A sediment rating curve developed as a linear regression on logged values which is back-transformed must be corrected for the bias introduced by the log transformation. This article shows that the variances are identical for linear regressions based on values of logged load and logged concentration from the same data set. This means that the bias correction factoss 101.1513σ2 for the back-transformed regressions are equivalent. Therefore a back-tranoformed log regression based on loads corrected for bias gives identical sedimett discharges to a back-transformed log regression on concentrations corrected for bias. Regression equations from gauging stations in two neighbouring basins in Costa Rica confirm this conclusion.
Mean loads for individual discharge classes were plotted on diagrams in log scales to find the points where the sedimett rating curve changes direction. When sediment rating curves were developed on logged mean concentrations, water discharge weighted mean concentrations had to be determined in order to produce equations comparable to those on logged mean loads. Consequently, discharge weighted mean concentrations must be used in a plot to determine the change in direction of a sedimett rating curve and to check the goodnsss of fit of a regression developed by any model employing concentration as the dependent variabe.
- Received February 17, 1997.
- Revision received July 7, 1997.
- Accepted July 15, 1997.
- © IWA Publishing 1997