J35.3A Assessment of the Simulations of Global-scale River Flows from an Integrated Hydrological Modeling Framework

Wednesday, 10 January 2018: 9:00 AM
Room 18A (ACC) (Austin, Texas)
Wen-Ying Wu, Univ. of Texas at Austin, Austin, TX; and Z. L. Yang, P. Lin, and D. Maidment

Gauged streamflow data provide an integrated measure of hydrologic processes over the entire watershed. However, current understanding of large-scale river flows is limited because of insufficient gauged observations. Therefore, an integrated hydrological modeling framework is needed to take advantage of gridded meteorological forcing, land surface modeling, channeled flow modeling, ground observations, and satellite remote sensing. Launched in August 2016, the National Water Model successfully incorporates weather forecasts to predict river flows for more than 2.7 million rivers across the continental United States, which transfers a “synoptic weather map” to a “synoptic river flow map” operationally. In this study, we apply a similar framework to a high-resolution global river network database, which was developed from a hierarchical Dominant River Tracing (DRT) algorithm, and runoff output from the Global Land Data Assimilation System (GLDAS) to a vector-based river routing model (The Routing Application for Parallel Computation of Discharge, RAPID) to produce river flows from 2001 to 2016 using Message Passing Interface (MPI) on TACC's Stampede system. In this simulation, global river discharges for more than 177,000 rivers are computed every 30 minutes and outputed every 3 hours. The modeling framework’s performance is evaluated with various observations including river flows at more than 400 gauge stations globally. Although the model exhibits an overall good performance, there are mean biases in available water storage, evapotranspiration and runoff. The time delay from cross-correlation between modeled and observed flows shows that land cover characteristics and river geometry parameters should be considered as they affect flow travel time on land and within channels, respectively. The low spatial correlation in cold regions during spring indicates the inadequate treatment of snowmelt and soil freeze–thaw processes in the land surface model. Directions for future research will be outlined and discussed.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner