Comparative study of geo-statistical and multivariate models for air temperature interpolation in central and northern regions of Iran

Air temperature is one of the major variables required for agroclimatic classifications. For spatial zoning of temperature point observations, the interpolation approaches in which the horizontal and vertical gradients are included may be applied. In this research, the skill of Kriging, Co-Kriging, geographically weighted regression and Linear Multivariate Regression was evaluated for the interpolation of the monthly mean temperature values using the data of 56 synoptic stations located in the northern and central regions of Iran. The results of the statistical analysis indicated that the geographically weighted regression have the greatest difference with the other methods in month of December, with root mean square error (RMSE) equal to 0.83 °C, Based on the RMSE values of all months, the geographically regression method (with RMSE of 1.26°C) is the most suitable approach for temperature spatial zoning in this region. and then linear multiple regression method with RMSE of 2.24 °C, Kriging with RMSE of 2.52 °C and Cokriging with highest RMSE of 2.86 °C were ranked second to fourth, respectively. Besides, it is concluded that for high altitude areas where almost no weather station exist, the geographically weighted regression method provided the most accurate interpolated data of the air temperature.