The influence of autocorrelated errors on Catchment Modelling Systems

Engineers Australia
Publication Type:
Conference Proceeding
The 29th Hydrology and Water Resources Symposium: Water Capital, 2005, pp. 1 - 8
Issue Date:
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail2008007592OK.pdf1.05 MB
Adobe PDF
The calibration of catchment modelling systems (CMSs) is regarded as an important part of the modelling process to ensure the reliability and robustness of the parameters. This typically involves minimising an objective function that measures the goodness of fit between observed and simulated hydrographs, with the mean squared error (MSE) objective function being one of the most popular. The focus of this paper is on the assumptions that underlie this objective function, and in particular, on the assumption that the model errors should be independently distributed. Using the US EPA Storm Water Management Model (SWMM) applied to the Powells Creek Catchment in Sydney, Australia as a case study, it is shown that the model errors were autocorrelated for each of the ten historical rainfall/runoff events that were analysed, and that autocorrelation could be removed by fitting a lag-one autoregressive [AR(1)] model to these errors. When examining the influence of the autocorrelated errors on the model performance, however, it was found that the removal of autocorrelation did not result in any observable improvement, and in fact worsened model performance in some cases. The results of this study therefore suggest that caution should be exercised when using an autoregressive model to meet the assumption of independence during model calibration.
Please use this identifier to cite or link to this item: