How Data-Driven Rain Garden Design Can Mitigate Climate Change Impacts

Stormwater that runs off buildings and paved streets frequently carries high levels of pollutants into rivers and lakes, some of which serve as sources of drinking water. Rain gardens, green infrastructure initiatives designed to mimic natural systems such as native prairies, help keep urban runoff out of sewers and surface waters, suppressing stormwater’s negative effects.

Constructing rain gardens to deliver maximum benefit, however, takes more than just soil, a shovel, and a green thumb. Using machine learning, Northwestern Engineering’s Kimberly Gray is working to make sure urban green infrastructure is designed to be ecologically restorative and mitigate climate change.

Gray, Roxelyn and Richard Pepper Family Chair in Civil and Environmental Engineering, and her team collected data on a dozen rain gardens in Evanston. They then compared the rain gardens to natural prairies to identify which characteristics contribute to the best results.

“We’ve taken ecological and chemical data and studied it using machine-learning techniques to identify clusters of the most beneficial areas,” Gray said. “We then determined what are the most important attributes that lead to both successful water retention and biodiversity enhancement. In this way we identify when they are most like a natural system, and what causes them to perform differently.”

Over the course of four years, the team collected data on various factors, including biodiversity, soil quality, pollination rates, and seasonal effects such as temperature and rainfall. They paid close attention to how each rain garden performed throughout different weather conditions, documenting how seasonal changes influenced their effectiveness in capturing stormwater and supporting local flora and fauna. By studying these spaces across seasons, they aimed to see which characteristics made the rain gardens function more like natural prairies, which are known for their resilience and capacity to sequester carbon.

Gray’s team then compared the data from these rain gardens as a function of various features (e.g., stormwater source) and with data from a local prairie. This comparison was essential to identify which rain garden features could best replicate the ecological benefits of native prairies.

“The idea is to bridge the gap between green infrastructure and natural ecosystems,” Gray said. “We’re trying to understand how rain gardens can function as ecological powerhouses within an urban setting.”

Leveraging machine learning

With a wealth of environmental data, it would be almost impossible to identify patterns manually, especially when considering the numerous variables involved, from plant species diversity to soil nitrogen levels. Machine learning, Gray said, allows the researchers to discern relationships and trends that may otherwise remain hidden.

“We used machine learning algorithms to analyze data across multiple variables simultaneously,” Gray said. “This approach lets us pinpoint which characteristics are most influential in enhancing rain garden performance.”

By feeding vast amounts of data into these algorithms, the team could uncover the key features that made the systems behave more like the natural systems they are designed to mimic. These insights then allow them to predict which rain garden designs would yield the most engineering, ecological and climate benefits. The machine learning models helped highlight certain features as critical for improving a rain garden's function. For example, rain gardens, in general, have much less diversity than native systems but there is also a large degree of variation among Evanston rain gardens. Although many local rain gardens were dominated by native vegetation the plant assemblages were of “resilient” quality and supported fewer high-quality species common in native prairies. 

Urban rain gardens, however, bear much greater plant density than native systems and surprisingly, can be much more attractive to pollinators than native prairies.  

“Details about plant diversity was a major finding of our study,” Gray said.  “The plants matter.  Diverse rain gardens aren’t just visually appealing — they are also better at supporting pollinators and maintaining soil quality, which are crucial for the garden’s long-term resilience.”

Another finding suggested that the deeper-rooted native plants were better at capturing, storing and pumping to the atmosphere stormwater, a vital feature as cities seek solutions to manage flooding in the face of more extreme weather events. The team delineated the degree to which rain gardens mimic prairie-like conditions and ideally, although rain gardens are small, patchy and randomly placed on the urban landscape, it is possible that in the future scientists can determine how they add up to be ecologically restorative, particularly of biodiversity, a key component of climate action. 

“What’s exciting about this approach is that we can optimize rain garden designs for multiple benefits: reducing stormwater runoff, supporting biodiversity, and even mitigating climate change by sequestering carbon,” Gray said.

A roadmap for resilient urban green spaces

Gray’s work has implications beyond Evanston’s rain gardens. By providing a data-driven blueprint, they are creating a framework that cities worldwide can use to develop urban green spaces that are not only beautiful but functionally powerful in addressing environmental challenges. With the ability to identify specific design features that enhance performance, municipalities can make evidence-based decisions on how to plan urban green infrastructure.

Gray sees the potential for this research to inform policy and planning in urban environments.

“Our goal is to offer cities a template that they can customize to their specific needs, based on solid scientific data,” she said. “This isn’t just about one rain garden or one city — it's about reimagining the urban landscape on a broader scale.”

Beyond its practical applications, Gray’s work underscores the importance of treating urban green spaces as ecosystems, capable of providing critical services for climate adaptation and ecological health. As the climate crisis accelerates, cities need every tool at their disposal to create resilient infrastructures that protect both people and nature.

Green infrastructure, like rain gardens, has the potential to absorb excess stormwater, capture carbon, and support local biodiversity, ultimately contributing to a more balanced urban ecosystem.

“The more data we collect, the more accurate our predictions become,” she said. “Each season and every rain garden add a piece to the puzzle.”