Resistance, resilience and adaptation to climate …...climate change will depend on their capacity...

i

Resistance, resilience and adaptation to climate

change in riparian ecosystems

Helen Amy White

BSc, MSc University of Otago

This thesis is presented for the degree of Doctor of Philosophy of The

University of Western Australia

School of Biological Sciences

Ecology

19 July 2017

ii

Thesis Declaration

I, Helen A. White, certify that:

This thesis has been substantially accomplished during enrolment in the degree.

This thesis does not contain material which has been accepted for the award of any

other degree or diploma in my name, in any university or other tertiary institution.

No part of this work will, in the future, be used in a submission in my name, for any

other degree or diploma in any university or other tertiary institution without the prior

approval of The University of Western Australia and where applicable, any partner

institution responsible for the joint-award of this degree.

This thesis does not contain any material previously published or written by another

person, except where due reference has been made in the text.

The work(s) are not in any way a violation or infringement of any copyright, trademark,

patent, or other rights whatsoever of any person.

The following approvals were obtained prior to commencing the relevant work

described in this thesis:

Access survey sites was granted to H. A. White from private landowners and the

Department of Parks and Wildlife. Access to disease risk areas was granted under

permit DON00243. The collection of floral specimens for identification was enabled

under DPaW permits (SW015930, SW016818, SW017712, CE004258, CE004742

and CE005178).

The work described in this thesis was funded by The Warren Catchments Council,

the UWA School of Biological Sciences, and CSIRO Land & Water.

This thesis contains published work and/or work prepared for publication, some of

which has been co-authored.

Signature:

Date: 22 March 2018

iii

Abstract

Increases in global temperatures and changes in precipitation regimes resulting from

global climate change are placing increasing stresses on ecosystems. In Mediterranean-

type climates and arid biomes in particular, rising minimum and maximum temperatures

coupled with declining precipitation regimes are increasing the frequency and duration of

droughts, heatwaves and fires; events predicted to increase in severity over the coming

decades. While climate effects on individual life history stages can appear small, the

cumulative impact on rates of population turnover might be the difference between

population persistence and decline. It follows, that the persistence of a species under

climate change will depend on their capacity for range expansion and migration with their

shifting climatic niche. Where the migration rate is limited by life history traits, such as

in sessile organisms with long generation times such as trees, or in landscapes where a

species’ optimal climatic niche becomes obsolete, persistence within the geographic

range depends much more on exposure to the changes, their ability to resist changes, and

in the longer term, to adapt and evolve in response to the altered conditions in situ.

In this thesis, I explore the concepts of ecosystem resistance, resilience and

adaptation to climate change in riparian ecosystems. I use the Warren River and its major

tributaries, the Tone River and Murrin Brook, to ‘transect’ a 1200 to 550 mm per annum

rainfall gradient of the Mediterranean-climate zone of southwest Western Australia. In

Chapter 2, I test the hypothesis, that the shallow groundwater and higher humidity of

riparian zones could provide pockets of favourable microclimates, or refugia, for species

as the wider region becomes uninhabitable, aiming to identify the degree to which the

local hydrological regime decouples the riparian assemblages from macroclimatic

drivers. Contrary to expectations, I show that the regional rainfall gradient plays a greater

role in determining the composition of riparian assemblages than any putative ‘buffering

effect’ of the local hydrological gradient, and instead suggests that the riparian

iv

assemblages are at greater risk from climatic changes than anticipated. To gain an

understanding of the responses of riparian communities to future aridification, in Chapter

3, I investigate the effects of streamflow decline on recruitment across the rainfall

gradient. I show that the relative distribution of mature and immature individuals of the

obligate and facultative riparian species are shifting in climate space, and ranges are

contracting at the drier, eastern extent for a number of species. Interestingly, one species,

the major canopy forming species of the riparian zones of the SWWA, Eucalyptus rudis,

demonstrates the widest distribution of any species examined, and it appears unaffected

by the streamflow deficit.

One strategy to increase climate resilience in natural communities is to selectively

harvest and transplant seed from regions that are historically similar in climate space to

the projected future climate of the restoration site (assuming adaptation to local

conditions). In Chapter 4, I investigate the potential for climate-adaptive seed sourcing

using a full reciprocal transplant experiment of the riparian tree, Eucalyptus rudis, to

identify mechanisms underpinning observed trait differentiation in the natural

populations spanning the Warren River Transect. I show that E. rudis responses are

highly plastic when transplanted to drier climates: seedlings sourced from high rainfall

sites were indistinguishable in responses traits from low-rainfall sourced seedlings. Under

wetter conditions, however, we identified conserved growth traits in maternal lineages

sourced from low-rainfall sites. This effect was only detected in individuals transplanted

400 mm pa greater than their source, a shift in climatic space which exceeds current

projections to 2090 for most of the catchment. I synthesise the major findings of the three

research chapters and discuss their implications and limitations. These findings are then

placed in the context of improving management practices, adaptation planning, and

climate-adaptive restoration practices in regions of the world undergoing aridification due

to climate change

v

Table of Contents

1 General Introduction ................................................................................................ 1

1.1.1 Anthropogenic climate change ................................................................ 1

1.1.2 Migration, adaptation or extinction ......................................................... 2

1.1.3 Planning for change ................................................................................. 4

1.1.4 Increasing aridity in Mediterranean-climate regions ............................... 7

1.2 Thesis Outline ..................................................................................................... 9

2 Will riparian zones act as hydrological refugia under climate change? Evidence

from a climosequence of hydrosequences in southwest Australia .................................. 13

2.1 Introduction ...................................................................................................... 13

2.2 Methods ............................................................................................................ 17

2.2.1 Study region .......................................................................................... 17

2.2.2 Vegetation surveys ................................................................................ 22

2.2.3 Spatial determinants of community composition .................................. 24

2.2.4 Environmental determinants of community composition ..................... 29

2.2.5 Statistical analysis ................................................................................. 39

2.3 Results .............................................................................................................. 41

2.3.1 Landform diversity ................................................................................ 41

2.3.2 Floristic diversity ................................................................................... 44

2.3.3 Environmental drivers of community composition ............................... 52

2.4 Discussion ........................................................................................................ 55

2.4.1 Macroclimate as the primary driver of community composition .......... 56

2.4.2 Cascading effect of climate and canopy community on the understorey

species assemblages ............................................................................................... 58

2.4.3 Implications for management under climate change ............................. 60

2.5 Supplementary material .................................................................................... 61

3 Evidence of range shifts in riparian plant assemblages in response to multidecadal

streamflow declines ......................................................................................................... 67

3.1 Introduction ...................................................................................................... 67

3.2 Methods ............................................................................................................ 71

3.2.1 Study system .......................................................................................... 71

3.2.2 Streamflow ............................................................................................ 73

3.2.3 Vegetation ............................................................................................. 75

3.2.4 Forest structure ...................................................................................... 78

3.2.5 Statistical analysis ................................................................................. 82

3.3 Results .............................................................................................................. 83

3.4 Discussion ........................................................................................................ 93

3.5 Supplementary material .................................................................................. 100

4 Does plasticity confer resilience to a drying climate? An experimental test of

genotype by environment interactions along a rapidly changing rainfall gradient ....... 105

4.1 Introduction .................................................................................................... 105

vi

4.2 Methods .......................................................................................................... 111

4.2.1 Study species ....................................................................................... 111

4.2.2 Experimental design ............................................................................ 113

4.2.3 Seed collection and seedling preparation ............................................ 116

4.2.4 Establishment and maintenance of transplant sites ............................. 117

4.2.5 Measured responses ............................................................................. 119

4.2.6 Rationale and methods of statistical analysis ...................................... 121

4.3 Results ............................................................................................................ 125

4.3.1 Effects of source site rainfall and maternal lineage on seed mass and

early growth under glasshouse conditions ........................................................... 125

4.3.2 Trait fixation versus plasticity in transplant sites ................................ 127

4.4 Discussion ...................................................................................................... 147

4.4.1 Trait plasticity as the dominant explanation for phenotype variation . 147

4.4.2 Implications for management .............................................................. 155

4.5 Supplementary Material ................................................................................. 157

5 Synthesis and Conclusions .................................................................................. 161

5.1.1 Riparian flora at risk ............................................................................ 162

5.1.2 Limits to buffering capacity of the river system ................................. 163

5.1.3 Keystone canopy assemblages ............................................................ 164

5.1.4 Increasing resilience via climate adaptive restoration ......................... 165

5.1.5 Conclusions ......................................................................................... 167

6 References ............................................................................................................ 168

vii

Acknowledgements

First and foremost, I am indebted to my supervisors, Raph and John and I cannot thank them

enough for their tireless enthusiasm, patience and good humour throughout the last four years.

This research was supported by an Australian Government Research Training Program (RTP)

Scholarship (Australian Post-Graduate award). The University of Western Australian (UWA)

top-up program, and the CSIRO Land & Water top-up scholarship program (formerly, Climate

Adaptation Flagship) and the generous support of the Warren Catchments Council, via a

Commonwealth Biodiversity Fund.

The supportive staff and board members of the Warren Catchments Council (WCC). In

particular, Mark Sewell, Kathy Dawson, Jenny Carley, Lee Fontanini and Andy Russel. The

WCC in conjunction with DPaW Science, Margaret Byrne, Tara Hopley and John Scott

(CSIRO) established the Warren and Donnelly River’s restoration project, which I was

welcomed in to. The group has been instrumental in this project, I am indebted to you all

sharing your knowledge, your enthusiasm and your passion for the Southern forests.

For granting site access, flora and fauna collection licence and identification services and

assistance I am grateful to the staff at the Department of Parks and Wildlife, Donnelly and

Wheatbelt regions and the WA Herbarium. Special mention to Ian Wilson for site information

& fire warnings and Terry Macfarlane and Michael Hislop for assistance in identification of

plant specimens.

AAM Geospatial Pty Ltd generously provided LiDAR survey at a reduced rate and in kind

support, with special thanks to Brummer Grobbelaar and Jay Thompson.

The communities of Manjimup, Pemberton, Kojonup and surrounds, and the many land owners

who granted site access. Special mention to Bill Bennit & Elaine Steele and to John Young who

granted me access to their private property to plant experimental plots and continue to access

sites, for going on 3.5 years now.

I whole-heartedly thank the many, many people who gave their time to help in the field over the

past four years: Paul Yeoh (CSIRO), Kathryn Bachelor (CSIRO), Lee Fontanini (WCC), Andy

Russel (WCC & Pemberton Hiking and Canoeing), Jenny Middleton, Steve Robinson, Sean

Tomlinson, Dwain Stevenson, Angela Eads, Alice Watt, Carly Wilson, Juliana Pille-Arnold,

Chris Chester, Mitchell Paterson, Julie Futter, Sue Swann, Peter Yeeles, Mark Murphy, Leanda

Mason, Paige Featherstone, Mikela Moretti & Lizzie Wiley. Special mention to Jen & Carly for

their continual willingness to come along and unwavering enthusiasm. For help in the

identification of plant specimens, Mary van Wees and Julie Ellery.

The staff past and present of the School of Biological Sciences, in particular, Rick Roberts and

the Tech-team.

Bruce Webber and the Ecosystem Change Ecology team at CSIRO Land & Water, Floreat have

provided an incredible amount of logistical support. Special mention to Paul Yeoh and Kathryn

Bachelor for their assistance, particularly in the early stages of putting together the translocation

experiment

The UWA ento lab group for many stimulating and thought provoking discussions, weekend

diversions and general banter over coffee.

Finally, to my family and friends, who have supported me always, for following me in to the

field, understanding my absence, and for your words of encouragement in the final few months.

viii

AUTHORSHIP DECLARATION: CO-AUTHORED PUBLICATIONS

This thesis contains work that has been prepared for publication.

Details of the work: Will riparian zones act as hydrological refugia under climate

change? Evidence from a climosequence of hydrosequences in southwest Australia

Location in thesis:

Chapter 2

Student contribution to work:

HAW, JKS and RKD conceived and designed the study. HAW undertook the field

work and collected the data. LiDAR was captured and the raw data was processed by

AAM Geospatial, further analysis by HAW. Raw stream gauge data obtained from the

Department of Water, and analysed by HAW. Statistical analysis HAW with guidance

from RKD. Preparation of the paper HAW, with contributions from RKD and JKS.

Co-author signatures and dates:

Helen A. White Raphael K. Didham John K. Scott

19 July 2017 19 July 2017 19 July 2017

Details of the work: Evidence of range shifts in riparian plant assemblages in response

to multidecadal streamflow declines

Location in thesis:

Chapter 3


HAW, JKS and RKD conceived and designed the study. HAW undertook the field

work and collected the data. LiDAR was captured and the raw data was processed by

AAM Geospatial, further analysis by HAW. Raw stream gauge data obtained from the

Department of Water, and analysed by HAW. Statistical analysis HAW with guidance

from RKD. Preparation of the paper HAW, with contributions from RKD and JKS.


Helen A. White Raphael K. Didham John K. Scott

19 July 2017 19 July 2017 19 July 2017

ix

Details of the work: Does plasticity confer resilience to drying climate? An

experimental test of G×E interactions along a rapidly changing rainfall gradient

Location in thesis:

Chapter 4


HAW, JKS and RKD conceived and designed the study. HAW collected seed,

identified obtained access and permission to planting sites. The seedlings were grown,

transplanted and maintained by HAW. Data collected by HAW. Statistical analysis

undertaken by HAW with guidance from RKD. Preparation of the paper HAW, with

contributions from RKD, JKS and BLW.


Helen A. White Raphael K. Didham John K. Scott Bruce L. Webber

19 July 2017 19 July 2017 19 July 2017 22 March 2018

Student signature:

Date: 19 July 2017

I, Raphael Didham certify that the student statements regarding their contribution to

each of the works listed above are correct

Coordinating supervisor signature:

Date:19 July 2017

Chapter 1: General Introduction

1

1 General Introduction

1.1.1 Anthropogenic climate change

Natural ecosystems are disappearing at unprecedented rates, due to the cumulative direct

and indirect impacts of anthropogenic activities (MEA 2005). Despite the high value that

society places on protecting native biodiversity, human population growth continues to

fuel the spread of urban and agricultural development into natural ecosystems. As a result,

ecosystems are becoming more limited in extent, and increasingly fragmented and

isolated, and truly ‘wild’ places, a rarity. What’s more, the growing threats to our

remaining ecosystems from land-use change are being exacerbated by global climate

change. At the last assessment, global land and ocean surface temperatures had increased

by an average of 0.85°C over the last century, due to increases in the atmospheric

concentrations of greenhouse gasses (IPCC 2014a). With rising mean global temperature,

ice sheets across both Greenland and Antarctica have lost mass, glaciers throughout the

world have retreated, and in conjunction with thermal expansion of the oceans, sea levels

have risen an average of 0.19 m worldwide (IPCC 2014a). Regionally, freshwater

hydrological cycles have been affected by glacial retreat (Peterson et al. 2002, IPCC

2014a), shifts in precipitation form and regime have been observed (Vörösmarty et al.

2000, Hope et al. 2006, Mariotti et al. 2008), and the frequency of heatwaves, droughts

and forest fires are on the rise (Allen et al. 2010, 2015, Veraverbeke et al. 2017), feeding

further carbon dioxide emissions (CO2) into the atmosphere. These changes pose a

significant risk to human populations, threatening vital ecosystem services ranging from

nutrient cycling, soil formation, and pollination services critical to primary production,

through to the capture and storage of freshwater, filtration of wastewaters and climate

regulation (Costanza et al. 1997, 2014, Hennessy et al. 2007, Pecl et al. 2017). In the face

of our rapidly changing climate, natural resource and land managers are beginning to put

in place strategies to reduce the vulnerability of ecosystems (IPCC 2014b), but we must

2

first understand the stresses imposed on the changing natural world and the demands of

society.

1.1.2 Migration, adaptation or extinction

The responses of species and ecosystems to changing climatic stresses are complex and

poorly understood at fine scales. The distribution of a species at any point in time is

determined not only by its phylogeographic history and abiotic environment, but also a

complex array of biotic interactions within its geographic and climatic space (e.g.

McDowell et al. 2011, Benavides et al. 2013, Brown and Vellend 2014, Alexander et al.

2015). With increases in global temperature, mismatches arise between the geographic

ranges of species and their optimal climatic conditions (i.e. their climatic niche, or

climatic envelope; Thuiller et al. 2005). While heatwaves and droughts over summer are

placing trees under undue stress (Allen et al. 2010), the small rises in winter temperatures

too, are reducing the frequency of freezing temperatures required for vernalisation and

triggering spring flowering (Cook et al. 2012) or germination (Mondoni et al. 2012).

Further, unseasonably warm-spells in the Arctic winter result in precipitation falling as

rain, which then encapsulates plants in ice, and damages sensitive flower and shoot buds

(Milner et al. 2016). While climate effects on individual life history stages can appear

small, the cumulative impact on rates of population turnover might be the difference

between population persistence and decline.

The persistence of a species under climate change, then, will depend largely on

their capacity to expand their range and migrate with their shifting climatic niche (Thomas

et al. 2004, Thuiller et al. 2005). Where the potential migration rate is limited by life

history traits, such as in sessile organisms with long generation times (e.g. trees or corals;

Davis and Shaw 2001, Davis et al. 2005, Aitken et al. 2008, Loarie et al. 2009) or in

landscapes where a species’ optimal climatic niche becomes obsolete (e.g. at high

elevations, coastal habitats or range restricted species on islands), persistence within the


3

existing geographic range depends much more on their exposure to the changes as well

as their ability to resist changes, and in the longer term, to adapt and evolve in response

to the altered conditions in situ (Davis and Shaw 2001, Nicotra et al. 2010, Hoffmann and

Sgrò 2011).

Although the possible outcomes for a species are often framed as simply as

migration, adaption or extinction, in reality, species responses to climate change are far

more complex and difficult to predict. The vulnerability of a species depends on the intra-

specific diversity, as much as the diversity of the climate and topography of the range it

inhabits, and also, how these have interacted to enable it to inhabit its current day

distribution. Across a wide-ranging species for example, individuals can express a variety

of phenotypes, such as the tall, single trunk form of low altitude trees in contrast to the

dwarf, gnarled forms of conspecifics at high altitude (Pryor 1956). Expression of a

particular phenotype can be genetically fixed or determined by varying plasticity in

response to environmental cues (Kawecki and Ebert 2004, Leimu and Fischer 2008,

Hereford 2009). The mechanisms driving the phenotypic variation observed throughout

a species range become important when we consider the response of that species to

climate change. In a species with highly plastic phenotypes, the current range of a species

might be a good estimate of its climate envelope, and consequently how much of a shift

in climate it can tolerate. Conversely, if the population is composed of several smaller

locally adapted populations, each population effectively has an optimal climatic envelope,

rendering the species far more vulnerable to small shifts in climate than would otherwise

be predicted based on its current range (Atkins and Travis 2010, Valladares et al. 2014).

While the breadth of a species or populations, climatic niche will largely determine the

rate at which it must migrate and adapt (processes which are not mutually exclusive;

Davis and Shaw 2001); a growing body of evidence suggests that topographic or

4

biological features may have the capacity to moderate the perceived rate of climate

change, enabling persistance of slower adapting species.

Refugia have been defined in this context as topographical or biological features

of the landscape which act to decouple local microclimates from the ambient climate and

thus ‘buffer’ organisms from regional climate shifts (Rull 2009, Dobrowski 2011, Keppel

et al. 2012, 2015, McLaughlin et al. 2017). Across mountainous terrain, valleys become

sinks for cooler air and moisture, for example, Daly et al. (2010) demonstrated that the

valleys in the Oregon Cascades were 6.5°C cooler on average than recorded at the ridges.

Similarly, McLaughlin et al. (2017) reported that humidity is up to 30% higher in the

valleys, than on the ridges in Californian mountains resulting in stark differences in

vegetation types. Furthermore, even across regions of relatively flat terrain, the magnitude

of variability in the microclimates in forested areas can be greater than predicted under

climate change scenarios (Lenoir et al. 2009, 2017). As an apparent consequence, lags

exist between understorey species distributions and geographic displacement expected

based on regional climate shifts (Bertrand et al. 2011, De Frenne et al. 2013). The reduced

exposure to ambient changes, even if it is only in small pockets of the landscape, has been

suggested as a mechanism to ‘buy time’ for slow migrating species as their optimal

climatic range shifts and allow for adaptation, and re-expansion in to the novel climate at

a later stage. For many species, particularly those with hard geographic limits to range

expansion, such as those bounded by oceans (e.g. Gynther et al. 2016) or mountain ranges

(e.g. Williams et al. 2003), failure to adapt will undoubtedly lead to extinction without

human intervention.

1.1.3 Planning for change

There is increasing awareness amongst land managers and conservation practitioners that

traditional approaches to species conservation, such as legislative protection of parcels of

land (i.e. in National Parks and Reserves) or management approaches that only target


5

single factors (e.g. invasive species, habitat loss, or pollution) are not going to be

sufficient to manage future synergistic interactions between multiple threats (Araújo et

al. 2004, Brook et al. 2008). In response, more integrative management and climate

adaptation strategies are being explored (McLachlan et al. 2007, Rout et al. 2013, Stein

et al. 2013, Lavorel et al. 2015, Prober et al. 2015, Aitken and Bemmels 2016). At the

more ‘passive’ end of the spectrum, networks of protected corridors aligned with climate

gradients are being developed to facilitate range expansion and migration through the

landscape (Hannah et al. 2002, Renton et al. 2012; e.g. http://www.gondwanalink.org/).

At the more proactive (and likely riskier) end of the spectrum, approaches such as the

intentional translocation of a species into the space occupied by its shifting optimal

climatic envelope has been suggested (McLachlan et al. 2007). This process, termed

‘assisted migration’ is suggested for species with limited capacity to migrate or adapt

without intervention due to species traits (e.g. poor dispersal, rarity, low fecundity, long

generational times; Hewitt et al. 2011), or where geographic barriers limit range

expansion (e.g. oceans, mountain ranges, mountain peaks).

Although assisted migration offers a potential solution for species faced with

almost certain extinction (Mitchell et al. 2013), there is a great deal of resistance and

scepticism towards the practice. The resistance is largely due to uncertainty in climate

predictions and the direction and magnitude of translocation required, as well as

potentially negative ecological consequences for the receiving environment (e.g. species

may become invasive; Ricciardi and Simberloff 2009, Hewitt et al. 2011, Webber et al.

2011). As an intermediate step, the artificial selection or genetic modification of

genotypes optimised for projected future climates (or ‘assisted gene migration’), has been

suggested as one way to increase the resistance and adaptive potential of populations to

climate change (Prober et al. 2012, Aitken and Whitlock 2013, Aitken and Bemmels

2016). The premise of assisted gene migration is to selectively harvest and transplant seed

http://www.gondwanalink.org/

6

from regions that are historically similar in climate space to the projected future climate

of the receiving site, introducing genotypes which may confer a greater resistance to

drought or high temperatures, and thereby increase climate resilience, but again is not

without risk (e.g. outbreeding depression, introduction of maladaptive traits; Aitken and

Whitlock 2013). This can be seen as just one of a broad range of ways in which the

conceptual, and practical, focus of ecosystem restoration efforts are shifting away from

more traditional restoration targets of recreating the presumed historical conditions at a

site. Instead, shifting to restoration projects angled towards creating functional, resilient

ecological communities which may be better able to withstand, and adapt to, future

climatic changes (Harris et al. 2006, Seavy et al. 2009, Davies 2010, Capon et al. 2013).

As with all of the strategies aimed at increasing resilience towards climate change, one of

the fundamental advances that needs to be made before we can even consider proceeding

with these practices, is certainty in our predictions of species range shifts and climate

driven limitations.

One limitation of current projections of species movements is that they largely

focus on temperatures; predicting latitudinal and altitudinal expansion under warming

(e.g. Thomas et al. 2004, Thuiller et al. 2005, Randin et al. 2009). Increasingly however,

observational data suggest that water availability is more prominent in driving range shifts

in plant distributions, and that the early footprint of global warming is far more complex

than just the simple poleward and elevational shifts predicted by broad scale, niche

models (Lenoir et al. 2010, McLaughlin and Zavaleta 2012, VanDerWal et al. 2013). For

example, in a sample of 86 tree species across the eastern United States, Fei et al. (2017)

demonstrated the anticipated northerly shift, but, also a westerly migration towards wetter

conditions that was 40% larger than the northerly range shift. Similarly, species are

reported to contract downslope (Crimmins et al. 2011, Rapacciuolo et al. 2014) and

around deeper, moister soils (McLaughlin and Zavaleta 2012) throughout California.


7

Moreover, the highest concentrations of drought-induced dieback the Jarrah forests of

southwest Western Australia (Brouwers et al. 2013b) are not being observed at the

warming or drying extent of the species range, as is predicted by niche models, but instead

in discrete patches within the landscape on shallow soils. To move forward on

implementing climate adaptation strategies in ecosystem restoration and management,

there is a pressing need to understand the drivers of range contraction and expansion, how

features of the landscape can increase or reduce exposure, and the adaptive capacity

inherent in the species and ecosystems which we intend to protect.

1.1.4 Increasing aridity in Mediterranean-climate regions

Rainfall by temperature interactions on vegetation communities are likely to be

particularly complex in Mediterranean-climate regions of the world, where the

occurrence and structure of vegetation is largely determined by a pronounced summer

drought. Characterised by mild, wet winters (when the majority of the mean annual

rainfall budget is received) and long, hot, dry summers, Mediterranean-climate regions

are also predicted to be among the most drastically altered by climate changes

(Vörösmarty et al. 2000, Wetherald and Manabe 2002, Giorgi 2006).

There are five major regions defined as having a Mediterranean climate:

California (USA), South Africa, Chile, southern Australia, and the Mediterranean basin

(Underwood et al. 2009). In contrast to much of the rest of the world, where precipitation

models are characterised by a high level of uncertainty, the atmospheric drivers of rainfall

over the Mediterranean-climate regions are relatively predictable (Mariotti et al. 2008)

and the climate projections across models show a high level of consistency in predicting

severe rainfall deficits (Klausmeyer and Shaw 2009, IPCC 2014a). Moreover, despite

covering just 2% of the total land surface area, these regions are estimated to contain up

to 20% of the world’s vascular plants (Cowling et al. 1996, Rundel et al. 2016) and are

classed among Myers et al. (2000) global biodiversity hotspots. The high levels of

8

diversity and endemism of the biome are attributed to the annual wet-winter, dry-summer

cycle, low soil fertility, and importantly, a long, stable evolutionary history (Cowling et

al. 1996, Hopper and Gioia 2004, Petit et al. 2005). By the end of the century, the current

extent of Mediterranean-type climate zones is predicted to contract in area, but overall

expand in extent. By contrast, the climate zones over southern Australia (South Australia

and the southwest Western Australia) are predicted to contract in area by 49-77% as the

inland extent increases in aridity (Klausmeyer and Shaw 2009).

The south-west of Western Australia (SWWA) has experienced one of the most

substantial rainfall declines observed worldwide (Hennessy et al. 2007, Petrone et al.

2010, Silberstein et al. 2012). In the 1970’s, a significant decrease in the frequency and

magnitude of wet weather systems was observed (Hope et al. 2006). The result has been

an average decline in mean annual rainfall of 10 to 16%, culminating in reductions of up

to 50% in surface runoff to rivers and water storage dams (Petrone et al. 2010). Future

climate projections for the region predict further declines in rainfall, and consequently

streamflow [out to 2090 (CSIRO and Bureau of Meteorology 2015) and to 2030 (Barron

et al. 2012, Silberstein et al. 2012) respectively] under all emission scenarios examined.

The climatic gradients in the SWWA are also unusual compared to other regions of the

world, in that there is a predictable, gradual decline in precipitation with increasing

distance from the coast, and an absence of any major confounding altitudinal or

temperature gradients (Anand and Paine 2002, Hopper and Gioia 2004). Moreover, the

rivers of SWWA neatly bisect this precipitation gradient, maximising the potential

contrast between the longitudinal precipitation gradient and local hydrological gradients.

This region thus affords an ideal opportunity for partitioning out the effects of rainfall on

species and ecosystems independent of altitudinal or temperature gradients that frequently

co-occur in other regions of the world.


9

1.2 Thesis Outline

In this thesis, I explore the concepts of ecosystem resistance, resilience and adaptation to

climate change in riparian ecosystem restoration. I use the Warren River and its major

tributaries, the Tone River and Murrin Brook, as a ‘transect’ across the regional rainfall

gradient of the Mediterranean-climate zone of SWWA. As one of the largest, most intact

river catchments in the region, the Warren River system transects an extensive 1210 to

530 mm per annum rainfall gradient through largely native forests and woodlands.

Through systematic vegetation surveys, and a large manipulative field experiment, I build

comprehensive datasets describing the riparian vegetation assemblages, the age-structure

of the most common and characteristic plant species, and examine intraspecific trait

variation within one keystone species, Eucalyptus rudis, along the length of the Warren

Catchment. I employ a series of ‘space-for-time’ substitution study designs across this

climate gradient as a means of testing ecosystem, species and intraspecific responses to a

range of predicted rainfall decline scenarios for the region up to 2090. Using high

resolution digital ground models obtained from an aerial LiDAR survey and stream

gauging data, I estimate the recent flood regime as well as the deficit in streamflow

observed since the 1970’s, and use these to test the major environmental drivers of

community composition and the degree of resistance of riparian communities to multi-

decadal streamflow declines.

In Chapter 2, I examine the importance of the regional climate drivers relative to

the local hydrological regime in explaining compositional shifts in riparian plant

assemblages. It has been widely hypothesised that the shallow groundwater and higher

humidity of riparian zones could provide pockets of favourable microclimates, or refugia,

for species as the wider region becomes uninhabitable. To test this hypothesis, I aim to

identify the degree to which the local hydrological regime can decouple the riparian

assemblages from macroclimatic drivers, thus acting as a hydrological refugium in the

10

face of further regional rainfall declines. I quantify both the canopy and understory

assemblages in over 300, 5 × 10 m plots, stratified along the longitudinal and transverse

gradients of the riparian zone, and examine the relative explanatory power of local

hydrological vs. macroclimatic drivers of vegetation composition using a hierarchical,

multivariate variance partitioning analysis.

Contrary to expectations, in Chapter 2 I show that the regional rainfall gradient

plays a greater role in determining the composition of riparian assemblages than any

putative ‘buffering effect’ of the local hydrological gradient, and instead suggests that the

riparian assemblages are at greater risk from climatic changes than anticipated. Given the

dependency of the riparian assemblages on the regional climate, and the magnitude of

rainfall decline forecast for SWWA, it appears that range shifts and changes in

assemblage structure are inevitable. To gain a more complete understanding of the

responses of riparian communities to future aridification, it is essential then, to examine

the interactive effects of regional climate change and alterations to local hydrological

regimes on individual species responses. In long-lived species such as trees and woody

shrubs, where mature individuals can be relatively resilient to environmental

perturbations, failure to recruit can be an early warning indicator of range contraction or

displacement.

In Chapter 3, I test the effects of streamflow decline on recruitment across a

longitudinal rainfall gradient for 17 species of trees and woody shrubs common to the

riparian zones of the SWWA. I quantify the change in historic vs recent streamflow

conditions using two selected 10-year periods for which data were available: 1980 to 1989

and 2001 to 2010. Although both periods come after the major 1970s ‘step decline’ in

precipitation, and therefore underestimate overall flow reduction, there could have been

a lag period between rainfall change and subsequent ecological impacts of flow reduction,

and therefore I assume that the 1980s period reflects relatively ‘low’ flow reduction, while


11

the 2000s period reflects ‘high’ flow reduction. To test whether the effects of flow

reduction on riparian and non-riparian plant species are exacerbated or buffered by

regional rainfall, I examine the distributions of juvenile and adult plants along gradients

of mean annual rainfall, recent streamflow and the change in streamflow

I show that the relative distribution of mature and immature individuals of the

obligate and facultative riparian species are shifting in climate space. Ranges appear to

be contracting at the drier eastern extent for a number of species. Interestingly, one

species, the major canopy forming species of the riparian zones of the SWWA,

Eucalyptus rudis, demonstrates the widest longitudinal distribution of any species

examined, and with a high proportion of juveniles observed throughout the catchment it

appears unaffected by the streamflow deficit. Across its range however, E. rudis

expresses extreme divergence of phenotype, from a taller, single trunk form with larger

leaves in the wetter extent of the catchment to a shorter, multi-stemmed ‘mallee’ form in

the drier extent, which raises the question of whether morphological traits in E. rudis have

differentiated across the gradient, or are responding more plastically to environmental

cues.

In Chapter 4, I elucidate the mechanisms of phenotypic divergence in E. rudis and

its apparent resistance to climate change by carrying out a fully reciprocal transplant

experiment examining the genotype by environment interaction. I sourced seed from 31

maternal trees at nine sites distributed across the full 1210 to 530 mm annual rainfall

gradient. After a short rearing period in the glasshouse, I transplanted the 1,880 seedlings

out into six, common gardens situated within natural riparian zones representing the

rainfall gradient and habitats typical of the catchment. Over an 18-month period, I

examined the survival, growth and leaf traits of the seedlings to tease apart the genotype

by environment interaction and differentiate fixed from plastic trait responses, and

consequently establish the potential for climate-adjusted seed provenancing.

12

Finally, in Chapter 5, I synthesise the major findings of the three research chapters

and discuss their implications and limitations. These findings are then placed in the

context of improving management practices, adaptation planning, and climate-adaptive

restoration practices in regions of the world undergoing aridification due to climate

change.

13

2 Will riparian zones act as hydrological refugia under climate change?

Evidence from a climosequence of hydrosequences in southwest

Australia

2.1 Introduction

Increases in global temperatures and changes in precipitation regimes resulting from

global climate change are placing increasing stresses on ecosystems. In Mediterranean-

type climates and arid biomes in particular, rising minimum and maximum temperatures

coupled with declining precipitation regimes are increasing the frequency and duration of

droughts, heatwaves and fires; events predicted to increase in severity over the coming

decades (Stocker et al. 2013, CSIRO and Bureau of Meteorology 2015). However, models

forecasting the effects of climate change on biological systems have largely been based

on regional climate trends (e.g. Thomas et al. 2004, Thuiller et al. 2005), and until recently

have ignored the capacity of finer-scale topographic or biological features of the

landscape to buffer local microclimates from changing regional averages (Ashcroft and

Gollan 2013, Ackerly et al. 2015, McCullough et al. 2016). Yet, it is precisely at these

fine scales that establishment, survival to maturity and subsequent reproductive output of

each individual is determined, especially in sessile organisms. There is now compelling

evidence that fine scale environmental heterogeneity may impart greater resilience to

climatic changes than is generally considered in species distribution models based on

regional climate shifts (Dobrowski 2011, Keppel et al. 2012, 2015, Lenoir et al. 2013,

2017, McLaughlin et al. 2017).

The climatic conditions experienced by a species can vary significantly

throughout its range. In addition to gradients in regional means, topographical and

biological features can alter climate at fine scales creating further heterogeneity across

14

the landscape. Moreover, the features driving this variation differ in their vulnerability to

regional climate changes. For example, the effects of rising temperatures on treeline

expansion is exacerbated on equatorial-facing slopes due to greater solar irradiance and

in convex valleys where seedlings are sheltered from cooler winds (Danby and Hik 2007,

Kullman and Öberg 2009). Topographic features may moderate rising temperatures

through coastal fogs (e.g. Ackerly et al. 2015), shading effects of pole-facing slopes (e.g.

McCullough et al. 2016), decoupling of water availability from the regional precipitation

regimes, such as in spring-fed desert pools (e.g. Ransley and Smerdon 2012) or

channelling and concentrating precipitation, such as in outcrops (e.g. Abbott 1984, Schut

et al. 2014). Biological features of the landscape may also serve to moderate climatic

extremes. For instance, the structural attributes of forest canopies are known to reduce

temperature variability, increase humidity and filter light levels for understorey

organisms (Beatty 1984, Whitmore 1989, Ashcroft et al. 2009, Ashcroft and Gollan 2013,

Lenoir et al. 2013). This buffering effect may be a significant contributing factor to the

substantial lags already observed between the distribution of herbaceous understorey

communities and changes in temperature and precipitation over recent decades (e.g.

Bertrand et al. 2011, Ash et al. 2017). The ability of biological features of the landscape

to create microrefugia depends not only on the degree to which they can decouple

microclimate from macroclimate, but also on their sensitivity to regional climate shifts

(Rull 2009, Dobrowski 2011).

The dual buffering effects of topography and vegetation structure are nowhere

more important than in the riparian zones of Mediterranean-climate and arid biomes with

highly seasonal precipitation regimes. The consistency in access to surface water

(Stromberg et al. 2005), the shallower groundwater (Lite et al. 2005, Stella et al. 2013)

and the higher humidity relative to the wider landscape (Ashcroft et al. 2009) have all

flagged riparian systems as important hydrologic and/or thermal refugia for many plants

Chapter 2: Riparian zones as refugia

15

(McLaughlin et al. 2017) as well as animals (Seavy et al. 2009, Seabrook et al. 2014,

Nimmo et al. 2015). Riparian zones are important for two functionally distinct groups of

species; those encroaching from the adjacent uplands to exploit the rich riparian

environment, but limited by the hydrological, erosive or depositional processes imposed

by proximity to the river channel, and the obligate riparian species adapted to these

stresses (Bendix 1994, Luo et al. 2008, Osterkamp and Hupp 2010, Gurnell et al. 2015).

At the same time, variation in macroclimatic (Karrenberg et al. 2003), altitudinal (Lite et

al. 2005, Yang et al. 2011) and geomorphological (erosional/depositional) gradients

(Tabacchi et al. 1998, Gurnell et al. 2015) across the longitudinal axis of the river system

(i.e. from headwaters to mouth) can also drive turnover in riparian assemblages at broader

scales than the local hydrological gradient. Thus, under a drying climate, the ability of

the riparian zone to buffer local precipitation deficits may also be tied to regional climatic

and hydrological or geomorphological changes upstream. While there have been a

number of studies comparing the relative impacts of longitudinal gradients versus local

hydrological (‘transverse’) gradients on the richness and composition of riparian

assemblages (e.g. Bendix 1994, Sagers and Lyon 1997, van Coller et al. 2000, Lite et al.

2005, Petty and Douglas 2010, Yang et al. 2011), none have attempted to determine the

degree to which the riparian zone can decouple the local assemblages from the regional

climate.

The south-west of Western Australia (SWWA) provides an ideal environment to

test the capacity of riparian zones to buffer regional climate change. At the regional scale,

the SWWA is undergoing one of the strongest precipitation declines in the world

(Hennessy et al. 2007), having already experienced a 10 - 16% decline since the 1970’s

(Bates et al. 2008) which has resulted in a three-fold reduction in surface water run-off

(Petrone et al. 2010, Silberstein et al. 2012). Furthermore, the declines in precipitation are

already having marked biological impacts, including crown-mortality (Matusick et al.

16

2012, 2013, Brouwers et al. 2013b) and structural vegetation shifts (Pekin et al. 2009,

Matusick et al. 2016) over wide areas of native woodlands. Climate projections out to

2090 predict further precipitation declines, coupled with increases in temperature likely

resulting in increased frequency of drought, fire and heatwaves (CSIRO and Bureau of

Meteorology 2015). Additionally, precipitation events are predicted to occur as fewer,

more-extreme events ultimately increasing the erosive potential of flood events (Barron

et al. 2012, Silberstein et al. 2012, Leigh et al. 2015). At the catchment scale, the climatic

gradients in the SWWA are also unusual compared to other regions of the world, in that

there is a predictable, graduated decline in precipitation with increasing distance away

from the coast, and an absence of any major confounding altitudinal or temperature

gradients (Figs. 2.1, 2.2, 2.3). Moreover, the rivers of SWWA transect this precipitation

gradient, maximising the potential contrast between the longitudinal precipitation

gradient and the transverse local hydrological gradients, in a region where the flora is still

largely intact (Fig. 2.1). This contrast between the longitudinal precipitation gradient and

the transverse hydrological gradients effectively creates a sequence of hydrologically

similar systems across a gradient of regional precipitation regimes: a ‘space for time’

proxy of an idealised ‘climosequence of hydrosequences’ (cf. Turner et al. 2017).

Already, the region faces rapid and ecologically significant climate changes, placing the

utmost urgency on understanding the underlying capacity of the system to buffer these

changes and promote resistance and resilience to climate changes.

Here, I used a mensurative catchment-scale experiment to investigate the relative

importance of longitudinal and local hydrological gradients of water availability on the

composition of riparian plant communities. I used a series of hierarchical variance

partitioning analyses (Borcard et al. 1992, Cushman and McGarigal 2002), to disentangle

the relative importance of the longitudinal precipitation gradients versus local

hydrological gradients in driving the canopy and understorey vegetation communities in


17

a typical SWWA river system, the Warren River Catchment. I hypothesised that: (1) the

local hydrological regime would buffer the riparian assemblages from the varying

regional precipitation gradient, and thus the regional climate gradients would play a lesser

role in determining vegetation composition than local hydrological gradients. Then to

elucidate the role of structural attributes of the canopy assemblage in buffering

understorey assemblages from the regional climate, I hypothesised that (2) understorey

assemblages would be more responsive to local hydrological conditions and variation in

the microclimatic conditions created by the canopy assemblage, rather than the specific

species composition of the canopy flora. Finally, in undertaking this study, I provide the

first systematic survey of the riparian flora of the Warren River Catchment, and to my

knowledge, of SWWA riparian flora, and thus provide a critical baseline from which to

assess changes across the region in the coming decades.

2.2 Methods

2.2.1 Study region

At just 130 km long in overland distance (275 km river distance), the Warren River

Catchment of SWWA transects a precipitation gradient ranging from less than 520 mm

per annum (mm pa) in the headwaters to over 1200 mm pa at the coast. With a gradual

elevational incline to 385 m (Fig. 2.2), the mean annual temperature is fairly consistent

across the catchment. The annual temperature range increases from 17.4°C to 23.6°C with

distance from the coast, reflecting the declining oceanic influence (Fig. 2.2).

The Warren River and its eastern most tributaries the Tone River and Murrin

Brook (Fig. 2.1) originate on the Darling Scarp and pass through four major upland

vegetation types, Wandoo woodlands, Southern Jarrah forest, Karri forest before

descending on to the Scott Coastal Plain and passing through the coastal heath and

woodlands. The headwaters are within the south-western range of the Wandoo woodlands

18

(Eucalyptus wandoo), which form open canopies with diverse shrub and annual herb

layers (Beard et al. 2013). The Jarrah (Eucalyptus marginata) and Marri (Corymbia

callophylla) forests dominate the landscape in regions over 650 to approximately

900 mm pa (Fig. 2.1, 2.2) and also form an open canopy with a similarly diverse

understorey. The Karri (Eucalyptus diversicolor) forests dominate the high rainfall

regions (>900 mm pa). Forming dense canopies at up to 70 m tall, the Karri forests

support a distinct, and dense understorey community, common species include Karri

Hazel (Trymalium odoratissimum subsp. trifidum) and Karri Oak (Allocasuarina

decussata). At about 30 m asl the river descends off the Darling Plateau on to the Scott

Coastal Plain, where the ancient dune systems are dominated by sclerophyll shrub and

heathlands on the higher ground and lower stature woodlands of Peppermint (Agonis

flexuosa) and Warren River Cedar (Taxandria juniperina) in the valleys.

Native vegetation remains over approximately two thirds of the catchment. The

majority of the vegetation is managed as state forest, supporting a native timber industry,

but large areas are also within reserves and National Parks (Fig. 2.1). The upper catchment

was subject to extensive clearing and conversion to arable land in the 1950s (Burvill 1997,

Smith et al. 2006), the native vegetation represented by scattered paddock trees and small

remnant blocks, often along roadsides and within the riparian zones. The removal of the

woodlands led to a rise in the water table, which drew salts to the soil surface.

Consequently, high soil salinity is now a major issue across the upper catchment,

particularly in low lying sumps and wetlands (Smith et al. 2006).


19

Fig

. 2.1

. The

War

ren

Riv

er C

atch

men

t in

the

south

-wes

t of

Wes

tern

Aust

rali

a. T

he

loca

tions

of

veg

etat

ion

surv

ey tra

nse

cts,

posi

tion

of

the

Dep

artm

ent

of

Wat

er G

aug

e st

atio

n (

DoW

2017)

and t

he

cover

age

of

nat

ive

veg

etat

ion

in

sta

te m

anag

ed r

eser

ves

(D

PaW

2017).

20

Fig. 2.2. The elevational and climatic gradients across the Warren River Catchment: (a)

elevation (Geoscience Australia, http://www.ga.gov.au/), and (b-d) interpolated climate

records for mean annual rainfall, mean annual temperature and annual temperature range

(respectively) for the period 1961-1990 (Hijmans et al. 2005).


21

Fig. 2.3. Temporal variation in climatic conditions and river flow in the lower Warren

River Catchment. Mean (±SE) monthly (a) maximum and minimum temperatures and (b)

rainfall at Pemberton (Fig. 2.1; Station Id. 009592; www.bom.gov.au, accessed Mar.

2017). (c) Mean (±SE) monthly discharge at Barker Road gauging station (Fig.2.1;

Station Id. 607220; http://www.water.wa.gov.au; accessed Nov. 2016)

http://www.bom.gov.au/

http://www.water.wa.gov.au/

22

2.2.2 Vegetation surveys

Riparian vegetation was sampled in a stratified random design from the lower reaches of

the Warren River, through to the upper tributaries of the Tone River and Murrin Brook

(hereafter referred to as the Warren River transect) (Fig. 2.1). The river transect was

stratified into five mean annual rainfall zones, in 200 mm isohyets: ≤ 600 mm, 600-

800 mm, 800-1000 mm, 1000-1200 mm and >1200 mm (Fig. 2.2). Within each zone, 20

potential survey locations, spaced at least 1 km apart and assigned to the true left or right

bank (facing downstream), were randomly generated along the length of the Warren River

transect in ArcGIS 10.3.1 (ESRI Inc. 2016). I determined the logistical feasibility of

sampling at each of these locations, with the goal being to survey 10 sites per zone. Sites

were rejected if: (1) the area was disturbed as a result of human infrastructure such as

roads or bridges; (2) there was evidence of herbicide use in the understorey; (3) the site

had recently been burned; or (4) the site was heavily impacted by grazing from domestic

livestock. To ensure that the vegetation surveys covered a representative range of

geomorphic zones, sites were classified into flood plains or steep banks. Once five sites

of either landform had been selected, all further sites of that landform were rejected.

