Inspiration
While brainstorming ideas for this hackathon, our breakthrough came in the most fitting place possible, walking along the Indianapolis Canal. At Purdue Indianapolis, the Canal is something of a landmark. There's a running joke among some students that you jump in when you graduate. But standing there, looking at the murky, algae-streaked water, that joke started to feel a little less funny. What we thought was a dirty canal in downtown Indy, turned out to be a symptom of one of the most widespread and underreported environmental crises on the planet: harmful algal blooms.
Every year, our global water systems are strangled by toxic algal blooms. They don’t just turn water green; they contaminate drinking water, kill aquatic ecosystems, and threaten public health at a scale most people never hear about. They cost the global economy over $8 billion annually in healthcare, lost tourism, and collapsed fisheries (World Bank 21). But the most staggering statistic isn't the dollar amount, but the accessibility gap. In the U.S., a single "No Swim" advisory is often a sign of a failed reactive system. In developing nations, there are often no signs at all. There is only the sudden death of livestock, the loss of a season's harvest, and communities left wondering why their water source turned into poison.
In the United States alone, localized events can devastate regional economies by $1 billion in lost property value and tourism (National Ocean Service). Currently, the "Gold Standard" for monitoring involves expensive physical sensors and laboratory testing. Even in America, this is not nationally adopted or consistently applied. In the Global South, where fish provide a primary protein source for millions, this level of infrastructure is a financial impossibility. These regions are forced to carry the burden of industrial and agricultural runoff with zero tools to hold polluters accountable (Ho et al. 567).
What it does
Our project leverages the billions of dollars already invested in satellite infrastructure, like NASA MODIS-Aqua, and pairs it with a RAG-based attribution engine. By having nations like the United States utilize and refine this satellite infrastructure, we take the financial burden of detection off of developing countries that cannot support it. We’ve built an end-to-end pipeline that identifies toxic algal blooms via satellite and then uses RAG to correlate those blooms with potential upstream sources for earlier detection.
How we built it
Starts with a simple location search, where our system uses a Haversine-based BallTree to pinpoint the nearest major water bodies
Once a target is identified, the model takes input of real-world chlorophyll-a concentrations as well as heavy wind/rain directly from our satellites. We pull 20 years of spectral history to understand exactly what 'normal' looks like for that specific grid cell
We then apply context through feature engineering - calculating seasonal residuals and Z-scores while layering in precipitation data to see if a spike of algae was triggered by a recent rain event.
To ensure the highest accuracy and eliminate false alarms, we use a three-way ensemble detection model: a statistical Z-score check, an Isolation Forest machine learning algorithm, and a spatial autocorrelation check that verifies the bloom is a physical cluster rather than sensor noise
When a bloom is confirmed, we query the OpenStreetMap Overpass API to identify every industrial facility, wastewater plant, and farm within a 25-kilometer radius. We feed this live metadata into Gemini 2.0 Flash, which generates a structured Community Report
Challenges we ran into
At first, we wanted to utilize Sentinel-3, another satellite infrastructure, but it required paid API keys to access some of the premium features we wanted to implement. So we decided to use NASA's MODIS-Aqua satellite.
Initially, when we were trying to find potential upstream sources of algal pollution, we only got garbage values or areas that didn't exist. We further looked into OpenStreetMap API to get more accurate representations of industrial/agricultural sources.
Accomplishments that we're proud of
We believe that clean water shouldn't depend on where you live. We aren't just detecting algae; we are providing a real, consistent, and global solution that identifies runoff sources before they destroy an ecosystem.
What we learned
Developing Bloompoint was a masterclass in the gap between raw data and actionable insight. We quickly learned that while the world isn’t lacking in environmental data, it suffers from a massive accessibility crisis. Shifting our focus from paid services like Sentinel-3 to NASA’s open-source MODIS-Aqua infrastructure taught us that high-performance detection doesn't have to be expensive to be effective; it just needs to be smarter.
What's next for Bloompoint
We want to diversify the user's input, so that we can take in inputs like a pin drop location or a name of a body of water, as opposed to coordinates (what we have now). We'd also like to implement a heatmap to show the concentrations of algae pollution as a distribution.
Log in or sign up for Devpost to join the conversation.