Early validation of the UND/STWRC ensemble modeling system

During the past year, a quasi-operational ensemble modeling system has been implemented within the University of North Dakota's Surface Transportation Weather Research Center in support of road weather research efforts funded by the Federal Highway Administration (FHWA) through the North Dakota Department of Transportation. This ensemble modeling system's current design includes six members utilizing the Weather Research and Forecasting (WRF) model at a 10km horizontal resolution and 35 vertical computational levels, with a variety of physical parameterization variations as well as two different initializing analyses.

The output from the ensemble modeling system is intended to be utilized by a series of statistical and heuristic algorithms as well as fine-scale blowing snow and roadway visibility schemes. Therefore it is important, if not critical, to be able to quantify the performance of the ensemble from both the perspective of traditional numerical weather prediction (NWP) skill metrics as well as metrics designed to examine the degree of “value added” by the ensemble through the uncertainty information and probabilistic forecasts that the ensemble can provide.

In this paper we will present early validation results of the UND/STWRC ensemble modeling system over a domain that encompasses the states participating in the FHWA Pooled Fund Study as of July 2005. Validation results presented will include those from traditional NWP skill metrics, ensemble-based skill metrics as well as time-series comparisons of ensemble member and mean output with field observations taken at the STWRC field site (described in a companion paper). The latter comparison provides additional information related to the utility of an ensemble at a relatively coarse resolution compared to the roadway scale, by examining the degree that the envelope of member solutions at the 10 km scale encompasses the observed state near the roadway.