Estimating the added value of seasonal forecasting systems through hydro-economic stochastic programming and model predictive control in the Jucar River Basin (Spain)

To better communicate forecast skill, there is a need to move beyond traditional statistical metrics and towards user-tailored metrics, i.e. targeting economic gain. Here, we highlight this need and present an investigation based on hydro-economic stochastic programming (in particular the SDDP algorithm) and model predictive control (MPC). SDDP provides long-term average benefit functions that are introduced as boundary conditions to a MPC, which further uses seasonal hydrological forecasts as inputs. Resulting economic benefits are compared and contrasted with traditional skill metrics to rank forecasting systems.

The method is applied to the Jucar River Basin (Spain), characterized by a distinct water use, irregular hydrology and multiannual droughts. Monthly-scale SDDP and MPC models are assessed using different systems: 1) average hydrological discharge (no forecast); 2) ensemble of historical streamflows; 3) post-processed (two different methods) seasonal hydrological forecasts from the E-HYPE pan-European hydrological model (based on ECMWF System 4 meteorological forecasts); 4) locally-adjusted hydrological models forced by bias-adjusted meteorological forecasts from Copernicus C3S (ECMWF SEAS5, UKMO GloSEA5 and MeteoFrance System6); and 5) perfect forecasts. System performance with current practices is also computed.

Results show that SDDP and MPC algorithms forced by seasonal forecasts could improve the current operating rules. Up to 75% of the gap between current practices and optimization fed by perfect forecasts could be covered by some of the systems analysed. Furthermore, improved skill based on the traditional metrics does not always translate into economic gains.

Acknowledgements:

This study has been supported by the European Union's Horizon 2020 research and innovation programme under the IMPREX project (grant agreement no: 641.811); the European Research Area for Climate Services programme (ER4CS) and the Agencia Estatal de Investigación (Ministerio de Ciencia, Innovación y Universidades) under the INNOVA project (grant agreement: 690462; PCIN-2017-066); and the postdoctoral program of Universitat Politècnica de València (UPV).