Species distribution modelling tools and databases to assist managing forests under climate change
Introduction
Managing trees under climate change is a great challenge for forestry in the present century (Millar et al., 2007). It will involve adjustments in ecological, social and economic systems (Spittlehouse and Stewart, 2003). These changes will need to be evidence-based and underpinned by appropriate models and data. Immense progress has been made in developing tools and databases to support forestry decision-making under climate change, but major challenges remain.
The aim of this review is to provide an overview of some of the key recent advances relevant to determining where particular tree species may be able to grow in either natural forests or plantations under current and future climatic conditions. The focus is mainly, but by no means entirely, on species of the genus Eucalyptus as they have advantages for climate change studies. For example, eucalypts have very poor dispersal capabilities (Booth, 2017a), so the survival of existing stands is the key issue for the present century. The great complications of assessing possible dispersal to and establishment at other sites are avoided. Eucalypt species and provenances have also been extensively tested outside their natural distributions, so knowledge of their climatic adaptability beyond their natural distributions is good at least for the commercially important species (Jacobs, 1981, Booth et al., 2015).
There has been a great increase in species distribution modelling (SDM) since the first package called BIOCLIM became available in 1984 (Nix, 1986, Booth et al., 2014). For example, Dyderski et al. (2018) identified 124 papers from the Web of Science published between 1996 and early 2017 that analysed more than 500 tree species distributions and produced projections based on at least one climate change scenario. Most of these papers focussed on threatened, invasive or plantation species. The contents of these papers, including abstracts and ten important features, were summarised in an Excel spreadsheet available as an appendix to the Dyderski et al. (2018) paper. For example, the geographic spread of the studies was as follows: Global (2), Africa (8), Asia (36), Australia (4), Europe (32), North America (38) and South America (4). Though widely used the SDM approach is not without some limitations for examining how tree species will respond to climate change (Keenan, 2015). However, these limitations may be less for eucalypt species than many other tree species (see Table 1).
If they are available the use of more detailed process-based models may be desirable for studying climate change impacts. But presently their applications are confined to a very small number of relatively well-known species of commercial importance (see, for example, the climate change analyses of E. globulus and E. nitens plantations by Battaglia et al., 2009, Battaglia and Bruce, 2017). In contrast, three SDM studies have examined the likely impacts of climate change on many Eucalyptus and closely related Corymbia species (which were previously classified as eucalypt species). These studies considered the likely impacts on 819 species (Hughes et al., 1996), 108 species (Butt et al., 2013) and 657 species (González-Orozco et al., 2016). The Hughes et al. (1996) paper did not project future distributions using a climate change scenario, so it was not listed by Dyderski et al. (2018), though the Butt et al. (2013) paper was listed. The González-Orozco et al. (2016) analysis projected future distributions and fell within their assessment period, but was missed by Dyderski et al. (2018). Including this analysis would have increased the number of tree species distributions assessed under climate change to over 1000. There are no other broad-scale analyses of the likely impacts of climate change on most eucalypt species using alternative approaches. Accordingly, as SDMs are the only source of climate change impact information for many eucalypt and other tree species it is worthwhile to recognize their limitations, but also to consider how they may provide useful information.
In the latest SDM study of eucalypt distributions González-Orozco et al. (2016) made it clear that they were aware of concerns about the ‘equilibrium assumption’. This implies that analysis of just the current natural distribution provides a reliable indication of a species climatic requirements under climate change. Various authors have pointed out the need to analyse not only the natural distribution, but also occurrences outside the natural distribution (see Booth, 2017b for a review of the ‘equilibrium assumption’). González-Orozco et al. (2016) attempted to address the issue by using models which avoid ‘over-fitting’ (see Booth, 2017c for a discussion of some of the limitations of the González-Orozco et al., 2016 analysis and how they might be overcome). Despite some limitations the paper provides much useful information, including maps of projected responses for the 657 species as part of the supplementary information. An advantage of the approach is that it can be applied to large numbers of species. A disadvantage, in comparison to use of data from both the natural distribution (i.e. realized niche sensu Hutchinson, 1957) and trials outside the natural distributions (c.f. fundamental niche sensu Hutchinson, 1957), is that it assumes all variables have climatic adaptability beyond the realized niche. The analyses in Booth et al. (1988) suggest that this is not the case, but more work is needed examining species climatic responses in locations outside their natural distributions.
The requirements for using SDM techniques to deliver “a global climatological audit to assist conservation and sustainable development” in forests under climate change were outlined by Booth (1991). The paper identified five key objectives related to climatic data, species distribution data, species and provenance trial data, descriptions of species climatic requirements and mapping of climatically suitable areas. These topics are used here as section headings. Within each section, progress is described and then desirable future objectives are outlined. An extra section has been added after climatic data to consider soil data. The biodiversity database known as the Atlas of Living Australia (ALA) has integrated elements of the five key objectives into one unified system freely available on the internet (spatial.ala.org.au) and this is described in a separate section.
Section snippets
Climatic data
The outstanding achievement in relation to global climatic data has been the development of the WorldClim database (Hijmans et al., 2005). This has recently been updated with WorldClim 2 (Fick and Hijmans, 2017) providing improved climatic databases for land areas of the world. These were developed by applying thin plate spline methods (Hutchinson, 1991, Hutchinson, 2013), which were originally used to prepare Australian interpolation surfaces for the BIOCLIM package. The WorldClim 2 website (//worldclim.org/version2
Discussion
A search on Google Scholar indicates that the uptake of SDM modelling techniques generally has been dramatic in recent years with many hundreds of studies completed in a wide range of terrestrial and marine environments (see Fig. 1). A number of key SDM-related issues have been identified including the use of presence only or presence/(pseudo-)absence data, errors and uncertainties associated with SDMs (Gould et al., 2014) and variable selection (Porfirio et al., 2014). These issues, and other
Acknowledgements
Funding for this review was provided by CSIRO. I am grateful for the helpful comments of Stephen Roxburgh, David Bush and Sadanandan Nambiar as well as the anonymous reviewers. Thanks also to Elisabeth Bui for providing information about SoilGrids250m database.
References (76)
- et al.
Assessment and prioritisation of plant species at risk from myrtle rust (Austropuccinia psidii) under current and future climates in Australia
Biol. Conserv.
(2018) Using biodiversity databases to verify and improve descriptions of tree species climatic requirements
For. Ecol. Manage.
(2014)- et al.
Niche analysis and tree species introduction
For. Ecol. Manage.
(1988) - et al.
Native forests and climate change: Lessons from eucalypts
For. Ecol. Manage.
(2015) - et al.
Advances in monitoring and modelling climate at ecologically relevant scales
Adv. Ecol. Res.
(2018) - et al.
The biodiversity and climate change virtual laboratory: where ecology meets big data
Environ. Modell. Softw.
(2016) - et al.
A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning
For. Ecol. Manage.
(1997) - et al.
Mapping rootable depth and root zone plant-available water holding capacity of the soil of sub-Saharan Africa
Geoderma
(2018) - et al.
Maximum entropy modelling of species geographic distributions
Ecol. Model.
(2006) - et al.
A computerised system for matching climates in ecology
Agr. Ecosyst. Environ.
(1985)
Impact of landscape predictors on climate change modelling of species distributions: a case study with Eucalyptus fastigata in southern New South Wales, Australia
J. Biogeogr.
Direct climate change impacts on growth and drought risk in blue gum (Eucalyptus globulus) plantations in Australia
Aust. Forestry
Identifying particular areas for potential seed collections for restoration plantings under climate change
Ecol. Manage. Restor.
Going nowhere fast: a review of seed dispersal in eucalypts
Aust. J. Bot.
Assessing species climatic requirements beyond the realized niche: some lessons mainly from tree species distribution modelling
Clim. Change
Impacts of climate change on eucalypt distributions: an examination of a recent study
Aust. Forestry
Why understanding the pioneering and continuing contributions of BIOCLIM to species distribution modelling is important
Austral Ecol.
BIOCLIM: the first species distribution modelling package, its early applications and relevance to most current Maxent studies
Diversity Distrib.
TREDAT: A tree crop database
Aust. Forestry
BIOCLIM – a bioclimate analysis and prediction system
Plant Protect. Quart.
Eucalypts face increasing climate stress
Ecol. Evol.
Phenology is a major determinant of tree species range
Ecol. Lett.
How much does climate change threaten European forest tree species distributions?
Glob. Change Biol.
Eucalypt Domestication and Breeding
Novel methods improve predictions of species’ distributions from occurrence data
Ecography
Elevated CO2 does not increase eucalypt forest productivity on a low-phosphorus soil
Nat. Clim. Change
Global planted forests thematic study: results and analysis
WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas
Int. J. Climatol.
Mapping Species Distributions: Spatial Inference and Prediction
A tool for identifying potential Eucalyptus nitens seed orchard sites based on climate and topography
South Forests
Species distribution modelling: contrasting presence-only models with plot abundance data
Sci. Report
Phylogenetic approaches reveal biodiversity threats under climate change
Nat. Clim. Change
A tool for simulating and communicating uncertainty when modelling species distributions under future climates
Ecol. Evol.
Habitat Suitability and Distribution Models with applications in R
Cited by (80)
Climate change drives contrasting shifts in fish species distribution in the Mekong Basin
2024, Ecological IndicatorsSystem dynamics modelling to simulate regional water-energy-food nexus combined with the society-economy-environment system in Hunan Province, China
2023, Science of the Total EnvironmentPrediction of wild pistachio ecological niche using machine learning models
2022, Ecological InformaticsCitation Excerpt :Literature reviews showed that many environmental variables may be required to develop the habitat suitability models (HSMs) in SDM studies (Abdelaal et al., 2020; Gilani et al., 2020; Rahmanian et al., 2021). There are several modeling techniques available to predict habitat suitability (HS) for tree species (Booth, 2018; Chaves et al., 2022; Koch et al., 2022). Machine learning models such as maximum entropy (MaxEnt), generalized linear model (GLM), flexible discriminant analysis (FDA), support-vector machine (SVM), multivariate adaptive regression splines (MARS), boosted regression trees (BRT) and random forest (RF) have been commonly used in modeling studies due to their ability to work with limited data by implementing powerful pattern recognition algorithms (e.g., Deb et al., 2017; Hosseinzadeh et al., 2018).