### BITB38 — South Brother
**CRD page:** [Three Brothers/Central Chagos Coral Surveys BITB38-BITB49 (Mar 2015)](https://coralreefdiagnostics.com/three-brotherscentral-chagos-20150321-to-0324)
**Date:** 2015-03-21
**GPS:** 6.16828°S, 71.54042°E
**Frames reviewed:** 44 of 44 (100% of `BiTb38-great site-finished`; all present, non-zero and readable). All 44 occur in the supplement's image table, but only 34 survive in the paired point/mask audit; ten consecutive files, P3210426–P3210437, are lost from that join through same-basename collisions. Those ten include the lobster den and several baitfish/reef frames.
**Coral cover:** 56.03% (formal transect surveys; **algal cover 33.78%**) · **published % bleached lookup:** 7.59% (`% bleaching (2nd training)`, Mayfield & Dempsey 2026 supplement). The 7.59% value is **not corroborated as a site-scale March bleaching event** by Anderson's field record or the full 44-frame review; see below.
**Reef type:** protected, emergent **lagoonal fringing reef**, inside the lagoon · **Temperature:** 29.7 °C · **Salinity:** 34.4 · **Survey effort:** 20 transects across four depth strata—D1 n=2, D2 n=6, D3 n=8, D4 n=4.
**Live hard coral cover across methods:** formal transects **56.03%**; the supplement's 44-image estimate **33.73%**; paired CoralNet points 27.3% and Coralscapes 29.3% on the 34 frames that survived the filename join. The paired deviations are −28.7 and −26.7 percentage points, but they do not demonstrate that imagery intrinsically under-read this reef: the denominators differ, 30.2% of paired points are `Unknown`, ten annotated frames are absent from the paired join, and the edited photographs are not documented as coincident with the formal transects.
**Formal benthic composition:** *Echinopora* 13.77%, *Acropora* 13.54%, *Heliopora* 6.65%, *Turbinaria* 6.05%, *Favia* 2.99%, *Porites* 2.39%, *Gardineroseris* 1.94%, *Pocillopora* 1.64%, *Stylophora* 1.37%, *Pavona* 1.18%, *Montipora* 1.12%, then 18 additional genera at <0.8% each. Algae is led by turf 17.13%, CCA 9.30% and sediment-laden turf 4.92%, with macroalgae 1.82% and erect algae 0.61%. Barren substrate is 7.35%; non-coral invertebrate cover is 2.84%.
**Representative photos:** P3210408.jpg, P3210417.jpg, P3210485.jpg, P3210402.jpg, P3210419.jpg, P3210435.jpg, P3210454.jpg

**BITB38 is a plate-coral reef, but not the tabular-Acropora monoculture described in the sparse pass.** Some frames do carry broad, horizontal tables compatible with *Acropora*, especially P3210478–P3210488, and P3210485 is the signature wide view of a metre-scale table canopy over massive and branching colonies. The earlier part of the dive is different. **P3210408 and P3210417** show hundreds of thin, upright-to-scroll plates packed into rosettes and overlapping shingles. Their fine corallite texture, strongly curled margins and the formal transect composition make them compatible chiefly with **plate-forming *Echinopora* and *Turbinaria***, not safely identifiable as tabular *Acropora* from the photographs alone. Other frames add branching *Acropora*-type thickets, blue-grey branching colonies compatible with *Heliopora*, massive and lobate *Porites*-type colonies, corymbose pocilloporids and a long tail of faviid, agariciid and other forms. The correct visual description is a **high-cover, taxonomically mixed plate-and-branching mosaic**.

**The 56.03% site mean hides one of the strongest depth turnovers in the Chagos set.** D1 averages 45.6% coral and is *Acropora*-led at **34.4%**. D2 is the low-cover, algae-dominated interval: **34.3% coral versus 57.27% algae**, with *Heliopora* at **17.5%**, turf at 36.43% and *Acropora* only 2.43%. D3 rises to **69.33% coral**, split among *Echinopora* 13.13%, *Acropora* 12.68%, *Heliopora* 8.6% and *Turbinaria* 4.55%. D4 reaches **74.9% coral** and is a different reef again—***Echinopora* 41.65% and *Turbinaria* 19.6%**, with *Acropora* only 4.65% and algae 12.25%. Two D4 transects are extraordinary: 88% and 94% coral, composed almost entirely of *Echinopora* (88% and 78%, with 16% *Acropora* in the latter). The photographic plate fields are therefore not merely compatible with the formal record; they are the visible expression of a deep Echinopora/Turbinaria-dominated stratum.

**Replication is sufficient to treat the turnover as real, but the reef is patchy within strata.** There are 20 transects, not a single low-effort snapshot. Yet individual D2 transects range from 8% to 60.8% coral, and D3 from 32.6% to 92%. Two D4 transects are ≥88% coral while the other two are 57–60%. A site-level sentence such as “near-continuous plates with almost no bare substrate” is true for P3210408/P3210417 and the highest-cover formal transects, but not for the whole site. **P3210476** is the counter-frame: camera equipment rests on pale rubble and sand among widely spaced coral heads. P3210475 and P3210477 are similarly open. The single 56% mean legitimately describes the total survey, while neither the dense plate fields nor the open rubble plain should be corrected away.

