ClimaGrid

heatmaps
city builder
Generated satellite imaging
The diffusion process done by the model. From right to left.

Name and School: Georgiy Derkachev, Lakshya Dhakar, Neil Mehra, Advik Vatsyayan – McNair academic Highschool Date: 07/10/26 Category: Sustainability - GreenSpan Urban Development Project Name: ClimaGrid Description: A website to help urban planners mock up quick designs and stress test them under real intense conditions.

Inspiration

Our inspiration for this project came from the company we chose- GreenSpan Urban Development. This company works to build net-zero cities, but they run into problems regarding scaling systems and preserving livability. We decided we could solve this issue using our system even after taking into account the natural and urban factors in the area. Additionally, sustainability is a growing issue in the real world as green spaces are decreasing at an alarming rate, while the population is rapidly increasing. We believe our solution can also have real world impacts that could alleviate the growing issue we are all facing.

What it does

Firstly, the user builds a grid city with four options for each square— water, green space, low density housing, or high density housing. Afterwards, the model will evaluate the urban heat island effect, energy usage, and pollution. Additionally, the grid is sent to a diffusion model which generates a satellite image of the city.

How we built it

We built this project using flask, typescript, python, react, and pytorch.

Challenges we ran into

Some of the challenges we ran into while creating this project were:

The backend was very unstable, rendering the site completely dysfunctional. The API crashed and went down, so we couldn't test the project for a few hours.

Accomplishments that we're proud of

Some of the accomplishments we are proud of are:

Designing a clean and easy-to-use site for users. Incorporating the many different heatmaps that correspond correctly to the original city grid map. Generating an accurate aerial map from the custom city built.

What we learned

One of the many things we learned was how to use diffusion models. Additionally, we learned how to create the many heatmaps we used in our project. We also learned how to link the frontend and backend with the machine learning models we made.

What's next for ClimaGrid

Our main focus for improving this product would be to successfully add a resource allocation model. This model will allocate the necessary and available resources to the areas that are deemed to need this resources to improve the well being of the nature and the people in the region. This model will significantly boost our product as it will add a logical and viable solution to the unique problems identified in each region.