WRF Simulation, Model Sensitivity, and Analysis of the December 2013 New England Ice Storm

Abstract

Ice storms pose significant damage risk to electric utility infrastructure. In an attempt to improve storm response and minimize costs, energy companies have supported the development of ice accretion forecasting techniques utilizing meteorological output from numerical weather prediction (NWP) models. The majority of scientific literature in this area focuses on the application of NWP models, such as the Weather Research and Forecasting (WRF) model, to ice storm case studies, but such analyses tend to provide little verification of output fidelity prior to use. This study evaluates the performance of WRF in depicting the 21-23 December 2013 New England ice storm at the surface and in vertical profile. A series of sensitivity tests are run using eight planetary boundary layer (PBL) physics parameterizations, three reanalysis datasets, two vertical level configurations, and with and without grid nudging. Simulated values of precipitation, temperature, wind speed, and wind direction are validated against surface and radiosonde observations at several station locations across northeastern U.S. and southeastern Canada. The results show that, while the spatially and temporally averaged statistics for near-surface variables are consistent with those of select ice-storm case studies, near-surface variables are highly sensitive to model when examined at the station level. No single model configuration produces the most robust solution for all variables or station locations, although one scheme generally yields model output with the least realism. In all, we find that careful model sensitivity testing and extensive validation are necessary components for minimizing model-based biases in simulations of ice storms.