Evaluation of Soil Moisture Simulations of the NOAA National Water Model over 5 Years of Retrospective Analysis

The impact of soil moisture initial conditions on prediction of hydrologic states has been a topic of research in hydrology for decades. Better representation of the soil moisture in the model is expected to improve streamflow forecasts and, therefore, we investigate the quality of the soil moisture simulations from National Water Model. The NOAA National Water Model (NWM) is an operational implementation of the community WRF-Hydro modeling system that provides short, medium, and long range forecasts of streamflow as well as distributed hydrological states and fluxes (including soil moisture) over the contiguous United States.

We verify the soil moisture simulations from a 5 year retrospective analysis (2011-2015) forced with NLDAS-2 for different versions of the NWM which has undergone some calibration. The verification is performed against the in-situ soil moisture observations obtained from International Soil Moisture Network (ISMN) that collates data from many soil moisture networks around the world, including USCRN, SCAN, SNOTEL, COSMOS, PBO_H2O, etc.

The model simulations exhibit high biases in the western and southeastern part of the US which translates into a wet bias in the streamflow simulations. Correlation coefficient of the soil moisture anomaly is another metric used in this study, where the anomaly is defined as the deviations of the soil moisture simulations from the climatology. In general, the results show high correlation coefficient in the eastern and central US except from southeast and reduces as moving toward west of the continental divide. Statistics are broken down by the land cover and soil type to investigate the accuracy of the simulations under different land cover or soil types. Seasonal statistics are also provided.