The spectral absorption feature behind floating plastic detection

Why there's no single wavelength, just a dip water doesn't have

Floating plastic doesn't have one magic color that lights up on a satellite image. What it has is a localized dip in reflectance, an absorption feature, tied to the carbon-hydrogen bonds in the polymer itself. Every plastic resin, whether it's polyethylene from a fishing float or polypropylene from a bottle cap, has C-H bonds that absorb light at specific points in the near-infrared and shortwave-infrared. Water doesn't have that bond structure, so it doesn't absorb there the same way. That contrast, plastic pulling down reflectance where water stays relatively flat or behaves predictably, is the whole signal.

In hyperspectral data this shows up as a clean, narrow absorption feature around 1215 nm. Most research teams don't have hyperspectral imagery on a useful revisit cycle, though. What they have is multispectral, with a handful of broad bands instead of hundreds of narrow ones. The 1215 nm feature falls between standard bands on sensors like Sentinel-2, so you can't read it directly. What you can do is use the near-infrared band as a proxy, because the broader absorption behavior around it still shifts how plastic reflects light compared to open water, sargassum, foam, or sun glint.

Why the near-infrared band carries the signal

This is where the near-infrared plastic signature actually does its work. Open water absorbs strongly in the NIR, so a clean water pixel reads dark there. Floating plastic, because of that underlying C-H absorption structure, reflects NIR light differently than the water beneath or around it. Researchers have built on this with combined-band approaches, most notably the Floating Debris Index, which pairs NIR against a baseline interpolated from red and shortwave-infrared bands to flag pixels that sit outside the normal water curve. This flags a spectral anomaly on the water surface consistent with floating debris, worth a closer look rather than a lab-grade material ID.

A real-world reflectance curve floating debris produces rarely matches a textbook plastic reflectance curve from a lab sample. Wet weathered plastic tangled with biofouling and trapped water produces a messier signal than a dry, clean polymer swatch on a bench. That anomaly is a population-level signal, visible across many pixels aggregating along convergence lines where windrows collect.

What this means if you're using band math to scope a target

If you're building or reading spectral band plastic detection outputs yourself, a few things are worth keeping in mind:

• The NIR band alone won't separate plastic from sun glint, foam, or certain algae blooms reliably. Index approaches that baseline against red and SWIR exist specifically to cut down on those false positives, not eliminate them.
• Windrows aggregate along fronts, convergence zones, and current lines, so a spectral anomaly that lines up with known oceanographic features is a stronger lead than one that doesn't.
• Multispectral sensors trade spectral resolution for swath width and revisit frequency. That's the right tradeoff for scoping a search area over open water. Confirming material type from orbit asks more of the sensor than it can give; that confirmation step still happens from a vessel or a drone.

Ground truth still requires a vessel or drone on site. A spectral read like this narrows down where a team should look before committing fuel, crew time, and vessel days to a search pattern. An event-driven read built on this kind of anomaly gives a planner a flagged aggregation worth checking, a starting point for the search rather than a finished answer.

If your team is trying to turn satellite spectral signatures into an actual search decision rather than a research exercise, Plastic Detection turns this kind of multispectral anomaly into an early read on where debris windrows are aggregating, so you can scope a target area before a vessel ever leaves the dock.