Quantify the Error Propagation of Satellite Precipitation by a Hydrologic Analytical Framework

The integration of satellite precipitation in hydrologic simulation provides a potentially viable solution in regions with scarce or no ground-based measurement networks. However, given the different physical principles in retrieval algorithms and different combination of sensors, the performance of satellite precipitation product in flood modeling is lacking consistency. Aiming at understanding the error propagation of satellite precipitation in hydrologic simulation, we propose the use of an analytical framework to transform the error in precipitation to error in catchment flood response. The analytical framework, originally developed to characterize the rainfall-runoff generation process at basin scale, predicts the difference in hydrograph properties (i.e. cumulative volume, centroid and dispersion) between the satellite- and reference-driven simulations by bias in catchment-average cumulative rainfall excess, mean and variance of catchment response time. Results show that under the smooth topographic setup of the study area, the bias in spatial and temporal correlation between rainfall and runoff generation are insignificant to difference in cumulative volume of flood event; difference in flood event centroid due to bias in delay from the runoff generation and runoff routing are of equal importance; difference in dispersion of flood event hydrograph is mainly controlled by the bias in the variability of runoff generation time. Sensitivity tests show that the analytical framework is able to reproduce the difference in cumulative volume of flood events. Meanwhile, the main weakness exhibited for the analytical framework is its increased random error in estimating the difference in flood event centroid and dispersion; this random error is shown to decrease as basin area increases.