Historical changes in annual peak flows in Maine and implications for flood-frequency analyses

Flood-frequency analyses use statistical methods to compute peak streamflows for selected recurrence intervals— the average number of years between peak flows that are equal to or greater than a specified peak flow. Analyses are based on annual peak flows at a stream. It has long been assumed that the annual peak streamflows used in these computations were stationary (non-changing) over very long periods of time, except in river basins subject to direct effects of human activities, such as urbanization and regulation. Because of the potential effects of global warming on peak flows, the assumption of peak-flow stationarity has recently been questioned. Maine has many streamgages with 50 to 105 years of recorded annual peak streamflows. In this study, this long-term record has been tested for historical flood-frequency stationarity, to provide some insight into future flood frequency.

Changes over time in annual instantaneous peak streamflows at 28 U.S. Geological Survey streamgages with long-term data (50 or more years) and relatively complete records were investigated by examining linear trends for each streamgage’s period of record. None of the 28 streamgages had more than 5 years of missing data. Eight streamgages have substantial streamflow regulation. Because previous studies have suggested that changes over time may have occurred as a step change around 1970, step changes between each streamgage’s older record (start year to 1970) and newer record (1971 to 2006) also were computed. The median change over time for all 28 streamgages is an increase of 15.9 percent based on a linear change and an increase of 12.4 percent based on a step change. The median change for the 20 unregulated streamgages is slightly higher than for all 28 streamgages; it is 18.4 percent based on a linear change and 15.0 percent based on a step change.

Peak flows with 100- and 5-year recurrence intervals were computed for the 28 streamgages using the full annual peak-flow record and multiple sub-periods of that record using the guidelines (Bulletin 17B) of the Interagency Advisory Committee on Water Data. Magnitudes of 100- and 5-year peak flows computed from sub-periods then were compared to those computed from the full period. Sub-periods of 30 years with starting years staggered by 10 years were evaluated (1907–36, 1917–46, 1927–56, 1937–66, 1947–76, 1957–86, 1967–96, and 1977–2006). Two other sub-periods were evaluated using older data (start-of-record to 1970) and newer data (1971 to 2006). The 5-year peak flow is used to represent small and relatively frequent flood flows in Maine, whereas the 100-year peak flow is used to represent large flood flows.

The 1967–96 sub-period generated the highest 100- and 5-year peak flows overall when compared to peak flows based on the full period of record; the median difference for all 28 streamgages is 8 percent for 100- and 5-year peak flows. The 1977–2006 and 1971–2006 sub-periods also generated 100- and 5-year peak flows higher than peak flows based on the full period of record, but not as high as the peak flows based on the 1967–96 sub-period. The 1937–66 sub-period generated the lowest 100- and 5-year peak flows overall. The median difference from full-period peak flows is -11 percent for 100-year peak flows and -8 percent for 5-year peak flows. Overall, differences between peak flows based on the sub-periods and those based on the full periods, generated using the 20 unregulated streamgages, are similar to differences using all 28 streamgages.

Increases in the 5- and 100-year peak flows based on recent years of record are, in general, modest when compared to peak flows based on complete periods of record. The highest peak flows are based on the 1967–96 sub-period rather than the most recent sub-period (1977-2006). Peak flows for selected recurrence intervals are sensitive to very high peak flows that may occur once in a century or even less frequently. It is difficult, therefore, to determine which approach will produce the most reliable future estimates of peak flows for selected recurrence intervals, using only recent years of record or the traditional method using the entire historical period. One possible conservative approach to computing peak flows of selected recurrence intervals would be to compute peak flows using recent annual peak flows and the entire period of record, then choose the higher computed value. Whether recent or entire periods of record are used to compute peak flows of selected recurrence intervals, the results of this study highlight the importance of using recent data in the computation of the peak flows. The use of older records alone could result in underestimation of peak flows, particularly peak flows with short recurrence intervals, such as the 5-year peak flows.