Satellites measure the ocean’s heat. But to know what heat did to a coral colony, someone still has to look at the coral. In Hong Kong, that looking has a remarkable, decades-long record — and ReefBeacon is built to respect both its value and its limits.
A twenty-year photographic memory
Every summer since 1997, volunteer divers participating in the coral survey program have documented reefs across 80+ survey sites, year after year. Two decades of survey photos form a rare long-term baseline: not a model’s guess about the past, but the past itself, frame by frame. That archive is exactly the kind of primary source our organization exists to work with.
Three grades and a confidence level
When ReefBeacon reads a survey photo, it does not output a single verdict. It grades bleaching severity on three levels — mild, moderate, severe — and attaches a confidence level to every judgement. When confidence drops, the case is flagged for human review rather than silently passed through. The point is not to replace the diver’s judgement; it is to make thousands of photos searchable and comparable while keeping a person in the loop exactly where the machine is unsure.
Why photos and satellites need each other
The two records answer different questions. Daily satellite data tells you when heat stress arrived and how long it stayed; survey photos tell you what the corals actually did in response. Read together, they turn “the water was hot in August” into “these sites bleached moderately, those recovered, these did not” — the knowledge conservation teams plan around. Read separately, each is half the story.
Honest about what a photo can say
A photo is a moment, not a diagnosis. Bleaching and recovery can move quickly; visibility, angle and lighting all affect what an image shows. ReefBeacon’s photo judgements therefore always carry their confidence level and never overwrite field observation — and heat-stress alerts remain what they are: warnings to go and look, not statements about what has already been lost.