Accessing sites within the more remote sections of rainfall zone 1000-1200 mm pa was

not possible due to steep granite rock faces and extremely dense forest, therefore only

nine sites were sampled in this zone. Furthermore, since the riparian regions of the >600

mm pa zone were narrow, 11 sites were sampled to increase the replication in quadrats.

The 50 selected sites were surveyed once each, over two consecutive summers, between

December 2013 and April 2014, and November 2014 to May 2015. Surveys were

undertaken over the summer months during periods of low or ceased flow to allow safe

access on foot or via kayak.

Permission to access survey sites was obtained from private landowners and the

Department of Parks and Wildlife (DPaW). Access to disease risk areas was granted


23

under permit DON00243 held by H. A. White. The collection of specimens for

identification was enabled under DPaW permits (SW015930, SW016818, SW017712,

CE004258, CE004742 and CE005178).

Once a site was deemed suitable, a transect was laid perpendicular to the river,

from the water’s edge to the end of the riparian zone. The end of the riparian zone was

visually determined by changes in topology and a shift in dominant vegetation type to

that of the surrounding landscape. Transects ranged in length from 5 m on steep banks up

to 95 m across extensive flood plains and billabongs (oxbow lakes). The coordinates of

the transect origin and end were recorded using a handheld Global Positioning System

(GPS, GPSMAP® 62s, Garmin). To aid in correcting GPS error during data processing,

we took site photos and the compass bearing of each transect. To sample the vegetation,

a row of consecutive 10 m wide by 5 m long quadrats were laid out spanning the length

of the transect, sampling the entire width of the riparian zone (i.e. a 5 m transect was fully

sampled with one quadrat, while a 25 m transect was sampled with five quadrats). All

trees, shrubs and perennial ground cover rooted within the quadrat were recorded.

Although more detailed measures were taken (i.e. cover classes on clonal species, counts

of trees and shrubs, see Chapter 3), only occurrence records were used in this analysis to

incorporate all species into the analysis. I excluded lianas as they do not simply conform

to the understorey or canopy classes used here, and annuals were mostly absent during

the summer season when sampling occurred.

As sampling was carried out during the summer months the majority of specimens

collected for identification were sterile. Where site access allowed, return visits were

made to a few sites during peak flowering in spring (September to November) to collect

flowering specimens to confirm identification. Identifications were confirmed against the

Western Australian Herbarium reference collection by H. A. White and WA herbarium

staff. Where possible specimens were identified to species or subspecies level, but where

24

there was any doubt (i.e. sterile or poor quality specimens) they were lumped at the genus

level. Table S2.1 contains a full species list with the identifications to the highest

taxonomic resolution available; all statistical analysis was undertaken on the reduced data

set. Nomenclature follows that of the Western Australian Herbarium

(https://florabase.dpaw.wa.gov.au).

2.2.3 Spatial determinants of community composition

Accurate spatial quantification of topography and vegetation structure over the length of

the Warren River transect was obtained using an aerial LiDAR (light detecting and

ranging) survey (Figs. 2.4; 2.9; 2.10). A 500 m strip covering the length the Warren River

transect (approximately 130 km2) was carried out from the 13th to the 16th of January 2015

by AAM Geospatial Pty Ltd from a fixed wing aircraft using a Q780 laser system with a

pulse rate frequency of 180 kHz. The laser returns had a horizontal accuracy of 0.55 m

and vertical accuracy of 0.30 m, and were supplied in x:latitude, y:longitude, and

z:elevation m above sea level; ‘point clouds’. The point clouds were classified

algorithmically by AAM into ground and vegetation points, and a 1 × 1 m resolution

digital ground model (DGM) was interpolated from the ground points. All points returned

from infrastructure (e.g. bridges and buildings) were removed from the dataset prior to

analysis.

The error in the GPS coordinates of vegetation quadrats was corrected against the

DGM, vegetation return points and field records, and the geographic coordinates at the

centre of each quadrat were obtained (projected Universal Transverse Mercator [UTM],

southern hemisphere, grid zone 50). Site T28, in the 1000-1200 mm pa rainfall zone was

removed from further analysis as it could not be confidently rectified to the DGM.

To objectively describe the spatial structure across the transects, and account for

non-independence of quadrats within transects, the coordinates of each quadrat centroid

https://florabase.dpaw.wa.gov.au/


25

were used to calculate distance-based Moran’s eigenvector maps (dbMEM) (Borcard and

Legendre 2002, Borcard et al. 2004, Dray et al. 2006). This method calculates a series of

orthogonal eigenvectors using a principal coordinates analysis (PCoA) on a truncated

Euclidean distance matrix on the geographic coordinates of the sampling sites. The

resulting eigenvectors provide a series of orthogonal vectors describing spatial variation

amongst sites from broad to fine scales. Developed as a method to incorporate spatial

processes into the analysis of ecological communities (e.g. spatial autocorrelation), they

can be used identify the proportion of variation of the community that could be spatially

structured independent of environmental variables or induced by environmental patterns.

The dbMEMs were calculated in the R package ‘adespatial’ (Version 0.0-7; Dray

et al. 2016) in R (Version 3.3.2; R Core Team 2016) using the following specifications.

A distance matrix was calculated on mean-centred, geographic coordinates of each

quadrat. As this distance matrix can be represented in two dimensions roughly equating

to the physical layout of the sites (e.g. the easting and northing coordinates, or even three

axes if the scale is large enough to incorporate the curvature of the earth) the matrix was

truncated to modify the relationships among distant sites. Thus, the distance matrix was

truncated by a threshold value 14.047 km, which represents the smallest distance required

to maintain links amongst all quadrats (i.e. the greatest distance between any two

neighbouring transects). This threshold was determined using a single link clustering

analysis on a Euclidian distance matrix generated using the mean centred geographic

coordinates, and extracting the greatest primary link distance. All values above this

threshold were replaced with an arbitrary, large distance constant (four times the

threshold distance) which acts to disconnect distant sites. The zero values along the matrix

diagonal were also replaced with the constant so as to remove the ‘connection’ otherwise

observed within a site and ensure that the neighbouring quadrats are defined as nearest

neighbours (Dray et al. 2006).

26

The PCoA on the truncated distance matrix produced n - 1 orthogonal

eigenvectors (dbMEMs) and associated eigenvalues from the modified distance matrix.

To reduce the number of dbMEMs for analysis, I tested the dbMEM set for spatial auto-

correlation (Moran’s I) using 999 random permutations, and retained only those which

were significantly positively (or negatively) autocorrelated, and therefore represent the

spatial patterns in the layout of the sampling sites from a broad (dbMEM1) to fine

(dbMEM214) scale (Fig. 2.5, Dray et al. 2006).


27

Fig

.2.4

. C

once

ptu

al l

ayout

of

the

step

s an

d p

roce

sses

use

d t

o g

ener

ate

the

envir

onm

enta

l an

d s

pat

ial

vec

tors

use

d i

n t

he

var

iance

par

titi

onin

g

anal

yse

s. T

he

spat

ial

com

ponen

ts i

n l

ight

blu

e, t

he

loca

lly

var

iable

envir

onm

enta

l co

mp

onen

ts i

n b

row

n (

Fore

st s

tru

cture

and h

ydro

logy)

and

com

ponen

ts d

escr

ibin

g l

ongit

udin

al c

lim

ate

gra

die

nts

in

ora

nge.

Air

bo

rne

LiD

AR

surv

ey

Dep

artm

ent

of

Wat

er

Dai

ly s

trea

m f

low

gau

ge

dat

a

Wo

rld

Cli

mat

e

dat

abas

e

Hig

h r

eso

luti

on p

oin

t

clo

ud

: co

mb

ined

veg

etat

ion &

gro

und

retu

rns

Dig

ital

surf

ace

mo

del

max

imu

m p

oin

t hei

ght

(1

m2)

Dig

ital

gro

und

mo

del

(1 m

2)

Gro

und

po

ints

only

Rec

urr

ence

in

terv

al

(ri_

mea

n,

hp

_ra

ng

e):

mea

n a

nd

ran

ge

of

the

pro

bab

ilit

y o

f in

und

atio

n (

at l

east

once

) d

uri

ng

a ca

lend

ar y

ear

Ca

no

py

hei

gh

t (c

py

_m

ax,

cpy

_m

ean

, cp

y_

va

r,

cpy

_cv

):

max

, m

ean,

var

iance

and

co

effi

cien

t o

f var

iati

on o

f ca

no

py

hei

ght

in t

he

quad

rat

area

mo

del

led

in t

he

cano

py d

igit

al

surf

ace

mo

del

Mo

del

dai

ly s

tage

hei

ght

by s

ite

elev

atio

n a

t gau

ge

stat

ions.

Then

, i

nte

rpo

late

dai

ly s

tage

hei

ght

at e

ach

tran

sect

lo

cati

on (

Fig

. 2

.6)

Cal

cula

te h

yd

rop

erio

d &

recu

rren

ce i

nte

rval

s b

ased

on

hei

ght

abo

ve

bas

e fl

ow

PC

A2

: C

oas

tal

infl

uen

ce:

dev

iati

ons

fro

m t

he

pri

mar

y

gra

die

nt

(8%

)

Geo

gra

phic

Co

ord

inat

es o

f

quad

rat

centr

e

(UT

M P

roje

ctio

n)

En

vir

on

men

tal a

nd

Cli

ma

tic

fact

ors

Sp

ati

al v

ecto

rs

15

db

ME

Max

es s

um

mar

isin

g s

pat

ial

var

iati

on a

mo

ngst

quad

rats

Dis

tance

-bas

ed

Mo

ran’s

eigen

vec

tor

map

s (d

bM

EM

)

Ret

ain d

bM

EM

wit

h s

ignif

ican

t

+/-

spat

ial

auto

-

corr

elat

ion

19

Bio

clim

atic

var

iab

les

des

crib

ing t

emp

erat

ure

and

rain

fall

(1

96

1-1

99

0)

PC

A a

nal

ysi

s su

mm

aris

es

dat

a

into

tw

o P

CA

axes

des

crib

ing

94

% o

f th

e var

iati

on

No

rmal

ise

elev

atio

n t

o

met

res

abo

ve

bas

e fl

ow

SO

UR

CE

PR

OC

ES

SIN

GF

AC

TO

RS

To

po

gra

ph

ic r

an

ge

an

d v

ari

an

ce:

the

range

(to

po

_ra

ng

e)

and

var

iance

(to

po

_v

ar)

of

the

elev

atio

n o

bse

rved

wit

hin

each

quad

rat

Fo

rwar

d s

elec

tio

n

of

db

ME

Mo

n

spec

ies

mat

rix

(RD

A)

Geo

gra

phic

co

ord

inat

es:

des

crib

e la

rge-s

cale

lin

ear

stru

cture

of

the

sam

pli

ng d

esig

n

PC

A1

: P

rim

ary r

ainfa

ll a

nd

tem

per

ature

gra

die

nt

acro

ss

catc

hm

ent

(86

%)

Hy

dro

per

iod

(h

p_

mea

n,

hp

_ra

ng

e):

mea

n a

nd

ran

ge

num

ber

of

day

s in

und

ated

annual

ly

Ele

va

tio

n (

ele_

mea

n):

Mea

n e

levat

ion o

f th

e q

uad

rat

abo

ve

sea

level

Pen

etra

tio

n r

ate

:p

en_

cpy

: T

he

per

centa

ge

of

LiD

AR

po

ints

ret

urn

ed b

elo

w 3

m a

s a

per

centa

ge

of

the

tota

l

retu

rns

in a

quad

rat;

pen

_sr

b:

the

per

centa

ge

of

retu

rns

bet

wee

n 0

.5m

and

3m

as

a p

erce

nta

ge

of

the

tota

l re

turn

s

bel

ow

3m

: p

en_

grd

: th

e p

erce

nta

ge

of

retu

rns

bel

ow

0.5

m

as a

per

centa

ge

of

the

tota

l re

turn

s.

No

rmal

ise

po

ints

to g

round

lev

el

28

Fig. 2.5. Distance-based Moran’s eigenvector maps (dbMEM’s). (a) The spatial auto-

correlation of the 214 dbMEM’s, as ranked from broadest to finest scale spatial variation.

Solid circles indicate that the dbMEM represents variation that is significantly spatially

autocorrelated, whereas open circle no significant spatial autocorrelation. A subset of the

significantly spatially correlated dbMEM are illustrated in (b), where graphs depict the

principal coordinate loadings (spatial eigenfunctions) of six representative dbMEMs

against the centred, geographic easting coordinate in metres.


29

2.2.4 Environmental determinants of community composition

To describe the topography of each quadrat, I used the DGM to calculate the mean

absolute elevation (ele_mean; in m asl). The coefficient of variation (topo_var) and range

(topo_range; in metres) of elevation within each 5 10 m quadrat were calculated from

a normalised elevation, adjusted to the lowest point in each transect (i.e. the transect origin

equalled 0 m). The topo_var variable describes topographic heterogeneity, and was

calculated as the variance in elevation within each quadrat, a greater value indicating a

higher diversity of micro-relief within the quadrat. The topo_range variable, calculated

as the difference between the highest and lowest pixel in a quadrat and was used as a

measure of overall steepness of the quadrat, with a higher value indicating steeper slopes

and values closer to 0 indicative of a flat plain.

2.2.4.1 Hydrology

To investigate species distribution patterns in relation to surface water and flood regime,

parameters describing ecologically important aspects of flow were estimated using the

LiDAR-generated DGM and maximum daily stage height data obtained from the West

Australian Department of Water (DoW) for the only four available gauging stations

situated in the main channel along the Warren and Tone Rivers (607220: Barker Road;

607003: Wheatley Farm; 607007: Bullilup; 607027: Hillier Road DoW,

http://water.wa.gov.au/maps-and-data/monitoring; accessed 7th November 2016; Fig. 2.1,

Fig. 2.6). A period of ten years from 1st January 2003 to the 31st December 2012 was

selected for analysis to encompass a range of recent hydrological conditions with

complete records (Fig. 2.6.b).

To re-construct a time series of daily water levels for each of the 49 transect sites,

a linear model (LM) was used in R 3.3.2 (R Core Team 2016) to predict stage height

(water level) as a function of elevation at the four gauge stations, and then interpolate

values for each adjacent transect site for each day of the ten-year period. First, the

http://water.wa.gov.au/maps-and-data/monitoring

30

minimum recorded stage height observed at each gauge station over the ten-year period

was subtracted from all records at the station to give a height (m) above the lowest

recorded water level, hereafter referred to as ‘base flow’ (i.e. normalising water level data

to the lowest summer standing water level or the dry channel at each station). Then a two-

day moving average (for each day plus one day prior) was calculated to account for the

lag in water moving through the catchment. Next, the LM (water level across the four

gauge stations as a function of elevation) was run for each day during the defined ten-

year period (Fig. 2.6.c). From this model, a predicted estimate of daily water level at each

transect could be extracted based on the elevation at the base of each transect (i.e. the

water’s edge) (open circles in Fig. 2.6.c). Finally, this process was repeated for each day

of the ten-year period to produce a predicted time series for each of the 49 transect sites

(Figs. 2.6. d-e). Figure 2.6 summarises, and shows examples of the methods used to

predict water height across the transect locations.

In estimating water levels across the catchment using this method, rather than a

full hydrodynamic model, we made three important assumptions. The first assumption is

that the minimum elevation in a cross section of the river at each transect site is 0 m,

which could be either the lowest elevation in a dry river bed or the elevation of the water

level of a permanent water body. The LiDAR survey was intentionally conducted during

summer when the river had ceased to flow and much of the upper half of the catchment

was dry. All estimates of water height / inundation area are then based on heights above

this 0 m mark (‘base flow’).

Second, because elevational gradients were shallow, we assumed that the water

surface was approximately linear and unimpeded (Brunner 2010). Figure 2.6.a

demonstrates the elevational rise of the catchment against river distance, and shows that

with the exception of the rapid rise from the Scott coastal sand plains on to the Darling

Plateau (~ 25 m asl), the Warren and Tone rivers have a relatively steady incline across


31

the catchment without major natural or artificial dams. To minimise the effects of

differences in water level slope, I initially interpolated water height between each pair of

gauge stations individually. However, estimates generated using this method resulted in

some highly irregular estimations. Data from Bullilup tended to have consistently higher

stage measures than the linear model would predict (e.g. Fig.2.6.c) which is possibly an

indication of water pooling near the gauge. When estimating the heights across gauges

however, this peak resulted in unrealistic estimates at locations both near the gauge station

(e.g. suggested frequent and extended periods of submergence of species known to be

intolerant of flooding) as well as extrapolations of water heights outside the gauge stations

(e.g. flood levels exceeding 5 m in the upper tributaries, or negative values). I therefore

decided to assume ‘average linearity’ across the entire catchment and take the approach

of a single linear model. While this method buffered some of the extreme values (Fig.

2.6. d-e), it successfully provided a metric which differentiated variation in hydroperiod

across a transect which, crucially, is independent of the site position within the catchment

(as opposed to elevation above base flow, or distance to river, which has a range of

possible values which increase in proportion to increasing catchment area above the site).

Using this method, it is important to note that estimations outside of the gauge station

range are extrapolated, and must be treated with greater caution.

The third assumption was that the lag between water level rises in the main

channel and the adjacent billabongs was biologically insignificant. The billabongs all

occurred within the higher rainfall regions, where the water tables rise substantially across

the entire zone over winter and soils become waterlogged regardless of surface water

depth. Further, since the billabongs were flooded for 2 to 3 months a year, the

miscalculation of one or two days from this lag was assumed to be minor with respect to

the stress placed on vegetation inhabiting these habitats.

32

The resulting estimated time series were used to generate ecologically relevant

flow metrics for each elevation point (at 0.1 m increments from 0.5 m above baseflow).

While there are a multitude of metrics describing ecologically relevant aspects of flow,

many are highly correlated. I therefore selected just two parameters to describe the

frequency and duration of inundation. The first, recurrence interval (RI) was calculated

as the probability that the specified elevation point was inundated at least once during any

one calendar year and described the frequency of inundation on an inter-annual scale. The

second, hydroperiod (HP) was a mean of the total number of days annually that the water

level equalled or exceeded the elevation point and represented the duration over which an

elevation point was saturated or completely submerged. The LM estimates of the flood

time series and the calculation of hydroperiods and recurrence intervals were all

performed in R version 3.3.2 (R Core Team 2016). In addition to the continuous values

generated for the main analyses, I also defined the following categorical recurrence

interval classes to assess sampling efficacy: (1) annual: quadrats experiencing a mean

recurrence interval of greater than or equal to 0.9 (i.e. 90% or greater chance of flooding

in any one year); (2) frequent: recurrence intervals greater than or equal to 0.5 but less

than 0.89; (3) uncommon: recurrence intervals greater than 0.00 but less than 0.49; and

(4) rare: plots that were not inundated over the 10-year period as estimated by the

recurrence intervals calculated for the period from January 2003 to December 2012.

Fig. 2.6. The process used to interpolate maximum daily water levels recorded at vegetation

sampling transect sites from four Department of Water gauge stations across the Warren River

catchment. (a) Elevation of gauge stations (crosses) and transects (open circles) by distance from

the river mouth. (b) Normalised daily maximum stage height of the four gauge stations for the

10-year period between 1st January 2003 and 31st December 2012. Grey bars mark the water

level on the three dates shown in (c); linear models of the water height as predicted by elevation

during a period of low flow (1st March 2002), medium flow (4th October 2006) and of high flow

(24th July 2009). Crosses indicate the interpolated and extrapolated water level heights at the

transect sites. Linear models as shown in (c) were calculated each day of the 10-year period to

estimate daily maximum water level at each individual transect site (in black lines), for example

(d) showing a period of high flow and (e) a period of low flow. ►


33

34

2.2.4.2 Climate

I obtained interpolated spatial layers for 19 standard bioclimatic variables describing

annual and seasonal variation in temperature and rainfall at approximately 1 km2

resolution from WorldClim (Table 2.1; Hijmans et al. 2005). As there was a high

correlation among temperature and rainfall variables along the elevational gradient of the

catchment, I used a principal components analysis (PCA) to create composite climate

axes (Table 2.2; Figs. 2.7; 2.8). The PCA was run on normalised data, using Euclidian

distances in the package, ‘vegan’ (Version 2.4-2; Oksanen et al. 2017) in R 3.3.2. The

PC1 gradient described 85.6% of the variation in the data, encompassing the major

temperature and rainfall gradients across the catchment, maximum and minimum

temperatures positively correlated with elevation while annual rainfall negatively

correlated with elevation (Fig. 2.8a). The PC2 gradient described a further 8%,

corresponding primarily to the abnormalities in the gradient which are observed over the

coastal plain to approximately 50 m asl (Fig. 2.8b), where higher mean annual

temperatures are observed closer to the coast and winter rainfall peaks and stabilises

(Fig 2.8).


35

Fig. 2.7. Principal components analysis (PCA) ordination of 19 bioclimatic variables

describing annual and seasonal variation in temperature and rainfall across transect sites

along the Warren River. Arrows indicate the gradient of greatest variation in each

environmental variable, and arrow length is proportional to the strength of the correlation

with PC axes 1 and 2. See Table 2.1 for variable codes.

Table.2.1. Definition of the 19 bioclimatic variables defined by (Hijmans et al. 2005)

Code Definition

bio_1 Mean annual temperature (°C)

bio_2 Mean diurnal range (mean of monthly (max temp - min temp); °C)

bio_3 Isothermality ((bio_2/bio_7)*100; °C)

bio_4 Temperature seasonality (standard deviation*100; °C)

bio_5 Max temperature of warmest month (°C)

bio_6 Min temperature of coldest month (°C)

bio_7 Temperature annual range (bio_5 - bio_6; °C)

bio_8 Mean temperature of wettest quarter (°C)

bio_9 Mean temperature of driest quarter (°C)

bio_10 Mean temperature of warmest quarter (°C)

bio_11 Mean temperature of coldest quarter (°C)

bio_12 Annual precipitation (mm)

bio_13 Precipitation of wettest month (mm)

bio_14 Precipitation of driest month (mm)

bio_15 Precipitation seasonality (coefficient of variation, mm)

bio_16 Precipitation of wettest quarter (mm)

bio_17 Precipitation of driest quarter (mm)

bio_18 Precipitation of warmest quarter (mm)

bio_19 Precipitation of coldest quarter (mm)

36

Fig. 2.8. Variation within the 19 Bioclimatic variables as described by PCA axes (a) PC1

(86%) and (b) PC2 (8%). PCA axes on the x-axis and the climatic variable on the y-axis.

Units are °C for temperatures and mm for rainfall. Abbreviations: temperature: temp.;

quarter: quart.; precipitation: precip. See Table 2.1 for variable definitions and units.


37

2.2.4.3 Forest structure

I measured the structural characteristics of the forest within each plot using LAStools

(Isenburg 2017) to extract vegetation metrics from the LiDAR point cloud. The

vegetation points were first normalised to the ground classified points to generate

vegetation heights above ground (Figs. 2.4; 2.9; 2.10). To describe canopy structure, I

generated a canopy surface model (CSM) from the height normalised point cloud using

the maximum height of all points within each 1 ×1 m pixel across the whole Warren River

transect. The maximum (cpy_max), the mean (cpy_mean) and the variance (cpy_var) of

the CSM were then calculated across the pixels within each of the quadrats. The laser

penetration rate was used to gauge a measure of vegetation densities across vegetation

strata (Leutner et al. 2012) and a proxy of light penetration through the strata (Lovell et

al. 2003). I defined two vegetation strata to separate the shrub from the tree layers,

defining all points between 0.5 m and 3 m as the shrub layer and points 3 m and above as

canopy vegetation. I excluded points within 0.5 m of the ground to allow for any

misclassification of vegetation points and vertical inaccuracy (after Müller et al. 2014).

The penetration through the canopy (pen_cpy) was defined as the percentage of the points

(both ground and vegetation points) below 3 m, divided by total number of points within

the quadrat. Similarly, the shrub layer penetration (pen_srb) was defined as the percentage

of the points returned from below 0.5 m divided by the sum of all the points below 3 m

(i.e. proportional to the number of points that passed through the canopy layer), and

finally, penetration to ground level (pen_grd) was the sum of all points returned from

below 0.5 m divided by the total number of points returned in the quadrat.

38

Fig. 2.9. The stages of LiDAR point cloud processing. All LiDAR points in m above sea

level and normalised in m above ground level (a-b), and the digital ground and canopy

surface models generated from the ground points (excluding vegetation), and the

normalised points respectively (c-d).


39

Fig. 2.10. A cross section through the LiDAR point cloud of transect T19. (a) All

vegetation and ground points coloured by m above sea level and (b) all vegetation and

ground points normalised to ground level and coloured as m above ground level. Note the

main river channel on the far left, and a seasonally inundated billabong devoid of

vegetation (see also Fig 3.2).

2.2.5 Statistical analysis

To assess the adequacy of the sampling effort across the riparian zone, quadrats were

divided into the four flood frequency classes, annual, frequent, uncommon, and rare and

species accumulation curves were calculated for each class using random addition of

quadrats with 999 permutations in package, ‘vegan’ in R 3.3.2.

To first examine the patterns of variation across the entire vegetation community

a transformation-based PCA (tbPCA) was conducted on a Hellinger transformed species

matrix truncated to presence-absence (PA). The Hellinger distance was used as it

accurately preserves distances among sites and has the property of being Euclidean,

making it appropriate for PCAs on species assemblages sampled along gradients

(Legendre and Gallagher 2001, Legendre and Legendre 2012).

A hierarchical variance partitioning analysis on the Hellinger-transformed species

assemblages was performed using redundancy analysis (RDA) and partial-RDA to

40

examine how the vegetation communities varied across space and environmental

gradients (Borcard et al. 1992, Anderson and Gribble 1998, Cushman and McGarigal

2002). At the first tier of the analysis, the variance in the vegetation assemblages was

partitioned between the spatial predictors, environmental predictors, and the variation

explained by both predictor sets i.e. spatially structured environmental drivers. Variation

in space was described by the dbMEM as well as by the mean-centred geographic

coordinates (easting, northing). The coordinates were included to describe any large-scale

linear gradient in species turnover across the study region and remove the necessity to

detrend the species assemblages (Borcard et al. 2004). Environmental variation was

described by the hydrological, climatic and forest structure components summarised in

Figure 2.4. Then, to investigate the relative contribution of the environmental components

in explaining the variation in the assemblages, a second-tier analysis was used to partition

the shared and independent variance among the different classes of environmental drivers.

The variance partitioning analysis was carried out as follows. First, the 16 factors

in the environmental predictor set were correlated against one another to identify and

remove collinearities (Pearson’s r > 0.7, Dormann et al. 2013). Second, the components

of both the spatial and environmental predictor sets were tested for significance in

explaining variation in the species assemblage. A global RDA was run using the

Hellinger-transformed species matrix to generate an adjusted R2 (Ra2; Peres-Neto et al.

2006) for each predictor set. Then, the 𝑅a2 and a significance criterion of α = 0.05,were

applied in the double-stop forward selection procedure (Blanchet et al. 2008) to reduce

the predictor set to those which significantly explained variation in the species

assemblage. Finally, using a series of RDA, partial-RDA and manual calculations to

determine the separate components [see Anderson and Gribble (1998) and Cushman and

McGarigal (2002) for the further details on the methods used to elucidate the different

components], the variation explained by the global model was split amongst the model


41

components, revealing the relative contribution of each predictor set to the patterns in

species assemblages across the Warren River transect. These analyses were carried out

using the packages ‘adespatial’ and ‘vegan’ in R 3.3.2.

To determine whether the canopy species responded to different drivers than the

understorey communities, the variance partitioning procedure was carried out on the

canopy and understorey species separately. Species records were split into canopy

species, defined here as species where the majority of the leaf tissue was above 3 m in

height, versus understorey species, including shrubs, perennial sedges, rushes and

grasses. Furthermore, to investigate the role of the canopy species assemblage (as distinct

from the structural vegetation metrics already included in the environmental predictor

set), canopy species composition was also incorporated as a predictor in the understorey

analysis. The canopy species assemblage was represented by the first 10 eigenvectors of

a Hellinger-transformed PCA analysis which together described 71.33% of variation in

species composition (Table 2.2). I reduced the number of quadrats in both the canopy,

and understorey variance partitioning analyses to include only quadrats where both

understorey and canopy species were recorded (to ensure that a measure of canopy species

composition was recorded for every understorey quadrat analysed).

2.3 Results

2.3.1 Landform diversity

The 49 spatially rectified transects encompassed 311 quadrats (5 × 10 m) along the length

of the Warren River transect and represented a gradient of rainfall and hydrological

conditions (Table 2.3). A total of 103 quadrats were classified as ‘annually’ flooding, 95

as ‘frequently’ flooding, while the less frequently flooded classes ‘uncommon’ and ‘rare’,

were represented by 73 and 40 quadrats respectively. The species accumulation curves

calculated for each of these riparian inundation classes indicate that sampling was

42

adequate across the three higher flood frequency classes (Fig. 2.11). The fourth class,

‘rare’ flood frequency, was less well sampled, as might be expected given that the survey

method was targeted towards riparian vegetation influenced by the river.

Table 2.2. The first 10 eigenvectors describing variation in the canopy species

assemblages of the Warren River transect using a transformation-based principal

components (tbPCA) analysis on Hellinger transformed presence-absence species

records. Bold text marks species best described by each of the axes.

*Exotic species

Species PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9 PC10

Agonis flexuosa -0.58 0.49 0.37 -0.08 0.28 0.10 -0.08 0.09 -0.25 0.06

*Acacia dealbata -0.03 0.01 0.01 0.01 -0.02 -0.04 -0.01 -0.01 -0.02 -0.01

Allocasuarina decussata -0.03 0.00 -0.01 0.03 0.06 0.00 -0.05 0.04 0.13 0.02

Allocasuarina huegeliana 0.00 0.00 0.00 0.00 -0.01 0.00 -0.01 -0.04 0.01 0.02

Banksia grandis -0.01 -0.02 -0.02 0.03 0.02 0.02 0.01 0.02 0.04 0.02

Banksia seminuda 0.04 -0.22 -0.36 0.42 0.58 -0.10 0.29 -0.06 -0.39 0.05

Callistachys lanceolata -0.15 0.07 -0.07 -0.20 -0.13 0.18 0.87 -0.04 0.31 0.03

Chorilaena quercifolia -0.01 0.00 0.00 0.00 0.00 -0.02 -0.01 0.00 0.00 0.00

Corymbia calophylla -0.05 0.00 -0.22 0.32 -0.59 0.35 -0.03 0.39 -0.33 0.23

Eucalyptus diversicolor -0.02 0.01 0.00 0.02 -0.03 -0.04 -0.03 -0.01 -0.03 -0.02

Eucalyptus marginata 0.01 -0.04 -0.09 0.06 -0.10 0.12 -0.12 -0.05 0.08 -0.85

Eucalyptus rudis 0.76 0.52 0.18 0.02 0.08 -0.04 0.07 0.12 0.00 0.09

Eucalyptus wandoo 0.00 -0.03 0.00 -0.01 -0.03 0.02 -0.04 -0.09 0.02 -0.11

Hakea oleifolia -0.06 -0.06 -0.07 0.31 0.31 0.28 -0.21 0.34 0.68 0.16

Melaleuca cuticularis 0.00 -0.01 -0.01 0.00 -0.01 0.00 -0.01 -0.02 0.00 -0.02

Melaleuca rhaphiophylla 0.12 -0.63 0.65 -0.09 0.01 -0.01 0.08 0.18 -0.08 0.09

Melaleuca viminea -0.01 -0.06 -0.10 -0.03 -0.15 0.05 -0.23 -0.71 0.14 0.39

Spyridium globulosum -0.04 0.02 0.03 0.00 0.08 0.11 -0.06 0.01 0.10 0.01

Taxandria juniperina 0.01 -0.10 -0.45 -0.68 0.10 -0.21 -0.16 0.38 -0.02 0.11

Trymalium odoratissimum

subsp. trifidum-0.19 0.08 0.04 0.32 -0.26 -0.82 0.05 0.14 0.22 0.03

*Pinus sp. 0.00 0.00 -0.01 0.01 0.00 0.01 0.01 0.01 0.00 0.01

Eigenvalues 0.21 0.12 0.07 0.06 0.05 0.05 0.04 0.04 0.04 0.02


43

Table 2.3. Sampling effort across gradients of mean annual rainfall and flood recurrence

intervals along the Warren and Tone Rivers. Numbers indicate total number of quadrats

sampled in each rainfall and flooding class. The reduced set of quadrats used in the

variance partitioning analyses in brackets. Annual: quadrats experiencing a mean

recurrence interval of greater than or equal to 0.9 (i.e. 90% or greater chance of flooding

in any one year); Frequent: recurrence intervals greater than or equal to 0.5 to 0.89 (i.e.

greater than 50%, but less than 90%, chance of flooding any one year); Uncommon:

recurrence intervals from 0 to 0.49 and Rare: plots that were not inundated over the 10-

year period as estimated by the recurrence intervals calculated for the period from January

2003 to December 2012.

Rainfall zone

Flood

frequency > 1200 1200 - 1000 1000 - 800 800 - 600 < 600

Annual 10 (6) 19 (13) 18 (10) 21 (12) 35 (12)

Frequent 35 (25) 14 (10) 17 (14) 17 (17) 12 (5)

Uncommon 34 (31) 13 (12) 15 (8) 4 (4) 7 (3)

Rare 8 (8) 15 (12) 7 (6) 7 (7) 3 (0)

Fig. 2.11. Species accumulation curves for quadrats of four inundation frequency classes.

The accumulation curves and error (+/- standard deviation, shaded) were calculated using

random addition of quadrats over 999 permutations.

Transects in the lower catchment were highly contrasting in landform. The

riparian zones ranged from 5 m wide at T08, up to 95 m wide at T13. In these lower river

sections (rainfall >800 mm pa), the main river was generally confined to a narrow

channel, physically separated from the flood plains by a bank, and the plains were formed

instead around billabongs. The riparian zone in these regions therefore had an undulating

44

landform, spanning one or more seasonally inundated pools, with extensive regions of

riparian vegetation and a complex array of hydrological regimes. In contrast, the erosional

zones, the hard granite rock sections and sections along the banks of the long pools tended

to be steep, and with minimal representation by characteristic riparian vegetation.

In the upper catchment (<800 mm pa), transects generally ranged from 5 m to

55 m in length. These narrower riparian zones meant fewer quadrats were sampled in the

upper catchment, but the landforms were less diverse than was observed in the lower

catchment. The majority had gradual elevational rises on both sides of the bank, often

without obvious distinction between erosional and depositional banks. In contrast to the

plains in the lower catchment, the river in the upper catchment floods directly out on to

the flood plains, where it was observed to pool in sumps, and remain saturated for the

duration of winter and near channel riparian vegetarian is likely exposed to high erosional

forces during high flow periods. Transect T97 was an exception to this, at 55 m wide, the

transect crossed wide, flat plains, likely experiencing saturated soils for extended periods

over the wet, winter season forming a wide wetland.

2.3.2 Floristic diversity

A total of 117 species were identified from 51 genera and 27 families (Table S2.1; Fig.

2.12). The success rate of specimen identification to species level was 99.46% in the trees

and shrubs, and 95.8% in the sedges, rushes and colonial shrubs. The most diverse

families were the Myrtaceae, represented by 6 genera and 13 species, Cyperaceae with 7

genera and 13 species and the Fabaceae represented by 4 genera with 13 species (Table

S2.1). The canopy layer was dominated by Myrtaceae in particular, representing 10 of the

13 species of canopy trees (Fig. 2.12). The remaining Myrtaceae, the majority of the

Fabaceae, and the Ericaeae, Proteaceae and Dilleniaceae dominated the highly diverse

shrub layer (Fig. 2.12). The common wetland plant families, Cyperaceae, Restionaceae


45

and Juncaceae were all well represented across the catchment, often observed at the

winter water level along the main channel as well as throughout the flood plains.

The species accumulation curves of the highest flood frequency class (Fig. 2.11)

and the observed patterns across individual species (Fig. 2.13), suggested that the

majority of species richness was at the outer limits of the riparian zone, with

approximately two thirds of the total species recorded in regions with hydroperiods less

than 25 days per annum (Fig. 2.13). Although there were fewer species in the areas

experiencing higher hydroperiods, many of these species were observed over greater

geographical ranges. The canopy species Eucalyptus rudis, Melaleuca rhaphiophylla, and

Banksia seminuda were observed across a range of elevations, rainfall conditions (Fig.

2.14) and generally with sites with longer hydroperiods (Fig. 2.13). Likewise in the

understorey species, the woody shrub Astartea leptophylla had an estimated mean

hydroperiod of over 120 days per annum (Fig. 2.13) with one of the widest elevational

ranges recorded of all shrub species (Fig. 2.14). The commonly observed sedge,

Lepidosperma persecans, and rushes Baumea juncea and Ficinia nodosa also ranged

across the catchment (Fig. 2.14), often in regions with long hydroperiods, though were

also commonly present in the uplands (Fig. 2.13).

Dissimilarity among the riparian assemblages of the Warren River transect largely

varied relative to quadrat flood frequency and rainfall zone, with the first two axes of a

PCA ordination showing clustering in relation to both variables (Fig. 2.15). Higher values

of PC1 and PC2 tended to reflect communities under more predictable, annual or biannual

inundation regimes within the lowest rainfall zones (Fig. 2.15a). Melaleuca

rhaphiophylla largely drove this pattern as it was the most common canopy species in the

upper catchment. The PC1 gradient, and to a lesser extent PC2, described much of the

overall species turnover across the catchment associated with the rainfall gradient (Fig.

2.15a). PC3 was particularly interesting, as it differentiated amongst the sites where the

46

invasive, blackberry, Rubus anglocandicans, dominated the understorey through the

lower half of the catchment (Fig. 2.15b). Lastly, PC4 discriminated structure among some

Fig. 2.12. Number of species per plant family by habit that were recorded across the 311

vegetation quadrats sampled in the riparian zone of the Warren River transect. Note these

values correspond to the lumped taxonomic groups used in the analysis, see Table S2.1

for a species list at the highest taxonomic resolution.

of the widespread upland species (e.g. Macrozamia riedlei, Corymbia calophylla and

Hibbertia species) versus the Juncus species and the sedge, Lepidosperma persecans

(Fig. 2.15b) more commonly observed in wetter soils, although these were not obviously

clustered by either flooding frequency or rainfall zone.


47

The canopy was largely composed of Myrtaceae trees, predominantly E. rudis

which was present at 33 of the 49 sampled transects, as well as M. rhaphiophylla (20

transects), Agonis flexuosa (21 transects) and Corymbia calophylla (16 transects). The

Proteaceous trees Banksia seminuda and Hakea oleifolia were also commonly

encountered, their presence recorded at 16 and 18 transects respectively. While a few

species were present across the majority of the catchment, there was a distinct gradient of

species turnover in the canopy along the length of the river system (Fig. 2.14; Table S2.1).

The canopy of the coastal plains was characterised by A. flexuosa, M. rhaphiophylla and

Taxandria juniperina, in addition to E. rudis. In the Karri forest, A. flexuosa continued to

be one of the most frequently occurring riparian species (Table S2.1) and was present at

almost every site in this vegetation class. Although not strictly canopy species, both

Callistachys lanceolata, and Trymalium odoratissimum subsp. trifidum were common

and formed a distinctive sub-canopy through the densely-vegetated Karri forest zones.

The riparian zones of the Jarrah and the Wandoo forests of the upper catchment were

dominated by M. rhaphiophylla, and also E. rudis and H. oleifolia. In the canopy of the

upper-most transects, T97 to T100, M. rhaphiopylla was replaced by M. viminea and

M. cuticularis (Fig. 2.14).

48

Fig. 2.13. Distribution of canopy (blue) and understorey (green) species sampled across

the Warren River transect, in relation to mean quadrat hydroperiod in days per year

inundated. Crosses indicate individual records of a species. The limits of boxes mark the

lower and upper quartiles (25%, 75%) centred around the median (bold centre line) of the

species distribution. Whiskers indicate the max and min (range). An ‘X’ preceding a

species name indicates an exotic species.


49

Fig. 2.14. Distribution of canopy (blue) and understorey (green) species sampled across

the Warren River transect, in relation to quadrat mean annual rainfall in mm per annum.

Crosses indicate individual records of a species. The limits of boxes mark the lower and

upper quartiles (25%, 75%) centred around the median (bold centre line) of the species

distribution. Whiskers indicate the min and max (range). An ‘X’ preceding a species name

indicates an exotic species.

50

Fig. 2.15. Transformation-based principal components analysis (tbPCA) of the complete

species assemblages of the Warren River transect riparian zones displaying the variation

explained by axes 1-2 in (a) and 3-4 in (b). The assemblage is described by presence-

absence records using the Hellinger transformation.


51

As with the canopy assemblage, there was substantial turnover in common

understorey species across the transect. Species that were commonly encountered on the

flood plains where the Warren River crosses the Scott Coastal Plain, and through the Karri

forest, were the sedges Carex (often C. appressa < 50 m asl) and Lepidosperma effusum

and L. persecans, as well as the rush, Leptocarpus thysananthus and a number of Juncus

species (Table S2.1). On the higher ground of the coastal plains the Ericaceae shrubs

Leucopogon obvatus subsp. revolutus, L. propinquus and Acacia pulchella were observed

frequently (Fig. 2.14; Table S2.1). A number of shrubs were encountered exclusively on

the coastal plain, including A. cochlearis, A. cyclops, L. parviflorus and Brachyloma

preissii (Table S2.1).

The understorey through much of the Karri forest riparian zone was dominated by

the invasive blackberry, Rubus anglocandicans. The shrubs Astartea leptophylla and

Taxandria linearifolia, were characteristic of these higher rainfall zones, most commonly

present adjacent to the main river channel or within the waterlogged areas of the

billabongs (Fig. 2.13).

Through the Jarrah forest, the understorey was characterised by Hakea lissocarpha,

Hibbertia commutata and Trymalium ledifolium var. rosmarinifolium. Additionally, the

shrubs L. obvatus subsp. revolutus, L. propinquus and A. pulchella which were commonly

observed on the coastal plain were also common through the Jarrah forest. Of note,

perennial vegetation in the understorey of these regions in the highest flood frequency

classes was sparse relative to that of the higher rainfall regions, and there was a greater

presence of woody flood debris (pers. obs.).

The understorey of the upper-most sites in the transect was also sparse, but Juncus

kraussii subsp. australinsus, Ficinia nodosa and a number of species of the glasswort

genus Tecticornia were present at low densities (Fig. 2.14; Table S2.1).

52

2.3.3 Environmental drivers of community composition

A reduced set of 215 quadrats containing both canopy and understorey species was

analysed in the variance partitioning analysis. The predictor set for the canopy species

analysis (hypothesis 1), included the geographic coordinates (𝑅𝑎2 = 0.1024), a subset of

12 of the 15 significantly autocorrelated dbMEM (global model 𝑅𝑎2 = 0.2790; reduced

model 𝑅𝑎2 = 0.2755), and 8 of 10 environmental predictors (global model 𝑅𝑎

2 = 0.1807;

reduced model 𝑅𝑎2 = 0.1793) including two climate predictors (PC1 and PC2), three

hydrological predictors (topo_var, hp_mean, hp_range) and three vegetation structure

predictors (pen_cpy, pen_srb and cpy_max). For the canopy community, the longitudinal

patterns across the length of the catchment (geographic coordinates) and in the geographic

layout at small through to large scales (dbMEMs), accounted for approximately 29.7% of

the total variation in canopy assemblages (Fig. 2.16a). The measured environmental

gradients explained a lower total amount of variation in the canopy assemblages (17.9%),

and the majority of this was spatially-structured environmental variation. Only 4.6% of

the total variation attributed to environmental components was independent of known

spatial gradients in the data (Fig. 2.16a). Finally, almost half of the spatial components of

variation could not be ascribed to measured environmental gradients in the canopy

communities.

I further partitioned the environmentally-structured variation in canopy community

dissimilarity into three main component classes representing forest structure, hydrology

and climate. Climate, which was itself largely spatially structured, explained most of the

environmental variation in canopy community turnover among plots (Table 2.4a; Fig.

2.16c). Although hydrology and forest structure also explained some variation in the

assemblages (Fig. 2.16c), a substantial proportion of the total variation in these drivers

was not independent from climate (Table 2.4a), and only 2.8% and 1.3% (respectively)

of the total variation could be independently ascribed to hydrology and forest structure.


53

Fig. 2.16. Hierarchical variation partitioning using redundancy analyses (RDA) on the

canopy and understorey species communities of the riparian zone along the Warren River

transect. (a-b) Percentage variation (adjusted R2) in the canopy and understorey

assemblages partitioned at the first tier, between spatial and environmental drivers. (c-d)

Percentage variation in the canopy and understorey assemblages (respectively)

partitioned at the second tier constrained by variation in environmental driver classes,

vegetation structure, hydrology, climate and canopy species. In grey, the percentage

variation explained within the intersection of space and environment, i.e. spatially

dependent component, and in blue/ green, the percentage explained by the environmental

driver independent of space. Note that bars indicate total variation explained by an

environmental driver and that shared with other environmental drivers, see Table 2.4 for

further partitioning within the environmental drivers.

54

Table 2.4. The variance in the canopy and understorey assemblages partitioned among

environmental driver classes, vegetation structure, hydrology, climate and canopy species

(percentage of total variance determined by RDA). Variance is partitioned within the

spatially independent (variation attributed to environment drivers only) and spatially

dependent component (variation attributed by both space and environmental drivers).

The predictor set for the understorey species analysis (hypothesis 2) included the

geographic coordinates (𝑅𝑎2 = 0. 06328) and a subset of 14 of the 15 significantly

autocorrelated dbMEM (global model 𝑅𝑎2 = 0.2108; reduced model 𝑅𝑎

2 = 0.2086).

Forward selection on the total environmental predictor set retained 6 of the original 10

environmental predictors (global model 𝑅𝑎2 = 0.2376; reduced model 𝑅𝑎

2 = 0.2332),

including the climate predictors (PC1 and PC2), two hydrological predictors (hp_mean,

and ri_range), two forest structure predictors (pen_srb and cpy_max) as well as the 10

PCA axes for variation in canopy species composition (cpy_PC1 to cpy_PC10). In

comparison to the canopy communities, the understorey communities were less spatially

structured, with approximately 23.3% of the total variation accounted for by the spatial

components (Fig. 2.16b). Although a greater proportion of the understorey community

was described by the environmental components (23.1%) than the canopy, half of this

environmentally driven variation was spatially independent, 11.5% (Fig. 2.16b).

