Climate change presents significant challenges for plants and plant selection. Temperatures are increasing, and rainfall is changing faster than plant species can adapt. This means that many species that we plant today may no longer cope with the future climate. This may be exacerbated in urban environments where impervious surfaces and the urban heat island effect enhance the stressful environments for urban plants. One novel method to select urban trees for future climates is based on climate niche analysis. Using global databases, the current climate niche of a tree species is extracted and compared against the predicted future climate at a given location. If the climate niche of a tree is not matched with the future climate niche of a city, then that tree species is deemed no longer suitable. These types of climate niche analysis are very common, and they are already influencing management and tree species selection decisions. For example, in the City of Melbourne over half of the current species have been deemed ‘unsuitable’ for a future climate based on climate niche predictions and species are getting replaced. Botanic gardens have developed climate niche tools for selecting ‘climate proof’ tree species. However, this climate niche approach is not without problems, and many potentially suitable tree species are being excluded because their distribution in nature imposes is likely a poor indicator of their tolerances. In this paper we critically examine this tree selection approach and highlight the “good, the bad and the ugly” of this and other methods and give examples how species selection could be improved.
Tree species selection for future climates is a significant issue for urban planners. Our climate is changing rapidly but tree species are long-lived organisms. Is the climate changing so fast that some species can no longer grow, thrive and survive? There is surprisingly little data if urban trees are increasingly stressed. Many municipalities collect urban tree inventory data, but at the same time have a poor documentation on the reasons for tree species poor health, removal or replacement. Consequently, there is very little empirical evidence that large proportions of the urban forest are impacted by changes in climate. It is thus most surprising that several recent studies based on climate niche analysis suggest that the majority of urban tree species are at significant risk from the impacts of climate change (Burley et al., 2019; Esperon‐Rodriguez et al., 2019; Kendall and Baumann, 2016). A recent study suggests that over 70% of tree species in urban centres will be at risk from projected changes in annual precipitation and mean annual temperature (Esperon-Rodriguez et al., 2022). These numbers are staggering and cause concern among urban foresters. However, a closer investigation of the study and the methodology would suggest that there is little evidence that trees in the urban forest will be severely impacted by changes in climate. There are key limitations in the methodology of climate niche analysis, flaws in the application of the safety margin concepts. Furthermore, there is a lack of consideration as to the tolerance mechanisms that many tree species possess to tolerate stress, and their capacity to adapt and acclimate to changing environmental conditions.
Climate niche modelling
A key methodological component of many recent studies on climate suitability of urban trees is climate niche analysis. This methodology calculates the climate niche of a species based on global occurrence databases like the Global Biodiversity Information Facility (GBIF) and climate data like WorldClim. The occurrence data of a species are matched to mean climate data for that location. The climate data are parameters like mean annual temperature or mean annual precipitation, but others like aridity indices or mean precipitation of the driest quarter are also used. The data are then plotted to evaluate the ‘climate niche’ of a species (see Figure 1). Key percentiles are calculated from the climate niche of each species to represent critical limits. For example, the 90th percentile for mean annual temperature has been used as an upper thermal limit. Species would be considered vulnerable if they were to be planted in a city which is predicted to have a mean annual temperature greater than the upper thermal limit. The future mean annual temperature of a city might be predicted to be 19.5 °C and this would be at the edge of the current occurrence records (at the 98th percentile), then that species would be vulnerable. If the predicted mean annual temperature would be 17.8 °C and that would correspond to the 70th percentile of distribution, then that species would not be vulnerable.
Figure 1. Example of a climate niche analysis of a tree species. The graphs show the number of records plotted against a climate parameter such as mean annual temperature or mean annual precipitation. Some publications assume that a threshold is reached when the climate of a city is outside the 10th or 90th percentile of the distribution for a given species, left figure (Esperon-Rodriguez et al., 2022). I.e. the current or future climate of that city is drier or hotter than most of the occurrence record for the species. Others are a bit more nuanced and assume high vulnerabilities if the future climate of a city is approaching the edges of the current distribution of a species, right figure (Kendall and Baumann, 2016).
Issues and limitations of climate niche modelling
At first glance this approach seems very logical and straight forward. We are lucky that we now have vast databases with occurrence records for many global plant species and that these can be matched with excellent climate records. Thus, the data are regarded as valid.
However, all climate niche analyses have several limitations. The data provide only information where plants currently or historically occur. But they do not provide any information about the locations where the species cannot occur. This is an important distinction: the absence of a datapoint does not mean that the species cannot exist in that climate or that part of the globe. It only means that it has not been observed there.
