Global search · setting & hydrologyAugust 2026

Are there other Rebel Reefs?

Tela Bay should not work. A turbid, river-fed, nutrient-loaded embayment is where coral reefs are supposed to die, and instead it holds three of the highest-cover reefs in the entire Mesoamerican system. If a place built like this can carry 68 to 84% live coral, the obvious question is whether there are others, and whether anyone has looked.

So we described Tela's physical situation in numbers, deliberately ignoring anything about its ecology or its health, and searched every shallow tropical coast on Earth for the same build. Then we asked, for each match, whether there is a reef there and whether anyone has studied it.

In plain terms

Think of it as a fingerprint of the setting, not the reef. How enclosed the bay is, how steep and shallow the seabed, how narrow the shelf, how murky the water, how big the river, how flashy its flood season, what rock its catchment is made of. Nothing about coral at all. Then we look for that same fingerprint elsewhere, and only afterwards ask what is living there.

Every marker is a place the search compared against Tela. Colour shows what we concluded. Tap any marker for its numbers. Note that the three verified sites were confirmed not by satellites but by reading the local-language scientific literature.

Setting features from ETOPO1 bathymetry and NASA MODIS-Aqua ocean colour. Hydrology from RiverATLAS v10 (HydroATLAS). Heat from NOAA Coral Reef Watch. Reef records from GBIF, normalised by total occurrence records. Coral cover at Tela from the Healthy Reefs Initiative 2024 survey.

Tela's fingerprint

Against every other shallow tropical coastal cell on Earth, Tela is unusual in a specific combination. Each number below is how many standard deviations Tela sits from the global average.

The physical signature of Tela Bay. Structural features only. Nothing here describes the reef itself.
FeatureTela vs the worldMeaning
Enclosure (land within 10 / 25 / 50 km)+1.51 / +1.40 / +1.47Strongly embayed, a pocket rather than open coast
Local seabed relief+1.47Banks and drop-offs, not a flat shelf
Chlorophyll+0.82Productive, nutrient-fed water
Sediment exposure (Kd490)+0.64Murky
Shelf width / distance offshore−0.90 / −0.90Narrow shelf, deep water close in
River (Ulúa): catchment, flow, flood swing, karst21,882 km², 443 m³/s, 9.9×, 14%A large, seasonally flashy river on soft rock

Does the search work? The controls say yes

A search like this is only worth reading if it recovers places we already know are built like Tela, and correctly rejects places that are not. It does, and on purely physical grounds:

Control test. Known turbid, river-influenced systems versus known open clear-water systems, scored blind to their ecology.
PlaceSetting rankVerdict
Bocas del Toro, Panamatop 0.22%✓ turbid, river-fed, embayed
Varadero, Colombia (beside the Magdalena)top 0.82%✓ our closest published analogue
Faro Cazones, Cubatop 1.7%✓ but see the karst note below
Utila, Honduras19.7%✓ correctly rejected, clear water
Bonaire37.4%✓ correctly rejected, no rivers at all
Cordelia Bank, Honduras43.6%✓ correctly rejected

The last one matters most. Cordelia Bank is a genuine bright spot, roughly 54% coral cover, and the search puts it far from Tela. That is correct and it is the point: Cordelia is a healthy reef in clear water. It is not a physical replica. A search that ranked it highly would be tracking reef health, which is not what this measures.

Adding real river chemistry sharpened things further. Faro Cazones and both Cuban candidates sit on catchments that are 89 to 100% karst, against Tela's 14%. They are topological twins fed by chemically different water: karst catchments deliver high-alkalinity flow. Without the hydrology layer they looked like near-perfect matches.

What the search found

Ninety-one distinct systems survived the cut. Forty were profiled for hydrology, twelve checked for evidence of reef, and three investigated properly. Those three all turned out to hold coral.

The three verified matches. All confirmed through local-language literature, not satellites.
PlaceCoral cover reportedWhy it matters
Toliara, southwest Madagascar46.8 ± 3.4%The strongest parallel found anywhere. A monitored site sits 8 km from the mouth of the Onilahy, and the river's influence on the reef is already being studied there. There is a resident marine institute working the exact question we are asking at Tela.
Bengkulu, west Sumatra49.3%, 57.2%, 68.4%Three villages surveyed in Indonesian-language journals, 19 to 26 coral species. 68.4% is Cocalito-class. International databases held zero coral records for this coast.
Tana Delta, Kenya"rich coral reef patches"Reefs under seasonal freshwater plumes, with the Tana's sediment load swinging from 2,796 to 24,322 tonnes per day, a factor of 8.7× against the Ulúa's 9.9× flood swing.
The finding, and it is not the one we expected

None of these reefs is undiscovered. They are known locally and invisible internationally. The gap is language and data-sharing, not exploration. Every one of them has published science, and in Madagascar's case an institute studying river-influenced reefs, none of which surfaces in an English-language search or in the global occurrence databases.

That is more useful than a discovery would have been. It means there are existing datasets and existing research groups for the comparison Tela needs, and the next step is a set of emails rather than an expedition.

Two methods that failed, kept on the record

Both failures point at the same wall, and the wall is the interesting part.

Discarded approaches. Each was tested against Tela as a known positive before being trusted, and each failed that test.
MethodWhat happened
Global reef habitat mapping (Allen Coral Atlas)It reports no reef at Cocalito. Essentially no data coverage there at all, for a reef holding 68% live coral. Satellite habitat mapping needs water clear enough to see the bottom, so it is structurally blind in turbid water. We had briefly used its silence as evidence of absence. That was wrong and is withdrawn.
Sentinel-2 bottom signal and textureDesigned to survive turbidity by looking for patchy shallow bottom. Both muddy control sites beat both known reefs. The Ulúa river mouth returned the strongest bottom signal in the whole test, and muddy inshore water was three times more "textured" than Cocalito. Sediment plumes produce exactly the signal reef structure would.

Turbid-water reefs cannot be found from optical satellites. That is not a complaint about our implementation, it is the reason this class of reef goes unrecorded, and it is why the only method that worked was reading what local scientists had already written.

What this search does not cover

Reported plainly, because the honest answer to "did you find them all" is no.

Coverage limits of the global search. What was examined, and what was not.
LimitDetail
The verification is a thin tip~99,000 candidate cells → 91 systems → 40 with hydrology → 12 with reef evidence → 3 investigated
The cutoff is too tight, demonstrablyAn earlier pass put Varadero outside the shortlist despite it being a real, published analogue. Genuine matches are being missed between the cutoff and roughly the top 2%.
Resolution excludes small systemsThe grid is ~10 km. Tela's western pocket is about one cell. Any cove-scale reef smaller than that is invisible to this search, which is a real limitation given that the differences that matter at Tela occur over a kilometre.
Latitude and temperature cutsNothing outside 35°N-35°S was screened, and a warm-water floor removed 13,000 shallow cells, some of which are marginal reef environments that matter for climate refugia.
One targetThe whole search is anchored on Cocalito. Banco Capiro or Punta Sal would each return a different list.

The next cheap improvement is loosening the cutoff and running hydrology on roughly seven hundred systems instead of ninety. The expensive one, and the one that would matter most, is repeating the whole search at finer resolution, because that is the only way to see reefs the size of Cocalito.

Method, code and the full candidate list are described on Methods & Data. If you work on any of the coasts named here, we would like to hear from you, see Get in touch.