bloomguard

Main Page

🧠 Inspiration

We are doing this because as new students in the McMaster community we were looking for off campus housing for next year and many landlords made a point of informing us whether or not the tap water was drinkable especially during algae blooming seasons. This piqued our curiosity and led us to conduct research upon this issue of algal bloom. Then, we were presented with the opportunity of Delta Hacks where we could put our research to use.

⚙️ What it does

Bloomguard is an early-warning and response dashboard for harmful algae blooms that turns satellite data into a concrete action plan. Bloomguard automatically analyzes a series of satellite images over Lake Erie, using a quantitative spectral index to detect elevated chlorophyll levels associated with early-stage algae growth. The system generates a map highlighting high-risk areas and condenses this information into a simple risk score, allowing operators to instantly understand bloom severity without interpreting raw satellite data. Once risk is detected, bloomguard predicts where the bloom will move next by calculating a trajectory cone based on marine wind data. This forecast shows the probable direction and distance the bloom will drift over the next 24 hours, creating a time window for intervention. An AI analyst then converts the technical outputs—risk level, location, and movement—into a short operational report explaining the situation and outlining a response strategy in plain language. If action is approved, bloomguard responds by generating a drone mission plan around the high-risk area. The system visualizes a flight path to help dispatch the drones, and completes the workflow from detection → prediction → response in a single interface. The result is not just awareness, but preparedness: bloomguard helps decision-makers act early, precisely, and deliberately—before an algae bloom becomes a public-health or ecosystem crisis.

🛠️ How we built it

We built this project using a React frontend and Flask backend. We used multiple APIs such as the Standard Weather API for wind data, and the Gemini API with the 2.0 Flash model for our tactical overview. We sped up the process using Cursor and Gemini models to guide us through the build, and set up a Git workflow to manage our edits and collaborate. Our team roles were split into two engineers and two researchers/designers. This way, we made progress on multiple aspects of the project simultaneously, and kept things easy to manage with only two engineers working together.

⚠️ Challenges we ran into

One challenge we ran into was when the development server kept calling the Mapbox API every time we moved the map ui, like an infinite loop. In the beginning, it drained a lot of our tokens until we managed to fix that bug. Hopefully it doesn't amount to a lot of money. Secondly, when planning how to integrate our third party softwares and APIs, we realized that using the satellite data API is too demanding for our computers, so we ended up using a dummy image to represent the spike in chlorophyll. Building the rest of our project around that change forced us to rethink certain parts of our plan, because we didnt expect to need to use dummy data or hardcode certain values. It took extra time for our team.

🏆 Accomplishments that were proud of

One of our key accomplishments was successfully integrating real-time environmental data into bloomguard’s prediction pipeline. By calling live wind and marine conditions, we were able to calculate a dynamic trajectory cone that shows where an algae bloom is likely to move next. We also explored and compared multiple algorithms for drift estimation, allowing us to balance accuracy with computational simplicity for a real-time system. In addition, we integrated Gemini to analyze processed data and generate clear, actionable reports, transforming complex satellite and environmental signals into understandable operational insights.

🔍 What we learned

We learned that prediction is just as critical as detection when managing harmful algae blooms. Detecting algae only shows where the problem is now (which is often too late) but forecasting where the bloom will drift using wind and surface conditions creates the real opportunity for timely action. The trajectory cone transforms a static detection map into a forward-looking decision tool. We also learned that prevention directly breaks the acidification cycle: Prevention breaks the acidification cycle. We learned that water acidification is often driven by the bacteria that eats the algae, not just atmospheric CO₂. By neutralizing blooms before they occur, our project helps prevent oxygen depletion, bacterial overgrowth, and the formation of acidic conditions in aquatic ecosystems.

🌱 What’s next for bloomguard

Next, we plan to move bloomguard closer to real-world deployment through strategic partnerships. This includes working with Phoslock to align our simulated response with proven clay-based mitigation methods, and partnering with existing agricultural and environmental drone providers rather than building hardware from scratch. We also aim to expand bloomguard beyond a single lake by scanning additional locations and adapting the system to different coastal and freshwater environments. Finally, we see strong potential in partnering with organizations such as Waterwise to support water stewardship initiatives and integrate bloomguard into broader water-quality management programs.