Nonstationarity in Records of Extreme Precipitation Frequency for the Upper Midwest

Extreme precipitation estimates, particularly for rare storms with large return periods like the 100-year storm, are a useful tool for watershed managers and engineers to evaluate the flood risk for their communities. Although national-level estimates are publicly available, these estimates do not consider temporal dynamics (i.e., nonstationarity) of extreme precipitation. Modern observation records for individual stations extend fewer than 150 years, which poses challenges for accumulating sufficient observations of such rare events in order to run statistical tests or identify trends over time. This research examines the historical record of 1,218 long-term (≥70 years) precipitation records from weather stations in the Upper Midwest to identify trends in the fraction of stations that report an annual maximum precipitation above various thresholds. First, we identify significant (p<0.01) nonstationarity in the aggregate frequency of occurrence of top 50% and top 10% annual maximum precipitation events, as well as in the occurrence of largest recorded event at each station (∼97-year storm). Second, our results show that a piecewise linear fit with two inflection points occurring during the years 1920 and 1974 better models observed extreme event nonstationarity than a simple linear fit, where the rate of change for aggregate frequency of extreme event occurrence increases at both break points. The Dust Bowl period in the American Midwest explains the break point around 1920, and the break point near 1974 coincides with surface temperature increases for the continental United States. This model predicts that the storm depth assigned to the 97-year recurrence interval, using an individual station’s annual maximum series and a stationary assumption, would actually have a 71-year recurrence interval in 2024 (95% confidence interval: 54–100).