Identifying spatial clustering in change points of streamflow across the contiguous U.S. between 1945 and 2009

Much of the work investigating sudden changes in streamflow in the U.S. has used only a small subset of all available gage data and has identified only a single change point in each gage's period of record. In this paper, we apply a change point detection and clustering algorithm that uses all U.S. Geological Survey flow gages with near‐continuous records, detects multiple change points in annual streamflow, and groups change points into geographic clusters which are not predefined by any political or hydrologic boundaries. We identify 17 spatially distinct change point clusters, 14 of which are related to concurrent changes in precipitation. Several geographic regions display multiple clusters, indicating multiple change points in time. The presence of abrupt changes in streamflow suggests that natural variability in the climate signal may be dominating observed streamflow variations in the last 60 years in many locations in the contiguous U.S. Key Points Examines concurrent changes across multiple nearby stream gages to minimize influence of nonclimatic factors on change points Regional‐scale change points in streamflow tend to correspond to concurrent changes in precipitation Many regions undergo several change points in streamflow over the study period Plain Language Summary Traditional analyses of changes in mean annual streamflow usually look for long‐term trends, gradual changes in magnitude of mean annual streamflow that occur over many years. However, in some cases, mean annual streamflow changes rapidly. In just a few years, the magnitude of the mean annual streamflow can shift from one mean level to another by a step‐like change. These rapid shifts are sometimes referred to as a change points. This paper looks for concurrent change points across 1500+ stream gages with continuous streamflow data since the 1940s. Many prior analyses have looked at change points on individual stream gages. When only a single stream gage is analyzed, it is difficult to distinguish whether a change point is due to changes in climate drivers of more direct changes like modifications to the instruments used to measure streamflow. By analyzing concurrent changes across many gages, we presume that we can identify change points due only to widespread shifts in climate. While the exact cause of a change point in streamflow is not often known, it is often considered to be due to natural variations in the climate drivers that influence streamflow. Thus, identifying change p