Science of the Total Environment 685 (2019) 451–462

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Substantial declines in urban tree habitat predicted under climate change

Hugh Burley a,⁎, Linda J. Beaumont a, Alessandro Ossola a, John B. Baumgartner a, Rachael Gallagher a,
Shawn Laffan b, Manuel Esperon-Rodriguez c, Anthony Manea a, Michelle R. Leishman a
a
Department of Biological Sciences, Macquarie University, NSW 2109, Australia
b
School of BEES, The University of New South Wales, UNSW, Sydney, NSW 2052, Australia
c
Hawkesbury Institute of Environment, Western Sydney University, NSW 2751, Australia

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Climate change likely to cause declines

in suitable habitat for urban trees.
• New urban habitat likely gained by
some tree species due to climate
change.
• Tree species in warmer cities likely to
decline more than trees in cooler cities.
• Industry and government need climate-
ready tree species selection.

a r t i c l e i n f o a b s t r a c t

Article history: Globally, local governments are increasing investment in urban greening projects. However, there is little consid-
Received 28 March 2019 eration of whether the species being planted will be resilient to climate change. We assessed the distribution of
Received in revised form 17 May 2019 climatically suitable habitat, now and in the future, for 176 tree species native to Australia, commonly planted
Accepted 19 May 2019
across Australia's Significant Urban Areas (SUAs) and currently grown by commercial nurseries.
Available online xxxx
Species' occurrence records were obtained from inventories and herbaria, globally and across Australia, and com-
Editor: Elena Paoletti bined with baseline climate data (WorldClim, 1960–1990) and six climate scenarios for 2030 and 2070 using cli-
matic suitability models (CSMs). CSMs for each species were calibrated and projected onto baseline and future
Keywords: scenarios. We calculated changes in the size of climatically suitable habitat for each species across each SUA,
Urban trees and identified urban areas that are likely to have suitable climate for either fewer or more of our study species
Horticulture industry under future climate.
Species selection By 2070, climatically suitable habitat in SUAs is predicted to decline for 73% of species assessed. For 18% of these
Climate change species, climatically suitable area is predicted to be more than halved, relative to their baseline extent. Generally,
Vulnerability
for urban areas in cooler regions, climatically suitable habitat is predicted to increase. By contrast, for urban areas
Species distribution models
in warmer regions, a greater proportion of tree species may lose climatically suitable habitat. Our results highlight
changing patterns of urban climatic space for commonly planted species, suggesting that local governments and
the horticultural industry should take a proactive approach to identify new climate-ready species for urban
plantings.
© 2019 Published by Elsevier B.V.

⁎ Corresponding author.
E-mail address: hugh.burley@mq.edu.au (H. Burley).

https://doi.org/10.1016/j.scitotenv.2019.05.287
0048-9697/© 2019 Published by Elsevier B.V.
452 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
1. Introduction
Globally, government authorities are placing great emphasis on the
environmental and socio-economic importance of urban trees and veg-
etation (Luttik, 2000; Chiesura, 2004). In particular, planting trees is in-
creasingly considered a cost-effective method for improving the
adaptability and sustainability of cities to projected changes in climate
(Espeland and Kettenring, 2018). Trees can further mitigate urban envi-
ronmental issues that will likely be exacerbated by climate change, such
as the urban heat island effect [UHI, Norton et al., 2015; Lanza and Stone,
2016] and air pollution (Grote et al., 2016; Livesley et al., 2016). As a
consequence of these benefits, local authorities are increasing their in-
vestment in planting as well as monitoring tree and forest assets across
urban environments (Endreny, 2018).
Despite these financial commitments, there is little consideration of
whether the current selection of species commonly used in urban plant-
ings will be resilient to changes in climate projected to occur throughout
the lifespan of each tree species (Ordóñez and Duinker, 2015), which
may be decades to a century. Thus, as temperatures continue to rise in
the 21st century, urban forestry and planning efforts must identify spe-
cies that are most likely to thrive in future climates, to maximise plant-
ing success and return on investment in the long term.
The biological fingerprint of climate change is already apparent
across natural environments, with species' range shifts being observed
for large number of taxa, including vascular plants [e.g. Parmesan and
Yohe, 2003; Serra-Diaz et al., 2014; Guo et al., 2018]. However, most as-
sessments of broad-scale species responses to climate change have fo-
cused on species living in natural environments, or those that are
economically important (e.g. crops or invasive species). Far less atten-
tion has been paid to trees in urban environments, although the charac-
teristics of these environments (such as impervious surfaces and the
UHI effect) can exacerbate climatic extremes (Savi et al., 2015). Changes
to the growth of urban trees, likely due to shifting climate, have already
been reported [e.g. see Lanza and Stone, 2016; Nitschke et al., 2017;
Pretzsch et al., 2017]. This is compounded by the fact that structure
and composition of urban forests vary along climate gradients likely
due to the differing ability of species to survive biophysical conditions
(Ossola and Hopton, 2018a). Analyses of species' current environmental
distributions in natural and urban settings show that cool-origin trees
tend to be planted in warmer cities, and vice-versa (Kendal et al.,
2018). While heat stress might be exacerbated in urban areas due to
the UHI (Corburn, 2009), irrigation may mitigate moisture stress
(Jenerette et al., 2016; Vogt et al., 2017). In this way, the effects of cli-
mate change on trees might be superimposed on other characteristics
the urban biophysical environment, potentially leading to shifts in the
distribution of suitable climatic space across urban areas.
Climate suitability models (CSMs, also termed species distribution
models) are widely used tools for forecasting climate impacts on either
single species, or suites of species (Elith and Leathwick, 2009). These
correlative models assess the relationship between the location of indi-
viduals of the species and the environmental characteristics of those lo-
cations, thus assuming that the species' environmental tolerances are
described by the location of previously recorded individuals (Franklin
and Miller, 2010). CSMs can then be used to map the current distribu-
tion of suitable habitat for a species, identify suitable areas beyond the
species' known occupied range, and assess how suitability may change
under past or future climate scenarios (Baumgartner et al., 2018).
The framework for utilising CSMs in natural environments is well
developed [e.g. Guisan and Zimmermann, 2000]. However, the applica-
tion of these models for species in urban areas is less common [see
Rhodes et al., 2006; Akasaka et al., 2015; Beninde et al., 2016 for exam-
ples]. CSMs provide a means for undertaking a relatively rapid assess-
ment of species' realised tolerances to climate. Further, CSMs enable
habitat suitability to be estimated based on high-level biophysical filters
for species (i.e. climatic conditions), and regardless of local factors that
can affect species' occurrence in highly modified and managed urban
landscapes (such as management with irrigation, fertilisation, etc.). In
this way, predictions from CSMs facilitate the assessment of the sensi-
tivity of species to climate change under limited input of resources
(e.g. water), which is critical as such resources are costly for local gov-
ernment authorities, and will be also impacted by climate change.
