On the statistical attribution of the frequency of flood events across the US Midwest

The frequency of flood events has increased across most of the U.S. Midwest in the past 50-70 years; however, little is known about what is driving these changes. Using an observation-driven approach, we develop a statistical framework to attribute the changes in the frequency of flood peak events to changes in the climate system and to land use / land cover. We focus on 287 U.S. Geological Survey sites with at least 50 years of daily discharge measurements between the second half of the 20th century and the present. Our analyses are performed at the seasonal level and consider five predictors: precipitation, temperature, antecedent wetness conditions, agriculture, and population density. Even though we use simple models, we are able to reproduce well the interannual variability in the frequency of flood events as well as the overall long-term tendencies. Results indicate that precipitation and antecedent wetness conditions are the strongest predictors, with the role of the latter increasing as we lower the threshold for the event identification. Temperature is an important predictor only in the northern Great Plains during spring, where snow-related processes are most relevant. Population (as a proxy of urbanization) and agriculture are less important compared to the climate predictors.