Uncertainty and Sensitivity Analysis in Reservoir Modeling: a Monte Carlo Simulation Approach

Water resource modelling plays a crucial role in water resources management, but it involves many inherent uncertainties. This research investigates how epistemic uncertainties affect reservoir water budgets, projecting forward over a 30 year period using Monte Carlo simulation. It encompasses long-term variations in water demand, reservoir volume, precipitation, evaporation and inflow, while also considering siltation processes, reservoir dredging, population growth, reduced water consumption, and the hydrological impacts of climate change. The research focuses on fifty reservoirs in a semi-arid region of Brazil. The findings demonstrate that some reservoirs consistently met their demands with high level reliability, even within a wide range of uncertainty. Conversely, reservoirs with morphohydric indices indicating a tendency toward water scarcity are significantly affected by input variability introduced through uncertainty analysis. An empirical model is proposed to estimate the probability of these reservoirs achieving the reference volume reliability of 90%, while considering the uncertainties of: annual average inflow, reservoir maximum volume and annual demand. Sensitivity analysis reveals that reservoir inflow and demand are the two most influential variables affecting a reservoirs’ ability to meet its demand. For over exploited reservoirs, variations in these variables strongly impact the volume reliability. This research provides a valuable tool for estimating the likelihood of reaching a 90% volume reliability, while taking into account the inherent uncertainties in the modeling process. Additionally, it identifies key variables that have the most influence on the reservoirs’ ability to meet its demand. Notably, this study conducts uncertainty and sensitivity analyses in the context of physical and hydrological reservoir features for a large number of reservoirs, adding novelty to the research field.