Env | Space Env ∩ Space Env | Space Env ∩ Space

Hydrology 2.82% -0.39% 1.50% 0.48%

Structure 1.25% 1.44% 0.50% 0.36%

Climate 0.63% 7.10% 1.79% 0.18%

Canopy 6.19% 4.85%

Hydrology & Structure -0.07% 0.06% 0.06% 0.12%

Structure & Climate 0.03% 2.46% 0.17% 0.69%

Climate & Hydrology -0.07% 1.00% 0.03% 0.01%

Canopy & Hydrology 1.19% -0.15%

Canopy & Structure 0.09% 0.38%

Canopy & Climate 0.03% 2.79%

Structure & Climate & Hydrology -0.01% 1.68% 0.00% -0.10%

Structure & Hydrology & Canopy 0.08% 0.00%

Structure & Climate & Canopy -0.06% 1.65%

Hydrology & Climate & Canopy -0.06% 0.03%

Hydrology & Climate & Canopy & Structure 0.00% 0.54%

Total variation explained 4.58% 13.35% 11.50% 11.82%

(b) Understory(a) Canopy


55

The environmentally-structured component of variation in understorey

communities was further partitioned into four main determinants, forest structure,

hydrology, macroclimate and canopy species composition (Table 2.4b; Fig. 2.16d).

Canopy species composition explained the vast majority of understorey community

turnover among plots, contributing 17.8% of the 23.1% environmentally-structured

variance (Table 2.4b; Fig. 2.16d). Reinforcing the importance of the space and climate

variables in the canopy assemblage analysis (above), the majority of the variance in the

understorey assemblage that was explained by canopy species composition was not

independent of climate, particularly within the spatially-structured environmental

component (Table 2.4b). A further 6.19% of the variance in the understorey assemblages

covaried with changes in canopy species composition, independent of space, vegetation

structure, climate or hydrology; indicative of associations and interactions with canopy

species (Table 2.4b). By contrast, the variation explained exclusively by forest structure,

climate and hydrology was small, accounting for 0.50%, 1.79% and 1.50% respectively

(Table 2.4b). Somewhat surprisingly for a riparian zone community, the hydrological

variables explained very little variance in understorey composition, either independently

or in combination with other model components (Table 2.4b; Fig. 2.16d)

2.4 Discussion

Riparian zones have the potential to provide climate refugia for species that may

otherwise face displacement by rapid warming and drying of the climate. The intrinsic

capacity for riparian buffering, however, depends on the degree to which local

microclimate is effectively decoupled from regional climate changes. I examined the

extent to which local hydrology can buffer plant assemblages from climatic influences

across a strong regional rainfall gradient, using a novel ‘climosequence of

hydrosequences’ in southwest Australia. Contrary to expectations I show that the regional

rainfall gradient plays a greater role in determining the composition of riparian

56

assemblages than any putative ‘buffering effect’ of the local hydrological gradient. This

trend was stronger for the canopy communities than for the understorey communities,

despite the fact that canopy species might be expected to be accessing groundwater

resources, with lower seasonal variability in hydrological stress. Even for the understorey

communities, which ought to have a greater dependency on surface water availability,

variation in species composition was overwhelmingly attributed to the effects of regional

climate rather than local hydrological regime.

To examine the potential ‘sheltering’ role of the forest canopy in decoupling local

understorey microclimate from the ambient macroclimate, the canopy community and its

structure were included as drivers of understorey species composition. Although, the

forest canopy was shown to explain a significant proportion of variation in understorey

composition, the principal mechanism mediating this association was not found to be

through canopy structural influences on microclimatic control as I hypothesised.

Surprisingly, I found little evidence that local hydrological or environmental gradients

were driving community composition in either canopy or understorey communities,

suggesting that the riparian zone has a limited capacity to buffer community change in

the face of regional climate changes. The wider implications of these results for

management are discussed within the context of future projected rainfall declines in

southwest Australia.

2.4.1 Macroclimate as the primary driver of community composition

I partitioned compositional variation in riparian plant communities among the transverse

hydrological drivers and longitudinal climatic drivers of community composition along a

gradient spanning the length of the Warren River Catchment. Although there was a high

stochastic component of variation in community composition among sites, I detected

significant assemblage structuring based on both the hydrological and regional climate

gradients represented in the longitudinal axis. Of these factors, the greatest proportion of


57

explained variation in community composition could be attributed to the longitudinal

climatic drivers, with over double the influence of the hydrological drivers. This

contradicts, to a large extent, my first hypothesis that the local hydrological regime would

buffer riparian communities from variation in the regional climate regime. Although there

was a small number of obligate riparian species which had wider longitudinal

distributions (e.g. Astartea leptophylla, Eucalyptus rudis, Banksia seminuda), the

majority of the species inhabiting even the higher flood frequency plots, turnover was

high right across the regional climate gradient.

At face value, similar longitudinal gradients of species turnover have been

recorded in riparian assemblages in other river systems, but the relative influence of

macroclimate on these trends has been masked by collinear altitudinal, temperature or

strong erosional gradients (Lyon and Sagers 1998, Karrenberg et al. 2003, Yang et al.

2011). By contrast, in the Warren River system I was able to partition macroclimate as

the only major factor varying longitudinally along the catchment because the river

traverses a relatively flat landscape gradient (approximately 0.14%) with comparatively

little variation in temperature and stream power. It might be expected that the

communities would be similar along the longitudinal axis, if the mesic river environment

was decoupling the communities from regional rainfall gradient and alternatively, floral

communities would be largely driven by the transverse hydrological gradient as has been

observed elsewhere (e.g. Guadiana River of Portugal; Aguiar et al. 2006). Instead, I

observed a high compositional turnover in riparian communities along the Warren River

transect. This result is somewhat surprising, as although the river ceases to flow over

summer, still-water pools remain throughout much of the lower two thirds of the Warren

Catchment in summer, and in the deeper pools of the upper catchment, indicating that the

water table is relatively shallow during summer. Given that, the canopy species in

particular would be expected to have sufficiently deep root systems to gain access to

58

groundwater even in regions where surface water dries up (Hubble et al. 2010, Capon et

al. 2016). The results tend to suggest that a substantial proportion of the community is

not utilising these deeper water sources and is instead largely dependent on surface water

conditions. In a functional sense, this is more comparable to the riparian communities of

perennial rivers in temperate regions of Australia (Hancock et al. 1996, Lyons et al. 2000,

Warfe et al. 2014). Interestingly, in other regions where water is not a growth limiting

factor, shallow root systems have evolved in response to low nutrient conditions (Lamont

1982), and this may be a factor in understanding riparian responses in southwest Australia

where soils are known to be extremely nutrient poor (Lamont 1982, Hopper and Gioia

2004, Turner et al. 2017). Alternatively, riparian communities may largely be composed

of facultative riparian species that lack traits to withstand the physiological stresses of

annual waterlogging (Davison 1997, Jackson and Colmer 2005) and restrict root systems

to the upper soil layers, leaving them susceptible to an increasing frequency of summer

droughts.

2.4.2 Cascading effect of climate and canopy community on the understorey species

assemblages

The forest canopy is known to regulate microclimatic conditions within the forest interior

and effectively decouple the conditions experienced by understorey plants and animals

from external macroclimatic conditions (Ashcroft and Gollan 2013, Frey et al. 2016).

This decoupling effect has been shown to operate at ecologically relevant scales, and

generate a significant lag in species range shifts in herbaceous forest species in response

to regional climate shifts (Bertrand et al. 2011). In addition, a number of studies in

riparian systems have demonstrated that understorey communities respond to different

environmental drivers than their associated overstorey communities, often with a greater

dependence on hydrology (Lyon and Sagers 1998, Decocq 2002, Lyon and Gross 2005).

Thus, under hypothesis 2, I expected that the understorey communities of the Warren


59

River Catchment would be both highly dependent on the local hydrological regime, and

highly dependent on surrounding forest structure, as a proxy for variation in local

microclimatic conditions (rather than being influenced by the species identities of trees

in the surrounding canopy). Instead, over three-quarters of the total explained portion of

variation in understorey species composition could be attributed to the species

composition of canopy trees, represented by the canopy tbPCA axes, and not by structural

variation in canopy architecture. This suggests that the understorey communities are

tightly linked to variation in canopy species assemblages, but not to forest structural

conditions that might reflect microclimatic buffering effects of the forest canopy.

Furthermore, all three of the climate, canopy composition and forest structure variables

(and their shared components of variation), had greater explanatory power over

understorey species composition than local hydrology. This strongly suggests that local

buffering of microclimate and hydrology is not a strong determinant of understorey

species composition, as had been predicted in hypothesis 2.

The large spatially-structured proportion of shared variation among climate,

canopy species composition, and (to a lesser extent) forest structure is consistent with

previous findings that the SWWA’s forests show a graduated shift from tall, dense, high

biomass forests with high productivity in the high rainfall regions of the lower catchment,

to low-stature, open woodlands with slow growth and low productivity in the drier

extremes of the upper catchment (Pekin et al. 2009, Brouwers and Coops 2016). In

contrast to expectations, however, the patterns of turnover in regional forest composition

along rainfall gradients filtered through to the understorey communities. Given the near-

natural condition of these forests and the high degree of specialisation and endemism of

plants in relation to soils type, nutrient availability and climate, the high covariance in the

turnover of both understorey and canopy species could indicate convergence in adaptation

to the same extrinsic environmental driver, such as macroclimate or water logging, or in

60

unmeasured drivers such as soil type or soil salinity (Cowling et al. 1996, Hopper and

Gioia 2004). Alternatively, the variance explained by the canopy could indicate direct

associations between canopy and understorey species such as allelopathy, or facilitation

[e.g. hydrologic or nutrient redistribution (Prieto et al. 2014)], or through common

mycorrhizal associations (Lamont 1982). The underlying mechanisms driving this

association, and particularly the extent to which the association holds under the climate

change warrants further attention.

Here I considered rainfall to be the parameter mostly likely to be driving

vegetation changes along the longitudinal axis of the river. As a correlative study

however, it is important to point out that this gradient could equally reflect any number

of multifaceted, and collinear parameters predicted under climate change, including

increases in severity of summer drought (or winter frosts, with decreased cloud cover),

increases in surface water intermittency, greater depths to ground water and/or wetted

soils (Lite and Stromberg 2005) and greater extremes in flooding cycles (Alexander and

Arblaster 2009, Leigh et al. 2015). Although the turnover in riparian communities

observed here could be determined by a number of proximate mechanisms linked to

climate variation (and which might vary in their importance among species), the

cumulative outcome is unexpectedly high turnover in riparian communities ‘specialised’

to different climatic conditions.

2.4.3 Implications for management under climate change

By 2030, rainfall is predicted to decline by a further 15% in the SWWA (Hope et al.

2015), leading to further declines in surface run off of 12% to 40% (Silberstein et al.

2012), over and above the 16% decline already experienced since the 1970s (Petrone et

al. 2010), with a further deficit of 5 to 75 days per year when the river ceases to flow

altogether (Barron et al. 2012). As one of the largest rivers in south-west Australia, these

changes are predicted to be catastrophic for freshwater flora and fauna (Barron et al. 2012,


61

Beatty et al. 2013, Ogston et al. 2016). The changes currently being observed in SWWA

are an early outlook of changes that are expected in semi-arid and Mediterranean-type

climate regions across the globe (Klausmeyer and Shaw 2009, Underwood et al. 2009).

An understanding of the impacts of climate change on riparian systems is critical in

identifying regions which may provide a hydrological refuge not only for plants, but also

the fauna that inhabits them and should be prioritised for conservation protection

(Seabrook et al. 2014, Nimmo et al. 2015). Here, I used a space for time substitution

approach to show that changing precipitation regimes (along a spatial gradient) explained

a greater proportion of variation in species turnover than local hydrological regimes,

indicating that local environmental conditions are not decoupled from macroclimatic

gradients, and local riparian buffering will not ensure community resistance in the face

of climate change (Dobrowski 2011, Keppel et al. 2012, McLaughlin et al. 2017). These

results add to the growing body of literature suggesting that significant range shifts and

changes in assemblage structure are inevitable in the forests of SWWA given the

magnitude of rainfall decline, with significant range contraction predicted at the drier,

north-eastern extent of species ranges in particular (Pekin et al. 2009, Hamer et al. 2015a,

Matusick et al. 2016). The results presented here suggest that these effects could be

equally severe in riparian communities, not just in upland communities, even despite

having access to a less temporally variable water source. Exactly how these range shifts

will manifest is likely to depend on species specific responses. In long-lived species such

as trees and woody shrubs, where mature individuals can be relatively resilient to

environmental perturbations, failure to recruit can indicate early signs of a range

contraction. The results presented indicate that there is an urgent need to assess the impact

of climate shifts on recruitment in riparian species along drying climate gradients.

2.5 Supplementary material

62

T

ab

le S

2.1

. A

com

ple

te l

ist

of

spec

ies,

id

enti

fied

to t

he

hig

hes

t ta

xonom

ic l

evel

by p

rese

nce

at

each

of

49 s

ample

d t

ranse

ct l

oca

tions

along t

he

War

ren

and T

one

Riv

ers.

Lif

eform

s ar

e ab

bre

via

ted t

o H

: H

erb, T

: T

ree,

Sg:

Sed

ge,

F:

Fer

n, S

b:

Shru

b, R

: R

ush

, G

: G

rass

, G

T:

Gra

ss t

ree,

Cy:

Cyca

d.

Fam

ily:

Sp

ecie

s

T01

T02

T04

T05

T06

T08

T10

T13

T15

T19

T21

T28

T29

T30

T31

T32

T33

T37

T39

T41

T42

T43

T45

T48

T49

T50

T54

T55

T60

T62

T66

T68

T70

T71b

T71c

T73

T75

T78

T80

T84

T87

T89

T90

T91

T93

T94

T97

T98

T99

T100

Aiz

oace

ae

Ca

rpo

bro

tus

mo

dest

us

H1

Asp

ara

gace

ae

Lo

ma

nd

ra ?

pre

issi

iH

3

Casu

ari

nace

ae

All

oca

sua

rin

a d

ecu

ssa

taT

218

1

All

oca

sua

rin

a h

ueg

eli

an

aT

6

Ch

en

op

od

iace

ae

Atr

iple

x h

yp

ole

uca

H2

Tecti

co

rnia

sp.

H2

6

Tecti

co

rnia

lep

ido

sperm

aH

24

Tecti

co

rnia

perg

ran

ula

ta

?su

bsp

perg

ran

ula

ta

H6

2

Cyp

era

ce

ae

*C

yp

eru

s sp

.S

g1

*C

yp

eru

s co

ng

est

us

Sg

12

11

2

*C

yp

eru

s era

gro

stis

Sg

22

13

4

Ba

um

ea

ju

ncea

Sg

43

26

28

45

68

46

28

Ba

um

ea

va

gin

ali

sS

g7

Ca

rex

sp.

Sg

11

Ca

rex a

pp

ress

aS

g12

10

15

Ca

rex t

ere

tica

uli

sS

g11

Cya

tho

ch

aeta

sp.

Sg

6

Cya

tho

ch

aeta

cla

nd

est

ina

Sg

44

6

Fic

inia

no

do

saS

g5

16

110

23

29

24

39

34

4

Lep

ido

sperm

a e

ffu

sum

Sg

36

92

415

42

211

11

15

13

43

3

Lep

ido

sperm

a p

ers

eca

ns

Sg

93

77

610

74

24

511

24

210

36

22

24

31

19

11

31

Lep

ido

sperm

a s

qu

atu

m s

. la

t.Sg

51

21

Lifeform

12

00

mm

pa

+1

200

to

100

0 m

mpa

> 6

00 m

mpa

8

00 t

o 6

00

mm

pa

10

00

to

800

mm

pa


63

Tab

le S

2.1

. C

onti

nued

.

Fam

ily:

Sp

ecie

s

T01

T02

T04

T05

T06

T08

T10

T13

T15

T19

T21

T28

T29

T30

T31

T32

T33

T37

T39

T41

T42

T43

T45

T48

T49

T50

T54

T55

T60

T62

T66

T68

T70

T71b

T71c

T73

T75

T78

T80

T84

T87

T89

T90

T91

T93

T94

T97

T98

T99

T100

Cyp

era

ce

ae

co

nt.

Tetr

ari

a o

cta

nd

raS

g2

Tetr

ari

a s

p.

Jarr

ah F

ore

stS

g2

De

nn

sta

ed

tiace

ae

Pte

rid

ium

esc

ule

ntu

mF

27

12

10

12

22

47

23

91

33

41

411

11

94

46

63

Dil

len

iace

ae

Hib

bert

ia s

p.

Sb

14

15

2

Hib

bert

ia c

om

mu

tata

Sb

71

22

13

Hib

bert

ia c

un

eif

orm

isS

b14

35

85

21

21

Hib

bert

ia r

acem

osa

Sb

16

Eri

cace

ae

Bra

ch

ylo

ma

pre

issi

iS

b6

Leu

co

po

go

n c

ap

itell

atu

sS

b4

Leu

co

po

go

n i

nte

rsta

ns

Sb

5

Leu

co

po

go

n o

bo

va

tus

subsp

. re

vo

lutu

sS

b4

12

24

129

96

Leu

co

po

go

n p

arv

iflo

rus

Sb

41

5

Leu

co

po

go

n p

rop

inq

uu

sS

b41

11

10

52

13

41

10

13

1

Leu

co

po

go

n v

ert

icil

latu

sS

b3

21

41

1

Fab

ace

ae

Aca

cia

spp.

Sb

2

Aca

cia

co

ch

lea

ris

Sb

12

26

Aca

cia

cyclo

ps

Sb

13

*A

ca

cia

dea

lba

taT

13

Aca

cia

exte

nsa

Sb

2

Aca

cia

pu

lch

ell

aS

b38

13

11

8#

Aca

cia

sa

lig

na

Sb

13

1

Aca

cia

uro

ph

yll

aS

b3

5

Bo

ssia

ea

lin

op

hyll

aS

b1

111

12

Ca

llis

tach

ys

lan

ceo

lata

T10

44

11

42

24

12

71

21

Lifeform

12

00

mm

pa

+1

200

to

100

0 m

mpa

10

00

to

800

mm

pa

80

0 t

o 6

00

mm

pa

> 6

00 m

mpa

64

T

ab

le S

2.1

. C

onti

nued

.

Fam

ily:

Sp

ecie

s

T01

T02

T04

T05

T06

T08

T10

T13

T15

T19

T21

T28

T29

T30

T31

T32

T33

T37

T39

T41

T42

T43

T45

T48

T49

T50

T54

T55

T60

T62

T66

T68

T70

T71b

T71c

T73

T75

T78

T80

T84

T87

T89

T90

T91

T93

T94

T97

T98

T99

T100

Fab

ace

ae

co

nt.

Ga

stro

lob

ium

bil

ob

um

Sb

117

Ga

stro

lob

ium

pra

em

ors

um

Sb

8

Ho

vea

ell

ipti

ca

Sb

32

44

28

11

6

Hae

mo

do

race

ae

An

igo

zan

tho

s fl

avid

us

H5

71

15

25

5

Co

no

styli

s sp

.H

21

Co

no

styli

s a

cu

lea

ta s

ubsp

.

acu

lea

taH

1

Co

no

styli

s ?se

rru

lata

H1

41

He

me

rocall

idace

ae

Dia

nell

a s

p.

H1

Dia

nell

a r

evo

luta

var.

revo

luta

H1

Jo

hn

son

ia l

up

uli

na

H1

Irid

ace

ae

Pa

ters

on

ia s

p.

H1

Ju

ncace

ae

Ju

ncu

s sp

.R

11

*Ju

ncu

s a

cu

tus

R1

Ju

ncu

s a

ma

bil

isR

32

712

1

Ju

ncu

s k

rau

ssii

subsp

.

au

stra

lin

sus

R2

21

46

Ju

ncu

s su

bse

cu

nd

us

R14

24

1

Ju

ncu

s p

all

idu

sR

12

42

48

712

1

Ju

ncag

inace

ae

Cycn

og

eto

n s

p.

H1

1

Cycn

og

eto

n l

inea

reH

1

> 6

00 m

mpa

Lifeform

12

00

mm

pa

+1

200

to

100

0 m

mpa

10

00

to

800

mm

pa

80

0 t

o 6

00

mm

pa


65

Tab

le S

2.1

. C

onti

nued

.

Fam

ily:

Sp

ecie

s

T01

T02

T04

T05

T06

T08

T10

T13

T15

T19

T21

T28

T29

T30

T31

T32

T33

T37

T39

T41

T42

T43

T45

T48

T49

T50

T54

T55

T60

T62

T66

T68

T70

T71b

T71c

T73

T75

T78

T80

T84

T87

T89

T90

T91

T93

T94

T97

T98

T99

T100

Myrt

ace

ae

Ag

on

is f

lexu

osa

T76

89

7#

261

16

651

314

75

95

#56

13

89

559

Ast

art

ea

lep

top

hyll

aS

b1

39

24

354

10

716

831

610

82

24

81

75

510

2

Co

rym

bia

ca

lop

hyll

aT

23

13

210

21

11

43

92

11

Eu

ca

lyp

tus

div

ers

ico

lor

T2

11

1

Eu

ca

lyp

tus

ma

rgin

ata

T2

14

21

21

Eu

ca

lyp

tus

rud

isT

11

23

28

613

12

17

32

113

22

27

33

210

44

411

14

31

12

13

8

Eu

ca

lyp

tus

wa

nd

oo

T3

34

Mela

leu

ca

cu

ticu

lari

sT

50

Mela

leu

ca

in

ca

na

Sb

94

529

70

23

12

45

16

#12

Mela

leu

ca

rh

ap

hio

ph

yll

aT

31

34

812

41

524

822

11

15

46

52

33

Mela

leu

ca

vim

inea

T69

4

Pa

rase

ria

nth

es

lop

ha

nth

aS

b

Ta

xa

nd

ria

ju

nip

eri

na

T24

510

110

5

Ta

xa

nd

ria

lin

ea

rifo

lia

Sb

11

23

4

Pin

ace

ae

*P

inu

s sp

.T

1

Po

ace

ae

Au

stro

stip

a s

p.

G11

Pri

mu

lace

ae

Sa

mo

lus

jun

ceu

sH

1

Pro

teace

ae

Ba

nk

sia

gra

nd

isT

19

Ba

nk

sia

sem

inu

da

T10

21

61

21

71

15

824

22

21

Ha

kea

lis

soca

rph

aS

b4

53

Ha

kea

ole

ifo

lia

T5

347

26

12

13

31

21

13

113

16

213

Ha

kea

pro

stra

taS

b3

15

Pers

oo

nia

lo

ng

ifo

lia

Sb

41

Pro

teaceae s

eedlin

g (

unid

ent.

)-

31

> 6

00 m

mpa

Lifeform

1200 m

mpa +

1200 t

o 1

000 m

mpa

1000 t

o 8

00 m

mpa

800 t

o 6

00 m

mpa

66

Tab

le S

2.1

. C

onti

nued

.

Fam

ily:

Sp

ecie

s

T01

T02

T04

T05

T06

T08

T10

T13

T15

T19

T21

T28

T29

T30

T31

T32

T33

T37

T39

T41

T42

T43

T45

T48

T49

T50

T54

T55

T60

T62

T66

T68

T70

T71b

T71c

T73

T75

T78

T80

T84

T87

T89

T90

T91

T93

T94

T97

T98

T99

T100

Re

sti

on

ace

ae

Lep

toca

rpu

s d

ep

ila

tus

R1

21

34

Lep

toca

rpu

s th

ysa

na

nth

us

R1

22

11

Lep

toca

rpu

s ?th

ysa

na

nth

us

R3

Rh

am

nace

ae

Sp

yri

diu

m g

lob

ulo

sum

T22

31

Try

ma

liu

m l

ed

ifo

liu

m v

ar.

rosm

ari

nif

oli

um

Sb

62

Try

ma

liu

m o

do

rati

ssim

um

subsp

. tr

ifid

um

T17

14

742

56

437

18

92

98

Ro

sace

ae

*R

osa

ru

big

ino

saS

b1

11

16

*R

ub

us

an

glo

ca

nd

ica

ns

Sb

71

24

414

16

216

24

212

43

11

20

*R

ub

us

ulm

ifo

liu

sS

b3

Ru

tace

ae

Ch

ori

laen

a q

uerc

ifo

lia

T2

San

tala

ce

ae

Exo

ca

rpo

s o

do

ratu

sS

b1

Exo

ca

rpo

s sp

art

eu

sS

b1

1

Th

ym

ela

eace

ae

Pim

ele

a c

lava

taS

b15

11

Vio

lace

ae

Hyb

an

thu

s fl

ori

bu

nd

us

?su

bsp

. fl

ori

bu

nd

us

Sb

3

Xan

tho

rrh

oe

ace

ae

Xa

nth

orr

ho

ea

gra

cil

isG

T2

2

Xa

nth

orr

ho

ea

pre

issi

iG

T3

43

24

1

Zam

iace

ae

Ma

cro

zam

ia r

ied

lei

Cy

51

14

11

22

22

24

44

34

12

12

Lifeform

1200 m

mpa +

1200 t

o 1

000 m

mpa

1000 t

o 8

00 m

mpa

800 t

o 6

00 m

mpa

> 6

00 m

mpa

67

3 Evidence of range shifts in riparian plant assemblages in response to

multidecadal streamflow declines

3.1 Introduction

Already, the small rise in mean global temperature resulting from anthropogenic climate

change is ecologically visible in forest ecosystems. Marked phenological shifts in the

timing of bud break, flowering and senescence have been reported across North America

and Europe (Vitasse et al. 2010, Reyer et al. 2013), as well as increases in episodic

mortality events linked to increasingly frequent heatwaves and drought (Allen et al. 2010,

2015). For many species, survival over the coming decades will depend on their ability

to adapt to the new climatic conditions in situ, or shift geographic range to maintain their

climatic optimum (Parmesan 2006, Aitken et al. 2008, Dawson et al. 2011). In stark

contrast to mobile organisms where analyses of distributional shifts have been shown to

match climatic shifts (Chen et al. 2011), sessile organisms such as plants, particularly

those with longer generation times like woody shrubs and trees, can be much more

constrained in their responses to climate change (Lenoir and Svenning 2015).

In a strict sense, determination of range shifts requires temporally-replicated data

over relevant time scales (e.g. Bertrand et al. 2011, Feeley et al. 2011, Telwala et al. 2013,

Máliš et al. 2016). In lieu of such datasets, indications of potential range shifts in plant

species have been inferred by examining the skew in abundance distributions (Murphy et

al. 2010, Groom 2013), or by exploiting the long generation times and comparing the

distribution of juveniles relative to the adult population (Lenoir et al. 2009, Galiano et al.

2010, Zhu et al. 2012, 2014, Máliš et al. 2016, Fei et al. 2017). In comparing the

distribution of juvenile to adults, the assumption is made that the range inhabited by new

68

recruits into the population is representative of the optimal climatic envelope within

current climate space, while the distribution of adults represents a suboptimal climate

envelope that characterised historical conditions (Lenoir et al. 2009). For example, range

expansion is typically first observed as the establishment of seedlings beyond the former

adult range, such as upward shifts along elevational temperature gradients (Galiano et al.

2010, Elliott and Kipfmueller 2011, Vitasse et al. 2012), while range contraction can

manifest as recruitment failure at range margins (Zhu et al. 2012, Bell et al. 2014), where

a lack of new reproductively mature individuals will eventually render a population

inviable. Although adult mortality events are more typically taken as ‘conclusive’

evidence of range contraction, they tend to only be evident for long-lived species

following catastrophic disturbance events (Allen et al. 2010, Brouwers et al. 2013a,

Matusick et al. 2013, Stella et al. 2013), or with acute biotic stressors (Galiano et al.

2010), whereas recruitment and seedling mortality are more sensitive to incremental

changes in environmental conditions (Lloret et al. 2009, Bell et al. 2014).

Outside of alpine and high latitudes regions, where temperature rises are lifting

elevational and latitudinal tree lines (e.g. Feeley et al. 2011), there is little available

information on potential range shifts in lowland or low-latitude ecosystems, and little

focus on climatic gradients other than temperature (Lenoir and Svenning 2015, Fei et al.

2017). In the few studies that do exist, patterns emerging suggest more complex

interactions with moisture changes than the simple elevational and poleward shifts

projected by temperature rises (Rapacciuolo et al. 2014, Máliš et al. 2016, Fei et al. 2017).

It is perhaps unsurprising that moisture demands might exacerbate temperature-

dependent range shifts, when rises in temperature increase the atmospheric moisture

demand (vapour pressure deficit; Breshears et al. 2013), and induce cavitation and

hydraulic failure (Choat et al. 2012). In line with these observations, the role of

topographic and hydrological features in reducing the exposure, and buffering organisms

Chapter 3: Range shifts in riparian plants

69

from regional climates is receiving increasing attention (Dobrowski 2011, Keppel et al.

2012, 2015, Lenoir et al. 2017, McLaughlin et al. 2017). In regions experiencing a

warming and drying climate, riparian zones are predicted to buffer the suboptimal

climatic envelope, affording species more time to adapt to the new environmental

conditions.

The buffering effect of riparian systems in the face of climate warming hinges on

the riparian corridors remaining as hydrological refugium in the landscape. However, in

regions with warming and drying climates, the hydrological regime of riparian systems is

also under threat (Barron et al. 2012). Decades of monitoring the downstream impacts of

flow reduction in regulated rivers (e.g. dammed rivers, or rivers with high water

extraction), has shown almost universally that there is overall narrowing of the river

channel, encroachment of upland species, and decline in the distribution of obligate

riparian species, both through competition with upland species and reduction of suitable

habitat (Busch and Smith 1995, Shafroth et al. 2002, Tockner and Stanford 2002, Lite

and Stromberg 2005, Stromberg et al. 2010, Bejarano et al. 2011, 2012). These impacts

of flow regulation in managed systems are predicted to mirror the changes expected in

natural systems under climate-induced flow reduction in the future (Horner et al. 2009,

Seavy et al. 2009, Stella et al. 2013, Stromberg et al. 2013). While a number of studies

have demonstrated the impacts of climate change on riparian communities, these have

generally only been studied at a local scale (e.g. Stella et al. 2013), and excluded the

responses of upland species (e.g. Bejarano et al. 2012) across contrasting regional climate

gradients. To gain a more complete understanding of the responses of riparian

communities to drying and warming climates, it is essential to take an integrative

approach, examining the interactive impacts of regional climate change and the alteration

of local hydrological regimes on recruitment failure and distributional range shifts in co-

occurring riparian and upland species.

70

The south-west of Western Australia (SWWA) has experienced one of the most

substantial rainfall declines observed worldwide (Hennessy et al. 2007, Petrone et al.

2010, Silberstein et al. 2012). In the 1970’s, a significant decrease in the frequency and

magnitude of wet weather systems was observed (Hope et al. 2006). The result has been

a 16% decline in rainfall, culminating in reductions of up to 50% in surface runoff to

rivers and water storage dams (Petrone et al. 2010). Future climate projections for the

region predict further declines in rainfall, and consequently streamflow [out to 2090

(CSIRO and Bureau of Meteorology 2015) and to 2030 (Barron et al. 2012, Silberstein et

al. 2012) respectively] under all emission scenarios examined. As the major climatic

driver of vegetation types across the region, rainfall declines are predicted to shift optimal

climatic envelopes for plant species in a south-westerly direction (Hamer et al. 2015a).

Although geographic range shifts in SWWA plant species have not been explicitly

reported, declines in crown health and crown mortality (Brouwers et al. 2013a, 2013b,

Evans et al. 2013, Matusick et al. 2013), as well as shifts in dominant structural form (i.e.

from a tall single trunk, to shorter multi-trunked forms; Matusick et al. 2016) and

reductions in primary productivity (Brouwers and Coops 2016) been observed in several

keystone Eucalyptus species, providing early indications that the region’s flora is under

stress. Although there are no field studies showing comparable evidence of impacts for

riparian communities of the SWWA (but see, Groom et al. 2001, Froend and Sommer

2010), the fact that there has been a three-fold decline in streamflow per unit change in

rainfall (Silberstein et al. 2012) suggests that riparian plant communities are likely to be

at greater risk than upland communities.

In this study, I test the ecological impacts of multidecadal streamflow declines on

the riparian plant communities of SWWA. I examine the distribution and frequency of

immature versus mature individuals of riparian and upland species inhabiting the riparian

zones in response to changing local hydrological and regional rainfall distributions. In


71

doing so, I investigate whether there is potential recruitment failure in response to climate

change reduced flows, and whether effects of flow reduction are exacerbated or buffered

by regional rainfall, in riparian and upland species. I hypothesise that due to a higher

sensitivity to surface water availability, the observed range of immature individuals will

have contracted relative to the observed range of the adult population. Furthermore, I

expect that the mismatch in distribution of immature versus adult plants will vary among

functional groups, and will be greatest in obligate riparian species that are restricted to

near-channel habitats (but buffered within areas of high regional rainfall), less severe in

facultative riparian species that are also known to utilise adjacent upland habitats, and

predominately absent in upland species (i.e. buffered by streamflow).

3.2 Methods

3.2.1 Study system

The Warren River, and its major tributaries the Tone River and Murrin Brook (hereafter

the Warren River transect) of the SWWA are cumulatively about 275 km in length. Along

the length of the catchment there is only a shallow topographical gradient to a maximum

elevation of 385 m asl (Fig. 2.2; Geoscience Australia, www.ga.gov.au, accessed 23 May

2016), thus average annual temperatures vary little across the catchment (between 14.3°C

and 15.7°C; Fig. 2.1). Instead, rainfall is considered the most significant climatic driver

of vegetation distributions across the region (Fig. 2.2). In the headwaters, historical mean

annual rainfall is approximately 550 mm pa, and rainfall incrementally increases to

approximately 1200 mm at the river mouth (Fig. 2.2). Four major vegetation types are

observed across this gradient. The river originates in the inland Wandoo woodlands

(Eucalyptus wandoo, < 650 mm pa) of western Wheatbelt region, passes through the

southern Jarrah and Marri woodlands (Eucalyptus marginata and Corymbia calophylla,

650 – 900 mm pa), then the tall, dense Karri forests of the higher rainfall regions of the

Darling scarp (Eucalyptus diversicolor, > 900 mm pa), before descending on to the Scott

http://www.ga.gov.au/

72

Coastal plain where Agonis flexuosa is the dominant canopy species. The majority of the

land use in the lower two-thirds of catchment is in national parks or native forestry

reserves bordering, but not overlapping with, the riparian zone (Fig. 2.1), while the upper

third has been subjected to extensive clearing for agriculture.

Daily interpolated rainfall records for Australia were obtained from the Australian

Bureau of Meteorology (BoM; licenced to UWA) from the early 1900s. To quantify the

step decline in precipitation observed in the 1970s (Hope et al. 2006), mean annual

rainfall was calculated for two periods: 1901 to 1960 (historical) and 1970 to 2010

(recent). The percentage change in rainfall between the two periods was calculated, but

was highly negatively correlated with historical rainfall (Pearson’s r = -0.9; Fig. S3.1),

therefore, only mean annual historical rainfall is included in the analysis.

The Warren River transect encompassed a rainfall gradient ranging from over

1200 mm pa at the mouth, to less than 550 mm pa in the headwaters. To ensure that

vegetation sampling encompassed a representative range of these rainfall conditions, the

locations of sampling sites were stratified by rainfall isohyet, defining five strata, ≤

600 mm, 600-800 mm, 800-1000 mm, 1000-1200 mm and >1200 mm (Fig. 2.2). Within

each stratum, 20 potential survey locations, spaced at least 1 km apart and assigned to the

true left or right bank (facing downstream), were randomly generated in ArcGIS 10.3.1

(ESRI Inc.). The logistical feasibility of sampling a site was determined on the first site

visit using a set of predefined criteria (see Section 2.2.2), with the goal being to survey

10 sites per zone. Sites were discarded where infrastructure, agricultural or management

practices may have interfered with seedling establishment or where a site was

inaccessible.

To describe the topography of the riparian zone at a high resolution, an aerial

LiDAR (light detection and ranging) survey was undertaken across the length of the


73

Warren River transect. The point clouds were algorithmically separated into ground and

non-ground points, and the ground points were interpolated to generate a 1 × 1 m pixel

digital ground model (DGM) with a horizontal accuracy of 0.55 m and a vertical accuracy

of 0.30 m.

3.2.2 Streamflow

Streamflow records were obtained from the Western Australian, Department of Water

(DoW; water.wa.gov.au/maps-and-data/monitoring; accessed 7th November 2016; Fig.

2.1) for four gauge stations situated along the Warren and Tone Rivers to calculate the

flow regimes at each sampling site. Two 10-year periods, 1980 to 1989 and 2001 to 2010

were selected to estimate past and recent conditions. Unfortunately, continuous flow

records are not available for periods predating the 1970s ‘step decline’ in precipitation,

so the selected periods are likely to underestimate overall flow reduction. However, it is

possible that there could have been a lag period between rainfall change and subsequent

ecological impacts of flow reduction, therefore I assume that the 1980s period reflects

‘low’ impacts of flow reduction, while the 2000s period reflects ‘high’ impacts of flow

reduction. Importantly, further significant shifts have been observed in streamflow since

the 1980s (Petrone et al. 2010).

Of the four gauging stations available along the Warren River transect, three had

continuous data for the selected periods while the fourth, uppermost gauging station on

the Tone River, Hillier Road, (DoW ID: 607027; 251 m asl) was only established in 2002.

While the analysis could have been carried out on just the three stations with complete

gauging records, the gauge at Bullilup (DoW ID: 607007; 201 m asl) tends to record

higher water levels than expected (see discussion section 2.2.4.1), and interpolation from

just the three sites in a preliminary analysis resulted in over-estimation of water height in

the upper catchment. In order to retain the Hillier Road gauge data in the estimates of

flow regime, a linear model (LM) was developed to estimate missing data in the historical

http://water.wa.gov.au/maps-and-data/monitoring

74

records at Hillier Road based on the known (recent) relationship between Bullilup and

Hillier Road flow regimes. I constrained both the upper and lower limits of the model to

include records between 0.5 m and 2 m at Hillier Road gauge station. The lower limit of

0.5 m was imposed to remove records tracking evaporation once flow had ceased

(≈ 0.25 m at Hillier Road; Fig. 3.1a), and to account for inaccuracies in the DGM. The

upper limit of 2 m was imposed based on the breakdown in the relationship between stage

heights at Bullilup and Hillier Road above this value (Fig. 3.1a). The remaining subset of

stage height data was log transformed, and modelled using a LM in R, version 3.3.2 (R

Core Team 2016) (Fig.3.1b). Stage height was estimated from the Bullilup records for the

two selected periods. The estimated stage heights at Hillier Road were assigned values of

0.5 m or 2 m where stage height at Bullilup was outside the minimum or maximum range

of the model, respectively.

Fig. 3.1. Stage height (m) at a DoW gauges, Bullilup (ID: 607007) and Hillier Road (ID:

607027) on the Tone River. (a) Stage height above base flow at the two gauging stations.

The red lines indicate the upper (2 m) and lower (0.5 m) limits of the data used to (b)

model the relationship between the two stations. The grey box in (a) delineates the

approximate area of data displayed in (b). The linear model coefficients and model fit

(adjusted R2) are annotated on the plot, and the shaded area indicates 95% confidence

interval of the predicted model.


75

Using the estimated Hillier Road data, and the records from Bullilup and the two

lower Warren gauges (Barker Road and Wheatley Farm, Fig. 2.1), a linear model was

then constructed to model the water height above baseflow as a function of elevation for

each day of the two 10-year periods (see Section 2.2.4.1 for detailed methods). The water

height at each sampling site was interpolated from the elevation of the lowest point in the

channel (or summer water level) taken from the DGM to estimate a time series of water

heights at each sampling site (Section 2.2.4.1; Fig. 2.6). The mean hydroperiod (HP;

defined here as the mean number of days per year that the water level equals or exceeds

the elevation) and recurrence interval (RI; the probability that the water level equals or

exceeds the elevation at least once during any calendar year) were calculated for each

0.1 m increment from 0.5 m above baseflow, and greater. Finally, the differences in HP

and RI between the historical and recent rainfall periods were calculated, and the resultant

differences, ΔHP (change in hydroperiod) and ΔRI (change in recurrence interval), were

retained alongside recent RI and recent HP, respectively, as predictors in statistical

models.

3.2.3 Vegetation

In total, 49 survey sites were visited once during two consecutive summers, December

2013 to April 2014, and November 2014 to May 2015. At each site, a 10 m wide transect

was run from the water’s edge out to the width of the riparian zone (varying in length

from 5 to 90 m, depending on the width of the riparian zone). All trees and shrubs rooted

within the transect were recorded and identified to species level following the

nomenclature of the Western Australian Herbarium (https://florabase.dpaw.wa.gov.au/).

As the majority of the trees and woody shrubs in the region (predominantly Myrtaceae

and Proteaceae) retain a woody capsule/fruit after seed set, each plant was searched for

the presence of fruit or flowers to assess whether an individual was reproductively

immature or mature, and its reproductive state was recorded as a binary response, 1 or 0


76

respectively. I used a binary response of immature/ mature status (rather than age-class

structure based on heights or stem diameters) to estimate broad ‘recruitment’ trends over

a longer timeframe and reduce the temporal bias of a single time point sample (Dixon et

al. 2002).

The geographic coordinates of the transect were marked using a GPS unit

(GPSMAP® 62s, Garmin) and the location of each plant within the transect recorded to

the nearest 0.5 m with a tape measure (Fig.3.2). To account for GPS error, the transect

coordinates were spatially rectified to the DGM, LiDAR point clouds of the vegetation

and field pictures. The position of each plant within a transect was then spatially adjusted

to the rectified transect position to obtain corrected geographic coordinates, elevation

(m asl), and elevation relative to the transect origin (i.e. above base flow), allowing the

hydrological regime to be calculated for each individual plant.


77

Fig. 3.2. Examples of survey transect measurements for T5, T19 and T84 (see Fig 2.1 for

locations within catchment). Transect/quadrat positions were rectified using aerial

imagery, field photos and LiDAR generated digital ground models. The elevation of each

tree and shrub was calculated relative to the transect origins. Forest structure was

quantified within buffer zones of 2.5 m for individual plants and 100 m for transects to

account for variation in surrounding conditions that might influence species occurrence

or age structure. Note DGM is scaled to elevation in meters above sea level, whereas the

CSM is in meters above ground height. The river in the CSM is also in white, where no

vegetation points were recorded.

78

3.2.4 Forest structure

To account for variation in surrounding land use and microclimate on seedling

establishment, I quantified the structure of the forest at a landscape scale around each

transect and at a local scale around each individual tree. The forest structure in the

landscape was described by an area encompassing each transect plus a 100 m surrounding

buffer (Fig.3.2). Structure was measured using ground normalised LiDAR point clouds

and a 1 × 1 m resolution canopy surface model (CSM) describing the maximum canopy

height in each pixel (see Section 2.2.4.3 for further details). From the point clouds, I

obtained the maximum point height, and the laser penetration rates through six vertical

height strata: the penetration rate to 24 m, penetration through 24 to 16 m, 16 to 8 m, 8 to

3 m, 3 to 0.5 m and penetration to ground level (≤ 0.5 m). A further four metrics were

obtained from the CSM: the range, mean, coefficient of variation (CV) and variance (var)

in maximum canopy height across each transect and buffer zone. The metrics calculated

at this scale described the structure of the forest across the transect as well as the

surrounding landscape, i.e. a riparian zone backing on to cleared farm land showed high

penetration through the sub-canopy layers and to ground level and a high variance in the

canopy surface models (CSM), whereas a transect in pristine, dense karri forest had low

penetration through the majority of canopy and sub-canopy layers, with high mean

canopy heights, with low variability.

At the individual scale, a circular buffer with a radius of 2.5 m was generated

around each tree and shrub (Fig. 3.2). The structure at this scale was used as a proxy for

localised site microclimate, whereby density of forest strata and laser penetration rates to

ground level might be used as a proxy for variance in humidity and light conditions

(Lovell et al. 2003, Leutner et al. 2012). Similar to landscape structure, I obtained the

laser penetration rates through six vertical height strata: the penetration rate to 24 m,

penetration through 24 to 16 m, 16 to 8 m, 8 to 3 m, 3 to 0.5 m and penetration to ground


79

level (≤ 0.5 m) for each tree and shrub. In addition, the mean and the maximum height of

the CSM was calculated for buffer zone of each individual.

Two principal component analyses (PCA) were run separately for the landscape

and microstructure variable sets in ‘vegan’ package (Version 2.4-2; Oksanen et al. 2017)

in R 3.3.2 (R Core Team 2016) to manage collinearities and reduce the number of

predictors. Predictor variables were normalised prior to analysis. At the transect scale, the

PC1 and PC2 axes accounted for 89% of the total variation observed in forest structure.

PC1 (hereafter, T_PC1) described 73% of the variation in canopy height and density,

where increases in canopy height were observed with decreasing penetration rates

through most vegetation strata (Fig. 3.3a). PC2 (hereafter, T_PC2) described a further

16% of variation and represented deviations from these trends, largely in shrub layer

density and canopy height (Fig. 3.3b). At the individual tree scale, the axes PC1 and PC2

accounted for 70% of the total variation. PC1 (hereafter I_PC1) described 52% of the

total variation, largely in canopy height (Fig. 3.4a), while PC2 described 18% of the

variation, representing penetration rates through the sub-canopy layers and to ground

level (I_PC2, Fig. 3.3b).

80

Fig. 3.3. Variable loadings on (a) PC1 and (b) PC2 axis of a principal coordinates analysis

ordination at the transect level (T). Forest structure is measured in laser penetration (pen.)

rates through vegetation strata, obtained from LiDAR point clouds and the range, mean,

coefficient of variation (CV) and variance (Var) of the canopy height obtained from a

canopy surface model (CSM).


81

Fig. 3.4. Variable loadings on (a) PC1 and (b) PC2 axis of a principal coordinates analysis

ordination at the individual tree and shrub level (I). Forest structure is measured in laser

penetration (pen.) rates through vegetation strata, obtained from LiDAR point clouds and

the range, mean, coefficient of variation (CV) and variance (Var) of the canopy height

obtained from a canopy surface model (CSM).

82

3.2.5 Statistical analysis

As most species recorded were rare, and in frequencies too low to test for differences in

age structure, I selected only species of trees and woody shrubs that were sufficiently

abundant (> 50 individuals) to investigate the effects of changing rainfall and

hydrological regimes on recruitment.

