Livestock and Land-Use Surveys in Sub-Saharan Africa

8/7/2019 Livestock and Land-Use Surveys in Sub-Saharan Africa

1/46


2/46

Livestock and Land-useSurveys in Sub-SaharanAfricaW illiam Wint an d Da vid B ou rnEnvironm ental Research Group Oxford

(ERGO)Oxfam Research Paper 11

Oxfam Research Papers report the findings of original research,commissioned by Oxfam (UK and Ireland) to support its

overseas programme of relief and development.Oxfam (UK and Ireland)


3/46

Oxfam (UK and Ireland ) 1994A catalogue record for this book is available from the British lib ra ry .ISBN 0 85598 284 5

Published by Oxfam (UK and Ireland)274 Banbury Road, Oxford, 0X 2 7DZTypeset in 11 pt. New Century SchoolbookPrinted by Oxfam Print UnitOxfam is a registered charity no. 202918

This book converted to digital file in 2010


4/46

ContentsAcknowledgements ivIntroduction 1

W hat determines livestock num bers? 1Unreliable statistics? 2Carry ing capacity 3

A unique data set: the ERGO database 3D ata collection, coverage an d con tents 4Sites surveyed and da ta collected 5Accuracy of the data - are they reliable? 6D ata processing and analysis 7

Prelim inary findings 7Relationships between livestock and environment 7Sum ma ry of preliminary findings 10Implications 12

Recom mendations for further study 15References 17Appendix 1 Aerial survey sampling 24Appendix 2 Su mm ary of surveys carried o ut since 1980 and included w ithin

the ERGO data set 25Appendix 3 Regression and correlation analyses - basic principles and a

worked example 26Appendix 4 Relationship between livestock and environment 30Appendix 5 Stepwise mu ltiple regression 35


5/46


6/46

AcknowledgementsThis repo rt draw s on data collected in the field over the last 13 yea rs. The au thorsgratefully acknowledge the efforts and assistance provided by all the local andexp atriate staff concerned, as well as the staff of the numerous na tional and inter-national institutions that have supported, commissioned or funded the surveys,with particular thanks to the International Livestock Centre for Africa, underwhose aegis many of the techniques used were developed and tested. The authorswish to tha nk the people who read and commented on various drafts of the report,includ ing Jo hn Rowley of Oxfam; David Rogers, Mike Packer, Simon H ay, MalcolmCoe and Clive Hambler of the Department of Zoology, Oxford University; BrianWilliams of the London School of Hygiene and Tropical Medicine; Richard Baker ofthe UK Central Science Laboratory, Harpenden; and Camilla Toulmin and IainScoones of the Intern ationa l In stitute for Environm ent and Development, London.This paper was written for Oxfam by William Wint and David Bourn of theEnv ironmental R esearch Group, Oxford. It ha s been edited and prepared for publi-cation by Olivia G raham.


7/46


8/46

IntroductionFor many years now, it has been apparent that the extensive savannah regions ofAfrica are in a state of transition. They are no longer 'natural' ecosystems, but aresubject to increasing pressures on land use, as a result of the conflicting interests offarmers, p astoralists, foresters, and conservationists. A major debate is tak ing placeon how these areas can be developed sustainably, and how an appropriate balancebetween crops, livestock, trees, and wildlife can be established and ma intained.In recent years there has been an increasing awareness that arid rangelands havebeen progressively degraded by drought and over-exploitation. Until recently, thefocus of this awareness has been on the advance of the desert into degraded areas.Toulmin (1993) has no ted tha t the te rms of the lan d degradation debate have changedover the past few years from 'combating desertification' to 'improving natural-resource management in dry land regions'. In her view this is more than just achange in terminology. It has helped to shift a ttention to the institu tions responsibleat local and national levels for managing the ways in which natural resources areactually used.This is a sensible approach, but at the same time it must be recognised that soundmanagement depends on reliable information, and that not enough is known abouteither the current state of natural resources in many sub-Saharan countries, or howthe situation has changed over time. Without more and better information, it is farfrom certain that any strategies designed to improve matters will succeed. With thegrowth of human populations there has been an expansion of agriculture and atransfo rma tion of vegetation and land use , which have not been adeq uatelymonitored and which are poorly documented.This paper presents the findings of a preliminary analysis of data from systematiclow-level aerial surveys and complementary ground studies of livestock and land usein Mali, Niger, Nigeria, Sudan and Chad. The to tal area covered was more than 1.5million square kilometres. The surveys were carried out between 1980 and 1993, inboth wet and dry seasons. The analysis of the data provides an independent andobjective assessment of resources across a range of agro-climatic zones. The paperidentifies some fascinating and significant environmental correlates of livestockdistribution, and highlights features worthy of further investigation.What determ ines livestock numbers?This is not a straightforward question, and different researchers have come to bothcomplementary and divergent views on it. In a regional study of wildlife and pastoralsystems in Eastern and Southern Africa, Coe et al. (1976) found strong correlationsbetween animal numbers , mean annual ra infa l l , and above-ground pr imaryproduction.They noted th at the re were more animals in six pastoral are as of Kenya surveyed byWatson (1972) than in other areas with equivalent rainfall. Many domestic animalshad died in the six areas during the 1973/74 drought, and this was interpreted as anindication that the previous stocking levels had been above the long-term carryingcapacity.


9/46

Whilst acknowledging that other factors, such as soil fertility and ground wateravailability, also influence primary production, and hence carrying capacity at thelocal level, Coe et al. (1976) argued that rainfall w as the main determ inant of animalnu m ber s ac ross a broad rang e of sav ann ah reg ions rece iv ing 160-700mmprecipitation per annum. The underlying thrust of their argument was that wildlifesystems, having evolved over millions of years, provide as good an indication as any ofhow many anim als might be sustainably supported in the medium to long term.Extending this approach to tsetse-free zones of Ethiopia, Bourn (1976) showed thatcattle numbers increase with:

hum an population d ensity;rainfall;altitude.Tsetse-infested areas of Ethiopia generally support much lower cattle densities tha nthose which are tsetse-free. This pattern is repeated across the continent. In tsetse-free countries national cattle numbers increase with mean annual rainfall; there aresubstantially fewer animals in tsetse-infested countries; and countries withsignificant num bers of trypano-tolerant cattle have intermediate n um bers of anim als.Building on these findings and incorporating additional sources of information, Bell(1982 and 1985) drew attention to the important modifying influence of soil-nutrientstatus on the general relationship between plants and animal populations and wateravailability.In a regional assessment of South American rangelands in Argentina and Uruguay,Oesterheld et al., (1992) found that domestic ungulate numbers for a given level ofabove-ground primary production (which can be equated with rainfall) were greaterthan those of wild herbivores in natural ecosystems world-wide, and concluded that'elementary' (by which they meant extensive) animal husbandry practices raise thecarrying capacity of rangelands for large stock.However, this conclusion has been disputed by Fritz and Duncan (1993), who wereunable to find any significant difference between pastoral grazing areas and naturalwildlife site s, for example game reserves, in sa vannah regions of Africa. Ins tead , th eirana lysis , similar to th at of Bell, showed that, for any given rainfall level, soil type andnutrient status (presumably reflecting the quality of available grazing) were the keydeterm inan ts of animal numbers in both pastoral and wildlife system s.Unreliable statistics?Can we rely on national livestock population figures? These are derived from varioussources of information, including vaccination and dip-tank records, agricultural fieldsurveys, tax retu rns, and slaughter, marketing, and trade data.Where countries have undertaken major vaccination campaigns, against rinderpestfor example, this obviously gives an indication of animal population size, but onlyrarely are all animals vaccinated, and an assumption must therefore be made aboutthe proportion which were not treated. Also, there is always the possibility th at some


10/46

animals were vaccinated more than once. Major disease-control campaigns aretargeted at single species, usually cattle or small ruminants. Other species oflivestock are rarely included, and estimates of their population size tend to be mostuncertain.No stan dar d method of assessment exists and m ethods vary from country to country.The reliability of livestock population figures obviously depends on the quality of theda ta collected an d the validity of underlying assumptions. FAO publishes a n annu alcompilation of national livestock statistics derived from information provided bymember states, but these figures are rarely based on objective field surveys and arenot necessarily reliable.Carrying capacityIn recent years, there has also been a great deal of thinking about the followingquestions:1 Is it possible to calculate or define a maximum livestock density (often known a s'carrying capacity*) that an area can permanently support?2 Is the idea of a fixed carrying capacity applicable to pasto ral (or agrop astoral)systems w here livestock managem ent is dom inated by risk-aversion strategies?3 W hat exactly is mean t by the term 'desertification', which is so commonly used inthe li teratu re and is increasingly a focus of development policy?In recent years the usefulness of the concept of carrying capacity has been called intoquestion by many writers. Behnke et al. , (1993) suggest that whilst it may betheoretically possible to define a maximum sustainable stocking level for an area ofland as a preliminary indicator of potential overstocking, the idea of carrying capacityis largely irrelevant to the formulation of livestock policy. This is partly because it isnot possible to regulate animal numbers effectively given current systems ofcommunal land -tenure and opportunistic pastoral management in Africa.Others argue that the idea of a fixed and constant sustainable stocking level isfundamentally flawed when applied to pastoral systems. This is because there is toomuch ann ual v ariation in productivity of any one area, and because livestock are notgenerally kept in one place for the whole year round. In particular, there is a need toestablish new ways of predicting natural variability and the ways in which pastoralmanagement adapts to it, and also to develop more appropriate techniques forassessing ra nge land carrying capacity.So, conventional theories and recommended management practices derived from theidea of carry ing capacity are increasingly thought to be inappropriate for Africa.

