Extended Abstract

During the HWT project, precipitation objects, i.e. cohesive areas of precipitation, based onensemble probability (probability of precipitation > 50% for 12.7mm in a 3 hour period) were identifiedfor the entire 6 weeks of the project. Then total counts of both forecast and observed objects wereaccumulated for 3 hourly lead times and are shown in Figure 1. Since the issue time of all forecasts is0Z, the lead time also indicates the time of day. The model forecast too many objects in the earlierhours (3, 6, and 9), and too few objects in the remaining hours. A similar plot of the total object areas(not shown) shows that the objects cover more of the domain than the forecasts at all lead times except3 hours, when the areas are similar. The Brier Scores for all lead times are quite small, ranging from0.017 to 0.009. Since Brier Score is similar to mean squared error, these small values indicate a goodforecast.In addition to examining precipitation, the HWT experiment also examined radar reflectivity andradar echo top. The reflectivity is in dBz, while radar echo top is the height of the 18 dBz isopleths.Prior to this experiment, MODE had not been used directly on radar fields. This project confirmed thatobject based verification is appropriate for these fields. An example is shown in Figure 2. The locationand orientation of the reflectivity and echo top forecasts match well with the observations, though theforecasts are shifted slightly too far to the southwest. Forecast areas are somewhat too small. Thisanalysis is for a single case, but this type of information can also be derived for longer periods.Figure 2: Top: Observed (left) and simulated (middle) composite reflectivity valid on June 8, 2010 at 15 UTC.Corresponding MODE objects (right) for observed field (blue line) and forecast field (red and blue solid objects).Bottom: Observed (left) and simulated (middle) height of 18dBZ reflectivity field, called Radar Echo Top, valid on-191-