Improving Snow and Streamflow Simulation in the National Water Model By Leveraging Advanced Mesonet Observations from the Mountains of New York State

Snow water storage is a critical aspect of river forecasting and water management strategies across most mountainous basins in the United States (US). Despite its importance, the current operational version of the NOAA National Water Model (NWM) initializes snow states of the forecast using a simple deterministic analysis cycling every hour. This approach can lead to substantial errors in snowpack characterization within the model as biases in forcing and snowpack physics accumulate within the modeled snow state throughout the winter season. Furthermore, the extent of this error is largely unknown and varies according to region due to regional differences in climatological forcing and snowpack history. Performance of the NWM, and the underlying Noah-MP land surface model, at simulating snow over the mountains of the northeastern US are poorly characterized, in part due to limitations in the observational network.

This study works to improve snow state initialization and prediction within the NWM by taking advantage of observations from the New York State Mesonet (NYSM). To achieve this outcome, we are working to: 1) optimize the snow physics parameterization suite in the NWM to improve streamflow prediction in New York State and assess snow-state uncertainty as related to snow parameterizations, and 2) implement and demonstrate snow data assimilation of in situ snow depth and snow water equivalent observations through the ensemble Kalman filtering framework within NCAR’s Data Assimilation and Research Testbed. We leverage unique high-quality automated snow and meteorological observations from the NYSM that include stations located in the Adirondack Mountains, Catskill Mountains, and Tug Hill Plateau regions. These are supplemented with manual snow course observations, used to ground-truth the NYSM automated observations and characterize spatial variability and station representativeness. Initial results will be presented using data from the 2019-2020 snow season to better constrain uncertainty in snow physics and demonstrate the impact of assimilating in situ snow observations into the NWM.