Understanding low flow characteristics and the factors that influence the variability and change of low flow is important for water resources planning and development. In this work, low flow statistics of regulated and unregulated streams in the state of North Carolina were updated with streamflow data through 2019. Expressed as unit flow, low flows vary from highest in the mountain region to lowest in the coastal region. About 22% of the streams considered show a significant downward trend, but considerable low-frequency variability confounding trends were also observed. Moreover, temporal changes are evident in the 7Q10 parameter when computed over different time windows. Compared to 7Q10 computed for the entire historic record, 61% of streams show a decrease for the 1980–2019 period, whereas 15% show an increase. Similarly, 74% of streams show a decrease for the period from 2000 to 2019, whereas 20% show an increase. Considering regulation status, the long-term pattern for both regulated and unregulated streams is similar; however, regulated streams show a lower volatility at shorter time scales, possibly due to managed release of water from storage. The relationship between regional index time series of groundwater-depth low flow and the Atlantic Multidecadal Oscillation (AMO) was examined to determine whether low-frequency climate modes can account for the long-term pattern in low flows. Consequently, a significant correlation was found between AMO and groundwater-depth low flow, such that positive AMO is associated with lower groundwater-depth low flow and vice versa, particularly for the Piedmont region. Predictive equations for annual low flows at the ecoregion level were developed using a pool of predictors comprising antecedent streamflow, indices of large-scale climate modes, and groundwater depth. Springtime average streamflow and AMO were selected as the primary predictors of low flow for coastal and Piedmont regions, whereas springtime average streamflow and the November–December–January average Oceanic Niño Index were used as the primary predictors for the mountain region. The predictions were disaggregated to US Geologic Survey gauge locations to test their utility at local scale. The relative root mean square error (RMSE) of the disaggregated predictions was <23% at 79% of the stations, between 24% and 43% at 10% of the stations, and greater than 44% at 1% of the stations. The remaining 10% of stations showed large RMSEs. This latter percentage is characterized by smaller drainage basins and intermittent flows, suggesting the prediction models are not applicable to drainage basins smaller than roughly 20 square km and intermittent streams.

Keywords: Low flow, Unit flow, Eco-Regions, Prediction, AMO, ONI, 7Q10, Regulated Stream, Unregulated Stream.