Stochastic Hydrology Revisited

Thispaper reviews some of the stochastic methods used in applied hydrology over thepast thirtyyears, theperiod during which thepower and availability of computers grew rapidly, and methods of time-series modelling and simulation came into use which hadpreviously been computationally prohibitive. Where stochastic methods are used to estimate thefrequencies with which extreme hydrological events (floods, droughts) will occur in thefuture, these methods assume that hydrological processes are stationary, so that rainfall and runoff recordsfrom past years can be used to estimate how often extreme events will occur in thefuture. But where there are changes in land use or climate, hydrologicalprocesses also change, and the past may not be a good guide to thefuture. In South America, there have been extensive changes in land use during the last thirtyyears, and there is increasing evidence that climate is also changing. Standard hydrological procedures, such as estimating annual events with T-year return period, and regionalization of annualfloods, then become inappropriate. Thepaper argues that under conditions of climate and land-use change, good assessment of thefuture frequency of extremes must await better knowledge of thephysical processes that determine the behaviour of atmosphere and the oceans.