In this study, we assessed the baseline (1960–1990), short term
(2030) and long term (2070) distribution of climatically suitable habitat
for 176 native Australian tree species that are commonly planted across
Australia's urban landscapes and are also currently sold in commercial
plant nurseries. These species are a representative sample of the
continent's urban tree flora, and thus are indicative of how urban trees
may fare under climate change. In doing so, we asked: 1) which tree
species are more likely to experience increases and decreases in the dis-
tribution of climatically suitable habitat across urban areas in the short
(2030) and long term (2070); and 2) which urban areas are more likely
to experience substantial increases or decreases in the number of spe-
cies with suitable habitat by 2030 and 2070? We expect that urban
areas in cooler regions will experience greater gains in the number of
species with climatically suitable habitat than urban areas in warmer re-
gions, consistent with findings from previous studies modelling large
suites of species occurring in natural environments on the Australian
continent [e.g. O'Donnell et al., 2012]. We also identify key challenges
associated with the use of CSMs for species in urban environments,
and further discuss the implications for urban forestry, planning and in-
dustry efforts in order to create future urban forests that are more resil-
ient to climate change.
2. Methods
2.1. Defining significant urban areas (SUAs)
Our spatial units for estimating species' climatic suitability are
Australia's Significant Urban Areas (SUAs). SUAs are defined by the
Australian Bureau of Statistics (ABS) as containing at least 10,000 people
within a single labour market (ABS, 2018). The boundaries for these
SUAs were obtained from the Australian Bureau of Statistics (ABS,
2018) and refer to the year 2016. Combined, the 82 SUAs span
~46,211 km2, and range in size from ~46 km2 (City of Sale, Victoria) to
6189 km2 (City of Melbourne, Victoria).
2.2. Tree species selection
To analyse horticulturally significant Australian species, we obtained
a list of plant species currently grown by the nursery industry across
Australia, including a count of nurseries known to be growing each spe-
cies within each state or territory (M Plumber, pers. comm). Duplicate
records or records that contained topiary (i.e. shaping) descriptions
for a species were condensed into a single record for that species. The
remaining species records were designated an origin (native or exotic)
using a range of state-based botanic garden and government databases
(e.g. NSW-PlantNET: http://plantnet.rbgsyd.nsw.gov.au/, SA-eFlora:
http://www.flora.sa.gov.au/, WA-FloraBase: https://florabase.dpaw.wa.
gov.au/). The exotic species records were then removed from the list.
We then created a spatial database of urban tree inventories in
Australia by contacting local government authorities (see Table S1 in
supporting information for further details). This resulted in
georeferenced tree inventory data for 44 local councils spanning 49
SUAs. The 400 most frequently reported tree species in the tree species
database were then intersected with the nursery list, creating a list of
248 native tree species. We consider these 248 species as both com-
monly planted across Australian urban areas and currently sold in the
Australian horticultural market. In order to standardise taxonomy, the
final 248 tree species list was checked against the backbone taxonomy
of the Global Biodiversity Information Facility (GBIF, www.gbif.org,
accessed on 15/10/2018), and then against The Plant List (TPL) using
the Taxonstand package [version 2.1, Cayuela et al., 2012] in the R
 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
language [Version 3.5.1, R Core Team, 2018]. The accepted names from
the TPL taxonomy were then used for this analysis, while varieties, cul-
tivars and sub-species were considered at species level.
2.2.1. Species occurrence data
For each of the species included in this study, we downloaded global
occurrence records from GBIF and the Atlas of Living Australia (ALA,
www.ala.org.au) using the rgbif and ALA4R packages in R [version
3.5.1, R Core Team, 2018] [https://github.com/AtlasOfLivingAustralia/
ALA4R, Chamberlain and Boettiger, 2017]. We then undertook an inten-
sive spatial data-cleaning process to remove spatially invalid or suspect
records that, if retained, can cause species' climate niches to be
miscalculated. Spatially invalid records, those collected before 1950, du-
plicate records, those records within 10 km of capital cities (which are
likely to be herbarium records, and thus not reflective of the species
realised niche), herbaria and biodiversity institutions, and individual re-
cords N300 km from other records for that species were removed using
the CoordinateCleaner package [Zizka et al., 2019, version 1.0–7,
resulting in ~22% of ~ 2.2 million records for 248 species being re-
moved]. Although this restriction removed some cultivated records, nei-
ther the ALA nor GBIF have comprehensive information about species'
cultivated status. As the tree inventory data was collected at the stem
level by professional arborists, it gives a more comprehensive and ro-
bust representation of the ‘urban niche’ than do natural occurrence re-
cords alone. Thus, we combined the georeferenced tree inventory
records with the GBIF and ALA records.
Due to Australia's history of urbanisation – which was largely con-
centrated in the temperate south east – arid and tropical regions have
fewer SUAs and tree inventories than SUAs in temperate and subtropi-
cal regions. Consequently, we report our results for only the 82 SUAs
Fig. 1. Map of 82 Australian Significant Urban Areas (SUA, plotted in pink) included in this study, with the largest SUA in each state and territory plotted in blue symbols. According to the
Köppen classification of Kriticos et al. (2012), all these SUAs fall within ‘temperate’ zones. (For interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)
453
found in Australia's five ‘temperate’ Köppen zones, according to the
Köppen classification developed by Kriticos et al. (2012), using the
kgcv R package (v.1.0.0.2, see Fig. 1). Note that this ‘temperate’ classifi-
cation includes mild temperate climates in subtropical regions – for ex-
ample Mackay and Rockhampton in Queensland – through to cooler
temperate climates in Tasmania. We then calibrated climate suitability
models (CSMs) within Australia's temperate SUA's for the 248 species,
using the modelling approach summarised below (Section 2.4).
2.3. Climate data
We obtained baseline climate data from the WorldClim Database
[worldclim.org/bioclim, Version 1.4, Hijmans et al., 2005]. WorldClim
comprises 19 bioclimatic variables, summarised for the period
1960–1990, of which we used eight for model calibration: 1) annual
mean temperature, 2) temperature seasonality, 3) maximum tempera-
ture of the warmest month, 4) minimum temperature of the coldest
month, 5) annual precipitation, 6) precipitation seasonality, 7) precipi-
tation of the wettest month, and 8) precipitation of the driest month.
These variables were chosen to capture climate averages, seasonality
and extremes, all of which have been identified as important variables
for predicting climatically suitable habitat for plants (Bradie and
Leung, 2017). Importantly, when averaged across the species analysed,
the tree inventory data shows the urban climate niche (quantified using
tree inventory records) is effectively a subset of the ‘natural’ niche
(quantified using occurrence records from GBIF and ALA, see Appendix,
Fig. S1).
When assessing the impacts of climate change, projections from
multiple global climate models should be used to account for
between-model variation (Beaumont et al., 2008; Taylor et al.,
454 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
2012; Intergovernmental Panel on Climate, 2014; Baumgartner
et al., 2018). We utilised a subset of six global climate models rec-
ommended by CSIRO's Climate Change in Australia report (for Rep-
resentative Concentration Pathway 8.5 (RCP 8.5), see https://www.
climatechangeinaustralia.gov.au/en/, accessed 20/09/2018). These
models have been shown to perform better than others in their
ability to simulate historical Australian climate (Table 1), hence
more confidence can be placed in their projections of future cli-
mate. For each of these six GCMs, we downloaded monthly maxi-
mum and minimum temperature and monthly precipitation from
CSIRO for 2030, 2050 and 2070, at a spatial resolution of 30 arc-
seconds (~1 km). From these data, we calculated the eight biocli-
matic variables for both time periods and re-projected the data to
the Australian Albers Equal Area Conic projection (EPSG:3577) at
1 km × 1 km resolution to match the baseline climate, using R
[version 3.5.1, R Core Team, 2018].
2.4. Modelling approach
We modelled climatic suitability under baseline conditions
(1960–1990) using the Maxent algorithm (Phillips et al., 2006; Phillips
and Dudík, 2008; Elith et al., 2011) within the dismo R package
[Hijmans et al. (2016), version 1.1–4]. Maxent is a machine learning, cor-
relative approach to modelling climatic suitability that is generally con-
sidered superior to other algorithms when true absence data are
unavailable (Elith et al., 2006). The resulting map generated by Maxent
illustrates how climatic suitability for the modelled species varies across
the landscape, with suitability in grid cells ranging from 0 (highly un-
suitable) to 1 [highly suitable, see Elith et al., 2011; Merow et al., 2013
for more details of the Maxent algorithm].
Models were calibrated using most of the default Maxent settings, al-
though hinge and threshold ‘features’ [mathematical transformations/
simplifications of predictor variables, Elith et al., 2011] were disabled
to reduce overfitting the models. In addition to occurrence records,
Maxent requires ‘background’ data that characterise the surrounding
environment in which the modelled species occurs. Our background
points comprised random samples of up to 70,000 cells from the pool
of cells that (a) contained occurrence records in ALA for one or more
plant species, (b) fell within 200 km of records for the target species
[i.e. a buffered target-group background, Elith and Leathwick, 2007;
Phillips and Dudík, 2008], and (c) were within the same Köppen climate
zone as the modelled species' known occurrences [using the Köppen
Table 1
Six climate models utilised for this project to assess potential impacts of climate change on
native Australian tree species planted in urban areas (all for Representative Concentration
Pathway 8.5, CSIRO/BOM 2015).
Climate Description of projection for 2070
model
CanESM2 Forecasts an extremely hot, dry future, with warming N4 °C
throughout much of Australia. Annual precipitation is projected to
decline in central and Western Australia, and increase in north--
east Queensland, with few changes in the south-east.
ACCESS1.0 Projects a hot, dry future. Warming exceeds 2.5 °C across most of
Australia, and N 3.5 °C in central Australia. Drying is projected over
most areas.
MIROC5 Projects moderate warming, not exceeding 3 °C, and slight changes
in annual precipitation with declines in NE Queensland and SW
Australia.
HadGEM2 Projects a hot future with warming typically N2.5 °C. Annual
precipitation is projected to increase in central Australia and
decline in most other places.
NorESM1 Projects moderate warming, with most of the continent exceeding
2 °C. Little change in annual precipitation is projected, particularly
in the south-east, although there is drying in SW Australia.
GFDL-ESM2M Projects a hot, very dry future, with warming in central regions
exceeding 3.5 °C. Drying is projected across most of the continent,
and exceeding a 20% decline in many areas.
classification from Kriticos et al., 2012]. This approach was utilised to
ensure that both the occurrence and background points had similar spa-
tial biases, a key requirement for improving Maxent model calibration.
2.4.1. Assessing model performance
The performance of each model was estimated by calculating the av-
erage Area Under the Receiver Operating Characteristic curve (Swets,
1988) and the True Skill Statistic (TSS) on test data through five-fold
cross-validation. This involved splitting the occurrence data for each
species into five random subsets of close to equal size (i.e. folds), fitting
the model to four of the five folds and predicting to the fifth. This pro-
cess was repeated until each fold was used four times for model fitting
and once for model evaluation (Stone, 1974). Subsequently, CSMs
were fitted a final time using the complete set of species data.
It is important to emphasize that models can have high performance
indicators with respect to test/training data, yet still over-project suit-
ability (i.e. although the model projects some cells to be suitable,
given ecological knowledge of the species these cells are unlikely to be
suitable in reality). This may indicate that models are encountering
novel conditions. To explore this, we calculated Multivariate Environ-
mental Similarity Surfaces [MESS, Elith et al., 2010; Baumgartner et al.
(2017)]. Although MESS maps indicated that the models were not
projecting to novel states for individual variables, this approach does
not consider novel combinations of variables. An alternate explanation
for over-projection is that suitable conditions for the species are influ-
enced by environmental variables not considered in the model [such
as edaphic conditions, see Barry and Elith, 2006]. As such, we visually
assessed all maps of current suitable habitat, and excluded from our
analysis models whose resulting maps contained substantial over-
projection. Although this approach is somewhat subjective, excluding
questionable models in this way adds more rigour to our analysis than
if we were to rely solely on AUC and TSS values.
This expert validation procedure suggested that for 72 species, addi-
tional/alternative climate variables (beyond the readily available biocli-
matic variables) may be required in order to adequately characterise the
species' realised climate niche. Hence, the following results are based on
data for the remaining 176 species with robust models only. The CSM
for each of these species was projected onto the six climate scenarios
for two future time periods (2030 and 2070), using the project function
in the rmaxent package [https://github.com/johnbaums/rmaxent;
Baumgartner et al., 2017].
2.4.2. Summarising CSM output
For each species, we generated maps illustrating the projected distri-
bution of suitable climate under baseline and future conditions. In these
maps, a grid cell is given a probability value between 0 (highly unsuit-
able) and 1 (highly suitable). Using a species-specific threshold – the
10th percentile training presence logistic threshold – based on the
weighting of different model errors (‘commission’ errors, where a grid
cell is classified as suitable when it is not, versus ‘omission’ errors,
where a grid cell is classified as unsuitable when it is suitable) the
maps were converted into binary representations of regions that are
less or more suitable (0 or 1). Henceforth, we refer to these regions as
‘no to low suitability’ and ‘suitable’ regions, respectively. We suggest
that this terminology more clearly articulates the subjective nature of
threshold selection.
For each future time period, the six maps representing the six cli-
mate scenarios were overlaid and summed, such that the value of a
grid cell could range from 0 (no to low suitability in all climate scenar-
ios) to 6 (suitable in all climate scenarios). Maps were then re-coded
to indicate whether grid cells were classified as suitable in the majority
of climate scenarios (i.e. at least four of the six scenarios). These grid
cells are most likely to be climatically suitable for the modelled species
in the future. We then calculated percent changes to the size of suitable
climate in terms of (i) overall change in size, (ii) loss of baseline suitable
areas and (iii) gain of newly suitable areas. Finally, we intersected
 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462 455

Fig. 2. Examples of horticulturally-significant tree species for which climatically suitable area is predicted to expand (Pongamia pinnata, Indian beech), shift (Melia azedarach, Chinaberry)
and shrink (Syzygium smithii, Common lilly pilly) for 2030 (black areas) and 2070 (orange areas), relative to the baseline (1960–1990) period (grey areas). The centroids of Australia's
‘temperate’ Significant Urban Areas (as classified by Kriticos et al. (2012)) are plotted in pink on the baseline map. Blue arrows indicate predicted species movements under climate
change. Note that ‘plantings’ refers to records from tree inventories, ‘records’ refers to occurrence records within the Atlas of Living Australia (ALA) and Global Biodiversity Information
Facility (GBIF), ‘growers’ refers to the number of Australian nurseries recorded in our industry dataset as selling that species. We classify Indian beech and Chinaberry as being at low
risk from climate change, as climatically suitable habitat for these species expands or remains widespread, respectively, until at least 2070. In contrast, climatically suitable habitat for
the Common lilly pilly is predicted to contract under future climate, making this species at potentially higher risk from climate change. (For interpretation of the references to color in
this figure legend, the reader is referred to the web version of this article.)