A second limitation is that the occurrence data do not provide any information about the number of plants observed at the one observation point. It could be a single tree on the side of the road, or it could be a forest with 100,000 trees. However, in the database these two observations are equally weighted, they just count as one data point. Conversely, there are other cases where many individual tree occurrences have been uploaded for the same location (e.g. Platanus x acerifoila for New York City), thereby biasing the occurrence vs climate record to one climate point. We also do not have information about the health status of the observation, if the tree was healthy, free of any damage or a poor and struggling specimen. And of course, we assume that the people collecting the data, which can be anyone, are properly trained in species identification.
The third and probably most significant limitation is that in most instances the occurrence record reflects the sum of all biological processes that led to the natural establishment of a plant at that location. In other words, it reflects the realised climate niche in nature (Soberon and Nakamura, 2009). This is displayed in Figure 2. Under natural conditions a plant will undergo a range of natural processes from flowering, to seed setting, seed distribution, seed germination and plant establishment (Young et al., 2005). All these processes are influenced by climate and each of the processes has a distinct climate niche. Some have broader climate niches than others, and it is usually the establishment niche that sets the climatic limit to where plants naturally occur, as a young plant is very vulnerable to climate during the time from seed germination to the establishment. However, a mature plant usually has much greater climatic tolerances and thus has a much broader climate niche. The climate that a plant can tolerate is called the fundamental climate niche. The issue is that the fundamental niche of most plants is unknown and that the climate niche that can be calculated from occurrence records reflects the realised niche of a species (Booth, 2018). Most records are observations of botanists in nature where plants would have undergone the natural processes described above.
Figure 2. Example of the various climate niches at different stages of the life cycle of a plant. The reproductive niche (green) indicates the climate suitable for processes such as flowering and seed setting. The dispersal niche (blue) is the climate where seeds can be distributed. The establishment niche (red) is the climate in which seeds germinate and a young plant establishes, which is usually the smallest niche as it is the most vulnerable phase of a plant. The adult niche is the climate in which a mature plant can exist, and this is usually much greater than the other niches. However, in nature and without human intervention the realised climate niche is set by the narrowest niche, in this case the establishment niche. Data are for illustration purposes only, modified after (Young et al., 2005).
However, there are a few examples of species where global occurrence data more likely represent the fundamental niche. These are species that have been grown commercially in plantations and this has increased their distribution relative to their natural distribution. The natural range of Pinus radiata was a very restricted climate near Monterrey in California – thus its name Monterrey pine (Rogers, 2002). Now it is a major soft wood species in Australia and New Zealand covering a vast climate niche. The range of mean annual temperature within the natural distribution of Eucalyptus cladocalyx (sugar gum) in Australia was 14-17 °C, but it was grown successfully at some sites in Africa where the mean annual temperature was as high as 21 °C (Booth 2017). The native distribution of Eucalyptus globulus, the Tasmanian blue gum, is on the island of Tasmania and the southern ranges in Victoria. Hence, it had a very narrow geographical and climatic distribution in its realised niche in a cold and high rainfall climate. Any climate niche analysis would classify this species as highly vulnerable to increases in temperature or decreases in rainfall. However, today it is grown as a plantation species in almost every continent and inhabits hot and dry Mediterranean climates of California, South Africa, Portugal and Spain.
These examples show that the fundamental climate niche of a species is invariably much greater than the realised climate niche that can be discovered from the global databases. For predictions of climate suitability in natural environments this may not matter as natural processes still dictate species distribution, thus the realised niche is relevant. However, in urban centres all the processes that determine the ‘realised niche’ are inconsequential. Tree species are planted as 2-4 year old saplings and nurtured for a period time until they are established. The use of the Global Urban Tree Inventory database (GUTI, Ossola et al., 2020) is one step towards a better approximation of the fundamental niche, but it still has the same limitations. GUTI contains occurrence data of 4,734 tree species in 473 urban areas. However, if a tree species does not occur in a certain urban area, it does not mean that it cannot grow there. It just means that it has not been planted there, or that it was not recorded.
Issues and limitations of climate data
Another major limitation of climate niche analysis is the type of climatic data that are used. These are invariably mean climate data, such as mean annual temperature, mean annual precipitation, mean temperature of the driest month, or mean precipitation of the driest quarter. These values give an idea of the general climatic conditions, if the climate is warm, moist, if there are dry periods or frost periods. But these values are bad predictors of the climatic stresses or extremes that urban trees will experience. Trees will not be impacted or killed by mean annual temperature or mean annual precipitation. They are killed by long periods of drought, unseasonal periods of drought, heatwaves in the middle of a drought period, storms, floods or pest and disease attacks during periods of stress. None of these can or will be measured by the mean climatic parameters used in climate niche analyses.