A uniqu e data set: the ERGO databaseThe rest of this paper concentrates on the somewhat surprising findings of apreliminary analysis of a database assembled by The Environm ental R esearch Group


11/46

animals were vaccinated more than once. Major disease-control campaigns aretargeted at single species, usually cattle or small ruminants. Other species oflivestock are rarely included, and estimates of their population size tend to be mostuncertain.No stan dar d method of assessment exists and m ethods vary from country to country.The reliability of livestock population figures obviously depends on the quality of theda ta collected an d the validity of underlying assumptions. FAO publishes a n annu alcompilation of national livestock statistics derived from information provided bymember states, but these figures are rarely based on objective field surveys and arenot necessarily reliable.Carrying capacityIn recent years, there has also been a great deal of thinking about the followingquestions:1 Is it possible to calculate or define a maximum livestock density (often known a s'carrying capacity*) that an area can permanently support?2 Is the idea of a fixed carrying capacity applicable to pasto ral (or agrop astoral)systems w here livestock managem ent is dom inated by risk-aversion strategies?3 W hat exactly is mean t by the term 'desertification', which is so commonly used inthe li teratu re and is increasingly a focus of development policy?In recent years the usefulness of the concept of carrying capacity has been called intoquestion by many writers. Behnke et al. , (1993) suggest that whilst it may betheoretically possible to define a maximum sustainable stocking level for an area ofland as a preliminary indicator of potential overstocking, the idea of carrying capacityis largely irrelevant to the formulation of livestock policy. This is partly because it isnot possible to regulate animal numbers effectively given current systems ofcommunal land -tenure and opportunistic pastoral management in Africa.Others argue that the idea of a fixed and constant sustainable stocking level isfundamentally flawed when applied to pastoral systems. This is because there is toomuch ann ual v ariation in productivity of any one area, and because livestock are notgenerally kept in one place for the whole year round. In particular, there is a need toestablish new ways of predicting natural variability and the ways in which pastoralmanagement adapts to it, and also to develop more appropriate techniques forassessing ra nge land carrying capacity.So, conventional theories and recommended management practices derived from theidea of carry ing capacity are increasingly thought to be inappropriate for Africa.

A uniqu e data set: the ERGO databaseThe rest of this paper concentrates on the somewhat surprising findings of apreliminary analysis of a database assembled by The Environm ental R esearch Group


12/46

Oxford (ERGO). The database is derived from standardised methods of informationcollection over a wide variety of environments across Africa, from the arid rangelandsof the Sahel to the humid forest zone.The database consolidates the records of a series of systematic low-level aerialsurveys and complementary ground stud ies conducted in M ali, Niger, Nigeria, S udanand Chad since 1980. These surveys were carried out by three groups: ERGO, theInternational Livestock Centre for Africa (ILCA), and Resource Inventory andManagement (RIM).The surveys cover a total area of more than 1.5 million square kilometres, most ofwhich was surveyed in both wet and dry seasons. Some 13,000 geographically co-ordinated sample points are included within the database. Survey records oflivestock, human settlement, vegetation and land use are associated with eachsample point. Other attributes include mean annual rainfall, agro-climatic zone andthe presence or absence of tsetse.Data collection, coverage and conten tsThe analysis is based on information from three primary sources:* standard cartography, for mean ann ual rainfall, plan t growing period, soils, andtsetse , used to stratify the survey information;* sys temat ic low- leve l aer ia l surveys , for the direct observation of pastorallivestock, huma n se ttlemen t (rooftops), vegetation and land use ;* c o m p l e m e n t a r y g r o u n d s u r v e y s , for information about village livestockcomposition and local production system s.Low-level aerial surveysAerial surveys are normally conducted from a height of between 400 and 2,000 feetabove the ground. They are designed to collect numerical information about targetpopulations and d istributions over large, often remote, areas w ithin a sho rt period oftime. The survey technique is usually based on a systematic flight pattern whichprovides uniform coverage of an entire region and enables a geographically co-ordinated gridded database to be established. Sample counts are tak en from each gridcell. Key features of this technique are its repeatability, and t h at it does not rely onany previous knowledge of the area concerned. Repeat surveys at critical times of theyear allow seasonal changes in distribution and abundance to be determined. Formore detailed information about the aerial survey techniques u sed, see Appendix 1.Ground surveysThe ground surveys which were used were designed to collect information whichcannot be obtained from aerial reconnaissance, (such as the number of animals keptunder shelter) and to provide data which can be combined with aerial counts to giveestimates of variables that neither technique can address on its own. An example isthe number of poultry kept in rural vil lages: air counts give the number ofhabitations in a sample grid; ground counts provide the number of birds perhabitation . Combining the two figures gives village poultry population estim ates and


13/46

distributions.Ground surveys can also be used to differentiate between the large ruminants thatare kept in association with human settlement, and those that are found away fromvillages. This distinction can be made in one of two ways. Either the aerial surveyobservers m ake a separate record of animals seen w ithin village boundaries, or theyexclude such livestock from the aerial counts altogether. If the second course isadopted, then the estimates of total livestock numbers are calculated by adding thenumbers derived from aerial counts to those derived from the ground survey andaerial habitation records.This integrated air/ground approach provides an objective basis for resourceassessment and a good understanding of local production systems. Since 1986, allsurveys carried out have included a ground element.Sites surveyed and data co llectedThe locations of the surveys carried out are shown in Fig ure 1 below. Details of thedate, area and funding agency for each survey, together with a summary of thecategories of information collected are found Appendix 2.

Figure 1: Location of survey zon es

Areas Surveyed

1: Mali surveys2: Nigersurveys3: Chad surveys4: Sudan Gezira5: Sudan Red Sea6: Nigeria surveys

LivestockAll surveys included in the ERGO data set counted cattle, camels and smallruminants visible from the air. From 1986 onwards, village animals were alsocounted by complementary ground surveys (see above).


14/46

HabitationRural village and p astoralist (i.e. perm anent and temporary) habitation can be easilyassessed and differentiated from the air. In the majority of surveys, each of these twocategories were broken down into a num ber of structu ral types which may reflect theethnicity of the ir builders: for example the Fulan i ruga s of West Africa, the Beja ten tsof Sud an, or th e Tiv hu ts of Nigeria. For the purposes of this analysis, however, onlyvillage habitation has been included in the dataset. This is because it has beenassumed that the pastoralist habitation will always be found where the pastorallivestock are, and therefore will not shed any light on the relationship betweenhuman habitation and livestock numbers.Village dwellings a re considered to represen t either agropastoralists, arable farmers,or the inhabitants of village and towns. Dwellings in larger towns and cities (morethan 5,000 dwellings or covering an area of more than 10 square kilometres) havealso been excluded from the present data.Vegetation and land useLand use estimations in different surveys are less directly consistent than the otherassessments. In the earlier studies, relatively few categories were considered generally cultivation and some measure of grazing. The later surveys recordedvegetation categories that can be amalgamated into seven basic types: land withinth e cu ltivation cycle (active cropping plus fallow), bar e ground, grass land , scrub, openwoodland, dense woodland, and forest. In a number of surveys grass cover was alsoassessed. This differs from grassland in that the scrub and woodland categories, aswell as grassland, support grass. Where grass cover was not measured specifically, anindex has been calculated and tested in areas where grass cover was specificallyassessed in order to confirm its relevance an d accuracy.Accuracy of the data are they reliable?Though the vege ta t ion type was usua l ly a s sessed v i sua l ly , r a the r thanphotographically, comparisons of vegetation types which are constant from season toseason were assessed consistently in repeat surveys, suggesting that the methodyields accurate results. For example, in the Bahr el Ghazal Region of Chad, woodyvegetation was assessed in 1991 at 14.3 per cent and in 1993 at 13.6 per cent. InGongola State, Nigeria, grassland w as estimated in 1983 and 1984 at 14 per cent a nd17 per cent respectively, and in Niger State, Nigeria, cultivation was estimated in1989 and 1990 as 20 and 21 per cent respectively. These comparisons providereasonable grounds for assuming th at vegetation and land-use d ata are sufficientlyprecise to be meaningful.Livestock and habitation estimates are also made visually, but corrected for observererror by comparing visual estimates with counts from photographs. A *bias' is thencalculated, and the correction factor applied to all observer counts over 10. Calculatedobserver biases are usually less than 5 per cent, and commonly between 1 per centand 2 per cent. Repeat surveys of perma nent h abitation types, which should notchange much from year to year, show surprisingly little variation. For example, in th eBahr el Ghazal region in Chad the 1991 and 1993 estimates for permanenthabitations were 79,700 and 82,800 respectively, and for Niger State, Nigeria for 1989and 1990 were 17.5 and 17.8 pe r squa re kilometre respectively.


