Abstract
Abstract. To extend geographical coverage, refine spatial resolution, and improve modeling efficiency, a computation- and data-intensive effort was conducted to organize a comprehensive hydrologic data set with post-calibrated model parameters for hydro-climate impact assessment. Several key inputs for hydrologic simulation – including meteorologic forcings, soil, land class, vegetation, and elevation – were collected from multiple best-available data sources and organized for 2107 hydrologic subbasins (8-digit hydrologic units, HUC8s) in the conterminous US at refined 1/24° (~4 km) spatial resolution. Using high-performance computing for intensive model calibration, a high-resolution parameter data set was prepared for the macro-scale variable infiltration capacity (VIC) hydrologic model. The VIC simulation was driven by Daymet daily meteorological forcing and was calibrated against US Geological Survey (USGS) WaterWatch monthly runoff observations for each HUC8. The results showed that this new parameter data set may help reasonably simulate runoff at most US HUC8 subbasins. Based on this exhaustive calibration effort, it is now possible to accurately estimate the resources required for further model improvement across the entire conterminous US. We anticipate that through this hydrologic parameter data set, the repeated effort of fundamental data processing can be lessened, so that research efforts can emphasize the more challenging task of assessing climate change impacts. The pre-organized model parameter data set will be provided to interested parties to support further hydro-climate impact assessment.
Highlights
With the advance of high-performance computing and more abundant historical observation data, hydrologists and water resource engineers are better equipped to improve the scale, resolution, and accuracy of hydrologic simulation
Average daily maximum temperature (Tmax), daily minimum temperature (Tmin), annual total precipitation (P ), and average wind speed (W ) from 1980 to 2008 for each of the 2107 HUC8s were computed for comparison
Both Daymet and Maurer are closer to PRISM, but North American Regional Reanalysis (NARR) is more divergent than the other data sets
Summary
With the advance of high-performance computing and more abundant historical observation data, hydrologists and water resource engineers are better equipped to improve the scale, resolution, and accuracy of hydrologic simulation. A statistical model can generally simulate hydrologic variables well with fewer predictors, the assumption of stationarity may be questionable in a changing environment in which many hydrologic processes are expected to be disrupted (Milly et al, 2008) Under such conditions, historical relationships may not provide fully accurate information about future streamflow and water availability. One example is ANN (and the various related machine learning algorithms) These types of advanced statistical methods are extremely powerful in forecasting reservoir outflows with minimal observation, the physical relationships among various predictors cannot be interpreted (Govindaraju and Rao, 2000); this hinders their direct extension across different locations and climate patterns. These methods may not be suitable choices for climate-related research
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
