First results from Version 7 TRMM 3B43 precipitation product in combination with a new downscaling–calibration procedure

Accurate precipitation data at high spatial and temporal resolution is deemed necessary for many hydrological and water management applications, and especially in data scarce river basins and regions where strong competition for water resources prevails. In this study we used a new downscaling–calibration procedure of the freely accessible Version 7 of TRMM (Tropical Rainfall Measuring Mission) 3B43 product, in conjunction with limited rain gauge data sets, to generate improved monthly pixel-based precipitation data at higher spatial resolution (1 km). The spatial downscaling from 0.25° to 1 km grids was achieved by using site-specific non-linear relationships between annual precipitation and annually averaged NDVI (Normalized Difference Vegetation Index). The calibration was based on Geograhical Difference Analysis (GDA) and Geographical Ratio Analysis (GRA). The new integrated procedure was tested for Lake Tana Basin (LTB) in Ethiopia with a humid climate (rainfall 1395 mm/yr during 1998–2004) and for the Caspian Sea Region (CSR) in Iran with a semi-arid climate (rainfall 442 mm/yr during 1999–2003). The best 1 km annual precipitation data were achieved through downscaling followed by GDA calibration for most cases. The monthly fractions derived from the un-calibrated TRMM 3B43 product can be used to disaggregate 1 km annual precipitation to 1 km monthly precipitation. The disaggregated 1 km monthly precipitation has not only significant improvement in the spatial resolution, but also good agreements with rain gauge data were achieved for both LTB (R2 = 0.87, RMSE = 56 mm, MAE = 32 mm and Bias = 0.01) and the CSR (R2 = 0.79, RMSE = 23 mm, MAE = 16 mm and Bias = 0.14). A similar calibration procedure using rain gauges at monthly time scale did not improve the level of performance.