Regional-scale river flow modeling using off-the-shelf runoff products, thousands of mapped rivers and hundreds of stream flow gauges

Abstract

The permanent rivers and streams of the United States - i.e. the " blue lines" from the United States Geological Survey (USGS) maps - are available in a Geographic Information System dataset called NHDPlus in which network connectivity for approximately three million reaches is known. The USGS also measures stream flow at about twenty thousand gauges whose locations are known exactly on NHDPlus rivers. The Routing Application for Parallel computatIon of Discharge (RAPID) is a river network model that can be run on NHDPlus if given inflow to rivers from runoff. The second phase of the North American Land Data Assimilation System (NLDAS2) provides runoff estimates from four land surface models over the continent. One could therefore combine RAPID with NLDAS2 runoff to compute stream flow in all NHDPlus " blue lines" and compare computations with USGS observations over the United States. The work presented here investigates such a combination but focuses on the Texas Gulf Coast Hydrologic Region as a test case for regional-scale river flow modeling using off-the-shelf runoff products, thousands of mapped rivers and hundreds of stream flow gauges. Despite initial difficulties in the determination of water inflow from gridded NLDAS2 runoff datasets to NHDPlus rivers, the " blue line" approach is shown to be advantageous partly because hundreds of gauges are easily located on river networks and used throughout basins. Such use of a large number of gauges allows for new investigations related to the analysis of flow computations and to the optimization of model parameters thereby bridging the gap between continental-scale environmental modeling and local-scale planning and decision making. The quality of NLDAS2 runoff relative to locally-calibrated data is assessed through comparison between computed and observed river flow. With the exception of large outliers at a few gauges, the best NLDAS2 runoff products perform similarly to the regional dataset. Such outliers prevent the successful optimization of river model parameters but a careful selection of what stations to use for optimization is shown to overcome this difficulty. While river model parameters determined separately for each runoff dataset allow obtaining best results, a single set of parameters is sufficient to assess the relative quality of runoff datasets. The detailed NHDPlus description of river reaches also allows investigating the impact of increased spatial variability in model parameters which - following optimization of model computations - suggests that those parameters of reaches located on the main stems of large rivers tend to have the largest influence on the quality of flow simulations. © 2013.