Inspiration
Global warming has accelerated the melting of glaciers and ice has driven the sea levels to increase. This increase is already affecting billions worldwide. We wanted to provide a way for people who may be affected to get a better understanding of their environment by creating a visually easy-to-understand website and providing them with a way of making better-informed decisions by seeing predictions for the near future.
What it does
Displays data from the past 50 years and predicted data for the next 75 years using machine learning to predict how the coastline may change each year based on previous data. Zooming in on a certain zone will allow the user to see the shift more clearly. A slider can be used to go back in time or go forward into the future, and a button can be used to animate the coastline from the year chosen until the maximum future scenario.
How we built it
We are a team of 4 engineers, and all of us had a specific focus on the project. One of us focused on finding and extracting the data, and then on optimizing the delivery of the data. 2 of us focused on the frontend and the backend of the project to make it visually appealing and add our novel features. Our final engineer focused on developing and training the machine learning model.
Challenges we ran into
- Finding a reliable, high-quality dataset of coastline position was extremely challenging, and even after finding the right one, we had to significantly transform the data.
- We had to put in a lot of work to ensure that we served the right data to the user, with the right fidelity, in a reasonable amount of time.
- We put a lot of work into extracting the insights from the output of our model and presenting them in a clean and impactful manner.
- The new model we were training for less noisy data ended up crashing in the middle of the night while all of us were sleeping.
Accomplishments that we're proud of
Firstmost, we are proud to have participated and finished our project! We are proud of:
- Developing and training a working transformer to predict coastline movement in such a short time.
- The data structures we developed to store and serve coastline data.
- Creating a functional and clean interface to interact with our data.
What we learned
We learned how to work with real-world geospatial data and leverage algorithms and data structures to efficiently serve vast, complex bodies of data.
What's next for Receding Coastlines
We want to move from Florida to include the entirety of the US. We also want to train it on additional data to see how other environmental factors such as temperature may affect the rate of coastline movement.
Log in or sign up for Devpost to join the conversation.