Constraints on the impact of radar rainfall data assimilation on forecasts of cumulus convection

Experience from operational trials of assimilation of radar data in kilometre-scale numerical weather prediction models (operating without cumulus parametrization) shows that the positive impact of the radar data on convective precipitation forecasts typically decays within a few hours, although certain cases show much longer impact time-scales. In this work the impact time of radar data assimilation is related to characteristics of the meteorological environment. Three cases of convection over southern Germany with different synoptic conditions are investigated (one case with two data assimilation cut-off times), each with an ensemble of ten forecasts at 2.8 km horizontal resolution based on different initial and boundary conditions from a global forecast ensemble. Control forecasts are compared with forecasts where radar rainfall data are assimilated using latent heat nudging. The impact time of the radar data on total precipitation is quantified, and found to correlate well with a convective time-scale that measures the rate at which convection is responding to changes in large-scale forcing. Short impact times were associated with short convective time-scales that are characteristic of equilibrium convection. In this regime the statistical properties of the convection are constrained by the large-scale forcing, and effects of the radar data are lost within a few hours as the convection rapidly returns to equilibrium. When the convective time-scale was large (non-equilibrium conditions), the impact of the radar data was longer since convective systems were triggered by the latent heat nudging and were able to persist for many hours in the very unstable conditions present in these cases. The impact of the assimilated radar data on the location of precipitation was assessed using the equitable threat score (ETS) and the displacement and amplitude score (DAS). The impact times for these measures were consistently shorter than for total precipitation, but again shortest for the equilibrium conditions. Copyright (c) 2011 Royal Meteorological Society