NC Water Watch

Inspiration

I was inspired by a movie I watched a while back called Dark Waters(2019). It is about a lawyer going after the infamous chemical company DuPont, whose non-stick pots and pans were known use 'teflon', a type of PFAS chemical. Teflon and other PFAS chemicals are known to last for extremely long periods in the environment before breaking down and cause horrific health effects in humans such as birth defects in the children of affected pregnant women or heightened risk of certain cancers. My home state of North Carolina happens to be one of the worst in terms of PFAS contamination and my city's water exceeds the EPA's safe limit for these chemicals. So that's why I felt obligated to create the NC Water Watch.

What it does

I'm going to break my project down in 3 parts:

Interactive map

Powered by machine learning, users can search through the map for their city and find if their water is polluted with PFAS chemicals, along with the most likely reason as to why it might be polluted.

Quizzes

Users can quiz themselves on common questions and misconceptions about PFAS chemicals and the way they spread. Each time they get a question right they get 10 points and a better chance to land a spot on the leaderboard :)

Email their representatives

Users can input some basic info about themselves and then with a click of a button, email those who represent them in the North Carolina State Legislature as well as in Congress who have genuine power to vote on bills that can help contain the releasing & spreading of these chemicals.

How we built it

Interactive map

I built this by collecting a small dataset of water sources from the nc pfas network, then tested on other bodies of water to see if they were polluted using supervised machine learning and logic from this study done in California. Some of the features I used to predict if they were polluted or not include distance from industrial polluters, amount of AFFF firefighting foam used in a 25km radius, distance from military bases, and many more. Predicting using those, I achieved a final accuracy of 94.87% and an AUROC of 94.84% using the light gradient boosting machine. After that I had 130+ water sources across North Carolina which were used by over 6 million people and my model's prediction on whether they were polluted with pfas or not. I finally put all that data on an interactive map which included all the water sources locations, the population it serviced, whether it was polluted, and if so why it was polluted(ex: nearby military base or industrial plants). TLDR: Used supervised learning to predict if body of water was polluted or not with pfas, I achieved a 94.87% accuracy on a test set using a light gradient boosting machine. Put all data on map.

Quizzes

Little bit less of a word jumble than above but I built this using django to store a user's points and add to a users points. Also used a little javascript and ajax to send a request to the backend to add points to a user when they got a question right.

Email their Representatives

I built this using the Google Civics API which retrieved a user's representatives in their state as well as in congress, from there I used yagmail to send an email to those representative's emails with the user's information attached and a pre built message.

Challenges we ran into

I had issues into getting an accurate ML model for the interactive map, for a while it was stuck around the 70-80% range before I had to collect a bunch more data and use different models till I could get around the 95% mark. ## What we learned I learned a lot about the data collecting process and about using django and javascript together. Even outside of coding, I learned a lot more about PFAS chemicals than I knew coming in, how they spread, how they affect people, etc.

What's next for NC Water Watch

I hope to predict pollution accurately in every body of drinking water across NC to cover the 4 million other North Carolinians whose water source wasn't covered. As well as maybe extend to other states in the future, if I had infinite time for this competition I would've done it for the entire United States.