An Evaluation of ensembles based upon MAPLE precipitation nowcasts and NWP precipitation forecasts

Skills of precipitation forecasts from radar based nowcasts, Numerical Weather Prediction models and ensembles of the above two are evaluated. MAPLE ( McGill Algorithm for Precipitation nowcasting by Lagrangian Extrapolation) uses variational echo tracking, a semi-Lagrangian advection scheme, scale-based filtering and appropriate rescaling of the filtered nowcast fields to generate precipitation nowcasts. In this study, MAPLE algorithms are applied to continental scale US radar composites of rainfall reflectivity to generate 5-min step nowcasts. The 5-min step radar reflectivity nowcasts are converted to rainfall rate maps using a Z-R relationship and accumulated to generate 1-h nowcast precipitation accumulations for up to 12 hours. One hour precipitation accumulations from GEM and WRF models are considered. Ensembles of MAPLE nowcast accumulations and model forecast accumulations are generated using a weight scheme bases upon the climatological or long term average of Critical Success Index (CSI) of each individual component. US radar composites, available every 5 minutes are accumulated to generate 1 hour precipitation totals and these are used as verification maps. Critical Success Index is used to asses the relative skills of various types of forecasts. Results, based upon several cases in the August-September 2004 period and January-March 2005 period, confirm that radar-based MAPLE nowcasts have better skill than NWP forecasts in the 1 to 6 hour period and indicate that MAPLE-NWP ensembles have better skill than either MAPLE or NWP forecasts in the 4 to 12 hour period on average.