Repository logo

Mixed populations flood frequency analysis in the mid-Atlantic region of the United States


In many parts of the United States, floods at a single site are caused by multiple mechanisms. Flood mechanisms can broadly be classified as meteorologic, land surface processes, and disturbances. These non-homogeneous flood series are typically referred to as mixed populations. While the two latest revisions of federal flood frequency guidelines, published in 1982 and 2019, identified the treatment of mixed populations as an area of future research, no quantitative guidance exists on the classification of flood events or the incorporation of flood types into frequency analyses. Without quantitative guidance on the treatment of mixed populations in flood frequency analyses, there is the potential for considerable variability in frequency-based flood estimates. The treatment of a flood series resulting from a mixed population violates the assumption that floods at a site are independent and identically distributed. To avoid this issue, separate statistical models should be fit to floods arising from different mechanisms and the resulting curves should be combined to produce flood quantiles. Mixed population and flood typing literature has focused primarily on the western United States. In comparison, the Mid-Atlantic region of the United States is characterized by complex meteorology and numerous flood causal mechanisms but has been studied less frequently in mixed population literature. In addition to the absence of guidance of mixed populations, flood frequency approaches in the United States use the annual maximum series, or the maximum flow each year. The peaks-over-threshold, or partial duration series, approach to selecting flood events has the potential to increase the information content in a flood series which is important when subdividing flood events by causal mechanism. The objectives of the study are to examine: (1) flood typing based on gridded meteorologic products, (2) the advantages of using partial duration series over annual maximum series when performing a flood frequency analysis with consideration of flood type, and (3) the difference in flood quantiles from the proposed combined population methodology compared to those from the current mixed flood frequency analysis. An automated flood classification procedure was developed using gridded meteorologic products. The automated classification procedure was validated manually using historic storm publications. Flood frequency analyses were performed using both partial duration and annual maximum flood series. The method is applied within the Lehigh River watershed in eastern Pennsylvania. While the flood frequency analysis results varied across the watershed, separation of flood series by causal mechanism generally resulted in higher flood quantiles than those obtained from mixed flood series. Design floods based on the treatment of flood series as homogeneous are likely underpredicting event magnitudes. In addition, quantitative guidance on separation of flood events by causal mechanism and treatment of flood type subsets within frequency analyses is needed to produce more reliable flood estimates.


2023 Summer.
Includes bibliographical references.

Rights Access

Embargo expires: 08/28/2025.


mixed populations
flood frequency analysis


Associated Publications