On Risk Analysis of Water Resources Systems Under Non-Stationary Conditions

Abstract

Traditionally, uncertainty-related analyses of water resources systems, such as flood frequency analyses for mitigation, are performed under stationary conditions, where, statistical properties, such as the means and variances of random variables involved are assumed to be constant with time. In some cases, due to natural and artificial influences, hydrometric data are reportedly experiencing shifts, trends or other changes, even on an annual time scale. In identifying sustainable management solutions for water resources systems it is important to recognize impacts of such changes on risks of system failure. This information may be particularly valuable for long-term planning of water resources projects. Methods of assessing risks of water resources systems are summarized herein. The work identifies risk analyses for systems with different characteristics, static or dynamic, and nonrepairable or repairable. It is shown that a stochastic point process is an effective tool for risk analyses of systems characterized by non-stationary conditions. Risk analyses of repairable systems with long-term non-stationarities, representative of many cases in the water resources engineering, have not been extensively investigated. A marked inhomogeneous alternating renewal process is shown to be suitable for such cases, and the discussion of this process presented herein provides a foundation for further exploration of its applicability.