| Water management infrastructure is currently designed and operated under the assumption of stationarity (Milly et al., 2008). This assumption implies natural change and changes due to human influence are small enough to be ignored, such that variables like annual peak streamflow do not change over time. Furthermore, flood frequency analysis normally requires that peak flow data is homogeneous, independent and stationary. This paper reveals that annual peak streamflows are not stationary, especially in highly urbanized areas of the country. Nearly all previous research dealing with flood trend detection involved watersheds which were not heavily influenced by humans. We take a different approach by looking at all gaged rivers in the northeastern United States. A hydrologic database of 1,312 gages in the northeastern United States was compiled to analyze trends in instantaneous peak streamflows. After performing a parametric t-test on the slope of the linear relationship between the natural log of annual peak streamflow and time, it was determined that approximately 258 gages (19.7%) indicated positive trends in flood events. Magnification and reduced recurrence intervals were developed to investigate how the presence of increasing trends will affect future storm events and return periods. Results indicate that the 10 year storm is reduced, on average, to a 6.1 year storm; the 25 year storm is reduced, on average, to a 13.8 year storm; the 100 year storm is reduced, on average, to a 48.5 year storm. Lastly, impacts of the magnified flow events and reduced return periods are discussed from a design perspective. |
