

Table 2: Performance traits in the breeding objective (Isberg et al., 2003) Selection Objective Objective Units

Skin Grade: Percentage first grade Increase 1%

Breeder Output: Viable hatchlings per clutch (NoHatch) Increase 1

Juvenile Survival: Mortality, hatchlings to slaughter Decrease 1 %

Slaughter Age: Days to slaughter Decrease 1 week

Weekly Feed Consumption: Hatchlings to slaughter Decrease 1 gram

1 For a more detailed description of trade in crocodilian skins, see MacNamara et al. (2003).

98

Skin grade directly influences farm income, since there is a strong price differential between first and lower grade skins. Increasing the number of viable hatchlings per clutch (NoHatch) and juvenile survival (and hence output) provides an opportunity for producers, as price takers, to increase revenue. An increase in growth rate would be useful, as the longer the production period, the greater the variable cost per juvenile. Including feed consumption as a selection objective takes increased feed requirements following genetic improvement into account (Jones, 1982) Thus, these selection objectives represent cost-reducing objectives, whereas increasing NoHatch and juvenile survival represent revenue-increasing objectives. 2. Methods 2.1 The estimation of economic values There is no consensus on the optimal method for estimating EVs, although a feature common to most is the use of a profit function. The conventional method estimates the EV of a trait as the partial derivative, or rate of change, of a profit function with respect to that trait, with all other traits held constant at their average values (Goddard, 1998). Profit functions are also used when EVs are estimated as the absolute change in profit, following re-optimisation of key management variables after genetic improvement. EVs are often described as representing the value to the investor of a marginal improvement in the genetic characteristics in the breeding objective (Hazel et al., 1994; Brascamp et al., 1985; Smith et al., 1986), where it is assumed that the manager seeks to use the levels of genetic characteristics under selection (and of non-genetic factors) that maximise profit (Melton et al., 1993). Under the assumption that the goal of the manager is profit maximisation, then microeconomic theory provides a theoretical foundation upon which to base the estimation of the EVs (Melton et al., 1993). Additionally, in order to capture the dynamic nature of genetic improvement, the production system model should address concepts such as the existence of returns to scale in agriculture and the nature of technological change. Therefore, the estimation of EVs can be analysed in the context of an economic problem. 2.2 Genetic improvement in the context of economic theory of the firm Several approaches to estimating EVs have been developed under the conditions of constrained profit maximisation. Examples include Amer and Fox (1992), who incorporated genetic improvement into neoclassical production theory; McArthur (1987), who used a production function to derive EVs that included changes to managerial practices to exploit improved livestock; and Melton et al. (1993), who estimated the EVs under the conditions of constrained profit maximisation. The method of Amer and Fox (1992), which formed a basis for the estimation of EVs in this study, was extended in Amer et al. (1994a) using a generalised Cobb-Douglas production function. The benefit of using this functional form, compared with the linear profit function, is the ability to incorporate the law of diminishing returns to inputs, and returns to scale. This enabled the production relationships between inputs and outputs in the long-run to be more closely modelled (Amer et al., 1994a). Theoretically, the Cobb-Douglas production function will also allow for input re-optimising of input combinations following genetic improvement, so that the EVs represent the absolute change in optimal profit. Consequently, in order to develop a production function for crocodiles, it is necessary to address the concept of returns to scale and the response of the profit maximising producer to a change in their inputs. An important issue in agriculture is the existence of returns to scale in agriculture and is applicable since genetic improvement programs are long-term initiatives. The rate of improvement depends upon many factors, including the amount of genetic variation, intensity of selection, and the length of the generational interval1. Since an appropriate model should include the effects of increasing the scale of the enterprise on the production system, it is necessary to assume a magnitude for the scale economies. For example, a farm may benefit from increasing returns to scale because workers are able to specialise in certain tasks, or use more sophisticated machinery (Pindyck and Rubinfeld, 2001). Conversely, a farm may experience decreasing returns to scale if the scale of the enterprise leads to difficulty in coordination and communication between managers and workers. Overall, the presence of economies (or diseconomies) of scale depends on the potential efficiency gains (or losses) from increasing the scale of the enterprise. There is some disagreement whether the production model should display constant or decreasing returns

1 estimated as 13 years in crocodiles; Chapter 9

99

to scale. Some authors (Amer and Fox, 1992; McArthur, 1987) argue that constant returns to scale do not model the average cost trend of farms because farms faced decreasing marginal returns to an increase in scale. On the other hand, the techniques employed by Smith et al. (1986) when estimating EVs were based on the assumption of constant returns to scale. However, empirical studies have suggested that there was no evidence to suggest the existence of diseconomies of scale for larger enterprises (Debertin, 1986,; Chavas, 2001). As such, constant returns to scale may be taken as a plausible assumption. That genetic improvement is analogous to technological change is evident in the theoretical approaches of Amer and Fox (1992) and McArthur (1987). In these cases, the EVs represented the increase in net profit from the cost savings and increased production associated with genetically superior animals. In the context of production theory, the outcome of genetic improvement on farm-level production, in terms of the improving existing inputs, is the improvement of existing animal production characteristics (Doll and Orazem, 1984). The implication of this conclusion is that genetic improvement, as a form of technological change, often leads to increased output regardless of whether the new technology is targeted at reducing the costs of an input or increasing its productivity. For a further description of the effects of technological change on production decisions, see Doll and Orazem (1984). Briefly, technological change that improves an input will increase the useage of the input, depending on cost, and thus affect the level of output, depending on how the costs associated with the input are changed. The technological change may take the form of a reduction in the cost of the input, an increase in the marginal productivity of the input, or as an increase in output at every level of input use without changing the marginal productivity of the input. 2.3 The methodology of estimating economic values The method chosen for estimating EVs was the absolute change in profit, based on profit maximising behaviour. The key proposition of this approach is that the EVs are estimated as the change in profit when management variables had been reoptimised following genetic improvement. The reasoning underlying this method was that producers would make new management decisions to exploit the genetically superior animals. This was because the full benefits of genetic improvement can only be realised through new management decisions (Amer et al., 1994b; 1997) and, therefore, when modelling EVs, the optimum management should be modelled for each genotype. In comparing the two approaches (EVs as the rate of change of profit or the absolute change in profit), Goddard (1983) indicated that when EVs had been calculated as the rate of change of profit, it was unnecessary to continually re-optimising management variables due to the small nature of genetic change. While it would seem convenient to accept Goddard’s proposition, there are some qualifications. When larger genetic changes are considered, Goddard (1983) indicated that management variables should be reoptimised for the genetically superior animals. Additionally, the simple linear profit function would no longer be suitable. Instead, using a profit function based on a non-linear production function would result in EVs that were no longer constant, but dependent on the levels of the genetic characteristics (Melton et al., 1993). Furthermore, Bright (1991) determined that the use of a simple profit function to derive EVs as the partial derivative was only acceptable for EVs derived under short-run conditions. When addressing the issue of constraining enterprise size, as advocated by Smith et al. (1986), rescaling was rejected because the EVs generated do not include the economic benefits associated with increased production. This is despite a brief study of profit maximising behaviour indicating that an increase in output was often the profit maximising response to technological change (Doll and Orazem, 1984). Moreover, rescaling resulted in EVs of sufficient difference to reduce the efficiency of the selection index, even for small trait changes (Amer et al., 1994a). Constraining management-controlled variables to reflect enterprise size would only be appropriate for defining EVs in the short run, when producers are constrained by fixed or limiting resources, such as pasture availability or purchased feed in dollars (Koots and Gibson, 1998b), or under a quota (Harris and Freeman, 1993). However, genetic improvement programs are long-term initiatives, under which conditions all fixed inputs are variable. Estimating EVs as the absolute change in profit will result in the closest approximation of the true EVs. Furthermore, by re-optimising management variables such as slaughter age, the EVs include the benefits of management decisions taken to exploit genetically-superior animals. Deriving EVs as the partial

