全球变化知识资源中心

The effects of climate change on biodiversity are increasingly well documented, and many methods have been developed to assess species' vulnerability to climatic changes, both ongoing and projected in the coming decades. To minimize global biodiversity losses, conservationists need to identify those species that are likely to be most vulnerable to the impacts of climate change. In this Review, we summarize different currencies used for assessing species' climate change vulnerability. We describe three main approaches used to derive these currencies (correlative, mechanistic and trait-based), and their associated data requirements, spatial and temporal scales of application and modelling methods. We identify strengths and weaknesses of the approaches and highlight the sources of uncertainty inherent in each method that limit projection reliability. Finally, we provide guidance for conservation practitioners in selecting the most appropriate approach(es) for their planning needs and highlight priority areas for further assessments.

The Earth has warmed by about 0.74 °C in the last 100 years, and global mean temperatures are projected to increase further by 4.3 ± 0.7 °C by 2100¹. Agricultural expansion, over-exploitation and introduction of invasive alien species have been the main drivers of biodiversity loss in the recent past, but several lines of research suggest that climate change could become a prominent, if not leading, cause of extinction over the coming century², both via direct impacts on species and through synergies with other extinction drivers^{1, 3}. Species have already responded to recent climatic shifts^{4, 5, 6, 7, 8}, and various attempts have been made to assess the potential risks to biodiversity posed by climate change over coming decades^{9, 10, 11}.

To assess the threats to a species posed by climate change one must have information regarding its vulnerability, which is defined by the IPCC as 'the predisposition to be adversely affected'¹². Although there is currently no broad consensus in the scientific literature regarding the definition of 'species' vulnerability', it is generally accepted that this is a function of both intrinsic and extrinsic factors¹³, and assessments often consider exposure, sensitivity and adaptability in combination^{13, 14}. Exposure is the magnitude of climatic variation in the areas occupied by the species¹⁵. Sensitivity, which is determined by traits that are intrinsic to species, is the ability to tolerate climatic variations, while adaptability is the inherent capacity of species to adjust to those changes^{14, 15}. Attempts at projecting the effects of climate change on species have used both different currencies (that is, the range of measures used to assess species' climate change vulnerability) and divergent approaches for identifying the most vulnerable taxa. Because of this lack of consensus by the conservation community, a formal comparative evaluation is necessary to guide sensible choices of the most appropriate technique(s) for assessing species' vulnerability.

Here we provide the first comprehensive review of currencies and approaches that have been used to assess species' vulnerability to climate change, based on a total of 97 studies published between 1996 and 2014 (with >70% of the studies published during the past five years). We describe the four dominant currencies of species' climate change vulnerability assessments and provide examples of how these have been applied. Three broad categories of approaches plus three combinations thereof were identified, and we describe each, examining how they address uncertainties, and discuss their key limitations. Finally, we provide guidance for practitioners. Through these analyses, we aim to help conservationists select appropriate approaches for assessing species' vulnerability, such that climate change adaptation responses are as solidly based as possible.

We conducted a systematic literature search using ISI Web of Knowledge. Keywords were selected to identify studies on climate change (climate change^*, global warming^*, sea-level rise^*, elevated CO₂^*, drought^*, cyclones^*, CO₂ concentration^*) impacts (population reduction^*, range changes^*, range shift^*, turnover^*, extinction risk^*, extinction probability^*) that led to vulnerability assessments (vulnerability^*, sensitivity^*, adaptability^*, exposure^*) based on different types of approaches (mechanistic^*, SDM^*, correlative^*, trait-based^*, criteria^*, niche models^*). We then selected the most representative papers (in terms of both spatial and temporal scales, and taxa). Studies differed widely in taxonomic coverage, birds being the most frequently considered taxon, followed by mammals and plants, while non-insect invertebrates were seldom assessed (Fig. 1). Additionally, spatial scales of application and authors' interpretations of the concept of vulnerability varied extensively. More than 60% of the studies were developed at local scale, while only 4% of the papers assessed species' vulnerability globally (Fig. 1). As a result, numerous species have been assessed in only part of their range and their estimates of vulnerability may therefore be unrealistic.

Birds are the most analysed taxon, followed by mammals and plants, while invertebrates other than insects have seldom been assessed. Colours represent the spatial scale of the assessments. Regional scale is defined as describing the range of 10⁴−10⁷ km², while scales smaller than 10⁴ km² are referred to as local scales.

There is no standard way to assess a species' vulnerability to climate change, and the type of information needed (for example, range extent, population size) will determine which approaches are most appropriate.

