Predicting mammal species richness and distributions: testing the effectiveness of satellite-derived land cover data

Mapping species richness and distributions is an important aspect of conser vation and land use planning, but the time and cost of producing maps from field surveys is prohibitive. It is useful, therefore, if mappable environm ental variables, from a readily accessible source, can be used as surrogate s for species attributes. We evaluated the power of satellite-derived land cover information, from the Land Cover Map of Great Britain, to predict spe cies richness and occurrences of terrestrial mammals in one hundred 10 x 10 km quadrats, from four regions of Britain. The predictive power of the lan d cover data was relatively poor - with a few exceptions, land cover explai ned less than half of the variation in mammal species richness and occurren ce in regression models. Predictive power was considerably stronger when re gions were analyzed separately than when analyzed together, and best fittin g models varied between regions and between mammal taxa. Predictive power w as also affected (positively or negatively depending on taxon) when PCA-ord inated land cover variables were used as predictors. The predictive strengt h of the land cover data was probably limited mostly by the high proportion of British mammal species with geographic distributions changing rapidly a nd independently of land cover (and hence the non-saturation of preferred h abitats), and to a lesser extent by shortcomings in the mammal and land cov er data, and the influence of landscape factors other than land cover on ma mmal distributions. The results suggest that regional stratification is ess ential when attempting to predict species richness and distributions, even across relatively limited areas such as Great Britain. We conclude that cau tion is necessary in using results from environmental information systems s uch as this as a basis for conservation and land use planning decisions.