Dead Pixel: Closing the Accountability Gap in Coastal Light Pollution

SMathHacks 2026 | Data Science Track

Tagline: You can't fine a pixel.

Originality

Most light pollution research stops at measuring the problem. Dead Pixel asks the next question: who is responsible?

Artificial Light at Night (ALAN) bleaches coral. Ayalon et al. (2019, Global Change Biology) demonstrated that corals exposed to ALAN undergo oxidative stress and photosynthetic impairment, with blue and white LED spectra causing the most severe damage (Ayalon et al. 2021, Frontiers in Physiology). In Hawaii, coral reefs generate approximately $800 million annually in marine tourism revenue (NOAA Fisheries), and Lin et al. (2023, Nature Sustainability) showed that tourists are drawn specifically to areas with high live coral cover. Hawaii's nearshore reefs contribute $364 million in annual added value (Davidson, Hamnet, & Minato, 2003).

The science is established. The economic stakes are clear. But current satellite monitoring operates at 500-meter resolution, and our analysis found that a single pixel in Waikiki contains 171 businesses. The satellite produces one brightness number for all 171. Individual attribution is impossible. We call this the Accountability Gap, and it is the fresh angle Dead Pixel takes: not "is light pollution bad?" but "why can't anyone be held accountable for it, and what would it take to change that?"

Our approach is unique in three ways:

1. Quantifying the Gap: No prior work has counted the businesses inside a single satellite pixel to demonstrate why enforcement fails.
2. Calculus-Driven Analysis: We applied a slope field concept from calculus coursework to 12 years of radiance data, treating each pixel's history as a derivative to show where the problem is accelerating.
3. The Hardware Bridge: We designed a $2.88 sensor that closes the gap the data science exposes, allowing us to turn analysis into action.

Impact

Dead Pixel has direct real-world implications for environmental enforcement, economic protection, and municipal policy. We developed three layers of original data science:

• Layer 1 — Business Density Per Pixel: We mapped over 5,000 businesses along Oahu's coastline and overlaid them onto a 500m × 500m pixel grid matching satellite resolution. Of over 2,400 coastal pixels, 328 contain at least one business. The densest pixel contains 171, proving quantitatively that satellite data cannot support individual enforcement.

• Layer 2 — Radiance Trend Analysis: We computed the year-over-year radiance slope for each coastal pixel across 12 years of NASA satellite data. The magnitude is asymmetric: the fastest-brightening pixel is increasing at more than double the rate of the fastest-dimming pixel. The largest increases cluster directly over the Honolulu–Waikiki corridor, on top of the densest reef zones.

• Layer 3 — Reef Proximity Risk Map: We integrated NOAA reef shapefiles and calculated the distance from each pixel to the nearest reef. Our highest-risk pixel contains 171 businesses just 290 meters from live coral. A second notable pixel contains 88 businesses only 3.7 meters from the reef.

The Hardware Bridge: The LuxBox is a $2.88 reef-level light sensor deployed in a wireless mesh network at 30-to-50-meter intervals. If two adjacent sensors detect a blue-spectrum spike but their neighbors are dark, enforcement agencies can localize the source to a specific 30-meter property line. 75 sensors cover all of Waikiki for approximately $250. For the first time, a lawmaker could attach a GPS coordinate to a fine.

Execution

• Datasets: 12 years of NASA VIIRS annual nighttime radiance composites, NOAA benthic habitat shapefiles for Oahu's coral reefs, and coastal business data from OpenStreetMap via Overpass Turbo.
• Tools: Python, Berkeley datascience library, GeoPandas, h5py, NumPy, Matplotlib, Shapely, JupyterLab.
• Methodology: Layer 1 counts businesses per pixel using spatial containment. Layer 2 fits a linear regression to each pixel's 12-year radiance history to compute a slope. Layer 3 normalizes business count and reef proximity into a composite risk score: Risk Score = (Normalized Business Count + Normalized Reef Proximity) / 2.

Hardware Bill of Materials (BOM)

Component	Cost/Unit	Justification
Microcontroller	$0.55	Wholesale pricing via Alibaba
Wireless Radio	$0.55	Long-range mesh networking
RGB Color Sensor	$0.99	Blue-spectrum detection
HDPE Housing	$0.26	Zero-leaching, salt-spray resistant
Solar Panel	$0.20	Autonomous power
Polycarbonate Cap	$0.19	Optical transparency
Charging Port	$0.07	Configuration and backup
Battery	$0.05	Nighttime operation
Photoresistors	$0.01	Ambient light sensing
Wiring	$0.01	Internal connections
Total	$2.88

Hackathon Prototype vs. Ideal Device

We built this prototype to prove one thing: we can catch the "blurry" light pollution that satellites miss. While our current unit is a "lab version" optimized for a demo, every choice was a stepping stone toward a massive, real-world deployment.

Sensing the Stressor: Right now, the prototype uses a standard photoresistor. It’s great for a live demo because it reacts instantly to a flashlight, but it’s a blunt instrument. Based on Ayalon et al. 2019, 2025, we recognized that corals don't just react to "light"—they specifically suffer from high-energy blue-spectrum light (450nm-490nm). Our final model replaces this with a TCS34725 RGB sensor—a specialized eye that can isolate that specific "blue spike" from hotel floodlights.

The Bare-Metal Pivot: For this presentation, we’re powered by a USB-C cable for 100% reliability. However, a reef doesn't have a computer outlet. Our production model uses an Arduino REV 3 architecture, but the prototype uses an Arduino Circle MKR IoT as it easily wires with our photoresistor. By stripping away power-hungry LEDs and voltage regulators, we drop the idle current into the micro-amp range, allowing the final "LuxBox" to run indefinitely on a tiny solar panel and supercapacitor.

From WiFi to the Swarm: A single sensor is just a data point; a "swarm" is a map. The prototype is a standalone unit, but our deployment model utilizes a LoRa (Long Range) Mesh Network. Unlike WiFi, which fails in high-humidity coastal air, LoRa provides the "muscle" to slice through salt spray. This mesh network allows our nodes to share data and triangulate a light source down to a 30-meter property line.

Sustainable Shells: We 3D-printed this prototype in PLA because it’s fast for iteration, but it isn't marine-stable. Our production units are engineered from High-Density Polyethylene (HDPE). It is UV-stable, 100% recyclable, and zero-leaching, ensuring our hardware won't break down and become the very microplastic pollution we’re trying to prevent.

Salient Takeaways

• The Problem: Hawaii's $800 million reef tourism economy is being stressed by artificial light, and satellite data can't tell us who's responsible.
• The Finding: 171 businesses inside a single 500-meter satellite pixel in Waikiki, sharing one brightness reading.
• The Line: You can't fine a pixel.
• The Solution: $250 covers all of Waikiki with 75 ground-level sensors to identify specific polluters.
• Entrepreneurial Viability: Dark Sky enforcement generates fine revenue that funds further reef restoration—the system pays for itself.

References

• Ayalon, I., et al. (2019). "Red Sea corals under ALAN undergo oxidative stress and photosynthetic impairment." Global Change Biology.
• Ayalon, I., et al. (2021). "The Endosymbiotic Coral Algae Symbiodiniaceae Are Sensitive to ALAN." Frontiers in Physiology.
• Lin, B., et al. (2023). "Coral reefs and coastal tourism in Hawaii." Nature Sustainability.
• NOAA Fisheries. "Coral Reefs in the Pacific."
• Davidson, Hamnet, & Minato (2003). Hawaii Coral Reef Initiative Research Program economic valuation.

Built With

• arduino
• github
• json
• python