Topography modulates air and soil temperature and gives rise to local climates that may strongly deviate from regional macroclimates. These local climatic conditions determine the distribution of many organisms but are hardly accounted for by coarse-grained macroclimate grids. Downscaling macroclimate grids or interpolating high-resolution climate surfaces from weather station data has thus become a staple in ecological research. Against this background, topographically complex territories pose a challenge, as downscaling and interpolation techniques are susceptible to high topographic heterogeneity. Simultaneously, these regions will act as refugia and preserve biodiversity amidst the challenges of climate change. This contrast renders choosing the technique to derive local climatic conditions under high topographic complexity critically relevant.In this study, we compared how interpolation and downscaling techniques predict the local climates of topographically heterogeneous territories. We interpolated monthly average temperatures from weather stations in a Southern Italy mountain region and downscaled WorldClim and CHELSA macroclimatic grids from their native 30 arcsecs to 10 m resolution. We carried out the interpolation and downscaling procedures by employing four commonly used statistical algorithms and by including physiographic descriptors (altitude, northness, eastness, distance to the coast, and monthly average of daily clear-sky insolation time) known to determine local climatic conditions. We compared the techniques' predictive performance via leave-one-out cross-validation. We extended the comparison over the study area to identify where the techniques most strongly diverged and evaluated how interpolation and downscaling techniques fare against environmental extrapolation.Although the interpolations scored the best values across all error metrics, interpolating weather station data and downscaling WorldClim yielded essentially equivalent predictions of local climates in the cross-validations. Differences between the two techniques emerged during the spatial comparison, where downscaling WorldClim produced biased and less detailed monthly climate surfaces. On the contrary, weather station interpolation was affected by environmental extrapolation in the study area's most internal and topographically heterogeneous landscapes. Downscaling CHELSA severely over-smoothed local climatic conditions during both cross-validations and spatial comparisons, possibly due to lacking climate variability in the native product itself.We demonstrated that technique and data source choice affect the prediction of local climates in topographically heterogeneous territories. Despite its limitations, we argue that interpolating weather station data better accounted for local-scale topographic effects. These effects form an integral part of the microrefugia that will shelter mountain communities from warming climates. We thus recommend carefully considering the techniques chosen to unravel the intricate climatic mosaics of topographically heterogeneous territories.
Read full abstract