**The coral-form point and mask classes are too coarse to resolve this formal assemblage.** On the 34 paired frames, the point set attributes 54.7% of live coral points to *Acropora*, 20.9% to *Porites*, 1.4% to *Pocillopora* and 23% to “other hard coral.” Coralscapes makes “other hard coral” its largest live form at 40.5%, ahead of tabular *Acropora* at 18.9%, massive/meandroid forms at 16%, non-tabular *Acropora* at 12.8% and *Pocillopora* at 6.3%. The formal data explain the disagreement: *Echinopora*, *Heliopora* and *Turbinaria* together make up **26.47 percentage points—47% of all formal coral cover**—but none has a dedicated point or mask class. Much of the image-model “other hard coral,” and probably some apparent “tabular Acropora,” is the site's defining non-Acropora plate fauna.

**The animal record is exceptional even though ten of its best frames are absent from the paired audit.** **P3210419** is a close lionfish compatible with *Pterois*, fins extended against a red sponge- and CCA-covered crevice; species is not asserted. **P3210434 and P3210435** show a den containing at least six large spiny lobsters with striped legs, long pale antennae and patterned carapaces, compatible with painted spiny lobster (*Panulirus versicolor*) but retained as a morphological compatibility. P3210430–P3210432 establish the overhang and approach to the den. The water column repeatedly becomes a moving wall of small silversides or baitfish: P3210427–P3210429, P3210432, P3210437, P3210445, P3210449–P3210450 and **P3210453–P3210454**. **P3210448 and P3210456–P3210457** add close schools of silver trevallies over the reef. These are conspicuous-fauna observations, not abundance estimates.

**No crown-of-thorns starfish, disease lesion or clear corallivore predation scar is visible in any of the 44 frames.** The 3.89% mean dead-coral mask and 5.06% rubble estimate are consistent with old exposed framework and the open rubble/sand patches, not an acute mortality front. Pale plate rims recur throughout, but their regularity around otherwise pigmented living plates is compatible with active marginal growth and image lighting.

**BLEACHING — the published 7.59% lookup is not supported as a reef-wide March observation.** Anderson's field record is no bleaching on March or April dives, with bleaching first observed in May. Across all 44 photographs, the overwhelming plate, branching and massive assemblage is pigmented brown, olive, green, blue-grey, cream and mauve. The point audit assigns 15 of 1,020 paired points (1.47% of all points) to bleached coral across ten frames, while the mask estimates only **0.099% bleached area**. Many point calls are demonstrably off-target: P3210419 is a lionfish macro; P3210445 is mostly baitfish, diver and open water; P3210448 and P3210456 are trevally frames; P3210476 is camera equipment on rubble. Coralscapes assigns zero or near-zero bleaching to most of them.

Two frames deserve a narrower caveat. **P3210402** contains several pale or partly white branching colonies amid a dense, otherwise pigmented assemblage; recent branch mortality, local paling, natural pale colour and flash cannot be separated confidently. **P3210487** is a close small branching colony with cream-white distal branches over darker brown inner tissue; three bleached points and 1.59% mask bleaching agree that it is pale, but growth tips, partial bleaching and recent tissue loss remain competing explanations from one image. These are individual/local cases, not evidence that 7.59% of coral across this reef was bleached. The old sentence calling bleaching “moderate, consistent with 7.6%” has therefore been removed.

**Coverage and provenance are central to the method discrepancy.** The supplement contains all 44 BITB38 image rows and reports 33.73% coral cover, whereas the paired audit retains only 34 frames. The missing run P3210426–P3210437 is not a random ten-frame slice: it includes a coral macro, three baitfish/wall frames, the approach to the overhang and the complete lobster-den sequence. Their correct paired pixel-level outputs are unavailable because of basename collisions, so merely changing a site key in the joined table would not reconstruct the analysis. Within the retained set, about **308 of 1,020 points (30.2%) are `Unknown`**, and the Chagos point set is predominantly unconfirmed CoralNet robot output. This is an image-analysis provenance problem, not a human annotator's error.

<b>Assemblage composition.</b> Twenty formal transects give the preferred site estimate: 56.03% coral, 33.78% algae, 7.35% barren substrate and 2.84% non-coral invertebrates. The assemblage is led almost equally by <i>Echinopora</i> (13.77%) and <i>Acropora</i> (13.54%), followed by <i>Heliopora</i> (6.65%) and <i>Turbinaria</i> (6.05%), with a pronounced D1-to-D4 transition from Acropora through Heliopora to Echinopora/Turbinaria dominance. Across the 34 paired images, CoralNet points give 27.25% live hard coral and Coralscapes 29.33%; those products omit ten annotated folder frames and cannot resolve the formal plate-coral genera. The photographs strongly corroborate the reef's high structural complexity and plate dominance, but the formal transects—not the paired image subset—support the site-level cover and genus values.

**Flags:** (1) The sparse pass misidentified the reef as a near-continuous tabular-Acropora field. Formal transects show *Echinopora* 13.77% > *Acropora* 13.54%, with *Heliopora* 6.65% and *Turbinaria* 6.05%; D4 is 41.65% *Echinopora* plus 19.6% *Turbinaria*. (2) The published 7.59% bleaching lookup is not visually corroborated at site scale; P3210402 and P3210487 are ambiguous local cases, while multiple point calls sit on lionfish, baitfish, trevally, gear and rubble frames. (3) Ten consecutive images P3210426–P3210437 are absent from the paired audit through filename collisions, including the lobster-den sequence. (4) The reef ranges from 8% to 94% coral among transects; dense plate fields and open rubble plains are both real. (5) No COTS, disease or clear predation scars are visible. (6) No per-frame depth data are available.

**Benthic data:** [Download BITB38 benthic survey data (CSV)](../benthic_data/BITB38_benthic.csv) — scleractinian cover by genus with full names, plus substrate and algal categories (bare substrate, CCA, turf, macroalgae, invertebrates), from the LOF BIOT-1 survey (Mayfield & Dempsey 2026, *Oceans*, supplemental file S1).