MODELING OF THREE WATERSHEDS FROM MISSISSPPI, SOUTH KOREA, AND NEW ZEALAND USING MULTIPLE PRECIPITATION PRODUCTS

Remotely sensed precipitation products have increasingly been used to support hydrologic modeling. This study used five different types of precipitation data to support streamflow modeling in three basins from Mississippi (Skuna; 1634 sq km), South Korea (Guem; 3064 sq km), and New Zealand (Punehu; 41 sq km). All hydrologic models within a watershed are identical except for the input precipitation data. The hydrologic model selected for this study was the Soil and Water Assessment Tool (SWAT). Precipitation products include two ground-based products: rain gauge and the NEXRAD based Multisensor Precipitation Estimator (MPE - only in Mississippi). Additionally, remotely sensed precipitation data include two products derived from NASA's Tropical Rainfall Measurement Mission (TRMM); TRMM 3B42 research and real-time, as well as the Climate Prediction Center's morphing technique (CMORPH). TRMM research has a ground bias correction; whereas, TRMM real-time and CMORPH are pure satellite products without corrections. Model performance was evaluated using standard hydrology metrics including mass balance error (MBE) and Nash-Sutcliffe efficiency coefficients (NS) obtained by directly comparing simulated and observed streamflow. In the Skuna Basin simulation (2008 to 2010) based on MPE (MBE=0%; NS=0.72), rain gauge (MBE=4%; NS=0.60), and TRMM research (MBE=1%; NS=0.66) yielded acceptable results unlike the real-time version of TRMM. In the Guem Basin (2005-2007) all products yielded acceptable simulations with MBE (±22%) and NS >0.80. Finally, in the small Punehu watershed (2005-2007) both rain gauge (MBE=0%; NS=0.62) and TRMM research (MBE=7%; NS=0.54) yielded acceptable results unlike the pure satellite products. In summary ground-based precipitation data and TRMM research products yielded acceptable simulations the three examined basins. Conversely, real-time satellite precipitation products do not produce acceptable modeling in all examined watersheds.