100

derivative fails to consider the influence of genetic gain on the production of farms, in terms of their use of resources. The success of the conventional method is further limited by the use of rescaling, which undervalues productivity increasing genetic improvement, and constraining management variables to reflect the enterprise size. This may result in arbitrarily chosen constraints that do not reflect a production system in the long run (Visscher et al., 1994). Further issues that arise regardless of the method used to calculate the EVs include: the perspective of the breeding objective, the derivation of EVs that are equivalent for all bases of evaluation, the treatment of fixed costs, and the discounting of future revenues and costs. Briefly, the perspective is unlikely to affect the specification of the profit function if market signals are not distorted, although difficulties may arise if the goals of the industry and producer (or investor) differ. EVs will be relatively equal regardless of the perspective or unit of evaluation if economic profit equals zero, which will be the case if all costs, including the normal profit that is a return to management, are specified per unit of output as for a long run planning horizon. For a further discussion of these issues, see Goddard (1998). 2.4 The production data for farmed saltwater crocodiles As outlined previously, saltwater crocodiles are intensively farmed for their skins, which are manufactured into luxury leather goods such as handbags and shoes. The industry-preferred belly width is 35-45 centimetres, which corresponds to a slaughter age of three years1. The largest component of the operating costs of crocodile farms is feed, comprising 42-45 per cent (Treadwell et al., 1991). This is because in terms of production per unit time, crocodiles are very inefficient, despite their high feed conversion efficiency (Webb, 1989). The other main component is labour, which is around 40 per cent of operating costs (Treadwell et al., 1991). Since industry data were unavailable, the production and cost function components of the model were based on the production system of Janamba Croc Farm. For a description of the husbandry practices of Janamba Croc Farm, see Chapter 2. The price component of the profit function was taken from representative prices, within the ranges described in Table 2. Table 3 displays relevant production characteristics, taking the form of mean performance levels for breeders and juveniles. Operating and capital costs are displayed in Table 4. Wages and feed prices for Janamba Croc Farm were used, although they are confidential and may not be displayed. Other operating costs were adapted from the CrocProfit program (DPI, 2001), whilst capital costs were derived from McKelvie and Treadwell (1991). 2.5 The production function for farmed saltwater crocodiles The estimation of the EVs as the absolute change in profit, as mentioned above, was the preferred method for calculating the EVs for farmed saltwater crocodiles. This seemed appropriate if the profit maximising behaviour of producers following technological change were used as a guide in determining the response of output levels to genetic improvement. A Cobb-Douglas type production function was chosen to represent the production system of saltwater crocodiles. Conventional Cobb-Douglas production functions have commonly been used to represent the non-linear relationship between output and changes to the traits under selection (Bright, 1991; Amer et al., 1994a). The Cobb-Douglas production function is defined as

βα21 xAxY =

where x1 and x2 are two variable inputs in production, α and β are the partial production elasticities for inputs x1 and x2, and A is a scale variable. The key features of the Cobb-Douglas production function are the exponents, α and β, and the scale variable A. Through the exponents α and β, the function is able to reflect the law of diminishing marginal returns to inputs (Debertin, 1986), whereas the sum of the exponents, (α + β ), indicates the returns to scale parameter of the farm enterprise. The constant term, A, in the Cobb-Douglas production function represents the impact of fixed inputs on the production function,

1 see Chapter 5

101

although an alternative definition is that A represents the current state of the production technology (Debertin, 1986).

Table 3: Mean performance levels for production characteristics of saltwater crocodile breeders and juveniles (at Janamba Croc Farm)

Production Traits Units Performance Level

Crocodile Numbers

Average number of juveniles slaughtered per year No. 2500

Breeding Pairs No. 34

Number of stages in the production system No. 3

Juvenile Traits

Average belly width cm 37.63

Meat Yields Tail fillet % 0.4

Body % 0.6

Productive attributes Days to slaughter days 1054

Total feed consumption kg 101.92

Mortality stage 1 % 8

Mortality stages 2 & 3 % 3

Breeder Traits

Reproductive attributes Females nesting % 97.5

NoHatch No. 31.25

Productive attributes Weekly feed consumption kg / pair 5.882

An extension of the Cobb-Douglas production function was used in this study. However, instead of defining management-controlled variables as the arguments in the production function, such as labour and capital, the Cobb-Douglas type production function used the selection objectives. This is because quantities of the physical inputs used are dependent on requirements to sustain a predetermined level of genetic performance (Tess et al., 1983). This in part reconciles the conventional and production theory approaches, as the profit function of the conventional approach takes the genetic values of the selection objectives as input and generates profit as output (Goddard, 1998).

The production function was defined per breeding pair as γβα321 xxAxY =

where the scale parameter, A, is NoHatch per breeding pair, x1 is the slaughter age, x2 is juvenile survival, x3 is feed consumed per week per juvenile and Y is the output in number of juveniles slaughtered per breeding pair. In accordance with the conclusion that constant returns to scale are a plausible assumption, the sum of the partial elasticities of production (α + β + γ ) was constrained to equal one. α and γ were set equal to zero, and β was set equal to one. In this scenario, the number of juveniles slaughtered per breeding pair is directly dependent on the selection objectives of juvenile survival and breeder output, whilst feed consumed was assumed independent of the other traits. The production function was also calculated using alternative partial production elasticities. However, as the true values of α, β and γ were

102

Table 4: Operating and capital costs for a saltwater crocodile enterprise

Source of Cost AU$ Wages* Feed Prices (per kg)* Operating Costs** Fuel 2000 Repairs and maintenance 20000 Electricity 40000 Accounting and legal 1500 Administrative 3000 Phone 2500 Vehicle registration 500 Vehicle insurance 1000 Other insurance 1000 Council rates 3000 Water charges 500 Water pumping license 500 Salt (medicinal) 4000 Chemicals (medicinal) 500 Chemicals (cleaning) 2000 Pen cleaning equipment 1000 Miscellaneous 4000 Capital Costs (purchase for 53 breeding pairs)*** Office / storage shed 30000 Incubator 18000 Grower pens 180000 Breeder pens 147750 Meat mincer (new) 5000 Freezer (x2) 40000 Cool room 10000 Pressure pump 20000 Nissan Patrol ute 25000 Tractor and backhoe 10000 Trailer 500 4 wheel motorbike 6000 Slasher 2000 Miscellaneous tools 5000