15/46

Data processing and analysisAs can be seen from the table of surveys in Appendix 2, a number of areas weresurveyed during both dry and wet seasons, and some were surveyed during aparticular season in more than one year. When examining information from severalsites and times, two distinct analytical approaches are possible. One can either lookat each individual survey dataset, draw conclusions for each one, and then build amore general picture; or one can try first to establish a general picture, and theninvestigate the degree to which the results from specific areas confirm the overallpatterns.It is the second approach which has been used here. This is based on the assumptionth at any relationships w hich are strong enough be statistically significant across thewide range of land types and ecological conditions that have been surveyed are offundamental importance to livestock management. For this preliminary analysis,data from duplicate surveys have been averaged to give a single figure for eachvariable for each grid. This intentionally smooths out any seasonal or annualvariation , so as to provide an in itial overview of the da ta. In due course, it is plannedthat the data be disaggregated, and the effects of seasonal factors investigated.Similarly, it was decided that this preliminary analysis should look at livestock as asingle variable, rather than treat individual species separately. This is done byconverting all species of livestock into Tropical Livestock Units (TLU): one TLU isequivalent to one camel, 1.43 cattle, two equines, or ten small ruminants. Thisapproach has two main advantages. First, the results from areas with differentspecies compositions can be compared directly; second, these results can then berelated to carrying capacities (expressed in kilograms per square kilometre) workedou t for different ecosystems part icula rly th ose whe re wild anim als form asub stantia l element of the grazing herbivore biomass.It should be emphasised that both wild animals and horses and donkeys have beenexcluded from these analyses. Wild animals were counted in all the surveys carriedout, but were so infrequently recorded as to be insignificant. Equine numbers wererecorded only in the late r surveys, and so their exclusion allows da ta from the earliersurveys to be included in the analyses. (Even where equines are common theyrepres ent only a small fraction of the total livestock numbers.)

Preliminary findingsRelationships between livestock and environmentThe correlations between livestock density and different environmental factors showsome interesting, and rather startling relationships. For those who are interested inthe statistical mechanics, Appendix 3 provides a description of the terms andprocedures and a worked example. Appendix 4 gives the graphical representation ofeach relationship noted below, with degrees of variation being indicated by the bars.The important thing for non-statisticians to retain is that throughout the analysis,and for each of the relationships examined, there is considerably less than a one inten thou sand chance th at they are false.


16/46

Data processing and analysisAs can be seen from the table of surveys in Appendix 2, a number of areas weresurveyed during both dry and wet seasons, and some were surveyed during aparticular season in more than one year. When examining information from severalsites and times, two distinct analytical approaches are possible. One can either lookat each individual survey dataset, draw conclusions for each one, and then build amore general picture; or one can try first to establish a general picture, and theninvestigate the degree to which the results from specific areas confirm the overallpatterns.It is the second approach which has been used here. This is based on the assumptionth at any relationships w hich are strong enough be statistically significant across thewide range of land types and ecological conditions that have been surveyed are offundamental importance to livestock management. For this preliminary analysis,data from duplicate surveys have been averaged to give a single figure for eachvariable for each grid. This intentionally smooths out any seasonal or annualvariation , so as to provide an in itial overview of the da ta. In due course, it is plannedthat the data be disaggregated, and the effects of seasonal factors investigated.Similarly, it was decided that this preliminary analysis should look at livestock as asingle variable, rather than treat individual species separately. This is done byconverting all species of livestock into Tropical Livestock Units (TLU): one TLU isequivalent to one camel, 1.43 cattle, two equines, or ten small ruminants. Thisapproach has two main advantages. First, the results from areas with differentspecies compositions can be compared directly; second, these results can then berelated to carrying capacities (expressed in kilograms per square kilometre) workedou t for different ecosystems part icula rly th ose whe re wild anim als form asub stantia l element of the grazing herbivore biomass.It should be emphasised that both wild animals and horses and donkeys have beenexcluded from these analyses. Wild animals were counted in all the surveys carriedout, but were so infrequently recorded as to be insignificant. Equine numbers wererecorded only in the late r surveys, and so their exclusion allows da ta from the earliersurveys to be included in the analyses. (Even where equines are common theyrepres ent only a small fraction of the total livestock numbers.)

Preliminary findingsRelationships between livestock and environmentThe correlations between livestock density and different environmental factors showsome interesting, and rather startling relationships. For those who are interested inthe statistical mechanics, Appendix 3 provides a description of the terms andprocedures and a worked example. Appendix 4 gives the graphical representation ofeach relationship noted below, with degrees of variation being indicated by the bars.The important thing for non-statisticians to retain is that throughout the analysis,and for each of the relationships examined, there is considerably less than a one inten thou sand chance th at they are false.


17/46

The first and strongest association shown by the data indicates that livestockaggregate in the greatest numbers where there is most cultivated land. Not only isthere an even rise in livestock density with cultivation levels, but the degree ofvaria tion is quite low, and the statist ical levels of significance a re aston ishingly high.Not surprisingly, the relationship between animal numbers and the density ofpermanent human settlement (the number of rooftops per square kilometre) is alsovery strong, although, interestingly, it is slightly weaker than the association withcropping levels, and the variation, particularly for the more intensely croppedcategories, is a little greater.Animals numbers are also strongly correlated with rainfall , though not in astraightforward linear fashion as ar e the previous two relationships. The driest ar eas(with between 37.5 and 250mm precipitation per year) support very low livestocklevels an average of 1.28 TLU (320kg) per square kilometre. However, a modestincrease in rainfall (between 250mm and 500mm precipitation per year) is associatedwith a fourteen-fold rise in livestock levels to 18.29 TLU (approximately 4600kg) persquare kilometre. Livestock levels reach a peak of 21 TLU per square kilometre ataround 825mm, and then drop fairly evenly as it gets wetter.It is not only received wisdom, but intuitively reasonable, that domestic livestockshould be concentrated where th ere is an abundance of natura l grazing. But, not onlydo the data not support this hypothesis, they actively suggest that higher livestockdensities are found in a reas with less rangeland. However, although this relationshipis significant, it is weak, and a possible explanation is that livestock numbers are sovariable at specific levels of rangeland that no real trend can be reliably discerned.However, total livestock units do rise consistently with the levels of grass coverwithin all natu ral vegetation categories. More animals are found in areas with moregrass. But th ere is a lot of variation, particularly a t th e h igher levels of grass cover,and accordingly, the relationship is comparatively weak compared to the relationshipbetween livestock and cultivation. Variation in levels of grass cover explains only 10per cent of the variation in livestock density, as compared with the 50 per centexplained by either cultivation or habitation levels.Given that 'total livestock units' incorporates both those animals that are managedpastorally, and those tha t ar e closely associated with human settlement, perhaps it isto be expected th at the p rimary determ inants of livestock d istribution are those mostclosely linked to human activity namely cultivation and settlement levels. If thetraditionally held views are correct, and animals go to where there is grazing,examining these same relationships for the more pastorally m anaged livestock shouldreveal weaker relationships between animal distr ibution patterns and humanactivity.This is indeed the case. Both cultivation and habitation levels are less closelycorrelated with pastoral livestock units than with total livestock units. Nevertheless,the relationship between pastoral animals and cultivation levels is still substantiallythe strongest of those examined. As with total livestock un its, pastoral livestock uni tsrise with cropping intensity, though the increase levels off if more than half the landarea is cultivated.

8


18/46

However, the link between pastoral animal distribution and permanent humanhabitation levels appears to be more complex. Pastoral animal numbers rise to amaximum at habitation densities of around 20 rooftops per square kilometre, thenfall again in the more populous areas. Accordingly, despite its stati stica l significance,this relationship is sufficiently weak to be of little help in predicting pastorallivestock abundance.The natural environmental factors most closely related to pastoral animal numbersare annual rainfall and percentage grass cover. Pastoral animal numbers rise to amaximum at rainfall levels of around 825mm, and then decline to zero at about2500mm. This pattern is similar to that described for total animal numbers, exceptthat pastoral livestock appear to be absent from the very wettest areas, presumablybecause of the threat of vector-borne and other diseases.The distributions of rangeland and pastoral animal numbers are not significantlycorrelated. Given the significant, if weak, relationship between rangeland and totalanimal numbers, this result is perhaps somewhat unexpected as it would seemjustifiable to assum e th at pa storal anim als are more closely linked to rangeland tha nare village livestock. The lack of association between pastoral animal distributionsand rangeland suggests that the association between total animal numbers andrangelan d should be viewed with some reservation.The presence of grass in all natural vegetation categories, as represented by grasscover, is, however, positively correlated to pastoral animal numbers. Where there ismore grass, there are more pastoral animals, though an increase above 50 per centgrass cover appears not to be accompanied by a similar increase in animal densities.Statistically, the relationship is weak, especially when compared to that withcultivation, primarily because the re is a lot of variability in animal densities at bothhigh and low levels of grass cover.It is clear from the results presented above that the distributions of total animalnumbers and of pastoral animals are linked in very similar ways to the extent ofhuman activity and the abundance of natural grazing. Both livestock categories areclosely linked to cultivation rath er th an na tura l grazing.


19/46

Summary of preliminary findings;Relationship (all livestock)1 Livestock aggregate ingreatest num bers wherether e is most cultivated land.2 Livestock aggregate where the reis perman entnum an settlement.3 Animal numbers correlatewith rainfall.4 Fewer livestock are found in areaswith more rangeland.variable th at no5 More animals are found in a reaswith more grass cover.

