PySWMM: The Python Interface to Stormwater Management Model (SWMM)

A model-based analysis for trapping suspended sediment in stormwater inlets of urban drainage network

Raised awareness of environmental constraints in recent decades has led stormwater management to incorporate quality components and focus on the treatment of urban runoff water at pollutant source areas. This study evaluated the impact of a developed type of sediment trap, installed into stormwater inlets, on the total suspended solids (TSS) load in an urban city center catchment in Finland. The objective was to outline a modelling approach to assess efficiency of the traps to treat TSS originating from different land uses (green areas, pavement, parking, roof, street, and other areas not belonging to the main land uses). A Storm Water Management Model (SWMM) parametrization of a 5.87 ha catchment in the Lahti city center, Finland was utilized as the computation engine. The model had separate subcatchments for each land use, allowing the use of literature-based Event Mean Concentrations (EMC) to estimate the TSS pollutant washoff for the land uses. A method to assess the individual stormwater inlet pollutant loads and potential removal effect of the sediment traps was introduced. The hydrological and TSS load simulations covered a period of 6 months. The stormwater network inlets installed with sediment traps were ranked according to their potential removal of TSS. One out of five EMC sets was selected to be representative of the urban land uses in the study site (green areas 75 mg/l, pavement 46 mg/l, parking 44 mg/l, roof 20 mg/l, street 64 mg/l, other 46 mg/l). The simulation results showed the influence of land uses on the pollutant load and revealed the optimal set of locations for the sediment traps. Additionally, the effect of regular maintenance intervals on the pollutant load, given a maximum storage capacity of the traps, was explored. The results showed a large variation in TSS removal depending on the inlets chosen for the sediment traps, with removal rates ranging from about 0 % to 10 % of catchment TSS load. The maximum TSS removal was 63 %, which was the reported efficiency of the traps. These results highlighted the need for an informed decision when selecting trap locations. Streets and parking lots were the largest TSS contributors, with stormwater inlets on streets being the desired sediment trap locations. While the absolute level of simulated TSS load was found to be dependent on the EMCs, the ranking of sediment trap locations was similar for the simulations with different EMC data sets.

Tunable linear feedback control of urban drainage systems using models defined purely from data

Real-time and model-predictive control promises to make urban drainage systems (UDS) adaptive, coordinated, and dynamically optimal. Though early implementations are promising, existing control algorithms have drawbacks in computational expense, trust, system-level coordination, and labor cost. Linear feedback control has distinct advantages in computational expense, interpretation, and coordination. However, current methods for building linear feedback controllers require calibrated software models. Here we present an automated method for generating tunable linear feedback controllers that require only system response data. The controller design consists of three main steps: (1) estimating the network connectivity using tools for causal inference, (2) identifying a linear, time-invariant (LTI) dynamical system which approximates the network, and (3) designing and tuning a feedback controller based on the LTI urban drainage system approximation. The flooding safety, erosion prevention, and water treatment performance of the method are evaluated across 190 design storms on a separated sewer model. Strong results suggest that the system knowledge required for generating effective, safe, and tunable controllers for UDS is surprisingly basic. This method allows near-turnkey synthesis of controllers solely from sensor data or reduction of process-based models.

New optimization strategies for SWMM modeling of stormwater quality applications in urban area

