The quantitative precipitation forecasting component of the 2010 NOAA Hazardous Weather Testbed Spring Experiment
Spring Experiment participants used high resolution convection allowing model guidance and observations to forecast the probability of exceeding 0.5 in and 1.0 in of precipitation during two near term 6 hour periods. The deterministic high resolution guidance featured a variety of Weather Research and Forecasting (WRF) models with both Advanced Research WRF (ARW) and Nonhydrostatic Mesoscale Model (NMM) physics cores and grid spacing of 1-4 km. These models were contributed by the National Severe Storms Laboratory (NSSL), the Environmental Modeling Center (EMC), the National Center for Atmospheric Research (NCAR), and the University of Oklahoma's Center for Analysis and Prediction of Storms (CAPS). In addition to the deterministic guidance, the experiment also featured output from a 26 member Storm Scale Ensemble Forecast system (SSEF) with 4 km grid spacing provided by CAPS. Ensemble forecast guidance included mean precipitation, probability matched mean precipitation, point exceedance probabilities, and neighborhood exceedance probabilities.
The performance of the high resolution models relative to their operational counterparts was evaluated both subjectively and objectively. Subjective evaluations by participants showed that the NSSL WRF-ARW and the CAPS SSEF consistently provided better forecast guidance than the North American Mesoscale model (NAM) and the Short Range Ensemble Forecast system (SREF), respectively. In fact, over both forecast periods approximately 57% of the NSSL WRF-ARW forecasts and 65% of the CAPS SSEF forecasts were found to improve upon comparable operational guidance. The ensemble mean precipitation from the CAPS SSEF typically provided realistic depictions of both the areal coverage and maximum precipitation amounts. Even with the advent of high resolution modeling, participants found that warm season QPF remains challenging due to initial condition errors, misforecast boundaries, and model biases. Examples of major high resolution model successes as well as these challenges will be shown.