H11D-1367
Spatially-distributed Calibration of Two Macroscale Hydrologic Models Across the Columbia River Basin

Monday, 14 December 2015
Poster Hall (Moscone South)
Oriana Chegwidden, University of Washington Seattle Campus, Seattle, WA, United States, Mu Xiao, University of California Los Angeles, Los Angeles, CA, United States, David E Rupp, Oregon Climate Change Research Institute, Corvallis, OR, United States, Matt R Stumbaugh, University of Washington, Department of Civil and Environmental Engineering, Seattle, WA, United States, Joseph Hamman, Applied Physics Laboratory University of Washington, Seattle, WA, United States, Ming Pan, Princeton University, Princeton, NJ, United States and Bart Nijssen, University of Washington, Seattle, WA, United States
Abstract:
Hydrologic models are often calibrated to streamflow observations at discrete points along a river network. Even if the area contributing to each flow location is discretized into multiple model elements, the calibration parameters are typically adjusted uniformly, either by setting them to the same value or transforming them in the same way (for example, multiply each parameter value by a given factor). Such a procedure typically results in sharp gradients in calibrated parameters between neighboring subbasins and disregards parameter heterogeneity at the subbasin scale. Here we apply a streamflow disaggregation procedure to develop daily, spatially-distributed runoff fields at the same resolution as the model application. We then use these fields to calibrate selected model parameters for each model grid cell independently.

We have implemented two hydrologic models (the Variable Infiltration Capacity model and the Precipitation Runoff Modeling System) across the Columbia River Basin plus the coastal drainages in Oregon and Washington at a subdaily timestep and a spatial resolution of 1/16 degree or ~6km, resulting in 23,929 individual model grid cells. All model grid cells are calibrated independently to the distributed runoff fields using the shuffled complex evolution method and the Kling-Gupta Efficiency (KGE) as the objective function. The KGE was calculated on a weekly time step to minimize the effects of timing errors in the disaggregated runoff fields. We will present calibrated parameter fields and then discuss their structure (or lack thereof), which can provide important insight into parameter identifiability and uncertainty.