Relationship (pastoral livestock only)6 P astoral livestock numbers rise withcropping intensity bu t level off oncemore than half the land is cultivated.7 Pas toral livestock num bers rise to amaximum at 20 rooftops per sq.km.,and fall as area becomes more populous.8 Distributions of rangeland andpastoral animals not significantlycorrelated.

9 Pastoral livestock num berspositively correlated to grass cover(presence of grass in all n atu ralvegetation categories); above 50% grasscover, no further increase inlivestock nu mbers.

Strength of relationshipVery strong.

Not as strong as (1) andgreater variability whereland is more intensely cropped.Not a linear relationship:numbers peak at 825mm.Significant but very weak.

Significant, bu t much les sso than (1) and (2).

Very strong.

Significant, bu t rath er weak.

Much weaker relationshipthan that with cultivation.

Comments

Possible thatnumb ers are soreliable trendcan be found.N.B. This cate-gory is tota lgras s cover andincludes areasof woodland,scrub etc.

Unexpected.Suggests thatassociationbetween totallivestock unitsand rangeland[see (4) above]should be viewedwith reservation.

10


20/46

However, it is possible that the apparent relationships are camouflaging unknowninteractions an d associations between the different variables included in t he analysis.For example, it may be the case that both cultivation and grazing availability areactually determined by rainfall, and that the apparent relationships between humanactivity and livestock levels are merely a by-product of a stronger cause and effectlink between livestock and rainfall.This problem can be addressed by multiple stepwise regression analyses, whichpr ior i t i s e the c o r r e l a t i ons , and t he n c om pe ns a t e for h i t h e r t o u n s e e ninterrelationships between the predictor variables. An i l lustrative example isdescribed in Appendix 5.The results are shown below in Table 1. These strongly confirm the assertions putforward above that human activity, in the form of either cultivation or settlementdensity, is consistently the most important predictor of both total and pastorallivestock distributions. Second most important is rainfall. Measures of na tura lgrazing, either rangeland or grass cover, contribute little to the explanation of animalnumbers.Table 1: Multiple regressions of livestock biomass with environmentalvariables(Figures in Brackets are% variance explained at each step of theregression)DependentVariableMean Total TLUR^.7 112: DF=8,3264

Mean Pastoral TLUR2= 0.4585: DF= 6,4029Proportion PastoralR2 = .3090: DF = 7,2545

Bestpredictor+.3600Hab

(60)

+.3516Cul(34)

-.3470Hab(18)

2n dp red ic tor

+.2997Rain(+2%)

+.4873Rain(+8%)

+.1453Rain(+8%)

3rdpredictor+.2466Cul

(+2%)

+.1396GrLand(+3%)

+.1295GrCov(+2%)

4thp red ic tor

+.1536GrCov(+2%)

+.1189GrCov(+1%)

+.1255Cult(+3%),

OtherSignificantCorrelates-Open,+GrLand,+FoWd,-Dense-Open,+FoWd+GrLand,-Open+FoWd

11


21/46

It is arguable that rainfall is the primary cause of the distribution patterns of thedifferent vegetation types, and so is effectively a duplicate measure of the variabilityin vegetation category levels. Similarly, it could be argued that habitation density is aduplicate measure of cultivation. Accordingly, the analyses were repeated, oncewithout rainfall, and again excluding both rainfall and habitation density (seeAppendix Tables 5.1,5.2).In both sets of repeated analyses, cultivation remained the primary predictor ofanimal density, and grass cover added comparatively l i t t le to the statis t icalsignificance of the re lationsh ips identified.This confirms the overriding importance of the levels of human activity as a predictorof livestock nu mb ers.ImplicationsBoth bivariate and multivariate analyses suggest that livestock numbers are moreclosely associated with the consequences of hum an activity either cultivation orhab itation levels th an they are with any of the environ men tal characteristicsavailable in the ERGO database. In particular, the extent or distribution ofnatural grazing, either expressed as rangeland or as grass cover, is shown tobe of l imited relevance to l ivestock nu mbers.This pattern of relationships also holds true for pastorally managed livestock, andindeed the analyses suggest that for these animals, it is cultivation rather thanhabitation levels that is of primary significance.The strength of the relationships, in statistical term s, is startling given the bread th ofecological conditions, geographical regions, and land-use type from which the data aredrawn, and leaves very little room for doubt that the links are real and not a statisticalartefact Further, the similarity between the trends established for all livestock andpastorally managed livestock suggest that there is no reason to view the latter as aspecial case, affected by factors specific to the pastoralist management system.This conflicts with many currently held views that assume pastoralist animals to bemore tightly linked to natural vegetation than to human activity. Traditionally,pastoralists are supposed to move between areas of natural grazing as the seasonsdictate, and to make use of the resources represented by permanent cultivation onlyat certain times of the year generally towards the end of the d ry season when the reis not much natura l grazing available.This apparent discrepancy between theory and fact can be reconciled by severalpossible explanations, four of which are given below:1 The traditional theories are, and always have been, wrong.This is unlikely to be the case, not least because at some point in the past, humanpopulation and the attendant cultivation levels were substantially lower than theyare at present, and thus were unlikely to be widespread enough to be a significantresource for livestock.

12


22/46

2 The current results are wrong. Because the current analyses are concerned withannual mean livestock levels, and cover a wide range of ecological regions, theyconceal the fact that pastoralism and natural grazing may be linked in some areas butnot others.This idea is also difficult to support, given the statistical evidence. However, it isdistinctly possible that an investigation of seasonal or species-specific factors maybring to light certain conditions or regions where the primary hum an/livestock linksare over-shadowed by other factors.3 The traditional links were, at some time in the past, tenable, but are no longeroperative.There is little doubt that livestock management systems have changed over the lasttwo decades. There have been repeated and extensive droughts in the more aridregions covered by this study, and as a result, livestock ownership patterns havechanged so that many previously nomadic pastoralists no longer own the stock theylook after. Further, there is mounting, if circumstantial, evidence of increasingpasture degradation in the drier parts of the Sahel, either through overgrazing, orthroug h the effects of an expansion of cultivation th at is th e inevitable consequence ofincreasing human populations. Thus it seems likely that the predictability of naturalgrazing as a resource has declined.It is generally accepted that human populations have also expanded into previouslyunpopulated areas, particularly into the wetter areas, thereby clearing the naturalvegetation for cultivation (see above). As a result, regions that were originallysuitable for livestock only during the dry season, when, for example, the risk oftrypanosomiasis was low, have been opened up to livestock in general, and cattle inparticular, throughout the year. This has meant that animals that used to migrate inand out of the wetter areas on a seasonal basis, are now ju st as likely to rem ainperm anently in the m ore southern locations, rath er tha n move north , if only becausethe predictability of either natural grazing or cultivation is greater in the wetterare as tha n i t is in the dryer ones. This has had t he effect of relocating the centre oflivestock distribution southwards, and so away from the arid rangelands in th e no rth;a drift that may have been reinforced by the degradation of the natural dry-landgrazing resources.The southward relocation has had two consequences. First, there has been anincrease in livestock populations within ecological zones that are less seasonallydeficient in natural grazing than are the arid lands to the north. Second, there hasbeen a reduction in the need for long-distance transhumance between wet and dryseason grazing areas. Thus the traditional distinction between management systemsused by pastoralis ts and agropastoralis ts have become increasingly blurred.Pastoralists are settling and growing crops, agropastoralists are acquiring livestockand sending them on transhu man ce, though generally for relatively short distances.This tren d h as led to a closer involvement of stock owners with th e cash economy, andthus an increase in the number of settled pastoralists or, conversely, a decline inthe num ber of pastoralists who are nomadic.

13


23/46

4 The links between livestock and cultivation established by the current analyses aremisleading, and there is an unknown factor related to cultivation that is the causalfactor determining livestock distributions.Given that regression analyses can only identify associations and cannot be used toestablish causal relationships, it is not possible to state with any certainty that it iscultivation pe r se th at attra cts livestock. It is possible, indeed likely, tha t where th ereis cultivation, there will also be patches of grazing or browse that are suitable forlivestock. This is particularly true in areas which are extensively (as opposed tointensively) cultivated with large amounts of marginal fallow or recently cultivatedland. Also, in areas where cultivation is limited to specific areas, such as Wadis, thepresence of cropping and g rass a re likely to coincide.It may therefore be the case that the livestock are not concentrating where thecultivation is , bu t rat he r where the edges of the cultivation are .Whatever the correct explanation for the livestock/cultivation link (and it probablycontains elements of all those suggested above), it seems very likely that extensiverangelands do not support large numbers of l ivestock, and that animals areconcentrated in areas where cropping levels are highest.W hether it is the actual cultivation, or the marg inal resources associated with it t ha tare the attraction is debatable and, in the current context, unresolvable. However, tosome extent, this is irrelevant as far as the development implications a re concerned.The fact is that animals are found in the largest numbers in or near humansettlement, rathe r than in remote rangelands.If human populations continue to increase, as they are almost certain to, andcultivation carries on expanding into areas so-far unpopulated, the livestock/cultivationinterface can only become more important. It also seems likely that as levels ofcultivation rise, it will increasingly come to represent the m ost predictable resource forlivestock. In this way, the need for risk-aversion strategies, such as long-distancetranshum ance to seasonally available patches of natu ral grazing th at are heavily andunreliably dependent on rainfall pa tterns, will be reduced.Accordingly, it seems essential that future development planning should address thisinteraction as a priority. There are many topics of possible relevance in this context,some of which are already on many development agendas. The potential for conflictbetween pastora lists an d arable farmers over, for example, crop trampling or disputedaccess to water sources, has received some attention. However, it seems likely thatthe changes in stock ownership patterns combined with recent clearance of naturalvege ta t ion may c rea te addi t iona l ba r r ie r s to pas tora l i s t / f a rmer re la t ions .Traditionally, pastoralists have negotiated access to crop residues in return formanure and milk, and these arrangements have persisted between clans from year toyear. If many herds are no longer owned by the herders, then it is probable that theindividuals m anaging the flrnn>a1awill change from year to year. If the arable farmeris only recently settled, then the possibility of maintaining permanent arrangementsis reduced, and the po tential for disagreement increased.The concentration of animals in or around extensively cultivated areas may also

