Flood forecasting for the upper and middle Odra River basin

Abstract

The objective of the authors’ work in the area of flood forecasting and distributed modelling is to determine how different model formulations and different rainfall inputs contribute to forecast accuracy and uncertainty. To address these issues within the EU project FLOODRELIEF, a comprehensive distributed catchment model has been formulated and tested for the upper and middle Odra River. Within the FLOODRELIEF project multiple distributed modelling approaches, including the model presented here, are being developed for the Odra basin. The objectives of developing these different modelling approaches are; to evaluate the performance of these different models in representing the catchment processes and river flows, to examine the effect of model structure on the character of flow simulation and prediction uncertainty and to examine the sensitivity of these models to different rainfall input The Odra basin was selected for these analyses as a flood-prone catchment representing highly developed European catchments where comprehensive modelling of the river system, flood plains, polder subsystems, and structures as well as rainfall-runoff and snowmelt processes in the tributary catchments are required. Flood forecasting in the Odra requires both fast and reliable simulations for this complicated river basin and therefore a careful balance between accurate representation of the catchment flood processes, the flood wave movement and inundation extent and the need for rapid forecasts This paper describes the formulation, calibration, validation and real-time implementation of an operational distributed model for flood forecasting for the upper and middle Odra. This application shows that the MIKE 11 distributed model is able to reproduce the large-scale rainfall-runoff processes and the propagation of the flood wave through the main river channel system including the catastrophic flood of July 1997. This model is therefore suitable as a reference for the evaluations of different models and model structures in our investigations of simulation and forecast uncertainty. A model post-audit was carried out which identified cases where simplified representations in the rainfall-runoff models that performed well in the calibration, requires revision for the validation period. This type of post-audit analysis is extremely valuable in evaluating model performance and ensuring continuing improvement in flood forecasting accuracy.