Can climate data from METEOSAT improve wildlife distribution models?

Global climate change generated by human activities is likely to affect agroecosystems in several ways: reinforcing intensification in northern and western Europe, and extensification in the Mediterranean countries. If we are to predict the consequences of global warming for wildlife, distribution models have to include climate data. The METEOSAT temporal series from EWBMS offers an attractive alternative to using climatic surfaces derived from ground stations. The aim of this paper is to test whether this climatic satellite data can improve the distribution models obtained previously by Suárez-Seoane et al. using habitat variables for three agro-steppe bird species: great bustard, little bustard and calandra lark in Spain. Rainfall, radiation balance, evapotranspiration and soil moisture images were incorporated together with the other variables used as predictors in the published stepwise GAM models. Changes in the predicted distributions from the habitat only and climate-habitats models were assessed by reference to the CORINE land cover categories. Inclusion of climatic variables from METEOSAT led to statistically superior models for all three species. There were large differences in the climatic variables selected and the original variables dropped among the species. Evapotranspiration variables were the most frequently selected. Maps of the differences between the habitat and climate-habitat models showed very different patterns for the three species. Inclusion of climate variables led to a wider range of land cover types being deemed suitable. Despite the statistical superiority of models, care is needed in deciding whether to use climatic variables because they may emphasize the fundamental rather than the realized niche. Used together, however, habitat and climate models can provide new insights into factors limiting species distributions and how they may respond to climate change.