Modelling flow detention through temporary storage in the landscape using Artificial Neural Networks - A case study of Lake Mälaren

Abstract: Several strategies can be used to mitigate flood risks, many of them involves construction of embankments around threatened locations or attempts to spread the flood wave in time and space. In this thesis, the effects of a mitigation strategy called eco-flooding are studied in the catchment of Lake Mälaren. The methodology involves a hypothetical storage in the terrain used as an off-line detention basin for the peak flow. Artificial Neural Networks (ANN) is used as a modelling tool and the potential of using it for flow detention modelling is discussed. Two models were constructed, one describing the entire catchment of Lake Mälaren and one describing the sub-basin of Kolbäcksån. Results show that very large surfaces would have to be submerged in order to achieve a significant flow reduction, even in the relatively small sub catchment of Kolbäcksån. Finding the optimal combination of timing, duration and size of the diversion is also important for the mitigation success. The modelling technique showed to have some drawbacks, for example difficulties with introducing a reservoir into the ANN-model.