Build-up/wash-off models were originally developed for small-scale laboratory facilities with uniform properties. The effective translation of these models to catchment scale necessitates the meticulous calibration of model parameters. The present study combines the Mat-SWMM tool with a genetic algorithm (GA) to improve the calibration of build-up and wash-off parameters. For this purpose, Mat-SWMM was modified to equip it with the capacity to provide comprehensive water quality analysis outcomes. Additionally, this research also conducts a comparative examination of two distinct types of objective functions in the optimization. Rather than depending on previous literature, this study undertook a numerical campaign to ascertain an appropriate range for the relevant parameters within the case study, thereby ensuring the optimization algorithm's efficient functionality. This research also implements an integrated event calibration approach, i.e., a novel method that calibrates all rainfall events collectively, thus improving systemic interaction representation and model robustness. The findings indicate that employing this methodology significantly enhances the reliability of the outcomes, thereby establishing a more robust procedure. The first objective function (TSS instantaneous less squared difference function, OF 1), which is widely employed in the literature, was designed to minimize the difference between observed and predicted instantaneous Total Suspended Solids (TSS) concentrations. In contrast, the second function (mass and mass peak consistency function, OF 2) considers integral model outputs, i.e., the overall mass balance, the time of the peak mass flow rate, and its intensity. The analysis of the outputs revealed that both objective functions demonstrated sufficient performance. OF 1 provided slightly better performance in predicting the TSS concentrations, whereas OF 2 demonstrated superior ability in capturing global event characteristics. Notably, the optimal parameter set identified through OF 2 aligned with the physically plausible ranges traditionally recommended in technical manuals for urban catchments. In contrast, OF 1's optimal set necessitated an expansion in the acceptable parameter ranges. Finally, from a computational burden viewpoint, OF 1 demanded a significantly higher number of function evaluations, thus implying an escalating computational cost as the range expands. Conversely, OF 2 necessitated fewer evaluations to converge toward the optimal solution.

Do baseline assumptions alter the efficacy of green stormwater infrastructure to reduce combined sewer overflows?

Green stormwater infrastructure (GSI) is growing in popularity to reduce combined sewer overflows (CSOs) and hydrologic simulation models are a tool to assess their reduction potential. Given the numerous and interacting water flows that contribute to CSOs, such as evapotranspiration (ET) and groundwater (GW), these models should ideally account for them. However, due to the complexity, simplified models are often used, and it is currently unknown how these assumptions affect estimates of CSOs, GSI effectiveness, and ultimately planning guidance. This study evaluates the effect on estimates of CSOs and GSI effectiveness when different flows and hydrologic processes are neglected. We modified an existing EPA SWMM model of a combined sewer system in Switzerland to include ET, GW, and upstream inflows. Historical rainfall data over 30 years are used to assess volume and duration of CSOs with and without three types of GSI (bioretention basins, permeable pavements and green roofs). Results demonstrate that neglect of certain flows in modelling can alter CSO volumes from -15 % to 40 %. GSI effectiveness also varies considerably, resulting in differences in simulated percent of CSO volume reduced from 8 % to 35 %, depending on the GSI type and modeled flow or process. Representation of GW within models is particularly crucial when infiltrating GSI are present, as CSOs could increase in certain subcatchments due to higher GW levels from increased infiltration. When basing GSI planning decisions on modeled estimates of CSOs, all relevant hydrologic processes should be included to the extent possible, and uncertainty and assumptions should always be considered.

Deep reinforcement learning challenges and opportunities for urban water systems

The efficient and sustainable supply and transport of water is a key component to any functioning civilisation making the role of urban water systems (UWS) inherently crucial to the wellbeing of its customers. However, managing water is not a simple task. Whether it is ageing infrastructure, transient flows, air cavities or low pressures; water can be lost as a result of many issues that face UWSs. The complexity of those networks grows with the high urbanisation trends and climate change making water companies and regulatory bodies in need of new solutions. So, it comes as no surprise that many researchers are invested in innovating within the water industry to ensure that the future of our water is safe. Deep reinforcement learning (DRL) has the potential to tackle complexities that used to be very challenging as it relies on deep neural networks for function approximation and representation. This technology has conquered many fields due to its impressive results and can effectively revolutionise UWS. In this article, we explain the background of DRL and the milestones of this field using a novel taxonomy of the DRL algorithms. This will be followed by with a novel review of DRL applications in the UWS which focus on water distribution networks and stormwater systems. The review will be concluded with critical insights on how DRL can benefit different aspects of urban water systems.

