Attribution of monotonic trends and change points in peak streamflow across the conterminous United States using a multiple working hypotheses framework, 1941–2015 and 1966–2015

The U.S. Geological Survey has a long history of leading flood-frequency analysis studies. These studies play a critical role in the assessment of risk, protection of lives, and planning and design of flood protection infrastructure. Standard flood-frequency analysis is based on the assumption of stationarity—that is, that the distribution of floods at a given site varies around a particular mean within a particular envelope of variance (and skew) and that these parameters of the underlying statistical distribution representative of the floods do not vary over time. Gradual or abrupt changes in one or more of the distributional parameters are called nonstationarities and violate the underlying assumptions of current U.S. Federal Government guidelines for flood-frequency analysis. Uncertainty exists as to what degree of violations calls for the use of a modified method for flood-frequency analysis and what the modified method(s) should be.

When deciding whether to perform nonstationary flood-frequency analysis and choosing a method for such analysis, it is important to understand the causes of the nonstationarity. Gradual or abrupt changes in distributional properties of floods may be the result of numerous factors, such as regulation, diversion, land-use change, or climate change.

In the interest of developing a cohesive national approach for better understanding the causes of nonstationarities and incorporating potential or observed changes into flood-frequency estimates, subject-matter experts from the U.S. Geological Survey and cooperators worked together to develop a multiple working hypotheses framework for making attributions and a common vocabulary for making provisions of confidence. Seven regional teams of these experts used ancillary datasets and institutional knowledge to evaluate plausible causes for monotonic trends and change points in annual peak-streamflow data for the conterminous United States that had been identified in an earlier phase of the project.

The first chapter of this professional paper describes the development of a list of the potential attributions, presents a literature review of the potential attributions, describes the regional approach, summarizes insights obtained from the attribution process, and suggests future research. The other chapters provide the methods used for attribution in the seven regions—Pacific Northwest, Upper Plains, Midwest, Northeast, Southwest, South-Central, and Southeast—and summarize the regional patterns of nonstationarities.