14


24/46

contribute significantly to land degradation. Stock movement can trigger gullyerosion, as can be seen on the Mambila highlands of Nigeria, and intensive grazingmay compound th e effects of over-cultivation in m argina l a reas.It will be particularly interesting to establish the specific characteristics of croppedland th at is attra cting the anim als: the lack of trypanosomiasis; the crops themselves;the ancillary grazing and browse available. If the la st of these, could arable farmersbe induced to manage their land to increase these feed resources more efficiently, inorder to promote livestock production?Such issues a re obviously relevant to all areas th at are cultivated, but are likely to beparticularly important in regions where cultivation is limited to small patches: forexample, the riverine (fadama) cropping common in the semi-arid regions; the Wadiand Khor or lakeside cropping in the very dry are as; or in irrigated cultivation. In allsuch systems, the areas of human activity are likely to act as oases, and anyinteraction between farmers and pastoralists will be intensified. This will especiallybe the case if the settled farmers themselves own animals, and their crops attractlivestock from elsewhere.If, as this analysis suggests, there is little distinction between the factors that areassociated with pastoralist and village (i.e. farmers') livestock, then the questionarises as to whether development planne rs should tre at pa storalists as sepa rate fromfarmers. In regions where livestock numbers are high, it seems probable thatmaintaining such distinctions will become increasingly unnecessary, as theundeniable trend appears to be away from pastoralism towards agropastoralism.Adhering to a distinction of decreasing relevance, either through a reluctance toabandon familiar concepts, or perhap s through a di strus t of regional overviews, mayresult in substantial development resources being devoted to denying the inevitable.It would seem preferable to use the limited resources available to ameliorate thediscomfort of those who are most affected by the changes th at are tak ing place.A cautionary note must be added to this discussion. It is easy to conclude fromoverviews such as this that pastoralism is no longer the significant force in livestockmanagement that it once was. This may be the case to a degree. However, justbecause animal densities may be higher in cultivated areas, this is not to say thatthere are no livestock where there is no cultivation. The data presented above showthat the mean livestock density in areas with no cultivation is about 3 TLU persquare kilometre, equivalent to 3 camels, approximately 3 catt le or 30 smallruminants. Since much of the Sahel and sub-Sahelian Africa is uncultivated, thisdensity represe nts very substan tial numbers of livestock, and despite amoun ting to arelatively small fraction of the total (in the region of 10 per cent), must still besignificant enough to war rant attention . These livestock also, of course, represe nt thelivelihoods of large numbers of people.

Recommendations for further studyThe present analyses have deliberately adopted the overview approach as the leastcomplex way of exploring the d ata available. It is suggested th at , if significant result semerge despite the diversity of areas and conditions covered, then the relationships

15


25/46

contribute significantly to land degradation. Stock movement can trigger gullyerosion, as can be seen on the Mambila highlands of Nigeria, and intensive grazingmay compound th e effects of over-cultivation in m argina l a reas.It will be particularly interesting to establish the specific characteristics of croppedland th at is attra cting the anim als: the lack of trypanosomiasis; the crops themselves;the ancillary grazing and browse available. If the la st of these, could arable farmersbe induced to manage their land to increase these feed resources more efficiently, inorder to promote livestock production?Such issues a re obviously relevant to all areas th at are cultivated, but are likely to beparticularly important in regions where cultivation is limited to small patches: forexample, the riverine (fadama) cropping common in the semi-arid regions; the Wadiand Khor or lakeside cropping in the very dry are as; or in irrigated cultivation. In allsuch systems, the areas of human activity are likely to act as oases, and anyinteraction between farmers and pastoralists will be intensified. This will especiallybe the case if the settled farmers themselves own animals, and their crops attractlivestock from elsewhere.If, as this analysis suggests, there is little distinction between the factors that areassociated with pastoralist and village (i.e. farmers') livestock, then the questionarises as to whether development planne rs should tre at pa storalists as sepa rate fromfarmers. In regions where livestock numbers are high, it seems probable thatmaintaining such distinctions will become increasingly unnecessary, as theundeniable trend appears to be away from pastoralism towards agropastoralism.Adhering to a distinction of decreasing relevance, either through a reluctance toabandon familiar concepts, or perhap s through a di strus t of regional overviews, mayresult in substantial development resources being devoted to denying the inevitable.It would seem preferable to use the limited resources available to ameliorate thediscomfort of those who are most affected by the changes th at are tak ing place.A cautionary note must be added to this discussion. It is easy to conclude fromoverviews such as this that pastoralism is no longer the significant force in livestockmanagement that it once was. This may be the case to a degree. However, justbecause animal densities may be higher in cultivated areas, this is not to say thatthere are no livestock where there is no cultivation. The data presented above showthat the mean livestock density in areas with no cultivation is about 3 TLU persquare kilometre, equivalent to 3 camels, approximately 3 catt le or 30 smallruminants. Since much of the Sahel and sub-Sahelian Africa is uncultivated, thisdensity represe nts very substan tial numbers of livestock, and despite amoun ting to arelatively small fraction of the total (in the region of 10 per cent), must still besignificant enough to war rant attention . These livestock also, of course, represe nt thelivelihoods of large numbers of people.

Recom mendations for further studyThe present analyses have deliberately adopted the overview approach as the leastcomplex way of exploring the d ata available. It is suggested th at , if significant result semerge despite the diversity of areas and conditions covered, then the relationships

15


26/46

m ust be strong enough to apply despite the variability th at is bound to be a feature ofthe data collected.This preliminary study must therefore be viewed as a first step. Having identifiedsome significant tr end s, the analysis can then be extended to find out whethe r thesesam e tre nds ar e equally significant w ithin subsets of the information available. Thiscould be done in several ways:1 The overview approach could be continued, with additional information beingincorporated into the data set. This additional information could either be from otherare as for which comparable da ta are available, such a s Senegal, Zimbabwe, Botswanaor Kenya, where aerial surveys have been in use for some time, or it could describeadditional environmental factors such as soil type, drainage, and altitude. It wouldalso be possible to use economic or animal productivity data as variables, instead ofanimal biomass.Another possibility would be to perform some kind of sensitivity analysis on therelationships identified at present. For example, one could look at the likelyconsequences of a 10 per cent increase in human populations on the numbers oflivestock.It should also be possible to identify regions th at a re particularly heavily stocked, andso at risk of being overexploited.2 The relationships identified so far could be examined at a more local level, whichcould be eith er ad min istrative or eco-climatic, to see if thi s gave a similar picture. Forexample, one could look at whether cultivation is as important to livestock in thehumid zone as in the driest areas, or in Chad as well as Sudan; and it might bepossible to see if there were conditions under which some other parameter were thebest predictor of livestock distribution.3 The influence of seas on ally on the distribution of livestock an d it s association withpredictor variables could be examined. This would be of particular inte rest as more ofthe available data could be incorporated into the analyses, specifically those fromsurveys where figures for a single season are available.4 The factors associated with the distribution of the different livestock species couldbe investigated. For example, it is possible, indeed probable, that camels are notassociated with the same environmental factors as small ruminants. Assessing thespecies composition of the livestock in different environmental and socio-economicconditions would be of considerable intere st, and would allow plann ers to identify theconditions un der which particular species predominate.5 The present information could be used as a baseline for comparison with livestockdata from earlier estimates. This would establish trends in population levels overtime, or provide for an assessment of future trend s if suitable dat a become availablein due course. However, this avenue of study is likely to be the most problematicbecause of the difficulty of ensuring compatibility between estimates made bydifferent techniques.

16


27/46

References

BelL R H V (1982) The effect of soil nutrient availability on community structure inAfrican ecosystems'. In: The Ecology of Tropical Savannas (eds. Huntley B J and BH W alker) pp . 193-216, Berlin: Springer.

