The Use of A Modified Ebert-mcbride Technique To Evaluate Mesoscale Model QPF As A Function of Convective System Morphology During IHOP 2002

Abstract

Verification of quantitative precipitation forecasts (QPFs) made by fine-grid numerical models for mesoscale convective systems (MCSs) can be prob-lematic. Traditional verification statistics severely penal-ize a precipitation system that may have been forecast with a small positional error or incorrect shape, with resultant low correlation coefficients, high root mean square errors (RMSEs), and poor values of categorical statistics (Ebert and McBride 2000; Baldwin and Wandishin 2002). Numerous approaches have been applied to deal with the deficiencies of traditional verifi-cation methods (e.g., Du et al. 2000; Zepeda-Arce et al. 2000; Bullock et al. 2004). One such approach is the Ebert-McBride technique (EMT), which employs the concept of matching individual forecast and observed areas (Ebert and McBride 2000, hereafter EM2000). The technique utilizes contiguous rain areas (CRAs), defined as the areas of contiguous observed and fore-cast rainfall enclosed within a specified isohyet. A dis-placement is performed by an objective pattern match-ing technique to optimally align the forecast with the observations. The EMT method was originally applied to synoptic-scale precipitation systems over Australia. The current study adapts the EMT to warm-season MCSs occurring over the central U.S. (Rogers et al. 1998) with Betts-Miller-Janjic (BMJ) con-vective parameterization (Janjic 1994), the Penn State/NCAR MM5 model version 3.5 (Grell et al. 1995), and the WRF model version 1.3 (Skamarock et al. 2001). Both the MM5 and WRF were run with no con-vective parameterization and initialized with a “Hot Start” procedure (McGinley and Smart 2001) developed for the NOAA Forecast Systems Laboratory, Local Analysis and Prediction System (Albers et al. 1996). 2.1 Overview of EMT The EMT uses CRAs as a way to determine error statistics. These CRAs are made up of the union of ob-served and forecast rainfall areas which exceed a user-specified threshold amount. An optimal displacement vector is then determined by translating the forecast area over the observed area, typically by either maxi-mizing correlation coefficient or by minimizing the total squared error. The forecast is permitted to shift within an expanded box enclosing the CRA (the maximum distance allowed between the forecast and observed areas beyond which it is assumed the two areas are unrelated). Several user-defined parameters can be adjusted to define the temporal and spatial scale of the CRA, the pattern matching process, and how verification statistics are calculated. The International H2O Project (IHOP) that took place from 16 May to 26 June 2002 was designed to help improve the understanding and prediction of QPF. High-resolution model datasets produced for this project offered the opportunity to investigate precipitation fore-cast accuracy as a function of convective system mor-phology. Analysis from the EMT objective verification measures in concert with an observed morphological classification scheme revealed systematic errors for certain types of MCSs.