Aims: The availability of precipitation data plays an important role in many meteorological, hydrological and applications.
Materials & Methods: In this study, to improve precipitation maps and increase the accuracy of precipitation maps, linear regression, multivariate, and Kriging subsets were used. The data from 14 meteorological stations and IMERG images in the period of 20 years (2001 to 2020), digital elevation model, Latitude and Longitude maps were used. At first, based on regression in Minitab software, the relationship between air and ground parameters was taken. Finally, with the interpolation methods and based on the error coefficients, the best equations for predicting precipitation were determined and the spatial distribution of precipitation was obtained.
Findings: According to the results, six out of 13 models were selected because of low RMSE and high R2, R, and NS. In regression models where only one climatic or edaphic parameter was used, forecast accuracy was reduced. But in the models that were used in the regression elevation, Longitude, Latitude and IMERG parameters in combination with interpolation methods, the extracted data matched the real data with a slight difference. In this study, instead of the average of the input parameters, the maps of each parameter were used, increasing the accuracy of the forecast model to R2=0.8.
Conclusion: results showed that combining satellite precipitation products with interpolation methods led to a more accurate estimate of precipitation in the points without recording data will be precipitated and the multivariate regression method will be more accurate than the linear gradient.