Bell, R H V (1985) 'Carrying capacity and off take quotas'. In: Conservation andWildlife Management (eds. Bell R H V and E McShane Caluzi) Washington: USPeace Corps.Be hn ke, R H, I Scoone s and C Kerven (1993) (eds) Range Ecology atDisequilibrium: new models of natural variability and pastoral adaptation inAfrican savannas, Overseas Development Institute, International Institute forEnvironm ent an d Development and Commonwealth S ecretariat.Bourn D M (1978) 'Cattle, rainfall and t set se in Africa', J. Arid Envir. 1:49-61.Bou rn D M and W int G R W (1992) 'Nigerian National Livestock Resource Survey:The IT Implication in Implementation and in Future Use'. Paper presented to the1992 Summer Colloquium of the BCS Developing Countries Specialist Group,London, England.Coe, M J, D H di m m ing and J Ph illipson (1976) 'Biomass and production oflarge African herbivores in relation to rainfall and primary production', Oecologia22, 341-354.Cole, R (1989) (ed) Measuring Drought and Drough t Impacts in Red Sea Province,Oxfam Research Paper 2, Oxford: Oxfam.Environmental Research Group Oxford Ltd (ERGO) (1987) 'Photo-interpreta-tion and Data Processing of Integrated Low Level Aerial Photography andRadiometry'. Report on Special Services Agreement for the Food and AgricultureOrganisation (FAO) of the U nited N ations, Rome, Italy.Environmental Research Group Oxford Ltd (ERGO) (1990) 'Assessment of theFederal Government of Nigeria's Renewable Resource Monitoring Capabilities'.Report to the African Ministerial Conference on the Environment, United NationsEnvironm ent Programme (UNEP), Nairobi, Kenya.Environmental Research Group Oxford Ltd (ERGO) (1990) ' IntegratedLivestock Su rveys of Red Sea Province, Sudan'. Report to Oxfam, Oxford, Eng land .Fritz, H and P Dun can (1993) 'Large herbivores in rangelands', Nature 364,292-3.Ma rriott F C and W int G R W (1985) 'Sampling and Statistics in Low Level AerialSurvey'. Report prepared by Resource Inventory and Management Ltd, for theInte rnat ional Livestock C entre for Africa (ILCA/CIPEA), Addis Ababa, Eth iopia .

17


28/46

M illigan K R N, Ajayi S S an d Afolayan T A (1978) 'Aerial Surveys in the Study ofAnimal Populations and Range Conditions ' . Paper given at the 4th AnnualConference of the N igerian Society of Animal Production , Ife, Nigeria.Milligan K R N, Bourn D M and C hachu R (1979) Dry and Wet Season Patterns

of Cattle a nd Land U se in Four Regions of the N igerian Subhum id Zone'. Report tothe International Livestock Centre for Africa (ILCA/CIPEA), Subhumid ZoneProgramme, Kadun a, Nigeria.Milligan K R N, Bourn D M and Chachu R (1979) 'Ground Reconnaissance andProposals for Aerial Survey of Potential International Livestock Centre for AfricaCase Study Area ' . Report to the International Livestock Centre for Africa(ILCA/CIPEA), Subhumid Zone Programme, Kaduna, Nigeria.Milligan K R N, Keita M and de Leeu w P (1982) 'Recensement A 6rien Saisonnierdu C heptel et Types de Paysage du D elta Central du Niger au Mali'. Rapport F inal.

Internationa l Livestock Centre for Africa (ILCA/CIPEA), Kad una, N igeria.Milligan K R N, Synge B A, Sule B and von Kaufmann R R (1980) 'TheAbundance and Distribution of Cattle on the Jos Plateau, Nigeria'. Report to theFederal Livestock Department, Lagos, Nigeria.Milligan K R N, Wint G R W, Bourn D M and Rayson F (1984) 'Dry SeasonAerial Surveys Over Eigh t Regions of the Nigerian Subhum id Zone, March - April1984', Report to the International Livestock Centre for Africa (ILCA/CIPEA),Subhum id Zone Programme, Kaduna, Nigeria.Milligan K R N (1980) 'The Abundance and Distribution of Cattle on the JosPla teau , N ige r ia ' . Repor t by In te rna t iona l L ives tock Cent re for Af r ica(ILCA/CIPEA), Kaduna, Nigeria to the Federal Livestock Department (FLD),Lagos, Nigeria.M illigan K R N (1981) The Distribution of People, Livestock and Ecological Systemsin the Central Region of the Pastoral Zone of Niger'. Report to the United StatesAgency for International Development (USADD), Niamey, Niger.M i l l i gan K R N ( 1981 ) 'The Use of Aerial Survey for Mapping LivestockDistribution'. Internal Report prepared for the International Livestock Centre forAfrica (ILCA/CIPEA), Addis Ababa, Ethiopia.Milligan K R N (1982) Dry Season Distribution Patterns of Livestock and HumanPopulation and Environmental Conditions in a Pilot Intervention Area near TchinTabaradene, Niger'. Preliminary Report, submitted to United States Agency forInternational Development (USAID), Niamey, Niger.M illigan K R N (1982) 'Recensement Agrien des Populations H um aines et Animateset des Conditions Mgsologiques d'une Region du Centre de la Zone Pastorale duNiger'. International Livestock Centre for Africa (ILCA/CIPEA), Kaduna, Nigeria.M illigan K R N (1982) Hecensement Aerien des Populations Humaines et Animates

et Morphologies du Paysage pendant la Saison des Pluies d'une Region du Centre

18


29/46

de la Zone Pastorale du Niger' (English and French versions). InternationalLivestock Centre for Africa (ILCA/CIPEA), Kaduna, Nigeria.Mil l igan K R N (1983) An Aerial Reconnaissance of Livestock and Hum anPopulation in Relation to Land Use and Ecological Conditions in the SORD U

Project Area of Southern E thiopia, JEPSS Research Report no. 5, Addis Ababa,Ethiopia.Milligan K R N (1983) Dry Season Patterns of Livestock and H uman Distribution inthe Gourma Region of Mali, Addis Ababa, Ethiopia: In ternational Livestock Centrefor Africa (ILCA/CIPEA).Milligan K R N (1983) Dry Season Patterns of Livestock and Hum an Distribution inthe Office du Niger Region of Mali, Addis Ababa, Ethiopia: Interna tional LivestockCe ntre for Africa (ILCA/CIPEA).Milligan K R N and de Leeuw P (1983) 'Low-altitude aerial surveys in pastoralsystems research', International Livestock Centre for Africa (ILCA/CIPE A) Bulletin16, Addis Ababa, Eth iopia: ILCA/CIPEA.Milligan K R N and Keita M (1981) 'Livestock and Ecological Conditions in theMali Delta Region'. Report to the International Livestock Centre for Africa(ILCA/CIPEA) / L'Operation de Developpement de l'Elevage dans la Region duMopti (ODEM), Bamako, Mali.Milligan K R N and Keita M (1982) Etudes Airiennes CompUmentaires du Cheptelet Paysage dans Quatre Zones Pastorales de la Region du D elta Central du N iger au

Mali, Kad una , Nigeria: Interna tional Livestock Cen tre for Africa (ILCA/CIPEA).Oesterheld M, O E Sala and S J McN aughton (1992) "Effect of animal husbandryon herbivore carrying capacity a t a regional scale', Nature 356, pp.234-236.Reso urce In ventory a nd M anagement (RIM) (1983) livestock and Land Use inSouthern Gongola S tate'. Preliminary Wet Season R eport to the N ational LivestockProject Unit (NLPU), Kaduna, Nigeria.Resource Inventory and Management (RIM) (1983) 'Livestock Producers ofSouthern Gongola State'. Final report, Wet Season Groundwork, to the NationalLivestock Project Un it (NLPU), Kaduna, N igeria.Re source Inventory a nd M anagement (RIM) (1983) Treliminary Assessment ofthe Distribution of Cultivation and Sample Estimates of Cropped Land in BauchiSta te D uring the 1983 Growing Season'.Resource Inventory and Management (RIM) (1984) 'Cropping and Land UsePat te rns in Azare Region, Bauchi Sta te , Niger ia ' . Repor t to the Federa lDep artm ent of Rural D evelopment, Federal M inistry of Ag riculture, W aterResources and Rural Development, Nigeria.

