Refining predictions of climate change impacts on plant species distribution through the use of local statistics

Abstract

Bioclimate envelope models are often used to predict changes in species distribution arising from changes in climate. These models are typically based on observed correlations between current species distribution and climate data. One limitation of this basic approach is that the relationship modelled is assumed to be constant in space; the analysis is global with the relationship assumed to be spatially stationary. Here, it is shown that by using a local regression analysis, which allows the relationship under study to vary in space, rather than conventional global regression analysis it is possible to increase the accuracy of bioclimate envelope modelling. This is demonstrated for the distribution of Spotted Meddick in Great Britain using data relating to three time periods, including predictions for the 2080s based on two climate change scenarios. Species distribution and climate data were available for two of the time periods studied and this allowed comparison of bioclimate envelope model outputs derived using the local and global regression analyses. For both time periods, the area under the receiver operating characteristics curve derived from the analysis based on local statistics was significantly higher than that from the conventional global analysis; the curve comparisons were also undertaken with an approach that recognised the dependent nature of the data sets compared. Marked differences in the future distribution of the species predicted from the local and global based analyses were evident and highlight a need for further consideration of local issues in modelling ecological variables.

Introduction

One of the many concerns about climate change is its impact on living organisms. A focus of considerable recent attention has been the influence of changes in climate on the spatial distribution of species as well as its associated impacts on biodiversity and its conservation (Araujo et al., 2005, Akcakaya et al., 2006). From the range of approaches available for modeling the impacts of climate change on species distributions (e.g. Hamann and Wang, 2006, Austin, 2007, Botkin et al., 2007) considerable use has been made of approaches based on bioclimate envelope models (e.g. Berry et al., 2002, Pearson et al., 2006, Garzon et al., 2007).

Bioclimatic envelope models are typically based on the correlation between observed species distributions and climatic conditions. The approach is based on the ecological niche theory, often correlating the observed distribution of a species, its realised niche, with key climatic variables (Pearson and Dawson, 2003, Luoto et al., 2005). These models are attractive particularly as once a relationship has been established between the distribution of a species and a set of climate variables it may then be used to project the future distribution of the species under climate change scenarios (Pearson and Dawson, 2003, Luoto et al., 2007). The approach does, however, ignore important variables that can impact greatly on species distributions such as the effects of variation in inter-species composition and dispersal mechanisms as well as impacts of barriers to migration (Pearson et al., 2002, Pearson and Dawson, 2003, Brooker et al., 2007). The migration of some tree species, for example, is unlikely to keep pace with the change predicted by models as a function of climate change (Iverson et al., 2004). Additionally, the value of bioclimate envelope models often varies with scale. At local scales the models are often inaccurate and require the inclusion of additional variables, such as land cover, in order to provide accurate predictions of species distributions (Luoto et al., 2007). The predictive accuracy of models also often varies between species being, for example, more accurate for species with narrow rather than wide niches (Tsoar et al., 2007). Furthermore, many of the models used often fail to address important technical issues such as the impacts of spatial autocorrelation (Hampe, 2004, Dorman, 2007). Indeed, the importance of spatial issues is widely recognised with calls made for spatial models (Dorman, 2007).

Despite their simplicity and dependence on empirical relationships the bioclimatic envelope modelling approach can provide a valuable initial assessment of likely climate change impacts, especially if used at coarse spatial scales where macro-climate variation has the most impact on species distributions, making it a popular tool in macroecological research (Pearson and Dawson, 2003, Pearson and Dawson, 2004, Luoto et al., 2005). Used with due regard to their limitations, bioclimatic models can provide useful information for assessing the impacts of climate change on species distribution (Pearson et al., 2002, Heikkinen et al., 2006a). This article aims to show how the basic approach may be refined through the use of local statistical techniques and add to the rare literature providing empirical evidence of the value of bioclimatic envelope modelling (Araujo et al., 2005). A secondary aim is to highlight a concern with a widely used method for evaluating the significance of differences in model predictions. Model outputs are often evaluated and compared through evaluation of the area under the receiver operating characteristics (ROC) curve. In many studies the same sample of cases is used to derive the set of model outputs to be compared and so the common assumption of independent samples that underlies standard approaches is inappropriate. A variety of approaches for the comparison of correlated ROCs, based on a modification of the technique used widely, are, however, available (e.g. DeLong et al., 1988, Toledano, 2003, Wang and Gatsonis, 2007) and one is illustrated in the work reported.