To test whether the frequency of immature and mature individuals differed along

rainfall gradients or with shifts in flow regime, generalised linear mixed models (GLMM)

were fitted for each species using a binomial distribution and a logit link (Bolker et al.

2009) in R version 3.3.2, package, ‘lme4’ (Version, 1.1-12; Bates et al. 2015). The

variables describing transect level (T_PC1 and T_PC2) and individual level (I_PC1 and

I_PC2) variation in forest structure were included as fixed covariates to control for

variation in land use and microclimatic conditions independent of the hydrological

parameters. To account for non-independence of individuals sampled within a transect, a

random intercept was included for transect identity.

In the models, hydrological conditions were defined using combinations of five

fixed predictors RI, ΔRI, HP, ΔHP, and historical rainfall, plus their interactions.

However, RI and HP were highly correlated (Pearson’s r = 0.78; Fig. S3.1), suggesting

that individuals that were regularly inundated also tended to be inundated for longer

durations. Ecologically, however, the two parameters could represent quite different

limitations on an individual. For example, RI provides an indication of the regularity with

which an individual is exposed to surface water, but could also be indicative of the

regularity with which hydrochorously dispersed seed is deposited. By contrast, in

estimating the number of days per year that an individual is inundated, HP provides an

indication of the physiologically stressful, often anoxic conditions imposed by prolonged

saturation (Jackson and Colmer 2005). Therefore, rather than discarding one set of the


83

collinear parameters, a two-phase model selection approach was used to identify the

parameter set which best described recruitment in each species.

At the first phase, global models were constructed to test the effects of HP (HP,

ΔHP, rainfall and interactions) and RI (RI, ΔRI, rainfall and interactions) separately. The

global models were simplified using model selection procedures comparing Akaike

Information Criterion for small sample sizes (AICc) in ‘MuMIn’ package in R (Version

1.15.6; Barton 2016). For each parameter set, the most parsimonious model within 2 AIC

units of the top model was selected as the ‘best fit’ model (Arnold 2010). Then at the

second phase, the AIC of the best RI model and the best HP model were compared, and

the final model was taken as the model with the lowest AIC out of either model set (Table

S3.1). Prior to analysis all of the continuous predictors and covariates were centred and

scaled to 2 standard deviations (Gelman 2008). Models were assessed for over-dispersion,

however no adjustment was necessary. Model fit was assessed using Nakagawa and

Schielzeth (2013) R2 approach.

3.3 Results

A total of 4089 individuals representing 56 species of trees and woody shrubs were

identified across 49 sites sampled along the Warren River transect (Table S2.1). Of these,

17 species accounted for 80% (3256 individuals) of the total number of individuals

recorded (Table 3.1). Moreover, 44% of all individuals were reproductively immature,

although the proportions differed substantially between species (Table 3.1). Of the 17

species recorded in sufficient abundances to analyse demographically, six demonstrated

significant relationships with the examined hydrological and rainfall gradients (Tables

S3.1; 3.2a; 3.2b).

84

Table 3.1. The hydrological conditions prevailing within the sampled range of the most

abundant woody trees (T) and shrubs (S) of the Warren River transect as described by the

10%, 50% and 90% percentiles of mean annual rainfall (mm pa), flood recurrence interval

(probability of flooding in any one year) and hydroperiod (mean number of days flooded

annually). The total number of individuals recorded is listed under n, the value in brackets

indicating the number of individuals classed as immature. Functional groups were

allocated using the recurrence interval and the hydroperiod estimated during this study as

well as with accounts in the literature and within taxonomic descriptions, groupings

defined by Rood et al. (2010).

Of the six obligate riparian species recorded in sufficient numbers to model,

models failed to converge for three species, Melaleuca cuticularis, Melaleuca viminea

and Taxandria juniperina due to the narrow range of observed variation in responses to

predictors. Both M. cuticularis and M. viminea were recorded in just two and three

n 10% 50% 90% 10% 50% 90% 10% 50% 90%

Melaleuca cuticularis M 16 T Obligate 538 538 538 1.0 1.0 1.0 139 139 139

Im. 57 534 534 538 1.0 1.0 1.0 139 140 140

Melaleuca viminea M 38 T Obligate 538 538 538 0.5 1.0 1.0 6 36 139

Im. 53 538 538 538 0.5 0.5 1.0 4 6 139

Taxandria juniperina M 34 T Obligate 1190 1204 1214 0.3 0.4 1.0 1 5 121

Im. 21 1214 1214 1214 0.4 0.5 0.8 2 7 18

M 131 T Obligate 549 661 781 0.4 0.8 1.0 3 37 136

Im. 13 661 697 1214 0.4 0.9 1.0 4 78 127

Eucalyptus rudis M 79 T Obligate 544 852 1214 0.0 0.3 1.0 0 2 66

Im. 127 549 809 1214 0.0 0.4 1.0 0 3 102

Astartea leptophylla M 170 S Obligate 804 1166 1203 0.4 0.8 1.0 2 16 123

Im. 91 928 1204 1214 0.4 0.8 0.9 3 48 79

Banksia seminuda M 50 T Faculative 697 760 1058 0.1 0.4 0.8 0 3 59

Im. 44 716 760 1214 0.0 0.4 0.8 0 3 18

M 200 T Faculative 1054 1188 1214 0.0 0.1 0.5 0 1 8

Im. 537 1054 1188 1214 0.0 0.0 0.4 0 0 3

Hakea oleifolia M 67 T Faculative 697 781 1214 0.0 0.3 0.5 0 1 11

Im. 86 697 1214 1214 0.0 0.1 0.4 0 0 3

M 55 T Faculative 1098 1190 1210 0.0 0.3 0.5 0 1 9

Im. 9 1064 1190 1190 0.0 0.0 0.4 0 0 3

M 150 T Upland 760 1098 1217 0.0 0.0 0.7 0 0 14

Im. 87 760 760 760 0.7 0.8 0.8 14 27 37

Melaleuca incana M 688 S Upland 661 661 852 0.0 0.0 0.8 0 0 20

Im. 92 640 697 852 0.0 0.3 0.8 0 1 40

Acacia pulchella M 21 S Upland 1188 1188 1214 0.0 0.4 0.4 0 2 3

Im. 30 661 1188 1188 0.0 0.3 0.4 0 2 3

Hovea elliptica M 39 T Upland 964 1204 1217 0.0 0.0 0.1 0 0 1

Im. 51 1098 1204 1217 0.0 0.0 0.4 0 0 4

M 63 S Upland 661 1188 1214 0.0 0.3 0.4 0 1 3

Im. 36 661 1188 1214 0.0 0.3 0.4 0 1 3

M 12 S Upland 697 974 1213 0.0 0.0 0.1 0 0 1

Im. 45 675 760 1188 0.0 0.0 0.2 0 0 1

M 22 S Upland 1188 1188 1188 0.1 0.3 0.4 1 1 3

Im. 42 1188 1188 1204 0.0 0.3 0.4 0 1 2Hibbertia cuneiformis

Mature/

ImmatureSpecies

Agonis flexuosa

Trymalium

odoratissimum subsp.

Leucopogon obovatus

subsp. revolutus

Melaleuca

rhaphiophylla

Callistachys lanceolata

Leucopogon propinquus

Tree/

Shrub

Functional

group

Rainfall Recurrence interval Hydroperiod


85

transect sites, respectively, with extremely narrow rainfall ranges of just 534 to 538

mm pa. The flood plains inhabited by these species have low elevational variation thus

the estimated HPs and RIs varied little within a site. Hydroperiod ranged from 139 to 140

days in M. cuticularis and although HP appears more variable in M. viminea, ranging

from 4 to 139 days, the majority of individuals fell within a much narrower range of

values. In both of these species, juveniles were recorded in high proportions (Table 3.1)

across the surveyed sites. Taxandria juniperina was recorded in six transects, all within

a narrow rainfall band (1190 to 1214 mm pa), with varied HPs and RIs (Table 3.1). The

frequencies of juveniles detected was strongly biased to one transect (T04; Fig. 2.1) where

20 of the total 21 immatures were recorded, thus limiting the power of the model.

For the three remaining obligate riparian species where statistical models did

converge on reliable model estimates, only one species responded significantly to

variation in hydrological parameters. Astartea leptophylla was the only obligate riparian

species to show significant differences in age-class structure along the examined

gradients (Table 3.2a, Fig. 3.5a). The proportion of immature A. leptophylla individuals

increased with greater ΔRI (Table 3.2; Fig.3.5a). This effect was stronger in regions with

higher rainfall, whereas populations at the low rainfall edge of its range had a

comparatively lower proportion of immature to mature individuals, irrespective of ΔRI

(Fig.3.5a). By contrast, Melaleuca rhaphiophylla was recorded right across the

catchment, with population densities greatest near the river mouth and the upper

Catchment. The GLMM identified differences in age-class structure with HP, ΔHP and

their interaction, but due to the very low number of immature individuals recorded, these

parameters were not statistically significant (Tables S3.1; 3.2a). Interestingly,

Eucalyptus rudis was well represented by both mature and immature individuals and

demonstrated the widest rainfall range of the riparian species examined in this study

86

(Table 3.1), but once again age class structure did not differ significantly in relation to

the local hydrological gradients, or regional rainfall gradient (Tables S3.1; 3.2a).

The four facultative riparian species, Banksia seminuda, Agonis flexuosa, Hakea

oleifolia (and to a lesser extent, Callistachys lanceolata) revealed relationships with

differing aspects of the flow regime and the rainfall gradient (Tables S3.1; 3.2a). Over

most of the rainfall range inhabited by B. seminuda, the frequency of immature

individuals was higher in areas with low RI (Fig. 3.7b). Under high rainfall conditions, a

similar pattern was observed, although the proportion of immature individuals was

significantly higher than observed under lower rainfall conditions, regardless of position

on the flood plain (Fig. 3.7b). Agonis flexuosa was the most commonly encountered

species of the higher rainfall regions (>900 mm pa), observed in high abundances within

and across sites (Tables S2.1; 3.1). In the GLMM, RI, ∆RI, their interaction, and

interactions with rainfall, all had significant influences on the relative frequency of

immature vs mature A. flexuosa individuals (Table 3.2a; Fig. 3.5c, d). Unlike

A. leptophylla (an obligate riparian species), in which RI declines had a predominantly

positive effect on juvenile frequency, the greatest reductions in frequency of juvenile

A. flexuosa tended to occur in the areas experiencing the greatest declines in flood

recurrence intervals (Figs. 3.5c, d). For instance, in areas with low recent flood recurrence

intervals (i.e. upland areas experiencing floods 1 in 10 years or fewer), greater declines

in recurrence interval through time (∆RI) resulted in a greater reduction in the proportion

of immature A. flexuosa individuals and this effect was more severe in low rainfall zones

than high rainfall zones (Table 3.2a; Fig 3.5c). By contrast, and similar to the trend in

A. leptophylla, in riparian areas with relatively high RI (i.e. flooding 2 in 5 years), greater

declines in flood recurrence intervals (∆RI) actually resulted in an increase in the

proportion of immature individuals, at least in high rainfall to medium zones but not the

low rainfall zone (Table 3.2a; Fig 3.5d). Substantial declines in the hydroperiod for


87

Hakea oleifolia (reductions exceeding 20 days per year, Fig. 3.6) have resulted in a recent

hydroperiod which ranges from 0 to 6 days per year (10th to 90th percentile, Table 3.1).

The riparian zones that underwent the greatest HP declines had the lowest proportion of

immature Hakea oleifolia individuals (Fig. 3.6), regardless of rainfall zone or recent HP

(Table 3.2a). Although the best fit models for C. lanceolata detected differences in age-

class structure with rainfall, RI and ∆RI, there were very low frequencies of immature

individuals overall (Table 3.1), which severely limited the power of the analyses, and the

resulting trends were not significant.

Just two of the upland species showed differences in recruitment along the

hydrological and rainfall gradients, Trymalium odoratissimum subsp. trifidum (Table 3.2;

Fig. 3.5b) and Melaleuca incana (Table 3.2; Fig. 3.7a). For the other five common upland

species (the Fabaceae shrubs, Acacia pulchella and Hovea elliptica, the Ericaceae heaths

Leucopogon obovatus subsp. revolutus and L. propinquus, and the Dilleniaceae shrub

Hibbertia cuneiformis), the proportion of immature to mature individuals was relatively

consistent across the hydrological and rainfall gradients (Table 3.1) and fitted models

containing rainfall and/or flow regime predictors failed to provide greater explanatory

power than the null models (Table 3.2b). Under higher rainfall conditions, the upland,

sub-canopy tree, T. odoratissimum subsp. trifidum, also showed a slight increase in the

proportion of immature to mature individuals in regions with the greatest ΔRI (Table

3.2b; Fig. 3.5b); an effect similar to that observed in the higher rainfall extent of both

A. flexuosa and A. leptophylla. At the lower extent of the rainfall range for

T. odoratissimum subsp. trifidum, this effect was somewhat skewed (Fig. 3.5b) by the

presence of 82 immature individuals within a single transect (T66; Fig. 2.1), out of a total

of 87 recorded across the catchment, all of which were present within hydrological

regimes higher than observed in the adult population (Table 3.1). Finally, the frequency

of immature to mature M. incana did not differ within the narrow rainfall range it inhabits

88

(Table 3.1). Although the adult M. incana population was observed within the rarely

flooded zones (RI = 0.0; Table 3.1), a higher frequency of immature individuals was

found within lower lying areas of the riparian zone which experienced more frequent

flooding (RI = 0.3, i.e. flooded 1 in 3.3 years; Tables 3.1; 3.2b; Fig. 3.7a).

Table 3.2. Generalised linear mixed effects models testing the relative frequency of

immature to mature individuals of (a) riparian and (b) upland species along the riparian

zones of the Warren River transect as a function of mean annual rainfall (Rn), and either

hydroperiod (HP) plus change in hydroperiod (ΔHP) or recurrence interval (RI) plus

change in recurrence interval (ΔRI). Variation in forest structure is described at transect

and individual level as the covariates, T_PC1 and T_PC2, and I_PC1 and I_PC2,

respectively. The proportion change in variance (PCV) for the random effect (transect

identity) is calculated between the null and final models. The Akaike Information

Criterion (AICc) is a measure of fit scaled to the number of parameters in the model.

R2GLMM(m) is the marginal variance explained by all fixed factors and R2

GLMM(c) is the

conditional variance explained by both fixed and random factors (Nakagawa and

Schielzeth 2013). NA indicates a term was not tested due to collinearities within the fixed

predictor set. In species where the model fit was not better than the null model (Table

S3.1), results are shown for the null model only. Model coefficients highlighted in bold

indicate significant predictors. Note, some models failed to converge due to insufficient

variation within the tested environmental variables, or age classes and are therefore not

presented for Melaleuca cuticularis, Melaleuca viminea, and Taxandria juniperina.

†denotes an obligate riparian species. ►


89

(a)

Rip

ari

an

sp

ecie

sM

ela

leu

ca

rha

ph

iop

hyll

a†

Ast

art

ea

lep

top

hyll

a†

Eu

ca

lyp

tus

rud

is†

Ag

on

is fle

xu

osa

Ba

nk

sia

sem

inu

da

Ha

kea

ole

ifo

lia

Ca

llis

tach

ys

lan

ceo

lata

Fix

ed

eff

ects

n =

144

n =

261

n =

206

n =

737

n =

94

n =

153

n =

64

Inte

rcep

t (n

ull

)-3

.29

[-5

.39

, -1

.18

]-1

.73

[-3

.25

, -0

.21

]0.4

5 [

-0.1

4, 1.0

3]

0.6

4 [

0.0

6, 1

.22

]-0

.95 [

-2.5

3, 0.6

3]

-0.8

4 [

-1.9

1, 0.2

5]

-1.9

[-3

.14

, -0

.72

]

Inte

rcep

t (fu

ll)

-9.9

0 [

-20.0

8, 0.2

9]

-1.0

1 [

-1.9

9, -0

.03

]0.5

7 [

-0.1

7, 1.3

1]

-0.1

8 [

-1.0

2, 0.6

6]

-0.9

3 [

-2.1

7, 0.3

1]

-4.0

8 [

-8.0

6, -0

.09

]

I_P

C1

3.5

4 [

-0.4

0, 7.4

7]

-1.4

1 [

-2.3

7, -0

.45

]-1

.55

[-2

.77

, -0

.32

]

I_P

C2

T_

PC

1N

AN

A-3

.21 [

-5.1

7, -1

.25]

NA

T_

PC

2N

AN

AN

AN

AN

A

HP

-4.4

7 [

-10.4

6, 1.5

2]

ΔH

P-1

.34 [

-11.7

2, 9.0

4]

1.2

9 [

0.3

6, 2

.22

]

HP

: Δ

HP

-23.7

8 [

-57.3

2, 9.7

7]

RI

0.5

8 [

-0.0

6, 1.2

1]

-1.7

4 [

-2.9

9, -0

.50

]-4

.24 [

-9.5

1, 1.0

4]

ΔR

I-1

.28

[-2

.17

, -0

.39

]0.4

1 [

-0.1

5, 0.9

6]

-3.4

3 [

-7.9

2, 1.0

7]

Rn

2.1

6 [

0.2

1, 4

.10

]0.2

6 [

-1.1

5, 1.6

7]

4.2

9 [

1.9

5, 6

.64

]N

A-3

.03 [

-7.1

6, 1.1

1]

RI:

ΔR

I-1

.84

[-2

.79

, -0

.89

]

RI:

Rn

2.7

0 [

1.2

8, 4

.12

]

Rn

: Δ

RI

-1.4

7 [

-2.8

5, -0

.08

]

VC

fo

r ra

nd

om

eff

ects

(Tra

nse

ct)

106.6

2.8

72

1.4

64

2.2

48

0.2

831

2.8

82.0

16

VC

fo

r F

ixed

eff

ects

63.2

22.4

01.0

53.3

00.9

07.7

9

PV

C(T

ran

sect

)-2

185.1

%60.9

%-6

6.3

%93.4

%-2

9.8

%-1

682.5

%

R2

GL

MM

(m)

0.0

%28.0

%16.0

%48.0

%12.8

%59.5

%

R2

GL

MM

(c)

0.1

%61.6

%50.1

%52.1

%53.5

%74.9

%

AIC

c(N

ull

)86.7

247.3

257.7

773.2

109.5

184.3

55.9

AIC

c(F

ull

)83.7

233.7

739.6

98.2

175.1

50.2

Tab

le 3

.2.

90

(b)

Up

lan

d s

pe

cie

sA

ca

cia

pu

lch

ell

aH

ibb

ert

ia

cu

neif

orm

isH

ovea

ell

ipti

ca

Leu

co

po

go

n o

bo

va

tus

su

bsp

. re

vo

lutu

s

Leu

co

po

go

n

pro

pin

qu

us

Try

ma

liu

m o

do

rati

ssim

um

su

bsp

. tr

ifid

um

Mela

leu

ca

in

ca

na

Fix

ed

eff

ects

n =

51

n =

64

n =

90

n

= 5

7n

= 9

9n

= 2

37

n =

780

Inte

rcep

t (n

ull

)-0

.54 [

-4.8

7, 3.7

9]

0.8

7 [

-0.0

4, 1.7

8]

0.5

2 [

0.7

5, 1

.79

]1

.11

[-0

.18

, 2

.39

]-0

.69

[-1

.34

, -0

.05

]-3

.77

[-6

.33

, -1

.20

]-1

.14

[-2

.24

, -0

.04

]

Inte

rcep

t (fu

ll)

1.7

7 [

0.1

7, 3

.36

]-2

.05

[-2

.89

, -1

.14

]-2

.16

[-3

.67

, -0

.66

]

I_P

C1

-1.6

3 [

-3.2

6, -0

.001]

0.8

8 [

0.2

4, 1

.51

]

I_P

C2

1.7

7 [

0.2

8, 3

.25

]

T_

PC

1N

AN

A

T_

PC

2N

AN

A

HP

NA

ΔH

PN

A

HP

: Δ

HP

RI

NA

1.1

0 [

0.1

0, 2

.10

]

ΔR

IN

A-0

.36 [

-1.4

5, 0.7

2]

Rn

-5.5

5 [

-7.1

0, -4

.00

]N

A

RI:

ΔR

I

RI:

Rn

Rn

: Δ

RI

-2.9

8 [

-5.2

5, -0

.71

]

VC

fo

r ra

nd

om

eff

ects

(Tra

nse

ct)

12.1

31.5

67

0.6

446

1.8

65

0.3

155

0.0

4.4

21

VC

fo

r F

ixed

eff

ects

0.7

810.9

10.7

6

PV

C(T

ran

sect

)-8

23.9

%100.0

%-4

9.5

%

R2

GL

MM

(m)

13.9

%76.8

%9.0

%

R2

GL

MM

(c)

41.7

%76.8

%61.2

%

AIC

c(N

ull

)68.9

86.0

127.1

55.4

132.7

144.8

446.8

AIC

c(F

ull

)83.4

134.9

437.8

Tab

le 3

.2. C

onti

nued

.


91

Fig. 3.5. Variation in the relative frequency of immature (1) to mature (0) individuals of

tree and shrub species modelled as a function of mean annual rainfall (mm pa,

percentiles), recent flood recurrence interval (for the period 2001 to 2010), and the change

in recurrence interval (between two the ten-year periods 1980 to 1989 and 2001 to 2010).

Flood recurrence interval is the probability that individuals are likely to be inundated at

least once in any one calendar year, where 1 indicates annual flooding and 0 indicates

individuals were never flooded. The fitted lines (± 95% confidence intervals) represent

the 10th, 50th and the 90th percentiles of model predictions from binomial generalised

linear mixed effects models. Note that percentiles for each predictor vary between species

because each species is distributed over a distinct rainfall or hydrological range. Models

are presented for (a) the riparian shrub Astartea leptophylla; (b) the understorey tree

Trymalium odoratissimum subsp. trifidum; and the canopy tree Agonis flexuosa under (c)

the 90th percentile and (d) the 10th percentile of the recent recurrence interval.

92


Hakea oleifolia modelled as a function of the change hydroperiod, i.e. the mean number

of days per year that individuals were flooded in two contrasting ten-year periods 1980

to 1989 and 2001 to 2010. The fitted line (± 95% confidence intervals) represents model

predictions from a binomial generalised linear mixed effects model.


(a) Melaleuca incana and (b) Banksia seminuda modelled as a function of rainfall and

recent flood recurrence interval (periods between 2001 to 2010). Flood recurrence

interval is the probability that an individual is inundated at least once during a calendar

year, i.e. 1 flooded annually to 0, did not flood over the selected period. The fitted line (±

95% confidence intervals) represents model predictions from binomial generalised linear

mixed effects models.


93

3.4 Discussion

Recruitment failure at a species range margin can be indicative of a change in climatic

optima and advanced warning of an impending climate change induced range shift. In

comparison to temperature induced shifts, there are few examples of rainfall induced

shifts globally, in part due to high uncertainty in rainfall predictions (Lenoir and Svenning

2015), but also greater complexity of species range determinants in lowland species where

moisture tends to be a greater limiting factor than temperature. Utilising one of the

world’s most striking geographically-stratified rainfall gradients, that has undergone one

of the greatest observed declines in recent rainfall, I tested the effect of recent streamflow

decline on the age-class structure of riparian plant species in SWWA. I show that the

relative frequencies of immature versus mature individuals of a number of riparian

species differ significantly with the magnitude of divergence from the historical

hydrological regime. At the drier (low rainfall) margins of species ranges, declines in

streamflow were a key driver of reduction in the frequency of immature individuals,

indicative of recruitment failure and impending range contraction at the range margins.

At the higher rainfall margins of species ranges, however, juvenile abundance actually

increased in response to streamflow declines in a number of species, suggesting that they

are expanding their ranges into riparian habitats where they were historically limited by

high flood inundation regimes. In contrast to riparian species, the majority of the upland

species examined here, show little in the way of recruitment responses to changing

hydrological gradients or regional rainfall gradients. This consistency in recruitment

could indicate that the river may be stabilising recruitment processes across the current

distributions of upland species from the regional rainfall declines (i.e. buffering

individuals from climate change). Here, I discuss these findings and their implications for

ongoing management and species conservation in a region projected to face further,

significant rainfall declines.

94

3.4.1.1 Geographic shifts in climatic optima of riparian species

The declines in streamflow observed over the past 30 years have resulted in a marked

change in recruitment for riparian species in response to declining flood recurrence

interval (ΔRI) or reductions in hydroperiod (ΔHP). Declines in RI interacted with rainfall

for many species, with low rainfall conditions exacerbating declines in recruitment,

whereas recruitment increased under high rainfall in localities with the greatest reductions

in RI. In the facultative riparian species, A. flexuosa, for example, the relative frequency

of immature individuals declined significantly with decreasing flood recurrence interval,

except where recent RI was high in the regions of the catchment under greater rainfall.

The decline in frequency of juveniles was more apparent at the lower rainfall extent.

Similarly, the obligate, and facultative riparian species, A. leptophylla and B. seminuda

too, show lower frequencies of juveniles at the lower limits of their rainfall range. While

in both facultative species A. flexuosa and B. seminuda, juveniles were present

throughout their range, albeit in lower proportions in the lower rainfall extent, there were

no juvenile A. leptophylla recorded above transect T54, at ca 850 mm pa, despite adult

populations ranging out to T80, at ca 640 mm pa. Additionally, and potentially of greater

concern, the percentage of juveniles across the sampled populations of M. rhapiophylla

and C. lanceolata was just 9% and 14% respectively. Cumulatively, the results presented

here demonstrate a lower density of juveniles at the drier extent of these species ranges

and possibly indicative of a contraction in range and a shift in their climatic optima

(VanDerWal et al. 2009). Failure to recruit in the drier extent could be attributed to the

lower rainfall itself, or to the greater intermittency of surface waters at the lower rainfall

sites (Stromberg et al. 2005). In the Murray-Darling basin in South Australia, Jensen et

al. (2008) found that although E. camaldulensis, and E. largiflorens seedlings germinated

readily in flood debris following floodwater recession, the survival rates of seedlings were

significantly higher in rain triggered germination events and with greater availability of


95

surface waters. Alternatively, a comprehensive examination of the rate of soil moisture

draw-down in the establishment of US riparian species (i.e. during the post spring flood

peak drying period), showed that seedlings of obligate riparian species were sensitive to

the rate of water drying and retreat of soil moisture (Stella and Battles 2010a, Stella et al.

2010b).

The major assumption of examining distributions at a single time point to deduce

range mismatch between juvenile and adult populations, is that differences are indicative

of a shift in climatic optima rather than the natural divergence between the recruitment

niche and the adult niche (Grubb 1977). In a recent study examining range mismatch

between seedlings and adult forest trees across Slovakia, Máliš et al. (2016) showed that

the differences in range between age-classes was vastly different, but critically, stable

over a 30-year resurvey period strongly suggesting ontogenic shifts in niche requirement

rather than climate induced range shifts. This phenomena is particularly apparent in

riparian systems, where early establishment is highly dependent on surface, and shallow

soil water until root systems gain access to permanent groundwater sources (Mahoney

and Rood 1998, Stella et al. 2010b). Moreover, mature vegetation has the potential to

significantly alter its own flow regime over its lifetime by redirecting currents and altering

depositional processes (Dixon et al. 2002, Corenblit et al. 2007, Merritt et al. 2010,

Osterkamp and Hupp 2010). Here, by including estimates of recent hydroperiod and

recurrence interval as independent parameters from the observed changes over time, my

results strongly suggest that it is the change in streamflow rather than (or in addition to)

the absolute streamflow driving the range mismatch. Further investigation however,

would be beneficial to determine the nature of the declining rates of recruitment in the

low rainfall regions of the catchment. While the relationships with streamflow decline

suggest recruitment failure could be due to drying conditions restricting seedling

establishment as discussed here, it could also be attributed to lower seed production from

96

stressed mature plants, lower pollination rates owing to fewer individuals (e.g.

A. leptophylla had lower population densities in the lower rainfall extent of its range;

Edmands 2007), changes to biotic interactions, or a number of these factors acting

synergistically.

3.4.1.2 Evidence for narrowing of the riparian corridor

In contrast to the reductions observed in juvenile abundance observed under lower rainfall

conditions, declining flood frequencies under higher rainfall conditions, increased the

relative proportion of juveniles to mature individuals particularly in A. leptophylla and

A. flexuosa. Increases in seedling abundances or vegetation density and cover have been

observed widely as a result of flow reduction due to damming or water extraction

(Shafroth et al. 2002, Gordon and Meentemeyer 2006), particularly within facultative

species (Rood et al. 2010). The initial increase in vegetation cover post-damming, is

principally attributed the increases in the areas suitable for seedling establishment with

declining flood waters, i.e. moist, damp sediments, as well as a reduction in the erosive

flows seasonally clearing establishing seedlings (Mahoney and Rood 1998, Taylor et al.

1999, Johnson 2000, Polzin and Rood 2006, Stella et al. 2010a). Elsewhere, initial

increases in seedling abundance following the reduction of streamflow has resulted in a

higher density of vegetative cover, and a narrowing of the river channel (Rood et al.

2010). The increases in the proportions of juveniles observed in the areas of greatest

deficit, strongly suggests the riparian corridor may be beginning to narrow; a repeat of

survey of selected sites in the future would confirm this.

3.4.1.3 Stability in the upland populations

In five of the seven upland species examined, the distribution of immature and mature

individuals did not differ with regard to metrics describing aspects of streamflow or with

the regional rainfall gradient. Notwithstanding the hydrological parameters, rainfall is

considered one of, if not the most important abiotic determinants of species distribution


97

in the region (Lyons et al. 2000, Hopper and Gioia 2004, Gioia and Hopper 2017).

Consistency in the proportion of immature to mature individuals across the rainfall

gradient indicates stable range margins within the riparian zones. In species where the

surveyed area included the eastern most limits of their distribution such as the upland

shrubs L. propinquus, L. obovatus subsp. revolutus and H. elliptica (see

https://florabase.dpaw.wa.gov.au) these results may be indicative of the river buffering

species from regional rainfall declines observed to date (Reside et al. 2014, McLaughlin

et al. 2017). Although, further examination of age-class structure across the non-riparian

extent of their ranges is required to substantiate these findings. While there was no rainfall

effect on frequency of juveniles of M. incana, a higher proportion of juveniles were

observed in the high RI riparian platforms relative to adults. The lack of significance in

ΔRI however, indicate that these results reflect differences in niche requirements of

juveniles (Grubb 1977, Mahoney and Rood 1998). Alternatively, that M. incana could be

at equilibrium where seedlings germinate in less optimal conditions, but fail to become

established as sites flood intermittently, and juveniles are cleared before they reach

maturity (Johnson 2000).

3.4.1.4 Resilience in a keystone riparian species

Of the species examined, obligate riparian species E. rudis demonstrated the widest,

longitudinal distribution. Curiously, none of the hydrological or climatic parameters

examined here, including the covariables describing light and microclimate variation and

surrounding forest structure, explained patterns in juvenile establishment. As probably

the most iconic riparian species of the SWWA, the fact that neither HP and RI were

significant in describing the of the age-class structure is surprising, but confirms the

results of Pettit et al. (2001) who suggested that E. rudis can establish anywhere on the

floodplain. In contrast to the other species examined here, E. rudis demonstrates a shift

in its phenotype across its range. Under higher rainfall conditions it grows into a tall (up


98

to 30 m), single trunk form, in contrast to the ‘mallee’ like form with multi-stemmed

trunk, rarely over 15 m high, and with tougher, more sclerophyllous leaves (A. Watt.

Unpublished data) common in the upper tributaries. The stark contrast in phenotype

across the extent of the range provides a mechanism to explain the wide distribution of

the species. Whether this variation is indicative of genetically distinct, locally adapted

populations or plastic responses to the environmental conditions (Nicotra et al. 2010,

Hoffmann and Sgrò 2011) warrants further attention if we are to understand the apparent

resilience to streamflow and rainfall declines observed to date.

3.4.1.5 Implications for climate change and management

Over the past 30 years, the riparian vegetation of the Warren Catchment has been

subjected to reductions in mean hydroperiod of up 27 days per year and sites are becoming

inundated over fewer winters (with a deficit of up to 3 years out of 10). The results

presented here demonstrate that these declines are undoubtedly affecting recruitment in a

number of functionally important riparian species. At the dryer extent of species ranges,

declines have resulted in lower proportions of juveniles to mature individuals, potentially

presenting early warnings of a longitudinal range contraction. While at the higher rainfall

extent of the catchment, increasing frequencies of juveniles on riparian plains

experiencing declines in RI indicate the expansion of the riparian vegetation on to areas

previously uninhabitable, and potentially narrowing of the river channel. Downscaled

climate models over the SWWA project declines of between 5 and 75 fewer flow-days

per year by 2030, on top of the deficits already observed (Barron et al. 2012). Given the

apparent shifts in climatic optima already observed here, further flow reductions are likely

to significantly impact the riparian vegetation. How these impacts manifest remains to be

seen, but the results presented indicate that we will likely observe a significant

longitudinal contraction of range of the riparian species. Moreover, as the majority of the

riparian species (both facultative and obligate) did not show an upper rainfall limit to their


99

distribution, i.e. all species were observed within the lower reaches of the river, there is

almost no potential for compensatory range expansion.

Despite these observations, I do not anticipate a complete collapse of the riparian

flora for a number of reasons. First, the projections for summer rainfall are highly

uncertain (Hope et al. 2015) but, have the potential to ease summer drought conditions

for seedlings or trigger sporadic wide-scale recruitment events. For example, a cyclonic

depression observed during January 1982 is believed to be responsible for a large

recruitment event of M. rhaphiophylla, and E. rudis throughout SWWA (Pettit et al.

2001). As species with serotinous seed storage (canopy storage), they have seed available

much of the year to exploit unseasonably wet events (Pettit and Froend 2001a). Second,

even with significant reductions in river flows, the river is unlikely to cease to flow

completely (Barron et al. 2012) thus habitat will be available to the riparian species, albeit

over a smaller geographic range. Instead, I expect we will see a compositional change to

a greater proportion of mesic, facultative and upland species as further reductions in the

inhibitory high flow events are observed (Merritt and Poff 2010, Stromberg et al. 2010,

2012). Finally, in contrast to many of the rivers cited here, the Warren River is free-

flowing in that the main channel itself is not dammed, thus there is limited potential to

intervene and ensure ecological flows are sufficient to maintain riparian vegetation (as

prescribed elsewhere, e.g. Merritt et al. 2010, Poff et al. 2010, Stella et al. 2010, Miller et

al. 2013). Instead, if we face the severe streamflow declines projected under the higher

emissions scenarios, increasing the resilience of these species may require more proactive

measures.

100

3.5 Supplementary material

Table S3.1. Model selection using AICc scores to compare generalised linear mixed

effects models testing the proportion of immature to mature individuals as a function of

mean annual rainfall (Rn) and inundation (comparing the fit of a hydroperiod model,

containing hydroperiod (HP) and change in hydroperiod (ΔHP), versus the fit of a

recurrence interval model, containing recurrence interval (RI) and change in recurrence

interval (ΔRI)). The HP and RI measures were highly collinear so could not be included

in the same model. Variation in forest structure is described at transect and individual

level as the covariates T_PC1 and T_PC2, and I_PC1 and I_PC2, respectively. k denotes

the number of parameters included in the model, AICc is a measure of fit scaled to the

number of parameters in the model. Models with the lowest AICc denote the best model

fit: the most parsimonious model < 2 AIC was selected, and is indicated in bold.

Model - Acacia pulchella k Log likelihood AICc Δ AICc

I_PC2 3 -30.54 67.60 0.00

T_PC1 + I_PC2 4 -29.52 67.90 0.31

T_PC1 + I_PC1 4 -29.74 68.35 0.75

I_PC1 3 -30.96 68.43 0.83

Rn 4 -30.05 68.96 1.37

Null 2 -32.43 69.11 1.51

I_PC1 + I_PC2 4 -30.24 69.35 1.75

Model - Agonis flexuosa

I_PC1 + Rn + RI + ΔRI + Rn: RI + Rn: ΔRI + RI: ΔRI 9 -359.84 737.92 0.00

Rn + RI + ΔRI + Rn: RI + Rn: ΔRI + RI: ΔRI 8 -361.81 739.81 1.89

I_PC1 + Rn + RI + ΔRI + Rn: RI + Rn: ΔRI + RI: ΔRI + RI: ΔRI: Rn 10 -359.78 739.86 1.94

Null 2 -384.58 773.17 35.25

Alternate best fit

HP + ΔHP + Rn + ΔHP: Rn 6 -367.69 747.49 -

Model - Astartea leptophylla

I_PC1 + Rn + ΔRI 5 -111.87 234.00 0.00

I_PC1 + T_PC1 + Rn + ΔRI 6 -111.10 234.50 0.55

I_PC1 + Rn + RI + ΔRI + Rn:ΔRI 7 -110.57 235.60 1.61

I_PC1 + Rn + RI + ΔRI 6 -111.77 235.90 1.89

I_PC1 + I_PC2 + Rn + ΔRI 6 -111.81 236.00 1.98

I_PC1 + Rn + ΔRI + Rn:ΔRI 6 -111.81 236.00 1.98

Null 2 -121.63 247.32 13.35

Alternate best fit

T_PC1 + I_PC1 + Rn 5 -114.63 239.50 -

Model - Banksia seminuda

T_PC1 + Rn + RI + ΔRI + RI:ΔRI 7 -41.69 98.68 0.00

T_PC1 + Rn + RI 5 -44.08 98.85 0.17

T_PC1 + T_PC2 + Rn + RI 6 -43.02 99.00 0.32

T_PC1 + T_PC2 + Rn + RI + ΔRI + RI:ΔRI 8 -40.74 99.18 0.50

T_PC1 + I_PC1 + Rn + RI 6 -43.62 100.20 1.52

T_PC1 + Rn + RI + ΔRI + Rn:ΔRI + RI:ΔRI + RI:Rn + RI:ΔRI:Rn 10 -38.79 100.24 1.56

T_PC1 + Rn + RI + ΔRI 6 -43.67 100.30 1.62

T_PC1 + T_PC2 + Rn + RI + ΔRI 7 -42.51 100.31 1.63

Null 2 -52.77 109.68 11.00

Alternate best fit

T_PC1 + HP + Rn 5 -44.48 99.63 -


101

Table S3.1. continued.

Model - Callistachys lanceolata k Log likelihood AICc Δ AICc

Rn + RI + ΔRI 5 -20.08 51.20 0.00

Rn + RI + ΔRI + Rn: ΔRI 6 -19.11 51.69 0.49

I_PC1 + Rn + RI + ΔRI 6 -19.30 52.08 0.88

Rn + RI + ΔRI + RI: ΔRI 6 -19.80 53.08 1.88

Rn + RI + ΔRI + Rn: RI 6 -19.82 53.10 1.90

I_PC2 + Rn + RI + ΔRI 6 -19.83 53.13 1.93

Null 2 -25.96 56.11 4.91

Alternate best fit

ΔHP 3 -23.99 54.38 -

Model - Eucalyptus rudis

ΔRI 3 -125.06 256.24 0

ΔRI + Rn + ΔRI:Rn 5 -123.15 256.59 0.35

I_PC2 + ΔRI 4 -124.35 256.91 0.67

I_PC2 + ΔRI + Rn + ΔRI:Rn 6 -122.33 257.09 0.85

T_PC1 + I_PC2 + ΔRI 5 -123.40 257.10 0.86

T_PC1 + ΔRI 4 -124.59 257.38 1.15

Null 2 -126.83 257.72 1.48

T_PC1 + I_PC2 + ΔRI + Rn 6 -122.68 257.78 1.55

T_PC1 + ΔRI + Rn 5 -123.76 257.81 1.57

T_PC1 + ΔRI + Rn + ΔRI:Rn 6 -122.73 257.88 1.65

T_PC1 + I_PC1 + ΔRI + Rn + ΔRI:Rn 7 -121.70 257.96 1.72

Alternate best fit

I_PC2 3 -126.10 258.33 -

Model - Hakea oleifolia

I_PC1 + HP + ΔHP 5 -82.26 174.93 0

I_PC1 + ΔHP 4 -83.54 175.35 0.42

I_PC1 + HP + ΔHP + HP: ΔHP 6 -82.14 176.86 1.93

Null 2 -90.13 184.35 9.42

Alternate best fit

I_PC1 + RI 4 -85.40 179.08 -

Model - Hibbertia cuneiformis

I_PC2 3 -38.72 83.85 0.00

I_PC2 + I_PC1 4 -37.61 83.90 0.05

I_PC2 + Rn + RI + ΔRI + Rn: RI + Rn: ΔRI 8 -32.91 84.44 0.59

I_PC2 + ΔRI 4 -38.46 85.59 1.74

Null 2 -41.02 86.25 2.40

Model - Hovea elliptica

Rn + RI + Rn: RI 5 -57.70 126.12 0.00

I_PC1 + Rn + RI + Rn: RI 6 -56.61 126.24 0.12

RI 3 -60.28 126.84 0.72

Null 2 -61.54 127.23 1.11

I_PC1 3 -60.60 127.47 1.36

I_PC1 + RI 4 -59.61 127.70 1.58

Model - Leucopogon obovatus subsp. revolutus

Null 2 -25.71 55.64 0.00

I_PC2 + Rn 4 -23.65 56.07 0.42

Rn 3 -24.86 56.18 0.53

I_PC2 3 -24.93 56.31 0.66

Rn + RI 4 -24.04 56.85 1.20

T_PC2 + Rn 4 -24.35 57.48 1.83

I_PC1 3 -25.58 57.62 1.98

Model - Leucopogon propinquus

Null 2 -64.36 132.84 0.00

I_PC2 3 -63.82 133.90 1.06

I_PC1 3 -64.21 134.68 1.84

Rn 3 -64.21 134.68 1.84

T_PC1 3 -64.30 134.84 2.00

102

Table S3.1. continued.

Model - Melaleuca cuticularis k Log likelihood AICc Δ AICc

NA - model convergence failure

Model - Melaleuca incana

I_PC1 + I_PC2 + RI 5 -213.78 437.63 0.00

I_PC1 + RI 4 -214.90 437.85 0.21

I_PC1 + I_PC2 + RI + ΔRI 6 -212.99 438.10 0.47

I_PC1 + I_PC2 + RI + ΔRI + RI: ΔRI 7 -212.23 438.61 0.98

I_PC1 + RI + ΔRI 5 -214.55 439.18 1.55

I_PC1 + RI + ΔRI + RI: ΔRI 6 -213.72 439.55 1.92

Alternate model - Melaleuca incana

I_PC1 + I_PC2 + ΔHP 5 -213.76 437.60 0.00

I_PC1 + ΔHP 4 -215.12 438.29 0.68

I_PC1 + I_PC2 + HP + ΔHP 6 -213.40 438.91 1.31

I_PC1 + HP + ΔHP 5 -214.47 439.02 1.42

Null 2 -221.40 446.82 -

Model - Melaleuca rhaphiophylla

I_PC1 +I_PC2 + HP + ΔHP + HP: ΔHP 7 -33.98 82.77 0

I_PC2 + HP + ΔHP + HP: ΔHP 6 -35.44 83.48 0.71

I_PC1 + HP + ΔHP + HP: ΔHP 6 -35.87 84.36 1.59

Null 2 -41.37 86.83 5.27

Alternate best fit

ΔRI 3 -40.76 87.70 -

Model - Melaleuca viminea


Model - Taxandria juniperina


Model - Trymalium odoratissimum subsp. trifidum

I_PC1 + Rn + ΔRI + Rn: ΔRI 6 -61.46 135.29 0.00

I_PC1 + Rn 4 -64.61 137.39 2.10

Rn + ΔRI + Rn: ΔRI 5 -63.88 138.03 2.73

I_PC1 + Rn + ΔRI 5 -64.51 139.28 3.98

Rn 3 -67.54 141.19 5.90

I_PC1 3 -67.68 141.46 6.16

I_PC1 + ΔRI 4 -66.85 141.88 6.58

Rn + ΔRI 4 -67.76 143.70 8.40

Null 2 -70.40 144.84 9.55

ΔRI 3 -69.44 144.97 9.68

Alternate best fit

I_PC1 + Rn 4 -64.61 137.39 0.00


103

Fig

. S3.1

. Correlatio

n m

atrix am

ong

fixed

pred

ictors, recen

t hydro

perio

d (H

P_00.s), ch

ange in

hydro

perio

d (H

P_d.s), recen

t recurren

ce interv

al

(RI_

00.s)

and

chan

ge

in

recurren

ce in

terval

(RI_

d.s),

histo

rical m

ean

annual

rainfall

(r1901

_60_m

.s) an

d

chan

ge

in

mean

an

nu

al rain

fall

(declin

e_w

ide.s). A

ll pred

ictor v

alues are cen

tred an

d scaled

to 2

stand

ard d

eviatio

ns. V

alues in

dicate P

earsons co

rrelation

coefficien

t (r).

105

4 Does plasticity confer resilience to a drying climate? An experimental

test of genotype by environment interactions along a rapidly changing

rainfall gradient

4.1 Introduction

The persistence of species in the Anthropocene will depend on their capacity to migrate

with shifting climatic conditions, or to adapt and evolve in situ (Davis et al. 2005, Jump

and Peñuelas 2005). For species with long generation times and limited dispersal ability,

such as many tree species, it is becoming increasingly apparent that the pace with which

the climate is changing far exceeds the potential compensatory rate of migration (Davis

et al. 2005, Aitken et al. 2008, Franks et al. 2014). The vulnerability of long-lived species

to climate change therefore depends not only on their genetic variability and evolutionary

potential to adapt, but crucially their inherent ability to withstand imminent changes in

the short term.

Across the extent of a species range, individuals can express a variety of

phenotypic forms such as the tall, single trunk form of low altitude trees in contrast to the

dwarf, gnarled forms of conspecifics at high altitude (Pryor 1956). Likewise,

phenological differences occur in the date of bud break of deciduous trees in response to

temperature gradients across latitudes (Schreiber et al. 2013) and altitudes (Vitasse et al.

2013). This variation in phenotype can result from genetically fixed differences among

populations or by phenotype plasticity in response to environmental cues. The

mechanisms leading to evolution of trait fixation or plasticity are poorly understood, but

are known to vary between species and populations (Kawecki and Ebert 2004, Leimu and

Fischer 2008, Hereford 2009).

