H111-0006
Impact of Calibration Objective Function on the Performance of National Water Model
Friday, 11 December 2020
Poster
Xia Feng1, Erin Towler2, Yuqiong Liu3,4, Arezoo Rafieeinasab5, Brian A. Cosgrove6, Katelyn FitzGerald5, Laura Read5, Wanru Wu3, David Gochis2, Aubrey L Dugger5, Trey Flowers7 and Rachel McDaniel8, (1)UCAR/NWC, Tuscaloosa, AL, United States, (2)NCAR, Boulder, CO, United States, (3)NOAA/NWS/OWP, Silver Spring, MD, United States, (4)LEN Technologies, Oak Hill, VA, United States, (5)National Center for Atmospheric Research, Boulder, CO, United States, (6)NOAA/NWS Office of Water Prediction, Silver Spring, MD, United States, (7)NOAA/NWS Office of Water Prediction, National Water Center, Tuscaloosa, AL, United States, (8)NOAA/NWS/OWP/National Water Center, Tuscaloosa, AL, United States
Abstract:
The NOAA/OWP’s operational National Water Model (NWM) produces high-resolution forecasts of various hydrologic and land surface variables across the contiguous United States (CONUS) and Hawaii. It serves as a valuable resource to support the missions of the NWS and its federal partners. Acknowledging significant uncertainty in model parameters due to highly heterogeneous catchment properties, simplified physical representations, and scaling of field measurements, several NWM parameters are estimated through a calibration procedure. During calibration, model parameters are adjusted to optimize the objective function that measures the deviation between simulated and observed system behavior. This study aims to investigate the choice of objective function in parameter selection and its effect on the predictive ability of the calibrated NWM.
To achieve this goal, we selected four objective functions, including Normalized Nash-Sutcliffe efficiency (NNSE), Kling-Gupta efficiency (KGE), Lamontagne-Barber efficiency estimator of KGE (LBEm-KGE) and an event based multi-objective measure. These diverse objective functions were selected to investigate their impact on the full range of hydrologic applications, and to assess the performance of high flows, which are a critical focus. We plan to calibrate the NWM to the observed streamflow following the same optimization protocol used in the calibration of NWM V2.1. The optimization iterations will be carried out for about ninety river basins covering a wide range of physiographic and hydroclimatic conditions throughout the CONUS, Hawaii and Puerto Rico. The accuracy of NWM calibrated by different objectives will be compared with each other and with the V2.1 results to provide insight on the merits of each objective function in terms of optimization efficiency, parameter identifiability and most importantly the adequacy of the model in representing a variety of hydrologic responses. The findings of this work will help guide the selection of the objective function and optimization approach for the calibration process of the next version of NWM, V3.0, currently under development and scheduled for deployment into operations at the end of 2021.
