
By Anasa Laude, LEED AP BD+C
Years ago, I was part of a student-faculty team called Greenproofing, funded by NOAA and co-founded with interdisciplinary team of faculty including Kevin Foster, now deputy dean at the Colin Powell Center. We worked alongside Gioya DeSouza-Fennelly, a science teacher at Teachers College, who advised the project and helped bridge the work between college and high school students. Our mission was straightforward: install and maintain green roofs on schools and affordable housing sites across the city. We installed and maintained a roof at Lantern Group, a housing provider serving formerly homeless residents and youth aging out of foster care, with DYCD and CCNY students doing the hands-on work. The work was community-driven and grounded in real places. We were thinking about heat, air quality, and stormwater. We were not thinking about AI.
I’m returning to that work now with different eyes. Not to romanticize it, but to ask what it would look like if students today could engage with green roof design at the concept level, using the tools available to them now, including AI. What follows is a framework for a multi-day unit that teachers can adapt for environmental engineering, soil science, botany, math, or physics. It works as a standalone unit or as an interdisciplinary thread across classes.
What Green Roofs Actually Do
Before getting to the AI piece, it helps to understand what a green roof is doing and why it’s worth teaching. A green roof is a layered system: a waterproof membrane, a drainage layer, a growing medium, and vegetation. Together, those layers reduce stormwater runoff, lower surface temperatures through evapotranspiration, filter air, and add insulation to the building below. In dense urban areas, they’re one of the most effective interventions for managing heat at the neighborhood scale.
What makes green roofs interesting for classroom use is that they sit at the intersection of multiple disciplines. The growing medium is a soil science question. The plant selection is a botany question. The thermal performance is a physics question. The load calculations are a math and structural question. That’s before you get to the planning, community context, and systems thinking.
Where AI Enters the Picture
AI tools are changing what’s possible in green roof planning, and they’re doing it in ways that translate well into classroom learning.
At the planning stage, AI-assisted mapping tools can analyze satellite imagery to identify rooftops that are structurally viable, appropriately oriented, and large enough to be worth treating. In a classroom context, students can explore publicly available building and zoning data and think through what criteria they would use to rank candidate sites. That’s a real planning problem, and it teaches spatial reasoning, data interpretation, and prioritization.
At the design stage, AI tools can help model thermal performance, water retention, and plant coverage across different configurations. Students can use simplified modeling prompts to ask: what happens to runoff if we increase the depth of the growing medium? What happens to the heat island effect if we plant sedums versus grasses? These aren’t hypothetical questions. They’re the same questions practitioners are working through.
At the management stage, AI supports real-time monitoring of soil moisture, temperature, and plant health through sensor networks. This is where the data science connection lives: interpreting sensor data, identifying anomalies, and making decisions based on what the data shows.
Green Roofs and Solar: A Natural Pairing
One concept worth introducing explicitly is the relationship between green roofs and solar panels. On a conventional roof, solar panels heat up in direct sun and lose efficiency as temperature rises. A green roof moderates that surface temperature, which can improve solar panel performance. Some building systems integrate both, with panels elevated above the growing medium so both systems coexist.
For students, this is a useful design constraint to work with. How do you allocate roof space between vegetation and panels? What’s the tradeoff between maximizing solar generation and maximizing stormwater management? These are authentic engineering design challenges that don’t require a construction budget to think through seriously.
Bringing It Into Specific Classrooms
The unit works differently depending on the subject, but the connective tissue is the same: students are designing or evaluating a system, making decisions based on evidence, and accounting for tradeoffs.
In a soil science class, the focus is on the growing medium itself. Green roof substrates are engineered: they need to be lightweight, well-draining, and capable of supporting plant life with minimal organic matter. Students can compare engineered substrates to natural soils, examine drainage rates, and think through how substrate depth affects both plant selection and building load.
In a botany class, plant selection becomes the central problem. Green roofs in temperate climates often rely on sedum and other succulents because they’re drought-tolerant and shallow-rooted. But plant selection is also a function of climate, sun exposure, and maintenance capacity. Students can research plant communities, design planting plans, and discuss how biodiversity on a rooftop compares to a ground-level garden.
