Applications of conditional value-at-risk to water resources management.

Abstract

In this thesis I develop mathematical models of freshwater resources and assess the application of a risk measure, Conditional Value-at-Risk, as a criterion for making decisions on the allocation of these resources. The nature of hydrological systems is such that they are well represented by stochastic models. The models considered are: time simulation; stochastic and deterministic linear programming; and stochastic dynamic programming. The hydrological applications are: draw down of dams; allocation and blending of water resources; operation of a small-scale solar-powered desalination plant; and insurance against fishery and crop shortfall. In water resource applications, optimisation models usually have the goal of maximising expected return, or utility, but here I demonstrate that the minimisation of the risk metric is a relevant additional criterion to expected return for water resource management.