Improving probabilistic ensemble forecasts of convection through the application of QPF-POP relationships

Abstract

<p>Quantitative precipitation forecasts provide an accumulated precipitation amount for a given</p> <p>time period, and accurate forecasts depend on the correct prediction of areal coverage,</p> <p>timing, and intensity of precipitation. These forecasts are important to a variety of people for</p> <p>many different purposes, so expressing a likelihood of precipitation is also useful. Most</p> <p>simply, probabilities of precipitation are determined by considering the percentage of</p> <p>ensemble members forecasting precipitation greater than a specified threshold amount.</p> <p>Probabilities of precipitation can also be formed from quantitative precipitation forecasts</p> <p>through statistical post-processing. Past research has shown that there are many ways to</p> <p>post-process precipitation data, such as by binning the precipitation amounts, applying</p> <p>statistical calibration, and/or considering the percentage of an area receiving precipitation.</p> <p>The main goal of this study was to expand upon relationships between quantitative</p> <p>precipitation forecasts and probabilities of precipitation by developing new approaches that</p> <p>yield more accurate probabilities of precipitation than methods that are currently more</p> <p>commonly used. Ensemble forecasts from the 2007 and 2008 NOAA Hazardous Weather</p> <p>Testbed Spring Experiments were used to provide quantitative precipitation forecasts for</p> <p>various days. In the study, four main approaches were developed and tested extensively</p> <p>using Brier scores and other statistics. Brier scores for different approaches were compared</p> <p>to traditional methods of calculating probabilities of precipitation. It was shown at both 20</p> <p>km and 4 km grid spacings that new approaches were able to produce statistically significantly better forecasts than a traditional method that relies upon calibration of POP forecasts derived using equal-weighting of ensemble members. A deterministic approach developed during the study was also able to produce forecasts comparable to those of the calibrated traditional method.</p>