The recently enhanced National Hydrography Dataset (NHDPlus) provides a “blue line” description of the river networks of the USA consisting of approximately 3 million river and water body reaches. One of the major assets of NHDPlus is that the connectivity between each reach of large river networks is known hence allowing using the “blue line” rivers from maps as computing elements for a river model and providing streamflow computations everywhere in large river network, including ungaged rivers. The US Geological Survey National Water Information System provides stream flow observations at approximately 20,000 gages in the USA and these represent a valuable tool to assess runoff computations by land surface models at many places, not only the at river mouth. The Routing Application for Parallel computatIon of Discharge (RAPID, David et al. 2011) showed an example on how to use the NHDPlus river network along with runoff from a land surface model calibrated locally and with USGS gage measurements in the Guadalupe and San Antonio Rivers Basins in Texas (26,238 km2). Over the past two decades, the atmospheric science community has produced several continental-scale gridded meteorological datasets that cover millions of square kilometers over 30+ years at temporal and spatial resolutions on the order of 1-3 hours and 10-50 kilometers respectively (i.e. NARR, NLDAS2, etc.). These datasets contain computations by land surface models which typically calculate a number of variables including some related to the terrestrial water balance. In particular, the combination of surface and subsurface runoff has been extensively used to compute river flow at the continental scale using gridded river networks. It is therefore tempting to use such meteorological datasets along with the now available “blue lines” river models to support the development of advanced river models to be applied at the continental scale. Several scientific questions lie behind such work. Is the quality of continental-scale runoff datasets from land surface models sufficient for such regional- to continental scale river flow computations using mapped rivers? Where can runoff computations be improved? What are the advantages and drawbacks of using “blue line” river networks? Should a river model be optimized separately for each runoff dataset? Should one use all available stations to train a river model? Does it make a difference to use a dry or wet year when optimizing river flow parameters? The study proposed uses the Texas Gulf hydrologic region (464,135 km2) during the years 2000 to 2007 as a test case for answering these questions using runoff products from the second version of the North American Land Data Assimilation System (NLDAS2) and from a regionally-calibrated version of the Noah land surface model with multi-parameterizations options (Noah-MP, Niu et al. 2011, Yang et al. 2011).