Distributional changes. To assess the impacts of climate change on species, current and future distributions can be projected using either mechanistic or correlative niche models (both approaches are discussed below), which relate environmental conditions to species' physiological responses or occurrence data, respectively. Several analyses have provided examples of species likely to suffer range reductions in the twenty-first century^{16, 18}. For example, Vieilleident et al.²⁷ predicted that the Malagasy baobab Adansonia suarezensis is likely to go extinct before 2080 owing to an overall loss in suitable habitat. Changes in range size have usually been assessed by considering the climatic characteristics of current distributions and the projected distribution of these climatic conditions in future^{27, 28}. However, vulnerability might be exacerbated by other factors, including biotic interactions, reduced adaptive evolutionary response and dispersal ability. Several studies have incorporated dispersal ability into predictions of future range changes, by contrasting scenarios of no dispersal with unlimited dispersal^{29, 30, 31}, by estimating average or maximum potential dispersal distances^{16, 18, 24}, or by explicitly simulating metapopulation dynamics including dispersal events^{32, 33}. For example, Schloss et al.¹⁸ suggested that 87% of Western Hemisphere terrestrial mammals will probably experience a reduction in their climatically suitable area, with 20% of these species being particularly vulnerable due to their limited dispersal ability.

Population changes. A different set of modelling approaches uses predictions of population trends to inform risk assessments³⁴. Quantified population changes can be based on direct observations, indices of abundance^{34, 35, 36}, reporting rates used as proxies for abundance³⁷, or they can be inferred from declines in extent of occupied or suitable habitat^{34, 38}. Examples of population changes that have been observed over recent decades include declines in long-distance avian migrants to Dutch forests; these declines have likely been driven principally by temperature changes in spring³⁵. Also, a decrease in ice coverage has led to a reduction in polar bear (Ursus maritimus) numbers in the southern Beaufort Sea³⁹. Some approaches to projecting future population sizes incorporate past population trends into mechanistic models^{39, 40, 41}, and consider the effects of changes in model parameters (for example, distribution patterns, life history, climatic conditions). This type of approach has also been applied to a population of American marten (Martes americana) in North America, where explicit population models simulated a 40% decline in the population due to climate change by 2055⁴².

Extinction probability. One synthesis estimated that between roughly 20 and 30% of species assessed are likely to be at increasingly high risk of extinction in the face of increasing global warming¹². Extinction probability has been calculated for populations of species with known life-history characteristics, such as the emperor penguin (Aptenodytes forsteri)⁴¹, Arizona cliffrose (Purshia subintegra)⁴³, spring–summer chinook salmon (Oncorhynchus tshawytscha)⁴⁴ and polar bear (Ursus maritimus)³⁹, by using population viability analyses^{41, 43}, demographic models^{39, 44, 45}, or evolutionary models⁴⁶. These methodologies combine population fluctuations with changing environmental parameters in order to estimate extinction probability within a given time interval. For example, Fordham et al.⁴⁵ modelled the predicted abundance of the Iberian lynx (Lynx pardinus) under three climate scenarios by integrating temperature and precipitation data, prey availability and management interventions, and predicted that climate change may drive this species to extinction within the next 50 years. This work relied upon a thorough understanding of the species' biology and of demographic dynamics related to extinction risk. However, as most species lack such detailed data, extinction risk due to climate change tends to be quantified only for better-known species.

Vulnerability indices and other relative scoring systems. Vulnerability indices are quantitative indicators of the relative vulnerability of species. The data derived from the currencies discussed above, and from trait-based vulnerability assessments (TVAs), can be used to obtain scores¹⁴, categories³⁴ or indices⁴⁷, which are often easier for scientists and practitioners to interpret and use to identify species at risk within their focal areas. Foden et al.¹⁴, for example, classified birds, amphibians and corals into two vulnerability categories (low or high). One limitation of indices and scores is that they do not provide any direct measures of the expected impact on species, that is, they are not expressed in terms of any of the currencies otherwise used to assess species' vulnerability (for example, range reductions, extinction probability, population decline).

Different approaches are used to assess species' vulnerability to climate change. These approaches can be placed in four classes: correlative, mechanistic, trait-based, and combined approaches.

Correlative approaches. Distributional changes are typically estimated through the use of correlative models that aim to represent the realized niche of a species^{48, 49}. Correlative models relate observed geographic distribution of a species to current climate; resultant models are then applied to climate projections to infer potential climatically suitable areas for a given species in the future. Species' distribution can be presence-only data^{17, 22}, presence/absence⁵⁰ or abundance observations⁵¹, based either on fieldwork or specimen records^{22, 52}. Correlative models have been applied to species at scales ranging from local to global^{19, 53} (Fig. 1), and have been widely used to explore the vulnerability of vertebrates (including birds^{36, 52, 54}, mammals^{17, 28, 38}, amphibians^{30, URL:}