On quality indicators for radar-based river flow forecasts

Although there has been considerable progress in improving the quality of radar estimates of precipitation, including developments in radar technology, the unpredictable nature of radar errors continues to discourage many operational hydrologists from using radar data quantitatively as input to models. Indeed, the tendency has been to wait for all errors to be removed. Unfortunately it is unlikely that all errors will be removed all of the time, even with new promising radar technologies such as dual polarisation systems. In this paper a procedure is proposed and partially tested for deriving a quality index from single polarisation radar based on the use of a ‘peaks-over-threshold’ (POT) parameter for gauged catchments. The procedure is tested for several different catchments and for different rainfall types. For ungauged catchments a different approach is proposed based on the fitting and analysis of an autoregressive model to rainfall time series. This procedure is tested using the radar data distributed to users in real time during major flood events in the Midlands, UK. The practicality of applying these techniques to radar data in an operational environment is discussed.