Spatial interpolation of the deuterium and oxygen-18 composition of global precipitation using temperature as ancillary variable

To aid our research in geographical forensic provenancing and food authentication we have developed high resolution prediction maps of the annual mean deuterium and oxygen-18 composition of modern precipitation. The maps have a spatial resolution of 10′ (∼ 20 × 20 km at the equator) and cover the global land surface excluding the Antarctic. To achieve this, the relation between various temperature related variables and the isotopic composition of modern precipitation was explored using a combination of high resolution climate maps and global isotope records from the GNIP database. This revealed that the isotopic composition of precipitation is somewhat better correlated to the temperature during the coldest – often driest – period of year than the temperatures during the warmest – often wettest – period of year (especially below 0 °C). Although the reason for this effect is not directly clear, the temperature during the coldest quarter is used as ancillary variable in simple kriging with varying local means (SKlm). In SKlm, only the residual isotope values from the regression with T_cq are kriging interpolated, which are then added to the predicted isotope map based on T_cq. Because the ancillary variable explains the bulk of the isotopic variation (R² = 0.79–0.85), the deuterium and oxygen-18 maps mainly reflect the large scale global temperature pattern. More local isotope effects are accounted for by the interpolation of the residual values. This study furthermore shows that surface temperature better explains the global isotopic variation compared to a combination of latitude and altitude (R² = 0.68–0.69). Yet, at very low temperatures (< − 40 °C) our maps might underestimate the true isotope signal. The new isotope maps and the maps of the 95% confidence intervals can be downloaded from www.waterisotopes.org.