Improving Model Hydrology Using Antecedent Moisture Modeling (AMM): Joliet, IL Granite St Case Study

This presentation will tell the story of modeling completed near Granite Street in the City of Joliet in order to size needed sanitary storage. The traditional RTK Unit Hydrograph method provided unsatisfactory results in calibration, resulting in recommendations that required a lot of modeler's judgment. Many alternative models were considered to improve calibration and the Antecedent Moisture Model (AMM), while relatively untested by industry, was identified as the only model which could really fit the patterns of the flow meter data. This presentation will explain the AMM model at a high level and show how with a more accurate model the City was able to save $8 million in budgeted improvements. The City of Joliet, Illinois operates a combined sewer system. The City has received regulatory pressure to reduce the frequency of activation of its Combined Sewer Overflow (CSO) locations, which discharge untreated sewage to the Des Plaines River during rain events in which the combined sewer system does not have sufficient capacity. In 2010 the City agreed to a Long Term Control Plan with the Illinois Environmental Protection Agency (IEPA) that laid out a five phase plan with the goal of ensuring that by 2023 no CSO location would activate more than 20 times over any 5-year period (which is more or less an average of 4 CSOs per year plus a small factor of safety). In 2020 the City had completed the first three phases and performed additional modeling to check their progress and verify sizing and design details of the remaining two phases. Phase V of the Long Term Control Plan was intended to decrease the activation frequency at CSO 004 which serves the Granite St. basin and tentatively called for a 411,000 gallon storage tank plus a 7,600 gallons per minute (gpm) pump station, which was projected to cost roughly $8 million when indexed for construction costs. Prior modeling had been completed for the system using the RTK method. A majority of the flow in the Granite St. basins has been documented to come from sewer laterals and connected foundation drains, sources that are notorious for infiltration varying dramatically depending on antecedent moisture conditions. Prior modeling had calibrated the model with a focus toward events with the largest flow responses (i.e., the largest events and the ones with the highest antecedent moistures). This intentionally resulted in over-predicting flow responses for smaller events in order to be conservative. But it was not clear how much conservatism was appropriate. In desperately looking for a model method which would provide sufficient accuracy to make recommendations, the author came across the Antecedent Moisture Model (AMM). The AMM model was developed and used by Robert Czachorski for nearly two decades, but only recently were the equations released into the public domain. AMM is an empirically-calibrated method, similar to the RTK Unit Hydrograph method, but whereas RTK assumes a fixed capture coefficient (or percent capture), AMM allows parameters to model variability in the capture coefficient by antecedent moisture and by season. AMM has been particularly successful in modeling sanitary sewer infiltration, for which prior methods perform poorly. In 2021 the model was re-calibrated using the Antecedent Moisture Modeling (AMM) method. The new model allowed for a much better calibration, which adequately predicted both smaller events and larger events. The AMM model successfully represented the increasing percent capture rates for different soil conditions and resulted in a spectacularly better calibration. The AMM method also improved the model's ability to predict seasonal differences in infiltration, specifically the lower response during the dry summer months. After calibration, a long-term continuous simulation was conducted using a 13-year rainfall record at a nearby rain gauge to establish the existing level of protection. The model simulation showed the maximum number of activations of CSO 004 during any 5-year period to be 13, less than the limit of 20 CSOs/5 years. This indicated that Phase V was not necessary and that, as of completion of Phase III, This result was surprising, but additional quality control and comparisons between the two models showed it was reasonable. It appears the months for which meter data was available in 2015 and 2020 displayed larger moisture factors than supposed when compared to the long-term record. This skewed modeler judgment about the needed conservatism when calibrating using the RTK method. Indeed, CSO activation records from the first two years following completion of Phase III show the activation of CSO 004 is in line with the predictions of the AMM-calibrated model. Using the AMM model, which adequately represents antecedent moisture and seasonal effects, allowed modelers to predict CSO activation frequency with much more confidence, which reduced the need for conservatism and showed the Phase V project was not needed at all. The City of Joliet has submitted the findings of its modeling and has received approval to remove Phase V from the Long Term Control Plan, saving the City an estimated $8 million. Figure: A comparison of the two calibrations for the GRANITE-1 meter showing the May 17, 2020 rain event (the largest rain event in the calibration period). The AMM model (red) is a very good fit to the meter data (blue) and clearly represents the increasing percent capture rates as the rain event wears on and soil moisture increases. The RTK model (green) does not adequately model this dynamic.