Comparing nowcastings of three severe convective events by statistical and NWP models

Two precipitation nowcasting methods are compared. The first method uses the NWP model COSMO with assimilation of radar reflectivity and satellite data. The second applies the statistical nowcasting model (NOW) to observed radar, lighting and satellite data, and prognostic data from a low-resolution NWP model. The nowcasting methods are compared for three convective events of locally heavy precipitation.The model forecasts are verified and compared giving priority to the NOW model since forecast formulations for the considered models differ and as such the form used by the NOW model is generally applied. A simple categorical verification of the forecasts sensitive to the position of the precipitation area is used. A small shift of the forecast area may therefore significantly decrease the skill score.Results show that both methods have limitations. The NOW model is able to predict further development of already observed precipitation but the forecast is not satisfactory when there are no indications of convective activity contained in the predictors. The physical model does not have this limitation, but the forecasts significantly depend on the initial and boundary conditions and on assimilated data. When the assimilation is not used, the COSMO model forecasts are typically worse than the NOW model forecasts.