A sensitivity analysis applied to morphological computations

Abstract

In river engineering morphological predictions have to be made to study the implications of changes in a river system due to natural causes or human interference. It regards here time-depending processes. Characteristic parameters of the river have to be forecasted both in time and space. The morphological processes, however, are extremely complex and therefore a substantial degree of schematization is required before e.g. mathematical models can be applied to obtain the predictions wanted. Information about available mathematical models can e.g. be obtained from Jansen (1979) and Klaassen et al (1982). The present physical-mathematical formulation of the morphological problems involved is incomplete. For instance the variation of the width B(x,t) cannot yet be predicted. Therefore in this paper the restriction is made that (relatively) unerodible banks are present. The description of the problem in two (horizontal) space dimensions has not yet led to mathematical models that are used on a routine basis. During the last two decades or so, however, one-dimensional models have been developed gradually to useful tools for practical problems. In these models average values across the width of the river are predicted for the waterlevel h(x,t), the bedlevel z(x,t) and consequently for the water depth a(x,t). There is concern, however, regarding the accuracy of these predictions. Very few possibilities exist to calibrate and verify the mathematical models for a particular river. There are a large number of error sources of which here basicly two will be discussed. Sediment transport s(x,t) has to be predicted from the local hydraulic conditions. Alluvial roughness, for instance expressed in the Chezy coefficient C(x,t), also has to be forecast locally. The available transport predictors and roughness predictors are based on the presence of steady uniform flow. Hence there is already a potential source of errors in applying these predictions in a mathematical model with nonsteady and non-uniform flow. The two types of predictions are linked. In many transport predictors (transport formulae) the alluvial roughness has to be known. As future conditions are considered, this roughness also has to be predicted.