Direct Human Interventions Drive Non-Stationarity in Annual Peak Streamflow Patterns Across the United States

Abstract

Abstract Understanding the factors driving non-stationarity in annual peak streamflow, hereafter referred to as peakflow, remains pivotal amid climate change and direct human interventions1,2. Utilizing extensive streamflow observations from 3907 United States Geological Survey (USGS) stations, we have detected significant trends in 34% of these stations. Among these, two-thirds exhibit decreasing trends distributed across the United States, while the remaining one-third show increasing trends, predominantly in the Northeast and Great Lakes regions. Most USGS stations (84%) are influenced by direct human interventions such as water management and land use changes. Employing high-resolution climate and land-use data along with geospatial analytics, this study reveals urbanization and water management as the primary drivers, followed by agriculture and climate change. Urbanization emerges as the principal driver of peakflow trends in the Texas-Gulf, California, and Mid-Atlantic regions, accounting for up to 62%, 44%, and 32% of the variance, respectively. Water management explains most of the variance in the Tennessee (37%) and Ohio River Basins (30%). In the Upper Colorado River Basin, both agricultural and water management play significant roles, explaining up to 28% and 24% of the variance, respectively. Additionally, agricultural land use explains 17% of the variance in the Great Lakes region. Climate contributes modestly in the Rio Grande (15%) and California (11%) regions. Despite their extensive number of climate realizations (large ensemble), the latest generation of climate and earth system models inadequately captures these human-induced factors, limiting their predictive accuracy. By demonstrating the outsized influence of human interventions on peakflow trends and inadequacies in current climate models, our findings stress the imperative of integrating water management and urbanization effects into climate models for more accurate water predictions.