Appropriate flood forecasting for reservoir operation


Xiaohua Dong

Marjolein Dohmen-Janssen

Martijn Booij

Suzanne Hulscher


University of Twente


KNAW and University of Twente


December 2000 – December 2004


Appropriate modelling seeks for a complexity-accuracy-uncertainty consistent system that is as simple as possible, but compatible for its task. Appropriate flood forecasting methods have a sound practical background, they are oriented to satisfy the requirements of reservoir operation. The specifications of the models are determined by the physical characteristics of the river basin and the requirements from the reservoir operation, i.e., which hydrological processes should be considered. In order to appropriately operate a single model, a number of influencing factors need to be specified: spatio-temporal resolution, model complexity, uncertainties, and their mutual relationships and consistencies.


The objective of this research is to answer the following questions: How to find the appropriate flood forecasting models, and how to apply them in an appropriate way to fulfill the requirements of reservoir operation.


First the appropriate flood forecasting models are selected. The criteria of selecting the appropriate flood forecasting models are first of all based on requirements from the reservoir operation, and secondly the physical reality of the river basin which will determine if there is a realistic solution toward the requirements.

Second the appropriate application of the models is made by:

−Appropriate spatio-temporal resolution of input data. Sampling the rainfall data properly (both spatially & temporally) depends not only on the local precipitation patterns in the basin, but also on the hydrological response characteristics therein.

−Appropriate model complexity. The amount of the subbasins in the model and the hydrological processes included in each subbasin determine the descerning of “lumped” and “distributed” models, whose appropriate scale need to be identified.

−None of the uncertainties derived from the aspects mentioned above should dominate the output uncertainty. A consistent system could tolerate the variation of uncertainties from the sources, and prevent significant effect on output uncertainty.


X.H.Dong, M.J.Booij and C.M.Dohmen-Janssen, 2003. Appropriate spatial sampling resolution of precipitation for flood forecasting, NCR-days 2003. Nov 06-08, 2003. Roermond, the Netherlands.

X.H.Dong and C.B. Vreugdenhil, 2002. Balancing between generalization and over-fitting: ANN-based modeling for flow forecasting, Proceedings of second international symposium on flood defence (ISFD’2002), Beijing, Sept 10-13, 2002.