An evaluation of using GPM satellite products to forecast flash floods in mountainous catchments

Sammanfattning: NASA's GPM satellite rainfall products IMERG early run and IMERG late run were evaluated for their applicability in flash flood forecasting. The selected catchment was the upper Han River basin, a mountainous catchment with low precipitation in winter and heavy precipitation during summers located in central China. The evaluated period spanned from 2015-03 to 2016-03 and is limited by the IMERG data history. The hydrological model that was used to simulate river floods was the THREW model. The IMERG data was first compared to data from rain gauges at an hourly 0.1x0.1 degree grid scale. The comparison criteria were correlation (r2 ), relative root mean squared error (RMSE) and mean bias (BIAS). The errors between IMERG and rain gauges were also compared against elevation. The results showed that the correlation varied from 0.3 to -15 (late run) and 0.33 to -21 (early run), relative RMSE varied from 730 % to 4100 % (early run), bias varied from -25 % to 370 % (early run) and -19 % to 410 % (late run), Then, using the bias factor correction method, the two IMERG data sets were corrected against the rain gauge data set. Simulations of river flows during 2015-03 to 2016-03 with the prepared precipitation data sets showed that the uncalibrated IMERG data performed with a Nash-Sutcliffe Model efficiency of -8 and -10 (early and late) while rain gauges performed 0.19. The calibrated IMERG performed -0.14 and -0.21 (early and late) which was a considerably improvement from the uncalibrated IMERG. Simulation results showed that while the near-real time products IMERG early and IMERG late have potential to estimate rain events that generate floods, calibration of the IMERG products are necessary in order to set reasonable results.