Estimating the unit hydrograph and effective rainfall from observed output

The high uncertainty in estimated areal rainfall and potential evapotranspiration, as well as the observed streamflow means that it is difficult to identify an appropriate model structure and calibrate the parameter values. Improvements in model identification requires minimizing the impacts of the uncertainty in these quantities. This paper presents recent work on using observed streamflow and rainfall data to gain insight into the shape of the unit hydrograph (impulse response function), as well as estimating both the unit hydrograph and effective rainfall. The first method looks at the potential information available from using a Fourier deconvolution approach for estimating the unit hydrograph. While it would be normal to take a straight forward approach to deconvolution (using the rainfall and streamflow timeseries as the inputs), the inputs in this case were the correlation function of streamflow with rainfall and the autocorrelation of rainfall (Croke et al, 2011, Hydrology Research, 42(5), 372-385). The advantage of this approach is that most of the temporal variation in the response signature is averaged before undertaking the convolution, resulting in reduced noise in the resulting deconvolved response curve. The resulting impulse response function includes the unit hydrograph response curve as well as the impact of wetting/drying of the catchment. The signal at negative lags is entirely due to the wetting/drying processes, while the signal for positive lags is a combination of the two sources. This allows investigation of both parts of the catchment response to be explored. Examples of the approach applied to salt transport as well as streamflow generation will be presented.