Backstory
It was December 2023, and I was back home for winter break. I figured it'd be as good a time as any to take on my first real machine learning project (not counting that questionable exoplanet prediction model from one year prior). So I decided to sign up for the NASA Pale Blue Dot Challenge, where participants must use Earth Observation satellite data to create a visualization that advances the UN's Sustainable Development Goals (SDGs).
Taking inspiration from a service project that I had worked on in Cacao Colon, Honduras, where the only source of clean water was a drilled groundwater well, I imagined an ML tool that could provide free access to groundwater data; thus, alleviating surveying costs for communities like Cacao Colon. And there was my project idea.
I made a little forum post to see if anyone might be interested in joining my project, and to my delight, people all around the world started responding. Before long, I found myself at the helm of Team Viva Aqua — a motley group of 4 engineers, data scientists, geomaticians, and anthropologists hailing from Argentina, Senegal, and the US.
The Project
Note: Detailed documentation for this project is available on our GitHub
With only 5 weeks left till the deadline, there was little time to waste. I put together a plan, laid it out on our Gantt chart, and we were off. The first week, I spent a lot of time researching, as we learned more about the problem and tried to identify a specific region to focus on for our visualization. We ended up choosing Gambia, due its lack of access to clean water and availability of open-source data.
Then, we downloaded the data and got to the tedious task of cleaning and preprocessing all of it. I even made my own Python library, climateservAccess, to help with this. But despite our best efforts, we fell behind schedule, and by the time we finished compiling our dataset of ~6600 pts, there was only 1 week left.
That was also the point where my laptop decided to stop working. So while I tried and failed to fix my broken laptop, Francisco pulled through and got the ML models trained. Then, we took input data for West Gambia, ran it through our best model, and at last we had groundwater level estimates.
By that time, I had managed to get a new laptop, so I visualized the data using kepler.gl, put it on our website, and we were (almost) done — except for our report. So we grinded that out, submitted everything with minutes left to spare, and celebrated over a call. Mission complete!
Aftermath
Two months later, the news broke...
WE HAD WON!!
Or more accurately, we were 1 of 5 teams who won Best Overall, along with the grand prize: an absolutely unforgettable, 10-day, all-expenses paid Space Study Program to Huntsville, Alabama and Washington DC.
In June, we were also invited to present our project at UNOOSA's COPUOS 2024 (Committee on the Peaceful Uses of Outer Space) - which, needless to say, was an amazing experience.
And of course, all of this has been fantastic. However, the work is not done. Aqua Viva still has a long way to go to reach its original vision, and in order to do that, we're going to need help. So if you or someone you know is interested in contributing to this project, or just learning more about it, please let me know!
Aqua Viva poster displayed at COPUOS 2024
Presentation slides and images from COPUOS 2024