456 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462 457

species' maps with Significant Urban Areas (SUAs) from the 2016 ABS
Australian Census. We then calculated the area predicted to be climati-
cally suitable within the 82 SUAs under current climate, and the extent
to which this area may change under the future scenarios. Preliminary
analyses demonstrated that environmental space tends to be poorly
characterised by Maxent background samples in arid/tropical SUAs, fur-
ther justifying the exclusion of these SUAs from results (see
Section 2.2.1).
Finally, we fitted generalised additive models (GAMs) to assess the
relationship between both the mean annual temperature, and the max-
imum temperature of the warmest month (1960–1990) within an SUA
– averaged across grid cells – and (a) the percent of species lost (i.e. per-
cent of species with suitable climate within the SUA during the baseline,
that lacked suitable climate in the future) and gained (i.e. percent of
species with suitable climate within the SUA in the future, but not dur-
ing the baseline) and (b) the percent of the SUA's area that changed over
time from (i) unsuitable to suitable or (ii) suitable to unsuitable aver-
aged over all species. We used the mgcv R package [version 1.8.4,
Wood, 2011; R Core Team, 2018] to fit GAMs, using a restricted maxi-
mum likelihood approach (i.e. REML). Note that we also fitted GAMs
using annual precipitation (see Figs. S2–S5 in supporting information)
but the relationships show no pattern.
3. Results
Models for all 176 species had AUC and TSS values sufficiently high
to indicate acceptable predictive power [median AUC = 0.869 (±0.6)
and median TSS = 0.6 (±0.09), see Table S1 in supplementary material,
Swets, 1988]. Overall, the spatial extent of climatically suitable habitat is
predicted to decline from an average of 16,152 km2 (±9082 km2) in the
baseline period to 13,043 km2 (±8924 km2) by 2030, and 12,300 km2
(±,016 km2) by 2070. Of the 176 species, 73% are predicted to experi-
ence declines in the extent of climatically suitable habitat across their
respective SUAs: 18% are predicted to lose N50% of this habitat by
2030, while 34% are predicted to lose N50% by 2070. Fourteen species
(8%) were projected to have losses exceeding 90% of baseline climati-
cally suitable habitat [e.g. Callitris oblonga (Pygmy cypress pine) and
Acacia fimbriata (Brisbane golden wattle)]. In contrast, 11 species are
predicted to experience increases in the extent of climatically suitable
habitat of N50%, e.g. Ficus platypoda (Rock fig) and Brachychiton rupestris
(Queensland bottle tree). Generally, suitable habitat is predicted to shift
poleward. As such, species are likely to experience an extension of their
southern (colder) range margin (e.g. Pongamia pinnata and Melia
azedarach in Fig. 2), and/or a contraction of their northern (warmer)
margin (e.g. Melia azedarach and Syzygium smithii in Fig. 2). However,
some species – particularly those found along coastal regions of south-
east Australia – may experience contraction at both range margins
(see Syzygium smithii, Fig. 2).
SUAs were predicted to contain suitable climate for between 10 and
139 of the 176 species for the baseline period, with an average of 74 spe-
cies (±30). By 2070, this average is predicted to decline to 63 species (±
30) per SUA, with ten SUAs (eight in eastern Queensland and two in
south-west Western Australia) predicted to have suitable climate for
50% fewer species (Fig. 3). However, 21 of the 82 SUAs may contain suit-
able climate for more species by 2070, than during the baseline period,
including Orange in NSW (50% more species), Wangaratta in Victoria
(25%), and all five of the Tasmanian SUAs in this study (6–24%). Note
that these SUAs predominantly occur in cooler regions.
In general, the spatial extent of climatically suitable habitat across
SUAs was predicted to decline for the 176 species analysed. That is, av-
eraged across their representative species, 12.3% (±5.8%) of each SUA's
currently suitable area will no longer be suitable by 2070, although pre-
dicted gains of 6.3% (±4.3%) occur elsewhere within the SUA. SUAs with
warmer baseline mean annual temperatures are predicted to have
fewer tree species that experience increases in the spatial extent of suit-
able habitat, compared to cooler SUAs (Fig. 4). Similarly, the percentage
of tree species predicted to experience losses in suitable habitat is
greater in warmer SUAs (e.g. 80% in Rockhampton, Queensland) than
in cooler SUAs (e.g. 13% in Burnie, Tasmania). These patterns of change
were consistent regardless of the area and population of the SUAs
(Fig. 4).
4. Discussion
4.1. Urban trees and climate change
4.1.1. Predicted changes to climatically suitable habitat for horticultural
tree species
Climate suitability models (CSMs) for 176 commonly-planted, na-
tive Australian tree species show that climate change will likely reduce
climatically suitable habitat within the majority of the 82 temperate and
subtropical urban areas included in this study. Patterns in the loss of cli-
matically suitable habitat for urban species are broadly similar for pre-
dictions for plant species in natural ecosystems. That is, suitable
habitat for urban trees is generally contracting, and shifting pole-
wards, similar to trends in natural populations [e.g. Hickling et al.,
2006; O'Donnell et al., 2012; Butt et al., 2013]. This overarching pattern
is driven by the projection of a poleward shift in climate space in the un-
derlying climate models: hence, generally within temperate and sub-
tropical regions, species' low-latitude (here, northernmost) range
margins are predicted to contract, while their high-latitude (southern-
most) margins may extend into new areas.
Predicted shifts in climatically suitable habitat mean that urban
areas in cooler regions (for example, the city of Orange in central west-
ern New South Wales, and the city of Launceston in northern Tasmania),
are likely to experience smaller reductions in the number of species
with suitable habitat than urban areas in warmer regions (such as the
cities of Mackay and Brisbane in Queensland) (Fig. 5). This predicted
broad pattern of decreasing habitat along a gradient of increasing tem-
perature supports previous studies based on realised climate niches
(Jenerette et al., 2016; Kendal et al., 2018), which also demonstrated
that temperature is a key filter on trees in urban environments. Hence
our results, combined with those of other studies quantifying climate
niches of species (Jenerette et al., 2016; Kendal et al., 2018; McBride
and Laćan, 2018) indicate that a substantial proportion of the most com-
mon urban trees face climate change risks where they are currently
planted.
