Flood & drought forecasting

Flood and drought forecasting

Research developed at Irstea in Antony has focused on the development of flood forecasting systems, based on continuous rainfall-runoff models.

The hourly GRP model was introduced in the SOPHIE2 platform, which is used by most of the national local flood forecasting centers (SPC) in France (work carried out as part of the “Multi-risk” Program DGPR – IRSTEA).

Research questions on flood forecasting are studied in detail in our current research topics.

Our research interests are:

• to develop reliable methods for probabilistic streamflow forecast from ensemble weather predictions to better anticipate flood events and improve reservoir inflows management (Zalachori et al., 2012),
• to quantify the uncertainties associated with the forecasts (PhD thesis F. Bourgin, 2014; van Pham, 2011),
• to improve the quality of streamflow forecasts by assimilating new data and developing new techniques for model updating in real-time (Berthet, 2010),
• to develop methods for flood forecasting in poorly gauged or ungauged catchments (Randrianasolo et al., 2011),
• to develop multi-model approaches in hydrological forecasting (Seiller et al., 2012; Velazquez et al., 2011),
• to study the communication of uncertainties in the hydrologic forecasting chain and the value of probabilistic forecasts (Ramos et al., 2010, 2012).

Our works are carried out in partnership with SCHAPI, Météo-France, EDF, Laval University, University of Twente, among others, and contribute to the international initiative HEPEX.

Drought forecasting

While flood forecasting deals with short lead times (few hours up to few days), forecasting of low flows is carried out for lead times ranging from several weeks to several months. It is a matter of anticipating potential limitations on water resources availability given the current soil moisture state of a catchment.

Due to the uncertainties on future precipitation, the forecast is performed following a probabilistic approach, based on the use of various rainfall scenarios (observed in the past or obtained with a stochastic rainfall generator). This method allows to estimate confidence intervals on probable future discharges for the target forecast period.

Our research focuses on:

• improving the ability of hydrologic models to simulate low flow periods and anticipate drought situations (Pushpalatha et al., 2011, 2012),
• contributing to a better management of water resources in multi-purposes reservoirs by developing tools to help in low flow management and climate change adaptation (PhD thesis L. Crochemore, 2016).