CLIMATE VARIABILITY AND FLOOD FREQUENCY AT DECADAL TO MILLENNIAL TIME SCALES

Among climate specialists the prevailing view is that climate system behavior routinely leads to nonstationary time series of its component elements. Since climatic processes govern probability distribution properties, from which weather and hydrologic events are "drawn", it is expected that flood series would similarly exhibit nonstationarity. Noting changes at decadal to millennial scales in flood stages, volumes, and temporal clustering, this conclusion is obvious to fluvial specialists, even considering changes in sediment supply, which too are related to climate. The evidence arises from well-dated floodplain and slackwater alluvial stratigraphy, and level variations in lakes fed mainly by floods. In semiarid areas, with high total annual discharge correlated with annual peak discharge, even tree-ring chronologies can furnish valuable and well-dated information on flood frequency variations. This is true wherever rare and persistent weather events are the source of exceptional tree growth and large floods. There is a growing body of paleohydrology and paleoclimate data and knowledge that can enhance our understanding of the climate system and how it varies. Even at decadal scales of the 20th century climate nonstationarity is found to cascade into flood records, where frequency and return statistics also change in time. The challenge lies in identifying suitable hydroclimatic areas where simultaneous signals of climatic parameters and floods are available. In some cases, lack of such recognition may simply reflect the real lack of an effect. In others, however, lack of findings may be ascribed to use of unsuitable tools or to lack of a search effort. Flood series are also often a result of drivers that have mixed or multiple climatic origins. In practice, it is easier and definitely more convenient to assume statistical stationarity, but scientifically this is not an accurate depiction of reality. Because the flood events of interest are in the future, flood frequency analysis is essentially an exercise in forecasting. The quality of forecasts of these future distributions in turn is a direct reflection of our understanding of how processes at all scales, local to global, lead to floods. Thus, efforts to recognize nonstationarity must involve expertise from climatology, hydrology, and geology.