106

Genotypes could be adapted to their local environment through the fixation of

advantageous traits or via varying degrees of plasticity. To be considered locally adapted,

individuals are expected to demonstrate greater fitness than non-local genotypes but also,

confer a fitness disadvantage outside of home conditions, meaning the trait is not

beneficial throughout their range. Broadly, trait fixation is predicted to occur in scenarios

of high spatial heterogeneity, low temporal variability and low gene flow (Kawecki and

Ebert 2004). Indeed, over steep altitudinal gradients, where environmental conditions

vary dramatically over small spatial scales, fixation in traits such as growth rate, cold

resistance (Pryor 1956), and resource allocation to defence (O’Reilly-Wapstra et al. 2013,

Gosney et al. 2016) versus storage (Gauli et al. 2015) have been observed in various

Eucalyptus species, resulting in highly site-specific structuring of genetic variation. Trait

plasticity, on the other hand, is expected under conditions of high environmental and

temporal heterogeneity, and where gene flow is high among populations, and importantly,

in the presence of reliable environmental cues that allow an organism to accurately match

their phenotype (Bradshaw 1965, Schlichting 1986, Sultan and Spencer 2002, Kawecki

and Ebert 2004). For example, in Rana temporaria tadpoles originating from different

locations across an island archipelago, Lind et al. (2010) showed that the magnitude of

plasticity in ontogenic development rate was positively correlated with both habitat

heterogeneity and rate of gene flow among islands. However, translating these patterns

into predictions about the relative contribution of plasticity vs fixation in trait expression

across natural systems has proven more difficult. While both mechanisms behind trait

expression potentially offer resilience to environmental change, the identification of the

underlying mechanisms is critical to elucidating climate adaptation strategies.

While the fixed expression of traits adapted to the environmental conditions in a

particular part of a species geographic range confers an advantage under current climates,

under novel future climates the same traits have the potential to become maladaptive. The

Chapter 4: Plasticity in E. rudis

107

selection of particular genotypes that express adaptations, such as greater water use

efficiency or drought resistance however, could be spread throughout a species range to

enhance overall population viability under novel climates (Aitken and Whitlock 2013,

Prober et al. 2015, Aitken and Bemmels 2016, Montwé et al. 2016). In contrast, plasticity

of plant traits, particularly in long lived species, is predicted to increase population

resistance to climatic changes by buffering the immediate effects of changing climate and

extending the time frame over which a species is able to persist and adapt (Chevin et al.

2010, Nicotra et al. 2010, Reed et al. 2011, Valladares et al. 2014). Although there has

been some discussion regarding the adaptive potential of trait plasticity over longer time

scales (i.e. in buffering selective processes; Ghalambor et al. 2007, Crispo 2008, Chevin

et al. 2010), it is the arguments for phenotypic plasticity playing a pivotal role in

accelerating ecological and microevolutionary change that are gaining traction (West-

Eberhard 2005, Pigliucci et al. 2006, Lande 2009, Chevin and Lande 2010, Chevin et al.

2010, 2013, Nicotra et al. 2010, Dewitt 2016, Levis and Pfennig 2016). However,

generalised predictions and models testing how these mechanisms may facilitate

adaptation to environmental change are limited by a lack of empirical data or unifying

predictive patterns (Nicotra et al. 2010, Valladares et al. 2014, Levis and Pfennig 2016).

In forest ecosystems across the world, increases in temperatures (largely via

raising water demand), and decreases in precipitation are predicted to be the greatest

drivers of tree mortality over the coming decades (Allen et al. 2010). While experimental

examination of the mechanisms underpinning phenotypic trait variation in trees over both

altitudinal (Pryor 1956, Vitasse et al. 2010, 2013, Gauli et al. 2015, Mathiasen and

Premoli 2016) and latitudinal ranges (Schreiber et al. 2013, Benomar et al. 2016, Montwé

et al. 2016) are beginning to shed light on species responses to temperature (Alberto et al.

2013), responses to water availability along natural rainfall gradients are difficult to

isolate and show highly variable responses (Gibson et al. 1995, Li et al. 2000, Cornwell

108

et al. 2007, Richter et al. 2012, Mclean et al. 2014, Breed et al. 2016), particularly with

respect to finer scale patterns and plasticity (Mclean et al. 2014). This challenge might

stem, at least partially, from the fact that much of our understanding of genotype by

environment (G×E) interactions in trees is derived from forestry trials that were originally

designed to identify stock for plantations and to conduct broad scale assessments of

genetic variability across wide (often continental scale) species distributions (e.g. Warren

et al. 2006, Montwé et al. 2016, and reviewed in Alberto et al. 2013). While these studies

are often representative of a wide range of source climates, they typically lack replication

in transplanted environmental space with which to assess the extent of trait plasticity (but

see Wang et al. 2006), or whether there are thresholds at which a change in environment

drives the fixation (or flexibility) of the examined traits.

In an applied sense, the manipulation of these ecological and evolutionary

processes forms the basis of management strategies put forward to increase resilience and

adaptive capacity in restored ecosystems (Prober et al. 2015, Christmas et al. 2016). At

the forefront of adaptive restoration and reforestation planning, climate provenancing (i.e.

assisted gene migration) strategies propose to selectively harvest seed from regions of

climatic space that are similar to the projected future climates at the transplanting site

(Aitken and Whitlock 2013, Prober et al. 2015). The practice aims to utilize the

heritability of specific traits, such as those that confer greater drought resistance

(Dutkowski and Potts 2012, Breed et al. 2016, Montwé et al. 2016) or increased

productivity under higher temperatures (Schreiber et al. 2013, Montwé et al. 2016), in

order to increase resilience to longer term environmental changes. While for decades the

commercial forestry industry has been undertaking these practices to select populations

suited to regions outside of the natural range (e.g. Illingworth lodgepole pine experiment;

Wang et al. 2010), it is only in the last decade that the practice has gained traction in the

conservation and restoration literature (Prober et al. 2015, Aitken and Bemmels 2016). In


109

theory, individuals carrying an adaptive trait are transplanted to the at-risk population,

where the trait confers an advantage over local individuals and spreads through the

population. The practice is not without risk, however, because it can inadvertently lead to

the introduction of correlated traits that might be maladaptive to other local conditions,

such as soil type or herbivore defence at the transplant site, even if an accurate climatic

match is made. There is also the risk that assisted migration of locally-adapted genotypes

might lead to outbreeding depression and potentially a decline in the viability of the

targeted population (Aitken and Whitlock 2013). To identify populations that differ

significantly in climatic space, it is common practice to search widely over geographic

gradients that might have little or no genetic connectivity (Gibson et al. 1995, Li et al.

2000, Montwé et al. 2016). While the first generation of transplanted individuals might

flourish under the ‘optimal’ climatic conditions, the hybrid offspring of two genetically

distant individuals can have low genetic compatibility and low viability (Edmands 2007,

Aitken and Whitlock 2013). Therefore, in spite of the widespread occurrence of locally-

adapted populations (Leimu and Fischer 2008, Hereford 2009), climate provenancing

should not necessarily be broadly implemented without assessing the risk to the

populations in need of protection. Instead, in reviewing the practice of climate

provenancing, Aitken and Whitlock (2013) suggested that due to high uncertainty it ought

to be limited to species with long generation times with limited short term evolutionarily

potential, and species of high economic, ecological or conservation significance.

Riparian trees perform significant economic and ecological services (Costanza et

al. 1997, Davies 2010) and despite their arguably, stronger dependence on moisture

availability than many widespread forest species, they have received surprising little

scientific attention in this space internationally (but see Gibson et al. 1995, Dillon et al.

2015). In regions where the climate is drying, in particular the Mediterranean-type and

semi-arid climate zones, streamflow too is declining, placing riparian flora under stress

110

(Chapter 3). The south-west of Western Australia (SWWA) is experiencing one of the

most clearly defined and rapidly changing rainfall decline trends observed worldwide

(Hennessy et al. 2007). Already the region has suffered a 10 to 16% decline in mean

annual rainfall (Bates et al. 2008) which has culminated in declines in stream flow by up

to 50% since the 1970’s (Silberstein et al. 2012). By 2030, projections suggest further

declines of up to 13% in mean annual rainfall, and to 40% decline in annual runoff

(Silberstein et al. 2012), suggesting that the water-dependent ecosystems may be

particularly at risk and stressing the urgency identifying solutions to increase their

adaptive capacity.

Here, I use a series of reciprocal transplant experiments to test the mechanistic

basis of phenotypic trait variation observed across the significant rainfall gradient in the

SWWA, but within the narrow geographic distances of individual river catchments.

Across almost the complete ombrographic distribution of woodland ecosystems in

southwest Australia (ca 400 – 1400 mm pa; www.ala.org.au), Eucalyptus rudis,

dominates the riparian canopy community, yet varies dramatically in phenotypic traits

with changing precipitation regime. Under conditions of higher rainfall, E. rudis typically

has a tall (20 – 30 m) single-stemmed growth form (Fig. 4.1a), with large leaves, whereas

in low rainfall regions it typically has a shorter (5 – 15 m), multi-stemmed ‘mallee’ like

growth form (Fig.4.1b) with smaller, more sclerophyllous leaves. The visible phenotypic

differences observed within a single catchment provides the opportunity to, separate

environmental structuring of phenotypic variation from the broad geographic structuring

of population divergence. Here, I focus on a continuous population spanning the main

river channel of the Warren River Catchment (Fig. 4.1c), which encompasses

approximately 5% of E. rudis geographic range, but 75% of its rainfall range (Fig. 4.1c).

I aim to (1) identify whether morphological traits in E. rudis seedlings are fixed,

potentially indicating adaptation to their source rainfall or responding plastically to

http://www.ala.org.au/


111

environmental cues at the transplant site; (2) whether plasticity is favoured over a fixed

response (or vice versa), under source environments of greater (or lesser) stress; while (3)

investigating the presence of dry adapted genotypes that could be utilized for climate-

adaptive provenancing in future restoration projects (and, therefore, increase the adaptive

capacity of the system).

4.2 Methods

4.2.1 Study species

Eucalyptus rudis is widely distributed across the SWWA and is one of the major canopy

forming species of the riparian zone. Water availability appears to drive distribution,

either directly as rainfall or indirectly in seasonally flooded wetlands or rivers, and it is

only rarely observed in regions with lower than 400 mm rainfall per annum (Fig. 4.1).

The population along the Warren River and its major tributary the Tone River, is

distributed more or less continuously along the riparian zone (Table S2.1), where it

inhabits flood plains and seasonally damp regions that have had historically low

conversion rates for agriculture. While spatially continuous, populations of other

Eucalyptus species have been shown to be effectively independent at spatial scales over

50 km (Bloomfield et al. 2011, Breed et al. 2012) suggesting that there might be the

possibility of genetic divergence across E. rudis populations of the Warren River

Catchment (ca 130 km overland). In contrast to woodland species where seed dispersal

distances are typically small (Gauli et al. 2014), E. rudis exhibits hydrochory (seed

dispersal via flowing water) offering the potential for much greater dispersal distances,

albeit only in a downstream direction, during flooded periods (Pettit and Froend 2001a,

2001b). As genetic analysis has not yet been undertaken in E. rudis, however, actual gene

dispersal distances are unknown.

112

Figure 4.1. Study system in the south west Western Australia (SWWA). (a) The taller,

single trunk growth form of Eucalyptus rudis typical in the mid to high rainfall regions

and (b) shorter, multi-stemmed, mallee like growth form of the lower rainfall regions. (c)

Distribution within native range of E. rudis (Atlas of Living Australia 2016;

www.ala.org.au) across the river basins of the SWWA. (d) Location of seed source sites

(black) and experimentally transplanted sites (green) within the Warren River Catchment.

http://www.ala.org.au/


113

4.2.2 Experimental design

To test for potential genotype by environment (G E) interactions structuring phenotypic

differences among populations of E. rudis along the Warren and Tone rivers, seeds of 31

trees (maternal lineages) were collected from nine source sites. The seeds were then

germinated and grown under glasshouse conditions, before transplanting seedlings into

six common-garden experimental field sites located within natural riparian zones (Fig.

4.1d; Fig. 4.2). While this is a similar G E interaction approach taken in the majority of

common-garden transplant experiments on long-lived species, it should be noted that my

study (like most others involving trees) cannot strictly separate genotype differences from

potential maternal effects on seed resource investment since experimentally separating

these mechanisms requires manipulation over multiple generations (Kawecki and Ebert

2004, e.g. Ǻgren and Schemske 2012, Halbritter et al. 2015) or hand pollination and

selective crossing (e.g. Lopez et al. 2003, Rix et al. 2012). Instead, I refer to genotypic

differences in the sense that the measured traits carry variation that can be attributed to

the maternal lineage.

The full reciprocal transplant design (Fig. 4.2) adapted the two main aspects of

Kawecki and Ebert's (2004) approach to evaluating evidence for locally adapted

genotypes their ‘local versus foreign’ response model, relative to adaptive plasticity in

their ‘home versus away’ model. The home versus away model compares the plasticity

of trait expression in each maternal lineage grown under its home-site conditions relative

to the different environmental conditions experienced across the gradient of transplant

sites. While this comparison demonstrates differences in the traits of each maternal

lineage in response to each level of the novel environment, it does not explicitly test the

trait responses between lineages. To test differences among lineages I used Kawecki and

Ebert's (2004) local versus foreign model. Under their definition, a genotype is considered

to be locally adapted if it outperforms non-local genotypes under natal conditions.

114

Therefore, by comparing the trait responses of the locally-sourced maternal lineages at

each transplantation site against trait responses of maternal lineages from the foreign

source sites I am able to replicate this comparison and test each lineage for differentiation

among genotypes. Moreover, as this rainfall gradient encompasses (or exceeds) the range

of predicted future rainfall decline estimates for 2030, I test whether the populations

currently have the plasticity or resistance to withstand the predicted changes using a space

for time substitution approach. This also allows me to identify whether the ‘local’ source

consistently has an advantage [which is Kawecki and Ebert's (2004) strict definition of

locally adapted meta-populations].

Just as seedlings may respond to altered abiotic conditions by differentially

allocating resources between above and below ground biomass or across a leaf

ontogenetic gradient, it has been suggested that plants may face a trade-off in resource

allocation towards defence over growth in resource poor environments (Coley et al.

1985). For a species such as E. rudis, which is susceptible to seasonally heavy insect

attack (Clay and Majer 2001), the potential loss of leaf tissue resources to insects in poorly

defended individuals may be great enough to mask any other climatic effects under

examination. Moreover, changes to the biotic community such as insect herbivores are

proposed to be at least as threatening as changes to the abiotic environment under climate

change (e.g. Galiano et al. 2010). Thus, to examine whether such a trade-off exists

between growth and defence (and/or if the balance differs among source environments),

an insecticide treatment was applied to test the potential indirect consequences of rainfall

decline on plant fitness.


115

Figure 4.2. Conceptual layout of the translocation experiment, showing transplant sites

T1200 to T550 (where T denotes a transplant site, and the numeric is the approximate

mean annual rainfall at the site, see Fig. 4.1), and maternal lineages (M) nested within

source sites S537 to S1214 (where the numeric indicates mean annual rainfall at the

source site, S, for each maternal lineage, Fig. 4.1). Insecticide was randomly applied to

50% of each M within each transplant site (indicated by intact vs chewed leaves). In the

‘home versus away’ statistical models, trait expression within seedling sources across

transplant sites was compared to estimate plasticity (outlined by the red box). In the ‘local

versus foreign’ statistical models, trait expression of seedling sources from across the

catchment were compared within each transplant site to estimate local adaptation

(outlined by the yellow box). The ‘home’ site allocated for each source is indicated by

blue shading.

116

4.2.3 Seed collection and seedling preparation

Eucalyptus rudis seeds were collected from naturally pollinated wild populations along

the Warren and Tone Rivers between June and December 2013 with permission from the

Western Australian Department of Parks and Wildlife (DPaW; permits: CE004258,

SW015930) and private land owners. Collection sites, hereafter source sites, were spaced

at distances greater than 5 km apart to reduce the likelihood of sampling closely related

individuals among source populations. Within a source site, seed was collected from up

to five trees, and each tree is hereafter referred to as a maternal lineage (Fig. 4.2). Within

source sites, no limitations were set on distances between maternal trees due to the

difficulty in finding seed-bearing trees. Fruit was collected using both ground searches

for recently-fallen branches and a 10 m pole saw to obtain fruiting branches directly from

the canopy. Tall, dense forests in the mid to high rainfall regions created logistical barriers

to canopy seed collection resulting in lower replication in these regions than the dry-

sourced provenances (Figs. 4.1, 4.2). Seeds were collected from nine source sites,

totalling 31 maternal lineages (Fig. 4.2). A summary of environmental conditions at

source sites is provided in Table 4.1. Sources are coded ‘S’ for source with a numeric

estimate of the mean annual rainfall at the site, e.g. S1214 (Figs. 4.1, 4.2).

Table 4.1. Mean climate conditions at transplant sites and seed source sites (Hijmans et

al. 2005).

Transplant site - T550 - T700 - - T800 T1050 - T1150 T1200 -

Source site S538 S549 S547 S697 S781 S809 - - S1166 - S1204 S1214

Rainfall (mm pa) 538 549 547 697 781 809 809 1054 1166 1204 1204 1214

Rn. of warmest qt. (mm) 51 51 51 61 65 66 66 72 78 75 75 75

Rn. of coldest qt. (mm) 239 252 263 354 391 396 396 492 538 553 553 556

Temperature (°C) 15.1 15.1 15.2 15 15.2 15.3 15.2 15.2 15 15.4 15.4 15.6

Temp. of warmest qt. (°C) 20.3 20.3 20.3 19.8 19.9 19.9 19.8 19.7 19.3 19.5 19.5 19.6

Temp. of coldest qt. (°C) 10.3 10.3 10.4 10.6 11 11.1 11 11.1 11.2 11.7 11.7 12.1


117

Seeds were sown under common-garden conditions in a glasshouse at CSIRO

Floreat, in Perth, Western Australia (Fig. 4.1a) on the 24th and 25th January 2014. After

separating seeds from fruit and chaff, 100 seeds per source were individually sown on

potting mix (Baileys Premium potting mix, AKC Pty Ltd) in forestry tubes (40 ×120 mm).

Seeds that failed to germinate, germinated and failed to produce true leaves, or perished

within two weeks of sowing, were replaced on the 14th and 15th of February 2014 (if

sufficient seed was available). Additionally, a haphazard selection of 25 seeds from 30

(of 31) sources was weighed to test for differences in maternal investment (Sartorius M3P

microbalance, precision 0.001 mg).

Seedlings were watered as required with an automated reticulation system and

seed tray positions were rotated every 7 to 10 days to reduce bias due to tray positions in

the glasshouse. Over the five months in the glasshouse seedlings were fertilized (Searles®

Flourish, Native plants) as required and treated once with a low residual surface fungicide

(Tebuconzole 430c, 4Farmers Pty. Ltd). At the end of this period seedlings with fewer

than six true leaves were excluded from the trial. The remaining seedlings were

transported to field station at Manjimup, close to the field transplant sites (Fig. 4.1a, d),

and hardened outside for at least seven days prior to planting. A permit to plant on Crown

land was obtained from DPaW, and transplanted soil was tested by the vegetation health

service (DPaW) which confirmed the absence of environmentally detrimental soil

pathogens such as Phytophthora cinnamomi.

4.2.4 Establishment and maintenance of transplant sites

Transplant sites were established at six locations along the Warren and Tone Rivers

encompassing the full rainfall range of the source sites (Fig. 4.1). Sites were selected to

have a local topography representative of flood plain habitat and a vegetation community

typical for each climatic locality; i.e., an open understorey, the presence of a natural

canopy (including E. rudis) and no evidence of agricultural grazing. An irregular planting

118

grid was established at each of the transplant sites between existing vegetation and

topographic features. Seedlings were spaced at least 1 m away from adjacent

experimental seedlings or other naturally occurring woody vegetation, and at least 0.5 m

from non-woody plants. The climatic conditions at each of the transplant sites are

summarised in Table 4.1. Transplant sites are coded ‘T’ for transplant and a numeric in

the site code indicating the approximate mean annual rainfall at the site T1200, T1150,

T1050, T800, T700 and T550 (Fig. 4.2).

Seedlings from each maternal lineage were stratified by height and randomly

assigned to transplant sites, then randomly split between insecticide treatment and the

water control, and designated a random planting position. At planting, replication for each

maternal lineage ranged from four to eight seedlings within each insecticide treatment at

each transplant site. Lineages which did not produce enough viable seedlings to meet a

minimum four-plant threshold in each treatment and at all sites were transplanted to fewer

sites, prioritising sites nearest to its source and to both the highest and lowest rainfall

transplant sites. Transplanting took place during winter 2014 (23rd June to 3rd July) when

the soil was saturated (as evidenced by runoff and rising river levels). The seedlings were

removed from the tubes and hand planted into wedge-shaped holes with minimal

disturbance of existing groundcover vegetation. Each seedling was labelled using white

printed tags pegged into the ground adjacent to each seedling. Seedlings were not caged

and were thus exposed to wild mammalian herbivores.

Site visits to apply insecticide and monitor survival, growth and herbivore activity

were made fortnightly from September 2014 after the winter high-water levels receded

and allowed access to the transplanting sites. For transplant sites T1200, T1150, T1050,

T800 and T550 insecticide treatments began on the 11th to the 13th September for the

majority of seedlings (partially submerged plants were not treated until the following

visit). High water levels at T700 restricted access until the 28th to 30th September 2014.


119

In the insecticide treatments, a synthetic pyrethroid insecticide with a short half-life

(TEMPO® Residual Insecticide, Bayer, diluted to 6 mL/L, active component:

Betacyfluthrin) was applied to reduce the risk of run-off contaminating the waterways. A

water control was applied to non-insecticide seedlings at the same time as the insecticide

treatment was applied. Both the insecticide and the water control were applied in

quantities adjusted by plant size, by spraying each plant until all the leaf surfaces were

visibly wet. Treatments were applied at fortnightly intervals until January 2015 when high

seedling mortality at the low rainfall sites, principally due to kangaroo grazing, resulted

in a downscaling of the experimental design.

4.2.5 Measured responses

The responses of seedlings to experimental transplantation were measured using survival,

vertical growth (total height, excluding branches) as well as two leaf traits, leaf area (LA,

cm2) and specific leaf area (SLA; m2 leaf area per kg leaf dry mass; Pérez-Harguindeguy

et al. 2013). Height of the primary stem was selected as an easy to measure, non-

destructive estimate of biomass which has been shown to vary in Eucalyptus species

under similar experimental conditions (O’Brien and Krauss 2010, Breed et al. 2016). Leaf

traits, LA and SLA, were selected as traits that are both highly responsive to local

environment and climate. In particular, SLA scales positively with a number of measures

including those which increase photosynthetic rates, but decreases with aridity/ water

availability gradients and leaf longevity (Pérez-Harguindeguy et al. 2013), and thus was

hypothesised to be under strong selection in this system.

Survival was monitored at each site visit. All seedlings that perished between

transplanting in June 2014 and the first assessment in September 2014 were excluded

from survival analysis, as mortality was almost certainly due to flood damage or failure

to transplant successfully, and was thus not considered to be a response to experimental

treatments.

120

Growth was measured as the difference in height between specified periods, where

height was taken as the distance (to nearest 0.5 cm) from the base of the stem to the top

of the highest meristem. Height of each seedling was measured in the glasshouse prior to

planting in June 2014, and during site visits on 13th to 15th of October, 1st to 7th of

December 2014 and 9th to 12th December 2015. Two growth intervals were used for

analysis.

First, early growth after establishment at the transplant site was analysed using the

difference between October and December 2014 measurement periods. This standardised

period was used in order to exclude potential idiosyncratic bias due to variable timing of

sowing dates, inundation periods and insecticide treatment application (amongst other

factors) during the June to September 2014 period immediately post-transplant (when no

finer-scale observations of plants could be made due to inaccessibility of sites during

winter flooding). The magnitude of mammalian browsing differed greatly within and

between transplant sites during the first growth measurement period (moderate to high at

T1150, T800 and T700, low at T1050 and non-detectable at T1200 and T550). To control

for these differences, all individuals browsed between mid-October and December (often

indicated by post-browsing branching into multiple meristems) were excluded from this

initial growth analysis.

Second, longer-term growth over the complete timeframe of the experiment was

analysed using the period between measurements taken at planting in June 2014 and

December 2015, encompassing 18 months of growth in situ across three growing seasons

(spring 2014, spring and autumn 2015). Monitoring across the latter 12 months was

infrequent and mammalian browsing and branching was common across all sites except

the highest rainfall site, T1200, so all individuals except those showing signs of

continuous, heavy browsing (as indicated by browsing of the main stem, often to near


121

ground level, and failure to retain leaves to full expansion) remained in this coarser-scale

analysis of overall growth.

The LA and SLA metrics (Pérez-Harguindeguy et al. 2013) were measured from

a single leaf per seedling at two time points, first in December 2014, at the end of the first

growing season in situ and coinciding with the first growth period analysed, and the

second in December 2015, at the completion of the experiment after 18 months in situ.

The youngest, fully-expanded and toughened leaf with the least damage by herbivores or

pathogens was cut from each seedling, excluding the petiole, and stored in a sealed plastic

bag in a cold box. Leaves were arranged on a white grid-scaled Perspex board and pressed

flat with a second board of 10 mm clear Perspex and photographed as soon as possible,

no later than 12 hours after collection. All photographs were manipulated in Adobe

Photoshop (Version 2015.1.2, Adobe Systems Software Ltd.) to correct for camera angle

and lens distortion and LA was measured using Image J’s particle analyser function

(Version 1.48, NIH). Leaves were stored in the freezer after photographing until drying.

Leaves were dried for 72 h at 65°C until mass remained constant and then cooled in sealed

plastic bags with silica gel prior to weighing (A&D ER180A electronic balance, precision

0.0001 g).

4.2.6 Rationale and methods of statistical analysis

4.2.6.1 Effects of source rainfall and maternal lineage on seed mass and early growth

under glasshouse conditions

To determine whether maternal investment in seed mass might underpin source-site

rainfall effects on seedling growth, I tested the effect of source rainfall on mean seed mass

of each maternal lineage, and the consequent effect of variation in mean seed mass on

mean seedling growth under glasshouse conditions prior to transplant. Data were not

available on growth outcomes for individual seeds, as only a sub-sample of seeds was

weighed for each maternal lineage. First, I used a linear model (LM) to test the

122

dependence of seed mass on rainfall at seed source, then, a second LM tested the

additional independent effect of source rainfall on seedling height after accounting for the

seed mass effect. The analysis was performed in R (Version 3.2.5, R Core Team 2016).

Simplification of the full models was undertaken using model selection procedures

comparing model fit using maximum likelihood estimation and Akaike Information

Criterion adjusted for small sample sizes (AICc) in the ‘MuMIn’ package (Version

1.15.6., Barton 2016). The most parsimonious (least complex) model within 2 AICc units

of the top model (i.e. the model with the lowest AICc value) was selected as the ‘best’

model (Arnold 2010).

4.2.6.2 Testing for trait fixation versus plasticity in transplanted gardens

4.2.6.2.1 Survival

Survival was analysed as a binomial response for each individual (dead/alive) at

December 2015, after 18 months in situ and at the completion of the experiment. I tested

the effects of source site rainfall (S, source effect), transplant site rainfall (T, environment

effect), and insecticide treatment (I, control versus treated) as well as their interactions on

survival using a generalised mixed effect model (GLMM) with a binomial distribution

and logit link (Quinn and Keough 2002, Bolker et al. 2009). A significant main effect of

source or transplant sites would indicate differential survival rates among sources

regardless of transplant site, or differential survival rates between transplant locations

irrespective of source, respectively. A significant interaction term would indicate

significant variation in seedling survival across transplant sites depending on their source

environment. Note, however, that even though the significant interaction term indicates

differences in the survival of source provenances across different transplant sites, the

determination of a population as ‘locally adapted’ requires a population to exhibit both

an advantage at home, and a disadvantage under novel climates (Kawecki and Ebert

2004). A significant insecticide treatment effect in isolation would suggest that insect


123

attack significantly impacts survival regardless of source or transplant site. An interaction

with source site would indicate differential effects of insecticide treatment depending on

source environment, indirectly testing for differences in susceptibility to insect-caused

mortality. Finally, an insecticide by transplant rainfall interaction effect tests for

differences in insect herbivory pressure among sites. GLMMs were run in package ‘lme4’

(Version 1.1-12, Bates et al. 2015). Prior to running the analyses all non-binary predictors

were mean centred and scaled to two standard deviations (Gelman 2008, Schielzeth

2010). Transplant site, and maternal lineage nested within source site, were specified as

random components to account for non-independence of multiple seedlings measured

within maternal lineages within sites (Blanquart et al. 2013). The residuals were checked

for over-dispersion, but no adjustment was necessary. Model selection was carried out as

is described for LMs and model coefficients for the best model were estimated using

restricted maximum likelihood estimation, and model fit was assessed using the

Nakagawa and Schielzeth (2013) R2 approach.

4.2.6.2.2 Home versus away – a test for plasticity to environmental variation

The environmental plasticity hypothesis was tested by calculating trait differences in

height growth, LA and SLA between seedlings grown at the transplant site nearest their

source (i.e. their ‘home’ site), and seedlings of the same maternal lineage transplanted

away from home (Fig. 4.2: home vs away, HvA). The response differential was calculated

as the absolute response of each individual seedling minus the mean response of all the

locally-sourced seedlings at the seedling’s source. For seedlings transplanted away from

their home site, this differential represents the mean and variability of individual

responses to each level of novel climate across the gradient, relative to the mean response

of all the maternal lineages at its source site under source climates. For seedlings planted

at their home site, it represents the variability among individuals about the mean of the

locally sourced seedlings growing under source conditions. Responses that centre about

124

zero would indicate low environmental plasticity, whereas responses that differ from zero

would indicate plasticity. A positive value would indicate higher trait values than

expected under source conditions, and a lower value would indicate a lower trait value. I

tested the effect of source rainfall, transplant site rainfall and insecticide on the response

differentials using linear mixed effects models (LMM) with a Gaussian distribution in

package ‘lme4’ (Version 1.1-12, Bates 2005), using the methods and interpretation

described for the GLMMs. The residuals were assessed for homogeneity and normality

and the differentials were log transformed (i.e. creating a log response ratio (Hedges et

al. 1999) where required to meet these assumptions).

4.2.6.2.3 Local versus foreign – a test for differences in trait fixation across

populations

Differences in trait fixation across provenances were tested for by calculating differences

in height growth, LA and SLA between individual seedlings of foreign and local maternal

lineages within each common garden transplant site (Fig. 4.2: local vs foreign, LvF). The

response differential was calculated as the absolute response of each individual seedling

minus the mean response of all locally-sourced seedlings within each transplant site. For

foreign maternal lineages this differential represents the mean and variability of

individual responses relative to the mean response for all local lineages, and for local

lineages it represents variability of local individual responses around the local population

mean. A positive value for the response differential would indicate that the foreign

individuals have higher trait values than the local population mean, and suggest that the

local population does not carry traits advantageous to their home site. A negative value

would suggest a local advantage, in that the local provenance out-performs the foreign

individuals under home site conditions (Kawecki and Ebert 2004). Analyses are as

described for the HvA model.


125

Seedlings at transplantation site T800 experienced unusually low growth and high

mortality relative to that observed at both lower and higher rainfall sites, so data from this

site were treated with caution (only 3 out of 288 plants survived to the end of the

experiment, and there was no identifiable cause of mortality). Analyses where T800 had

undue influence on the resulting models (e.g. across transplant site comparisons,

including survival analysis and the HvA trait analyses) were re-run with and without T800

data as a sensitivity analysis to the main analyses (the full analysis is presented in

Supplement Figs. S4.1–2; Tables S4.1-2). In the HvA analysis, maternal lineages sourced

closest to T800 were reallocated the ‘nearest’ transplantation site as determined by mean

annual rainfall. As the LvF analysis was a within-site analysis and all responses were

measured relative to a mean generated from T800, the idiosyncratic site-level variation at

T800 had less influence on overall model results.

4.3 Results

4.3.1 Effects of source site rainfall and maternal lineage on seed mass and early

growth under glasshouse conditions

Seed mass increased significantly (and linearly; Table 4.2a) with mean annual rainfall at

source site (Fig. 4.3a; Table 4.3a), despite wide variation in seed mass among maternal

lineages sourced from within collection sites (Fig. 4.3a). Maternal lineages with greater

seed mass also tended to have taller seedlings (Table 4.3b), even after accounting for seed

mass variation (Fig. 4.3b, Table 4.3b). For maternal lineages from lower rainfall regions,

growth tended to be higher than predicted based on their low seed mass, whereas for

maternal lineages from higher rainfall regions growth tended to be lower than predicted

based on their typically higher seed mass (Table 4.3b).

126

Table 4.2. Model selection using Akaike Information Criterion (AICc) scores for fitted

linear models testing Eucalyptus rudis (a) seed mass as a function of source rainfall

(model 1) and (b) height pre-transplant as a function of source rainfall and seed mass

(model 2). The modelled predictors source rainfall and seed mass were tested for linear

(Lin) and quadratic (Quad) relationships with the responses. k denotes the number of

parameters included in the model, AICc is a measure of fit scaled to the number of

parameters in the model and AICc weight is an estimate of the likelihood of the model.

Models with the lowest AICc and greatest AICc weight denote the best model fit: the

most parsimonious model < 2 AIC was selected, and in bold.

k AICc Δ AICc AICc weight

(a) Model 1: seed mass

Source(Lin) 3 163.31 0.00 0.67

Source(Quad) 4 164.70 1.39 0.33

(b) Model 2: seedling height

Seed mass(Lin) x Source(Quad) 7 157.88 0.00 0.82

Seed mass(Lin) 3 163.31 5.43 0.05

Seed mass(Quad) x Source 7 163.89 6.01 0.04

Source x seed mass(Lin) 5 164.43 6.55 0.03

Source + seed mass(Lin) 4 164.70 6.82 0.03

Seed mass(Lin) + Source(Quad) 5 165.95 8.06 0.01

Seed mass(Quad) + Source(Quad) 10 166.94 9.06 0.01

Source(Lin) 3 168.85 10.96 0.00

Table 4.3. Linear models of Eucalyptus rudis (a) seed mass as a function of source

rainfall (model 1) and (b) seedling height pre-transplant as a function of source rainfall

and seed mass (model 2). ‘b’ indicates the coefficient ± 95% confidence interval. Terms

that were statistically significant are highlighted in bold text.

b [± 95% CI] t-value P

(a) Model 1: seed mass

Intercept 0.24 [0.21, 0.27] 19.152 <0.001

Source(Lin) 0.09 [0.04, 0.14] 3.633 0.001

(b) Model 2: seedling height

Intercept 11.60 [8.93, 14.27] 8.511 <0.001

Seed mass 23.22 [12.52, 33.91] 4.255 <0.001

Source(Lin) 0.89 [-3.01, 4.78] 0.447 0.659

Source(Quad) 4.56 [-5.54, 14.65] 0.884 0.386

Seed mass: Source(Lin) 15.74 [4.14, 27.34] 2.659 0.014

Seed mass: Source(Quad)

-64.24 [-99.82, -28.66] -3.539 0.002


127

Figure 4.3. General linear models of Eucalyptus rudis (a) mean seed mass of each

maternal lineage against historical mean annual rainfall at the source site. The fitted line

represents the predicted (± 95% CI) relationship from model 1 (Table 4.3a). (b) Partial

(independent) effect of source site rainfall on mean seedling height of each lineage after

five months in the glasshouse. The fitted line represents the predicted seedling height (±

95% CI) relationship from model 2 (Table 4.3b) holding seed mass constant at the 20th

and 80th percentile.

4.3.2 Trait fixation versus plasticity in transplant sites

4.3.2.1 Survival

Survival rates through the first spring, growing season up to December 2014 were high,

ranging from 96% at T700, to 99% at T550 (Fig. 4.4), excluding the deaths of seedlings

attributed to flooding and establishment failure (between June and September 2014).

Mortality began to increase during the dry summer period following the December 2014

measurements, particularly at sites T800 and T550, largely attributed to grazing from

rabbits and kangaroos respectively (data not presented). Mortality rates slowed over the

wet, winter season of 2015, plateauing through spring 2015 at all sites except T800

leading up to the final assessment in December 2015. The high mortality observed at site

T800 (where by December 2015 only three heavily grazed seedlings remained out of 288

planted at the site) substantially altered the modelled relationships (Tables 4.4; 4.5, Figs.

128

4.5, S4.1). For the five transplant sites excluding T800, survival increased linearly with

increasing transplantation site rainfall (Table 4.5, Fig. 4.5). Source also had a significant

effect on seedling survival, with the seedlings sourced from the lowest rainfall sites

experiencing greater survival rates, regardless of transplant site (Table 4.5b).

Furthermore, survival did not differ significantly between insecticide treated seedlings

with a reduced invertebrate herbivore load versus the non-insecticide treated control

plants (Table 4.4b).

Figure 4.4. Percentage survival of seedlings transplanted to trial sites and treated with a

water control (solid line) or insecticide treatment (dashed line) for the period from

planting in July 2014 to final measurement in December 2015. Transplant site names

indicate the approximate mean annual rainfall at the transplant site. Grey shading

indicates the period over which insecticide treatment was maintained at fortnightly

intervals. The vertical red dashed lines indicate timing of seedling measurements in

December 2014 and 2015.


129

Table 4.4. Model selection using Akaike Information Criterion (AICc) scores to compare

generalised linear mixed effects models testing variation in survival among transplanted

Eucalyptus rudis seedlings as a function of source rainfall (S), transplant site rainfall (T)

and insecticide treatment (I). Model set (a) tests survival across all transplant sites, and

model set (b) excludes the aberrant site T800 (see text for details). The most parsimonious

model >2 was selected and is indicated in bold. Further detail on the abbreviations and

terms used is presented in Table 4.2.

k AICc ΔAICc k AICc ΔAICc

(a) Including T800 (b) Excluding T800

S*T2 9 1925 0.00 S*T2 9 1881 0.00

S*T 7 1928 2.68 S*T 7 1881 0.13

S 5 1929 3.45 T 5 1883 2.26

T 5 1930 4.85 S2*T 9 1885 4.16

Null 4 1931 5.11 T*I 7 1886 4.47

I 5 1932 6.66 S 5 1886 4.62

S2*T 9 1932 6.71 Null 4 1887 6.23

S*I 7 1932 6.98 S*T*I 11 1887 6.24

T*I 7 1932 7.04 I 5 1889 7.70

S*T*I 11 1934 8.79 S*I 7 1889 8.04

S*T2*I 15 1934 8.88 S*T2*I 15 1890 8.73

S2*T*I 15 1939 14.04 S2*T*I 15 1893 11.71

S2*T2*I 21 1940 14.25 S2*T2*I 21 1895 13.40

Figure 4.5. Survival at 18 months post-transplant for Eucalyptus rudis seedlings planted

in experimental gardens at different points along a rainfall gradient. A shift of 0 mm

rainfall denotes the home site of each source provenance (S538-S1214, labelled according

to mean rainfall per annum at the source site). Each point represents the proportion of

each maternal lineage surviving at each transplant site. The fitted lines are the model

predictions from a generalised linear mixed model (± 95% Confidence intervals)

excluding transplant site T800 due to low sample size (Table 4.5b). Note that overlapping

points are offset along the x axis to reduce overlap.

130

Table 4.5. Generalised linear mixed models testing variation in Eucalyptus rudis seedling

survival at December 2015 (18 months in situ) as a function of source rainfall (Source)

and transplant site rainfall (Transplant) for (a) all transplant sites and (b) transplant sites

excluding the aberrant site T800 (see text for details). The proportion change in variance

(PCV) for the random effects components in the model (maternal lineage (ML) nested in

source site and transplant site) is calculated between the null and final models. The Akaike

Information Criterion (AICc) is a measure of fit scaled to the number of parameters in the

model. R2LMM(m) is the marginal variance explained by all fixed factors and R2

LMM(c) is the

conditional variance explained by both fixed and random factors (Nakagawa and

Schielzeth 2013). The intercept in the full model is the survival of the driest sourced

seedlings at the driest transplant site, without insecticide.

(a) Survival - Including T800 (b) Survival - Excluding T800

n = 1855 n = 1567

Fixed effects b [± 95% CI] b [± 95% CI]

Intercept -2.30 [-4.03, -0.58] -0.35 [-0.81, 0.11]

Source 0.09 [-0.46, 0.64] -0.36 [-0.67, -0.05]

Transplant (Lin) 2.47 [0.55, 4.40] 1.48 [0.64, 2.33]

Transplant (Quad) 5.50 [-0.36, 11.35]

Source : Transplant (Lin) 0.20 [-0.28, 0.68] 0.36 [-0.09, 0.81]

Source : Transplant (Quad) -1.66 [-3.32, -0.01]

VC for random effects

ML/ Source 0.090 0.088

Source 0.000 0.000

Transplant 1.289 0.243

VC for Fixed effects 1.834 0.678

PCV(ML/ Source) 4.18% 7.44%

PCV(Source) 100.00% 100.00%

PCV(Transplant) 59.24% 71.68%

R2glmm(m) 28.20% 15.78%

R2glmm(c) 49.41% 23.46%

AIC(Full model) 1925 1881

AIC(Null model) 1931 1887

4.3.2.2 Trait-mean variation in absolute height, LA and SLA

Transplantation of seedlings across the rainfall gradient elicited large differentiation in

growth and leaf trait responses between transplantation sites within just six months of

transplanting (Figs. 4.6a, c, e). All three responses, height, LA and SLA, increased

significantly with increasing mean annual rainfall across the transplantation sites (Figs.

4.6a, c, e). Seedlings at the highest rainfall site (T1200) were the tallest by December

2014, averaging 14.6 ± 0.7 cm (mean ± standard error (SE)) in insecticide treated plants


131

and 13.4 ± 0.7 cm in control plants. The greater seedling height at T1200 was also

accompanied by larger leaf area (LA control, 19.6 ± 1.5 and insecticide 20.8 ± 1.4 cm2)

and less sclerophyllous leaves (SLA control 25.9 ± 0.5 and insecticide 28.0 ± 0.5 m2kg-1)

after this first growing season in situ (Figs. 4.6c, e). As expected, seedlings at the lowest

rainfall site, T550, were substantially shorter in both control (4.2 ± 0.3 cm) and

insecticide treated seedlings (5.4 ± 0.3 cm) with smaller leaf area (LA control: 6.2 ± 0.5

and insecticide: 7.9 ± 0.6 cm2) and more sclerophyllous leaves (SLA averaging 15.7 ± 0.3

and 16.8 ± 0.3 m2kg-1 in control and insecticide treatments respectively; Fig 4.6e).

The trends in height and LA observed in the first growing season continued

through to the final measurement in December 2015 (18 months in situ; Fig. 4.6b, d, f).

Height 18-months post-transplant averaged 63.4 ± 3.1 cm (control) and 78.8 ± 3.1 cm

(insecticide) at the highest rainfall site, T1200, which was approximately three times taller

than observed at the lowest rainfall site, T550, at 23.3 ± 2.5 cm (control) and

25.5 ± 2.9 cm (insecticide). Likewise, mean LA across all of the transplant sites increased

between the December 2014 and 2015 measurements (Fig. 4.6c, d). Interestingly, SLA

was lower in 2014 than observed in 2015 and became less variable both within and among

transplant sites (Fig. 4.6e, f). It is also worth noting, that differences in all three responses

among the higher rainfall transplantation sites T1200, T1150 and T1050 became more

uniform over time.

132

Figure 4.6. Summary statistics for seedling traits measured in December 2014 (6 months

post-transplant) and December 2015 (18 months post-transplant): (a, b) plant height, (c,

d) leaf area, and (e, f) specific leaf area. Dark grey bars indicate the water control

treatment, and pale grey bars indicate the insecticide treatment. Box plots represent the

median and interquartile range, whiskers represent ± 1.5 × interquartile range, and points

represent outlier values outside this range.


133

4.3.2.3 Trait plasticity – home versus away model

A significant transplant site rainfall effect was detected for height, LA and SLA (Tables

4.6, 4.7), indicating a high degree of plasticity in plant traits. From six-months post-

transplant, maternal lineages that were shifted from lower rainfall source regions towards

higher rainfall transplant sites had significantly greater height-growth (Fig. 4.7a, Table

4.7), LA (Fig. 4.7b, Table 4.7), and SLA (Fig. 4.7c, Table 4.7) than the mean of their

siblings grown under home site conditions. Conversely, maternal lineages shifted from

higher rainfall source sites to lower rainfall transplant sites showed a reduction in height-

growth (Fig. 4.7a, Table 4.7), LA (Fig. 4.7b, Table 4.7), and SLA (Fig 4.7c, Table 4.7)

from that observed under home site conditions. The magnitude of the plasticity effect

appeared to differ depending on source site lineages for the growth response, but not for

LA or SLA, as shown by the significance of the interaction between transplant and source

rainfall terms (Table 4.7). Likewise, exposure to insect herbivores significantly reduced

growth differentials relative to siblings at the home site, but insecticide treatment did not

significantly affect LA or SLA (Table 4.7).

Overall, the modelled parameters accounted for substantial variation in the

differentials of height-growth and SLA relative to home conditions, with marginal R2

values of 31% and 50% for growth and SLA respectively. In contrast, variation in LA had

lower model fit particularly in 2014 (R2 = 19%, Table 4.7). The random variance

components explained by factors such as idiosyncratic performance advantage of some

maternal lineages over others, or site-to-site variation were uniformly small (explaining

an additional 3.31 to 13.66%, Table 4.7). Maternal lineages shifted from high rainfall

sites to drier sites nearly always showed response trait differentials lower than their home

site mean, whereas maternal lineages from dry source sites which were shifted towards

wetter sites showed substantial trait overlap with their home mean (i.e. trait conservatism)

134

in growth and LA, but not in SLA (Fig. 4.7a, c, e). Of the traits measured, SLA showed

the most consistently plastic response to transplantation across the rainfall gradient.

Unfortunately, low seedling survival in the mid to low rainfall regions of the

catchment (Figs. 4.4, 4.5), meant that mean local trait responses were calculated from

fewer individuals in 2015. The resulting differentials were much more variable and

heavily influenced by individual seedlings responses, and should therefore be treated with

caution. With this caveat in mind, responses over the later 12 months of the experiment

(as measured in December 2015) were fit by broadly similar models (Table 4.7), and the

positive relationship between increasing transplant site rainfall and increasing level of

trait expression also held true in 2015 (Fig. 4.7). As noted above, the three higher rainfall

transplantation sites had a more uniform response in 2015 than observed in 2014, this

pattern was evident in LA within the high rainfall sources, S1166, S1204 and S1214,

where small reductions in rainfall did not reduce LA, but transplantation to sites with 400

mm pa less rain than their source sites did cause LA to decline significantly (Table 4.7

Fig. 4.7d).