The effect of green infrastructure on resilience performance in combined sewer systems under climate change

Climate change is currently reshaping precipitation patterns, intensifying extremes, and altering runoff dynamics. Particularly susceptible to these impacts are combined sewer systems (CSS), which convey both stormwater and wastewater and can lead to combined sewer overflow (CSO) discharges during heavy rainfall. Green infrastructure (GI) can help mitigate these discharges and enhance system resilience under historical conditions; however, the quantification of its effect on resilience in a future climate remains unknown in the literature. This study employs a modified Global Resilience Analysis (GRA) framework for continuous simulation to quantify the impact of climate change on CSS resilience, particularly CSOs. The study assesses the efficacy of GI interventions (green roofs, permeable pavements, and bioretention cells) under diverse future rainfall scenarios based on EURO-CORDEX regional climate models (2085-2099) and three Representative Concentration Pathways (2.6, 4.5, 8.5 W/m2). The findings underscore a general decline in resilience indices across the future rainfall scenarios considered. Notably, the total yearly CSO discharge volume increases by a range of 145 % to 256 % in response to different rainfall scenarios. While GI proves effective in increasing resilience, it falls short of offsetting the impacts of climate change. Among the GI options assessed, green roofs routed to pervious areas exhibit the highest adaptive capacity, ranging from 9 % to 22 % at a system level, followed by permeable pavements with an adaptation capacity between 7 and 13 %. By linking the effects of future rainfall scenarios on CSO performance, this study contributes to understanding GI's potential as a strategic tool for enhancing urban resilience.

Green infrastructure optimization considering spatial functional zoning in urban stormwater management

Green infrastructure (GI) is used as an alternative and complement to traditional urban drainage system for mitigating urban stormwater issues mainly caused by climate change and urbanization. The combination of hydrological model and optimization algorithm can automatically find the optimal solution under multiple objectives. Given the multi-functional characteristics of GI, choosing the optimization objectives of GI are critical for multiple stakeholders. This study proposes a GI optimization method considering spatial functional zoning. Based on the basic conditions, the study area is divided into the flood risk control zone (FRCZ) and the total runoff control zone (TRCZ). The integrated model coupling hydrological model and optimization algorithm is applied to obtain the Pareto fronts and corresponding non-dominated solutions. The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method is used to support the decision-making process. The optimal solution obtained for the FRCZ achieves a flood risk reduction rate of 60.49% with an average life cycle cost per year of 0.20 × 108 Chinese Yuan (CNY); The optimal solution obtained for the TRCZ achieves a total runoff reduction rate of 22.83% with an average life cycle cost per year of 0.17 × 108 CNY. This study provides a reference for stakeholders in GI planning and design.

Monitoring and control of smart urban drainage systems using NB-IoT cellular sensor networks

Narrow-Band-Internet of Things (NB-IoT) is a promising wireless communication technology for harsh signal propagation conditions, including underground connectivity. The purpose of this study is to evaluate the real-life signal quality of an NB-IoT cellular sensor network for the monitoring and control of urban drainage systems. The implementation of the control systems toward smart infrastructures increase the need for low-cost and low-energy-consumption underground sensing devices. A measurement campaign was conducted in Rakvere, Estonia, to analyze the NB-IoT network's coverage and underground signal quality. The results were used to analyze the flood risk reduction potential in the pilot catchment, considering the quality of NB-IoT cellular network coverage. The signal measurements showed that in the case of plastic manholes, the best cellular transmitter placement is 0.5 m below the manhole cover, and at the ground level NB-RSRP reference signal strength of -104 dBm or better must be ensured to have a link budget with optimal energy consumption for communication. The results of flood risk reduction showed that excluding control locations with insufficient signal strength increased the number of major flooding nodes by 3.9% and mean flooding duration by 4.2% which is a relatively small change and characterizes the flexibility of the developed control system.

Practice makes the model: A critical review of stormwater green infrastructure modelling practice

Green infrastructures (GIs) have in recent decades emerged as sustainable technologies for urban stormwater management, and numerous studies have been conducted to develop and improve hydrological models for GIs. This review aims to assess current practice in GI hydrological modelling, encompassing the selection of model structure, equations, model parametrization and testing, uncertainty analysis, sensitivity analysis, the selection of objective functions for model calibration, and the interpretation of modelling results. During a quantitative and qualitative analysis, based on a paper analysis methodology applied across a sample of 270 published studies, we found that the authors of GI modelling studies generally fail to justify their modelling choices and their alignments between modelling objectives and methods. Some practices, such as uncertainty analysis, were also found to be limited, despite their necessity being widely acknowledged by the scientific community and their application in other fields. In order to improve current GI modelling practice, the authors suggest the following: i) a framework, called STAMP, designed to promote the standardisation of the documentation of GI modelling studies, and ii) improvements in modelling tools for facilitating good practices, iii) the sharing of data for better model testing, iv) the evaluation of the suitability of hydrological equations for GI application, v) the publication of clear statements regarding model limitations and negative results.

