### BITB44 — Three Brothers
**CRD page:** [Three Brothers/Central Chagos Coral Surveys BITB38-BITB49 (Mar 2015)](https://coralreefdiagnostics.com/three-brotherscentral-chagos-20150321-to-0324)

**23 Mar 2015**
**GPS:** 6.03978°S, 71.55077°E
**Setting:** protected (leeward) lagoonal submerged bank reef, inside the lagoon, 29 m
**Coral cover:** 40.5% (15 formal transects, D3–D4; **algal cover 27.9%**) · **published image bleaching:** 6.0% (`Mean(%bleaching)` column)
**Live hard coral cover across methods:** formal transects 40.5% · CoralNet points 28.6% · Coralscapes masks 28.1% (36 frames). These products have different sampling footprints and denominators and should not be averaged.
**Temperature:** 29.7 °C · **Salinity:** 34.6
**Photo review:** 25 of the 36 frames in `BiTb44-fnished`, a systematic sample spanning the whole dive sequence — P3230388, 390, 392, 400, 401, 409, 411, 413, 415, 421, 422, 430, 431, 433, 435, 436, 444, 446, 447, 448, 450, 451, 453, 455, 463. The 11 unexamined frames (389, 391, 399, 410, 412, 414, 432, 434, 445, 449, 454) are interleaved between reviewed frames throughout the set, so no stretch of the dive is unrepresented.
**Representative photos:** P3230400.jpg, P3230409.jpg, P3230436.jpg, P3230421.jpg, P3230455.jpg, P3230463.jpg

BITB44 is the only one of the three Three Brothers sites surveyed on 22–23 Mar 2015 that sits **inside the lagoon**: a protected, leeward submerged bank reef at 29 m, and it carries the highest coral cover of that group. **40.5% coral against 27.9% algae ranks it 49th of 109 Chagos sites for cover — just above the archipelago median — and 92nd of 109 for algal cover, well into the least algae-dominated quartile.** The algal budget is unremarkable in composition: crustose coralline algae (PCCATS) 10.7%, 73rd of 109, and fleshy turf (PTTS) 6.8%, 84th of 109. What is distinctive is the substrate: **bare substrate accounts for 17.7% of the benthos** while non-coral invertebrates take only 14.0%. Coral, algae, bare rock and invertebrates sum to exactly 100%, so nothing is unaccounted for. This is a protected, high-cover reef with a lot of open pavement and comparatively little weed — and the photographs show exactly what occupies that pavement, with wide expanses of pink crustose coralline crust carrying dense juvenile coral recruitment (P3230444, P3230455).

Coral cover is spread across **25 nonzero genera, 50th of 109 Chagos sites**, and the 25 genus percentages sum to 40.50% of the benthos — an exact match to the published cover figure. All values below are percent of total benthos, not percent of coral. The assemblage is strongly dominated by two genera: ***Porites* at 18.93% of the benthos — 47% of all the coral at the site** — and *Acropora* at 9.36% (23%). Those two alone hold 70% of the coral present, and the drop to third place is steep: *Pocillopora* 3.04% (8%), *Montipora* 2.30% (6%), *Platygyra* 2.04% (5%), *Favia* 1.27% and *Favites* 0.79%. Every remaining genus is under 0.55% of the benthos — *Pavona*, *Astreopora*, *Montastrea*, *Stylophora*, *Leptastrea*, *Acanthastrea*, *Psammocora*, *Galaxea*, *Cantharellus*, *Gardineroseris*, *Fungia*, *Goniastrea*, *Hydnophora*, *Ctenella*, *Echinopora*, *Cyphastrea*, *Leptoseris* and *Oxypora*. This is a two-genus reef with a long, thin tail, and the imagery matches: massive and digitate *Porites* boulders and tabular/corymbose *Acropora* are the structural backbone of almost every wide frame (P3230388, P3230390, P3230400, P3230409, P3230411, P3230431, P3230433, P3230463).

**The two depth strata differ far more in composition than in cover, and the deep stratum is where the site becomes a monoculture.** Coral cover barely moves — 39.0% at D3 (13 transects) and 42.0% at D4 (2 transects) — but everything around it does. At D3 the assemblage is genuinely mixed and near-balanced between its leaders: ***Porites* 13.37% of the benthos (34% of coral there) against *Acropora* 12.22% (31%)**, with *Montipora* 4.11%, *Pocillopora* 2.58% and *Favites* 1.58% behind them, across **25 genera**. At D4 the balance breaks: ***Porites* rises to 24.50% — 58% of all coral on that stratum — while *Acropora* falls by nearly half to 6.50%**, *Montipora* collapses from 4.11% to 0.50%, *Platygyra* climbs from 0.58% to 3.50%, and the genus count drops from 25 to **just 9**. The substrate shifts in step: **algal cover nearly halves with depth, 36.2% at D3 to 19.5% at D4, while bare substrate more than triples, 8.3% to 27.0%**, and non-coral invertebrates fall from 16.5% to 11.5%. The site-level figures of 18.9% *Porites* and 9.4% *Acropora* therefore describe neither stratum: they average a diverse, algae-covered, near-co-dominant shallow reef against a deep stratum that is more than half *Porites* growing on largely bare rock. **That deep contrast rests on only two transects, and the two disagree sharply** — one records 35% coral with **20% dead coral and 38% bare substrate**, the other 49% coral with no dead coral recorded and 16% bare; the 20% dead-coral reading is the highest at either site depth here and should not be smoothed away. Across all 15 transects coral cover spans 25–64%, so within-depth patchiness comfortably exceeds the 3-point difference between depth means.

