years been evidenced primarily in low flows and not in flood flows? Are there inconsistencies in observed precipitation extremes and streamflow extremes, and if so, why? What are the causes of pervasive increases in occurrence of low flows, for example, in the upper Midwest? To what extent is land cover, as contrasted with climate change, the cause of observed streamflow changes?
Assumptions on the occurrence of major hydrologic events to analyze extremes are based on the notion of stationarity, yet observational evidence increasingly shows that this assumption is untenable.
Stationarity represents the idea that hydrologic systems fluctuate in an unchanging envelope of variability (i.e., the mean and the degree of variability of hydrologic time series do not change over time). Water management systems have been traditionally designed based on this assumption. Therefore, it is critical to the protection of life and property to understand if and how these assumptions are being violated (Milly et al., 2008). From a scientific standpoint, fluctuations in stage heights and flood flows over the historical past constitute a natural experiment, with particular realizations that have in some cases been unexpected and changing over time. A good example is the American River in California, where over the past ~100 years the 5 largest three-day peak flood volumes all occurred in the second half of the record, as had 10 of the largest 13 (see also NRC, 1999). In other words, this hydrologic system no longer operates within its expected unchanging envelope of variability. The statistical distribution of flood volumes that represent this system has become non-stationary.
Bulletin 17B of the Interagency Advisory Committee on Water Data (IACWD, 1982), titled Guidelines for Determining Flood Flow Frequency, details a set of data-based methods that allows one to define flood potential. This document is the current standard in the United States, but it has not been updated since the early 1980s. The workshop participants broadly agreed that although Bulletin 17B was concerned about non-stationarity, a remedy is not well addressed in the document. Because the available evidence (at that time) indicated that major climate-induced changes occur on the scale of thousands of years, Bulletin 17B assumed that floods are unaffected by the shorter-term changes that have been documented in the context of anthropogenically induced climate change. Participants discussed the United States’ unmet need for new flood-frequency guidelines that draw on advances in hydrologic and climate science over the past 25 years, an observation that is supported by presentations and agreement at a previous COHS workshop (NRC, 2008). Regular revision of the Bulletin 17B guidelines as modeling and understanding of relevant phenomena improves would also be valuable. Regardless, continuing to use the assumption of stationarity in designing water management systems is no longer practical or defensible.