Rainfall-derived infiltration and inflow (RDII) into sanitary sewers is known to be a major contributor to sanitary sewer overflow (SSO) occurrences and water-in-basement (WIB) complaints. Modeling of sanitary sewer systems is commonly employed to investigate these problems, using continuous or single event simulations. Continuous simulation can simulate RDII more effectively for planning than single-event simulations by incorporating antecedent moisture conditions (AMC) directly, rather than using assumed AMC as necessary for single event simulations. AMC is represented by monthly initial abstraction (IA) parameters in SWMM5, which is used with the unit hydrograph parameters (RTK) to continuously simulate RDII. Monitoring limitations often hinder accurate calibration of these parameters, and assumed values need to be used. Understanding both the spatial and temporal variation of the empirically derived unit hydrograph parameters, including both total RDII capture fraction (R) and IA, is important for accurately establishing assumed values for these model parameters to obtain robust simulation results.
This paper presents a statistical analysis of spatial variation of total R and IA and their relationship with physical system factors such as pipe density (length of pipes per acre), pipe age, land use, vegetation coverage, soil type, etc., provided for a project in Columbus, Ohio. R and IA were obtained from continuous calibration of a system-wide model (Sewer System Capacity Model Update 2006 – SSCM MU 2006). Model calibration used flow monitoring data, radar-rainfall and rain gage rainfall data collected during 2008 and 2009. The model was calibrated with seasonal (dormant and growth seasons) RTK and seasonal IA parameters to simulate RDII. Global Moran’s I and Anselin Local Moran’s I tests were performed on seasonal total R and IA to detect spatial auto correlation and clusters and outliers. Multivariate regression analysis was performed to evaluate the relationship of total R and IA with various physical system factors.
The results showed significant spatial auto correlation of both total R and IA (SWMM5 Dmax parameter) for both dormant and growth seasons. Total R showed more significant spatial auto correlation than Dmax. Regression analysis of total R and Dmax with the physical factors revealed a strong relationship of total R with pipe density and pipe age. Adjusted R2 exceeding 70% for both seasons was achieved. However, no significant relationship was detected between Dmax and the physical factors. The significant relationship found in this study has benefitted Columbus’s SSCM MU 2006 model, as a means of estimating total R for basins that cannot be calibrated directly by using a calibrated basin nearby or downstream basins with similar pipe density and pipe age. Further testing on other systems may reveal whether this finding is broadly applicable or only applies to the Columbus, Ohio system studied.