Diffuse attenuation coefficient: why it caps satellite seagrass detection
If you've spent any time with ocean color data, you've run into Kd, usually written Kd(PAR) or Kd(490). It's the diffuse attenuation coefficient, and it's the single number that does the most to decide whether a satellite pass over your bay will show you a seagrass meadow or just show you water.
What Kd measures
Light entering the water column doesn't travel in a straight beam the way it does through air. It scatters off suspended sediment, gets absorbed by dissolved organic matter, and gets eaten by phytoplankton. Kd quantifies how fast that happens, in units of inverse meters. A Kd of 0.2 m⁻¹ means light intensity drops by a fixed fraction for every meter of depth, following an exponential decay curve. Small Kd, clear water, light gets deep. Large Kd, murky water, light dies out fast.
For anyone running a benthic survey, Kd is the direct driver of your effective mapping depth, sometimes called the Secchi depth or euphotic depth depending on which convention your agency uses. A sensor trying to pick up the reflectance of seagrass blades on the seafloor is working with whatever light made it down to the canopy and back up through the water column a second time. Double attenuation, same Kd. That's why benthic signal fades so much faster with depth than people expect from a single Secchi reading.
Why this caps satellite seagrass detection specifically
Seagrass mapping from a multispectral pass works because healthy meadows and bare sand or mud reflect light differently enough in the green and red bands that you can tell them apart. But that spectral contrast has to survive the round trip through the water column before it reaches the sensor. In a clear oligotrophic lagoon with Kd(PAR) around 0.1-0.15 m⁻¹, that signal can still be workable at 8-10 m. Push into a turbid estuary mouth with river input and resuspended fines, and Kd can climb past 0.5 m⁻¹, which collapses usable detection to a couple of meters before the water column simply dominates the pixel.
Turbidity sets a hard depth ceiling, and no amount of band math recovers signal that never reached the sensor in the first place. That's the part that catches planners off guard when they're scoping a blue-carbon baseline. A meadow sitting at 4 m in a glassy back-bay channel might be fully mappable. The same meadow at 4 m near a dredge spoil ground or just downcurrent of a creek mouth after a rain event might be invisible that week. The meadow hasn't changed; the water column has.
A few things follow from this in practice. First, tidal and seasonal timing matter more than people assume. Kd on a flood tide after a storm can be several times what it is on a calm neap tide in the dry season, so the same site can look completely different in two passes three weeks apart. Second, Kd is wavelength-dependent: shorter blue wavelengths usually penetrate deeper in clear water, while longer red wavelengths attenuate fast almost everywhere.
That's part of why benthic mapping tools lean on the green band for depth and on the red edge for separating plant material from substrate. Third, when you're writing up a baseline extent map for a habitat protection zone, water clarity at the time of capture is as much a methods note as your sensor resolution is. A reviewer asking why a known meadow edge didn't show up that year is almost always asking a water-clarity question.
None of this means deep or turbid water is off the table for a baseline program. It means the honest framing is extent and relative density in the zone where light reaches the bottom and comes back, not a guarantee of full-column penetration everywhere in the estuary. That's the premise behind an annual multispectral pass over your coastal waters that FlyPix turns into a seagrass extent and density layer, built around what the water column actually lets through rather than around a number that ignores it.
If you're scoping a pilot bay or estuary for a blue-carbon or habitat baseline, it's worth getting on the early-access list before you lock in survey dates.