Radar-based nowcasting is of crucial importance for providing warnings of imminent severe weather conditions, such as flooding, storms, cyclones and hurricanes, that can endanger populations and damage strategic economic infrastructure.
The forecast of precipitation is, however, a most challenging task. Since it is governed by complex microphysical processes, the exact circumstances for a cloud to form precipitation are still not fully understood. But also on a macro-scale, the formation of precipitation is linked to very complex dynamical processes of the atmosphere. Moreover, precipitation can be a very local phenomenon, e.g., in the case of an isolated convective event, acting on scales that are sometimes even smaller than the grid scale of the most recent Numerical Weather Prediction (NWP) models. Due to these incompatible scales of precipitating structures and (most) NWP models, the capabilities of the existing NWP models in the prediction of precipitation at a specific location are rather limited. Moreover, an incomplete data assimilation in the stage of the initialisation of an NWP model can lead to unreliable precipitation predictions for short lead times of 0-6 h, since the model does not cover adequately enough the present situation.
The abovementioned shortcomings of NWP models indicate that Quantitative Precipitation Forecasts (QPFs) for short lead-times clearly require a different strategy. Radar observations provide precipitation maps with high spatial (typically 1 km) and temporal (5 min) resolution, and are therefore ideally suited to act as the basic building blocks of operational precipitation nowcasts.