Short-term flood risk prediction: A comparison of the Cox Regression Model and a Conditional Distribution Model


  • M. R. Futter,

  • J. A. Mawdsley,

  • A. V. Metcalfe


Two models for estimating the risk of a flood exceeding some critical threshold within a few days (Smith and Karr, 1986; Ettrick et al., 1987), which take account of the season and prevailing catchment conditions, have recently been published. The models are fitted to a 1000-year synthetic data set, to compare the results with empirically determined risk estimates. After some modifications to the Smith and Karr method both models demonstrated reasonable accuracy. A second comparison is then made using data from a United Kingdom catchment. The Smith and Karr model was further modified to allow risk estimates for up to 30 days ahead to be made. Both models were applied to both data sets for 30 day ahead estimates, and the results demonstrate the sensitivity of the risk to prevailing conditions at the beginning of the period. The assumptions, data requirements, and accuracy of the models are compared and discussed.