Runoff modelling under noise-corrupted rainfall conditions

Abstract

Hydrological models are now routinely used in planning, design, operation, and control of water resources systems. However, all models, no matter how complex, are approximations of the real world and consequently are subject to various levels of errors. The analysis of uncertainty in hydrological models can provide valuable insight into the limitations and advantages of various surface runoff models. The benefits derived from such an analysis are many: first, it provides the modeller with a direct estimate of runoff prediction errors under specific rainfall conditions; second, it enables the user to analyze the trade-offs between different rainfall-runoff models; and finally, it can provide useful information for the design of data collection systems designed to achieve a given level of performance. This paper describes the application of uncertainty analysis to rainfall-runoff modelling under noise-corrupted rainfall conditions. The statistical properties of surface runoff subject to noise-corrupted rainfall conditions are examined. Methods of analysis described in this paper include (a) derived probability distribution, (b) first-order analysis, and (c) Monte Carlo simulations. The techniques are applied to linear and nonlinear runoff models, including the unit hydrograph, and the Soil Conservation Service model. Key words: runoff, uncertainty, error analysis, statistics, stochastic modelling, first-order analysis, Monte Carlo simulation, SCS model, unit hydrograph.