As with all CSM studies, it is important to highlight several key ca-
veats. Firstly, due to the availability of quality occurrence and tree in-
ventory data in eastern Australia, the majority of the 176 tree species
we modelled are temperate (e.g. Syzygium smithii, see Appendix,
Table S2). It is possible that the responses of trees planted within trop-
ical and sub-tropical SUAs may drive overall patterns of species turn-
over in the warm temperate SUAs included in our study (e.g. city of
Rockhampton in Queensland). Secondly, we have only considered
macro-climatic variables in our CSMs. Clearly, other factors – from mi-
croclimate, extreme weather events and edaphic conditions, to

Fig. 3. Scatterplots of projected changes to the climatically suitable habitat of 176 commercial native Australian tree species within 82 of Australia's Significant Urban Areas (SUAs) (y axis)
vs. current mean annual temperature of the SUA (x axis, a-d), and maximum temperature of the warmest month (MAXT, e–h), for 2030 (lighter points) and 2070 (darker points). Panels a,
b, e and f show the number of species, within each SUA, for which climatically suitable habitat is absent under baseline conditions, but present in the future (grey = 2030, black = 2070),
expressed as a proportion of the number of species with suitable conditions in the baseline (i.e. species ‘gained’). Panels c, d, g and h show the proportion of species within each SUA with
suitable habitat under baseline conditions, that do not have suitable habitat in the future (i.e. species ‘lost’). The left panels are for all 82 SUAs, while the right panels are for the SUAs with
area N 200 km2 and with population N 80,000. DE = deviance explained from generalised additive models (GAMs) of species gain/loss (y axis) in each SUA vs. the SUA MAXT (x axis).
458 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462 459

Fig. 5. For each of Australia's largest ‘temperate’ Significant Urban Areas [SUA, based on the classification of Kriticos et al., 2012], we plot the predicted change in average % of cells lost
(orange), gained (green) and stable (light blue) across all 176 species, from baseline (1960–1990) to 2070. Australia's ‘temperate’ SUAs are plotted in pink on the baseline map. (For in-
terpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

phenotypic plasticity [e.g. Singer et al., 2016; Senner et al., 2018; (Benito
Garzon et al., 2019)] – will also influence the suitability of urban areas
for particular species. Thirdly, the ability to adjust otherwise limiting
variables, such as rainfall through the use of irrigation (Moore, 2016;
Vogt et al., 2017), may enable some species to survive in urban areas
that CSMs suggest are unsuitable (e.g. espaliered fruit trees). Locations
characterised by contractions in the range of climatically suitable habi-
tat predicted here may also be able to support urban species via addi-
tional watering or provision of shade, and our findings can assist in
identifying regions which may require new maintenance regimes in
the future. Revised maintenance requirements are just one response
to climate change in urbanised areas, and below we discuss further im-
plications for urban planning.
4.2. Planning for climate-ready urban vegetation
4.2.1. Urban forest planning and governance
There is knowledge of which species are likely to succeed or fail in
particular locations – and the effect of climate extremes – within both
the horticultural industry, and the local government agencies that select
and maintain urban trees. Horticulturalists recognise that some urban
tree species experience reductions in growth – or even mortality – dur-
ing extreme weather events such as heatwaves and droughts. For exam-
ple, dendrochronological analyses of four urban tree species in
Melbourne, Australia, demonstrated that drought had a significant, neg-
ative impact on radial stem growth (Nitschke et al., 2017). Such impacts
are likely to be exacerbated as climate in the southern hemisphere be-
comes warmer and drier [e.g. Whetton et al., 2012].
In addition, there is a growing body of evidence highlight changes to
growth rates of urban trees likely due to climate change [e.g. Pretzsch
et al., 2017; Jia et al., 2018]. For example, Pretzsch et al. (2017) found
that, on average, urban trees growing in 10 metropolises located across
a range of biomes have experienced accelerated growth since the 1960s.
However, comparisons of the growth rate of trees in urban and sur-
rounding rural zones indicated that the urban trees in cities experienc-
ing subtropical climates will likely experience climate change-driven
declines in growth before trees in temperate, Mediterranean or boreal
regions (Pretzsch et al., 2017). Such empirical studies provide important

Fig. 4. Scatterplots of changes to the percent of species with climatically suitable habitat within 82 of Australia's Significant Urban Areas (SUAs) (y axis) vs. current mean annual temper-
ature of the SUA (x axis, a–d), and maximum temperature of the warmest month (MAXT, e–h), for 2030 (lighter points) and 2070 (darker points). ‘Range gained’ refers to the proportion of
species in the future time period projected to experience a net increase in the spatial extent of suitable habitat within that SUA vs baseline, while ‘Range loss’ refers to the proportion of
species projected to experience a net decrease in spatial extent of suitable habitat within that SUA. The left panels are for all 82 SUAs, while the right panels are for the SUAs with area
N 200 km2 and with human population N 80,000. Deviance = deviance explained by generalised additive models (GAMs) of species range gain/loss (y axis) in each SUA vs. the SUA
MAT/MAXT.
460 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
data to compare with correlative methods (e.g. using Maxent), and in-
corporating physiological parameters such as growth rates into CSMs
is a key frontier of research [e.g. see Benito Garzon et al., 2019].
As the reality of climate change is increasingly being acknowledged,
urban planners are recognising the challenges that climate change
poses, particularly for trees, which represent a substantial investment
and management effort [e.g. McPherson et al., 2018]. Urban trees are
long-term assets, with some plantings intended to survive for decades
– periods over which climate change is projected to intensify. Yet, de-
spite the acknowledgement from local authorities of the importance
of urban forests in mitigating the impacts of climate change [e.g. UHI,
heatwaves, etc., see Norton et al., 2015], there is still little recognition
that green assets themselves will be impacted by the very factors they
are intended to mitigate [e.g. Ordóñez and Duinker, 2015].
4.2.2. Climate-ready species selection for the horticultural industry
The repercussions for the nursery industry of this study's predicted
losses and gains of climatically suitable habitat for horticultural plant
species are two-fold. Firstly, the selection of species based on hierarchi-
cal filters that identify climate as the pivotal biophysical limiting factor
would substantially improve outcomes for cities and industry. However,
while projections of an area having negligible suitability for a plant does
not preclude that species from being grown there, it does indicate that a
higher input of resources may be required to sustain that plant. For in-
stance, lethal temperature thresholds of species may be increased
under sufficient irrigation, and transpirational loss may be reduced by
strategic pruning [e.g. Li et al., 2003] or provision of shade. Specific
planting or microclimatic contexts may also permit some plants to sur-
vive in areas that are broadly climatically unsuitable (such as within
areas less affected by urban heat). Monitoring of tree health after ex-
treme events will also help to identify individuals experiencing stress.
Secondly, our results signify new opportunities for growers: species
that may have been resilient in horticultural plantings under previous or
current conditions may be unreliable in the future, while new opportu-
nities will emerge as suitable climate space appears beyond a species'
current range. There are in excess of 1500 native and exotic species
planted across Australian cities (unpublished data with this manu-
script). However, with N6000 native species of trees and large shrubs
(Gallagher et al. in prep), Australia has a large diversity of woody plants
that could be tested for horticultural potential. Indeed, increasing the di-
versity of urban forests has been signalled as a priority to increase resil-
ience to pest, diseases and climate change [e.g. Dale and Frank, 2014;
Ordóñez and Duinker, 2015]. Our analysis highlights some of these po-
tential species: if considerations such as climatic suitability were para-
mount in plant selection – rather than cultural or aesthetic factors, as
is often the case (Tyrväinen et al., 2003) – the native arboreal palette
within an urban area could be expanded considerably. For instance, na-
tive Australian species such as Corymbia ficifolia (Red-flowering gum)
are popular because of their habit and flowers, but there could well be
many similar, but underutilised, species in the Australian flora.