135

Table 4.6. Model selection using AICc scores to compare generalised linear mixed effects

models testing the variation in Eucalyptus rudis seedling traits as a function of source

rainfall (S), transplant site rainfall (T) and insecticide treatment (I) under the home versus

away model. Model selection is presented for the response in traits, height growth, leaf

area (LA) and specific leaf area (SLA) in December 2014 (6 months post-transplant) and

in December 2015 (18 months post-transplant). The models presented exclude

transplantation site T800 due to low sample size. k denotes the number of parameters

included in the model, AICc is a measure of fit scaled to the number of parameters in the

model. Models with the lowest AICc denote the best model fit: the most parsimonious

model < 2 AIC was selected, and in bold.

Growth 2014 k Log Likelihood AICc ∆AICc

I + S + S2 + T + S:T + S

2:T 11 -895 1813 0.000

I + S + S2 + T + I:T + S:T + S

2:T 12 -894 1813 0.246

I + S + S2 + T + S:T + S

2:T 12 -895 1814 1.042

I + S + S2 + T + T

2 + S:T + S

2:T + T

2:S 13 -894 1814 1.123

I + S + S2 + T + T

2 + I:T + S:T + S

2:T 13 -894 1814 1.314

I + S + S2 + T + T

2 + S:T + S

2:T

2 + S

2:T 13 -894 1814 1.468

I + S + S2 + T + T

2 + I:T + S:T + S

2:T + T

2:S 14 -893 1814 1.477

I + S + S2 + T + I:S + S:T + S

2:T 12 -895 1814 1.573

I + S + S2 + T + T

2 + I:T + S:T + S

2:T

2 + S

2:T 14 -893 1815 1.789

I + S + S2 + T + I:S + I:T + S:T + S

2:T 13 -894 1815 1.888

Null 5 -919 1849 35.821

Growth 2015

S + S2 + T 8 -781 1579 0.000

S + S2 + T + T

2 + T

2:S 10 -779 1579 0.095

S + S2 + T + T

2 + S

2:T

210 -780 1580 0.934

S + S2 + T + S:T 9 -781 1580 1.434

S + S2 + T + S

2:T 9 -781 1580 1.460

S + S2 + T + T

2+ S

2:T + T

2:S 11 -779 1580 1.478

S + S2 + T + T

2+ S:T + T

2:S 11 -779 1580 1.638

S + S2 + T + T

2 + S

2:T

2 + T

2:S 11 -779 1581 1.963

S + S2 + T + T

29 -781 1581 1.983

Null 5 -793 1596 17.240

Leaf Area 2014

I + S + S2 + T 9 -1750 3518 0.000

I + S + S2 + T + S:T 10 -1749 3518 0.017

I + S + S2 + T + I:S + S:T 11 -1748 3519 0.156

I + S + S2 + T + I:S 10 -1749 3519 0.173

S + S2 + T + S:T 9 -1751 3519 0.823

S + S2 + T 8 -1752 3519 0.833

I + S + S2 + T + S:T + S

2:I 11 -1749 3520 1.227

I + S + S2 + T + S

2:I 10 -1750 3520 1.229

I + S + S2 + T + S

2:T 10 -1750 3520 1.408

I + S + S2 + T + I:S + S

2:T 11 -1749 3520 1.565

I + S + S2 + T + S:T + S

2:T 11 -1749 3520 1.709

I + S + S2 + T + I:S + S:T + S

2:T 12 -1748 3520 1.855

136

Table 4.6. Continued.

Leaf Area 2014 cont. k Log Likelihood AICc ∆AICc

I + S + S2 + T + T

2 10 -1750 3520 1.895

I + S + S2 + T + I:T 10 -1750 3520 1.913

I + S + S2 + T + T

2 + S:T 11 -1749 3520 1.919

I + S + S2 + T + I:T + S:T 11 -1749 3520 1.927

Null 5 -1763 3536 17.368

Leaf Area 2015

I + S + S2 + T + T

2 + I:S + S

2:I + T

2:S 13 -647 1322 0.000

I + S + S2 + T + T

2 + I:S + I:T + S

2:I + T

2:S 14 -646 1322 0.043

I + S + S2 + T + T

2 + I:S + I:T + S

2:I + T

2:I + T

2:S 15 -645 1322 0.345

I + S + S2 + T + T

2 + I:S + S:T + S

2:I + T

2:S 14 -647 1322 0.584

I + S + S2 + T + T

2 + I:S + I:T + S:T + S

2:I + T

2:S 15 -646 1322 0.801

I + S + S2 + T + T

2 + I:S + I:T + S:T + S

2:I + T

2:I + T

2:S 16 -645 1323 1.050

I + S + S2 + T + T

2 + I:S + S

2:I + T

2:I + T

2:S 14 -647 1323 1.219

I + S + S2 + T + T

2 + I:S + S

2:I + S

2:T + T

2:S 14 -647 1323 1.347

I + S + S2 + T + T

2 + I:S + I:T + S

2:I + S

2:T + T

2:S 15 -646 1323 1.627

I + S + S2 + T + T

2 + I:S + S:T + S

2:I + T

2:I + T

2:S 15 -646 1323 1.738

I + S + S2 + T + T

2 + I:S + I:T + S

2:I + S

2:T + T

2:S + S

2:I:T 16 -645 1323 1.778

I + S + S2 + T + T

2 + I:S + I:T + S

2:I + S

2:T + T

2:I + T

2:S 16 -645 1323 1.908

Null 5 -670 1350 28.617

SLA 2014

I + S + T + I:S 9 -3708 7434 0.000

I + S + S2 + T + I:S 10 -3707 7435 0.783

I + S + T + I:S + S:T 10 -3708 7435 1.596

I + S + T + I:S + I:T 10 -3708 7436 1.841

I + S + S2 + T + I:S + S

2:T 11 -3707 7436 1.975

Null 5 -3740 7489 55.482

SLA 2015

I + S + S2 + T + I:S + S

2:I 11 -1392 2806 0.000

I + S + S2 + T + I:S + I:T + S

2:I 12 -1391 2806 0.416

I + S + S2 + T + I:S + I:T + S:T + S

2:I + S

2:T 14 -1389 2807 0.494

I + S + S2 + T + I:S + S:T + S

2:I + S

2:T 13 -1390 2807 0.582

I + S + S2 + T + I:S + I:T + S:T + S

2:I + S

2:T + S

2:I:T 15 -1388 2807 0.615

I + S + S2 + T + T

2 + I:S + S

2:I 12 -1391 2807 0.937

I + S + S2 + T + T

2 + I:S + S

2:I + T

2:S 13 -1390 2807 1.229

I + S + S2 + T + I:S + I:T + S

2:I 13 -1390 2807 1.323

I + S + S2 + T + I:S + I:T + S:T + S

2:I + S

2:T 15 -1388 2807 1.377

I + S + S2 + T + I:S + S

2:I + S

2:T 12 -1391 2807 1.401

I + S + S2 + T + T

2 + I:S + S:T + S

2:I + S

2:T 14 -1389 2808 1.508

I + S + S2 + T + I:S + I:T + S:T + S

2:I + S

2:T + S

2:I:T 16 -1387 2808 1.519

I + S + S2 + T + I:S + I:T + S

2:I + S

2:T 13 -1390 2808 1.562

I + S + S2 + T + T2 + I:S + I:T + S

2:I + T

2:S 14 -1389 2808 1.714

I + S + S2 + T + T

2 + I:S + S

2:T

2 + S

2:I + T

2:S 14 -1390 2808 1.886

I + S + S2 + T + I:S + S:T + S

2:I 12 -1392 2808 1.942

I + S + S2 + T + I:S + I:T + S:T + S

2:I + S

2:T + I:S:T 15 -1389 2808 1.996

Null 5 -1414 2838 31.926


137

20

14

(n

= 7

87

)2

01

5 (

n =

67

9)

20

14

(n

= 1

18

5)

20

15

(n

= 5

51

)2

01

4 (

n =

11

83

)2

01

5 (

n =

55

1)

Fix

ed e

ffects

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

Inte

rcep

t (F

ull

)0

.47

[0

.13

, 0

.81

]0

.45

[0

.08

, 0

.82

]0.1

2 [

-0.2

2, 0.4

6]

0.9

5 [

0.1

9, 1

.71

]2

.05

[0

.87

, 3

.23

]-5

.57

[-7

.53

, -3

.61

]

So

urc

e(L

n)

-0.1

6 [

-0.6

9, 0.3

7]

-0.2

3 [

-0.7

1, 0.2

5]

-0.1

8 [

-0.6

1, 0.2

5]

0.7

3 [

0.0

2, 1

.45

]-7

.70

[-8

.92

, -6

.47

]-1

1.5

1 [

-13

.70

, -9

.33

]

So

urc

e(Q

uad

)-1

.27

[-2

.45

, -0

.10

]-1

.70

[-2

.76

, -0

.63

]-1

.58

[-2

.60

, -0

.56

]-2

.51

[-4

.20

, -0

.82

]2

2.0

5 [

16

.41

, 2

7.6

8]

Tra

nsp

lan

t (L

n)

0.8

8 [

0.5

9, 1

.17

]0

.73

[0

.27

, 1

.20

]0

.65

[0

.28

, 1

.03

]0

.68

[0

.16

, 1

.19

]6

.90

[4

.94

, 8

.86

]3

.08

[1

.06

, 5

.09

]

Tra

nsp

lan

t (Q

uad

)-0

.59 [

-2.3

9, 1.2

1]

Insect

-0.2

4 [

-0.3

4, -0

.13

]-0

.53 [

-1.1

8, 0.1

1]

-0.6

0 [

-1.2

3, 0.0

3]

5.1

3 [

2.9

3, 7

.33

]

Sourc

e(L

n):

Tra

nsp

lant

0.4

2 [

0.1

2, 0

.71

]

Sourc

e(Q

uad

): T

ransp

lant

-0.7

5 [

-1.4

8, -0

.03

]

Sourc

e(L

n):

Inse

ct

-0.9

0 [

-1.6

3, -0

.18

]-2

.61

[-3

.85

, -1

.37

]3

.52

[1

.00

, 6

.05

]

Sourc

e(Q

uad

): In

sect

2.1

0 [

0.0

0, 4

.19

]-1

4.6

4 [

-21

.85

, -7

.43

]

Souce

(Ln):

Tra

nsp

lant (

Quad

)-1

.92

[-2

.87

, -0

.98

]

VC

fo

r ra

nd

om

eff

ects

ML

: S

ourc

e0.0

27

0.0

11

0.0

77

0.0

16

0.8

17

0.4

05

Sourc

e0.0

43

0.0

36

0.0

08

0.0

45

0.1

66

0.3

34

Tra

nsp

lant

0.0

15

0.0

73

0.0

47

0.0

72

1.2

73

1.3

65

Resi

dual

0.5

45

0.5

59

1.0

81

0.5

88

30.1

32

8.6

85

VC

fo

r F

ixe

d e

ffe

cts

0.2

76

0.2

72

0.2

89

0.2

56

35

.67

51

0.8

81

PV

C(M

L:

So

urc

e)-1

3.4

4%

0.6

2%

-5.8

3%

-6.1

4%

-5.8

9%

-7.7

1%

PV

C(S

ou

rce)

73.2

5%

86.4

9%

96.1

6%

81.6

7%

99.2

3%

97.3

6%

PV

C(T

ran

spla

nt)

91.1

8%

69.1

2%

74.4

3%

66.2

9%

92.7

1%

69.3

6%

PV

C(R

esid

ual

s)3.0

5%

0.0

2%

-0.0

4%

3.5

2%

1.6

8%

1.4

7%

R2

GL

MM

(m)

30

.51

%2

8.6

3%

19

.25

%2

6.1

4%

52

.42

%5

0.2

1%

R2

GL

MM

(c)

39

.89

%4

1.3

2%

28

.02

%3

9.8

0%

55

.73

%5

9.9

2%

AIC

c (

Fu

ll)

18

13

15

79

35

19

13

21

74

34

28

06

AIC

c (

Nu

ll)

18

49

15

96

35

36

13

50

74

89

28

38

Heig

ht

grow

thL

AS

LA

Table 4.7.

138

Table 4.7. Generalised linear mixed effects models testing variation in Eucalyptus rudis

seedlings traits as a function of source rainfall, transplant site rainfall and insecticide in

December 2014 (6 months post-transplant) and in December 2015 (18 months post-

transplant) under the home versus away model. The modelled responses are a deviance

in trait value of each individual from the mean trait value of each source grown under

conditions nearest to their source for traits (a, b) height growth (c, d), leaf area and (e, f)

specific leaf area. The models presented exclude transplantation site T800 due to low

sample size. The intercept in the full model represents the seedlings sourced at the lowest

rainfall site (S539) at the lowest rainfall transplant site (T550) without insecticide.

Coefficients in bold are statistically different from zero (P < 0.05), and the abbreviations

are as defined in Table 4.5.◄

Figure 4.7. The response of Eucalyptus rudis seedling traits to transplantation along a

rainfall gradient in December 2014 (6 months post-transplant) and in December 2015 (18

months post-transplant) under the home versus away model. The responses represent the

deviance in trait value of each individual from the mean trait value of each source grown

under conditions nearest to their source (0) for traits (a, b) height growth (c, d), leaf area

and (e, f) specific leaf area. A shift of 0 mm rainfall denotes the home site of each source

provenance (S538-S1214, labelled according to mean rainfall per annum at the source

site). The fitted lines are the model predictions from the generalised linear mixed models

presented in Table 4.6 (± 95% Confidence intervals). The models presented exclude

transplantation site T800 due to low sample size. ►


139

140

4.3.2.4 Trait fixation – local versus foreign model

As expected from the HvA plasticity results, there was little evidence that E. rudis

seedlings exhibited fixed height-growth, LA and SLA over the 18 months they were

grown in situ (Tables 4.8, 4.9; Fig. 4.8c). For most responses, the foreign maternal

lineages were indistinguishable from the mean of locally sourced lineages. Height-growth

of seedlings sourced from the lowest rainfall sites (S538, S549 and S547) however,

revealed a significant conservatism in growth (Table 4.9) when transplanted to higher

rainfall sites as shown by a negative growth differential relative to the local mean (Fig.

4.8a). This conservatism was not apparent in seedlings sourced from S696, which

receives just 150 mm greater rainfall per annum on average (Table 4.9; Fig. 4.8a). While

this conservatism in seedlings from low rainfall source sites was apparent across the

gradient of transplant sites, an interaction between transplant and source rainfall indicates

that it was more pronounced as the magnitude of shift in rainfall between source and

transplant sites increased (Table 4.9, Fig. 4.8a). Conversely, and somewhat surprisingly,

seedlings transferred from high rainfall sites to drier transplant sites were

indistinguishable in growth and leaf traits from locally sourced seedlings, regardless of

the magnitude of difference in rainfall across the gradient (Fig. 4.8).

Although seedlings with a reduced insect load (insecticide treated seedlings)

deviated from their local mean less than the control seedlings, neither seedlings from

different sources, nor seedlings placed in different transplant sites were significantly

influenced by insecticide treatment, as no interactions were detected with either source or

transplant site rainfall in the models (Table 4.9).

After 18 months in situ, foreign-sourced seedlings generally continued to show a

similar height-growth response compared to the local mean across all transplant sites

(Fig.4.8b). The most prominent difference between years was an interaction between

source rainfall and the quadratic component of transplant site rainfall, indicating that


141

seedlings from the mid to low rainfall regions exhibited lower growth than the local mean

in the mid rainfall regions, but not at either extreme (Fig. 4.8b). While this result must be

treated with caution due to low final replication (as discussed in HvA), this result suggests

a reduction in the conservatism from low rainfall sourced seeds at the highest rainfall

transplant site, T1200 (Fig. 4.8b) and that variability in growth was better explained by

transplant site differences by 18 months in situ.

142

Table 4.8. Model selection using AICc scores to compare generalised linear mixed effects

models testing the variation in Eucalyptus rudis seedling traits as a function of source

rainfall (S), transplant site rainfall (T) and insecticide treatment (I) under the local versus

foreign model. Model selection is presented for the response in traits, height growth, leaf

area (LA) and specific leaf area (SLA) in 2014 (6 months post-transplant) and again in

2015 (18 months post-transplant). The Terms and abbreviations are as described in Table

4.5. The most parsimonious model < 2 AIC was selected, and in bold.

Growth 2014 k Log Likelihood AICc ∆AICc

I + S + S2 + T + S:T + S

2:T 11 -2849 5720 0.00

I + S + S2 + T + T

2 + S:T + S

2:T + T

2:I 13 -2847 5720 0.39

I + S + S2 + T + S:T + S

2:I + S

2:T 12 -2848 5721 1.13

I + S + S2 + T + T

2 + S:T + S

2:I + S

2:T + T

2:I 14 -2846 5721 1.54

I + S + S2 + T + T

2 + S:T + S

2:T 12 -2848 5721 1.66

I + S + S2 + T + I:S + S:T + S

2:I + S

2:T 13 -2848 5722 1.92

I + S + S2 + T + T

2 + S:T + S

2:T

2 + S

2:T + T

2:I 14 -2847 5722 1.93

I + S + S2 + T + T

2 + I:T + S:T + S

2:T + T

2:I 14 -2847 5722 1.99

Null 5 -2874 5757 37.65

Growth 2015

I + S + S2 + T + T

2 + I:T + S

2:T

2 + S

2:I + T

2:I + T

2:S + S

2:T

2:I 16 -3353 6739 0.00

I + S + T + T2 + I:S + I:T + T

2:I + T

2:S + T

2:I:S 14 -3355 6739 0.44

I + S + S2 + T + T

2 + I:S + I:T + T

2:I + T

2:S + T

2:I:S 15 -3354 6739 0.77

I + S + S2 + T + T

2 + I:T + S

2:I + T

2:I + T

2:S 14 -3355 6740 0.87

I + S + S2 + T + T

2 + I:T + S

2:T

2 + S

2:I + T

2:I + S

2:T

2:I 15 -3354 6740 1.01

I + S + T + T2 + I:S + I:T + T

2:I + T

2:S 13 -3357 6740 1.17

I + S + S2 + T + T

2 + I:T + S:T + S

2:T

2 + S

2:I + T

2:I + T

2:S + S

2:T

2:I 17 -3353 6740 1.43

I + S + S2 + T + T

2 + I:S + I:T + T

2:I + T

2:S 14 -3356 6740 1.56

I + S + S2 + T + T

2 + I:S + I:T + T

2:S 13 -3357 6740 1.70

I + S + T + T2 + I:S + I:T + T

2:S 12 -3358 6741 1.89

I + S + T + T2 + I:S + I:T + S:T + T

2:I + T

2:S + T

2:I:S 15 -3355 6741 1.94

I + S + S2 + T + T

2 + I:T + S

2:T

2 + S

2:I + S

2:T + T

2:I + T

2:S + S

2:T

2:I 17 -3353 6741 1.98

Null 5 -3377 6764 25.51

Leaf Area 2014

S + T + S:T 8 -2395 4805 0.00

I + S + T + I:T + S:T 10 -2393 4805 0.00

I + S + T + I:T 9 -2394 4806 0.39

S + T 7 -2396 4806 0.47

I + T + I:T 8 -2395 4806 0.83

T 6 -2397 4806 0.86

I + S + T + T2 + I:T + S:T + T

2:I 12 -2391 4806 1.02

I + S + T + S:T 9 -2394 4806 1.20

I + S + T + T2 + I:T + T

2:I 11 -2392 4807 1.34

S + S2 + T + S:T 9 -2394 4807 1.58

I + S + S2 + T + I:T + S:T 11 -2392 4807 1.63

I + S + T 8 -2395 4807 1.65


143

Table 4.8. Continued.

Leaf Area 2014 cont. k Log Likelihood AICc ∆AICc

S + T + T2 + S:T 9 -2394 4807 1.73

I + T + T2 + I:T + T

2:I 10 -2393 4807 1.74

I + S + T + T2 + I:T + S:T 11 -2392 4807 1.75

I + S + T + I:S + I:T + S:T 11 -2392 4807 1.89

Null 5 -2399 4808 2.87

Leaf Area 2015

S + T + T2 + T

2:S 9 -1130 2278 0.00

S + T + T2 + S:T + T

2:S 10 -1130 2279 1.20

I + S + T + T2 + T

2:S 10 -1130 2280 1.45

S + S2 + T + T

2 + T

2:S 10 -1130 2280 1.90

Null 5 -1143 2296 18.08

SLA 2014

I + T + T2 + I:T + T

2:I 10 -4127 8274 0.00

I + S + T + T2 + I:T + T

2:I 11 -4127 8276 2.01

I + S + T + T2 + I:S + I:T + T

2:I 12 -4126 8277 2.60

I + S + S2 + T + T

2 + I:T + S

2:I + T

2:I 13 -4126 8277 3.44

I + S + T + T2 + I:T + T

2:I + T

2:S 12 -4127 8278 3.71

I + S + T + T2 + I:T + S:T + T

2:I 12 -4127 8278 3.78

I + S + S2 + T + T

2 + I:T + S:T + T

2:I 12 -4127 8278 3.90

I + S + T + T2 + I:S + I:T + T

2:I + T

2:S 13 -4126 8278 4.29

I + S + T + T2 + I:S + I:T + S:T + T

2:I 13 -4126 8278 4.38

I + S + S2 + T + T

2 + I:S + I:T + T

2:I 13 -4126 8278 4.49

I + S + S2 + T + T

2 + I:T + S

2:I + S

2:T + T

2:I 14 -4125 8279 4.75

I + S + T + T2 + I:S + I:T + T

2:I + T

2:S + T

2:I:S 14 -4125 8279 4.90

Null 5 -4141 8292 17.87

SLA 2015

I + S + T + T2 + I:S + I:T + T

2:I 12 -1456 2937 0.00

I + T + T2 + I:T + T

2:I 10 -1459 2937 0.06

I + S + S2 + T + T

2 + I:T + S

2:I + T

2:I 13 -1455 2938 0.26

I + S + T + T2 + I:S + I:T + T

2:I + T

2:S 13 -1456 2938 1.05

I + S + T + T2 + I:T + T

2:I 11 -1458 2939 1.29

I + S + S2 + T + T

2 + I:T + S

2:I + T

2:I + T

2:S 14 -1455 2939 1.36

I + S + T + I:S + I:T 10 -1459 2939 1.37

I + T + I:T 8 -1461 2939 1.44

I + S + S2 + T + I:T + S

2:I 11 -1458 2939 1.54

I + S + S2 + T + T

2 + I:S + I:T + T

2:I 13 -1456 2939 1.68

I + S + T + T2 + I:S + I:T + S:T + T

2:I 13 -1456 2939 1.90

Null 5 -1465 2940 2.64

144

R

espon

se

20

14

(n

= 8

92

)2

01

5 (

n =

68

8)

20

14

(n

= 1

33

1)

20

15

(n

= 5

83

)2

01

4 (

n =

13

29

)2

01

5 (

n =

58

3)

Fix

ed e

ffects

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

b [

± 9

5%

CI]

Inte

rcep

t (F

ull

)1

.66

[0

.13

, 3

.18

]-1

8.2

4 [

-26

.30

, -1

0.1

8]

-0.4

1 [

-0.6

0, -0

.23

]-1

.34

[-1

.86

, -0

.82

]-1

.43 [

-3.2

1, 0.3

5]

0.0

1 [

-0.5

6, 0.5

9]

So

urc

e(L

n)

4.4

1 [

2.4

4, 6

.38

]1

6.2

2 [

0.7

2, 3

1.7

2]

1.2

2 [

0.5

1, 1

.93

]

So

urc

e(Q

uad

)-7

.22

[-1

1.1

0, -2

.45

]

Tra

nsp

lan

t (L

n)

1.3

7 [

-0.5

9, 3.3

2]

-3.3

1 [

-10.7

3, 4.1

0]

-0.2

9 [

-0.5

6, -0

.01

]0

.54

[0

.10

, 0

.98

]2

.14

[0

.27

, 4

.01

]-0

.87 [

-1.8

9, 0.1

6]

Tra

nsp

lan

t (Q

uad

)3

9.5

3 [

21

.37

, 5

7.6

9]

2.6

9 [

1.2

4, 4

.13

]4.2

1 [

-1.4

4, 9.8

7]

Insect

-1.0

5 [

-1.8

2, -0

.28

]6

.10

[0

.61

, 1

1.5

9]

1.9

1 [

0.7

0, 3

.11

]0.0

3 [

-0.5

7, 0.6

2]

So

urc

e(L

n): I

nsect

8.4

4 [

-1.3

7, 18.2

4]

Tra

nsp

lan

t (L

n): I

nsect

10

.67

[0

.09

, 2

1.2

4]

-2.9

8 [

-4.1

5, -1

.82

]1

.18

[0

.06

, 2

.30

]

So

urc

e: T

ran

sp

lan

t (Q

uad

)-4

7.7

9 [

-86

.25

, -9

.33

]

So

urc

e(L

n): T

ran

sp

lan

t5

.99

[3

.72

, 8

.26

]

So

urc

e(Q

uad

): T

ran

sp

lan

t-1

1.0

0 [

-16

.56

, -5

.44

]

Tra

nsp

lan

t (Q

uad

): I

nsect

-5.6

2 [

-9.3

2, -1

.93

]

So

urc

e: T

ran

sp

lan

t (Q

uad

)-4

.22

[-6

.27

, -2

.16

]

VC

for r

an

dom

eff

ects

ML

: S

ou

rce

1.5

85

27.6

90

0.1

19

0.0

47

0.6

94

0.2

04

So

urc

e0.0

00

40.6

20

0.0

00

0.0

51

0.4

93

0.0

00

Tra

nsp

lan

t sit

e0.4

63

0.0

00

0.0

20

0.0

20

0.9

88

0.1

18

Resid

ual

33.8

53

993.8

30

2.0

81

2.7

78

28.4

35

8.6

41

VC

for f

ixed e

ffects

3.6

23

82

.35

10

.02

30

.24

20

.68

60

.11

3

PV

C(M

L:

So

urc

e)1.1

1%

0.5

4%

0.2

6%

26.2

0%

2.6

1%

-2.8

7%

PV

C(S

ou

rce)

<-1

00%

16.3

3%

≈ 0

.00%

-48.7

9%

-0.9

6%

≈ 0

.00%

PV

C(T

ran

spla

nt)

61.9

8%

<-

100%

49.0

3%

88.7

0%

-56.1

2%

-13.4

5%

PV

C(R

esid

ual

s)3.4

6%

3.3

4%

-0.0

1%

2.3

6%

1.8

1%

0.7

9%

R2

GL

MM

(m)

9.1

7%

7.2

0%

1.0

1%

7.7

0%

2.1

9%

1.2

4%

R2

GL

MM

(c)

14

.35

%1

3.1

6%

7.2

3%

11

.46

%9

.15

%4

.79

%

AIC

c (

Fu

ll)

57

19

67

40

48

06

22

78

82

74

29

39

AIC

c (

Nu

ll)

57

57

67

64

48

08

22

96

82

92

29

40

Heig

ht

grow

thL

AS

LA


145

Table 4.9. Generalised linear mixed effects models testing variation in Eucalyptus rudis


December 2014 (6 months post-transplant) and in December 2015 (18 months post-

transplant) under the local versus foreign model. The modelled responses are a deviance

in trait value in each individual seedling from the mean trait value of the local source at

the transplant site traits (a, b) height growth (c, d), leaf area and (e, f) specific leaf area.

The intercept in the full model represents the seedlings sourced at the lowest rainfall site

(S539) at the lowest rainfall transplant site (T550) without insecticide. Coefficients in

bold are statistically different from zero (P < 0.05), and the abbreviations are as defined

in Table 4.5. ◄

Figure 4.8. The response of Eucalyptus rudis seedling traits to transplantation along a

rainfall gradient in December 2014 (6 months post-transplant) and in December 2015 (18

months post-transplant) under the local versus foreign model. The responses represent the

deviance in trait value of each individual from the mean trait value of the local source for

each transplant site (0) for traits (a, b) height growth (c, d), leaf area and (e, f) specific

leaf area. A shift of 0 mm rainfall denotes the home site of each source provenance (S538-

S1214, labelled according to mean rainfall per annum at the source site). The fitted lines

are the model predictions from the generalised linear mixed models presented in Table

4.9 (± 95% Confidence intervals). The models presented exclude transplantation site

T800 due to low sample size. ►


147

4.4 Discussion

Variation in phenotypic traits across environmental gradients may be due to fixed genetic

differences, maternal effects and plastic responses to environmental stimuli. Using a

large-scale reciprocal transplant of E. rudis seedlings across a 660 mm rainfall gradient I

was able to isolate fixed from inducible traits at a fine spatial scale. I found that early life

history traits in E. rudis are highly plastic in response to variation in their rearing

environment, irrespective of the source environmental conditions of the maternal lineage.

Comparisons within each transplant site revealed that seedlings did not differ in their

measured leaf traits with respect to source site origin. For seedling height-growth rate,

responses were similarly strongly driven by rainfall at the transplant site, but I also

identified a weak but significant effect of source site origin on height-growth. Seedlings

of maternal lineages from the drier source site locations (rainfall <700 mm pa) expressed

a conservatism in height-growth when grown at the higher rainfall transplant sites,

suggesting these genotypes may have differentiated from the high rainfall populations. I

discuss these findings and their contribution to understanding the evolution and

maintenance of plasticity in early life history traits of long-lived species. Translating these

findings into on-the-ground management actions to conserve species and ecosystem

functioning will be crucial for critically important water-dependent ecosystems under

novel rainfall regimes.

4.4.1 Trait plasticity as the dominant explanation for phenotype variation

On transplantation to the six-common garden sites, the measured seedling responses,

height-growth, LA, SLA and survival, all showed consistent positive covariance with

transplantation site rainfall, regardless of seedling source. First, this validates the

experimental design, in that the reciprocal transplantation gradient successfully induced

a gradient of environmental stress, which I infer is driven by rainfall differences (and

associated proximate variation in hydrological stress). Second, it shows that the selected

148

traits measured in E. rudis had extremely sensitive responses to this stress, albeit in a

highly plastic manner. For seedling leaf traits (LA and SLA), I found no evidence of trait

fixation to any of the source environmental conditions examined. This result was

somewhat unexpected since a number of leaf traits, including SLA, have been shown to

express elements of trait fixation among provenances of Eucalyptus species grown under

common field (Warren et al. 2006, Mclean et al. 2014) and glasshouse conditions (Gibson

et al. 1995, Li et al. 2000, Gauli et al. 2015). For instance, in a glasshouse trial across

provenances of the wide-ranging, arid zone eucalypt, E. microtheca, Li et al. (2000) found

a significant negative relationship between SLA and decreasing rainfall at the source site.

Furthermore, this relationship was expressed under both high and low moisture

treatments, indicating that SLA is a relatively fixed trait in E. microtheca. Even in

E. tricarpa, where plasticity among provenances was observed over a similarly large

rainfall gradient in eastern Australia (Mclean et al. 2014), there was still evidence of

differential plasticity in leaf trait expression among provenances. For E. rudis in the

Warren catchment I found no comparable evidence of differentiation in leaf traits to

source rainfall, in terms of either trait mean values or differential plasticity among

lineages.

In contrast to leaf traits, I found that height-growth in low rainfall sourced

seedlings was significantly lower than the locally sourced provenances at high rainfall

transplant sites. This indicates that the dry-sourced seedlings are expressing conservatism

of growth under high resource conditions. The effect was only observed among the driest

source populations (sourced < 700 mm pa) and only measurable at the higher rainfall

transplant sites. Conversely, when high rainfall sourced seedlings were transplanted to

the low rainfall conditions, height-growth was indistinguishable from the locally sourced

seedlings, indicating a higher plasticity in height-growth for the provenances sourced

under high rainfall. Although height may not be explicitly under selection, it is commonly


149

used as a measure of performance among tree seedlings in lieu of destructive sampling

(O’Brien and Krauss 2010, Breed et al. 2016). If height-growth differences among

lineages are due to heritable differences in traits, this could be indicative of (1) an

alternative fixed, resource allocation strategy, or (2) selection for a reduction in height-

growth plasticity, as I discuss further below. Of course, robust detection and

determination of trait heritability among lineages would require the study of multiple

generations, selective crosses or genetic heritability studies (Lopez et al. 2003, Ǻgren and

Schemske 2012, Rix et al. 2012, Halbritter et al. 2015). As E. rudis is a long-lived species,

and there are currently no genetic data available for these populations, the possibility of

non-heritable maternal influences on growth rates among lineages requires further

consideration, but cannot be discounted here.

4.4.1.1 Maternal seed investment

Seed mass was shown to increase significantly with rainfall at source site, and further,

seed mass was a strong predictor of early height-growth under glasshouse conditions.

These initial observations under controlled, glasshouse conditions indicate that there

could be a strong element of variation in maternal investment in E. rudis seeds, leading

to a growth advantage in early establishment for seedlings of maternal lineages with high

seed mass. This result adds to the body of literature linking seed mass in Eucalyptus to

offspring vigour (López et al. 2000, Lopez et al. 2003, Harrison et al. 2014). There is the

possibility, then, that the conservatism observed in height growth of low rainfall sourced

seedlings six-months post-transplant could be a lagged effect of this differential maternal

investment, particularly given that the difference dissipated by 18-months post-transplant.

However, in the glasshouse trials, I showed that after accounting for the variation in

seedling height attributed to seed mass, seeds sourced from low rainfall regions produced

taller seedlings than predicted based on seed mass alone, and conversely, seedlings

sourced from the high rainfall regions were shorter than expected. This suggests that

150

variation in absolute seed mass due to maternal investment only partially explains the

observed conservatism in low rainfall seedlings, and there is almost certainly a significant

source provenance effect on early establishment traits due to genetic differentiation

among populations across the rainfall gradient.

4.4.1.2 Conservatism in seedling height-growth

4.4.1.2.1 Fixed allocation of resources towards unmeasured traits

Fixation of traits within populations might be expected under conditions of high spatial,

but low temporal variability, such as across strong, but consistent environmental gradients

with tougher selective pressures (Kawecki and Ebert 2004). The first potential reason for

a reduced relative growth rate in dry-sourced seedlings, when grown under higher

resource conditions could be due to lower resource capture efficiency, stemming from

traits selected under their source site conditions and which are ‘maladaptive’ for

conditions at the transplant site. In dry Mediterranean-type climates, one of the greatest

selective pressures in early establishment is survival through the first summer (Ruthrof et

al. 2010, Hallett et al. 2011), particularly for an obligate riparian species (Stella and

Battles 2010a, Stella et al. 2010a). I hypothesise that the dry sourced seedlings could have

a greater fixed allocation of resources towards taproot growth in order to gain access to

the water table and increase their likelihood of survival over the first summer season. At

higher rainfall sites, dry-sourced seedlings grew comparatively slower than locally-

sourced seedlings, which could then be a consequence of greater fixed allocation of

resources towards vertical water-seeking roots over finer surface roots targeting nutrient

capture (Lamont 1982, Hamer et al. 2015b). Alternatively, it could indicate a greater

overall fixed-allocation towards below-ground mass over the measured above-ground

mass. In glasshouse trials of E. camaldulensis (the functional equivalent of E. rudis in

riparian systems across eastern, northern and central Australia), seedlings sourced from

populations in semi-arid and dry-tropical regions had a greater proportion of biomass


151

allocated to root mass than seedlings sourced from the humid tropics (Gibson et al. 1995).

Similarly, glass house trials demonstrated that root mass to leaf area ratios in the arid zone

Eucalypt, E. microtheca, increased with decreasing rainfall at the provenance source (Li

et al. 2000). Moreover, if early allocation to below-ground mass is the mechanism behind

the lowered growth rates, it could also be an independent explanation for the dissipation

of the provenance effect at 18-months post-planting as roots have established and

presumably accessed the water table. In addition to root allocation, traits such as increased

wood density (resistance to cavitation, wilting; Stackpole et al. 2010, Freeman et al. 2013)

and greater lignotuber storage of carbohydrates (capacity to recover from disturbances

such as fire or drought; Whittock et al. 2003, Gauli et al. 2015) have been demonstrated

to vary among provenances, and could also explain the observed conservatism in growth

among low rainfall provenances. Unfortunately, no below-ground biomass allocation data

are available at the present time to tease apart these alternative hypotheses. Finally, while

alternative allocation of resources towards defence from insect herbivory was tested

indirectly via the insecticide treatment, the lack of significant treatment effect suggests

that total allocation of resources towards inducible defence may be minimal (O’Reilly-

Wapstra et al. 2013). Although, given the high variability in insect herbivory observed

across the rainfall gradient (pers. obs.), there may be variation in plant chemical defence

traits unmeasured here worthy of further investigation.

4.4.1.2.2 Selection for differential plasticity among provenances

The second potential reason for reduced relative growth rate of dry-sourced seedlings

under higher resource conditions could be selection for a reduction in height-growth

plasticity. Variation in the degree of plasticity among provenances has been observed in

E. tricarpa in eastern Australia (Mclean et al. 2014), as lower growth of drier provenances

relative to higher rainfall provenances of E. marginata (O’Brien et al. 2007), and a

number of other Eucalyptus species (Warren et al. 2006). For example, Baquedano et al.

152

(2008) compared plasticity between populations of water-limited Pinus halepensis

sourced from rocky outcrops, compared to those from the wider region in Spain, and

suggested that the conservative growth observed in the outcrop sourced plants when

placed under high resource conditions may be an adaptive response to limit the production

of large or inefficient structures that are too costly to maintain when conditions worsen.

Richter et al. (2012) also found some evidence of this while examining Pinus sylvestris

seedling responses to altered temperature and rainfall regimes in Europe. On exposing

Spanish-Mediterranean and Swiss-Alpine sourced seedlings (i.e. greater and lesser

severity in summer drought respectively) to a high spring-rainfall treatment, they

observed a greater plasticity in resource allocation among the Swiss seedlings (Richter et

al. 2012). Although the Swiss seedlings initially showed a higher shoot: root ratio during

early growth, they subsequently experienced a greater mortality rate during summer

drought; offering a potential mechanism for the selection of differential plasticity as a

trait in itself.

4.4.1.3 Plasticity in Mediterranean-type climates

Regardless of the mechanisms of the observed height-growth conservatism, the

overwhelming response of the E. rudis provenances examined in this experiment was

towards an extremely high level of environmental plasticity. What is particularly

interesting about this result is that the (weak) conservatism in height-growth observed in

dry provenances shows that populations are diverging in their adaptive strategies, and yet

extreme plasticity in leaf traits has been maintained (or evolved). Trait plasticity is

predicted to arise where gene flow is high and dispersal away from the maternal

environment is frequent, or where organisms live in extremely heterogeneous

environments (Sultan and Spencer 2002). Although E. rudis demonstrates hydrochorous

dispersal (seed dispersed via water; Pettit and Froend 2001a), it is unidirectional, thus

gene immigration rates via seed is not likely to be significant into populations in the upper


153

tributaries. Gene flow via pollen dispersal has not been explicitly examined in E. rudis.

However, based on evidence from E. wandoo that overlaps in the dry extent of the

E. rudis range examined here (and is also principally insect pollinated), gene flow is likely

to occur over fairly low distances, the majority of pollination events occurring for E.

wandoo within 1 km (Byrne et al. 2008). For instance, in E. pauciflora population genetic

structure was found to be largely independent at distances greater than ca 27 km (Gauli

et al. 2015), and in other Eucalyptus species this can be up to 50 km where vertebrate

dispersers are active (Breed et al. 2012). For E. camaldulensis, which inhabits an almost

identical riparian niche in eastern and central Australia, leaf traits (amongst other traits)

have been found to be largely fixed regardless of catchment location (ranging from 400

to 1200 mm pa rainfall; Gibson et al. 1995). Even if dispersal is occurring over distances

as large as 50 km in E. rudis, climate conditions vary gradually across the Warren River

Catchment and seeds are unlikely to dispersal to climatic conditions that are sufficiently

different to the maternal environment to warrant such high observed plasticity in traits.

Instead, I argue that the strong seasonal heterogeneity in climate drives plasticity

in E. rudis. I observed E. rudis growth to be greatest during the spring and autumn

periods, whereas growth was limited by cooler temperatures in winter and by water

availability in summer. High plasticity of leaf traits potentially allows seedlings to amass

a larger total surface area of ‘cheaper’ high-SLA leaves during good conditions (warmer

autumn and spring growing seasons when rainfall is high), then transition to tougher,

more sclerophyllous and water efficient, low-SLA leaves to carry seedlings over the

summer drought-like conditions when the river dries. As riparian trees that probably

respond most strongly to soil moisture change and which transitions gradually between

seasons, they are less likely to express maladaptive phenotypes in response to atypical

summer rainfall events. For example, on the examination of the rate of soil moisture draw-

down on Californian riparian trees, Stella and Battles (2010) demonstrated that

154

Populus fremontii subsp. fremontii seedlings not only reduced SLA in response to annual

summer soil moisture draw-down, but the reduction in SLA increased with the rate of

drawdown. Here, I measured the leaf traits during the same calendar week, at the end

spring. Following a much drier than average spring in 2015 SLA was substantially lower

across all transplant sites than observed in 2014 (www.bom.gov.au/). While this change

could be the result of an ontogenic shift in leaf morphology (Jordan et al. 2000), all

seedlings throughout the duration of this study were demonstrating the rounder juvenile

leaf morphology, suggesting it was more likely an environmental response. Plasticity in

E. rudis allows individuals to respond to temporal variability in length, as well as in the

date of onset (and cessation) of growing periods, but adjust form accordingly and with

reliable environmental cues when the season does change. Moreover, a high degree of

plasticity appears to be highly successful across the entire catchment, not just in the low

rainfall regions.

Identifying overarching trends in the occurrence of plasticity and local adaptation

in plant traits remains a critical aim in the prediction of species responses to climate

changes (Valladares et al. 2014); a problem made more challenging where traits differ in

the magnitude of plastic response to the same environmental conditions, as observed here

and elsewhere (Warren et al. 2006, Mclean et al. 2014). My results indicate that in long-

lived species, plasticity of traits in response to water availability may be more

advantageous in traits which are temporally flexible. Across European temperature

gradients, environmental cues are more commonly found to trigger phenological events

(e.g. leaf fall, bud break; Kramer 1995, Vitasse et al. 2010) and control physiological and

morphological leaf traits (Bresson et al. 2011) than genotype. In contrast, traits such as

growth rate and drought or frost resistance (i.e. susceptibility to xylem cavitation; Choat

et al. 2012) which are relatively inflexible over the lifetime of the individual have shown

http://www.bom.gov.au/


155

greater source-site fidelity (Stackpole et al. 2010, Dutkowski and Potts 2012, Montwé et

al. 2016).

4.4.2 Implications for management

As the climate of southern Australia dries, the frequency of droughts and heatwaves

increase, and ecosystems across SWWA are beginning to show signs of stress (e.g.

Chapter 3, this thesis; Pekin et al. 2009, Brouwers et al. 2013, Evans et al. 2013, Matusick

et al. 2013, Brouwers and Coops 2016), management of these systems over the coming

decades is going to require increasingly intensive and deliberate actions, simply to

maintain the current, degraded state of the landscape. A broader understanding of how

species respond to environmental changes will allow us to better predict how species

might respond to future climate change (e.g. Valladares et al. 2014) and implement

climate adaptation strategies to reduce the vulnerability of forest systems to climate

change (Aitken and Whitlock 2013, Prober et al. 2015). I transplanted E. rudis seedlings

into drier climates that mimicked rainfall declines of up to 54% (up to a 660 mm pa deficit

from pre-1970s rainfall). Incredibly, plasticity in growth and leaf morphology traits

enabled seedlings to tolerate the harsher, low rainfall transplant site conditions with no

measurable differences relative to the locally sourced seedlings, even after 18-months

post-transplant. Recent climate downscaling over the SWWA estimate declines in winter

rainfall of up to 28% of historical levels by 2030, and up to a further 13% (low emissions,

RCP2.6) to 44% rainfall decline (high emissions, RCP8.5) by 2090 (Silberstein et al.

2012, CSIRO and Bureau of Meteorology 2015). Under all but the highest emission

scenario, the plasticity in E. rudis traits identified in this study strongly suggests that we

will not see substantial population threat (or extinction) based on the current range and

capacity to respond to variation in environmental conditions. Although, further

examination of local adaptation in the responses of the low rainfall populations to the

drying climate will be required, such as the transplantation to sites outside of the current

156

range to determine the mechanisms defining the lower rainfall limit. While it does not

look like E. rudis is seriously threatened by the projected climatic changes, the phenotypic

character of riparian systems will be inexorably altered by this level of environmental

change, which may have cascading impacts on the communities it supports. A shift in the

canopy structure may significantly alter the microclimate for the associated understory

flora. Equally, a shift in mean leaf trait towards a lower turnover of tougher, of higher

SLA leaves, for example, could have severe effects on the productivity of both the soil

and freshwater systems though lower input, but also longer decomposition times in soils

and freshwater environment (Madeira et al. 1995, Ribeiro et al. 2002). Finally, divergence

within the low rainfall, ‘stressed’, populations of the Warren Catchment, coupled with a

significantly lower survival rate in seedlings, suggest that selection under these dry

populations may be acting at a faster rate than throughout the rest of catchment; where it

is most required.


157

4.5 Supplementary Material

Figure S4.1. Survival at 18 months post-transplant for Eucalyptus rudis seedlings planted

in experimental gardens at different points along a rainfall gradient. A shift of 0 mm

rainfall denotes the home site of each source provenance (S538-S1214, labelled according

to mean rainfall per annum at the source site). Each point represents the proportion of

each maternal lineage surviving at each transplant site. The fitted lines are the model

predictions from a generalised linear mixed model (± 95% Confidence intervals)

including transplant site T800 (Table 4.5a). Note that points are jittered along the x axis

to reduce overlap.

158

Table S4.1. Model selection using AICc scores to compare generalised linear mixed

effects models testing the variation in Eucalyptus rudis seedling traits as a function of

source rainfall (S), transplant site rainfall (T) and insecticide treatment (I) under the home

versus away model. Model selection is presented for the response in traits, height growth,

leaf area and specific leaf area in 2014 (6 months post-transplant). These models include

transplant site T800, which was excluded from the main analysis due to low sample size.

