The use of a modified Ebert-McBride technique to evaluate mesoscale model QPF as a function of convective system morphology during IHOP 2002

The Ebert-McBride technique (EMT) is an entity-oriented method useful for quantitative precipitation forecast verification. In this talk, we describe how the EMT was modified to optimize its ability to identify contiguous rain areas (CRAs) during the 2002 Interna-tional H2O Project (IHOP) over the Central Plains of the U.S. The technique was used to identify systematic sources of error as a function of convective system morphology in three 12-km model simulations run over the IHOP domain: Eta, MM5, and WRF. To investigate these errors, a detailed radar-based morphological classification scheme was developed for all observed systems in which a CRA was identified.

Our results show that the Eta model produced average rain rates, peak rainfall amounts, and total rain volume that were lower than observed for almost all types of convective systems, likely due to the Eta's use of the Betts-Miller-Janjic (BMJ) convective parame-terization. Like the Eta, the MM5 and WRF underestimated rain volume for most MCS types. However, these models generated average rain rates and peak rainfall amounts that were larger than observed, implying a sizeable reduction in areal coverage. The results for the WRF and MM5 are consistent with previous observations of mesoscale models run with explicit microphysics and no convective parameterization scheme.

A decomposition of mean square errors (MSEs) into displacement, pattern (or variabil-ity), and volume components are shown as well. Two of the more striking results from the decomposition are illustrated. First, all three models forecast rainfall too far north-west for linear systems overall. This is likely related to an improper prediction of cold pool dynamics (i.e. the forecasts are too slow with cold pool development and thus MCS evolution). Finally, the average pattern error was found to be the largest component of error for most MCS types. The average pattern error for the Eta was much lower than the MM5 and WRF for continuous MCS types, a result consistent with its known lack of variability when using the BMJ convective scheme. Consequently, the Eta had signifi-cantly lower total MSE compared to the MM5 and WRF for these types.