* Confidential information from Janamba Croc Farm ** DPI Queensland (2001) *** McKelvie and Treadwell (1991)

unknown, in terms of the interactions between the selection objectives, the values used were chosen so as not to distort the number of slaughtered juveniles produced by the production function. 2.6 The cost function for farmed saltwater crocodiles The cost function was specified per breeding pair, and was dependent on the number of juveniles slaughtered per breeding pair. Consistent with the proposition that genetic improvement programs are long term initiatives, and the condition of Smith et al. (1986) that resources should be used efficiently in

103

production, the capital costs were specified as dependent on the number of juveniles in the production system. Thus, any increase in the number of juveniles in the production system as a result of genetic improvement resulted in a proportional increase in the fixed inputs. The cost function was defined as

''YvvYC += where C, the costs per breeding pair, is dependent on Y, the number of juveniles slaughtered per breeding pair and v is the cost per juvenile and includes all operating and capital costs. The costs per breeding pair also depend on the mortality costs, v’, which are the costs associated with juveniles that do not survive until slaughter, Y’. Juveniles that died in stage one (assumed to be year one) were assumed to have incurred 50 per cent of the operating costs in year one, excluding labour. Juveniles that died in stages two and three (assumed to be years 2 and 3) were assumed to have incurred all of the operating costs in year one and 50 per cent of operating costs associated with stages two and three, incurred in year two. The assumption that non-surviving juveniles incur 50 per cent of operating costs is consistent with the mortality costs assumed in other studies to determine EVs (De Vries, 1989; Koots and Gibson, 1998a). Operating costs per juvenile were specified on a per week basis, and capital and labour costs per juvenile were specified to be fixed for the duration of the production period. Therefore, only the operating costs, excluding labour, were affected for an improvement in the selection objective x1, slaughter age. As with the production function for saltwater crocodiles, the cost function was developed to exhibit constant returns to scale. In the functional form specified above, long-run average cost is equal to long-run marginal cost (Pindyck and Rubinfeld, 2001). Thus, the profit function for farmed saltwater crocodiles was defined as

''YvvYYPy −−=π

)'(')( 321321321γβαγβαγβαπ xxAxvxxAxvxxAxPy −−=

where A, is NoHatch per breeding pair, x1 is the slaughter age, x2 is juvenile survival, x3 is feed consumed per week per juvenile and Y is the output in number of juveniles slaughtered per breeding pair. P is the weighted average revenue per juvenile, v is the cost per juvenile and includes all operating and capital costs. v’ is the mortality costs, which are the costs associated with juveniles that do not survive until slaughter, Y’. The EVs were calculated as the absolute change in profit following genetic improvement in a selection objective, with the other selection objectives held constant at their pre-improvement levels of performance. By calculating the EVs as the absolute change in profit, the benefits of genetic improvement will include the benefits associated with cost reductions and increased output and revenue in accordance with genetic improvement being analogous to technological change. Due to the length of the production period, costs and revenue were discounted over the three years, specifying a discount rate of four per cent and assuming each stage of production is one year in length. The profit function was specified for the long-run, in which case the farm is maximising profit when economic profit equals zero, and all inputs are receiving a payment. This is consistent with the assumption that the farm is experiencing constant returns to scale when normal profit is included as a cost in the production process as a return to management (Varian, 1999). 2.7 Specification of costs and revenues Labour and operating costs were specified on a per juvenile basis. This was achieved by dividing costs by the average number of juveniles slaughtered each year (as not all juveniles slaughtered in a year are produced by breeders in the genetic improvement program), and then by the number of stages in the production system. This was to take into account that three clutches per breeder (on average) were in the production system at any one time. Juvenile feed consumption was assumed to be at a constant rate throughout the production period. The average price of juvenile feed was taken as a weighted average of the components of the diet in each of the three production stages. Each production stage was assumed to be one year in length.

104

Capital costs from McKelvie and Treadwell (1991) were transformed into purchase cost per breeding pair, by dividing by the number of breeding pairs in the McKelvie and Treadwell model. The capital costs were specified on a per juvenile basis through dividing by the number of stages in the production system and by the number of juveniles slaughtered per breeding pair, which was obtained from the production function. As a result, capital costs per juvenile depended on the specifications of the production function. As the capital costs derived from McKelvie and Treadwell were in 1991 dollars, per juveniles purchase costs were inflated to reflect 2003 values using the method in Trewin (2000). The annual depreciation on the purchase costs was calculated according to the salvage values in McKelvie and Treadwell (1991), and the depreciated values were used in the cost function. Revenue per juvenile was calculated as the sum of the revenue from skins and meat for an average slaughtered juvenile. Skin revenue was calculated for an average belly width of 37.63cm and the price was calculated as a weighted average depending on the proportion of skins in each grade. This was assumed to be 45 per cent first grade, 30 per cent second grade, and 25 per cent third grade. All prices were converted into Australian dollars using the exchange rate of AU$1 = US$0.6421 (August 20, 2003). Meat revenue was calculated for an average yield of 4.5kg per juvenile. The price was calculated as a weighted average depending on the proportion of meat in each cut and the average farm gate price (in Australian dollars) provided by Janamba Croc Farm. The EVs were calculated as the difference in profit following an improvement in each of the selection objectives. An improvement in NoHatch was an increase of one hatchling. An improvement in juvenile survival generated three separate EVs. Firstly, a one per cent fall in mortality in stage one; secondly, a one per cent fall in mortality in stages two and three combined; and thirdly, a one per cent fall in overall mortality. As overall mortality is multiplicative over the three production stages (that is, the mortality rates in stages two and three refer to the number of juveniles surviving stage one), a one per cent fall in mortality overall is larger than a one per cent fall in mortality in either stage individually. An improvement in the selection objective age at slaughter was a decrease in the production period of one week. An improvement in the selection objective of weekly feed consumed was a reduction in feed consumed by one gram per week. 3. Results and Discussion 3.1 Results for the base case The most important selection objectives, in terms of the magnitude of the EVs (Table 5), were found to be survival ($52.37 for a one per cent increase) and NoHatch per breeding pair ($41.95 for each additional hatchling). The magnitude of these results were within expectations, due to the large changes in performance levels achievable for each selection objective. In addition for juvenile survival, the economic benefits also include a reduction in mortality costs (the costs of raising an animal for no economic return). For survival in years two and three, the economic benefits associated with a reduced mortality cost indicate that emphasis should be placed on reducing mortality in these stages despite the higher mortality rates in year one. The EV for slaughter age was $25.68 for a reduction of one week. These were the costs associated with keeping an animal in production for an additional week. The low EV ($4.67) for decreasing weekly feed consumption can be attributed to the magnitude of the genetic change considered, since a reduction of one gram per week was a small trait change compared to the changes considered in the other selection objectives. 3.2 Sensitivity Analyses The EVs were also calculated for a range of alternative scenarios to determine their sensitivity to changes in economic and production system parameters. In addition, since no interactions between the selection objectives have been investigated1, sensitivity analyses were conducted to take into account some hypothetical interactions between production traits.

