What Is Diving Visibility? From Underwater Optics to How It's Measured

What Does "Visibility: 15m" Actually Mean?

Dive logs and shop blogs publish "visibility: 15m" every day — but this is not a standardized measurement. It's a diver's subjective estimate of horizontal visual range: the maximum distance at which a white target (another diver, a reef, a rope) can be distinguished. Note: this is horizontal, not vertical.

Scientific Measurement: The Secchi Disc

Oceanographers use a Secchi disc — a 30cm black-and-white disk lowered vertically until it disappears from view. This measures vertical transparency, distinct from a diver's horizontal visibility estimate.

The relationship is roughly: horizontal visibility ≈ Secchi depth × 1.5–2.5, though sun angle and conditions make exact conversion difficult.

Satellite Measurement: kd490 (Diffuse Attenuation Coefficient)

Our AI model uses kd490 — a value NASA computes daily from MODIS satellite data. It measures how rapidly light at 490nm (blue-green) attenuates per meter of water depth.

kd490 integrates chlorophyll (phytoplankton), suspended particles, and CDOM (yellow organic matter). More plankton = more scattering = higher kd490 = lower visibility.

Japan's Unique Dive Log Culture

Japanese dive shops quantifying visibility in meters every day is globally rare. Western shops typically report "conditions: good" — qualitative, not quantitative. Japan's industry standard of logging "visibility: 15m, temp: 19°C" daily has created the 46,000-observation dataset powering this site's AI. This density of dive visibility data is essentially impossible to replicate elsewhere.

Why the Same "10m" Can Look Different

• Sun angle: Midday light travels more directly, improving visibility
• Depth: Shallow water (<5m) appears clearer; deep (>30m) goes dark
• Water color: Green (high chlorophyll) vs blue (clear) feel different at same distance
• Particle type: Plankton bloom vs sand suspension scatter light differently

How AI Predicts Visibility

Our LightGBM model combines the following to predict 1–7 day visibility:

• Satellite data (kd490, chlorophyll: NOAA ERDDAP)
• Weather (wind, rain, pressure: Open-Meteo)
• Marine conditions (wave height/period/swell: Open-Meteo Marine)
• Seasonality (month/day sin/cos transforms)
• 3–7 day lag features

Best accuracy: IOP R²=0.824. Hardest site: Yonaguni R²=0.05 — its open-ocean, high-energy environment makes visibility nearly unpredictable from regional satellite data.

Summary

• Diver-reported visibility = subjective horizontal range estimate
• Oceanographic visibility = Secchi depth (vertical)
• Satellite measurement = kd490 (light attenuation coefficient)
• Japan's daily numerical logging is a unique global phenomenon
• 46,000 data points from this culture enable AI prediction otherwise impossible