Wednesday, 14 January 2004: 4:45 PM
Verification Techniques Appropriate for Cloud-Resolving NWP Models
The authors briefly summarize a number of emerging techniques for verifying highly episodic and localized features contained in NWP model forecasts, such as precipitation accumulated over short (hourly) time intervals. These techniques are viewed from the perspective of their ability to yield useful guidance about the sources of errors in models, and contrasted with traditional verification measures more appropriate for coarse-resolution models. The composite picture obtained from a variety of approaches is emphasized through examination of a particular period during which high-resolution (4-km grid) forecasts were produced by the Weather Research and Forecast (WRF) model during the Bow echo And MCV EXperiment (BAMEX). The authors focus on rain-area verification, with special attention to mesoscale convective systems (MCSs), during the period 15 May to 15 July, 2003. MCSs are defined as objects in both forecasts and observations and matched by a subjective process that serves as a benchmark for the construction of automated matching algorithms. Systematic errors in initiation, propagation and dissipation, as well as structure and intensity are obtained. A cross comparison of errors found herein with those obtained by other verification approaches (performed by other investigators) is performed to obtain a comprehensive set of error characteristics for the WRF model during the period of interest. Benefits of using verification approaches that emphasize the storm-scale (as opposed to the grid-scale) character of the model output are demonstrated.