19


30/46

Resource Inventory and Management (RIM) (1984) *Dry Season Aerial Surveyswithin the Nigerian Subhumid Zone'. Report to the International Livestock Centrefor Africa (ILCA/CIPEA), Addis Ababa, Ethiopia.Resource Inventory and Management (RIM) (1984) Dry Season Populations of

Cattle and Pastoralism in Gongola, Nigeria'. Report to the National LivestockProject Unit (NL PU), Kaduna, N igeria.Resource Inventory and Management (RIM) (1984) Tanning and Sett lementPatterns in Bauchi State , Nigeria ' . Report to the Bauchi State AgriculturalDevelopment Program me (BSADP), Bauchi, and to the W orld B ank.Resou rce In ventory an d Managem ent (RIM) (1984) livestock and Land Use inSouthern Gongola State, Nigeria'. Report to the National Livestock Project Unit(NLPU), Kad una, Nigeria.Resource Inventory and Management (RIM) (1984) 'Livestock Producers of theSubhumid Zone' . Report to the International Livestock Centre for Africa(ILCA/CIPEA), Subhumid Zone Programme, Kaduna, Nigeria.Resource Inventory and Management (RIM) (1985) 'A Review of Aerial SurveyFindings in West Africa'. Report to the International livestock Centre for Africa(ILCA/CIPEA), Addis Ababa, E thiopia.Reso urce In ventor y and Ma nagement (RIM) (1985) 'Conditions Pastorales dansle Niger Central suite la Secheresse de 1983/84'. Report to the InternationalLivestock Centre for Africa (ILCA/CIPEA), Addis Ababa, Ethiopia.Resource Inventory and Management (RIM) (1985) ' livestock, Land Use andHuman Hab itation in Three Nigerian G razing Reserves; Sorau, Gark ida and WawaZangi'. Report to the National livestoc k Project Unit (NLPU), Kaduna, Nigeria.Resource Inventory and Management (RIM) (1985) 'Livestock and HumanHabitation in the Gourma Region of Mali'. Report to the International LivestockCentre for Africa (ILCA/CIPEA), Bamako, Mali.Resource Inventory and Management (RIM) (1985) 'Livestock and HumanHabitation in the Office du Niger Region of Mali'. Report to the Internationallivesto ck Centre for Africa (ILCA/CIPEA), Bamako, Mali.Resource Inventory and Management (RIM) (1985) 'Livestock and Land Use inthe Subhum id Zone of Nigeria; A Synthesis of Ground and Air Surveys Conductedby RIM and ILCA in the Subhumid Zone of Nigeria ' . Dreft report to theInternational livestock Centre for Africa (ILCA/CIPEA), Kaduna, Nigeria.Resource Inventory and Management (RIM) (1985) 'Pastoral Conditions inCentral Niger Following the 1983/84 Drought ' . Report to the Internationallivestock Centre for Africa (ILCA/CIPEA), Addis Ababa, Eth iopia.Resource Inventory and Management (RIM) (1985) 'Populations de Betail et

Habi ta t ion Humaine dans la Region de Gourma en Mali ' . Repor t to theInte rna tional Livestock Cen tre for Africa (ILCA/CIPEA), Bam ako, Mali.

20


31/46

Resou rce Inventory and M anagement (RIM) (1985) 'Sampling and Statistics inLow Level Aerial Survey. Report to the International Livestock Centre for Africa(ILCA/CIPEA), Addis Ababa, Ethiopia.Reso urce Inventory and M anagement (RIM) (1985) 'West African Aerial Survey

Review' . Draf t Repor t to the Interna t ional Lives tock Centre for Afr ica(ILCA/CIPEA), Addis Ababa, Ethiop ia.Reso urce Inventory and M anagement (RIM) (1985) 'Wet Season Aerial Surveysof Cattle, Human Habitation and Cultivation in Selected Regions of the NigerianSubhumid Zone' . Report to the International Livestock Centre for Africa(ILCA/CIPEA), Addis Ababa, Ethiop ia.Resource Inventory and Management (RIM) (1986) 'Aerial Survey ComputerAnalysis. II: Mapping and Analysis Reference Guide' . For the InternationalLivestock Centre for Africa (ILCA/CIPEA), Addis Ababa, Eth iopia .Resource Inventory and Management (RIM) (1986) 'Aerial Survey ComputerAnalysis. II: Mapping and Analysis Reference Guide' . For the InternationalLivestock Centre for Africa (ILCA/CIPEA), Addis Ababa, Ethiopia.Res ou rce I n ven t ory an d M an agem en t ( RI M ) ( 1986 ) 'Data Analysis andStatistical Methods Used in Aerial Survey*. Report to the International LivestockCentre for Africa (ILCA/CIPEA), Addis Ababa, Ethiopia.Resource Inventory and Management (RIM) (1986) ' Integrated Air-GroundSurveys in the Nigerian Subhumid Zone'. Report to the International LivestockCentre for Africa (ILCA/CIPEA), Addis Ababa, Ethiopia.Reso urce Inventory and M anagement (RIM) (1986) 'Low Level Aerial Survey ofthe Gezira Region, Sudan'. Progress report to DEVCO, Eire.R e s o u r c e I n v e n t o r y a n d M a n a g e m e n t ( R I M ) ( 1 9 8 7 ) 'Gezira LivestockIntegration Study (4 volumes) ' . Report by RIM and the Ir ish GovernmentsDevelopment Agency (DEVCO) to the Gezira Rehab ilitation Project, Governm ent ofthe Republic of Sudan .Reso urce Inventory and M anagement (RIM) (1987) 'Refuge in the Sahel'. Reportto the Ministere des Ressources Naturelles et de l'Elevage, Republique du Mali.Resource Inventory and Management (RIM) (1988) 'Cattle, Cultivation andHuman Hab itation in th e Derived Savann a Zone of South W estern Nigeria'. Reportto the International Livestock Centre for Africa (ILCA/CIPEA), Subhumid ZoneProgramme, Ibadan.Resource Inventory and Management (RIM) (1989) livesto ck and Land Use inNiger and Anambra States, Nigeria (2 volumes)'. Report to the Federal LivestockDepartment (FLD), Abuja, N igeria.Resou rce Inventory and M anagement (RIM) (1989) 'Nigerian National Livestock

Census Proposals, Phases I and II'. Report to the Federal Livestock Department(FLD), Abuja, Nigeria.

21


32/46

Resou rce In ventory a nd M anagement (RIM) (1989) Thotographie Aerienne parTransec t dans l a C inquieme Region au Mal i ' . Repor t to rOpgra t ion deD6veloppement de lTSlevage dans la Region de Mopti (ODEM), Mali.Resource Inventory and Management (RIM) (1990) 'Handbook of Nigerian

Livestock Breeds'. Draft report to the Federal Department of Livestock and PestControl Services (FDLPCS), Abuja, Nige ria.Resource Inventory and Management (RIM) (1990) 'Nigerian LivestockResources Survey: Sample Strategy and Initial Findings'. Paper prepared for theNational Conference on the Nigerian Livestock Industry and Prospects for the1990s, Kaduna, 19th -21st November 1990.Reso urce Inventory a nd M anagement (RIM) (1990) 'Nigerian National LivestockCensus, Interim Report*. Report to the Federal Department of Livestock and PestControl Services (FDLPCS) (formerly the Federal Livestock Department (FLD)),New Se cretariat, Abuja, Nigeria.Resource Inventory and Management (RIM) (1991) 'Nigerian LivestockResources (5 volumes)'. Draft Report to the Federal Department of Livestock andPest Control Services (FDLPCS), Abuja, N igeria.Resource Inventory and Management (RIM) (1991) 'Woody Vegetation Cover andWood Volume Assessment in Nor the rn Nige r ia ' . Repor t to the Fores t ryManagement, Evaluation and C oordinating Unit, Federal Government of Nigeria.Resou rce Inve ntory a nd M anagem ent (RIM) (1992) *Low Level Aerial S urveys ofLivestock, Human Habitation and Pastureland in the Bahr el Ghazal Region,

Tchad' (English and French versions). Report to le Ministere de l'Elevage et derHydrauliqu e P astorale, NDjjamena, Tchad.Resource Inventory and Management (RIM) (1992) 'Nigerian LivestockResources (4 volumes)'. Report to the Federal Department of Livestock and PestControl Services (FDLPCS), Abuja, N igeria.Resou rce Inve ntory and M anagemen t (RIM) (1993) 'Low Level Aerial Surveys ofLivestock, Human Habitation and Rangeland in the "Zone d'OrganisationPastorale", Tchad' (English and French versions). Report to le Ministere del'Elevage e t des Ressources Animales, N'Djamena, Tchad.Toulmin C (1993) Com bating Desertification: Setting the Agenda for aGlobalConvention, International Institute for Environment and Development:Issues Paper 1993, IIED.von Kaufmann R and Blench R M (1987) Livestock Systems in Central Nigeria,Addis Ababa, Ethiopia: International livestock Centre for Africa (ILCA/CIPEA).von Kaufmann R and Blench R M (1987) 'The Subhumid Zone of Nigeria'. SystemStudy draft. International Livestock Centre for Africa (ILCA/CIPEA), SubhumidZone Programm e, Kaduna, Nigeria.

22


33/46

Watson, R M (1972) Results of Aerial Livestock Surveys of Kaputei Division,Samburu District and North-eastern Province, Statistics Division. Ministry ofFinance and Planning, Kenya.

23


34/46

Appendix 1: Aerial survey sampling

t117/ H i,\\ "\ w \The aircraft flies in parallel lines over the study area (i), and observers record fromfixed sample bands to each side. The flight lines are divided into equal sectors, tocreate a grid cell lattice, by which each record is located. Only those herds andsettlements which pass through the observation strip s are counted and photographed.The s trip w idth (W) is directly proportional to the flying height above ground (H) andis defined by externally m ounted viewing frames (ii) which are adjusted to delineatea band on the ground (A) which corresponds to the desired sample band width W atthe nominal flying height. Typically, W is 500m at 800 ft agl or 625m at 1,000ft agl.The air crew consists of a co-ordinator/navigator, pilot, and two observers orphotographers.