1 See Chapter 9

105

Feed

Con

sum

ed

(kg)

4.67

(0

.04)

4.63

(0

.04)

4.67

(0

.04)

4.

67

(0.0

3)

4.

58

(0.0

4)

4.76

(0

.04)

4.67

(0

.03)

4.

67

(0.0

4)

Surv

ival

(O

vera

ll)

52.3

7 (0

.42)

52.1

1 (0

.42)

48.7

2 (0

.44)

56

.39

(0.4

0)

50

.58

(0.4

3)

54.2

3 (0

.41)

55.1

6 (0

.41)

49

.57

(0.4

4)

Surv

ival

(Y

ears

2 &

3)

41.2

5

41.0

1

37.9

0

44.9

5

39.6

7

42.8

9

43.8

2

38.6

8

Surv

ival

(Y

ear

1)

22.7

8

22.5

4

19.2

5

26.6

9

21.4

9

24.1

3

25.4

9

20.0

7

Slau

ghte

r A

ge

25.6

8 (0

.21)

25.6

0 (0

.21)

25.6

8 (0

.23)

25

.68

(0.1

8)

24

.96

(0.2

1)

26.4

3 (0

.20)

25.6

8 (0

.19)

25

.68

(0.2

3)

Eco

nom

ic V

alue

s in

AU

$

NoH

atch

41.9

5 (0

.34)

Sens

itivi

ty A

naly

ses E

cono

mic

Val

ues i

n $

41.6

9 (0

.34)

31.5

5 (0

.29)

53

.44

(0.3

8)

38

.39

(0.3

2)

45.6

7 (0

.35)

49.9

2 (0

.37)

33

.97

(0.3

0)

Eco

nom

ic

Prof

it ($

)

37.5

1

29.4

8

-287

.49

396.

73

-49.

75

129.

34

286.

83

-211

.81

0.99

0.

01

+5%

-5%

0.05

0.03

+5%

-5%

Tab

le 5

: Eco

nom

ic v

alue

s for

the

base

cas

e an

d se

nsiti

vity

ana

lyse

s on

the

prod

uctio

n el

astic

ies, β

and γ,

and

the

econ

omic

par

amet

ers.

Val

ues i

n pa

rent

hese

s ind

icat

e th

e re

lativ

e co

ntrib

utio

n of

eac

h ec

onom

ic v

alue

to th

e ec

onom

ic se

lect

ion

inde

x fo

r tha

t sce

nario

.

Scen

ario

s

Bas

e C

ase

Prod

uctio

n Fu

nctio

n A

ssum

ptio

ns

Prod

uctio

n El

astic

ity X

2, β

= Pr

oduc

tion

Elas

ticity

X3, γ

=

Eco

nom

ic D

ata

Ass

umpt

ions

Exch

ange

Rat

e =

Dis

coun

t Rat

e =

Pric

e R

ecei

ved

for F

irst G

read

e Sk

ins =

106

Economic profit per breeding pair was highly sensitive to exchange rate movements, ranging between -$287.49 and $396.73 for a five per cent appreciation and depreciation in the exchange rate, respectively. Exchange rate movements only affected the EVs for the productivity increasing selection objectives (NoHatch, and survival). Thus, the increased importance of the other selection objectives, including juvenile survival, is due to the magnitude of the reduction in the value of the EV of NoHatch. A depreciation of the exchange rate had the opposite effect on the relative importance of each EV. Due to the length of the production period, the costs and revenues were discounted over the three production stages. The discount rate chosen was four per cent, and was varied by one percent for the sensitivity analyses. The implication of a higher discount rate (for example, five per cent) was that the costs and revenues occurring in the final year of the production period are valued less. NoHatch was the selection objective most affected by the higher discount rate, although the EVs of the remaining selection objectives also declined. This had the effect of only marginally altering the relative importance of the EVs. A discount rate of three per cent resulted in an approximately equal and opposite change in the EVs for the other selection objectives. The price for first-grade skins received by Janamba Croc Farm was varied by five percent. The importance of the price of first grade skins for farm revenue was apparent in the effect on economic profit per breeding pair, which ranged between -$211.81 and $286.83. The EV of NoHatch experienced the largest change at almost $8 for a five per cent increase or decrease in price. The influence of NoHatch was again evident with an increase in its EV sufficient to reduce the relative importance of the other selection objectives, despite an increase of almost $3 in the EV of survival overall following the five per cent price increase. The parameters of the production system were also subjected to sensitivity analyses (Table 6). EVs for the selection objectives were calculated under different cost conditions and for a range of possible performance levels of the production traits of the saltwater crocodiles at Janamba. Some of these scenarios are outlined below. McNamara et al. (2003) report that only 50 per cent of skins produced meet the requirements of a first-grade skin. Using the assumption of optimal efficiency, an estimate of 45 per cent first-grade skins was used in this study. To consider the effects of a change in the proportion of skins produced that were first grade, the proportion of first grade skins was varied by one, five and ten per cent. A five per cent increase or decrease in first grade skins changed the EV for NoHatch by approximately $10 in each direction, whilst a 10 per cent change in first grade skins changed the EV for NoHatch by more than $20 in each direction. Again, this had significant implications for the importance of each selection objective. Although the EV for survival was also affected by varying the percentage of first grade skins, the size of the change in the EV of NoHatch dominated these effects, so that the importance of the EVs for survival again fell. However, the effect on the EVs of slaughter age and feed consumed was greater. Similar outcomes occurred when the average belly width of skins was altered. The robustness of the EVs were tested against increases in production costs. Increases of five and 10 per cent in either labour costs, feed costs or the annual expenses (operating costs less feed and labour) resulted in a negative economic profit per breeding pair. As expected, economic profit was most sensitive to increases in feed costs, which Treadwell et al. (1991) had indicated were the main component of the operating costs. An increase in labour costs reduced the relative importance of the EV of NoHatch and increased the relative importance of the remaining selection objectives. This is because the EV for survival was marginally reduced by increased labour costs, as labour was not included in the mortality costs. An increase of five and 10 per cent in feed costs altered the EVs of all selection objectives. However, as the relative importance of slaughter age increased by 2.6 per cent for a 10 per cent increase in feed costs, and the relative increase in the importance of feed consumed was 0.6 per cent, this again highlights that productive efficiency per unit time is more important than per unit of food consumed. The EV of survival was only marginally reduced by increased feed costs. This is because the EV of survival includes the benefit of a reduction in mortality costs, which includes those associated with feed. The effect of an increase in the annual expenses, which comprised all operating costs excluding feed and labour, was found to be small.

