Advancing Autonomous Green Stormwater Infrastructure

Abstract

Flooding causes more damage and fatalities than any other natural disaster. Simultaneously, storms carry large quantities of pollutants into receiving waters. These challenges are compounded by urbanization and a changing climate. Traditional infrastructure solutions, such as larger storage basins and pipes, are cost prohibitive. Green infrastructure (GI) has been proposed as a nature-based alternative to new construction. GI includes assets such as rain gardens and bioswales, which are designed to capture runoff, treat pollutants, and infiltrate water into underlying soils. However, the scalability of these distributed solutions has yet to be vetted and measured at scale. Another promising solution, autonomous stormwater systems, leverage recent advances in wireless sensing, communications, and controls. Infrastructure assets are retrofitted with wireless sensors and controllable valves, enabling them to adapt to changing weather, flows, and pollutant loads. These controllable assets are coordinated at the system-scale to achieve flooding and pollutant objectives across watersheds. While these novel solutions have gained traction for traditional stormwater infrastructure, they have only just begun to be investigated for GI. Before autonomous technologies can be adopted for GI, several fundamental knowledge gaps must be closed. Broadly, these include a lack of understanding around how GI can be effectively and optimally measured at scale, as well as how pollutant transformations should be modeled to support real-time control. This dissertation addresses these knowledge gaps and presents foundational work towards enabling autonomous GI. The second chapter introduces an end-to-end data toolchain, underpinned by a wireless GI sensor network for continuously measuring real-time water levels. The toolchain automatically isolates storms in the sensor data to parameterize a dynamical system model of GI drawdown dynamics. The model outputs are then used to investigate the explanatory features of drawdown dynamics. We show how investments in monitoring networks support a more targeted and data-driven approach to GI design, placement, and maintenance. Investing in monitoring networks requires knowing where to place sensors and how many are needed. The third chapter introduces a sensor placement methodology for urban drainage networks using publicly available datasets and Gaussian Processes. The methodology is flexible enough to work for any stormwater sensor or spatial parameter of interest, has guarantees of optimality, and is computationally efficient. We show that the methodology maximizes information gained from the sensor network while minimizing its size. The fourth chapter addresses the lack of simulation tools necessary for modeling complex pollutant transformations affected by real-time control in urban drainage networks. A new water quality package, StormReactor, is introduced. StormReactor provides an open-source Python programming interface for simulating complex pollutant generation, treatment, and real-time control processes. Two case studies are presented to illustrate the fidelity of StormReactor and the potential of using real-time control for ecological benefits. Expanding upon these case studies, the fifth chapter evaluates the impact of real-time control on pollutant removal in GI. StormReactor is used to simulate real-time control of a real-world inspired GI to capture phosphorus. We show that real-time control not only provides a “digital” alternative to existing, passive GI upgrades, like soil amendments, but it also provides long-term flexibility. This flexibility enables stormwater managers to dynamically balance trade-offs in existing GI designs and aids in the larger goal of system-level control. The dissertation concludes with a broader discussion on how the discoveries made can support the development of a new generation of autonomous GI.