<p>Separate sewer systems (SSSs) assume effective separation of sewage and stormwater by design, thus expecting influent volumes at the receiving wastewater treatment plant (WWTP) to be minimally impacted by stormwater. So, the assessment of risks related to high influent volumes in WWTPs served by SSSs during extreme precipitation events has not been prioritized in research. However, a novel random forest (RF)-based modelling approach was able to diagnose the risk of high influent volumes due to rainfall-derived infiltration & inflow (RDII) ingressions for Indian WWTPs served by SSSs. The vulnerability of only the Indian WWTP could be attributed to stormwater ingressions through damages, cracks or gaps in sewer lines or through open or broken manhole covers. This resulted in de facto operation of SSSs as combined sewer systems (CSSs). Using covariates related to inflow history, daily temperature, precipitation and precipitation history from two candidate WWTPs located in India (33 m<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(^3\)</EquationSource> </InlineEquation>/d, tropical, monsoonal) and the United States respectively (95 m<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(^3\)</EquationSource> </InlineEquation>/d, extra-tropical), the model also identified persistence and periodicity in influent volumes for both plants. RF models offered robust predictive ability, with median R<InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(^2\)</EquationSource> </InlineEquation> values of 0.82 and 0.97 obtained for the Indian and US WWTPs respectively, and maximum values of 1 for both. Under hypothetical future scenarios of rainfall and population increase in India, the model predicted increasing modality in influent distributions. To further explore the transferability of the RF model to nearby locations, its ability to predict influent flows at a neighbouring WWTP served by an independent SSS was illustrated. Without requiring any covariate inputs from the new WWTP, standardized daily influent flows were predicted with 86% correct categorization of “high” (greater than mean flow) and “low” (lesser than mean flow) risk categories. This finding demonstrated a methodology to develop strategies for improving preparedness at the smaller, unmonitored WWTP before large rainfall events. Thus, the proposed study offers the first-ever approach for climate-informed WWTP management in data-limited countries.</p>

Investigating precipitation impacts on separate sewer flows in data-limited systems

期刊论文
发表日期：

2026年4月1日

查看全文

Hitesh Pande,
Bihu Suchetana

摘要

Separate sewer systems (SSSs) assume effective separation of sewage and stormwater by design, thus expecting influent volumes at the receiving wastewater treatment plant (WWTP) to be minimally impacted by stormwater. So, the assessment of risks related to high influent volumes in WWTPs served by SSSs during extreme precipitation events has not been prioritized in research. However, a novel random forest (RF)-based modelling approach was able to diagnose the risk of high influent volumes due to rainfall-derived infiltration & inflow (RDII) ingressions for Indian WWTPs served by SSSs. The vulnerability of only the Indian WWTP could be attributed to stormwater ingressions through damages, cracks or gaps in sewer lines or through open or broken manhole covers. This resulted in de facto operation of SSSs as combined sewer systems (CSSs). Using covariates related to inflow history, daily temperature, precipitation and precipitation history from two candidate WWTPs located in India (33 m \(^3\) /d, tropical, monsoonal) and the United States respectively (95 m \(^3\) /d, extra-tropical), the model also identified persistence and periodicity in influent volumes for both plants. RF models offered robust predictive ability, with median R \(^2\) values of 0.82 and 0.97 obtained for the Indian and US WWTPs respectively, and maximum values of 1 for both. Under hypothetical future scenarios of rainfall and population increase in India, the model predicted increasing modality in influent distributions. To further explore the transferability of the RF model to nearby locations, its ability to predict influent flows at a neighbouring WWTP served by an independent SSS was illustrated. Without requiring any covariate inputs from the new WWTP, standardized daily influent flows were predicted with 86% correct categorization of “high” (greater than mean flow) and “low” (lesser than mean flow) risk categories. This finding demonstrated a methodology to develop strategies for improving preparedness at the smaller, unmonitored WWTP before large rainfall events. Thus, the proposed study offers the first-ever approach for climate-informed WWTP management in data-limited countries.