Unfortunately, many species that are likely to be robust to climate
change may be impractical for urban plantings, due to horticultural
and/or logistical considerations (e.g. growth rate to market, profitabil-
ity, aesthetics, or risks such as limb drop). Thus to comprehensively as-
sess the suitability of a broad suite of new tree species, it will be
necessary to undertake assessment of climate suitability, as well as col-
late data on species' traits which capture other aspects of their physiol-
ogy, phenology and morphology. These may include traits requiring
detailed measurement, such as δ13C or gas exchange to determine
water-use efficiency, as well as drought and heat tolerance. Equally,
traits such as growth rate and canopy architecture may be important
for assessing potential longevity and suitable placement of tree species
as climate changes. These trait-based decisions will also need to be
constrained by the aesthetic needs of an urban planting (e.g. floral col-
our and characteristics, shade provision), as well as maintenance re-
quirements [e.g. root damage, pruning, irrigation – see Van Mechelen
et al., 2015]. Ultimately, combining the strengths and weakness of mul-
tiple lines of evidence [e.g. McKenney et al., 2007; Hunter, 2011;
Nitschke et al., 2017] into consistent frameworks for plant selection
[e.g. Gilbert, 2013; McPherson et al., 2018] is essential for planning, cre-
ating and maintaining urban ecosystems under climate change.
4.2.3. Application of CSMs in urban areas: Limitations and ways forward
The application of CSMs to urban areas is in its infancy, requiring fur-
ther development. The scant literature regarding urban CSMs [e.g. Caryl
et al., 2014; Beninde et al., 2016] is partly driven by the paucity of spatial
occurrence data in urban areas: in the past, scientists have been more
interested in recording species in natural landscapes or of economically
significant species. Consequently, many important horticultural plants
are difficult to robustly model using correlative methods (e.g. Magnolia
grandiflora, which is globally horticulturally significant, yet largely unre-
corded in the global occurrence databases https://www.gbif.org/
species/9605163). Harvesting of novel spatial data sources (e.g. global
urban tree inventories from local government authorities, citizen sci-
ence database like iNaturalist, etc.) could help address this bias, and im-
prove quantification of species' climatic niches. Shared benchmarks and
frameworks between decision-makers, stakeholders and researchers –
and the establishment of shared data platforms – would also increase
the wealth of data needed to devise and calibrate accurate CSMs across
large sets of species and urban areas [e.g. (1998); Janse and
Konijnendijk, 2007; Östberg et al., 2018].
Future applications of CSMs to urban forests could also aim to better
incorporate urban environmental drivers affecting trees (e.g. the UHI ef-
fect, or other drivers due to climate change, such as the CO2 fertilisation
effect). Inevitably, some species will be unable to tolerate conditions in
an urban area that models suggest will be suitable. For instance, species
may be exposed to temperatures beyond their physiological tolerance
as a result of the UHI or urban microclimatic effect (e.g. building heat ir-
radiance), or due to climatic extremes such as heat waves. Climate ex-
tremes are predicted to become more prevalent in Australia (Perkins-
Kirkpatrick and Gibson, 2017), but these were not considered in our
analysis due to paucity of data for urban areas. Nonetheless, CSMs pro-
vide a relatively rapid predictive framework to identify species less
likely to survive under future climate in a given region. This framework
could be utilised during tree replanting after removal (Ossola and
Hopton, 2018b), which would allow a significant reduction in the risk
associated with urban tree management in the medium and long
term. In summary, the creation of a region-specific climate-ready pal-
ette of species would allow a more systematic, cost-effective prioritiza-
tion of new plantings, green infrastructure and green spaces.
5. Conclusions
The number of current horticultural tree species with suitable cli-
mate within Australia's SUAs is predicted to progressively decline as
climate change intensifies, with more pronounced reductions in
urban areas which are currently warmest. Our results show that cli-
matic ‘opportunities’ for species differ markedly across the
Australian continent, indicating that a proactive approach is needed
to identify new climate-ready species and plantings for climate
change. The climate suitability model outputs we present provide a
useful baseline assessment of the impact of climate change on
urban tree species, but should not be used in isolation for urban for-
estry and planning. We advocate for an approach that combines
multiple lines of evidence to assess plant responses to climatic fac-
tors, particularly to shorter-term extreme events – including mea-
surements from plant traits and physiological experiments.
Combinatorial/synthetic approaches like this promote a more holis-
tic understanding of the resilience of urban plants and trees to cli-
mate change.
 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
Acknowledgements
The Which Plant Where project is funded by the Green Cities Fund,
as part of the Hort Frontiers Strategic Partnership Initiative developed
by Hort Innovation, with co-investment from Macquarie University,
Western Sydney University and the NSW Office of Environment and
Heritage, and funds from the Australian Government. We kindly ac-
knowledge Leigh Staas, Mathew Plummer and the city council officers
that kindly provided data and support.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.scitotenv.2019.05.287.
References
ABS, 2018. Significant urban area: delimitation of SUAs. Available at:. http://www.abs.
gov.au/ausstats/abs@.nsf/Latestproducts/7D88D2916BF4BBE3CA257A980013999D?
opendocument, Accessed date: 20 October 2018.
Akasaka, M., Osawa, T., Ikegami, M., 2015. The role of roads and urban area in occurrence
of an ornamental invasive weed: a case of Rudbeckia laciniata L. Urban Ecosyst. 18,
1021–1030.
Barry, S., Elith, J., 2006. Error and uncertainty in habitat models. J. Appl. Ecol. 43, 413–423.
Baumgartner, J.B., Wilson, P., Esperon-Rodriguez, M., 2017. Rmaxent: Tools for working
with Maxent in R. Version 0.8.3.9. https://github.com/johnbaums/rmaxent.
Baumgartner, J.B., Esperón-Rodríguez, M., Beaumont, L.J., 2018. Identifying in situ climate
refugia for plant species. Ecography 41, 1850–1863.
Beaumont, L.J., Hughes, L., Pitman, A.J., 2008. Why is the choice of future climate scenarios
for species distribution modelling important? Ecol. Lett. 11, 1135–1146.
Beninde, J., Feldmeier, S., Werner, M., Peroverde, D., Schulte, U., Hochkirch, A., Veith, M.,
2016. Cityscape genetics: structural vs. functional connectivity of an urban lizard pop-
ulation. Mol. Ecol. 25, 4984–5000.
Benito Garzon, M., Robson, T.M., Hampe, A., 2019. DeltaTraitSDMs: species distribution
models that account for local adaptation and phenotypic plasticity. New Phytol.
222, 1757–1765.
Bradie, J., Leung, B., 2017. A quantitative synthesis of the importance of variables used in
MaxEnt species distribution models. J. Biogeogr. 44, 1344–1361.
Butt, N., Pollock, L.J., McAlpine, C.A., 2013. Eucalypts face increasing climate stress. Ecol.
Evol. 3, 5011–5022.
