Introduction

Accurate spatial estimation of Precipitation and temperature is very important in hydrological models. Despite the development of automatic meteorological stations in recent years, obtaining reliable climate data in data-deficient areas is still a big challenge. Spatial estimation of climatic data is mainly done by geostatistical methods and satellite images. Nowadays, satellite products are widely accepted in the preparation of climatic maps. These products use satellite images to estimate temperature and rainfall data in points without data, and usually the provided data is accompanied by errors and needs to be recalibrated. It seems that the combination of covariates and satellite products can be effective in increasing the accuracy of climatic maps, especially in areas with complex topography such as Mazandaran province.

Materials and Methods

In this research, the accuracy of temperature and rainfall products of TRMM satellite was evaluated in Mazandaran province and the possibility of combining them with land features of latitude, longitude and elevation in the form of regression model was investigated. In this regard, the monthly rainfall data of 21 meteorological stations and 48 monthly and 4 annual images of TRMM products in 2014 and 2017 were used. The evaluation indicators are root mean square error (RMSE), mean bias error (MBE), mean absolute percent error (MAPE). Also, the annual temperature and rainfall maps of the province were drawn by satellite products and modified method.

Results and Discussion

The results showed that the TRMM products have a huge bias error, so that the amount of annual rainfall bias in some years reaches more than 180 mm per year. About the temperature products the underestimation error is more than 2 Celsius degrees. The correlation coefficients of land features and temperature and precipitation data in most of the months provided acceptable results and were significant in 95% confidence level. In general, the relationship between monthly temperature and, latitude and TRMM products was significantly positive in all the investigated months. in the case of altitude, the relationship was negative and strong. But the relationship between temperature and longitude was a little weaker than other covariates. Regarding the precipitation variable, satellite products have a positive and significant relationship in all the investigated months, and the altitude has a negative effect on precipitation data except in the spring months, but the latitude has a positive relationship in the cold months and in the warm months has almost a negative relationship, and no specific seasonal trend was found in the case of longitude. Also, the correlation coefficients of TRMM products with temperature and precipitation data was significant in 100 and 77% of the months, respectively. Investigating the possibility of combining the TRMM products with the latitude, longitude and altitude in a form of regression equation to estimate temperature and precipitation data showed that the hybrid method increased the accuracy of satellite productions, impressively and the error of rainfall and temperature products reduced by about 30% and 70%, respectively. But as expected, the spatial estimation error of precipitation data was higher than temperature in all investigated months. The annual rainfall maps of Mazandaran for the years 2014 and 2017 shows the higher accuracy of the hybrid methods compared to satellite products. So that it shows well the rainy area of the west coastline and also depicts the meridional and altitudinal gradients of precipitation in the Mazandaran province as well. Examining the annual isothermal maps showed that the drawn map with the correction method has a significant difference with the temperature product of the TRMM satellite and has well highlighted the ring of cold areas of Alborz Mountain range and the foothills of Damavand and Alam-Kouh peaks. Also, the modified map has correctly distinguished the temperate coasts of the southern Caspian Sea from the Middle Band and Upper Band regions. In addition, the higher temperature of eastern half of Mazandaran compared to the western half has shown well.

Conclusion

The results of the present research showed that the TRMM temperature and rainfall products alone do not have proper accuracy in the spatial estimation of climate data and have a large bias error, but their combination as a covariate, along with Longitude, latitude and altitude in a regression equation, improved the accuracy of temperature and rainfall maps, and can be used as a new post-processing method in modification of satellite products.