The Ebert-McBride Contiguous Rainfall Area (CRA) technique was used to evaluate the performance of the Eta model and a 12-km diabatically initialized ("hot start") version of the MM5 model during the International H20 Project (IHOP). The CRA technique was fine-tuned to optimize the pattern matching of forecasts with observations for the scales of precipitation systems observed during IHOP. A detailed morphological analysis of observed systems was performed using radar data for all CRAs identified in the IHOP domain. In addition, a morphological analysis was done to forecasted systems predicted by the MM5 using the model's hourly reflectivity output.

Around 200 convective systems were identified during the six-week period for each model. Systems were classified as being linear, linear bowing, nonlinear, isolated cells, discontinuous areal or nothing produced (e.g., a model showed a system that was not present in the observations). Linear systems were further classified as having trailing stratiform, leading stratiform or parallel stratiform regions, and the evolution characteristics were noted following the method of Bluestein-Jain. All rain systems producing a volume of water in a 6 h period roughly equivalent to 1 inch of rain over a 1 by 1 degree box were identified. Although traditional grid point to grid point objective skill measures were computed for the full domain, the focus of our analysis was on skill measures applicable to the identified CRAs. An analysis of these parameters, including mean square errors, correlation coefficients, improvements in correlation coefficient due to shifting of the forecast, volume errors, displacement errors, and pattern errors, will be presented as a function of the morphology. This analysis should reveal whether some types of systems, such as linear squall lines with trailing stratiform rain areas, were better predicted than other types within these models.