A generalized additive model for the spatial distribution of stable isotopic composition in Antarctic surface snow

Using the most recent compilation of surface Antarctic snow isotopic composition data and a 1 km resolution digital elevation model (DEM), a generalized additive model (GAM) was used to develop a new gridded map of stable isotopic composition in Antarctic surface snow. The model explained 93.9% of the observed variance in the δ18O data, 92.7% of the observed δD variance and 66.6% of the observed deuterium excess. The performance of the model was evaluated by applying different quantitative approaches to the residuals from a cross-validation test. This modeling approach is a useful tool for interpolating local surface snow isotopic composition over Antarctica both in terms of the low estimated errors and the possibility of understanding the processes that control stable isotopic composition in Antarctic surface snow, through the response curves of each independent variable. We also presented and interpreted the resulting spatial variation patterns of δ18O, δD and deuterium excess to better use in many hydrological and paleoclimate applications.