Nonstationarity of low flows and their timing in the eastern United States

Abstract. The analysis of the spatial and temporal patterns of low flows as well as their generation mechanisms over large geographic regions can provide valuable insights and understanding for climate change impacts, regional frequency analysis, risk assessment of extreme events, and decision-making regarding allowable withdrawals. The goal of this paper is to examine nonstationarity in low flow generation across the eastern US and explore the potential anthropogenic influences or climate drivers. We use nonparametric tests to identify abrupt and gradual changes in time series of low flows and their timing for 508 USGS streamflow gauging sites in the eastern US with more than 50 years of daily data, to systematically distinguish the effects of human intervention from those of climate variability. A time series decomposition algorithm was applied to 1-day, 7-day, 30-day, and 90-day annual low flow time series that combines the Box–Ljung test for detection of autocorrelation, the Pettitt test for abrupt step changes and the Mann–Kendall test for monotonic trends. Examination of the USGS notes for each site showed that many of the sites with step changes and around half of the sites with an increasing trend have been documented as having some kind of regulation. Sites with decreasing or no trend are less likely to have documented influences on flows. Overall, a general pattern of increasing low flows in the northeast and decreasing low flows in the southeast is evident over a common time period (1951–2005), even when discarding sites with significant autocorrelation, documented regulation or other human impacts. The north–south pattern of trends is consistent with changes in antecedent precipitation. The main exception is along the mid-Atlantic coastal aquifer system from eastern Virginia northwards, where low flows have decreased despite increasing precipitation, and suggests that declining groundwater levels due to pumping may have contributed to decreased low flows. For most sites, the majority of low flows occur in one season in the late summer to fall, as driven by the lower precipitation and higher evaporative demand in this season, but this is complicated in many regions because of the presence of a secondary low flow season in the winter for sites in the extreme northeast and in the spring for sites in Florida. Trends in low flow timing are generally undetectable, although abrupt step changes appear to be associated with regulation.