Geographically Weighted Regression: A Method for Mapping Isohyets in Gilan Province

So far several methods have been developed for mapping and interpolation of isohyets. one of the recently accepted methods is geographically weighting regression which is suitable for evaluation of spatial heterogeneity of dependent variable by using local regressions. In order to evaluate annually precipitation spatial variation، this study was conducted in Gilan province which precipitation is distributed non-uniform due to different environmental conditions. The results of geographically weighting regression method were compared with another interpolation methods including global polynomial، local polynomial، inverse distance weighting (IDW)، spiline، kriging and co-kriging and. In this study، average of 20 years annually precipitation data of 185 meteorological observations over Gilan Province and its neighboring stations was used for modeling of spatial distribution variations of mean annual precipitation by using other variables like elevation and position of points to the sea level. Cross validation technique was used to assessment accuracy of each interpolation methods. The result showed that geographically weighting regression method had minimum error with RMSE=147 and had significant difference with the kriging method which was in the second rank with RMSE=187. Finally the best method for mapping isohyets in Gilan province is geographically weighting regression method.