Water in a changing climate: implications for water resources design (Invited)

The design of water resource projects, whether it be a road culvert, storm water detention basin or reservoir is based on long term statistics of hydrologic variables, usually river discharge and precipitation. The assumption of statistical stationarity arises from the need in design to control the risks of failure for the built structures – too little water in the reservoir to meet water demands, controlling and capturing flood waters, making the levees high enough, and so forth. Methods in the statistical hydrology literature exist to assess and consider low frequency persistence and heavy tailed statistical distributions in water resources design. The problem is that procedures for estimating changes in the mean, variance and temporal correlation of the distributions, and incorporating these effects to reflect changing climate are non-existent or limited. While the climate change community relies on projections from climate models, it is unclear that such models capture observed statistical characteristics that impact water resources design, thereby resulting in structures with reliabilities different than their design values.

In this presentation, two aspects of the problem will be examined: (i) whether climate model simulations of current (20th C) and projected (21st C) climate captures the heavy tailed distributions observed by statistical hydrologists in analyses carried out in the 1970's and 80's (e.g. Matalas et al., 1975 Regional skew in search of a parent, Wat. Resour. Res., 11:815-826), and the implications of this for estimating the risk from extreme events; and (ii) whether climate model simulations of current (20th C) and projected (21st C) climate captures observed persistence in the river discharge records – the so called “Hurst Effect” (Hurst, 1951 Long-term storage capacity of reservoirs Trans. Am Soc. Civil Eng. 116:770:799). The degree of temporal persistence (low frequency correlation) significantly affects the reliability of reservoirs, and whether climate models capture such persistence in 20th C runs, and whether the persistence will increase or decrease in the 21st C is of fundamental importance to hydrology and climate science. Examples of the effect of these statistical characteristics on design will be presented.