Monthly Air Temperatures over Northern China Estimated by Integrating MODIS Data with GIS Techniques

Abstract

AbstractThe Three-North Shelter/Protective Forest Programme (TNSFP), the largest ecological afforestation program in the world, was launched in 1978 and will last until 2050 to improve ecological conditions in the Three-North regions of China. To manage the shelter forests sustainably, it is necessary to accurately estimate air temperature on a large scale, but the spatial distribution of ground meteorological stations is limited. A hybrid method was established by combining stepwise regression modeling and spatial interpolation techniques (SRMSIT) to construct the monthly mean, minimum, and maximum air temperatures (Tmean, Tmin, and Tmax, respectively) at a 1 km × 1 km grid size in the Three-North regions. Stepwise regression modeling was applied to construct the relationship between air temperatures (Tmean, Tmin, and Tmax—the dependent variables) and geographical and Moderate Resolution Imaging Spectroradiometer (MODIS) variables (the independent variables). Spatial interpolation techniques were used to correct the residual values. According to the factor analysis, three geographic (altitude, latitude, and continentality) and two MODIS variables [nighttime land surface temperature (LST) and normalized difference vegetation index] were selected in stepwise regression modeling, and nighttime LST was the most powerful remote sensing variable. The SRMSIT method, in which the spatial interpolation of the residuals was done with inverse distance weighting, achieved average root-mean-square error values at 0.86°, 1.10°, and 1.13°C for Tmean, Tmin, and Tmax, respectively. Therefore, the simple regression algorithms derived from the combination of remote sensing and geographical variables, together with residual interpolation techniques, have the potential to accurately estimate monthly air temperature in large regions.