Probabilistic modeling of the water availability in a large surface reservoir in the Netherlands

During dry spells, a large part of the Netherlands depends on water from the IJssel lake, a large surface water reservoir. Water is extracted for a number of purposes, such as irrigation, water quality, shipping and drinking water. Besides local precipitation, the main source of water flowing into the lake is the river IJssel; a distributary of the Rhine. During periods of low discharge and low precipitation, water shortages may occur, as the recent summer of 2018 showed. ​We develop a probabilistic model to simulate water availability in the lake during dry spells. We derive marginal distributions of precipitation, open water evaporation, river discharge and water intake from the surrounding region, based on a 101-year simulation of the deterministic Dutch national water model. We assess the plausibility of the resulting extreme tail of the distributions by comparing them to values based on the ECWMF seasonal reforecasting archive, which, when all ensemble members, years and lead times are combined, contains over 4,000 years of data. All correlations between the four terms are modeled using a four-dimensional copula. The resulting distributions of water availability show aggregated water shortages up to extremely dry (return periods in excess of 10,000 years) conditions. Lake level dynamics are, during dry conditions, dominated by high water demand from the surrounding region (caused by lack of local precipitation) and low IJssel river discharges. A coincidence of these two terms causes the most extreme shortages. Because model is conceptually relatively simple, it is able to run a large number of realizations and is potentially highly suitable for, for example, assessment of measure effectiveness.