Figure 3. Cumulative rainfall for the City of Melbourne (Australia) for all years since 2000. The graphs show a large interannual variation of rainfall, with a variation of 370 mm of annual rainfall in dry years (2019) and 840 mm of annual rainfall in wet years (2011). Data source: http://www.baywx.com.au/accumall.html
Even if mean climatic variables could be linked to tree performance in a meaningful way, they are not representative of the climate experienced by urban trees. The City of Melbourne in Australia has a mean rainfall of 650 mm according to the long-term average recorded in the main climate station near the city centre (Figure 3). However, the metropolitan area of Greater Melbourne experiences a mean annual rainfall of 450 mm in the west and over 1000 mm in the east. Thus, the actual variation of annual rainfall that tree species are potentially exposed to in Greater Melbourne can be much greater or smaller than the average. And within these values there is the year-to-year variation, which for the City of Melbourne ranges between 370 mm in dry years and over 800 mm of rainfall in wet years (Figure 3). Most urban trees tolerate this typical annual variation without any problems even though the mean annual rainfall of the extreme years would exceed the high or low thresholds (5th or 95th percentiles) of the species climatic niches.
Another issue is microclimate. The mean weather station of a city will not measure or detect the large variances that are created based on local microclimate. The central business district of many cities is often much warmer, has greater mean temperatures but also greater extreme temperatures due to the urban heat island effect. Contrary, trees in parks or near streams often experience much milder temperatures. In terms of rainfall, cities have many impervious surfaces, so much of the rainfall will not reach the root zone in densely built areas. On the other hand, cities also have additional sources of water that is accessible to trees, like leaking water pipes, or stormwater drains that get accessed by tree roots. The extent of this water access in most cities is unknown but also unquestionable. Thus, mean rainfall data are undoubtedly a very poor climate predictor for urban trees, as water access cannot be predicted for most urban trees.
Tree species issues
Most trees that are planted in urban areas have very little in common with their native counterparts. Many urban trees are hybrids, carefully bred for form and function to grow in the urban environment. They are often specific cultivars with limited or no genetic variability. The genetic source of trees is often unknown, or trees have been hybridised with other species so that there is no native equivalent (e.g. London plane, Platanus x acerifolia). Some urban trees are also grafted, where scion and rootstock are from different species or cultivars. Thus, it is difficult to assess to what degree the highly bred and cultivated urban tree would be represented by the wild-forms and therefore climate niche of the native trees of a species.
Tree stress response strategies
Climate envelope analysis in urban areas treats tree species equally in their response to climate and stress. As such, no allowances are made for diverse responses, adaptations, adjustments and acclimations. The climate niche method requires this over-simplification, but of course it is not based on reality. Tree species exhibit a large variety of responses and tolerances to stress, and these are difficult to predict and, in most cases, not well linked to climate parameters. The reason for this is that trees have evolved various mechanisms to drought or heat stress (Figure 4, (Levitt, 1980)). Some trees resist drought by avoiding drought stress through deep roots and access to deep soil moisture or groundwater. These trees maximise water uptake and are water spenders, but often have very vulnerable water conducting tissues. Others avoid drought stress through transpiration control; thus, these species minimize water loss and are water savers, and often have more vulnerable water conducting tissue. Other trees tolerate water deficit and drought stress by having more drought tolerant tissues and these species have less vulnerable water conducting tissues. Interestingly, species with various strategies can co-exist in the same environment (McDowell et al., 2008), although they can have different vulnerabilities. This highlights the flaw in using climate envelope as an indicator for tree vulnerability.
Figure 4. Schematic representation of the various strategies of plant responses to drought. One strategy of drought resistance is ‘escape’, where plants complete a life cycle when conditions are favourable, e.g. many desert annual herbs or geophytes. Drought ‘avoidance’ can be achieved by either ‘water spending’ where species with deep root systems access deeper soil moisture reserves and never experience drought. Eucalyptus camaldulensis, river red gum, is one example. Drought avoidance can also be achieved by ‘water saving’ or reductions of water loss, either by small leaves, carefully controlling transpositional water losses or by dropping leaves. Pinus edulis (pinyon pine, controls transpiration) or Platanus x acerifolia (London plane, drops leaves) are examples of this strategy. Plants that are drought ‘tolerant’ can withstand longer periods of drought by enduring water deficits thorough physiological or morphological traits or adjustments. Many eucalypts occurring in dryer environments, but also Callitris, have cavitation resistant tissues and are truly tolerating drought. Modified after (Levitt, 1980).
