On quality indicators for radar-based river flow forecasts

The use of real-time weather radar for operational flash flood warning has been hindered by a lack of information on the quality of river hydrographs derived using these data. In spite of considerable improvements in radar quality control procedures, including the advent of algorithms based upon multiparameter radar data, errors in radar estimates of rainfall are likely to remain, and vary considerably between rainfall types. In this paper a procedure is proposed for deriving a quality index based upon the use of a “Peak-Over-Threshold” (POT) parameter, and the fitting of an autoregressive model to rainfall or model river flow time series. The ‘goodness of fit' of the model with the POT values is taken as an indicator of the quality of the river flow forecasts. The procedure is tested using both stratiform and convective rainfall case studies which led to severe flooding. The importance of assessing whether hydrographs are reliable is discussed in the context provided by the case studies. The need for further work on quality indices is considered.