Deterministic nowcasting based on Lagrangian extrapolation of recent radar observations is a technique used in many operational and research centers and has proven to be useful at extending the anticipation with which flows can be forecasted with a rainfall–runoff model. However, the improvement in flow forecasts is not known a priory and case-dependent, mostly depending on the predictability of the rainfall field. In this work a probabilistic technique has been developed to automatically quantify the uncertainty in radar-based nowcasts obtained by Lagrangian extrapolation to be used in hydrology. The technique follows the ensemble approach: it generates a number of realizations of rainfall forecasts compatible with observations and that reserve the space and time structure of the rainfall field according to the String of Beads model. The results show that the technique reasonably reproduces the evolution of the rainfall field, but that, overall, the errors are underestimated in part due to the fact that the uncertainty in rainfall forecasts due to errors in the motion field is neglected in this current version of the method.

Peer Reviewed