عنوان مقاله [English]
Background and objectives: The most common methods used to estimate and spatialise precipitation data are geostatistic methods and satellite images. Today, satellite rainfall products have become very popular in preparing rainfall maps. These products use satellite imagery to estimate precipitation in places with no observed data and are usually associated with a large error and require calibration. Interpolation methods also estimate precipitation data using recorded data. It seems that the combination of interpolation methods and satellite images can be effective in increasing the accuracy of rainfall maps, especially in areas with complex topography such as Mazandaran province
Materials and methods: In this study, in order to evaluate different methods of estimating precipitation in Mazandaran province and combining satellite images with interpolation methods, precipitation data from 21 synoptic and rain gauge stations and 24 monthly images and 2 annual TRMM satellite images were used in 2012 and 2015 that the spatial resolution of this satellite product is 0.25 * 0.25 degrees. The studied interpolation methods included Kriging and Inverse Distance Weighting. Also the accuracy of rainfall products of TRMM satellite was investigated. In addition, to increase the accuracy of rainfall maps, multiple linear regression were used to combine satellite images with latitude, longitude and altitude covariate. The investigated methods were evaluated using the Root Mean Square Error and Mean Bias Error indices and regression analysis. Also the annual rainfall maps of the province for 2012 and 2015 were drawn and analysed.
Results: In this study, 5 theoretical semivariogram models were fitted to the data, that the spherical and exponential models was selected as appropriate models. Also the coefficient of determination of selected variogram model and the ratio of structured part to total variation showed a relatively strong variography anlysis and the effective range of precipitation data wasobtained about 80 km. Correlation coefficients of covariates and precipitation in most months provided acceptable results and were significant in almost more than 50% of the studied months. As a result, the coefficients of determination of four-dimensional gradient regression equation also showed satisfactory values, so that the used covariates explained between 10 and more than 70% of the precipitation variations. evaluation of investigated methods showed that interpolation models and TRMM satellite network data are not efficient in estimation of precipitation in the province and the use of covariates in the gradient method could reduce the error estimation of rainfall data by 20 to 40 percent. Investigation of Bias error showed that TRMM precipitation network data, despite good correlation with observational data, has about 5 times more underestimation error than other interpolation methods, but the combination of TRMM network data with other covariates in the 4-dimensional gradient method has reduced the MBE to zero. Regression analysis of the studied methods showed a significant advantage of the 4-dimensional gradient method that the slope of this method is 3 times more than the geostatistical methods, which shows the performance of this method in detecting low and high rainfall rings in Mazandaran province.
Conclusion: The results showed the superiority of the 4-dimensional gradient method in spatial rainfall estimation of Mazandaran province and showed the role of covariates in increasing the accuracy of rainfall maps, that use of the selected method reduced the estimation error of geostatistical methods and TRMM network data by 30% and 40% respectively. The results of this study showed that the combination of satellite raifall products with interpolation methods will lead to more accurate estimation of precipitation in highlands and the points with no recorded rainfall data.