Towards coordinated and robust real-time control: a decentralized approach for combined sewer overflow and urban flooding reduction based on multi-agent reinforcement learning

The real-time control (RTC) of urban drainage systems can make full use of the capabilities of existing infrastructures to mitigate combined sewer overflow (CSO) and urban flooding. Despite the benefits of RTC, it may encounter potential risks and failures, which need further consideration to enhance its robustness. Besides failures of hardware components such as sensors and actuators, the RTC performance is also sensitive to communication failures between the devices that are spatially distributed in a catchment-scale system. This paper proposes a decentralized control strategy based on multi-agent reinforcement learning to enhance communication robustness and coordinate the decentralized control agents through centralized training. To investigate different control structures, a centralized and a fully decentralized strategy are also developed based on reinforcement learning (RL) for comparison. A benchmark drainage model and a real-world drainage model are formulated as two cases, and the control agents are trained to control the orifices or pumps for CSO or flooding mitigation in each case. The three RL strategies reduce the CSO volume by 5.62-9.30% compared with a static baseline in historical rainfalls of the benchmark case and reduce the CSO and flooding volume by 14.39-21.36% compared with currently-used rule-based control in synthetic rainfalls of the real-world case. Benefitting from centralized training, the decentralized agents can achieve similar performance to the centralized agent. The decentralized control also enhances the communication robustness with smaller performance loss than the centralized control when observation communication fails, and provides a robust backup at the local level to limit the uncertainties when action commands from the centralized agent are lost. The results and findings indicate that multi-agent RL contributes to a coordinated and robust solution for RTC of urban drainage systems.

Predictive heuristic control: Inferring risks from heterogeneous nowcast accuracy

Urban Drainage Systems can cause ecological and public health issues by releasing untreated contaminated water into the environment. Real-time control (RTC), augmented with rainfall nowcast, can effectively reduce these pollution loads. This research aims to identify key dynamics in the nowcast accuracies and relate those to the performance of nowcast-informed rule-based (RB)-RTC procedures. The developed procedures are tested in the case study of Rotterdam, the Netherlands. Using perfect nowcast data, all developed procedures showed a reduction in combined sewer overflow volumes of up to 14.6%. Considering real nowcast data, it showed a strong ability to predict if no more rain was expected, whilst performing poorly in quantifying rainfall depths. No relation was found in the nowcast accuracy and the consistency of the predicted rainfall using a moving horizon. Using the real nowcast data, all procedures, with the exception of the one predicting the end of the rainfall event, showed a significant risk of operative deterioration (performing worse than the baseline RB-RTC), linked to the relative performance of the nowcast algorithm. Understanding the strengths of a nowcast algorithm can ensure the reliability of the RB-RTC procedure and can negate the need for detailed modelling studies by inferring risks from nowcast data.

Evaluation of uncertain signals' impact on deep reinforcement learning-based real-time control strategy of urban drainage systems

Real-time control (RTC) is a recognized technology to enhance the efficiency of urban drainage systems (UDS). Deep reinforcement learning (DRL) has recently provided a new solution for RTC. However, the practice of DRL-based RTC has been impeded by different sources of uncertainties. The present study aimed to evaluate the impact caused by the uncertainties on DRL-based RTC to promote its application. The impact of uncertainties in the measurement of water level signals was evaluated through large-scale simulation experiments and quantified using measures of statistical dispersion of control performance distribution and relative change of control performance compared to the baseline scenario with no uncertainty. Results show that the statistical dispersion of DRL-based RTC was reduced by 15.48%-81.93% concerning random and systematic uncertainties compared to the conventional rule-based control (RBC) strategy. The findings indicated that DRL-based RTC is robust and could be reliably applied to safety-critical real-world UDS.

