Authors: Keith Jennings – University of Vermont
Title: Can We Improve Precipitation Phase Partitioning in the National Water Model and NextGen Formulations? (Yes)
Presentation Type: Poster
Abstract: Hydrologic modeling in temperate and snowy regions requires accurate inputs of both precipitation rate and phase. While thousands of meteorological stations across the US deploy myriad types of gages to measure the former, there are relatively few direct observations of the latter. As such, hydrologic models typically rely on internal routines to partition precipitation into rain and snow. But how effective are these methods and what does it mean if we get the wrong answer? This presentation will review the types of precipitation phase partitioning methods, the data they require, and their relative pros and cons. We will focus on those methods deployed by the current National Water Model and the formulations (Noah-OWP-Modular and Snow-17) implemented for use in the Next Generation Water Resources Modeling Framework (NextGen). We will compare results from those approaches to observations from a unique dataset: crowdsourced precipitation phase reports from the Mountain Rain or Snow participatory science project. We will also show how these outcomes compare to precipitation phase estimates from three machine learning approaches: an artificial neural network, XGBoost, and random forest. Finally, we will suggest pathways for better predicting rain, snow, and mixed precipitation in the NextGen Framework.