AbstractsStatistics

Increasing attention has been drawn to the need for reliable streamflow estimates at ungaged locations under a range of climatic and hydrologic conditions. Climate projections for the northeastern United States over the 21st century – which include significant increases in temperature and precipitation – could have broad impacts on streamflows, potentially reducing the accuracies of existing streamflow models for the region. This thesis investigates recent changes in daily flow-durations in southern New England, and examines their influence on the reliability of the low-flow models for Massachusetts presented by Ries and Friesz (2000). An analysis of discharge data collected at gaging sites through water year 2012 revealed increases in nearly all flow durations at sites across southern New England since the mid-20th century, whereas very low flows (quantiles at or above the 95-percent exceedance probability) generally showed decreases, especially since the 1990s. Twenty-year moving streamflow quantiles at each of ten selected exceedance probabilities were examined for the periods of record of 16 streamflow-gaging stations in southern New England. The beginning of water year 1992 appeared to mark an inflection point in low-flow quantiles, before which very low flows were steady or increasing, and after which these flows showed near-universal decreases. While the observed peak in 20-year low-flow quantiles around 1992 may be due to the statistical method used to calculate the quantile trends, the inflection point could also be an indicator of when increasing evapotranspiration surpassed increasing precipitation as the principal climatic driver of changes in low flows in southern New England. The general upward translation of the flow-duration curve observed over the last 60 years is very likely linked to increases in annual precipitation during this period, while the decreases in very low flows are likely due to changes in climatic variables (increasing summer temperatures and evapotranspiration rates), and amplified by anthropogenic factors (greater areas of impervious surfaces and increasing rates of surface- and ground-water withdrawal). The data suggest that increasing precipitation rates have already caused the Ries and Friesz (2000) equations for the median low flows (Q50 to Q75) to become biased towards underestimation, and decreases in very low flows threaten to render the models for these flows biased towards overestimation in the coming decades. The streamflow quantile trends (for both the entire period of record of the gaging stations and just the post-1992 period) for each of the ten flow-durations of interest were extended into the future to the point where the corresponding Ries and Friesz (2000) model would fail (when actual flow durations would be outside the 90-percent prediction intervals for the estimated flows for greater than 10% of sites). The models for the lowest streamflows are estimated to lose validity by as early as 2018. Climate change is predicted to have significant effects on streamflow…