In a math class, the numbers are everywhere. Load calculations require understanding weight per square foot and structural capacity. Stormwater modeling involves volume, surface area, and absorption rates. Solar energy estimates involve angle, surface area, and efficiency percentages. These are applied problems that connect directly to what students are learning in algebra and geometry.
In a physics class, the focus can shift to thermal dynamics. How does evapotranspiration work and why does it cool surfaces? What is the difference between reflectance and thermal mass? How do green roofs compare to cool roofs and white roofs in terms of heat management? These questions ground thermodynamics in a real urban application.
A Note on AI’s Own Footprint
I’d be leaving something out if I didn’t name this: AI tools use energy, and large models use a lot of it. As Ray Garcia has reminded me, a tool is only as good as how we use it and the values we bring to the work. Part of teaching AI in this context is teaching students to ask that question themselves. If we’re using AI to design systems that reduce urban heat, what’s the energy cost of the AI itself? Is it running on renewable energy? Are we using it efficiently? These aren’t rhetorical questions. They’re design criteria.
That critical framing belongs in the unit, not as a caveat but as part of the inquiry.
So What Is AI Teaching Us Now?
That’s the question underneath all of this. And working through it in the context of green roofs has sharpened my thinking in ways I didn’t expect.
AI is teaching us how complex these systems really are. When you ask an AI to model a green roof, you have to specify everything: substrate depth, plant type, climate zone, roof orientation, building load. That specificity is the lesson. Students who try to prompt their way through a design problem quickly discover how many variables they were glossing over. The friction is productive. It reveals what they don’t yet know.
AI is teaching us where human judgment lives. The mapping, the modeling, the monitoring — AI can support all of it. What it can’t do is decide who the roof is for, or whether the community had a say in the design, or what it means for formerly homeless residents to have a garden above their heads. The limits of the tool point directly to what humans must bring to the work.
AI is teaching us that optimization is not the same as design. A model can maximize stormwater retention or solar output. It cannot account for a community’s relationship to a place. That gap isn’t a flaw to be fixed. It’s a design criterion — and teaching students to name it is one of the most important things we can do.
And AI is teaching us to ask better questions. In the classroom, the prompt is the lesson. Students who learn to interrogate an AI output — to ask why it recommended one substrate over another, or whether the energy cost of the model outweighs its benefit — are developing the same critical thinking they need to evaluate any data source, any system, any claim.
That’s what Greenproofing was always about, before any of us knew what a large language model was. You put students in proximity to a real problem, with real constraints, serving real people. The thinking follows. AI is just the newest set of constraints worth thinking through.
Questions Worth Sitting With
I want to name something before closing. Green infrastructure is not neutral. Green roofs are expensive. In New York City they’re far more common on luxury buildings than on affordable housing, even though the communities that need heat relief and flood mitigation most are often the ones least likely to have them. Green gentrification is real — the greening of a neighborhood can accelerate displacement. Permitting and building codes in many cities make even modest green infrastructure costly and slow. The average person anywhere in the world cannot afford an architect to help them design one.
AI carries its own version of this tension. It is largely controlled by a small number of companies with enormous power and unclear accountability. Its energy footprint is not trivial. And its benefits are not evenly distributed.
I still think both green infrastructure and AI can help us solve hard problems. I think that’s worth teaching. But I also think the most important thing we can teach students is how to hold the promise and the problem at the same time — and to ask who benefits, who decides, and who pays.
These aren’t questions with right answers. They’re questions worth exploring:
- Green roofs reduce heat and flooding. But who gets them, and who doesn’t? What does that tell us about how cities make decisions?
- If green infrastructure increases property values, what happens to the people who already live there?
- AI can help design and monitor green systems. But who controls those tools, and who has access to them?
- The average person can’t afford an architect or an AI consultant. Does that change what these tools are actually for?
- If we use AI to build more sustainable cities, but AI itself consumes significant energy, how do we weigh that tradeoff?
- What would it look like for a community to own and govern its own green infrastructure — and its own data?
That’s the full piece. The lesson breakdown chart from earlier pairs with it as a standalone resource teachers can download or reference separately. Want me to create a clean document version you can save and edit?











