Generalized Continental Scale Hydrologic Model Parameter Estimates: Application to a VIC model implementation for the Contiguous United States (CONUS)

Thursday, 18 December 2014: 11:35 AM
Naoki Mizukami1, Martyn P Clark2, Bart Nijssen3, Kevin Michael Sampson2, Andrew James Newman1 and Luis E Samaniego4, (1)National Center for Atmospheric Research, Boulder, CO, United States, (2)NCAR, Boulder, CO, United States, (3)University of Washington Seattle Campus, Seattle, WA, United States, (4)Helmholtz Centre for Environmental Research UFZ Leipzig, Leipzig, Germany
Parameter estimation is one of the biggest challenges in hydrologic modeling, particularly over large spatial scales. Model uncertainty as a result of parameter values can be as large as that from other sources such as the choice of hydrologic model or the choice of model forcing data. Thus far, parameter estimation has been performed in an inconsistent manner across the model domain, e.g., using patchy calibration or spatially constant parameters. This can produce artifacts in the spatial variability of model outputs, e.g., discontinuity of simulated hydrologic fields, difficulty with spatially consistent parameter adjustments, and so on. We implement a framework that is suitable for use across multiple model physics options to map between geophysical attributes (i.e., soil, vegetation) and model parameters that describe the storage and transmission of water and energy. Specifically, we apply the transfer functions that transform geophysical attributes into model parameters and apply these transfer functions at the native resolution of the geophysical attribute data rather than at the resolution of the model application. The model parameters are then aggregated to the spatial scale of the model simulation with several scaling functions - arithmetic mean, harmonic mean, geometric mean. Model parameter adjustments are made by calibrating the parameters of the transfer function rather than the model parameters themselves.We demonstrate this general parameter estimation approach using a continental scale VIC implementation at a 12km resolution. The VIC soil parameters were generated by a set of transfer functions developed with nation-wide STATSGO soil data. The VIC model with new soil parameters is forced with Maurer et al. 2002 climate dataset (1979-2008) and the simulation results are compared with the previous simulations with parameters used in past studies as well as observed streamflows at selected basins.