Despite frequent use of estimated flood discharges from urban and small rural catchments, the assessment of uncertainty in instantaneous discharge due to inaccurate head measurements is generally not performed. The present work discusses the impact of inaccurate head measurement on observed instantaneous flood discharge. The sources and sizes of uncertainties in head determination are discussed. A simple uncertainty model is developed in order to quantify the uncertainty in error-corrupted instantaneous and mean discharge. It is shown that uncertainty in mean discharge is less affected by uncorrelated head measurement error, while instantaneous discharges are substantially affected. Practical modelling using data from a Norwegian urban gauging station is undertaken. It is shown that the estimated annual maximum instantaneous flood discharge from this watershed has 95% confidence limits with sizes of 17–19% of the estimated discharge for average head measurement accuracy. Furthermore, it is shown that variability in head determination causes only a minor potential for bias in estimated instantaneous flood population statistics and quantiles.
Hydrometric gauging stations are often sited in reservoirs such as lakes and river pools since they possess favourable features for streamflow determination. By applying the equations which govern critical flow from a reservoir along with a spline representation of the known geometric characteristics of the reservoir outlet, a new method is developed for estimating the stage-discharge relationship in reservoir-like situations. The proposed rating curve model is amenable to least squares estimation using the available stage-discharge measurements. The proposed model is shown by theory to be general, and includes both normal uniform flow from a reservoir and the classical power-law rating curve model as special cases. Using geometric data and stage-discharge measurements from two Norwegian gauging stations, it is demonstrated that (a) the proposed model is more appropriate than the classical power-law rating curve model when performance is measured by the AIC statistic and the sizes of the jackknife standard deviations of the predicted discharges, and (b) the proposed model displays a superior performance over the power-law model in both upwards and downwards extrapolation.