Forecasting precipitation for hydroelectric power management: how to exploit GCM's seasonal ensemble forecasts

Abstract

AbstractThe EDF group is the biggest French electric power producer and distributor. Its activities are greatly related to weather and climate. In particular, optimal management of the hydroelectric power production system requires a good forecast of water resources, from several days to several months in advance. Currently, only climatology at the seasonal timescale is used for operational production management. Seasonal probabilistic forecasts would improve watershed management at some months' lead‐time if they are skilful enough. For this, two main problems have to be addressed: first, direct precipitation forecasts at this timescale have little, but positive, skill over Europe; second, the spatial scales of seasonal forecasting models are not adequate to predict local precipitation at the river basin scale. This study aims to evaluate the quality of seasonal forecasts of precipitation for 48 catchments in southern France. These are obtained by spatially downscaling global scale seasonal forecasts of geopotential height at 850 hPa. The method used is based on singular value decomposition and multiple linear regression. The statistical downscaling model is calculated from 45 years of observed local precipitation in the watersheds and geopotential fields from ERA40 re‐analysis data. The statistical model is then applied to the seasonal hindcasts from the DEMETER project. Two main results arise from this work. First, we show that it is possible to obtain useful and valuable information for EDF at the local scale from global seasonal averaged information. Second, we find that only a probabilistic multi‐model ensemble forecast approach provides useful information for EDF catchments, even with quite low skill, and that a deterministic approach, using only the ensemble mean of the forecasts, is not better than a forecast based on climatology. It has, nevertheless, to be pointed out that for operational purposes, being able to know that a forecast for a given location or date is not reliable is, in itself, valuable information. Copyright © 2007 Royal Meteorological Society