24


35/46

Appendix 2: Summary of surveys carried out since 1980 andincluded within the ERGO data setCountry Locality Agency* Date Area Grid Livestock1 Habitation* Vegetation

Km 2 Area Past Vill Cover8MALI

NIGER

SUDAN

CHAD

NIGERIA

Delta ILCADelta ILCADelta ILCAGourma ILCAGourma ILCAFifth IBRDRegionNRL USAIDNRL USAIDGezira IBRDGezira IBRDRed Sea OXFAMRed Sea OXFAMBahr IBRDEl GhazalZOP IBRDGongola IBRDStateGongola IBRDStateAnambra IBRDStateNiger IBRDStateNational IBRDNational IBRD

10-8003-8106-8203-S308-8407-87

09-8209-8502-8603-8603-8909-8908-9103-9307-8303-8405-8905-8903-9009-90

3494434944349448164081640102137

8155081550505693796911990011990059800147,60043875438751767559,600923,000923,000

83838382.382.381.7

82.382.356.356.310010010040 02525252540 0400

Ct,Sr,CmCt,Sr,CmCt,Sr,CmCm,Ct,SrCm,Ct,SrAll

Cm,Ct,Sr,Cm,Ct,SrAllAllAllAllAllAllCt,Sh,CmCt,Sh,CmAllAllAllAll

Pa,VilAll

Pa,VilAllAllAllAllAllAllGtGt Pa,AllAllAllAll

Pa,Vil

AllPa, VilPa, VilPa , VilPa, VilPa, VilPa, VilPa, VilVilPa, VilPa, VilPa, VilPa , Vil

Ct,Gr,OwGrAllAll

AllAllAllAllAllAllAllAllAllAllAllAll

1 Cm = Cam els; Ct = Cattle; Sr = Small Ruminants; Sh = Sheep; Gt = Go ats; All = IncludesEquines2 Pa = Pasto ralist; Vil = Village3 Ct = Cultivation; Gr = Grassland/Grass Cover; Ow = Open W oodland;All = Bare Ground, Grassland/Grass Cover; Scrub, Open and Dense Woodland; Forest.* ILCA: International Livestock Centre for Africa.IBRD: Intern ationa l B ank for Reconstruction and DevelopmentUSAID: United S tate s Agency for Inte rnatio nal DevelopmentZOP: Zone d'Organisation Pasto rale. NRL: Niger Rangelands Project

25


36/46

Appendix 3: Regression an d correlation analyses basicprinciples and a work ed exam pleGraphical plotsIn each case, the data have been 'binned ' or grouped into a num ber of categories, andplotted as means for each category, plus or minus two Standard Errors. Thiseliminates the confusion that a scatter plot of several thousand points would cause,and gives a preliminary indication of the variability of the livestock d ata w ithin eachcategory. The calculated regression lines, together with the proportion of varianceexplained, the sample size, and the statistical significance levels are shown at thebottom of each p lot.Linearised transformations of parameters included in the an alysesWhen the livestock biomass data are plotted against environmental or humansettlement variables, the lines produced are not necessarily straight, but may becurved in various ways. As the multivariate analysis techniques assume linearity,incorpora t ing some l inea r and some curv i l inea r r e la t ionships i s l ike ly tounderestimate the relevance of the curvilinear associations. Thus, in order to assessthe statistical relationships between the variables, these curves must first bestraightened or linearised'.To tak e a specific example, consider two param eters th at are intu itively closelyassociated habitation density a nd cultivation levels:Figure 3.1: % cultivation: habitation density

60 k

50*"I4 030

5 2010

10 30 50 70 80 +Habitation Density persq. km.

The plot given above shows th at cropping levels rise with ha bitation density, bu t in anon-linear fashion cultivation in tensity reaches a maximum in those areas wherethe re ar e more than 30 hab itations p er square kilometre. There is some evidence th atwhere habitation levels are highest, less land is cultivated perhaps the cultivationin such a reas tends to be more intensive, as opposed to extensive.

26


37/46

By converting both parameters to their logarithms, the line is straightened, and thestatistical significance of the regression line fitted through the point is muchincreased:Figure 3.2: % cultivation: habitation density logged values

5 ^3 5 3 1 0]68 (T88 TTT3 1^38 1.68 1.88 +Log 10 (Habitation Density)

LoglO (% Cultivation) = 0.3683 + 0.7475 LoglO (Habitation Density). R 2 = 0.478; DF 2, 4402;p 0 . 0 0 0 1Accordingly, all parameters except annual rainfall have been logged. Using thelogarithm of annual rainfall did not improve the degree of linearity, because the basicrelationship is in the form of an upside-down U. This can be straightened by anequation of the form y = a +bx +c/x. The constants a,b, and c were calculated usin g aniterative model provided by the statistical software package used .Bivariate correlationsThe bivariate correlations between the various parameters are shown in Table A2below. The four independent (i.e. predictor) variables most closely related areindicated by the figures in brackets. The significance of each relationship isdeterm ined by the value of the co rrelation coefficient (R).

27


38/46

Table 3.1: Bivariate correlation coefficients(Figures in brackets re presen t rank ed order of correlation)Independent Variable (x)

% C ultivation(Log Annual M ean)%Rangeland(Log Ann ual M ean)% Grass Cover(Log Annual M ean)Permanent Habitation(Log No/km2)Annu al Rainfall (mm)

RainfallLinearised for Total BiomassRainfallLinearised forPastoral Biomass% Forest & Woodland(Log)

% Open W oodland(Log)% Forest(Log)% Dense W oodland(Log)

TotalBiomass(LogTLU/km8)

RNP

I1 h-

a_ i

torl

Mao.caa>

20 .

10 -

5 :4

0 .w.

1 1 i

4i

1 1 1 1 1X> 20 30 40 50 60Mean Percentage Grass Cover 7 0 8 0

Log10 (Y)=0.2883 + 0.3284 Log10(x). R2=0.102; N=4685; P


45/46

Appendix 5: Stepwise multiple regressionContinuing with the example used in Appendix 3, this technique establishes theprimary relationship between cultivation and habitation, and then assesses thecorrelation between cultivation and the next most closely correlated parameter (inthis case linearised rainfall), whilst excluding the habitation/cropping interaction. Atthe next step, the method assesses the relationship between habitation and a thirdvariable, in the absence of the interactions with the first two identified predictorparameters. At each step it calculates the proportion of the variation in croppinglevels accounted for by the predictor variable being added to the analysis. If thisproportion increases markedly between each step, then the importance of the addedpredictor is substantial. If the proportion increases little or not a t all, then t he addedpredictor variable's importance is similarly slight. This process continues until theaddition of further predictor variables fails to increase the statistical validity of theequation generated.The equation produced is given below, and shows that five environmental parametersare significantly related to cultivation, and that together they explain approximatelytwo-thirds of the variation in cultivation levels. However, by far the strongest link isbetween cultivation and habitation density, which explains over half the variation.The add ition of a second predictor variable adds six per cent to the R squared value,and the inclusion of ext ra predictors adds little to the predictive value of the equation .LglO (% C ultivation) = 0.0348 + 0.6757LglO(Habitation Density) + 0.2697 fRainfall') 0.0592 Lgl(X% Forest) 0.1514 lgl0(% Grass Land) + 0.0743 LglO (% GrassCover).R

2=0.621, DF 6,3626; p0.0 001 .Step R2: Cultivation (55); Rainfall (61); % Forest (62); %Grass Land (62); % GrassCover (62)

From this, it can be stated with some assurance that habitation and rainfall are thetwo most im portan t pred ictors of the levels of cultivation.Livestock and environmental parameters repeated multiple regressionsWhen rainfall is removed from the multiple regression analyses, habitation density orper cent cultivation remain the most important predictors of the density of livestockbiomass:

35


46/46

Table 5.1: Multiple regressions of livestock and environmental variables,excluding rainfallDependent Variable Best predictor 2nd Predictor 3rd Predictor 4th Predictor Other

CorrelatesMean Total TLUR*=.6947: DF=5,3267 +.3746Hab(60) +.2924Cult(65) +..2687GrCov(69) -.055 0Open(6 9) +FoWd\-DenseMean Pastoral TLURM .415 5: DF=4,4031 +.4181Cult (34) +.1652GrCov(40) +.1775GrLand (41)+.0658GrCov(42)Proportion Pasto ralR2=..2998:DF=5,2547 -.3390H ab(18) +.1796GrCov(25) +.1439Cult(30) -.0214 Ope n (30) +For

In the absence of habitation density, biomass density is still most closely linked tohuman activity (i.e. cultivation levels), but the proportion of animals which werepastoral becomes primarily associated with grass cover. This is, however, somewhatmisleading as the overall predictive value of the result ing equations was verysubstantially lower than that of the analyses which included habitation levels, to theextent t hat 90 per cent of the variation in the proportions remained unexplained.

Table 5.2: Multiple regression of livestock and environmental variables,excluding rainfall and habitationDependent Variable Best predictor 2nd Predictor 3rd Predictor 4th Predictor OtherSignificantCorrelatesMean Total TLUR2=..6357 :DF4 ,3268 +.5455Cult(59) +.2734GrCov (63K0 4776 Open(63) +.0530For(63)Mean Pastoral TLUR2=0.4155:DF=4,4031 +.4181Cult (34) +.1652GrCov(40) +.1775GrLand (41) +.0658GrCov(42)

Proportion PastoralR^.0981: DF=5,2547 +.1909GrCov(6) -.0729Cult(7)) -.0339Open(9) -.0175 Dense (10) +For

Livestock and Land-Use Surveys in Sub-Saharan Africa

Documents

Transcript of Livestock and Land-Use Surveys in Sub-Saharan Africa