Combined sewer overflows and water quality before and after deep tunnels operation in Milwaukee, Wisconsin.

摘要

This study's objectives were twofold. The first was to investigate precipitating overflow events' factors before, and after the deep tunnels operation. The second objective was to quantify the deep tunnels' pollutant mitigative effects principally on the Milwaukee, Menomonee, and Kinnickinnic Rivers. Data from the Milwaukee Metropolitan Sewerage District's overflow and surface water quality monitoring program were used. In addition, precipitation records from gages at Mitchell Field and a couple of gages inside the combined sewer area (CSO) were also used.;In this study, Artificial Neural Network (ANN) modeling techniques were utilized to predict overflow events. The simulations were done separately for the periods before (1991–1993), and after (1994–1997) the deep tunnels became functional. Furthermore, the three river watersheds were also examined separately contributing to six case studies analyzed. ANN results indicated that more input parameters were necessary before overflow could occur after the deep tunnels' operation, compared to the period before. The findings also reflect an advantage of an ANN over using only total rainfall amount as predictor of overflow events.;Statistical analysis of water quality parameters (BOD, phosphorus, suspended solid, fecal coliform counts, zinc, and chloride) in the three rivers indicates that Menomonee River benefits the most from the deep tunnels. Fecal coliform counts inside the CSO area, and to a certain extent BOD and zinc levels, exhibit the most decline after 1994. These conclusions were based on t-test comparisons of regional averages incorporating spatial and temporal correlations calculations from the two periods. The results from the t-test were complemented and confirmed with results from Mann-Kendall test for trend. In contrast to fecal coliform, BOD, and zinc, suspended solids and chloride do not decrease after 1994.;There is a need for more river samplings close to overflow points and near the time of overflow events. Based on the results of spatial correlations for two adjacent river sampling stations, several stations were found to provide redundant information. Among these are sampling stations RI-2 and RI-3 (outside CSO), and RI-6 and RI-7 (inside CSO) of the Milwaukee River. Stations RI-21 and RI-22 (outside CSO) of the Menomonee River, and stations RI-18 and RI-19 (inside CSO) of the Kinnickinnic River are also redundant.