The terms and abbreviations are as described in Table 4.6. The most parsimonious model

< 2 AIC was selected, and in bold.

k Log

Likelihood AICc ∆AICc

Height growth – 2014

I + S + S2+ T + T2 + S:T + S2:T + T2:S 13 -988 2002 0.000

I + S + S2 + T + T2 + T2:S 11 -990 2002 0.459

I + S + S2 + T + T2 + I:T + S:T + S2:T + T2:S 14 -987 2002 0.472

I + S + S2 + T + T2 + I:T + T2:S 12 -989 2003 0.958

I + S + S2 + T + T2 + I:S + S:T + S2:T + T2:S 14 -987 2003 0.968

I + S + S2 + T + T2 + S2:T2 + T2:S 12 -989 2003 1.057

I + S + S2 + T + T2 + I:S + T2:S 12 -989 2003 1.430

I + S + S2 + T + T2 + I:S + I:T + S:T + S2:T + T2:S 15 -986 2003 1.538

I + S + S2 + T + T2 + I:T + S2:T2 + T2:S 13 -989 2004 1.577

I + S + S2 + T + T2 + S:T + S2:T2 + S2:T + T2:S 14 -988 2004 1.794

I + S + S2 + T + T2 + S:T + T2:S 12 -990 2004 1.906

I + S + S2 + T + T2 + S:T + S2:I + S2:T + T2:S 14 -988 2004 1.939

Null 5 -1016 2042 40.255

Leaf area – 2014

I + S + S2 + T + I:S + S:T + S2:I 12 -1917 3858 0.000

I + S + S2 + T + I:S + S2:I 11 -1918 3859 0.630

S + S2 + T + S:T 9 -1920 3859 0.877

I + S + S2 + T + T2 + I:S + S:T + S2:I 13 -1916 3859 1.241

S + S2 + T 8 -1922 3860 1.500

I + S + S2 + T + I:S + S2:I + S2:T 12 -1918 3860 1.682

I + S + S2 + T + I:S + S:T +S2:I + S2:T 13 -1917 3860 1.750

I + S + S2 + T + T2 + I:S + S2:I 12 -1918 3860 1.863

I + S + S2 + T + I:S + I:T + S:T + S2:I 13 -1917 3860 1.934

Null 5 -1934 3878 20.220

Specific leaf area – 2014

I + S + S2 + T + T2 + I:T + S2:I + T2:I 13 -4083 8192 0.000

I + S + S2 + T + T2 + S2:I + T2:I 12 -4084 8193 0.737

I + S + S2 + T + T2 + I:T + S2:I + S2:T + T2:I 14 -4082 8193 1.341

I + S + S2 + T + T2 + I:T + S:T + S2:I + T2:I 14 -4083 8194 1.756

I + S + S2 + T + T2 + I:S + I:T + S2:I + T2:I 14 -4083 8194 1.807

I + S + S2 + T + T2 + I:T + S2:I + T2:I + T2:S 14 -4083 8194 1.938

I + S + S2 + T + T2 + I:T + S2:T2 + S2:I + T2:I 14 -4083 8194 1.983

Null 5 -4124 8259 67.134


159

Table S4.2. Generalised linear mixed effects models testing variation in Eucalyptus rudis


December 2014 (6 months post-transplant) under the home versus away model. The

modelled responses are a deviance in trait value of each individual from the mean trait

value of each source grown under conditions nearest to their source for traits (a) height

growth, (b) leaf area and (c) specific leaf area. The models presented include

transplantation site T800, which was removed from the main analysis. The intercept in

the full model represents the seedlings sourced at the lowest rainfall site (S539) at the

lowest rainfall transplant site (T550) without insecticide. Coefficients in bold are

statistically different from zero (P < 0.05), and the abbreviations are as defined in Table

4.5.

Response (a) Height growth (n = 873) (b) LA (n = 1311) (c) SLA ( n = 1309)

Fixed effects b [± 95% CI] b [± 95% CI] b [± 95% CI]

Intercept 0.28 [-0.10, 0.66] 0.10 [-0.22, 0.42] 2.12 [-1.44, 5.66]

Source(Lin) 0.09 [-0.47, 0.66] -0.18 [-0.59, 0.23] -6.06 [-8.53, -3.60]

Source(Quad) -1.23 [-2.40, -0.06] -1.62 [-2.60, -0.65] -4.71 [-10.50, 1.08]

Transplant(Lin) 0.76 [0.51, 1.00] 0.68 [0.33, 1.04] 7.95 [4.35, 11.55]

Transplant(Quad) 0.52 [-0.21, 1.25] 3.77 [-7.16, 14.71]

Insect -0.23 [-0.33, -0.13] 1.07 [-0.29, 2.44]

Source: Transplant(Quad) -0.78 [-1.36, -0.20]

Insect: Source(Quad) -9.07 [-11.93, -6.21]

Insect: Transplant(Quad) 3.66 [0.13, 7.19]

VC for random effects

ML/ Source 0.017 0.073 0.807

Source 0.047 0.007 0.810

Transplant 0.017 0.042 4.484

Residual 0.544 1.057 29.095

VC for Fixed effects 0.262 0.300 39.206

PCV(ML/ Source) -25.86% -4.30% -27.88%

PCV(Source) 69.89% 96.79% 96.98%

PCV(Transplant) 89.34% 75.08% 77.84%

PCV(Residual) 3.11% -0.05% 3.13%

R2glmm(m) 29.52% 20.27% 52.69%

R2glmm(c) 38.70% 28.48% 60.89%

AIC (Full model) 2002 3859 8192

AIC (Null model) 2042 3878 8259

160

Figure S4.2. The response of Eucalyptus rudis seedling traits to transplantation along a

rainfall gradient in December 2014 (6 months post-transplant) under the home versus

away model. The responses are a deviance in trait value of each individual from the mean

trait value of each source grown under conditions nearest to their source (0) for traits (a)

height growth (b), leaf area and (c) specific leaf area. A shift of 0 mm rainfall denotes the

home site of each source provenance (S538-S1214, labelled according to mean rainfall

per annum at the source site). The fitted lines are the model predictions from the

generalised linear mixed models presented in Table S4.2 (± 95% Confidence intervals).

The models presented include transplantation site T800 which was removed from the

main analysis due to low sample size.

161

5 Synthesis and Conclusions

In the years since I defined the objectives for this thesis, the Great Barrier Reef has

experienced the largest bleaching event in recorded history (Hughes et al. 2017); large

craters have formed in the permafrost (Tesi et al. 2016); the incidence of lightening

ignited wildfires are increasing across Boral North America (Veraverbeke et al. 2017);

and currently, a 5,800 km2 iceberg is breaking off the Larsen C ice shelf in Antarctica,

with the potential to destabilise the entire ice shelf and the glaciers which feed it (Zhao et

al. 2017). Closer to home, the summer of 2017 bought heatwaves that broke the highest

temperature records throughout metropolitan Sydney and Brisbane and right across rural

south and eastern Australia (Bureau of Meteorology 2017) and this past June recorded

the lowest rainfall records over the much of the southwest of Western Australia (SWWA;

King 2017). Even if we manage to curb further emissions, we are looking at an increase

in mean global temperatures of at least 2°C, with extremes ranging well outside of the

historical climatic conditions (Solomon et al. 2009, IPCC 2014b) and ecosystems

globally, are showing signs of stress (e.g. Parmesan and Yohe 2003, Allen et al. 2010,

2015, Matusick et al. 2013, Hughes et al. 2017, Pecl et al. 2017). It is becoming

increasingly clear that the context under which decisions regarding the management of

natural resources and biodiversity are made, is changing.

In this thesis, I explored the concepts of ecosystem resistance, resilience and

adaptation to climate change in riparian ecosystems. I used the Warren River and its major

tributaries, the Tone River and Murrin Brook, as a ‘transect’ across the regional rainfall

gradient of the Mediterranean-climate zone of south-west Western Australia (SWWA). I

examined these concepts at a community (Chapter 2), species (Chapter 3), and intra-

specific (Chapter 4) level, with the overarching aim of determining the exposure and

sensitivity of the riparian flora to the recent rainfall declines as an indication of their

vulnerability to further declines predicted under climate change.

162

5.1.1 Riparian flora at risk

In Chapter 2, I show that variability in both the canopy and understorey assemblages of

the riparian zone are driven largely by the longitudinal, climate gradients rather than the

local hydrological regime, indicating that the riparian assemblages may be more

vulnerable to rainfall declines than initially anticipated. In examining the impact of the

streamflow deficits observed over just the past 30-years on the distribution of juvenile

trees and shrubs in Chapter 3, I confirmed their vulnerability. I presented evidence that

the reductions in hydroperiod and recurrence interval observed to date, have driven a

mismatch in the geographic ranges of the mature and immature populations of a number

of the common riparian species; an early warning of a contraction of their optimal climatic

niches. While the results of these first two research chapters are correlative, observational

datasets, they put forward a compelling case for the likelihood of significant climate

driven range shifts in the SWWA flora in the near future. Furthermore, as the majority of

the riparian species, both facultative and obligate, did not show an upper rainfall limit to

their distribution, i.e. many species were observed within the lower reaches of the river,

there is almost no potential for compensatory range expansion.

As discussed in Chapter 3, I do not expect that a complete loss of habitat for the

majority of the riparian species since there is little chance the river will cease to flow

completely (Barron et al. 2012, Silberstein et al. 2012), there will be pockets of riparian

habitat, albeit over a reduced range. The risk then might come in from greater competition

from encroachment of more facultative species (Chapter 3; Rood et al. 2010). For a

handful of species, particularly the obligate riparian species already restricted to the

highest rainfall regions of the catchment, such as Taxandria juniperina and T. linearfolia

that are likely dependent on permanent soil saturation, the eventuality of high emissions

scenarios may lead to local population extinction. In contrast to many rivers around the

world, the Warren is a free-flowing river thus there is limited capacity to prescribe

Chapter 5: Synthesis & Conclusions

163

ecological flows to ensure the health of the vegetation (Acreman and Dunbar 2004,

Arthington et al. 2006, Palmer et al. 2009, Poff et al. 2010, Stella et al. 2010b, Miller et

al. 2013). Instead, species survival may depend on engineered solutions or more active

management, such as maintaining small pockets of permanently saturated creek beds

downstream from small, independently managed onstream-farm dams (up to 420 of

which are spread across the Warren and the neighbouring Donnelly Catchment;

Department of Water 2012).

5.1.2 Limits to buffering capacity of the river system

In comparison to the riparian species, the majority of the upland species examined did not

demonstrate the recruitment failures that were observed in the riparian species. While this

could suggest that the riparian zones may have some capacity to buffer regional rainfall

gradients (Chapter 3), the extremely high turnover across the catchment suggests that

upland species are not utilising the water source (Chapter 2) and that ultimately the

capacity to buffer climate changes in the long term is limited. Follow up surveys would

be justified to see whether recruitment failure is occurring throughout the non-riparian

extent of their range, or more manipulative experiments to investigate the roots systems

and primary means of water capture across some of these species, could be used to

substantiate these patterns and determine whether they are accessing the higher water

tables in the riparian zone. Ultimately, these results indicate that the river systems of the

SWWA, and likely others globally have a limited capacity to buffer climate aridification.

It would be interesting to look at similar systems elsewhere, where the majority of the

rainfall is received in the headwaters such as some of the Mediterranean basin rivers, and

where the water source has a greater independence from the local climate (e.g.

(Karrenberg et al. 2003, Bruno et al. 2014).

164

5.1.3 Keystone canopy assemblages

I show that the understory communities of the Warren River are inexplicitly linked to the

canopy assemblages, moreover, that the association is driven by more than commonalities

in their hydraulic and climatic niches (Chapter 2). In Chapter 3, I looked at range

contraction at a species level, and show that although there were generalities in among

functional groups to streamflow deficits, each species presented slightly different realised

niches, with variable sensitivities in recruitment under changed hydrological regimes.

How shifts in species assemblages across the region will manifest then, will likely depend

on more than just shifts in the climatic optima, but also, changes in their associated species

assemblages. Further experimental examination of the mechanisms driving these species

associations is necessary to determine whether they are first, biotic or abiotically driven

(namely, soils), and second, investigate the consequences of mismatches among drivers,

including the keystone canopy species. Should a strong link between the species and their

dominant canopy assemblage be identified, and more importantly, the canopy species

identified as at severe risk (Hamer et al. 2015a) rather than resilient (Chapter 4; E. rudis),

strategies such as assisted migration could be considered as means to facilitate entire

assemblage migrations. If the association is an indirect abiotically driven association

however, i.e. via soils (Lamont 1982, Hopper and Gioia 2004, Hamer et al. 2015b, Turner

et al. 2017) such practices may be risky, but also futile.

On a more promising note, experimental examination of the apparent resilience

of the dominant canopy species, E. rudis via a large translocation experiment (Chapter 4)

demonstrated an incredibly high level of plasticity towards local climate conditions. This

result suggests that E. rudis will have the capacity to withstand the projected changes

throughout much the catchment and will likely not face climate driven extinction, under

even the highest emissions scenario. Additionally, this experiment showed that there has

been a divergence in the drier regions of the catchment, the nature of the adaptation


165

requires follow up research (i.e. examination of root structures), but the results presented

here indicate a dry adapted provenance. While this has the potential to be utilised towards

increasing adaptive capacity, more interestingly, it suggests the natural adaptive potential

of this, and potentially other Eucalyptus species may be high. Finally, although E. rudis

may not be susceptible to climate changes directly, the major limitation of this experiment

is determining the threats by defoliating insects. E. rudis is known to be susceptible to

defoliating insects which have caused mortalities in parts of its range (Clay and Majer

2001, Edwards 2010). Further experimental examination of the biotic interactions and

defence components may identify provenances with greater resilience to insect damage,

which might ultimately be important than drought tolerance (i.e. adaptation to insect

defence; O’Reilly-Wapstra et al. 2013, McKiernan et al. 2014, Bustos-segura et al. 2017).

5.1.4 Increasing resilience via climate adaptive restoration

While a substantial proportion of the Warren Catchment has intact riparian vegetation,

there are large pockets of riparian zones devoid of vegetation, damaged and degraded

from grazed by livestock, and completely cleared through farmland within the catchment,

as well as other SWWA catchments. Across the upper tributaries in particular, remnant

native forest blocks are highly fragmented and often devoid of understorey vegetation

due to agricultural grazing. Although in more limited extent, in the lower catchment too,

large areas adjacent to horticultural and agricultural parcels have been cleared, or where

declines of weed infestations have left sections of the understorey completely open (e.g.

Aghighi et al. 2014, Yeoh et al. 2016). Restoration of these areas can play a vital role in

increasing the resilience of species and the local industry to climate changes. From a

conservation perspective, increasing the connectivity between forest blocks can enhance

migration potential for plant species (Renton et al. 2012) and provide resilient habitat for

wildlife (Seabrook et al. 2014, Nimmo et al. 2015). From a societal perspective, a

functional riparian assemblage provides ecosystem services ranging from filtering runoff

166

and erosion control to economically valuable tourism and recreational activities. Ensuring

a functional riparian community into a drier future will require the active and deliberate

management of ecosystems.

While in comparison to the rest of the world, the global circulation patterns driving

the weather systems over SWWA are well understood and there is a high degree of

certainty in the declining winter and spring rainfall projections, increases in temperatures,

heatwaves and droughts (Hope et al. 2006, 2015), there is a high level of uncertainty in

other climate components, such as the sporadic summer and autumn rainfall patterns.

Equally, the severity of the changes we observe is highly dependent on the effectivity of

the mitigating actions taken in the next few critical decades to reduce the emission of

greenhouse gases in to the atmosphere. Thus, the paradox in planning for climate change

in ecological systems is in the fact that, with long intergenerational timeframes, the

success of risker adaptation strategies, such as assisted migration, is greater if

implemented early, but, the risk of interfering well in advance, with little certainty in the

degree of climatic change or in mitigating actions, could be perceived as riskier than doing

nothing (Dessai and Hulme 2007, Stein et al. 2013, Wise et al. 2014, Bradford and Bell

2016). The major goals of adaptation frameworks are to identify the threats to specific

processes from changing climates, but also, identify the point at which management

actions should be implemented (Dessai and Hulme 2004, 2007, Wise et al. 2014).

Successful planning for climate adaptation will depend a firm understanding of the

ecological processes, vulnerabilities, and tipping points; the research undertaken in this

thesis is a step towards this goal.


167

5.1.5 Conclusions

The role of river systems in providing refugia under climate change is not to be assumed.

Here, it was hypothesised that a river spanning a significant rainfall gradient it would

offer refuge from rainfall deficits for riparian vegetation. Instead, I found a riparian

assemblage far more reliant on regional rainfall gradients than the local hydrological

regime. What is more, the obligate riparian flora is beginning to show signs of a

geographic shift in their optimal climatic niche, indicating that the water dependent

species are incredibly vulnerable to rainfall declines. The high turnover in species

assemblages along the riparian corridors where species ought to be accessing ground

water, indicates species are highly dependent on surface water and potentially maladapted

to take advantage of the shallow surface waters of the riparian zone. The species of the

high rainfall regions appear to be expressing traits adapted to these high rainfall

conditions, rather than the resilient, dry hardened flora, expected under Mediterranean

climates. On a more positive note, I found an incredibly high plasticity to projected

rainfall declines in keystone riparian tree E. rudis. I suggest that the high plasticity in this

species may be an evolutionary response to the sharp contrast in seasonal conditions.

Moreover, I anticipate that the level of plasticity in this species presents resilience to the

coming changes and offers the potential for enhancing resilience in these vital

ecosystems. Finally, the results presented here form the baseline of our knowledge on

resilience and adaptation in riparian systems in SWWA and widen the broader

understanding of species responses to climate change.

168

6 References

Abbott, I. 1984. Ecological features of an outlying stand of jarrah (Eucalyptus

marginata) at Jilakin Rock, Western Australia. Journal of the Royal Society of

Western Australia 66:107–110.

Ackerly, D. D., W. K. Cornwell, S. B. Weiss, L. E. Flint, and A. L. Flint. 2015. A

geographic mosaic of climate change impacts on terrestrial vegetation: Which

areas are most at risk? PloS one 10:e0130629.

Acreman, M. C., and M. J. Dunbar. 2004. Defining environmental river flow

requirements – a review. Hydrology and Earth System Sciences 8:861–876.

Aghighi, S., L. Fontanini, P. . Yeoh, G. E. S. . Hardy, T. . Burgess, and J. . Scott. 2014.

A conceptual model to describe the decline of european blackberry (Rubus

anglocandicans), a weed of national significance in Australia. Plant Disease

98:580–589.

Ǻgren, J., and D. W. Schemske. 2012. Reciprocal transplants demonstrate strong

adaptive differentiation of the model organism Arabidopsis thaliana in its native

range. New Phytologist 194:1112–1122.

Aguiar, F. C., M. T. Ferreira, and A. Albuquerque. 2006. Patterns of exotic and native

plant species richness and cover along a semi-arid Iberian river and across its

floodplain. Plant Ecology 184:189–202.

Aitken, S. N., and J. B. Bemmels. 2016. Time to get moving: Assisted gene flow of

forest trees. Evolutionary Applications 9:271–290.

Aitken, S. N., and M. C. Whitlock. 2013. Assisted gene flow to facilitate local

adaptation to climate change. Annual Review of Ecology, Evolution, and

Systematics 44:367–388.

Aitken, S. N., S. Yeaman, J. A. Holliday, T. Wang, and S. Curtis-McLane. 2008.

Adaptation, migration or extirpation: climate change outcomes for tree

populations. Evolutionary Applications 1:95–111.

Alberto, F. J., S. N. Aitken, R. Alía, S. C. González-Martínez, H. Hänninen, A. Kremer,

F. Lefèvre, T. Lenormand, S. Yeaman, R. Whetten, and O. Savolainen. 2013.

Potential for evolutionary responses to climate change - evidence from tree

populations. Global Change Biology 19:1645–1661.

Alexander, J. M., J. M. Diez, and J. M. Levine. 2015. Novel competitors shape species’

responses to climate change. Nature 525:515–518.

Alexander, L. V, and J. M. Arblaster. 2009. Assessing trends in observed and modelled

climate extremes over Australia in relation to future projections. International

Journal of Climatology 29:417–435.

Allen, C. D., D. D. Breshears, and N. G. McDowell. 2015. On underestimation of global

vulnerability to tree mortality and forest die-off from hotter drought in the

Anthropocene. Ecosphere 6:art129.

Allen, C. D., A. K. Macalady, H. Chenchouni, D. Bachelet, N. McDowell, M.

Vennetier, T. Kitzberger, A. Rigling, D. D. Breshears, E. H. Hogg, P. Gonzalez, R.

Fensham, Z. Zhang, J. Castro, N. Demidova, J. H. Lim, G. Allard, S. W. Running,

A. Semerci, and N. Cobb. 2010. A global overview of drought and heat-induced

tree mortality reveals emerging climate change risks for forests. Forest Ecology

and Management 259:660–684.

Anand, R. R., and M. Paine. 2002. Regolith geology of the Yilgarn Craton, Western

Australia: Implications for exploration. Australian Journal of Earth Sciences 49:3–

169

162.

Anderson, M. J., and N. A. Gribble. 1998. Partitioning the variation among spatial,

temporal and environmental components in a multivariate data set. Australian

Journal of Ecology 23:158–167.

Araújo, M. B., M. Cabeza§, W. Thuiller, L. Hannah, and P. R. Williams. 2004. Would

climate change drive species out of reserves? An assessment of existing reserve-

selection methods. Global Change Biology 10:1618–1626.

Arnold, T. W. 2010. Uninformative parameters and model selection using Akaike’s

Information Criterion. Journal of Wildlife Management 74:1175–1178.

Arthington, A. H., S. E. Bunn, N. L. Poff, and R. J. Naiman. 2006. The challenge of

providing environmental flow rules to sustain river ecosystems. Ecological

Applications 16:1311–1318.

Ash, J. D., T. J. Givnish, and D. M. Waller. 2017. Tracking lags in historical plant

species’ shifts in relation to regional climate change. Global Change Biology

23:1305–1315.

Ashcroft, M. B., L. A. Chisholm, and K. O. French. 2009. Climate change at the

landscape scale: Predicting fine-grained spatial heterogeneity in warming and

potential refugia for vegetation. Global Change Biology 15:656–667.

Ashcroft, M. B., and J. R. Gollan. 2013. Moisture, thermal inertia, and the spatial

distributions of near-surface soil and air temperatures: Understanding factors that

promote microrefugia. Agricultural and Forest Meteorology 176:77–89.

Atkins, K. E., and J. M. J. Travis. 2010. Local adaptation and the evolution of species’

ranges under climate change. Journal of Theoretical Biology 266:449–457.

Baquedano, F. J., F. Valladares, and F. J. Castillo. 2008. Phenotypic plasticity blurs

ecotypic divergence in the response of Quercus coccifera and Pinus halepensis to

water stress. European Journal of Forest Research 127:495–506.

Barron, O., R. Silberstein, R. Ali, R. Donohue, D. J. McFarlane, P. M. Davies, G.

Hodgson, N. Smart, and M. Donn. 2012. Climate change effects on water-

dependent ecosystems in south-western Australia. Journal of Hydrology 475:473–

487.

Barton, K. 2016. MuMIn: Multi-Model Inference. R package version 1.15.6.

https://CRAN.R-project.org/package=MuMIn.

Bates, B. C., P. Hope, B. Ryan, I. Smith, and S. Charles. 2008. Key findings from the

Indian Ocean Climate Initiative and their impact on policy development in

Australia. Climate Change 89:339–354.

Bates, D. 2005. Fitting linear mixed models in R. Using the lme4 package. R News

5:27–30.

Bates, D., M. Mächler, B. M. Bolker, and S. C. Walker. 2015. Fitting linear mixed-

effects models using lme4. Journal of Statistical Software 67:1–48.

Beard, J. S., G. R. Beeston, J. M. Harvey, A. J. M. Hopkins, and D. P. Shepherd. 2013.

The vegetation of Western Australia at the 1:3,000,000 scale. Explanatory memoir.

Conservation Science in Western Australia 9:1–252.

Beatty, S. J. 1984. Influence of microtopography and canopy species on spatial patterns

of forest understory plants. Ecology 65:1406–1419.

Beatty, S. J., D. L. Morgan, and A. J. Lymbery. 2013. Implications of climate change

for potamodromous fishes.

Bejarano, M. D., M. Gonzalez del Tanago, D. Garcia de Jalon, M. Marchamalo, A.

Sordo-Ward, and J. Solana-Guterrez. 2012. Responses of riparian guilds to flow

170

alterations in a Mediterranean stream. Journal of Vegetation Science 23:443–458.

Bejarano, M. D., C. Nilsson, M. González Del Tánago, and M. Marchamalo. 2011.

Responses of riparian trees and shrubs to flow regulation along a boreal stream in

northern Sweden. Freshwater Biology 56:853–866.

Bell, D. M., J. B. Bradford, and W. K. Lauenroth. 2014. Early indicators of change:

Divergent climate envelopes between tree life stages imply range shifts in the

western United States. Global Ecology and Biogeography 23:168–180.

Benavides, R., S. G. Rabasa, E. Granda, A. Escudero, J. A. Hódar, J. Martínez-Vilalta,

A. M. Rincón, R. Zamora, and F. Valladares. 2013. Direct and Indirect Effects of

Climate on Demography and Early Growth of Pinus sylvestris at the Rear Edge:

Changing Roles of Biotic and Abiotic Factors. PLoS ONE 8:17–19.

Bendix, J. 1994. Scale, direction, and pattern in riparian vegetation-environment

relationships. Annals of the Association of American Geographers 84:652–665.

Benomar, L., M. S. Lamhamedi, A. Rainville, J. Beaulieu, J. Bousquet, and H. A.

Margolis. 2016. Genetic Adaptation vs. Ecophysiological Plasticity of

Photosynthetic-Related Traits in Young Picea glauca Trees along a Regional

Climatic Gradient. Frontiers in Plant Science 7:48.

Bertrand, R., J. Lenoir, C. Piedallu, G. Riofrío-Dillon, P. de Ruffray, C. Vidal, J.-C.

Pierrat, and J.-C. Gégout. 2011. Changes in plant community composition lag

behind climate warming in lowland forests. Nature 479:517–520.

Blanchet, G., P. Legendre, and D. Borcard. 2008. Forward selection of spatial

explanatory variables. Ecology 89:2623–2632.

Blanquart, F., O. Kaltz, S. L. Nuismer, and S. Gandon. 2013. A practical guide to

measuring local adaptation. Ecology Letters 16:1195–1205.

Bloomfield, J. A., P. Nevill, B. M. Potts, R. E. Vaillancourt, and D. A. Steane. 2011.

Molecular genetic variation in a widespread forest tree species Eucalyptus obliqua

(Myrtaceae) on the island of Tasmania. Australian Journal of Botany 59:226–237.

Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. H.

Stevens, and J. S. S. White. 2009. Generalized linear mixed models: a practical

guide for ecology and evolution. Trends in Ecology and Evolution 24:127–135.

Borcard, D., and P. Legendre. 2002. All-scale spatial analysis of ecological data by

means of principal coordinates of neighbour matrices. Ecological Modelling

153:51–68.

Borcard, D., P. Legendre, C. Avois-Jacquet, and H. Tuomisto. 2004. Dissecting the

spatial structure of ecological data at multiple scales. Ecology 85:1826–1832.

Borcard, D., P. Legendre, and P. Drapeau. 1992. Partialling out the spatial component

of ecological variation. Ecology 73:1045–1055.

Bradford, J. B., and D. M. Bell. 2016. A window of opportunity for climate-change

adaptation: easing tree mortality by reducing forest basal area. Frontiers in Ecology

and the Environment.

Bradshaw, A. D. 1965. Evolutionary significance of phenotypic plasticity in plants.

Advances in genetics 13:115–155.

Breed, M. F., N. J. C. Gellie, and A. J. Lowe. 2016. Height differences in two eucalypt

provenances with contrasting levels of aridity. Restoration Ecology 24:471–478.

Breed, M. F., K. M. Ottewell, M. G. Gardner, M. H. K. Marklund, M. G. Stead, J. B. C.

Harris, and A J. Lowe. 2012. Mating system and early viability resistance to

habitat fragmentation in a bird-pollinated eucalypt. Heredity 115:1–8.

Breshears, D. D., H. D. Adams, D. Eamus, N. G. McDowell, D. J. Law, R. E. Will, A.

171

P. Williams, and C. B. Zou. 2013. The critical amplifying role of increasing

atmospheric moisture demand on tree mortality and associated regional die-off.

Frontiers in Plant Science 4:Article 266.

Bresson, C. C., Y. Vitasse, A. Kremer, and S. Delzon. 2011. To what extent is

altitudinal variation of functional traits driven by genetic adaptation in European

oak and beech? Tree Physiology 31:1164–1174.

Brook, B. W., N. S. Sodhi, and C. J. A. Bradshaw. 2008. Synergies among extinction

drivers under global change. Trends in Ecology & Evolution 23:453–460.

Brouwers, N. C., and N. C. Coops. 2016. Decreasing Net Primary Production in forest

and shrub vegetation across southwest Australia. Ecological Indicators 66:10–19.

Brouwers, N. C., J. Mercer, T. Lyons, P. Poot, E. Veneklaas, and G. Hardy. 2013a.

Climate and landscape drivers of tree decline in a Mediterranean ecoregion.

Ecology and Evolution 3:67–79.

Brouwers, N., G. Matusick, K. Ruthrof, T. Lyons, and G. Hardy. 2013b. Landscape-

scale assessment of tree crown dieback following extreme drought and heat in a

Mediterranean eucalypt forest ecosystem. Landscape Ecology 28:69–80.

Brown, C. D., and M. Vellend. 2014. Non-climatic constraints on upper elevational

plant range expansion under climate change.

Brunner, G. 2010. HEC-RAS River Analysis System Hydraulic Reference Manual. US

Army Corps of Engineers, Hydrologic EngineeringCenter, Davis, CA.

Bruno, D., O. Belmar, D. Sánchez-Fernández, and J. Velasco. 2014. Environmental

determinants of woody and herbaceous riparian vegetation patterns in a semi-arid

mediterranean basin. Hydrobiologia 730:45–57.

Bureau of Meteorology. 2017. Exceptional heat in southeast Australia in early 2017.

Special Climate Statement 61:1–38.

Burvill, G. H. 1997. The last fifty years, 1929-1979. Page in G. H. Burvill, editor.

Agriculture in Western Australia 150 years of development and achievement 1892-

1979. University of Western Australia Press, Nedlands.

Busch, D. E., and S. D. Smith. 1995. Mechanisms associated with decline of woody

species in riparian ecosystems of the southwestern U.S. Ecological Monographs

65:347–370.

Bustos-segura, C., S. Dillon, A. Keszei, W. J. Foley, and C. Külheim. 2017.

Intraspecific diversity of terpenes of Eucalyptus camaldulensis (Myrtaceae) at a

continental scale. Australian Journal of Botany 65:257–269.

Byrne, M., C. P. Elliott, C. J. Yates, and D. J. Coates. 2008. Maintenance of high pollen

dispersal in Eucalyptus wandoo, a dominant tree of the fragmented agricultural

region in Western Australia. Conservation Genetics 9:97–105.

Capon, S. J., L. E. Chambers, R. Mac Nally, J. Robert, P. M. Davies, N. Marshall, J.

Pittock, M. Reid, T. Capon, M. Douglas, J. Catford, D. S. Baldwin, M.

Stewardson, J. Roberts, M. Parsons, and S. E. Williams. 2013. Riparian

ecosystems in the 21st century: hotspots for climate change adaptation?

Ecosystems:DOI: 10.1007/s10021-013-9656-1.

Capon, S. J., C. James, and A. George. 2016. Riverine trees and shrubs. Pages 119–

141in S. J. Capon, C. James, and M. Reid, editors.Vegetation of Australian

Riverine Landscapes: Biology, Ecology and Management. CSIRO Publishing,

Clayton South, Australia.

Chen, I., J. K. Hill, R. Ohlemüller, D. B. Roy, and C. D. Thomas. 2011. Rapid range

shifts of species of climate warming. Science 333:1024–1026.

Chevin, L. M., S. Collins, and F. Lefèvre. 2013. Phenotypic plasticity and evolutionary

172

demographic responses to climate change: Taking theory out to the field.

Functional Ecology 27:967–979.

Chevin, L. M., and R. Lande. 2010. When do adaptive plasticity and genetic evolution

prevent extinction of a density-regulated population? Evolution 64:1143–1150.

Chevin, L. M., R. Lande, and G. M. Mace. 2010. Adaptation, plasticity, and extinction

in a changing environment: Towards a predictive theory. PLoS Biology 8.

Choat, B., S. Jansen, T. J. Brodribb, H. Cochard, S. Delzon, R. Bhaskar, S. J. Bucci, T.

S. Feild, S. M. Gleason, U. G. Hacke, A. L. Jacobsen, F. Lens, H. Maherali, J.

Martínez-Vilalta, S. Mayr, M. Mencuccini, P. J. Mitchell, A. Nardini, J.

Pittermann, R. B. Pratt, J. S. Sperry, M. Westoby, I. J. Wright, and A. E. Zanne.

2012. Global convergence in the vulnerability of forests to drought. Nature

491:752–755.

Christmas, M. J., M. F. Breed, and A. J. Lowe. 2016. Constraints to and conservation

implications for climate change adaptation in plants. Conservation Genetics

17:305–320.

Clay, R., and J. Majer. 2001. Flooded gum (Eucalyptus rudis) decline in the Perth

metropolitan area: A preliminary assessment.

Coley, P. D., J. P. Bryant, and F. S. Chapin. 1985. Resource availability and plant

antiherbivore defense. Science 230:895–899.

van Coller, A. L., K. H. Rogers, and G. L. Heritage. 2000. Riparian vegetation-

environment relationships: complimentarity of gradients versus patch hierarchy

approaches. Journal of Vegetation Science 11:337–350.

Cook, B. I., E. M. Wolkovich, and C. Parmesan. 2012. Divergent responses to spring

and winter warming drive community level flowering trends. Proceedings of the

National Academy of Sciences of the United States of America 109:9000–9005.

Corenblit, D., E. Tabacchi, J. Steiger, and A. M. Gurnell. 2007. Reciprocal interactions

and adjustments between fluvial landforms and vegetation dynamics in river

corridors: A review of complementary approaches. Earth-Science Reviews 84:56–

86.

Cornwell, W. K., R. Bhaskar, L. Sack, S. Cordell, and C. K. Lunch. 2007. Adjustment

of structure and function of Hawaiian Metrosideros polymorpha at high vs. low

precipitation. Functional Ecology 21:1063–1071.

Costanza, R., R. D’Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S.

Naeem, R. V. O’Neill, J. Paruelo, R. G. Raskin, P. Sutton, and M. van den Belt.

1997. The value of the world’s ecosystem services and natural capital. Nature

387:253–260.

Costanza, R., R. de Groot, P. Sutton, S. van der Ploeg, S. J. Anderson, I. Kubiszewski,

S. Farber, and R. K. Turner. 2014. Changes in the global value of ecosystem

services. Global Environmental Change 26:152–158.

Cowling, R. M., P. W. Rundel, B. B. Lamont, M. K. Arroyo, and M. Arianoutsou. 1996.

Plant diversity in mediterranean-climate regions. Trends in Ecology and Evolution

11:362–366.

Crimmins, S. M., S. Z. Dobrowski, J. A. Greenberg, J. T. Abatzoglou, and A. R.

Mynsberge. 2011. Changes in climatic water balance drive downhill shifts in plant

species’ optimum elevations. Science 331:324–325.

Crispo, E. 2008. Modifying effects of phenotypic plasticity on interactions among

natural selection, adaptation and gene flow. Journal of Evolutionary Biology

21:1460–1469.

CSIRO, and Bureau of Meteorology. 2015. Climate Change in Australia Information for

173

Australia’s Natural Resource Management Regions: Technical Report. Australia.

Cushman, S. A., and K. McGarigal. 2002. Hierarchical, multi-scale decomposition of

species-environment relationships. Landscape Ecology 17:637–646.

Daly, C., D. R. Conklin, and M. H. Unsworth. 2010. Local atmospheric decoupling in

complex topography alters climate change impacts. International Journal of

Climatology 30:1857–1864.

Danby, R. K., and D. S. Hik. 2007. Variability, contingency and rapid change in recent

subarctic alpine tree line dynamics. Journal of Ecology 95:352–363.

Davies, P. M. 2010. Climate change implications for river restoration in global

biodiversity hotspots. Restoration Ecology 18:261–268.

Davis, M. B., and R. G. Shaw. 2001. Range shifts and adaptive responses to quaternary

climate change. Science 292:673–679.

Davis, M. B., R. G. Shaw, and J. R. Etterson. 2005. Evolutionary response to changing

climate. Ecology 86:1704–1714.

Davison, E. M. 1997. Are jarrah (Eucalyptus marginata) trees killed by Phytophthora

cinnamomi or waterlogging? Australian Forestry 60:116–124.

Dawson, T. P., S. T. Jackson, J. I. House, I. C. Prentice, and G. M. Mace. 2011. Beyond

predictions: biodiversity conservation in a changing climate. Science 332:53–58.

Decocq, G. 2002. Patterns of plant species and community diversity at different

organization levels in a forested riparian landscape. Journal of Vegetation Science

13:91–106.

Department of Water. 2012. Warren–Donnelly surface water allocation plan methods

report. Perth.

Dessai, S., and M. Hulme. 2004. Does climate adaptation policy need probabilities ?

Climate Policy 4:1–22.

Dessai, S., and M. Hulme. 2007. Assessing the robustness of adaptation decisions to

climate change uncertainties : A case study on water resources management in the

East of England. Global Environmental Change 17:59–72.

Dewitt, T. J. 2016. Expanding the phenotypic plasticity paradigm to broader views of

trait space and ecological function. Current Zoology 62:463–473.

Dillon, S., R. Mcevoy, D. S. Baldwin, S. Southerton, C. Campbell, Y. Parsons, and G.

N. Rees. 2015. Genetic diversity of Eucalyptus camaldulensis Dehnh. following

population decline in response to drought and altered hydrological regime. Austral

Ecology 40:558–572.

Dixon, M. D., M. G. Turner, and C. Jin. 2002. Riparian tree seedling distribution on

Wisconsin River sandbars: controls at different spatial scales. Ecological

Monographs 72:465–485.

Dobrowski, S. Z. 2011. A climatic basis for microrefugia: The influence of terrain on

climate. Global Change Biology 17:1022–1035.

Dormann, C. F., J. Elith, S. Bacher, C. Buchmann, G. Carl, G. Carré, J. R. G. Marquéz,

B. Gruber, B. Lafourcade, P. J. Leitão, T. Münkemüller, C. Mcclean, P. E.

Osborne, B. Reineking, B. Schröder, A. K. Skidmore, D. Zurell, and S.

Lautenbach. 2013. Collinearity: A review of methods to deal with it and a

simulation study evaluating their performance. Ecography 36:027–046.

Dray, A. S., G. Blanchet, D. Borcard, G. Guenard, T. Jombart, G. Larocque, P.

Legendre, N. Madi, and H. H. Wagner. 2016. adespatial: Multivariate Multiscale

Spatial Analysis:R package, Version: 0.0-7.

Dray, S., P. Legendre, and P. R. Peres-Neto. 2006. Spatial modelling: a comprehensive

174

framework for principal coordinate analysis of neighbour matrices (PCNM).

Ecological Modelling 196:483–493.

Dutkowski, G. W., and B. M. Potts. 2012. Genetic variation in the susceptibility of

Eucalyptus globulus to drought damage. Tree Genetics and Genomes 8:757–773.

Edmands, S. 2007. Between a rock and a hard place: evaluating the relative risks of

inbreeding and outbreeding for conservation and management. Molecular Ecology

16:463–475.

Edwards, K. 2010. Phytophthora species associated with declining Eucalyptus rudis in

the south-west of Western Australia By. Murdoch University.

Elliott, G. P., and K. F. Kipfmueller. 2011. Multiscale influences of climate on upper

treeline dynamics in the southern Rocky Mountains, USA: evidence of

intraregional variability and bioclimatic thresholds in response to twentieth-century

warming. Annals of the Association of American Geographers 101:1181–1203.

Evans, B., C. Stone, and P. Barber. 2013. Linking a decade of forest decline in the

south-west of Western Australia to bioclimatic change. Australian Forestry

76:164–172.

Feeley, K. J., M. R. Silman, M. B. Bush, W. Farfan, K. G. Cabrera, Y. Malhi, P. Meir,

N. S. Reviilla, N. M. R. Quisiyupanqui, and S. Saatchi. 2011. Upslope migration of

Andean trees. Journal of Biogeography 38:783–791.

Fei, S., J. M. Desprez, K. M. Potter, I. Jo, J. A. Knott, and C. M. Oswalt. 2017.

Divergence of species responses to climate change. Science Advances 3:e1603055.

Franks, S. J., J. J. Weber, and S. N. Aitken. 2014. Evolutionary and plastic responses to

climate change in terrestrial plant populations. Evolutionary Applications 7:123–

139.

Freeman, J. S., B. M. Potts, G. M. Downes, D. Pilbeam, S. Thavamanikumar, and R. E.

Vaillancourt. 2013. Stability of quantitative trait loci for growth and wood

properties across multiple pedigrees and environments in Eucalyptus globulus.

New Phytologist 198:1121–1134.

De Frenne, P., F. Rodríguez-Sánchez, D. A. Coomes, L. Baeten, G. Verstraeten, M.

Vellend, M. Bernhardt-Römermann, C. D. Brown, J. Brunet, J. Cornelis, G. M.

Decocq, H. Dierschke, O. Eriksson, F. S. Gilliam, R. Hédl, T. Heinken, M. Hermy,

P. Hommel, M. A. Jenkins, D. L. Kelly, K. J. Kirby, F. J. G. Mitchell, T. Naaf, M.

Newman, G. Peterken, P. Petrík, J. Schultz, G. Sonnier, H. Van Calster, D. M.

Waller, G.-R. Walther, P. S. White, K. D. Woods, M. Wulf, B. J. Graae, and K.

Verheyen. 2013. Microclimate moderates plant responses to macroclimate

warming. Pnas 110:18561–5.

Frey, S. J. K., A. S. Hadley, S. L. Johnson, M. Schulze, J. A. Jones, and M. G. Betts.

2016. Spatial models reveal the microclimatic buffering capacity of old-growth

forests. Science Advances 2:e1501392–e1501392.

Froend, R., and B. Sommer. 2010. Phreatophytic vegetation response to climatic and

abstraction-induced groundwater drawdown: Examples of long-term spatial and

temporal variability in community response. Ecological Engineering 36:1191–

1200.

Galiano, L., J. Martínez-Vilalta, and F. Lloret. 2010. Drought-induced multifactor

decline of Scots Pine in the Pyrenees and potential vegetation change by the

expansion co-occurring Oak species. Ecosystems 13:978–991.

Gauli, A., D. A. Steane, R. E. Vaillancourt, and B. M. Potts. 2014. Molecular genetic

diversity and population structure in Eucalyptus pauciflora subsp. pauciflora

(Myrtaceae) on the island of Tasmania. Australian Journal of Botany 62:175–188.

175

Gauli, A., R. E. Vaillancourt, T. G. Bailey, D. A. Steane, and B. M. Potts. 2015.

Evidence for local climate adaptation in early-life traits of Tasmanian populations

of Eucalyptus pauciflora. Tree Genetics and Genomes 11.

Gelman, A. 2008. Scaling regression inputs by dividing by two standard deviations.

Statistics in medicine 27:2865–2873.

Ghalambor, C. K., J. K. McKay, S. P. Carroll, and D. N. Reznick. 2007. Adaptive

versus non-adaptive phenotypic plasticity and the potential for contemporary

adaptation in new environments. Functional Ecology 21:394–407.

Gibson, A., E. P. Bachelard, and K. T. Hubick. 1995. Relationship between climate and

provenance variation in Eucalyptus camaldulensis Dehnh. Australian Journal of

Plant Physiology 22:453–60.

Gioia, P., and S. D. Hopper. 2017. A new phytogeographic map for the Southwest

Australian Floristic Region after an exceptional decade of collection and discovery.

Botanical Journal of the Linnean Society 184:1–15.

Giorgi, F. 2006. Climate change hot-spots. Geophysical Research Letters 33:L08707.

Gordon, E., and R. K. Meentemeyer. 2006. Effects of dam operation and land use on

stream channel morphology and riparian vegetation. Geomorphology 82:412–429.

Gosney, B. J., B. M. Potts, J. M. O. Reilly-wapstra, E. Vaillancourt, H. Fitzgerald, N.

W. Davies, and J. S. Freeman. 2016. Genetic control of cuticular wax compounds

in Eucalyptus globulus. New Phytologist 209:202–215.

Groom, P. K., R. H. Froend, E. M. Mattiske, and R. P. Gurner. 2001. Long-term

changes in vigour and distribution of Banksia and Melaleuca overstorey species on

the Swan Coastal Plain:63–69.

Groom, Q. J. 2013. Some poleward movement of British native vascular plants is

occurring, but the fingerprint of climate change is not evident. PeerJ 1:e77:DOI

10.7717/peej.77.

Grubb, P. J. 1977. The maintenance of species-richness in plant communities: the

importance of the regeneration niche. Biological Reviews of the Cambridge

Philosophical Society 52:107–145.

Gurnell, A. M., D. Corenblit, D. Garcia de Jalon, M. Gonzalez del Tanago, R. C.

Grabowski, M. T. O’Hare, and M. Szewczyk. 2015. A conceptual model of

vegetation-hydrogeomorphology interactions within river corridors. River

Research and Applications:doi: 10.1002/rra.

Gynther, I., N. Waller, and L. K.-P. Leung. 2016. Confirmation of the extinction of the

Bramble Cay melomys Melomys rubicola on Bramble Cay, Torres Strait: results

and conclusions from a comprehensive survey in August–September 2014.

Queensland Government, Brisbane.

Halbritter, A. H., R. Billeter, P. J. Edwards, and J. M. Alexander. 2015. Local

adaptation at range edges: Comparing elevation and latitudinal gradients. Journal

of Evolutionary Biology 28:1849–1860.

Hallett, L. M., R. J. Standish, and Hobbs. 2011. Seed mass and summer drought

survival in a Mediterranean-climate ecosystem. Plant Ecology 212:1479–1489.

Hamer, J. J., E. J. Veneklaas, P. Poot, K. Mokany, and M. Renton. 2015a. Shallow

environmental gradients put inland species at risk: Insights and implications from

predicting future distributions of Eucalyptus species in South Western Australia.

Austral Ecology 40:923–932.

Hamer, J. J., E. J. Veneklaas, M. Renton, and P. Poot. 2015b. Links between soil texture

and root architecture of Eucalyptus species may limit distribution ranges under

future climates. Plant and Soil 403:217–229.

