Optimisation meets the manager: a case study of the La Paz reservoir system

Managers must usually apply operating rules to optimise the use of water resources in a sustainable manner. Ideally a manager needs a set of near-optimal dynamic operating rules that are consistent with the objectives and level of risk set by the manager. The traditional approach for a reservoir is to develop fixed (static) rule curves based upon a statistical analysis. However, improved dynamic rules can be derived using optimisation techniques such as genetic algorithms. Also, simulation methods can be used. Here we show how both methods can be applied to generate near-optimal dynamic operating rules for a reservoir system used for drinking water supply in La Paz, Bolivia. In particular, we show how simple practical operating rules can incorporate the level of risk set by the manager. Further, these rules advise how quickly water levels should be altered when they are too high or too low.