Caryl, F.M., Hahs, A.K., Lumsden, L.F., Van der Ree, R., Wilson, C., Wintle, B.A., 2014. Con-
tinuous predictors of species distributions support categorically stronger inference
than ordinal and nominal classes: an example with urban bats. Landsc. Ecol. 29,
1237–1248.
Cayuela, L., Granzow-de la Cerda, I., Albuquerque, F.S., Golicher, D.J., 2012. Taxonstand: an
r package for species names standardisation in vegetation databases. Methods Ecol.
Evol. 3, 1078–1083.
Chamberlain, S.A., Boettiger, C., 2017. R Python, and ruby clients for GBIF species occur-
rence data. PeerJ Preprints 5, e3304v1.
Chiesura, A., 2004. The role of urban parks for the sustainable city. Landsc. Urban Plan. 68,
129–138.
Corburn, J., 2009. Cities, climate change and urban heat island mitigation: Localising
global environmental science. Urban Stud. 46, 413–427.
Dale, A.G., Frank, S.D., 2014. Urban warming trumps natural enemy regulation of herbiv-
orous pests. Ecol. Appl. 24, 1596–1607.
Elith, J., Leathwick, J., 2007. Predicting species distributions from museum and herbarium
records using multiresponse models fitted with multivariate adaptive regression
splines. Divers. Distrib. 13, 265–275.
Elith, J., Leathwick, J.R., 2009. Species distribution models: ecological explanation and pre-
diction across space and time. Annu. Rev. Ecol. Evol. Syst. 677–697.
Elith, J., Graham, C. H., Anderson, R. P., Dudík, M., Ferrier, S., Guisan, A., Hijmans, R. J.,
Huettmann, F., Leathwick, J. R., Lehmann, A., Li, J., Lohmann, L. G., Loiselle, B. A.,
Manion, G., Moritz, C., Nakamura, M., Nakazawa, Y., McC. M. Overton, J., Townsend
Peterson, A., J. Phillips, S., Richardson, K., Scachetti-Pereira, R., E. Schapire, R., Soberón,
J., Williams, S., Wisz, M. S. & Zimmermann, N E. (2006) Novel methods improve pre-
diction of species' distributions from occurrence data. Ecography, 29, 129–151.
Elith, J., Kearney, M., Phillips, S., 2010. The art of modelling range-shifting species.
Methods Ecol. Evol. 1, 330–342.
Elith, J., Phillips, S.J., Hastie, T., Dudík, M., Chee, Y.E., Yates, C.J., 2011. A statistical explana-
tion of MaxEnt for ecologists. Divers. Distrib. 17, 43–57.
Endreny, T.A., 2018. Strategically growing the urban forest will improve our world. Nat.
Commun. 9.
Espeland, E.K., Kettenring, K.M., 2018. Strategic plant choices can alleviate climate change
impacts: a review. J. Environ. Manag. 222, 316–324.
Franklin, J., Miller, J.A., 2010. Mapping Species Distributions: Spatial Inference and Predic-
tion. Cambridge University Press.
Gilbert, D., 2013. Urban tree species selection: an approach for adapting to a changing cli-
mate. Masters Guelph University, Ontario Available at:. https://atrium.lib.uoguelph.
ca/xmlui/bitstream/handle/10214/7504/Gilbert_Dayna_201309_MLA.pdf?se-
quence=3&isAllowed=y.
461
Grote, R., Samson, R., Alonso, R., Amorim, J.H., Cariñanos, P., Churkina, G., Fares, S., Thiec,
D.L., Niinemets, Ü., Mikkelsen, T.N., Paoletti, E., Tiwary, A., Calfapietra, C., 2016. Func-
tional traits of urban trees: air pollution mitigation potential. Front. Ecol. Environ. 14,
543–550.
Guisan, A., Zimmermann, N.E., 2000. Predictive habitat distribution models in ecology.
Ecol. Model. 135, 147–186.
Guo, F., Lenoir, J., Bonebrake, T.C., 2018. Land-use change interacts with climate to deter-
mine elevational species redistribution. Nat. Commun. 9, 1315.
Hickling, R., Roy, D.B., Hill, J.K., Fox, R., Thomas, C.D., 2006. The distributions of a wide
range of taxonomic groups are expanding polewards. Glob. Chang. Biol. 12,
450–455.
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high resolution in-
terpolated climate surfaces for global land areas. Int. J. Climatol. 25, 1965–1978.
Hijmans, R.J., Phillips, S., Leathwick, J., Elith, J., 2016. Dismo: Species Distribution Modeling.
R package version 1.1-4. http://CRAN.R-project.org/package=dismo.
Hunter, M., 2011. Using ecological theory to guide urban planting design: an adaptation
strategy for climate change. Landsc. J. 30, 173–193.
Intergovernmental Panel on Climate, C, 2014. Climate Change 2013 – The Physical Science
Basis: Working Group I Contribution to the Fifth Assessment Report of the Intergov-
ernmental Panel on Climate Change. Cambridge University Press, Cambridge.
Janse, G., Konijnendijk, C.C., 2007. Communication between science, policy and citizens in
public participation in urban forestry-experiences from the Neighbourwoods project.
Urban For. Urban Green. 6, 23–40.
Jenerette, G.D., Clarke, L.W., Avolio, M.L., Pataki, D.E., Gillespie, T.W., Pincetl, S., Nowak,
D.J., Hutyra, L.R., McHale, M., McFadden, J.P., Alonzo, M., 2016. Climate tolerances
and trait choices shape continental patterns of urban tree biodiversity. Glob. Ecol.
Biogeogr. 25, 1367–1376.
Jia, W., Zhao, S., Liu, S., 2018. Vegetation growth enhancement in urban environments of
the conterminous United States. Glob. Chang. Biol. 24, 4084–4094.
Kendal, D., Dobbs, C., Gallagher, R.V., Beaumont, L.J., Baumann, J., Williams, N.S.G., Livesley,
S.J., 2018. A global comparison of the climatic niches of urban and native tree popu-
lations. Glob. Ecol. Biogeogr. 27, 629–637.
Kriticos, D.J., Webber, B.L., Leriche, A., Ota, N., Macadam, I., Bathols, J., Scott, J.K., 2012.
CliMond: global high-resolution historical and future scenario climate surfaces for
bioclimatic modelling. Methods Ecol. Evol. 3, 53–64.
Lanza, K., Stone, B., 2016. Climate adaptation in cities: what trees are suitable for urban
heat management? Landsc. Urban Plan. 153, 74–82.
Li, K.T., Lakso, A.N., Piccioni, R., Robinson, T., 2003. Summer pruning reduces whole-
canopy carbon fixation and transpiration in apple trees. J. Hortic. Sci. Biotechnol. 78,
749–754.
Livesley, S.J., McPherson, G.M., Calfapietra, C., 2016. The urban forest and ecosystem ser-
vices: impacts on urban water, heat, and pollution cycles at the tree, street, and city
scale. J. Environ. Qual. 45, 119–124.
Luttik, J., 2000. The value of trees, water and open space as reflected by house prices in the
Netherlands. Landsc. Urban Plan. 48, 161–167.
McBride, J.R., Laćan, I., 2018. The impact of climate-change induced temperature increases
on the suitability of street tree species in California (USA) cities. Urban For. Urban
Green. 34, 348–356.
