The AzotoColumn is a bioretention system built for TSA's Engineering Design event, whose theme that year was "Managing the Nitrogen Cycle." Excess nitrogen from fertilizer runoff drives eutrophication and dead zones in nearby waterways, so our team designed a border-line filtration column — soil, sand, and gravel layered around nitrogen-fixing plants like clover — that a farm's runoff naturally flows into by gravity before a pump returns the cleaned water for irrigation. We named it the AzotoColumn after "Azoto," Greek for nitrogen.
We landed on the bioretention design after scoring it against two other concepts — an AI-driven soil-sensor network and selective catalytic reduction for vehicle exhaust — on cost, effectiveness, and maintenance, then spent most of the season iterating on a physical prototype: testing water flow and nitrogen levels, chasing down leaks, and tuning the soil layering until the numbers held up.
Nitrogen-rich fertilizer runoff is a major driver of eutrophication: it feeds algal blooms in nearby lakes and rivers that die off, consume dissolved oxygen, and leave "dead zones" where aquatic life can't survive. The same runoff can leach into groundwater and drinking supplies at levels linked to health risks like blue baby syndrome. Farms need a way to keep using nitrogen fertilizer without losing most of it to runoff.
- Had to be buildable and testable as a physical scale prototype within a school year, not just a paper design.
- Needed to hold water without leaking for accurate flow-rate and nitrogen-level testing.
- Had to stay affordable and installable on real farmland without disrupting existing operations.
- Needed to work across different soil types, rainfall levels, and farm layouts rather than one specific site.
Choosing bioretention over an AI sensor network or catalytic reduction
We scored our three candidate solutions — a bioretention filtration area, an AI-driven soil/moisture sensor network, and selective catalytic reduction for vehicle exhaust — in a Pugh chart against cost, effectiveness, accessibility, maintenance, and water resistance. Bioretention came out on top (+2) over the sensor network (0) and catalytic reduction (-2).
Reason:
The sensor network and catalytic reduction both required expensive, specialized hardware and ongoing maintenance, while bioretention reuses passive, low-cost materials (soil, sand, gravel, plants) that farmers can maintain themselves once installed.
Tradeoff:
Bioretention needs more physical space and upfront digging/layering work than a sensor-based system would, making it a harder fit for very small farms.
Switching to a pre-built container after chasing a 446 mL leak
Our first water reservoir was hand-cut plastic sheets sealed with hot glue and duct tape. During testing it lost 446 mL of water we couldn't account for. We isolated the cause by testing the mesh filter and tubing in controlled side-tests, ruling both out, before confirming the leak was coming from uneven, hand-cut seams in the container itself. Switching to a pre-built, slightly curved storage container cut the leak down to 50 mL, and a custom 3D-printed seal around the pump intake eliminated it entirely.
Reason:
Any leaked water directly corrupted our flow-rate and nitrogen-level test data, so a watertight reservoir wasn't a nice-to-have — it was a prerequisite for trusting any of our test results.
Tradeoff:
A pre-built container is less customizable to the prototype's exact internal shape than a hand-built one, which introduced its own fitting and stability issues we had to solve separately.
Concentrating perlite in the top third of the soil layer
After nitrogen testing showed elevated nitrate levels in the filtered water — a sign the soil wasn't retaining nitrogen well — we tested adding perlite (a lightweight, porous volcanic glass) at different soil depths. Placing it only in the top third of the layer improved ammonium infiltration into the soil while a perlite-free bottom layer kept porosity low there, limiting how much nitrogen could leach through into the collected water.
Reason:
Testing multiple placements directly, rather than assuming "more perlite is better," revealed a real depth-dependent tradeoff — top-heavy placement helped infiltration without also opening up leaching paths lower down.
Tradeoff:
This layering is more finicky to construct consistently than a uniform soil mix, and depends on getting the soil composition right in the first place.
- Filtration dropped nitrate from 25 ppm to 5 ppm and ammonium from 15 ppm to 3 ppm across our water samples, with pH staying stable through the process.
- Reduced prototype volume by 75% and cut reservoir leakage from 446 mL to effectively zero after redesigning the water container and pump seal.
- Estimated a full one-acre installation (excavation, lining, filtration media, pump, sensors, grading) at roughly $17,600 — expensive upfront, but offset over time by reduced fertilizer and irrigation costs.
- Placed 4th at the WTSA State Conference and went on to present the project at TSA Nationals.
- A structured comparison (Pugh chart) across cost, effectiveness, and maintenance made the case for a "low-tech" bioretention design over flashier AI-sensor or catalytic-reduction ideas.
- Physical prototypes surface failure modes a diagram never will — our worst leak came from hand-cut plastic seams we hadn't considered a risk until the data didn't add up.
- Iterating on where a material goes (perlite in the top third, not mixed throughout) can matter as much as which material is chosen at all.