by Kyle Jensen
As arid ecosystems have been recognized as being especially sensitive to climate change, they thus provide an appropriate system to assess the use of SDMs in estimating the threat of climate change to various species. Species distribution models (SDMs) can quantify relationships between species and environmental factors, and use this data to predict spatial distributions. SDMs are thus widely used to derive projections of species distribution under conditions of climate change. These models are correlative however, and as such are unable to identify causal species-environment relationships. They can only be used as supporting evidence for an existing hypothesis on factors affecting species distribution; as such the factors must be chosen as inputs for the SDM to function. Identifying the important climatic factors involved in determining the range of a given species is a key factor in assessing the potential effects of climate change on species distribution and extinction risk. Little research however has been done investigating the effects using alternative sets of climate predictor variables may have on the projections of SDMs. Pliscoff et al (2014) seek to examine this area of potential uncertainty, addressing the potential variability of SDM spatial projections and determination of extinction risks through the creation and analysis of several sets of environmental predictors. They found that by adjusting climate predictor variables they were able to significantly affect predictions of spatial distribution as well as, for the first time, extinction risk estimates. This implies greater variability in such studies than previously thought.
The study modelled the present and future potential distributions of 13 species of the shrubby Heliotropium L. sect. Cochranea, a group with a range centered in the Atacama Desert of South America. They are well studied which means that their entire distribution can be recorded, making current and future projections possible. Presence data for these species dating back to 1950 were compared against several climatic factors which were measured monthly over the course of a year. Six sets of variables were then developed, and eight modelling techniques were applied to each set of variables as well as to each species, resulting in 624 models. The effect of a set of variables on predictive performance and spatial projection were evaluated using generalized linear mixed models (GLMM). Climate change scenarios based on the six sets of variables were used to obtain future projections of distribution. Extinction risks were evaluated following the recommendations of IUCN, whereby predicted areas of occupancy (AOO) were incorporated into several GLMM analyses. Both range shifts and extinction risk were compared among the sets of climatic variables.
The use of different sets of climatic variables in SDMs significantly affected projections of spatial distribution, while having little effect on the predictive power of a given model. These uncertainties in distribution affect climate change projections, which in turn affect the IUCN estimates of extinction risk. This is the first time that the choice of climatic predictor variables has been shown to have impacts on extinction risk estimates. These newly recognized uncertainties should be further explored, and ought to be taken into account in future SDM studies.
Pliscoff, P., Luebert, F., Hilger, H. H., & Guisan, A. 2014. Effects of alternative sets of climatic predictors on species distribution models and associated estimates of extinction risk: A test with plants in an arid environment. Ecological Modelling, 288, 166-177.