ThermoTracker

Inspiration

Amishi was taking a class on the effects of climate disasters, so the prompt was really interesting to her, leading us to choose to do this project.

What it does

• Shows the weather forecast based on user location
• Uses LLM to generate suggestions to combat increasing urban heat
• Uses a clustering model in order to determine which areas are most susceptible to urban heat problems

How we built it

We used Streamlit to host the app, used Streamlit-Folium to create a visual map, and used geopy.geocoders with Nominatim to give address suggestions and information based on user text input, and used the NOAA weather api to get the weather and any current alerts in that area. We aggregated data using geopandas, and used sklearn to cluster and clean the data. We then used that data to determine whether or not the user's location is at high heat risk. Using the weather data from NOAA and the coordinates of their location, we also used Gemini to give household suggestions to improve heat resilience

Challenges we ran into

We have never worked with such large datasets or with geospatial data, so learning how to process that and learning how to efficiently deal with datasets of that size took up a lot of our time before we were able to use the data.

Accomplishments that we're proud of

We are proud that we were able to implement the large datasets and that we were able to use Gemini to make suggestions to the user

What we learned

As stated previously, we learned how to work with geospatial data and large datasets. We also learned how to use streamlit-folium to create interactive maps

What's next for ThermoTracker

We plan to use larger datasets and more predictor variables to expand the scope to the entire North Carolina area, and eventually make it countrywide.

Built With

• durham
• gemini
• geopandas
• matplotlib
• noaa-climate-data-online
• nominatim
• numpy
• pandas
• python
• shapely
• streamlit