Modelling Water Quality of Wet - Weather Emissions from Urban Drainage Systems

Combined sewer overflows (CSOs) remain a persistent environmental challenge in urban drainage systems worldwide. During wet-weather events, untreated mixtures of wastewater and stormwater are discharged into receiving waters, often carrying high loads of particulate pollutants. Among these, Total Suspended Solids (TSS) play a central role, as they act as carriers for a wide range of particle-bound contaminants. The temporal dynamics of TSS during CSO events are highly variable and frequently characterized by sharp intra-event concentration peaks. These short-term dynamics are particularly critical for Pollutionbased Real-Time Control (P-RTC), which aims to prioritize storage and treatment of the most polluted fractions of overflow discharges.

Despite decades of development in urban water quality modelling, existing modelling structures struggle to reliably reproduce intra-event TSS dynamics under realistic monitoring and data constraints. Detailed hydraulic models are often overparameterized and computationally demanding, while conceptual, empirical, and data-driven approaches tend to oversimplify key processes. Spatial heterogeneity and inter-event variability are frequently neglected. As a result, models may reproduce aggregated pollutant loads but fail to capture the timing and magnitude of concentration peaks that are operationally relevant for pollution control.

This PhD study systematically evaluates the performance of commonly used water quality modelling approaches across temporal scales. Particular emphasis is placed on intra-event dynamics through the introduction of a Peak-Weighted Dynamic Time Warping (PW-DTW) metric tailored to P-RTC requirements. The results indicate that the identified limitations are not primarily related to calibration strategy or parameter uncertainty, but rather to fundamental structural aspects of how particulate pollutant processes are represented in conventional water quality models.

Building on these structural limitations, a new modelling concept, the Localized Sediment Resuspension (LSR) model, is developed. The LSR framework inspired by evidence from previous research assumes that unresolved TSS peaks originate from sediment accumulation at hydraulically defined locations within the sewer network (hotspots), where resuspension is triggered once shear stress exceeds critical thresholds. Importantly in this study, these processes are treated as localized rather than uniform system-wide phenomena. The resulting hotspot contributions are transported downstream according to local hydraulic conditions and collectively create the pollution profile observed at the monitoring location. Parameter uncertainty and inter-event variability are addressed through a stochastic implementation.

Comparison with a calibrated detailed hydraulic model (baseline model) demonstrates that explicitly representing in-sewer sediment accumulation and resuspension, while respecting spatial heterogeneity and inter-event variability, substantially improves the reproduction of observed intra-event TSS peaks. A structured scenario analysis identifies a robust LSR configuration that achieves consistent improvements while avoiding overparameterization. Transferability tests in a second case study confirm that the modelling philosophy remains applicable beyond the original development site. Furthermore, incorporating sediment heterogeneity through particle size distribution (PSD) classes enhances predictive performance and enables interpretation of class-specific contributions to pollutant peaks.

Finally, when evaluated within a P-RTC framework, decisions supported by LSR-based flow and concentration predictions increase pollutant load capture compared to both flow-only strategies and baseline model–based flow and concentration strategies, demonstrating clear practical relevance for operational water quality management.