107

Feed

Con

sum

ed

(kg)

4.67

(0

.04)

4.67

(0

.03)

4.

67

(0.0

4)

4.67

(0

.03)

4.

67

(0.0

5)

4.67

(0

.04)

4.

67

(0.0

4)

4.

67

(0.0

4)

4.67

(0

.04)

4.

90

(0.0

4)

5.13

(0

.04)

4.

67

(0.0

4)

4.67

(0

.04)

Surv

ival

(O

vera

ll)

53

.13

(0.4

2)

56.1

8 (0

.40)

48

.56

(0.4

4)

59.9

9 (0

.39)

44

.75

(0.4

7)

54.4

0 (0

.41)

50

.33

(0.4

3)

51

.06

(0.4

3)

49.7

5 (0

.43)

52

.08

(0.4

3)

51.7

9 (0

.43)

52

.29

(0.4

2)

52.2

1 (0

.43)

Surv

ival

(Y

ears

2 &

3)

41.9

5

44.7

5

37.7

4

48.2

6

34.2

4

43.1

2

39.3

8

40.0

4

38.8

4

40.7

1

40.1

8

41.1

0

40.9

5

Surv

ival

(Y

ear

1)

23.5

2

26.4

8

19.0

9

30.1

7

15.3

9

24.7

5

20.8

1

21.5

2

20.2

5

21.3

9

20.0

0

22.4

0

22.0

1

Slau

ghte

r A

ge

25

.68

(0.2

0)

25.6

8 (0

.18)

25

.68

(0.2

3)

25.6

8 (0

.17)

25

.68

(0.2

7)

25.6

8 (0

.19)

25

.68

(0.2

2)

25

.68

(0.2

2)

25.6

8 (0

.22)

26

.69

(0.2

2)

27.7

0 (0

.23)

25

.96

(0.2

1)

26.2

4 (0

.22)

Eco

nom

ic V

alue

s in

AU

$

NoH

atch

44

.12

(0.3

5)

52.8

2 (0

.38)

31

.07

(0.2

8)

63.7

0 (0

.41)

20

.19

(0.2

1)

47.7

5 (0

.36)

36

.14

(0.3

1)

38

.21

(0.3

2)

34.4

8 (0

.30)

36

.86

(0.3

1)

34.9

6 (0

.29)

40

.54

(0.3

3)

39.1

4 (0

.32)

Eco

nom

ic

Prof

it ($

)

105.

51

377.

50

-302

.47

717.

48

-642

.45

218.

89

-143

.86

-79.

11

-195

.74

-184

.93

-407

.37

-6.3

3

-50.

18

+1%

+5%

-5%

+10%

-10%

+1cm

-1cm

+5%

+10%

+5%

+10%

+5%

+10%

Tab

le 6

: Eco

nom

ic v

alue

s for

the

base

cas

e an

d se

nsiti

vity

ana

lyse

s for

the

prod

uctio

n sy

stem

. Val

ues i

n pa

rent

hese

s ind

icat

e th

e re

lativ

e co

ntrib

utio

n of

eac

h ec

onom

ic v

alue

to th

e ec

onom

ic se

lect

ion

inde

x fo

r tha

t sce

nario

.

Scen

ario

s

Prod

uctio

n T

raits

Pr

ice

Rec

eive

d fo

r Firs

t Gra

de S

kins

=

(% 3

rd g

rade

skin

s con

stan

t)

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cos

ts o

f Pro

duct

ion

Labo

ur

Feed

Cos

ts

Add

ition

al A

nnua

l Exp

ense

s

108

A shortcoming of the model is that it does not take into account the possible interactions between the selection objectives and production traits. Therefore, EVs were calculated for scenarios that are simplistic examples of what these interactions might involve (Table 7). The EVs were first calculated for a scenario in which belly width increased by one centimetre, the percentage of first grade skins fell by one per cent and average feed consumption increased by one gram per week. This scenario increased the economic profit per breeding pair $144.42, which suggests that for small changes in skin attributes, larger skins are more important than producing marginally more first grade skins. In the second scenario, a possible relationship between slaughter age and skin attributes was proposed. Slaughter age was increased by four weeks as opposed to increasing feed, belly width was increased above the average by one centimetre and the percentage of first grade skins was reduced by one. The magnitudes of the EVs were only marginally affected, and the relative importance of each EV was also relatively unaffected. When the percentage of first grade skins was decreased by two per cent compared with the previous scenario, economic profit was found to be negative. In this scenario, it would not be profit maximising behaviour to keep the juveniles in the production system for the additional period. This reinforces the need for producers to maintain a high percentage of first grade skins, even at the expense of smaller skins. When the previous scenario was reversed to increase the percentage of first grade skins by one per cent, decrease average belly width per juvenile and reduce slaughter age by one week, an interesting result emerged. Compared to a change in economic profit per breeding pair of $46.36 for the reverse situation, economic profit was only increased by $25.05. This suggests that it is beneficial to increase slaughter age despite a small reduction in the average value of a skin. A further investigation of this result should be conducted to confirm the current findings. EVs were also calculated for a production function with alternative partial production elasticities, β = 0.99, α = 0.00, γ = 0.01. The alternative values for the partial production elasticities were chosen so that the number of juveniles slaughtered would not be overly distorted from the outcome in the base case. Except for the scenario when feed costs were increased by 10 per cent, any difference in the relative importance of the EVs was minor, and it can be concluded that the EVs were not sensitive to changes in the production function specifications. These results are displayed in Tables 8, 9 and 10. 4. Conclusions As the breeding goal is usually to increase farm profitability through the use of genetically superior animals as breeders, the value of the additional profit will depend on the characteristics that determine how saltwater crocodiles function as production units. The objective of this project was to determine the additional profit earned from a saltwater crocodile in production once it has experienced an improvement in a trait included the genetic improvement program. Furthermore, as the EVs function as weights for the estimated breeding values of each selection objective in an economic selection index, they provide an indication of the relative importance of each trait. By analysing genetic improvement in the context of the economics of the firm, the EVs represent the benefits of technological improvement in animal breeding. This is important, as in the competitive market producers are price takers, and as such can only increase profit through the utilisation of cost-reducing or productivity-increasing new technology. Therefore, the selection objectives of farmed saltwater crocodiles can be seen to represent cost-reducing genetic improvement, namely feed consumed per week and slaughter age, and productivity increasing genetic improvement, namely NoHatch and juvenile survival.