McKenney, D.W., Pedlar, J.H., Lawrence, K., Campbell, K., Hutchinson, M.F., 2007. Potential
impacts of climate change on the distribution of North American trees. BioScience 57,
939–948.
McPherson, E.G., Berry, A.M., van Doorn, N.S., 2018. Performance testing to identify
climate-ready trees. Urban For. Urban Green. 29, 28–39.
Merow, C., Smith, M.J., Silander, J.A., 2013. A practical guide to MaxEnt for modeling spe-
cies' distributions: what it does, and why inputs and settings matter. Ecography 36,
1058–1069.
Moore, G.M., 2016. The economic value of trees in the urban forest as climate changes. In:
Groening, G., Moore, E.E.F., Moore, G.M., Rayner, J.P. (Eds.), Acta Horticulturae. Inter-
national Society for Horticultural Science, pp. 1–12.
Nitschke, C.R., Nichols, S., Allen, K., Dobbs, C., Livesley, S.J., Baker, P.J., Lynch, Y., 2017. The
influence of climate and drought on urban tree growth in Southeast Australia and the
implications for future growth under climate change. Landsc. Urban Plan. 167,
275–287.
Norton, B.A., Coutts, A.M., Livesley, S.J., Harris, R.J., Hunter, A.M., Williams, N.S.G., 2015.
Planning for cooler cities: a framework to prioritise green infrastructure to mitigate
high temperatures in urban landscapes. Landsc. Urban Plan. 134, 127–138.
O'Donnell, J., Gallagher, R.V., Wilson, P.D., Downey, P.O., Hughes, L., Leishman, M.R., 2012.
Invasion hotspots for non-native plants in Australia under current and future cli-
mates. Glob. Chang. Biol. 18, 617–629.
Ordóñez, C., Duinker, P.N., 2015. Climate change vulnerability assessment of the urban
forest in three Canadian cities. Clim. Chang. 131, 531–543.
Ossola, A., Hopton, M.E., 2018a. Climate differentiates forest structure across a residential
macrosystem. Sci. Total Environ. 639, 1164–1174.
Ossola, A., Hopton, M.E., 2018b. Measuring urban tree loss dynamics across residential
landscapes. Sci. Total Environ. 612, 940–949.
Östberg, J., Wiström, B., Randrup, T.B., 2018. The state and use of municipal tree invento-
ries in Swedish municipalities – results from a national survey. Urban Ecosyst. 21,
467–477.
Parmesan, C., Yohe, G., 2003. A globally coherent fingerprint of climate change impacts
across natural systems. Nature 421, 37–42.
Perkins-Kirkpatrick, S.E., Gibson, P.B., 2017. Changes in regional heatwave characteristics
as a function of increasing global temperature. Sci. Rep. 7, 12256.
Phillips, S.J., Dudík, M., 2008. Modeling of species distributions with Maxent: new exten-
sions and a comprehensive evaluation. Ecography 31, 161–175.
Phillips, S.J., Anderson, R.P., Schapire, R.E., 2006. Maximum entropy modeling of species
geographic distributions. Ecol. Model. 190, 231–259.
462 H. Burley et al. / Science of the Total Environment 685 (2019) 451–462
Pretzsch, H., Biber, P., Uhl, E., Dahlhausen, J., Schütze, G., Perkins, D., Rötzer, T., Caldentey,
J., Koike, T., Con, T.V., Chavanne, A., Toit, B.D., Foster, K., Lefer, B., 2017. Climate change
accelerates growth of urban trees in metropolises worldwide/631/158/858/704/158/
2165 article. Sci. Rep. 7, 15403.
R Core Team, 2018. R: A Language and Environment for Statistical Computing. R Founda-
tion for Statistical Computing, Vienna, Austria.
Rhodes, J.R., Wiegand, T., McAlpine, C.A., Callaghan, J., Lunney, D., Bowen, M., Possingham,
H.P., 2006. Modeling species' distributions to improve conservation in semiurban
landscapes: Koala case study. Conserv. Biol. 20, 449–459.
Savi, T., Bertuzzi, S., Branca, S., Tretiach, M., Nardini, A., 2015. Drought-induced xylem cav-
itation and hydraulic deterioration: risk factors for urban trees under climate change?
New Phytol. 205, 1106–1116.
Senner, N.R., Stager, M., Cheviron, Z.A., 2018. Spatial and temporal heterogeneity in cli-
mate change limits species' dispersal capabilities and adaptive potential. Ecography
41, 1428–1440.
Serra-Diaz, J.M., Franklin, J., Ninyerola, M., Davis, F.W., Syphard, A.D., Regan, H.M., Ikegami,
M., 2014. Bioclimatic velocity: the pace of species exposure to climate change. Divers.
Distrib. 20, 169–180.
Singer, A., Johst, K., Banitz, T., Fowler, M.S., Groeneveld, J., Gutiérrez, A.G., Hartig, F., Krug,
R.M., Liess, M., Matlack, G., Meyer, K.M., Pe'er, G., Radchuk, V., Voinopol-Sassu, A.J.,
Travis, J.M.J., 2016. Community dynamics under environmental change: how can
next generation mechanistic models improve projections of species distributions?
Ecol. Model. 326, 63–74.
View publication stats
Stone, M., 1974. Cross-validation and multinomial prediction. Biometrika 61, 509–515.
Swets, J.A., 1988. Measuring the accuracy of diagnostic systems. Sci. Sci. 240, 1285–1293.
Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2012. An overview of CMIP5 and the experiment de-
sign. Bull. Am. Meteorol. Soc. 93, 485–498.
Tyrväinen, L., Silvennoinen, H., Kolehmainen, O., 2003. Ecological and aesthetic values in
urban forest management. Urban For. Urban Green. 1, 135–149.
Van Mechelen, C., Van Meerbeek, K., Dutoit, T., Hermy, M., 2015. Functional diversity as a
framework for novel ecosystem design: the example of extensive green roofs. Landsc.
Urban Plan. 136, 165–173.
Vogt, J., Gillner, S., Hofmann, M., Tharang, A., Dettmann, S., Gerstenberg, T., Schmidt, C.,
Gebauer, H., Van de Riet, K., Berger, U., Roloff, A., 2017. Citree: a database supporting
tree selection for urban areas in temperate climate. Landsc. Urban Plan. 157, 14–25.
Whetton, P., Hennessy, K., Clarke, J., McInnes, K., Kent, D., 2012. Use of representative cli-
mate futures in impact and adaptation assessment. Clim. Chang. 115, 433–442.
Wood, S.N., 2011. Fast stable restricted maximum likelihood and marginal likelihood es-
timation of semiparametric generalized linear models. J. R. Stat. Soc. Ser. B Stat
Methodol. 73, 3–36.
Zizka, A., Silvestro, D., Andermann, T., Azevedo, J., Duarte Ritter, C., Edler, D., Farooq, H.,
Herdean, A., Ariza, M., Scharn, R., Svantesson, S., Wengström, N., Zizka, V., Antonelli,
A., 2019. CoordinateCleaner: standardized cleaning of occurrence records from bio-
logical collection databases. Methods Ecol. Evol. 10, 744–751.