Evaluation and Comparison of Interpolation and Linear Regression Methods to Determine the Spatial Distribution of Precipitation in Chaharmahal and Bakhtiari Province, Iran

In the present study, simple and ordinary kriging methods, inverse distance and linear regression based on digital elevation model of the earth were evaluated for estimating annual rainfall using twenty-year statistics of precipitation data (1998-2018) in 33 rainfall stations in Chaharmahal and Bakhtiari province. For this purpose, first in ArcMAP, for each model in Kriging method, its variogram was calculated and using two-way evaluation technique, the error of the maps was estimated. The best method among geostatistical methods was conventional kriging method with Gaussian model. MAE, MBE and RMSE statistical indices for this method were 74.44, 0.48 and 93.72, respectively. Then, rainfall and altitude data of the stations were used using a linear regression model in Curve Expert environment. Finally, in order to determine the best model for spatial distribution of precipitation as well as comparing statistical and geostatistical methods, linear regression and ordinary kriging models were compared with each other and the MAE, MBE and RMSE statistical indices for regression method obtained were 115, 3 and 155, respectively. As a result, due to the accuracy, precision and error rate of the prepared maps, the most suitable method for interpolation of annual precipitation is the conventional kriging method with Gaussian model.