109

Feed

Con

sum

ed

(kg)

4.67

(0

.04)

4.79

(0

.04)

4.79

(0

.04)

4.55

(0

.04)

Surv

ival

(O

vera

ll)

53.6

1 (0

.41)

53.5

8 (0

.42)

52.8

0 (0

.43)

51.1

1 (0

.42)

Surv

ival

(Y

ears

2 &

3)

42.3

8

42.3

9

41.6

7

40.0

7

Surv

ival

(Y

ear

1)

23.9

6

22.8

8

22.1

2

22.6

5

Slau

ghte

r A

ge

25.7

1 (0

.20)

25.6

8 (0

.20)

25.6

8 (0

.21)

25.6

8 (0

.21)

Eco

nom

ic V

alue

s in

AU

$

NoH

atch

45.3

7 (0

.35)

42.2

3 (0

.33)

39.9

9 (0

.32)

41.5

5 (0

.34)

Eco

nom

ic

Prof

it ($

)

144.

42

46.3

6

-23.

44

25.0

5

+1cm

-1

%

+1g

+1cm

-1

%

+4w

eeks

+1cm

-2

%

+4w

eeks

-1cm

+1

%

-4w

eeks

Tab

le 7

: Eco

nom

ic v

alue

s for

the

base

cas

e an

d se

nsiti

vity

ana

lyse

s on

inte

ract

ions

bet

wee

n tra

its. V

alue

s in

pare

nthe

ses i

ndic

ate

the

rela

tive

cont

ribut

ion

of e

ach

econ

omic

val

ue to

the

econ

omic

sele

ctio

n in

dex

for t

hat s

cena

rio.

Scen

ario

s

Prod

uctio

n T

raits

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Wee

kly

Feed

Con

sum

ptio

n =

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Day

s to

Slau

ghte

r =

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Day

s to

Slau

ghte

r =

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Day

s to

Slau

ghte

r =

110

Feed

Con

sum

ed

(kg)

4.63

(0

.04)

4.64

(0

.04)

4.

63

(0.0

3)

4.

55

(0.0

4)

4.72

(0

.04)

4.63

(0

.03)

4.

64

(0.0

4)

Surv

ival

(O

vera

ll)

52.1

1 (0

.42)

48.5

2 (0

.44)

56

.09

(0.4

0)

50

.35

(0.4

3)

53.9

6 (0

.41)

54.8

7 (0

.41)

49

.36

(0.4

4)

Surv

ival

(Y

ears

2 &

3)

41.0

1

37.7

1

44.6

7

39.4

5

42.6

5

43.5

5

38.4

8

Surv

ival

(Y

ear

1)

22.5

4

19.0

5

26.3

9

21.2

7

23.8

7

25.2

1

19.8

6

Slau

ghte

r A

ge

25.6

0 (0

.21)

25.6

0 (0

.23)

25

.60

(0.1

8)

24

.88

(0.2

1)

26.3

5 (0

.20)

25.6

0 (0

.19)

25

.60

(0.2

3)

Eco

nom

ic V

alue

s in

AU

$

NoH

atch

41.6

9 (0

.34)

31.3

2 (0

.29)

53

.15

(0.3

8)

38

.15

(0.3

2)

45.4

0 (0

.35)

49.6

4 (0

.37)

33

.73

(0.3

0)

Eco

nom

ic

Prof

it ($

)

29.4

8

-294

.51

387.

58

-57.

36

120.

87

278.

03

-219

.06

0.99

0.

01

+5%

-5%

0.05

0.03

+5%

-5%

Tab

le 8

: Eco

nom

ic v

alue

s for

the

func

tion

with

alte

rnat

ive

parti

al p

rodu

ctio

n e

last

iciti

es a

nd se

nsiti

vity

ana

lyse

s for

the

econ

omic

as

sum

ptio

ns. V

alue

s in

pare

nthe

ses i

ndic

ate

the

rela

tive

cont

ribut

ion

of e

ach

econ

omic

val

ue to

the

econ

omic

sele

ctio

n in

dex

for t

hat

scen

ario

.

Scen

ario

s

Alte

rnat

ive

func

tion

Prod

uctio

n El

astic

ity X

2, β

= Pr

oduc

tion

Elas

ticity

X3, γ

= E

cono

mic

Dat

a A

ssum

ptio

ns

Exch

ange

Rat

e =

Dis

coun

t Rat

e =

Pric

e R

ecei

ved

for F

irst G

read

e Sk

ins =

111

Feed

Con

sum

ed

(kg)

4.63

(0

.04)

4.63

(0

.03)

4.

64

(0.0

4)

4.62

(0

.03)

4.

64

(0.0

5)

4.63

(0

.04)

4.

63

(0.0

4)

4.

63

(0.0

4)

4.64

(0

.04)

4.

87

(0.0

4)

5.10

(0

.04)

4.

63

(0.0

4)

4.63

(0

.04)

Surv

ival

(O

vera

ll)

52.8

7 (0

.42)

55.8

7 (0

.40)

48

.35

(0.4

4)

59.6

3 (0

.39)

44

.59

(0.4

7)

54.1

2 (0

.41)

50

.11

(0.4

3)

50

.82

(0.4

3)

49.5

3 (0

.43)

51

.85

(0.4

3)

51.5

8 (0

.45)

52

.04

(0.4

2)

51.9

7 (0

.43)

Surv

ival

(Y

ears

2 &

3)

41.7

1

44.4

7

37.5

6

47.9

3

34.1

0

42.8

6

39.1

7

39.8

3

38.6

4

40.5

0

39.9

9

40.8

7

40.7

3

Surv

ival

(Y

ear

1)

23.2

7

26.1

8

18.8

9

29.8

3

15.2

4

24.4

8

20.5

9

21.2

9

20.0

4

21.1

7

19.8

0

22.1

6

21.7

8

Slau

ghte

r A

ge

25.6

0 (0

.20)

25.6

0 (0

.19)

25

.60

(0.2

3)

25.6

0 (0

.17)

25

.60

(0.2

7)

25.6

0 (0

.19)

25

.60

(0.2

2)

25

.60

(0.2

2)

25.6

0 (0

.23)

26

.61

(0.2

2)

27.6

1 (0

.24)

25

.88

(0.2

1)

26.1

6 (0

.22)

Eco

nom

ic V

alue

s in

AU

$

NoH

atch

43.8

6 (0

.35)

52.5

3 (0

.38)

30

.84

(0.2

8)

63.3

8 (0

.41)

20

.00

(0.2

1)

47.4

7 (0

.36)

35

.90

(0.3

1)

37

.97

(0.3

2)

34.2

5 (0

.30)

36

.62

(0.3

1)

31.5

6 (0

.27)

40

.29

(0.3

3)

38.8

9 (0

.32)

Eco

nom

ic

Prof

it ($

)

92.2

7

368.

41

-309

.44

707.

33

-648

.37

210.

29

-151

.33

-86.

78

-203

.04

-192

.48

-414

.45

-14.

23

-57.

95

+1%

+5%

-5%

+10%

-10%

+1cm

-1cm

+5%

+10%

+5%

+10%

+5%

+10%

Tab

le 9

: Eco

nom

ic v

alue

s for

the

func

tion

with

alte

rnat

ive

parti

al p

rodu

ctio

n el

astic

ities

and

sens

itivi

ty a

naly

ses f

or th

e pr

oduc

tion

syst

em.

