Sensitivity of a semidistributed hydrologic model to rainfall estimation accuracy

Abstract

Radar based rainfall estimates, previously checked for radar artefacts, were merged with raingauge data to compute a series of hourly rainfall fields. These observed rainfall fields were sampled by the varied-density synthetic raingauge networks. A physically based semidistributed model (TOPMODEL) was used to simulate the resulting stream flow hydrograph. The basin response, based on the observed rainfall fields, was assumed to be the "ground truth', and other results were compared against it in terms of peak flow. The comparison showed that the space sampling error of precipitation caused inflated mean square errors of prediction in rainfall-runoff modeling. Furthermore, the temporal variability of the subbasins average precipitation was generally less than that of the computed rainfields based on a small number of gauges; this difference in variability propagates in rainfall-runoff modeling as a major term of the discrepancies. -from Authors