The complete Convergent Shoreline Node: optical fluorescence for cyanobacteria, lab-grade water chemistry, particulate air quality, and a triggered passive-sampler relay — with on-node machine-learning inference deciding when a reading matters.
No dev-board sensor does, at any price. AquaMesh measures surrogates — phycocyanin and chlorophyll-a fluorescence, turbidity, conductivity, pH, fine particulate — and uses the established USGS surrogate-monitoring paradigm to decide when conditions warrant gold-standard lab sampling.
When a proxy threshold trips, the node fires a relay that activates a passive sampler (POCIS for PFAS, an aerosol cassette for cyanotoxins) for later lab analysis. The innovation is intelligent timing of expensive chemistry, not a miracle sensor.
| Layer | Hardware | Target |
|---|---|---|
| Optical fluorescence | 617 nm + 470 nm excitation, Semrock bandpass, photodiodes, 24-bit ADC | phycocyanin (cyanobacteria) & chlorophyll-a |
| Water chemistry | Atlas EZO-pH, EZO-EC, turbidity, DS18B20 | storm-event & hydrochemical context |
| Air quality | Sensirion SPS30, Bosch BME688 | aerosolized cyanotoxin context |
| Triggered output | MOSFET-gated relay | activates POCIS / aerosol cassette |
A TensorFlow Lite Micro model on the ESP32-S3 / nRF class core classifies bloom likelihood locally, so the node transmits decisions rather than raw spectra — a genuine differentiator for the cyanobacteria use case.
Optical cross-talk under high turbidity, biofouling of wetted optics, and surrogate–target correlations that may fall below a usable R² are all real and unresolved. They belong in the research plan, not hidden from it. Validation against benchtop fluorometry and co-located reference stations is required before any performance claim.