Val

ues i

n pa

rent

hese

s ind

icat

e th

e re

lativ

e co

ntrib

utio

n of

eac

h ec

onom

ic v

alue

to th

e ec

onom

ic se

lect

ion

inde

x fo

r tha

t sce

nario

.

Scen

ario

s

Prod

uctio

n T

raits

Pr

ice

Rec

eive

d fo

r Firs

t Gra

de S

kins

=

(% 3

rd g

rade

skin

s con

stan

t)

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cos

ts o

f Pro

duct

ion

Labo

ur

Feed

Cos

ts

Add

ition

al A

nnua

l Exp

ense

s

112

Feed

Con

sum

ed

(kg)

4.63

(0

.04)

4.75

(0

.04)

4.75

(0

.04)

4.51

(0

.04)

Surv

ival

(O

vera

ll)

53.3

4 (0

.41)

53.3

3 (0

.42)

52.0

4 (0

.43)

50.8

6 (0

.42)

Surv

ival

(Y

ears

2 &

3)

42.1

3

42.1

5

40.9

7

39.8

4

Surv

ival

(Y

ear

1)

23.6

9

22.6

3

21.3

8

22.4

0

Slau

ghte

r A

ge

25.6

3 (0

.20)

25.6

0 (0

.20)

25.6

0 (0

.21)

25.6

0 (0

.21)

Eco

nom

ic V

alue

s in

AU

$

NoH

atch

45.1

0 (0

.35)

41.9

7 (0

.33)

38.2

6 (0

.32)

41.2

9 (0

.34)

Eco

nom

ic

Prof

it ($

)

136.

08

38.2

9

77.6

8

17.0

7

+1cm

-1

%

+1g

+1cm

-1

%

+4w

eeks

+1cm

-2

%

+4w

eeks

-1cm

+1

%

-4w

eeks

Tab

le 1

0: E

cono

mic

val

ues f

or th

e fu

nctio

n w

ith a

ltern

ativ

e pa

rtial

pro

duct

ion

elas

ticiti

es a

nd se

nsiti

vity

ana

lyse

s on

inte

ract

ions

bet

wee

n tra

its.

Val

ues i

n pa

rent

hese

s ind

icat

e th

e re

lativ

e co

ntrib

utio

n of

eac

h ec

onom

ic v

alue

to th

e ec

onom

ic se

lect

ion

inde

x fo

r tha

t sce

nario

.

Scen

ario

s

Prod

uctio

n T

raits

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Wee

kly

Feed

Con

sum

ptio

n =

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Day

s to

Slau

ghte

r =

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Day

s to

Slau

ghte

r =

Cha

nge

Ave

rage

Bel

ly W

idth

=

Cha

nge

Perc

enta

ge o

f Firs

t Gra

de S

kins

=

Cha

nge

Day

s to

Slau

ghte

r =

113

The EVs calculated for farmed saltwater crocodiles indicate that an increase in NoHatch will increase profit per breeding pair by $41.95, whilst reducing slaughter age by one week increases profit per breeding pair by $25.68. Increasing juvenile survival by one per cent increases profit per breeding pair by $52.37, whereas decreasing feed consumed per week by one gram produces a negligible increase in profit of $4.67. Although juvenile survival is the most important selection objective in terms of the magnitude of its EV, the relative importance of the each EV depends on the magnitude of the EV of NoHatch. Changes to parameters determining the costs and revenues of the production system have the greatest influence on the EV of NoHatch. This is obvious because its EV is almost entirely determined by the effect of genetic improvement on revenue per breeding pair. In contrast, the EV of survival was a combination of the cost savings from a reduction in mortality and the increased revenue from more juveniles reaching slaughter size. Feed consumed per week was the least important selection objective, caused mainly by the small marginal change (one gram per week) evaluated. It is important that the profit function used to calculate the EVs closely approximates true farm profit, as the efficiency of the economic selection index depends on the accuracy of the EVs (Goddard, 1998). The selection objectives should directly relate to all sources of revenue and costs, as the omission of an important trait might distort the eventual direction of selection, by exaggerating the economic benefits or ignoring the costs of selection for a certain trait. Harris and Freeman (1993) argued that EVs derived for current prices and costs were valid only if the production costs and market conditions were expected to be stable into the future. After assessing the impact of large changes in the EVs, Smith (1983) concluded that frequent revision of EVs to accommodate new husbandry techniques, changes in market conditions or the increased productivity of the improved livestock was unnecessary, as the affect on efficiency would be small. This can be confirmed for the EVs for farmed saltwater crocodiles through comparison with the EVs calculated in the sensitivity analyses. The changes in the economic parameters, production costs and productive attributes were specified to reflect possible future market and production system conditions that might be encountered by producers. The preceeding sensitivity analyses primarily affected the economic profit earned per breeding pair, the most extreme fluctuations occurring when the percentage of first grade skins was varied by 10 per cent. The same scenario had the most significant consequences for the relative importance of the EVs of the selection objectives NoHatch, slaughter age and survival. However, for the less extreme sensitivity analyses, the relative importance of each EV was stable and generally varied by less than four per cent. The magnitude of the EVs varied in response to changes in the function parameters, but given that the function of the EVs is to act as weights, the efficiency of the economic selection index is unlikely to be compromised. However, Smith (1983) stressed that whole life-time productive efficiency should be considered when defining the breeding objective. In the event that an important trait was left out, the likelihood that the efficiency of the index has been compromised increases. Such traits may include those related the reproductive life of female crocodiles and juvenile temperament. Furthermore, the breeding objective did not include meat production as a selection objective, as meat is only a by-product of skin production. The EVs were calculated using data made available from Janamba Croc Farm, and were considered representative of the production and marketing systems of Australian crocodile farms. A recommendation would be to survey the industry to obtain a more accurate perspective of the economic climate of the industry. In addition, a benefit-cost analysis to evaluate the overall impact of a genetic improvement is also recommended. Overall, the EVs indicate that farm profitability will be increased the most if genetic improvement is realised in the selection objectives that increase productivity; NoHatch and juvenile survival. Although the EVs for slaughter age and feed consumed per week are of a lesser magnitude, they represent significant opportunities for cost savings. If improvement in those selection objectives can be achieved rapidly, and for rates greater than the one unit of improvement measured in this study, it may be more profitable to direct selection towards animals that are superior in those traits. The usefulness of the EVs may be compromised by several factors. The most important factors are as follows. The EVs have been calculated using current prices, and thus represent current demand and

114

supply. Significant changes in world supply or demand could affect market conditions enough to require new EVs to be calculated. However, the indication is that prices for first grade saltwater crocodile skins will be stable into the future. McNamara et al. (2003) noted that demand in France, the major market for first grade skins, was strong, and that there was no indication that an increase in the supply of first grade skins would depress prices. However, unforseen shifts in demand, such as a backlash against animal leather by consumers, might drastically reduce the prices of crocodile skins. When the implications of a large demand shift are combined with the sensitivity of economic profit to exchange rate appreciations and the proportion of first grade skins produced, the vulnerability of the industry is evident. This is further compounded by the heavy dependence of the crocodile industry on fashion-based demand for crocodile leather, as the end use of skins is essentially confined to the manufacture of luxury leather items. Developments in the Australian industry may also curtail the usefulness of the EVs derived in this study. Growth in value adding in the domestic industry, especially through the expansion of tanning capabilities and the development of a branded product (MacNamara et al., 2003) might significantly affect skin revenue. Product diversification in terms of niche markets for meat and other by-products would require further traits to be included in the breeding objective, as producers earn a relatively larger proportion of income from sources other than skins.

