Bioclimatic mapping using geostatistical approaches: application to mainland Portugal

Abstract

AbstractSpatial interpolation of all the variables necessary for the bioclimatic classification of mainland Portugal, according to the system known as ‘Worldwide Bioclimatic Classification System’ (WBCS), was carried out. Such classification aims to establish a consistent relationship between worldwide vegetation types and easy‐to‐calculate bioclimatic indexes. Despite being relatively small, the Portuguese territory exhibits considerable climatic contrasts, ranging from mediterranean summer‐dry areas in the south, to hyper‐humid temperate mountains in the northwest. The territory has a somewhat sparse network of meteorological stations.In the scope of the WBCS, Compensated Thermicity Index, Annual Ombrothermic Index, Ombrothermic Index of the Summer Bimonth and of the Summer Trimester, Summer Compensated Ombrothermic Index and Positive Temperature were estimated throughout the study area using several interpolation techniques: multiple linear regression (MLR), ordinary kriging, kriging with external drift using three different drift models, and MLR followed by ordinary kriging of the residuals. The best geostatistical estimator for each variable was chosen taking into account both cross‐validation and visual inspection of the produced maps. The best estimates for the studied indexes were produced using multivariate geostatistical techniques: kriging with external drift (using multiple regression) and multiple regression followed by ordinary kriging of the residuals. Comparison among different geostatistical approaches for all interpolated indexes shows that in low‐sampling density areas, choosing the right set of predictor data is as important as choosing the right interpolation technique. Moreover, natural vegetation types were considered a reliable bioclimatic indicator—as an external criterion—for the evaluation of these maps, and confirmed these results.Finally, a bioclimatic diagnosis of the entire studied territory was accordingly derived by map algebra, from interpolations, resulting in two bioclimatic maps: those of ombrotypes and thermotypes. Copyright © 2009 Royal Meteorological Society