Using a Water Budget Approach to Quantify Inflow and Infiltration Impacts on Urban Streamflow

Human activities can have substantial impacts on watersheds, and a dominant, yet understudied, impact on urban watersheds is the inflow and infiltration (I&I) of water into sewage infrastructure. This study uses a water-budget approach to quantify the magnitude of I&I and its effects on streamflow. The analysis is performed over the 2013-2020 period on 90 watersheds in the Atlanta, Georgia USA metropolitan statistical area (MSA), which has a humid subtropical climate. The following annual totals are determined for each watershed: precipitation, water leakage from municipal sources, actual evapotranspiration (AET), water withdrawals, and observed stream discharge. AET is the most difficult component to estimate, therefore, multiple models are used to estimate AET totals, and reference watersheds are used to adjust the totals. Predicted discharge is estimated by subtracting known water outputs from the water inputs, and I&I was the difference between predicted discharge and observed discharge. The most I&I-impacted watersheds are those with the largest I&I to stream discharge ratios. Mean annual totals for precipitation and supply-pipe loss for those watersheds are 1,498 mm and 39 mm, respectively. Mean annual totals for AET, stream discharge, and I&I, are 737 mm, 534 mm, and 267 mm. Therefore, the mean I&I to streamflow ratio for the ten most I&I-impacted watersheds is 0.51 (i.e., I&I is 51% of streamflow). Mean population densities, percent developed, and percent imperviousness for the ten watersheds are as follows: 1,308 people per km², 60%, and 37%, respectively. I&I is strongly positively correlated with the above three urbanization variables. Regression analyses show that population density explains approximately 50% of the variation in I&I and is the best predictor of I&I. The most urbanized watersheds in the Atlanta MSA have relatively low population densities compared to typical urban watersheds globally, so it remains to be seen if the regression model can be used in locales with much higher population densities. Nevertheless, these results are supported by previous findings in the eastern United States and the results should be transferrable to most urban watersheds there, while the general approach for quantifying I&I should be applicable globally.