Long-Term Historical Rainfall-Runoff Modeling Using High-Resolution Satellite-based Precipitation Products

Monday, 15 December 2014: 10:35 AM
Hamed Ashouri, Phu Nguyen, Andrea R Thorstensen, Kuo-lin Hsu and Soroosh Sorooshian, University of California Irvine, Civil and Environmental Engineering, Irvine, CA, United States
This study evaluates the performance of a newly developed long-term high-resolution satellite-based precipitation products, named Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network – Climate Data Record (PERSIANN-CDR), in hydrological modeling. PERSIANN-CDR estimations are biased corrected using GPCP monthly climatology data. PERSIANN-CDR provides daily rainfall estimates at 0.25° x 0.25° grid boxes for 1983-2014 (delayed present). This newly released product makes it feasible to model the streamflow over the past 30 years. Three test basins from the Distributed Hydrologic Model Intercomparison Project - Phase 2 (DMIP 2) are chosen. Comparing with other satellite products, the Version 7 TRMM Multi-satellite Precipitation Analysis (TMPA) product is used. Stage IV radar data is used as a reference data for evaluating the PERSIANN-CDR and TMPA precipitation data. All products are scaled to 0.25° and daily spatiotemporal resolution. The study is performed in two phases. In the first phase, the 2003-2011 period where all the products are available is chosen. Precipitation evaluation results, presented on Taylor Diagrams, show that TMPA and PERSIANN-CDR have close performances. The National Weather Service (NWS) Office of Hydrologic Development (OHD) Hydrology Laboratory-Research Distributed Hydrologic Model (HL-RDHM) is then forced with the PERSIANN-CDR and the TMPA precipitation products, as well as the stage IV radar data. USGS Streamflow observations at the outlet of the basins are used as the reference streamflow data. The results show that in general, in all the three DMIP 2 basins the simulated hydrographs forced with PERSIANN-CDR and TMPA show good agreement, as the statistical measures such as root mean square error, bias, and correlation coefficient are close. In addition, with respect to the streamflow peaks, PERSIANN-CDR shows better performance than Stage IV radar data in capturing the extreme streamflow magnitudes. Based on the results from the first phase of the study and given the fact that PERSIANN-CDR covers the 1983-2014, in the second phase of the study we model the streamflow for the period of 1983-2014. The results will be presented in the meeting.