Hydrologic Evaluation of TRMM-3B42 and GLDAS-CLM Precipitation Products over a Mid-size Basin

Recently, a great deal of effort has been devoted to providing continuous precipitation data from satellites and coupled land-atmosphere models with quasi-global coverage. These model and satellite based precipitation products are key variables for hydrologic and water management applications. This study assesses the applicability of the NASA TRMM-3B42 (3hr/0.25 deg) and Global Land Data Assimilation System (GLDAS) (3hr/1.0 deg) quasi-global precipitation fields for streamflow simulation at the sub-catchment scale (≥ 100 km2) by forcing a hydrological model for the Susquehanna River Basin (~70,000 km2). Two major storm events, Hurricane Irene (20-28 Aug 2011) and Tropical Storm Lee (1-5 Sept 2011) that caused widespread floods in the study area were selected for this study. NOAA's National Centers for Environmental Prediction (NCEP) hourly multi-sensor precipitation analysis (stage IV) high resolution (1hr/4km) rainfall fields are used for forcing the reference hydrological model. The two coarser products (GLADS and TRMM-3B42) are compared against the NCEP reference rainfall at varying spatial scales. For example, when accumulated over the duration of Tropical Storm Lee and averaged over the region, 3B42 rainfall generally agrees with NCEP (only 11% overestimation), while GLDAS exhibits significant bias (48% underestimation). However, the hydrological modeling results using both the TRMM-3B42 and GLDAS products exhibit a general flow pattern that agrees with the NCEP based model simulation at larger basin scales, but errors magnify significantly at smaller scales. Model results are compared and presented along select river basins with contributing areas ranging from 100 to 70,000 km2 to illustrate how model results differ between precipitation products as function of basin scale. Although the study is based on two severe events, the framework developed herein is applicable for more detailed investigations, such as seasonal variation or climatological studies facilitated by long-term datasets.