Inspiration
The vast majority of natural disasters resulting in loss of life are water-related. We sought to consider an emerging issue that is currently underreported outside of immediate localities, yet could have significant ramifications on wider communities as the climate crisis develops.
What does our project do?
SnowPast is an XGBoost-powered ML solution that analyses historic data to predict how much water will be produced from melting snowpacks, providing municipal authorities and nongovernmental organizations with more information to prepare for outlier events. We hope that researchers may also use our model to create preventative measures for risks associated with water-related natural disasters in vulnerable areas.
How we built it
We used Python and TensorFlow to build our model and utilized Snotel datasets for training and validation. Our web app runs on a Django framework, deployed via Heroku, and we use MongoDB to store our data in an easily queryable format. The solution frontend is built with React and native CSS.
Challenges we ran into
Finding datasets and model architectures that were appropriate for the task was a challenge! We also struggled with receiving real-time data to show to the end user.
What are we proud of?
Finishing our project! We're extremely proud that we managed to completely implement a solution, especially given that this is a beginner hack.
What did we learn?
Full-stack web development, AI models, data parsing and pre-processing, the Django framework, MongoDB, snowpack formation and melt factors, how to pull an all-nighter, and how to get a great team room in Levine Hall.
What's next?
Continuing to use and build upon our prior knowledge to create meaningful real-world projects! It'd be great to find a larger dataset that contains information outside of California to improve our model's resilience, or to consider other area-based predictors for natural disasters or other phenomena affecting communities and ecosystems.
Log in or sign up for Devpost to join the conversation.