Generate_SWMM_inp: An Open-Source QGIS Plugin to Import and Export Model Input Files for SWMM

SWMM is an open-source model and software developed by the US EPA for the simulation of rainfall-runoff and routing in water bodies, sewer systems and wastewater infrastructures. It has been applied in numerous practical works and research projects. For a new SWMM model, objects such as nodes, links and catchments can either be drawn via SWMM’s graphical user interface (GUI) or specified manually in a plain text file in “.inp” format (“input file”). Since the required data regarding sewer geometries and river systems are usually available as spatial data in a GIS environment, there is a need for user-friendly interfaces for the model setup. SWMM contains neither an import function for geodata nor processing tools as provided in geographic information systems (GIS) such as the open-source software QGIS. Existing approaches were script-based or required commercial all-in-one products. We developed a free and open-source QGIS plugin to generate SWMM models from geodata and to import existing SWMM input files into QGIS. An application example is presented to demonstrate the basic features and usage of the plugin.

Optimizing the Control of Decentralized Rainwater Harvesting Systems for Reducing Urban Drainage Flows

The practice of rainwater harvesting (RWH) has been studied extensively in recent years, as it has the potential to alleviate some of the increasing stress on urban water distribution systems and drainage networks. Within the field, an approach of real-time control of rainwater storage is emerging as a method to improve the ability of RWH systems to reduce runoff and urban drainage flows. As applying real-time control on RWH tanks means releasing water that could be used for supply, applying controlled-release policies often hinders the RWH system’s ability to supply water. The suggested study presents an approach that has the potential to improve the capability of a distributed network of RWH systems to mitigate peak drainage flows while substantially reducing the impact on harvested rainwater availability. The suggested method uses a genetic algorithm to generate release policies, which are tailored for any given rain event and initial conditions. The algorithm utilizes the modeled drainage system’s response to a given rainfall pattern and manages to substantially reduce peak drainage flows with little impact on available rainwater when compared to the conventional no-release alternative and other active release methods.

Exploring the potential of utilizing unsupervised machine learning for urban drainage sensor placement under future rainfall uncertainty

Recently, advanced informatics and sensing techniques show promise of enabling a new generation of smart stormwater systems, where real-time sensors are deployed to detect flooding hotspots. Existing stormwater design criteria assume that historical rainfall frequency and intensity are reliable predictors to place real-time sensing devices. However, nonstationarity in rainfall due to climate change violates this assumption by disturbing hydrologic regimes and relocating flooding spots. This paper proposes a novel methodology of combining unsupervised machine learning (Agglomerative Clustering) and analysis of variance (ANOVA) to optimize the sensor placement under uncertain rainfalls. An urban drainage network located in Salt Lake City, Utah, USA, is chosen as the case study to demonstrate the application of the proposed method. Results show that: i) the proposed Agglomerative Clustering and ANOVA integrated approach can efficiently and accurately pinpoint sensor locations for drainage flooding detection; ii) rainfall uncertainty has limited impacts on the number of sensors, but it induces significant effects on sensor locations from the historical period (2000-2009) to the future period (2040-2049). By exploring the effects of climate nonstationarity on sensor placement, this work aims to help engineers and decision-makers better respond to the changing climates and rainfall extremes in urban drainage catchments.

Revealing the Challenges of Smart Rainwater Harvesting for Integrated and Digital Resilience of Urban Water Infrastructure

Smart rainwater harvesting (RWH) systems can automatically release stormwater prior to rainfall events to increase detention capacity on a household level. However, impacts and benefits of a widespread implementation of these systems are often unknown. This works aims to investigate the effect of a large-scale implementation of smart RWH systems on urban resilience by hypothetically retrofitting an Alpine municipality with smart rain barrels. Smart RWH systems represent dynamic systems, and therefore, the interaction between the coupled systems RWH units, an urban drainage network (UDN) and digital infrastructure is critical for evaluating resilience against system failures. In particular, digital parameters (e.g., accuracy of weather forecasts, or reliability of data communication) can differ from an ideal performance. Therefore, different digital parameters are varied to determine the range of uncertainties associated with smart RWH systems. As the results demonstrate, smart RWH systems can further increase integrated system resilience but require a coordinated integration into the overall system. Additionally, sufficient consideration of digital uncertainties is of great importance for smart water systems, as uncertainties can reduce/eliminate gained performance improvements. Moreover, a long-term simulation should be applied to investigate resilience with digital applications to reduce dependence on boundary conditions and rainfall patterns.

