A probabilistic nowcasting approach to precipitation start/stop time using a recursive probability calculator

Providing the start/stop time of precipitation at specific locations enables a wide range of consumers to benefit from the short term precipitation forecast. However, the accuracy of such forecast is limited by the predictability of small scale features in precipitation. The proposed solution at Pelmorex Media Inc. is to use an ensemble of forecasts to calculate a probability estimation based on consecutive runs of the Pelmorex radar nowcasting system. The system is based on McGill Algorithm for Precipitation Nowcasting Using Semi-Lagrangian Extrapolation (MAPLE) and produces short-term deterministic precipitation forecasts at 1 km resolution for the whole North America at every 10 minutes interval. The characteristic of the operational Pelmorex radar nowcasting system that distinguishes it from a normal NWP forecast is its high spatial and temporal resolution. The Recursive Probability Calculator (RPC) is an innovative method of merging forecasts from past runs to create statistical information about the current deterministic forecasts. For this study, more than 72 hours of nowcasting and radar data was recorded during 17, 18 and 19 April 2013 throughout several large scale rain storms that accompanied high wind gusts, scattered showers, and convective events. The deterministic nowcasting and probabilistic estimations were documented at 12 airport locations near major Canadian cities and then compared to METAR data. According to the experiments, the probabilistic method can be tuned by setting proper thresholds for start/stop times based on a statistically significant amount of real events. In conclusion, the probabilistic approach increases the expected accuracy, reduces sensitivity to radar false echoes, and adds flexibility to the operational design.