993 An Investigation of WRF Physics Schemes using a Massive Ensemble Simulation of the 1993 Superstorm

Wednesday, 25 January 2017
Matthew Vaughan, SUNY, Albany, NY; and R. G. Fovell and Y. Cao

This study analyzes the 12–14 March 1993 Superstorm through a massive ensemble modeling approach. The foci of the study are to produce the most accurate reconstruction of the observed storm and evaluate the impact of physics schemes on the simulated storm. The Weather Research and Forecasting model (WRF) Version 3.7, initialized using ECMWF Re-Analysis Interim (ERAI) data, is used to simulate the Superstorm with various physics options. The 60-km resolution model domain spans North America while a 20-km inner nest covers the eastern U.S. Microphysics, planetary boundary layer, shortwave radiation, longwave radiation, land surface, and cumulus parameterization schemes are varied to produce >1500 model runs with unique physics combinations.
The root-mean square error of sea level pressure (SLPer) with respect to ERAI analyses is calculated to select the “benchmark” case, defined as the ensemble member with the lowest SLPer over all simulation times. Additional metrics including the Northeast Snowfall Impact Scale (NESIS) and total snowfall over land are calculated to assess the impact of each ensemble member. To gauge the robustness of the benchmark physics combination, simulations of the 29 January 2015 and the 30 December 2013 snowstorms are performed to determine whether the benchmark physics schemes perform well for weaker events.
Preliminary results suggest the benchmark ensemble member had a higher NESIS value and higher snowfall amounts than most of the ensemble runs. This agrees with previous work using initial condition variations to obtain an ensemble of solutions suggesting the Superstorm was virtually as impactful as it could have been. Additionally, each category of physics parameterizations contained schemes with at least some significant differences in SLPer and NESIS values. Particularly, WRF single moment class 5 microphysics has a pronounced low snowfall bias, producing significantly lower NESIS values and total snowfall than other microphysics schemes.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner