Planning Level Modeling of E. coli levels in a Suburban Watershed Using PCSWMM
Abstract
Scajaquada Creek is an important urban stream in the Niagara River watershed. It is listed in the Department of Environmental Conservation’s (DEC) statewide 2010 Section 303d "List of Impaired Waters" because of bacteria and dissolved oxygen impairments related to sewer overflows and urban runoff. This project used PCSWMM to model hydrology of the headwater areas of Scajaquada Creek, from Lancaster to the Buffalo City line. Eight cross-sections were surveyed along Scajaquada Creek for model input and seven sub-basins were delineated. Daily flow data from a USGS gauge station located at the Buffalo City Line for the years 1989, 1990, 1992 and 1994 were used to calibrate and validate the model. Linear regression and the Nash Sutcliffe coefficient of efficiency were used to assess goodness-of-fit. Model parameter values were very stable from year to year and r2 between observed and modeled flows ranged from 0.63 to 0.66 and Nash Sutcliffe values ranged from 0.60 to 0.76. E. coli samples were collected at several sites along the Creek, but most of the sampling focus to date has been at the USGS gauge station. Geometric mean E. coli levels in all seasons were higher at this site for storm events as compared to dry weather samples. Geometric mean E. coli levels during warmer months (May-early September) were higher for both storm (6,700 cfu/100 mL) and dry weather (54 cfu/100 mL) samples compared to colder months (February-April). A first-order decay co-efficient approach was used to model bacteria levels with acceptable planning level results. This basic model did not explicitly represent CSO’s or storm water drainage systems for each municipality. However, such details could be included to further refine management decisions at the municipality level. Our experience with Scajaquada Creek showed that PCSWMM can effectively represent flow and water quality in a mixed land use river with results that are comparable to HSPF, but with less effort to operationalize the model.
This paper is only available in PDF Format:
View full text PDF