A climate niche analysis also assumes that trees are static in their climate responses and cannot acclimate. However, there is broad literature that many species can acclimate to higher or lower temperatures and e.g. adjust photosynthesis and respiration (Atkin and Tjoelker, 2003; Wang et al., 2020). This means that higher temperatures do not necessarily damage trees, in fact they can be advantageous. We assessed the temperature optimum of ecosystem photosynthesis (gross primary production) of different forest ecosystems in Australia using eddy covariance data from flux towers (Bennett et al., 2021). This showed that the ideal temperature for photosynthesis of most forests was well aligned with their mean annual temperature. However, the trees had a temperature buffer and could tolerate higher mean annual temperatures. We then modelled how these forests would perform under climate change to assess the climate vulnerability of each forest using the Australian CABLE model. Most forest ecosystems increased photosynthesis and biomass by 40% in a future climate by 2080 and only water limited Mediterranean woodlands showed lower increases or no change (Bennett, Arndt et al, unpublished data). However, no forest system was worse off, highlighting that the assumption of warmer climate equalling ‘risk’ or ‘bad outcomes’ is oversimplified and at times plain wrong.
Use climate niche modelling with caution
We caution against only using climate niche models for assessing tree species vulnerability and therefore future urban tree selection. As outlined above, the analysis is overly sensitive, and the applied vulnerabilities may be statistically justified but not biologically relevant. In fact, climate niche modelling can have negative consequences for tree species selection. Currently 21% of species in the city of Melbourne are assessed as potentially vulnerable to climate change (Kendall and Baumann, 2016). The species were labelled red, amber and green according to their vulnerability and this was taken literally by tree planners. One tree species that was tagged red (climate vulnerable) is Eucalyptus leucoxylon (yellow gum). The species is a drought tolerant woodland species that is outcompeted by faster growing eucalypts in wetter forests, resulting in a narrow climate niche. However, it is an ideal street tree: medium height, drought tolerant, robust. There is no evidence that it is vulnerable to climate change. However, the urban forest strategy of one Melbourne council labelled it drought and heat sensitive in its urban forest strategy (Council, 2017) and this species struggled to sell by commercial nurseries in Melbourne in subsequent years (H Mitchell personal communication). Thus, the outcome of climate niche analysis will likely lead to an over selection of species from warmer climates, while trees growing in colder climates will disappear from our cities. Many of these species will likely grow perfectly well in a future climate, but they are no longer planted. If there is scientific evidence that these ‘colder’ trees will perform badly in a warmer climate then by all means, deselect them from a future urban forest. However, there is still very little evidence to support the claim.
There are alternatives to climate niche analysis. Cities and urban planners can rely on the experience and expertise of their arborists and urban foresters as they assess the performance of urban trees on a regular basis. Expert opinion matters and in lieu of better scientific data would be preferrable to only relying on climate niche analysis data. It would be very powerful if urban tree assessments made by arborists in many towns and cities could be collected and archived in accessible databases so that data can be shared, similar to the way occurrence data is shared in the Global Biodiversity Information Facility (GBIF).
An alternate approach that relies on local tree assessment is the sister climate city analysis (Hancock, 2022). The urban forest of a city with a future analogue climate will be used to select tree species that are growing well in this future climate sister city. There are multiple climate analogue explorer tools for the US (US-Climate-Analogue-Explorer) or Australia (Climate-Analogues-Australia). However, this requires that each city has up-to-date inventories and a detailed health assessment of the tree species in that city.
Whilst there is a desire to develop easy ‘click and go models’ for selecting tree species for future climates in urban centres we caution against these models if they are only based on climate niche analysis. As we have outlined above the selection of a tree that is climate resilient is complex and depends on many variables. It depends on the tree’s drought and heat resistance strategy, its suitability as an urban tree, its ability to adjust or acclimate to different climate conditions. Much of this information is not known for most species, but much of it can be extracted from urban inventories. However, this legwork is required if urban foresters are to make good decisions for species selection for the future urban forest. Using simple climate niche models can be a part of the process in identifying widely distributed species – but not in assessing the proposed vulnerability of species.
We wish to thank Dave Kendall and Alessandro Ossola for many stimulating discussions on the topic of climate niche models, their advantages and limitations.