Modelling vegetation roughness for river basin management

Background

Several groups have already investigated the influence of vegetation on local flow conditions, both in laboratory flume experiments as in the field. As a result, different descriptions exist that relate physical vegetation parameters (such as stem width, density, flexibility, etc.) to an effective roughness coefficient. All these approaches to describe bed roughness in terms of readily measurable vegetation characteristics may be physically sound. However, it is not yet clear how much can be gained by introducing these methods in the commonly used river management tools. Ultimately, a prediction with the largest possible accuracy is desired, based on parameters that can (easily) be measured in the field. This does not necessarily mean that the largest level of detail is automatically what should be aimed at. For example, a large number of independent physical parameters may introduce a combined uncertainty larger than the uncertainty that results from a smaller number of empirical parameters.

A common practice in modern river management is to calibrate numerical hydraulic models through adjustment of ‘free’ friction parameters. Usually these parameters characterize the boundary roughness in different sections of the river channel. In the calibration procedure the roughness parameters are varied until the numerical model agrees with field data. In order to prevent usage of ‘unrealistic’ roughness values, limits are imposed on the allowed range of these calibration parameters. These limits (‘threshold values’) are usually based on knowledge of occurring boundary roughness in the field. By using such a calibration procedure it is implicitly assumed that only the uncertainty in boundary roughness accounts for the discrepancies between the model predictions and the field situation. In fact, an attempt is made to compensate for all the shortcomings of the numerical model by adjustment of the boundary roughness. Furthermore, a set of calibrated roughness values demands regular updating due to changing conditions in the field. Apart from possible changes in the unknown processes that are swept together in the roughness parameter, it is obvious that vegetation characteristics change over time. The forecasting ability of the model is thus limited by its (static) roughness description.

Objectives

The emphasis in this work is on the appropriate modelling of vegetation roughness: how to incorporate vegetation characteristics in numerical flow models, such that the level of detail of their description, and the corresponding parameter uncertainties, are in balance with the desired precision of predicted water levels. In order to answer this question, current developments in the fields of channel flow modelling, decision support systems and the hydraulic roughness of vegetation need to be combined.