Exploring the Spatial Impact of Green Infrastructure on Urban Drainage Resilience

This paper explores the spatial impact of green infrastructure (GI) location on the resilience of urban drainage systems by the application of exploratory spatial data analysis (ESDA). A framework that integrates resilience assessment, location sensitivity analysis and ESDA is presented and applied to an urban catchment in the United Kingdom. Three types of GI, namely a bioretention cell, permeable pavement, and green roof, are evaluated separately and simultaneously. Resilience is assessed using stress-strain tests, which measure the system performance based on the magnitude and duration of sewer flooding and combined sewer overflows. Based on the results of a location sensitivity analysis, ESDA is applied to determine if there is spatial autocorrelation, spatial clusters, and spatial outliers. Results show a stronger spatial dependency using sewer flooding indicators. Different GI measures present differences in spatial autocorrelation and spatial cluster results, highlighting the differences in their underlying mechanisms. The finding of conflicting spatial clusters indicates that there are trade-offs in the placement of GI in certain locations. The proposed framework can be used as a tool for GI spatial planning, helping in the development of a systematic approach for resilience-performance orientated GI design and planning.

Efficient integration of IoT-based micro storages to improve urban drainage performance through advanced control strategies

The smart rain barrel (SRB) consists of a conventional RB with storage volumes between 200 and 500 L, which is extended by a remotely (and centrally) controllable discharge valve. The SRB is capable of releasing stormwater prior to precipitation events by using high-resolution weather forecasts to increase detention capacity. However, as shown in a previous work, a large-scale implementation combined with a simultaneous opening of discharge valves clearly reduced the effectiveness. The aim of this work was to systematically investigate different control strategies for wet weather by evaluating their impact on sewer performance. For the case study, an alpine municipality was hypothetically retrofitted with SRBs (total additional storage volume of 181 m³). The results showed that combined sewer overflow (CSO) volume and subsequently pollution mass can be reduced by between 7 and 67% depending on rain characteristics (e.g., rain pattern, amount of precipitation) and an applied control strategy. Effectiveness of the SRBs increases with lower CSO volume, whereas more advanced control strategies based on sewer conditions can clearly improve the system's performance compared to simpler control strategies. For higher CSO volume, the SRBs can postpone the start of an CSO event, which is important for a first-flush phenomenon.

Deep Reinforcement Learning with Uncertain Data for Real-Time Stormwater System Control and Flood Mitigation

Flooding in many areas is becoming more prevalent due to factors such as urbanization and climate change, requiring modernization of stormwater infrastructure. Retrofitting standard passive systems with controllable valves/pumps is promising, but requires real-time control (RTC). One method of automating RTC is reinforcement learning (RL), a general technique for sequential optimization and control in uncertain environments. The notion is that an RL algorithm can use inputs of real-time flood data and rainfall forecasts to learn a policy for controlling the stormwater infrastructure to minimize measures of flooding. In real-world conditions, rainfall forecasts and other state information are subject to noise and uncertainty. To account for these characteristics of the problem data, we implemented Deep Deterministic Policy Gradient (DDPG), an RL algorithm that is distinguished by its capability to handle noise in the input data. DDPG implementations were trained and tested against a passive flood control policy. Three primary cases were studied: (i) perfect data, (ii) imperfect rainfall forecasts, and (iii) imperfect water level and forecast data. Rainfall episodes (100) that caused flooding in the passive system were selected from 10 years of observations in Norfolk, Virginia, USA; 85 randomly selected episodes were used for training and the remaining 15 unseen episodes served as test cases. Compared to the passive system, all RL implementations reduced flooding volume by 70.5% on average, and performed within a range of 5%. This suggests that DDPG is robust to noisy input data, which is essential knowledge to advance the real-world applicability of RL for stormwater RTC.