loading page

Evaluating Input Data and Rain Snow Separation Improvements to the National Water Model Simulation of Snow Water Equivalent
  • Irene Garousi-Nejad,
  • David Tarboton
Irene Garousi-Nejad
Stanford University, Stanford University

Corresponding Author:[email protected]

Author Profile
David Tarboton
Utah State University, Utah State University
Author Profile


We compared snowfall, and snow water equivalent (SWE) accumulation and ablation simulations from the WRF-Hydro model with the U.S. National Water Model (NWM) configuration against observations at a set of representative point locations from Snow Telemetry (SNOTEL) sites across the western U.S. We focused on the model’s partitioning of precipitation between rain and snow and selected sites that span the variability of the percentage of rain on snow precipitation events. Our results show that the NWM generally under-estimates SWE and tends to melt snow earlier than observations in part due to errors in the precipitation and air temperature inputs. We reduced some of the observed and modeled discrepancies by using SNOTEL snow-adjusted precipitation and removing air temperature biases, based on observations. These input changes produced an average 59% improvement in the peak SWE. Modeled peak SWE was further improved using humidity-dependent rain-snow-separation. Both dew point and wet-bulb parameterizations were evaluated, with the dew-point parameterization giving better overall improvement, reducing the bias in SWE by 18% compared to the NWM air temperature-based scheme. This modification also improved melt timing with the number of site years having difference between modeled and observed date of half melt from peak SWE six or more days reduced by 6%. These SWE magnitude and timing improvements varied when analyzed for each rain-on-snow percentage class, with generally better results at sites where most precipitation events fall either as snow or as rain, and less improvement when there is a mix of snow and rain-on-snow events.