12 th International Conference on Urban Drainage, Porto Alegre/Brazil, 11-16 September 2011 Hansen et al. 1 Flow Forecasting in Urban Drainage Systems using Deterministic Updating of Water Levels in Distributed Hydraulic Models L.S. Hansen 1,2 *, M. Borup 1 , A. Møller 3 , P.S. Mikkelsen 1 1 Department of Environmental Engineering (DTU Environment), Technical university of Denmark, Miljøvej, Building 113, Denmark 2 Krüger A/S, Gladsaxevej 363, 2860 Søborg, Denmark 3 DHI Denmark, Agern Allé 5, 2070 Hørsholm, Denmark *Corresponding author, e-mail Lisbets11@hotmail.com ABSTRACT There is a growing need for generating more precise model simulations of urban drainage systems, both in off-line and on-line situations. In order to generate these improved model simulations data assimilation tools are needed that make it possible to include system measurements in the models to eliminate some of the unavoidable discrepancies between model and reality. The latter can partly be achieved by using the commercial tool MOUSE UPDATE, which is capable of inserting measured water levels from the system into the distributed, physically based MOUSE model. This study evaluates and documents the performance of the updating procedure for flow forecasting. Measured water levels in combination with rain gauge input are used as basis for the evaluation. When compared to simulations without updating, the results show that it is possible to obtain an improvement in the 20 minute forecast of the water level in an updated node and in the 3 hour forecast of flow through a downstream node. Our results indicate that updating produces better forecasts when implemented in a network with slow flow dynamics and with measurements from basins that are located upstream and contribute significantly to the flow at the forecast location. KEYWORDS Urban drainage systems; flow forecasting; data assimilation; deterministic updating of water levels. INTRODUCTION For any urban runoff model there will always be some deviations between the observed and simulated flows and water levels within the system. This can be due to groundwater infiltration, leaky pipes, poor precipitation or flow measurements, etc. or due to the necessary assumptions used when making the model. For any online model that is used for decision making it is crucial to keep the model in touch with reality. For simple models, this task would usually be performed by a classical data assimilation tool like the Kalman filter (Kalman, 1960). The nature of distributed, physically based urban runoff models like MOUSE from DHI, however, makes the implementation of classical data assimilation tools impossible or extremely computationally burdensome. MOUSE UPDATE is a pragmatic and ready-to-use tool that assimilates water level or flow measurements into the MOUSE model, thus ensuring that the simulations are in accordance with the available measurements. It should thereby result in an improvement in model forecasting performance. The tool works by adding an additional flow to the model, which