Comparative evaluation of statistically downscaling Tropical Rainfall Measuring Mission (TRMM) and Global Precipitation Measurement (GPM) mission precipitation data: evidence from a typical semi-arid to arid environment

Abstract

This study aimed at comparing the performance of statistically downscaling Tropical Rainfall Measuring Mission (TRMM) 3B43 Version 7 and Global Precipitation Measurement (GPM) 3IMERGM Version 6 precipitation data from 25 km and 10 km resolutions, respectively, to 1 km resolution using Moderate Resolution Imaging Spectroradiometer (MODIS) MOD13A3 Version 6 Normalized Difference Vegetation Index (NDVI) data. Simple linear regression (SLR) and geographically weighted regression (GWR) were applied in a typical semi-arid to arid country—Jordan—using data for the year 2015. It was observed that all statistically downscaled outputs exhibited almost similar spatial distribution patterns. These patterns captured the typical decreasing behavior of precipitation in the country from northwest to south and east. They also provided more spatial detail of precipitation information than their raw counterparts. Unscaled and scaled validation measures of bias, precision, and accuracy relied on precipitation data for the year 2015 obtained from 137 rain gauges distributed all over the country revealed that: (1) GWR produced better results than SLR, (2) statistically downscaling GPM 3IMERGM Version 6 precipitation data outperformed statistically downscaling TRMM 3B43 Version 7 precipitation data, and (3) although, all statistically downscaled precipitation outputs overestimated rain gauge precipitation data, this overestimation was larger in the statistically downscaled TRMM 3B43 Version 7 precipitation outputs than in the statistically downscaled GPM 3IMERGM Version 6 precipitation outputs.