Caprices du climat et de l'hydrologie en Haute-Durance : nos prévisions d'apports plurimensuels sont-elles encore fiables ?

Click to increase image sizeClick to decrease image size

The water volume contributed by a river over the coming months is a function, on the one hand, of the drainage basin's initial hydrological status and, on the other, of meteorological events over the coming period. In the case of rivers experiencing winter conditions, the fraction of flow contributed by the melting season, explained by the amount of snow that fell several months previously, is enough to justify our attempt to perform long-term prediction of spring contributions in order to optimise management of mountain water reserves. Due to the lack of long-term meteorological predictions, other than precipitation and snow measurements taken during the accumulation season, EDF has adopted a pseudo-stationary climatic hypothesis: the years over which models are calibrated are assumed to be statistically representative of future conditions. However, we know that the climate experienced natural fluctuations in the past that could, one day, be overlaid with the tangible effects of changes that are anthropogenic in origin. In order to reassure ourselves of the rationality of long-term supply prediction tools, we therefore ensure here, using the example of the Durance basin at Serre-Ponçon, that climate and hydrological events have experienced a certain temporal homogeneity since the beginning of the model calibration period. The significant inter-annual variability highlighted makes it impossible to reject a 40-year period steadiness hypothesis and on such a reduced area. For example, over the past decade, snow cover and therefore the contribution from melting snow was rather small without this "anomaly" resulting in an overall significance greater than that noted in the past during a rather dry or rather wet decade. Finally, following a systematic analysis of rain, temperature, snow cover and flow records, supply forecasting model performance is then analysed in turn. At this time, the tools used are entirely suited to the current climate context.