115

References

Amer, P.R. and Fox, G.C. (1992), ‘Estimation of economic weights in genetic improvement using

neoclassical production theory: an alternative to rescaling’, Animal Production 54, 341-350.

Amer, P.R., Fox, G.C. and Smith, C. (1994a), ‘Economic weights from profit equations: appraising their accuracy in the long run’, Animal Production 58(1),11-18.

Amer, P.R., Kemp, R.A., Fox, G.C. and Smith, C. (1994b), ‘An economic comparison of beef cattle genotypes for feedlot traits at their optimal slaughter end point’, Canadian Journal of Animal Science 74, 7-14.

Amer, P.R., Emmans, G.C. and Simm, G. (1997), ‘Economic values for carcase traits in UK commercial beef cattle’, Livestock Production Science 51, 267-281.

Bright, G. (1991), ‘Economic weights from profit equations: appraising their accuracy’, Animal Production 53(3), 395-398.

Brascamp, E.W., Smith, C. and Guy, D.R. (1985), ‘Derivation of economic weights from profit equations’, Animal Production 40, 175-180.

Chavas, J.P. (2001), ‘Structural change in agricultural production: Economics, technology and policy’, in Bruce L. Gardner and Gordon C. Rausser (eds.), Handbook of Agricultural Economics volume 1A: Agricultural Production, Elsevier Science B.V., Amsterdam.

Debertin, D.L. (1986), Agricultural Production Economics, Macmillan, New York.

Department of Transport and Regional Services (DOTRS), (2001), Crocodile farming, in Investing in Northern Australia: Key Opportunities-Emerging Industries Sectors, Canberra. [http://www.dotrs.gov.au]

Department of Primary Industries (DPI), (2001). CrocProfit – Decision Tool for Farmers and Investors, Queensland. [http://www.dpi.qld.gov.au]

Doll, P.J. and Orazem, F. (1984), Production Economics: Theory with Applications, 2nd edn, John Wiley and Sons, Inc.

Goddard, M.E. (1983), ‘Selection indices for non-linear profit functions’, Theoretical Applied Genetics 64, 339-344.

Goddard, M. E. (1998), ‘Consensus and debate in the definition of breeding objectives’, Journal of Dairy Science 81(2), 6-18.

Harris, B.L. and Freeman, A.E. (1993), ‘Economic weights for milk yield traits and herd life under various economic conditions and production quotas’, Journal of Dairy Science 76, 868-879.

Hazel, L.N., Dickerson, G.E. and Freeman, A.E. (1994), ‘The selection index- then, now, and for the future’, Journal of Dairy Science 77(10), 3236-3251.

Isberg, S.R., Nicholas, F.W., Thomson, P.C., Barker, S.G., Manolis, S.C. and Moran, C. (2003), ‘Defining breeding objectives for saltwater crocodile genetic improvement programs’, In Proceedings of the Association for the Advancement of Animal Breeding and Genetics, Melbourne, Australia, 6-11 July 2003, Vol. 15, pp. 166-169.

Jones, L.P. (1982), ‘Economic aspects of developing breeding objectives: a specific example; breeding objectives for merino sheep’, in J.S.F. Barker, K. Hammond and A.E. McClintock (eds.), Future Developments in the genetic Improvement of Animals, pp119-136, Academic Press, London.

Koots, K.R. and Gibson, J.P. (1998a), Economic values for beef production traits from a herd level bioeconomic model’, Canadian Journal of Animal Science 78, 29-45.

116

Koots, K.R. and Gibson, J.P. (1998b), ‘Effects of production and marketing circumstances on economic values for beef production traits’, Canadian Journal of Animal Science 78, 47-55.

MacNamara, K., Nicholas, P., Murphy, D., Riedel, E., Goulding, B., Horsburgh, C., Whiting, T. and Warfild, B., (2003), ‘Markets for skins and leather- from goat, emu, ostrich, crocodile and camel industries’, RIRDC – Publ. No 02/142, Canberra.

Manolis, S.C., Webb, G. and Richardson, K. (2000), ‘Improving the quality of Australian crocodile skins’, RIRDC – Publ. No 00/21, Canberra.

McArthur, A.T.G. (1987), ‘Weighting of breeding objectives- an economic approach’, Proceedings of the Australian Association of Animal Breeding and Genetics 6, 179-187.

McKelvie, L. and Treadwell, R. (1991), ‘The economics of crocodile farming’, Proceedings of the Intensive Tropical Animal Production Seminar, Townsville.

Melton, B.E., Wilham, R.L. and Heady, E.O. (1993), ‘A note on the estimation of economic values for selection indices: a response’, Animal Production 59, 455-459.

Pindyck, R.S. and Rubinfeld, D.L. (2001), ‘Microeconomics,’ 5th edn, Prentice Hall International, Inc., New Jersey.

Smith, C. (1983), ‘Effects of changes in economic weights on the efficiency of index selection’, Journal of Animal Science 56(5), 1057-1064.

Smith, C., James, J.W. and Brascamp, E.W. (1986), ‘On the Derivation of Economic Weights in Livestock Improvement’, Animal Production 43, 545-551.

Stubbs, A. (1998), ‘Information systems for new animal industries’, RIRDC – Publ. No 98/139, Canberra.

Tess, M.W., Bennett, G.L. and Dickerson, G.E. (1983), ‘Simulation of genetic changes in life cycle efficiency of pork production. 1. A bioeconomic model’, Journal of Animal Science 56(2), 336-349.

Treadwell, R., McElvie, L and Maguire, C.B. (1991). ‘Profitability of selected aquaculture species’, ABARE discussion paper 91.11.

Trewin, D. (2000), ‘A guide to the consumer price index,’ 14th Series, Australian Bureau of Statistics, Report No. 6440.0.

Varian, H.R. (1999), ‘Intermediate Microeconomics: A Modern Approach,’ 5th edn, W.W. Norton & Company, Inc., New York.

Visscher, P.M., Bowman, P.J. and Goddard, M.E. (1994), ‘Breeding objectives for pasture based dairy production systems’, Livestock Production Science 40, 123-137.

De Vries, A.G. (1989), ‘A model to estimate the economic values of traits in pig breeding’, Livestock Production Science 21, 49-66.