176

Hancock, C. N., P. G. Ladd, and R. H. Froend. 1996. Biodiversity and management of

riparian vegetation in Western Australia. Forest Ecology and Management 85:239–

250.

Hannah, L., G. F. Midgley, T. Lovejoy, W. J. Bond, M. Bush, J. C. Lovett, D. Scott, and

F. I. Woodward. 2002. Conservation of Biodiversity in a Changing Climate.

Conservation Biology 16:264–268.

Harris, J. A., R. J. Hobbs, E. Higgs, and J. Aronson. 2006. Ecological restoration and

global climate change. Restoration Ecology 14:170–176.

Harrison, P. A., G. Bailey, R. E. Vaillancourt, and B. M. Potts. 2014. Provenance and

seed mass determines the seed germination sucess of Eucalyptus ovata

(Myrtaceae). Seed Science and Technology 42:466–472.

Hedges, L. V, J. Gurevitch, and P. S. Curtis. 1999. The meta-analysis of response ratios

in experimental ecology. Ecology 80:1150–1156.

Hennessy, K., B. Fitzharris, B. C. Bates, N. Harvey, M. Howden, L. Hughes, J.

Salinger, and R. Warrick. 2007. Australia and New Zealand. Pages 507–540in M.

L. Parry, O. F. Canziani, J. P. Palutikof, P. J. van der Linden, and C. E. Hanson,

editors.Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution

of Working Group II to the Fourth Assessment Report of the Intergovernmental

Panel on Climate Change. Cambridge University Press, Cambridge, UK.

Hereford, J. 2009. A quantitative survey of local adaptation and fitness trade-offs. The

American naturalist 173:579–88.

Hewitt, N., N. Klenk, A. L. Smith, D. R. Bazely, N. Yan, S. Wood, J. I. MacLellan, C.

Lipsig-Mumme, and I. Henriques. 2011. Taking stock of the assisted migration

debate. Biological Conservation 144:2560–2572.

Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones, and A. Jarvis. 2005. Very high

resolution interpolated climate surfaces for global land areas. International Journal

of Climatology 25:1965–1978.

Hoffmann, A., and C. Sgrò. 2011. Climate change and evolutionary adaptation. Nature

470:479–485.

Hope, P., D. Abbs, J. Bhend, F. Chiew, J. A. Church, M. Ekström, D. Kirono, A.

Lenton, C. Lucas, K. McInnes, A. Moise, D. Monselesan, F. Mpelasoka, B.

Timbal, L. Webb, and P. Whetton. 2015. Southern and south-western flatlands

cluster report. Page in M. Ekström, P. Whetton, C. Gerbing, M. Grose, L. Webb,

and J. Risbey, editors. Climate Change in Australia Projections for Australia’s

Natural Resource Management Regions. CSIRO and Bureau of Meteorology,

Australia.

Hope, P. K., W. Drosdowsky, and N. Nicholls. 2006. Shifts in the synoptic systems

influencing southwest Western Australia. Climate Dynamics 26:751–764.

Hopper, S. D., and P. Gioia. 2004. The southwest Australian floristic region: Evolution

and conservation of a global hot spot of biodiversity. Annual Review of Ecology,

Evolution, and Systematics 35:623–650.

Horner, G. J., P. J. Baker, R. Mac Nally, S. C. Cunningham, J. R. Thomson, and F.

Hamilton. 2009. Mortality of developing floodplain forests subjected to a drying

climate and water extraction. Global Change Biology 15:2176–2186.

Hubble, T. C. T., B. B. Docker, and I. D. Rutherfurd. 2010. The role of riparian trees in

maintaining riverbank stability: A review of Australian experience and practice.

Ecological Engineering 36:292–304.

Hughes, T. P., J. T. Kerry, M. Álvarez-Noriega, J. G. Álvarez-Romero, K. D. Anderson,

A. H. Baird, R. C. Babcock, M. Beger, D. R. Bellwood, R. Berkelmans, T. C.

177

Bridge, I. R. Butler, M. Byrne, N. E. Cantin, S. Comeau, S. R. Connolly, G. S.

Cumming, S. J. Dalton, G. Diaz-Pulido, C. M. Eakin, W. F. Figueira, J. P.

Gilmour, H. B. Harrison, S. F. Heron, A. S. Hoey, J. A. Hobbs, M. O.

Hoogenboom, E. V. Kennedy, C. Kuo, J. M. Lough, R. J. Lowe, G. Liu, M. T.

McCulloch, H. A. Malcolm, M. J. McWilliam, J. M. Pandolfi, R. J. Pears, M. S.

Pratchett, V. Schoepf, T. Simpson, W. J. Skirving, B. Sommer, G. Torda, D. R.

Wachenfeld, B. L. Willis, and S. K. Wilson. 2017. Global warming and recurrent

mass bleaching of corals. Nature 543:373–377.

IPCC. 2014a. Climate Change 2014 Synthesis Report. Contribution of Working Groups

I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on

Climate Change:1–112.

IPCC. 2014b. Summary for Policymakers. Pages 1–32in C. B. Field, V. R. Barros, D. J.

Dokken, K. J. Mach, M. D. Mastrandrea, T. E. Bilir, M. Chatterjee, K. L. Ebi, Y.

O. Estrada, R. C. Genova, B. Girma, E. S. Kissel, A. N. Levy, S. MacCracken, P.

R. Mastrandrea, and L. L. White, editors.Climate Change 2014: Impacts,

Adaptation and Vulnerability - Contributions of the Working Group II to the Fifth

Assessment Report. Cambridge University Press, Cambridge, UK.

Isenburg, M. 2017. LAStools: award winning software for rapid LiDAR processing.

http://www.cs.unc.edu/~isenburg/lastools/.

Jackson, M. B., and T. D. Colmer. 2005. Response and adaptation by plants to flooding

stress. Annals of Botany 96:501–505.

Jensen, A. E., K. F. Walker, and D. C. Paton. 2008. The role of seedbanks in restoration

of floodplain woodlands. River Research and Applications 24:632–649.

Johnson, W. C. 2000. Tree recruitment and survival in rivers: influence of hydrological

processes. Hydrological Processes 14:3051–3074.

Jordan, G. J., B. M. Potts, P. Chalmers, and R. J. E. Wiltshire. 2000. Quantitative

genetic evidence that the timing of vegetative phase change in Eucalyptus globulus

ssp. globulus is an adaptive trait. Australian Journal of Botany 48:561–567.

Jump, A. S., and J. Peñuelas. 2005. Running to stand still: Adaptation and the response

of plants to rapid climate change. Ecology Letters 8:1010–1020.

Karrenberg, S., J. Kollmann, P. J. Edwards, A. M. Gurnell, and G. E. Petts. 2003.

Patterns in woody vegetation along the active zone of a near-natural alpine river.

Basic and Applied Ecology 4:157–166.

Kawecki, T. J., and D. Ebert. 2004. Conceptual issues in local adaptation. Ecology

Letters 7:1225–1241.

Keppel, G., K. Mokany, G. W. Wardell-Johnson, B. L. Phillips, J. A. Welbergen, and A.

E. Reside. 2015. The capacity of refugia for conservation planning under climate

change. Frontiers in Ecology and the Environment 13:106–112.

Keppel, G., K. P. Van Niel, G. W. Wardell-Johnson, C. J. Yates, M. Byrne, L. Mucina,

A. G. T. Schut, S. D. Hopper, and S. E. Franklin. 2012. Refugia: Identifying and

understanding safe havens for biodiversity under climate change. Global Ecology

and Biogeography 21:393–404.

King, A. 2017. Autralia’s dry June is a sign of what’s to come. The Conversation.

Klausmeyer, K. R., and M. R. Shaw. 2009. Climate change, habitat loss, protected areas

and the climate adaptation potential of species in Mediterranean ecosystems

worldwide. PLoS ONE 4:e6392.

Kramer, K. 1995. Phenotypic plasticity of the phenology of seven European tree species

in relation to climatic warming. Plant Cell and Environment 18:93–104.

Kullman, L., and L. Öberg. 2009. Post-Little Ice Age tree line rise and climate warming

178

in the Swedish Scandes: A landscape ecological perspective. Journal of Ecology

97:415–429.

Lamont, B. B. 1982. Mechanisms for enhancing nutrient uptake in plants, with

particular reference to Mediterranean, South Africa and Western Australia. The

Botanical Review 48:597–689.

Lande, R. 2009. Adaptation to an extraordinary environment by evolution of phenotypic

plasticity and genetic assimilation. Journal of Evolutionary Biology 22:1435–1446.

Lavorel, S., M. J. Colloff, S. McIntyre, M. D. Doherty, H. T. Murphy, D. J. Metcalfe,

M. Dunlop, R. J. Williams, R. M. Wise, and K. J. Williams. 2015. Ecological

mechanisms underpinning climate adaptation services. Global Change Biology

21:12–31.

Legendre, L., and P. Legendre. 2012. Numerical Ecology. 3rd ed. Elsevier Science,

Amsterdam.

Legendre, P., and E. D. Gallagher. 2001. Ecologically meaningful transformations for

ordination of species data. Oecologia 129:271–280.

Leigh, C., A. Bush, E. T. Harrison, S. S. Ho, L. Luke, R. J. Rolls, and M. E. Ledger.

2015. Ecological effects of extreme climatic events on riverine ecosystems:

Insights from Australia. Freshwater Biology 60:2620–2638.

Leimu, R., and M. Fischer. 2008. A meta-analysis of local adaptation in plants. PLoS

ONE 3:1–8.

Lenoir, J., J.-C. Gégout, J.-C. Pierrat, J.-D. Bontemps, and J.-F. Dhôte. 2009.

Differences between tree species seedling and adult altitudinal distribution in

mountain forests during the recent warm period (1986-2006). Ecography 32:765–

777.

Lenoir, J., J. Gégout, A. Guisan, P. Vittoz, T. Wohlgemuth, N. E. Zimmermann, S.

Dullinger, H. Pauli, W. Willner, and J. Svenning. 2010. Going against the flow:

potential mechanisms for unexpected downslope range shifts in a warming climate.

Ecography 33:295–303.

Lenoir, J., B. J. Graae, P. A. Aarrestad, I. G. Alsos, W. S. Armbruster, G. Austrheim, C.

Bergendorff, H. J. B. Birks, K. A. Bråthen, J. Brunet, H. H. Bruun, C. J. Dahlberg,

G. Decocq, M. Diekmann, M. Dynesius, R. Ejrnæs, J. A. Grytnes, K. Hylander, K.

Klanderud, M. Luoto, A. Milbau, M. Moora, B. Nygaard, A. Odland, V. T.

Ravolainen, S. Reinhardt, S. M. Sandvik, F. H. Schei, J. D. M. Speed, L. U.

Tveraabak, V. Vandvik, L. G. Velle, R. Virtanen, M. Zobel, and J. C. Svenning.

2013. Local temperatures inferred from plant communities suggest strong spatial

buffering of climate warming across Northern Europe. Global Change Biology

19:1470–1481.

Lenoir, J., T. Hattab, and G. Pierre. 2017. Climatic microrefugia under anthropogenic

climate change: implications for species redistribution. Ecography 40:253–266.

Lenoir, J., and J.-C. Svenning. 2015. Climate-related range shifts – a global

multidimensional synthesis and new research directions. Ecography 38:15–28.

Leutner, B. F., B. Reineking, J. Müller, M. Bachmann, C. Beierkuhnlein, S. Dech, and

M. Wegmann. 2012. Modelling forest α-diversity and floristic composition - on the

added value of LiDAR plus hyperspectral remote sensing. Remote Sensing

4:2818–2845.

Levis, N. A., and D. W. Pfennig. 2016. Evaluating “Plasticity-First” Evolution in

Nature: Key Criteria and Empirical Approaches. Trends in Ecology and Evolution

31:563–574.

Li, C., F. Berninger, J. Koskela, and E. Sonninen. 2000. Drought responses of

179

Eucalyptus microtheca provenances depend on seasonality of rainfall in their place

of origin. Australian Journal of Plant Physiology 27:231–238.

Lind, M. I., P. K. Ingvarsson, H. Johansson, D. Hall, and F. Johansson. 2010. Gene flow

and selection on phenotypic plasticity in an island system of rana temporaria.

Evolution 65:684–697.

Lite, S. J., K. J. Bagstad, and J. C. Stromberg. 2005. Riparian plant species richness

along lateral and longitudinal gradients of water stress and flood disturbance, San

Pedro River, Arizona, USA. Journal of Arid Environments 63:785–813.

Lite, S. J., and J. C. Stromberg. 2005. Surface water and ground-water thresholds for

maintaining Populus-Salix forests, San Pedro River, Arizona. Biological

Conservation 125:153–167.

Lloret, F., J. Peñuelas, P. Prieto, L. Llorens, and M. Estiarte. 2009. Plant community

changes induced by experimental climate change: Seedling and adult species

composition. Perspectives in Plant Ecology, Evolution and Systematics 11:53–63.

Loarie, S. R., P. B. Duffy, H. Hamilton, G. P. Asner, C. B. Field, and D. D. Ackerly.

2009. The velocity of climate change. Nature 462:1052–1055.

Lopez, G. A., B. M. Potts, R. E. Vaillancourt, and L. A. Apiolaza. 2003. Maternal and

carryover effects on early growth of Eucalyptus globulus. Canadian Journal of

Forest Research 33:2108–2115.

López, M., J. M. Humara, A. Casares, and J. Majada. 2000. The effect of temperature

and water stress on laboratory germination of Eucalyptus globulus Labill . seeds of

different sizes 57:245–250.

Lovell, J. L., D. L. B. Jupp, D. S. Culvenor, and N. C. Coops. 2003. Using airborne and

ground-based ranging lidar to measure canopy structure in Australian forests.

Canadian Journal of Remote Sensing 29:607–622.

Luo, W. B., F. B. Song, and Y. H. Xie. 2008. Trade-off between tolerance to drought

and tolerance to flooding in three wetland plants. Wetlands 28:866–873.

Lyon, J., and N. M. Gross. 2005. Patterns of plant diversity and plant-environmental

relationships across three riparian corridors. Forest Ecology and Management

204:267–278.

Lyon, J., and C. L. Sagers. 1998. Structure of herbaceous plant assemblages in a

forested riparian landscape. Plant Ecology 138:1–16.

Lyons, M. N., G. J. Keighery, N. Gibson, and G. W. Wardell-Johnson. 2000. The

vascular flora of the Warren bioregion, south-west Western Australia: composition,

reservation status and endemism. CALMScience 3:181–250.

Madeira, M., M. B. Araújo, and J. S. Pereira. 1995. Effects of water and nutrient supply

on amount and on nutrient concentration of litterfall and forest floor litter in

Eucalyptus globulus plantations. Plant and Soil 168–169:287–295.

Mahoney, J. M., and S. B. Rood. 1998. Streamflow requirements for cottonwood

seedling recruitment - an integrative model. Wetlands 18:634–645.

Máliš, F., M. Kopecky, P. Petřík, J. Vladovič, J. Merganič, and T. Vida. 2016. Life

stage, not climate change, explains observed tree range shifts. Global Change

Biology 22:1904–1914.

Mariotti, A., N. Zeng, J.-H. Yoon, V. Artale, A. Navarra, P. Alpert, and L. Z. X. Li.

2008. Mediterranean water cycle changes: transition to drier 21st century

conditions in observations and CMIP3 simulations. Environmental Research

Letters 3:44001.

Mathiasen, P., and A. C. Premoli. 2016. Living on the edge: adaptive and plastic

responses of the tree Nothofagus pumilio to a long-term transplant experiment

180

predict rear-edge upward expansion. Oecologia 181:607–619.

Matusick, G., K. X. Ruthrof, N. C. Brouwers, B. Dell, and G. S. J. Hardy. 2013. Sudden

forest canopy collapse corresponding with extreme drought and heat in a

mediterranean-type eucalypt forest in southwestern Australia. European Journal of

Forest Research 132:497–510.

Matusick, G., K. X. Ruthrof, J. B. Fontaine, and G. E. S. J. Hardy. 2016. Eucalyptus

forest shows low structural resistance and resilience to climate change-type

drought. Journal of Vegetation Science 27:493–503.

Matusick, G., K. X. Ruthrof, and G. S. J. Hardy. 2012. Drought and heat triggers

sudden and severe dieback in a dominant Mediterranean-type woodland species.

Open Journal of Forestry 2:183–186.

McCullough, I. M., F. W. Davis, J. R. Dingman, L. E. Flint, A. L. Flint, J. M. Serra-

Diaz, A. D. Syphard, M. A. Moritz, L. Hannah, and J. Franklin. 2016. High and

dry: high elevations disproportionately exposed to regional climate change in

Mediterranean-climate landscapes. Landscape Ecology 31:1063–1075.

McDowell, N. G., D. J. Beerling, D. D. Breshears, R. A. Fisher, K. F. Raffa, and M.

Stitt. 2011. The interdependence of mechanisms underlying climate-driven

vegetation mortality. Trends in Ecology and Evolution 26:523–532.

McKiernan, A. B., M. J. Hovenden, T. J. Brodribb, B. M. Potts, N. W. Davies, and J. M.

O’Reilly-Wapstra. 2014. Effect of limited water availability on foliar plant

secondary metabolites of two Eucalyptus species. Environmental and Experimental

Botany 105:55–64.

McLachlan, J. S., J. J. Hellmann, and M. W. Schwartz. 2007. A framework for debate

of assisted migration in an era of climate change. Conservation Biology 21:297–

302.

McLaughlin, B., D. D. Ackerly, P. Z. Klos, J. Natali, T. E. Dawson, and S. E.

Thompson. 2017. Hydrologic refugia, plants and climate change. Global Change

Biology 23:2941–2961.

McLaughlin, B., and E. S. Zavaleta. 2012. Predicting species responses to climate

change: demography and climate microrefugia in California valley oak (Quercus

lobata). Global Change Biology 18:2301–2312.

Mclean, E. H., S. M. Prober, W. D. Stock, D. A. Steane, B. M. Potts, R. E. Vaillancourt,

and M. Byrne. 2014. Plasticity of functional traits varies clinally along a rainfall

gradient in Eucalyptus tricarpa. Plant, Cell and Environment 37:1440–1451.

MEA. 2005. Millennium ecosystem assessment. Ecosystems and Human Well-Being:

Biodiversity.

Merritt, D. M., and N. L. Poff. 2010. Shifting dominance of riparian Populus and

Tamarix along gradients of flow alteration in western North American rivers.

Ecological Applications 20:135–152.

Merritt, D. M., M. L. Scott, N. Leroy Poff, G. T. Auble, and D. A. Lytle. 2010. Theory,

methods and tools for determining environmental flows for riparian vegetation:

Riparian vegetation-flow response guilds. Freshwater Biology 55:206–225.

Miller, K. A., J. A. Webb, S. C. De Little, and M. J. Stewardson. 2013. Environmental

flows can reduce the encroachment of terrestrial vegetation into river channels: A

systematic literature review. Environmental Management 52:1202–1212.

Milner, J. M., Ø. Varpe, R. van der Wal, and B. B. Hansen. 2016. Experimental icing

affects growth, mortality, and flowering in a high Arctic dwarf shrub. Ecology and

Evolution 6:2139–2148.

Mitchell, N., M. R. Hipsey, S. Arnall, G. McGrath, H. Bin Tareque, G. Kuchling, R.

181

Vogwill, M. Sivapalan, W. P. Porter, and M. R. Kearney. 2013. Linking Eco-

Energetics and Eco-Hydrology to Select Sites for the Assisted Colonization of

Australia’s Rarest Reptile. Biology 2:1–25.

Mondoni, A., G. Rossi, S. Orsenigo, and R. J. Probert. 2012. Climate warming could

shift the timing of seed germination in alpine plants. Annals of Botany 110:155–

164.

Montwé, D., M. Isaac-Renton, A. Hamann, and H. Spiecker. 2016. Drought tolerance

and growth in populations of a wide-ranging tree species indicate climate change

risks for the boreal north. Global Change Biology 22:806–815.

Müller, J., S. Bae, J. Röder, A. Chao, and R. K. Didham. 2014. Airborne LiDAR reveals

context dependence in the effects of canopy architecture on arthropod diversity.

Forest Ecology and Management 312:129–137.

Murphy, H. T., J. VanDerWal, and J. Lovett-Doust. 2010. Signatures of range

expansion and erosion in eastern North American trees. Ecology Letters 13:1233–

1244.

Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca, and J. Kent.

2000. Biodiversity hotspots for conservation priorities. Nature 403:853–858.

Nakagawa, S., and H. Schielzeth. 2013. A general and simple method for obtaining R2

from generalized linear mixed-effects models. Methods in Ecology and Evolution

4:133–142.

Nicotra, A. B., O. K. Atkin, S. P. Bonser, A. M. Davidson, E. J. Finnegan, U.

Mathesius, P. Poot, M. D. Purugganan, C. L. Richards, F. Valladares, and M. van

Kleunen. 2010. Plant phenotypic plasticity in a changing climate. Trends in Plant

Science 15:684–692.

Nimmo, D. G., A. Haslem, J. Q. Radford, M. Hall, and A. F. Bennett. 2015. Riparian

tree cover enhances the resistance and stability of woodland bird communities

during an extreme climatic event. Journal of Applied Ecology 53:449–458.

O’Brien, E. K., and S. L. Krauss. 2010. Testing the home-site advantage in forest trees

on disturbed and undisturbed sites. Restoration Ecology 18:359–372.

O’Brien, E. K., R. A. Mazanec, and S. L. Krauss. 2007. Provenance variation of

ecologically important traits of forest trees: Implications for restoration. Journal of

Applied Ecology 44:583–593.

O’Reilly-Wapstra, J. M., A. M. Miller, M. G. Hamilton, D. Williams, N. Glancy-Dean,

and B. M. Potts. 2013. Chemical Variation in a Dominant Tree Species: Population

Divergence, Selection and Genetic Stability across Environments. PLoS ONE 8.

Ogston, G., S. J. Beatty, D. L. Morgan, B. J. Pusey, and A. J. Lymbery. 2016. Living on

burrowed time: Aestivating fishes in south-western Australia face extinction due to

climate change. Biological Conservation 195:235–244.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O’Hara, G. L.

Simpson, P. Solymos, M. H. H. Stevens, and H. Wagner. 2017. vegan: Community

Ecology Package:R package, version 2.4-2.

Osterkamp, W. R., and C. R. Hupp. 2010. Fluvial processes and vegetation - Glimpses

of the past, the present, and perhaps the future. Geomorphology 116:274–285.

Palmer, M. A., D. P. Lettenmaier, N. L. Poff, S. L. Postel, B. Richter, and R. Warner.

2009. Climate change and river ecosystems: Protection and adaptation options.

Environmental Management 44:1053–1068.

Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change.

Annual Review of Ecology, Evolution, and Systematics 37:637–669.

Parmesan, C., and G. Yohe. 2003. A globally coherent fingerprint of climate change

182

impacts across natural systems. Nature 421:37–42.

Pecl, G. T., M. B. Araújo, J. D. Bell, J. Blanchard, T. C. Bonebrake, I.-C. Chen, T. D.

Clark, R. K. Colwell, F. Danielsen, B. Evengård, L. Falconi, S. Ferrier, S. Frusher,

R. A. Garcia, R. B. Griffis, A. J. Hobday, C. Janion-Scheepers, M. A. Jarzyna, S.

Jennings, J. Lenoir, H. I. Linnetved, V. Y. Martin, P. C. McCormack, J.

McDonald, N. J. Mitchell, T. Mustonen, J. M. Pandolfi, N. Pettorelli, E. Popova, S.

A. Robinson, B. R. Scheffers, J. D. Shaw, C. J. B. Sorte, J. M. Strugnell, J. M.

Sunday, M.-N. Tuanmu, A. Vergés, C. Villanueva, T. Wernberg, E. Wapstra, and

S. E. Williams. 2017. Biodiversity redistribution under climate change: Impacts on

ecosystems and human well-being. Science 355:eaai9214.

Pekin, B. K., M. M. Boer, C. Macfarlane, and P. F. Grierson. 2009. Impacts of

increased fire frequency and aridity on eucalypt forest structure, biomass and

composition in southwest Australia. Forest Ecology and Management 258:2136–

2142.

Peres-Neto, P., P. Legendre, S. Dray, and D. Borcard. 2006. Variation partitioning of

species data matrices: estimation and comparison of fractions. Ecology 87:2614–

2625.

Pérez-Harguindeguy, N., S. Díaz, E. Garnier, S. Lavorel, H. Poorter, P. Jaureguiberry,

M. S. Bret-Harte, W. K. Cornwell, J. M. Craine, D. E. Gurvich, C. Urcelay, E. J.

Veneklaas, P. B. Reich, L. Poorter, I. J. Wright, P. Ray, L. Enrico, J. G. Pausas, A.

C. De Vos, N. Buchmann, G. Funes, F. Quétier, J. G. Hodgson, K. Thompson, H.

D. Morgan, H. ter Steege, M. G. A. van der Heijden, L. Sack, B. Blonder, P.

Poschlod, M. V Vaieretti, G. Conti, A. C. Staver, S. Aquino, and J. H. C.

Cornelissen. 2013. New handbook for standardised measurement of plant

functional traits worldwide. Australian Journal of Botany 61:167–234.

Peterson, B. J., R. M. Holmes, J. W. McClelland, C. J. Vörösmarty, A. I. Shiklomanov,

I. A. Shiklomanov, and S. Rahmstorf. 2002. Increasing river discharge to the

Arctic Ocean. Science 298:21712173.

Petit, R. J., A. Hampe, and R. Cheddadi. 2005. Climate changes and tree

phylogeography in the Mediterranean. Taxon 54:877–885.

Petrone, K. C., J. D. Hughes, T. G. Van Niel, and R. P. Silberstein. 2010. Streamflow

decline in southwestern Australia, 1950-2008. Geophysical Research Letters 37:1–

7.

Pettit, N. E., and R. H. Froend. 2001a. Availability of seed for recruitment of riparian

vegetation: a comparison of a tropical and a temperate river ecosystem in

Australia. Australian Journal of Botany 49:515–528.

Pettit, N. E., and R. H. Froend. 2001b. Variability in flood disturbance and the impact

on riparian tree recruitment in two contrasting river systems. Wetlands Ecology

and Mangement 9:13–25.

Pettit, N. E., R. H. Froend, and P. M. Davies. 2001. Identifying the natural flow regime

and the relationship with riparian vegetation for two contrasting Western

Australian rivers. Regulated Rivers: Research & Management 17:201–215.

Petty, A. M., and M. M. Douglas. 2010. Scale relationships and linkages between

woody vegetation communities along a large tropical floodplain river, north

Australia. Journal of Tropical Ecology 26:79–92.

Pigliucci, M., C. J. Murren, and C. D. Schlichting. 2006. Phenotypic plasticity and

evolution by genetic assimilation. The Journal of Experimental Biology 209:2362–

2367.

Poff, N. L., B. D. Richter, A. H. Arthington, S. E. Bunn, R. J. Naiman, E. Kendy, M.

183

Acreman, C. Apse, B. P. Bledsoe, M. C. Freeman, J. Henriksen, R. B. Jacobson, J.

G. Kennen, D. M. Merritt, J. H. O’Keeffe, J. D. Olden, K. Rogers, R. E. Tharme,

and A. Warner. 2010. The ecological limits of hydrologic alteration (ELOHA): A

new framework for developing regional environmental flow standards. Freshwater

Biology 55:147–170.

Polzin, M. L., and S. B. Rood. 2006. Effective distrubance: seedling safe site and patch

recruitment of riparian cottonwoods after a major flood of a mountain river.

Wetlands 26:965–980.

Prieto, I., F. I. Pugnaire, and R. J. Ryel. 2014. Water uptake and redistribution during

drought in a semiarid shrub species. Functional Plant Biology 41:812–819.

Prober, S. M., M. Byrne, E. H. Mclean, D. A. Steane, B. M. Potts, R. E. Vaillancourt,

and W. D. Stock. 2015. Climate-adjusted provenancing : a strategy for climate-

resilient ecological restoration 3:1–5.

Prober, S. M., K. R. Thiele, P. W. Rundel, C. J. Yates, S. L. Berry, M. Byrne, L.

Christidis, C. R. Gosper, P. F. Grierson, K. Lemson, T. Lyons, C. Macfarlane, M.

H. O. Connor, J. K. Scott, R. J. Standish, W. D. Stock, E. J. B. Van Etten, G. W.

Wardell-Johnson, and A. Watson. 2012. Facilitating adaptation of biodiversity to

climate change: a conceptual framework applied to the world’s largest

Mediterranean-climate woodland. Climatic Change 110:227–248.

Pryor, L. D. 1956. Variation in snow gum (Eucalyptus pauciflora Sieb.). Proceedings of

the Linnean Society NSW 81:299–305.

Quinn, G. P., and M. J. Keough. 2002. Experimental Design and Data Analysis for

Biologists. Cambridge University Press, New York.

R Core Team. 2016. R: A language and environment for statistical computing. R

Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-

project.org/.

Randin, C. F., R. Engler, S. Normand, M. Zappa, N. E. Zimmermann, P. B. Pearman, P.

Vittoz, W. Thuiller, and A. Guisan. 2009. Climate change and plant distribution:

Local models predict high-elevation persistence. Global Change Biology 15:1557–

1569.

Ransley, T. R., and B. D. Smerdon. 2012. Hydrostratigraphy, hydrogeology and system

conceptualisation of the Great Artesian Basin. A technical report to the Australian

Government from the CSIRO Great Artesian Basin Water Resource Assessment.

Canberra, Australia.

Rapacciuolo, G., S. P. Maher, A. C. Schneider, T. T. Hammond, M. D. Jabis, R. E.

Walsh, K. J. Iknayan, G. K. Walden, M. F. Oldfather, D. D. Ackerly, and S. R.

Beissinger. 2014. Beyond a warming fingerprint: Individualistic biogeographic

responses to heterogeneous climate change in California. Global Change Biology

20:2841–2855.

Reed, T. E., D. E. Schindler, and R. S. Waples. 2011. Interacting effects of phenotypic

plasticity and evolution on population persistance in a changing climate.

Conservation Biology 25:56–63.

Renton, M., N. Shackelford, and R. J. Standish. 2012. Habitat restoration will help some

functional plant types persist under climate change in fragmented landscapes.

Global Change Biology 18:2057–2070.

Reside, A. E., J. A. Welbergen, B. L. Phillips, G. W. Wardell-Johnson, G. Keppel, S.

Ferrier, S. E. Williams, and J. VanDerWal. 2014. Characteristics of climate change

refugia for Australian biodiversity. Austral Ecology 39:887–897.

Reyer, C. P. O., S. Leuzinger, A. Rammig, A. Wolf, R. P. Bartholomeus, A. Bonfante,

184

F. de Lorenzi, M. Dury, P. Gloning, R. Abou Jaoudé, T. Klein, T. M. Kuster, M.

Martins, G. Niedrist, M. Riccardi, G. Wohlfahrt, P. de Angelis, G. de Dato, L.

François, A. Menzel, and M. Pereira. 2013. A plant’s perspective of extremes:

Terrestrial plant responses to changing climatic variability. Global Change Biology

19:75–89.

Ribeiro, C., M. Madeira, and M. B. Araújo. 2002. Decomposition and nutrient release

from leaf litter of Eucalyptus globulus grown under different water and nutrient

regimes. Forest Ecology and Management 171:31–41.

Ricciardi, A., and D. Simberloff. 2009. Assisted colonization is not a viable

conservation strategy. Trends in Ecology and Evolution 24:248–253.

Richter, S., T. Kipfer, T. Wohlgemuth, C. Calderón Guerrero, J. Ghazoul, and B.

Moser. 2012. Phenotypic plasticity facilitates resistance to climate change in a

highly variable environment. Oecologia 169:269–279.

Rix, K. D., A. J. Gracie, B. M. Potts, P. H. Brown, C. J. Spurr, and P. L. Gore. 2012.

Paternal and maternal effects on the response of seed germination to high

temperatures in Eucalyptus globulus. Annals of Forest Science 69:673–679.

Rood, S. B., J. H. Braatne, and L. A. Goater. 2010. Responses of obligate versus

facultative riparian shrubs following river damming. River Research and

Applications 26:102–117.

Rout, T. M., E. Mcdonald-madden, T. G. Martin, N. J. Mitchell, H. P. Possingham, and

D. P. Armstrong. 2013. How to decide whether to move species threatened by

climate change. PLoS ONE 8:e75814.

Rull, V. 2009. Microrefugia. Journal of Biogeography 36:481–484.

Rundel, P. W., M. T. K. Arroyo, R. M. Cowling, J. E. Keeley, B. B. Lamont, and P.

Vargas. 2016. Mediterranean Biomes: Evolution of their Vegetation, Floras and

Climate. Annual Review of Ecology, Evolution, and Systematics 47:383–407.

Ruthrof, K. X., T. K. Douglas, M. C. Calver, P. A. Barber, B. Dell, and G. E. S. J.

Hardy. 2010. Restoration treatments improve seedling establishment in a degraded

Mediterranean-type Eucalyptus ecosystem. Australian Journal of Botany 58:646–

655.

Sagers, C. L., and J. Lyon. 1997. Gradient analysis in a riparian landscape: contrasts

among forest layers. Forest Ecology and Management 96:13–26.

Schielzeth, H. 2010. Simple means to improve the interpretability ofregression

coefficients. Methods in Ecology and Evolution 1:103–113.

Schlichting, C. D. 1986. The evolution of phenotypic plasticity in plants. Annual

Review of Ecology and Systematics 17:667–693.

Schreiber, S. G., C. Ding, A. Hamann, U. G. Hacke, B. R. Thomas, and J. S. Brouard.

2013. Frost hardiness vs. growth performance in trembling aspen: An experimental

test of assisted migration. Journal of Applied Ecology 50:939–949.

Schut, A. G. T., G. W. Wardell-Johnson, C. J. Yates, G. Keppel, I. Baran, S. E.

Franklin, S. D. Hopper, K. P. Van Niel, L. Mucina, and M. Byrne. 2014. Rapid

characterisation of vegetation structure to predict refugia and climate change

impacts across a global biodiversity hotspot. PLoS ONE 9:e82778.

Seabrook, L., C. Mcalpine, J. Rhodes, G. Baxter, A. Bradley, and D. Lunney. 2014.

Determining range edges: Habitat quality, climate or climate extremes? Diversity

and Distributions 20:95–106.

Seavy, N. E., T. Gardali, G. H. Golet, F. T. Griggs, C. A. Howell, R. Kelsey, S. L.

Small, J. H. Viers, and J. F. Weigand. 2009. Why climate change makes riparian

restoration more important than ever: recommendations for practice and research.

185

Ecological Restoration 27:330–338.

Shafroth, P. B., J. C. Stromberg, and D. T. Patten. 2002. Riparian vegetation response to

altered disturbance and stress regimes. Ecological Applications 12:107–123.

Silberstein, R. P., S. K. Aryal, J. Durrant, M. Pearcey, M. Braccia, S. P. Charles, L.

Boniecka, G. A. Hodgson, M. A. Bari, N. R. Viney, and D. J. McFarlane. 2012.

Climate change and runoff in south-western Australia. Journal of Hydrology

475:441–455.

Smith, M. G., R. N. M. Dixon, L. H. Boniecka, M. L. Berti, T. Sparks, M. A. Bari, and

J. Platt. 2006. Salinity Situation Statement: Warren River. Western Australian

Department of Water, Water Resource Technical Series No. WRT 32.

Solomon, S., G. Plattner, R. Knutti, and P. Friedlingstein. 2009. Irreversible climate

change due to carbon dioxide emissions. PNAS 106:1704–1709.

Stackpole, D. J., R. E. Vaillancourt, M. de Aguigar, and B. M. Potts. 2010. Age trends

in genetic parameters for growth and wood density in Eucalyptus globulus. Tree

Genetics and Genomes 6:179–193.

Stein, B. A., A. Staudt, M. S. Cross, N. S. Dubois, C. Enquist, R. Griffis, L. J. Hansen,

J. J. Hellmann, J. J. Lawler, E. J. Nelson, and A. Pairis. 2013. Preparing for and

managing change: climate adaptation for biodiversity and ecosystems. Frontiers in

Ecology and the Environment 11:502–510.

Stella, J. C., and J. J. Battles. 2010a. How do riparian woody seedlings survive seasonal

drought? Oecologia 164:579–590.

Stella, J. C., and J. J. Battles. 2010b. How do riparian woody seedlings survive seasonal

drought? Oecologia 164:579–590.

Stella, J. C., J. J. Battles, J. R. Mcbride, and B. K. Orr. 2010a. Riparian seedling

mortality from simulated water table recession, and the design of sustainable flow

regimes on regulated rivers. Restoration Ecology 18:284–294.

Stella, J. C., J. J. Battles, J. R. McBride, and B. K. Orr. 2010b. Riparian seedling

mortality from simulated water table recession, and the design of sustainable flow

regimes on regulated rivers. Restoration Ecology 18:284–294.

Stella, J. C., J. Riddle, H. Piégay, M. Gagnage, and M. L. Trémélo. 2013. Climate and

local geomorphic interactions drive patterns of riparian forest decline along a

Mediterranean Basin river. Geomorphology 202:101–114.

Stocker, T. F., Q. Dahe, G.-K. Plattner, L. V. Alexander, S. K. Allen, N. L. Bindoff, F.-

M. Bréon, J. A. Church, U. Cubash, S. Emori, P. Forster, P. Friedlingstein, L. D.

Talley, D. G. Vaughan, and S.-P. Xie. 2013. Technical Summary. Climate Change

2013: The Physical Science Basis. Contribution of Working Group I to the Fifth

Assessment Report of the Intergovernmental Panel on Climate Change:33–115.

Stromberg, J. C., K. J. Bagstad, J. M. Leenhouts, S. J. Lite, and E. Makings. 2005.

Effects of stream flow intermittency on riparian vegetation of a semiarid region

river (San Pedro River, Arizona). River Research and Applications 21:925–938.

Stromberg, J. C., S. J. Lite, and M. D. Dixon. 2010. Effects of stream flow patterns on

riparian vegetation of a semiarid river: implications for a changing climate. River

Research and Applications 26:712–729.

Stromberg, J. C., K. E. McCluney, M. D. Dixon, and T. Meixner. 2013. Dryland

riparian ecosystems in the American Southwest: sensitivity and resilience to

climatic extremes. Ecosystems 16:411–415.

Stromberg, J. C., P. B. Shafroth, and A. F. Hazelton. 2012. Legacies of flood reduction

on a dryland river. River Research and Applications 28:143–159.

Sultan, S. E., and H. G. Spencer. 2002. Metapopulation structure favors plasticity over

186

local adaptation. The American naturalist 160:271–283.

Tabacchi, E., D. L. Correll, R. Hauer, G. Pinay, A. M. Planty-Tabacchi, and R. C.

Wissmar. 1998. Development, maintenance and role of riparian vegetation in the

river landscape. Freshwater Biology 40:497–516.

Taylor, J. P., D. B. Wester, and L. M. Smith. 1999. Soil disturbance, flood management,

and riparian woody plant establishment in the Rio Grande floodplain. Wetlands

19:372–382.

Telwala, Y., B. W. Brook, K. Manish, and M. K. Pandit. 2013. Climate-induced

elevational range shifts and increase in plant species richness in a Himalayan

biodiversity epicentre. PLoS ONE 8:e57103.

Tesi, T., F. Muschitiello, R. H. Smittenberg, M. Jakobsson, J. E. Vonk, P. Hill, A.

Andersson, N. Kirchner, R. Noormets, O. Dudarev, and I. Semiletov. 2016.

Massive remobilization of permafrost carbon during post-glacial warming:1–10.

Thomas, C. D., A. Cameron, R. E. Green, M. Bakkenes, L. J. Beaumont, Y. C.

Collingham, B. F. N. Erasmus, M. F. De Siqueira, A. Grainger, L. Hannah, L.

Hughes, B. Huntley, A. S. Van Jaarsveld, G. F. Midgley, L. Miles, M. A. Ortega-

Huerta, A. T. Peterson, O. L. Phillips, and S. E. Williams. 2004. Extinction risk

from climate change. Nature 427:145–148.

Thuiller, W., S. Lavorel, M. B. Araujo, M. T. Sykes, and I. C. Prentice. 2005. Climate

change threats to plant diversity in Europe. Proceedings of the National Academy

of Sciences 102:8245–8250.

Tockner, K., and J. A. Stanford. 2002. Riverine flood plains: present state and future

trends. Environmental Conservation 29:308–330.

Turner, B. L., P. E. Hayes, and E. Laliberté. 2017. A climosequence of chronosequences

in southwestern Australia. bioRxiv preprint:doi: http://dx.doi.org/10.1101/113308.

Underwood, E. C., J. H. Viers, K. R. Klausmeyer, R. L. Cox, and M. R. Shaw. 2009.

Threats and biodiversity in the mediterranean biome. Diversity and Distributions

15:188–197.

Valladares, F., S. Matesanz, F. Guilhaumon, M. B. Araújo, L. Balaguer, M. Benito-

Garzón, W. K. Cornwell, E. Gianoli, M. van Kleunen, D. E. Naya, A. B. Nicotra,

H. Poorter, and M. A. Zavala. 2014. The effects of phenotypic plasticity and local

adaptation on forecasts of species range shifts under climate change. Ecology

Letters 17:1351–1364.

VanDerWal, J., H. T. Murphy, A. S. Kutt, G. C. Perkins, B. L. Bateman, J. J. Perry, and

A. E. Reside. 2013. Focus on poleward shifts in species’ distribution

underestimates the fingerprint of climate change. Nature Climate Change 3:239–

243.

VanDerWal, J., L. P. Shoo, C. N. Johnson, and S. E. Williams. 2009. Abundance and

the environmental niche: Environmental suitability estimated from niche models

predicts the upper limit of local abundance. American Naturalist 174:282–291.

Veraverbeke, S., B. M. Rogers, M. L. Goulden, R. R. Jandt, C. E. Miller, E. B. Wiggins,

and J. T. Randerson. 2017. Lightning as a major driver of recent large fire years in

North American boreal forests. Nature Clim. Change 7:529–534.

Vitasse, Y., C. C. Bresson, A. Kremer, R. Michalet, and S. Delzon. 2010. Quantifying

phenological plasticity to temperature in two temperate tree species. Functional

Ecology 24:1211–1218.

Vitasse, Y., G. Hoch, C. F. Randin, A. Lenz, C. Kollas, and C. Körner. 2012. Tree

recruitment of European tree species at their current upper elevational limits in the

Swiss Alps. Journal of Biogeography 39:1439–1449.

187

Vitasse, Y., G. Hoch, C. F. Randin, A. Lenz, C. Kollas, J. F. Scheepens, and C. Körner.

2013. Elevational adaptation and plasticity in seedling phenology of temperate

deciduous tree species. Oecologia 171:663–678.

Vörösmarty, C. J., P. Green, J. Salisbury, and R. B. Lammers. 2000. Global water

resources: Vulnerability from climate change and population growth. Science

289:284–288.

Wang, T., A. Hamann, A. Yanchuk, G. A. O’neill, and S. N. Aitken. 2006. Use of

response functions in selecting lodgepole pine populations for future climates.

Global Change Biology 12:2404–2416.

Wang, T., G. O’Neill, and S. N. Aitken. 2010. Integrating environmental and genetic

effects to predict responses of tree populations to climate. Ecological Applications

20:153–163.

Warfe, D. M., S. A. Hardie, A. R. Uytendaal, C. J. Bobbi, and L. A. Barmuta. 2014. The

ecology of rivers with contrasting flow regimes: Identifying indicators for setting

environmental flows. Freshwater Biology 59:2064–2080.

Warren, C. R., E. Dreyer, M. Tausz, and M. A. Adams. 2006. Ecotype adaptation and

acclimation of leaf traits to rainfall in 29 species of 16-year-old Eucalyptus at two

common gardens. Functional Ecology 20:929–940.

Webber, B. L., J. K. Scott, and R. K. Didham. 2011. Translocation or bust! A new

acclimatization agenda for the 21 st century? Trends in Ecology and Evolution

26:495–496.

West-Eberhard, M. J. 2005. Developmental plasticity and the origin of species

differences. PNAS 102:6543–6549.

Wetherald, R. T., and S. Manabe. 2002. Simulation of hydrologic changes associated

with global warming. Journal of Geophysical Research 107:4379.

Whitmore, T. C. 1989. Canopy gaps and the two major groups of forest trees. Ecology

70:536–538.

Whittock, S. P., L. A. Apiolaza, C. M. Kelly, and B. M. Potts. 2003. Genetic control of

coppice and lignotuber development in Eucalyptus globulus. Australian Journal of

Botany 51:57–67.

Williams, S. E., E. E. Bolitho, and S. Fox. 2003. Climate change in Australian tropical

rainforests: an impending environmental catastrophe. Proceedings Of The Royal

Society B: Biological Sciences 270:1887–1892.

Wise, R. M., I. Fazey, M. S. Smith, S. E. Park, H. C. Eakin, E. R. M. A. Van Garderen,

and B. Campbell. 2014. Reconceptualising adaptation to climate change as part of

pathways of change and response §. Global Environmental Change 28:325–336.

Yang, J., T. E. Dilts, L. A. Condon, P. L. Turner, and P. J. Weisberg. 2011.

Longitudinal- and transverse-scale environmental influences on riparian vegetation

across multiple levels of ecological organization. Landscape Ecology 26:381–395.

Yeoh, P. B., L. Fontanini, J. A. Carley, A. T. Russell, K. E. Dawson, J. K. Scott, C. Y.

M. Delaisse, and B. L. Webber. 2016. Landscape transformations following

blackberry decline in the south-west of Western Australia : successful restoration

or back to blackberry ? Proceedings of the Twentieth Australiasian Weeds

Conference.

Zhao, C., M. A. King, C. S. Watson, V. R. Barletta, A. Bordoni, M. Dell, and P. L.

Whitehouse. 2017. Rapid ice unloading in the Fleming Glacier region, southern

Antarctic Peninsula, and its effect on bedrock uplift rates. Earth and Planetary

Science Letters 473:164–176.

Zhu, K., C. W. Woodall, and J. S. Clark. 2012. Failure to migrate: Lack of tree range

188

expansion in response to climate change. Global Change Biology 18:1042–1052.

Zhu, K., C. W. Woodall, S. Ghosh, A. E. Gelfand, and J. S. Clark. 2014. Dual impacts

of climate change: Forest migration and turnover through life history. Global

Change Biology 20:251–264.

Resistance, resilience and adaptation to climate …...climate change will depend on their capacity...

Documents

Transcript of Resistance, resilience and adaptation to climate …...climate change will depend on their capacity...