Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
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
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
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
Student contribution to work:
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.
Co-author signatures and dates:
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
Chapter 1: General Introduction
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
Chapter 1: General Introduction
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
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.
Chapter 1: General Introduction
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.
Chapter 1: General Introduction
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
Chapter 1: General Introduction
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
Chapter 2: Riparian zones as refugia
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).
Chapter 2: Riparian zones as refugia
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).
Chapter 2: Riparian zones as refugia
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)
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
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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).
Chapter 2: Riparian zones as refugia
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.
Chapter 2: Riparian zones as refugia
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
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
Chapter 2: Riparian zones as refugia
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. ►
Chapter 2: Riparian zones as refugia
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).
Chapter 2: Riparian zones as refugia
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.
Chapter 2: Riparian zones as refugia
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).
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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.
Chapter 2: Riparian zones as refugia
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.
Chapter 2: Riparian zones as refugia
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.
Chapter 2: Riparian zones as refugia
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.
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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,
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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
Chapter 2: Riparian zones as refugia
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
Chapter 3: Range shifts in riparian plants
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
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
Chapter 3: Range shifts in riparian plants
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
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.
Chapter 3: Range shifts in riparian plants
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.
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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).
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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. ►
Chapter 3: Range shifts in riparian plants
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
.
Chapter 3: Range shifts in riparian plants
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
Fig. 3.6. Variation in the relative frequency of immature (1) to mature (0) individuals of
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.
Fig. 3.7. Variation in the relative frequency of immature (1) to mature (0) individuals of
(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.
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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
Chapter 3: Range shifts in riparian plants
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 -
Chapter 3: Range shifts in riparian plants
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
NA - model convergence failure
Model - Taxandria juniperina
NA - model convergence failure
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
Chapter 3: Range shifts in riparian plants
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).
104
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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).
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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. ►
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
145
Table 4.9. 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 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. ►
146
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
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.
Chapter 4: Plasticity in E. rudis
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
Chapter 4: Plasticity in E. rudis
159
Table S4.2. 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) 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
Chapter 5: Synthesis & Conclusions
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.
Chapter 5: Synthesis & Conclusions
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 finger- print 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.