A new interpolation method for Antarctic surface temperature

We propose a new methodology for the spatial interpolation of annual mean temperature into a regular grid with a geographic resolution of 0.01° for Antarctica by applying a recent compilation of the Antarctic temperature data. A multiple linear regression model of the dependence of temperature on some geographic parameters (i.e., latitude, longitude, and elevation) is proposed empirically, and the kriging method is used to determine the spatial distribution of regional and local deviations from the temperature calculated from the multiple linear regression model. The modeled value and residual grids are combined to derive a high-resolution map of surface air temperature. The performance of our new methodology is superior to a variety of benchmark methods (e.g., inverse distance weighting, kriging, and spline methods) via cross-validation techniques. Our simulation resembles well with those distinct spatial features of surface temperature, such as the decrease in annual mean surface temperature with increasing latitude and the distance away from the coast line; and it also reveals the complex topographic effects on the spatial distribution of surface temperature.