Optimal risk-based decisions by maximizing extraction of environmental information for streamflow forecasts

Water is fickle, water flows are unpredictable. Yet, we have never been in such dire need of accurate streamflow forecasts to manage water systems! Operating the latter means relying on a sequence of timely decisions to obtain optimal benefits and minimize risks and uncertainties. In the case of hydropower reservoirs for instance, the daily water releases should both maximize power production benefits, and offer sufficient flood-storage to guard against flood risks.
To effectively combine real-time information and understanding of physical processes involved in droughts and floods, Dr.Weijs chose to focus on the Val Ferret watershed in the Alps. By balancing information flows from intensive measurements and state-of the art models, Weijs’s methods will be essential to move from detailed physical understanding to probabilistic streamflow forecasts that can be produced cost-effectively and inform decisions reducing flood and drought risks.
My research focuses on analysing how detailed measurements of environmental variables such as rainfall, soil moisture and temperature, contribute information to forecasts of river flows further downstream and ultimately contribute to risk-based informed decisions. To perform this analysis, the framework of information theory is used to follow information in its flow from observation to decision. Taking this integrated view can lead to new insights about model calibration, sensor network design, and probabilistic forecasting.