# Exploring online sensor parameters as proxies for polar organic chemicals—An innovative approach for combined sewer overflow monitoring

## TL;DR

This study explores using online sensors to estimate organic chemical concentrations in combined sewer overflows, offering a cost-effective alternative to traditional lab methods.

## Contribution

The paper introduces a novel approach using sensor data to predict polar organic chemical concentrations in sewers during rain events.

## Key findings

- Indoor chemical concentrations in large catchments are well predicted by SAC254 nm and NH4-N with median relative errors of 32% and 29%.
- Road chemical concentrations can be predicted with similar accuracy across catchment sizes using level or flow measurements.
- Pesticide concentrations remain challenging to predict due to their variable behavior during rain events.

## Abstract

Combined sewer overflows (CSOs) can release toxic organic chemicals into surface waters during rain events. Currently, most overflow sites are not monitored because commonly used methods, such as automated grab sampling followed by laboratory analysis using liquid chromatography coupled with mass spectroscopy (LC-MS), are costly and time-consuming. Due to this monitoring gap, the dynamics of organic chemicals in CSOs remain poorly understood. This study explores the use of eight online sensor parameters as proxies for polar organic chemicals from different sources in combined sewer systems during wet weather. We used sensor and organic chemical data collected in three urban catchments of varying sizes. Correlations between chemicals from the same source and sensor parameters were calculated. In the largest catchment (160,000 inhabitants), indoor chemicals are strongly correlated with flow, electrical conductivity, spectral absorption coefficient at 254 nm (SAC254 nm), and ammonium (NH4-N). Additionally, linear regressions were developed to predict organic chemical concentrations from sensor data. Models based on SAC254 nm and NH4-N predict indoor chemical concentrations with median relative errors of 32% and 29%, respectively, in the large catchment. Prediction performance for road chemicals is independent of catchment size, with median relative errors ranging from 39% to 44%, using either level or flow measurements. However, the prediction of pesticide concentrations remains limited, as these chemicals exhibit diverse patterns across rain events. Overall, our results suggest that linear regression models can estimate indoor chemical concentrations in large catchments and road chemical concentrations in catchments of any size. However, for real-world implementation, further research is needed to refine calibration requirements and validate the models across diverse catchments. Nevertheless, these models are promising for cost-effective, long-term monitoring of organic chemicals and for mitigating the impact of CSO discharges.