Skip to the content.

Kd490 → Secchi: the optical theory

This page explains the single relationship at the heart of the Submarius water-clarity model: the conversion from satellite-derived diffuse attenuation coefficient (Kd490) to Secchi disk depth (a proxy for “vertical visibility”).

What’s Kd490?

The diffuse attenuation coefficient at 490 nanometres measures how much downwelling blue-green light is absorbed and scattered per metre of water column. Higher Kd = more attenuation = less light getting through = lower visibility.

Units: m⁻¹ (per metre).

Typical values:

Water type Kd490 (m⁻¹) Visual character
Open ocean (clear) 0.02 – 0.05 Deep blue, 30 m+ visibility
Coastal blue water 0.05 – 0.15 Blue-green, 10–25 m visibility
Coastal turbid 0.15 – 0.5 Green-grey, 3–10 m visibility
Estuarine / river-plume 0.5 – 5+ Brown-green, < 3 m visibility

Satellites measure Kd490 by inverting the water-leaving radiance spectrum — essentially, by reading the colour of the water from above and modelling what attenuation would produce that colour.

Submarius pulls Kd490 from three independent sources (VIIRS-SNPP, NOAA’s VIIRS+OLCI multi-mission gap-filled product, GOES-16 ABI processed via ACOLITE) and fuses them via inverse-variance weighting in log-Kd space.

What’s Secchi depth?

The Secchi disk is a 30 cm white disk lowered into water until it disappears from view. The depth at which it disappears is the Secchi depth (Z_SD). It’s the oldest standardised measurement of water clarity and the one divers most intuitively recognise.

Roughly: Secchi depth in metres ≈ vertical visibility a diver experiences. Horizontal visibility is typically 1.5–2× Secchi depth in clear water.

The classical “1.7/Kd” rule

For decades the standard relationship was:

Z_SD ≈ 1.7 / Kd(490)

This worked tolerably well for moderate-clarity coastal water but broke at the extremes — too pessimistic for very clear blue water, too optimistic for very turbid coastal/estuarine water.

The Lee 2015 mechanistic model

Lee, Z. P. et al. (2015) Secchi disk depth: A new theory and mechanistic model for underwater visibility, Remote Sensing of Environment 169, 139–149.

Lee derived a new equation based on contrast detection by the human eye at the most transparent wavelength of the water type, rather than fixed 490 nm. The simplified form:

Z_SD ≈ 1 / Kd_tr × ln( r_t − r_w_pc × R_d_pc / N_c_pc)

where Kd_tr is Kd at the transparent-window wavelength (which shifts with water type — ~480 nm for clear blue ocean, ~550–580 nm for turbid coastal water).

The logarithmic ratio is empirically nearly constant across diverse water types: 2.38 ± 0.03.

So the practical simplification:

Z_SD ≈ 2.38 / Kd_tr

Validated against N = 338 globally distributed in-situ measurements spanning 1 m to 30 m+ visibility:

R² = 0.96, ~18% mean absolute error.

This is the strongest published physical relationship between satellite-observable optical properties and human-perceived underwater visibility. It’s the foundation Submarius builds on.

Wavelength shifts by water type

The “transparent window” — the wavelength at which water absorbs and scatters light least — shifts with water composition. From Lee 2015 Figure 4:

Z_SD Transparent wavelength Perceived water colour
5 m 540 nm Green-yellow
10 m 530 nm Green
20 m 510 nm Blue-green
40 m 480 nm Blue

Practical implication: Kd at 490 nm (what the satellite measures most directly) is most accurate for clear water. For coastal or turbid water, Kd(490) underestimates the relevant attenuation, and a naive 2.38/Kd(490) calculation will over-estimate visibility.

This is why Submarius uses an adaptive coefficient — Case-1 vs Case-2 — rather than a fixed value.

Case-1 vs Case-2 water

Oceanographers classify water into two broad optical types:

Most published consumer-facing visibility products use 2.38 everywhere and silently overestimate coastal turbidity by 30–50%.

Submarius computes a case-2 factor from:

…then blends coefficients:

coeff = (1 − case2_factor) × 2.38 + case2_factor × 1.5
Z_SD  = coeff / Kd

A Florida east-coast surf zone with Kd = 0.18 maps to Secchi ~8 m (realistic) instead of the ~12 m a Case-1-everywhere model would produce. The Bahamas deep with Kd = 0.04 maps correctly to ~60 m.

Where Kd490 falls short

The relationship has known limits. We don’t claim performance outside them:

Citations and further reading

The Submarius blog post how Kd490 satellite data predicts water clarity covers this material at lower technical density for a non-oceanographer audience.