Stochastic Assessment of Water Resource System Vulnerability to Multi-year Drought

Abstract:

Global climate models suggest an increase in evapotranspiration in many parts of the world which is likely to cause an increase in drought severity, yet the weakness of climate models in modelling persistence of hydro-climatic variables and the uncertainties associated with regional climate projections mean that impact assessments based on climate model output may underestimate the risk of multi-year droughts. In this paper we propose a vulnerability-based approach to test water resource system response to drought. We generate a large number of synthetic streamflow series with different drought durations and deficits and use them as input to a water resource system model.

Two approaches to drought simulation are presented: (1) using a numerical streamflow generator based upon an optimal bootstrapping algorithm and (2) using a copula to characterise the joint probability distribution streamflow characteristics in successive months. Droughts with longer durations and larger deficits than the observed record are generated (1) by changing the objective function of the optimisation and (2) by perturbing the copula dependence parameter and by adopting an importance sampling strategy for low flows. In this way potential climate-induced changes in monthly hydrological persistence are factored into the vulnerability analysis.

The method is applied to the London water resource system (England) to investigate under which drought conditions severe water use restrictions would need to be imposed on water users. Results indicate that the water resource system is vulnerable to drought conditions outside the range of historical events. The vulnerability assessment results were coupled with climate model information to compare alternative water management options with respect to their vulnerability to increasingly long and severe drought conditions.