The photographs document a structurally varied reef with real high-cover ground. Broad tabular *Acropora* fields are the signature element — P3230390 and P3230400 both show multiple large gold-brown tables set in dense corymbose and branching stands, and P3230411 puts overlapping tables directly beside massive *Porites* boulders and pink CCA-encrusted framework. Large massive and digitate *Porites* colonies recur through P3230388, P3230409, P3230413 and P3230433, and P3230431 adds substantial grey-lavender foliose and columnar colonies to the mix. P3230435 is a single dense staghorn/branching *Acropora* thicket filling the frame. Against those, P3230392 and P3230446 are open counter-frames: low-relief pavement with scattered small colonies and visible bare and rubble ground. **P3230444 and P3230455 are the most informative structural frames — broad areas of dead, pink CCA-encrusted pavement (P3230444 is recognisably an old dead table) carpeted with small corymbose *Acropora* recruits.** They put a face on the site's 17.7% bare substrate and on the 20% dead-coral transect at D4: this reef has lost framework and is visibly recolonising it. Water clarity is good throughout the wide frames.

**Condition reads as locally, not uniformly, affected, and the published figure and the photographs agree in kind if not in detail.** `Mean(%bleaching)` records **6.0% for BITB44, 61st of 98 Chagos sites with a value** — mid-range severity for the archipelago — and both CoralNet training runs return 5.98%, so no run-selection divergence applies here. The photographs carry four instances of white or pale coral tissue, in ascending order of confidence. **P3230436 is the clearest: a *Pocillopora*-type colony in which roughly half the branch mass, a contiguous central-right block, is bone-white with branch structure fully intact, against normally brown tissue on the remainder** — textbook partial bleaching rather than old skeleton, with a small white patch on an adjacent massive *Porites* at frame left. **P3230430 shows the same signature on a corymbose *Acropora*-type colony**, with two discrete bone-white branch clusters set in otherwise pigmented tissue. **P3230401**, a macro looking down on a pale-cream corymbose colony growing on a large dark *Porites*, shows white tips over most of the colony and is markedly paler than everything around it — consistent with a bleaching or severely paled colony, though a naturally pale morph cannot be excluded from one still. **P3230435** shows numerous white branch tips throughout a large *Acropora* thicket; white apical tips are also normal rapid growth in *Acropora*, so this frame is recorded as ambiguous and not counted as bleaching. Most colonies in most frames are normally pigmented and there is **no coherent site-wide bleaching field**. **No COTS individual or feeding scar, no disease lesion, no tissue-loss margin and no corallivore bite scarring was seen in any of the 25 frames reviewed**; a prior review flagged a small COTS mask signal in P3230422 as false, and the present pass confirms it — P3230422 is an under-ledge shot of a spiny lobster.

Fauna includes two conspicuous set-pieces. **P3230421 documents an aggregation of at least three or four painted spiny lobsters (*Panulirus versicolor*-type — banded legs, long pale antennae) sheltering together beneath a sponge-encrusted overhang**, with P3230422 a closer frame of one animal. A single large, warty, pink-purple ambush predator consistent with a devil scorpionfish (*Scorpaenopsis*-type, identification kept conservative) is photographed repeatedly on open pavement across five frames — P3230447, P3230448, P3230450, P3230451 and P3230453 — from several angles; it is the same individual, not five animals. A map/honeycomb-patterned pufferfish (*Arothron*-type) sits mid-frame in P3230463. Butterflyfish, chromis and humbug damsels are constant background throughout the wide frames.

<b>Assemblage composition.</b> Formal transects give 40.5% coral, 27.9% algae, 17.7% bare substrate and 14.0% non-coral invertebrates across 25 nonzero genera, with *Porites* (18.9%) and *Acropora* (9.4%) holding 70% of all coral between them. The depth profile holds cover almost flat (39.0% at D3, 42.0% at D4) while composition changes completely: a 25-genus, near-co-dominant, algae-covered shallow stratum gives way to a 9-genus deep stratum that is 58% *Porites* on largely bare rock, with algal cover halving and bare substrate tripling. The photographs corroborate a tabular-*Acropora*-and-massive-*Porites* reef with extensive CCA-encrusted dead pavement under active juvenile recolonisation, localized partial bleaching in at least two colonies, a lobster aggregation, and no sign of COTS, disease or predation.

**Flags:** (1) **The D4 stratum rests on two transects and they disagree sharply.** One records 35% coral with **20% dead coral and 38% bare substrate**; the other 49% coral, 0% dead coral, 16% bare. Site cover is the unweighted mean of the two stratum means, so those two transects carry roughly half the headline 40.5%. The 20% dead-coral value is a real signal worth checking against the field sheet, not an averaging nuisance. (2) **The published bleaching figure is a COMPUTER-VISION product, not an in-situ observation.** It comes from CoralNet annotation of the informal survey images (Mayfield & Dempsey 2026, *Oceans*). Per-image human-vs-model agreement is far weaker for bleaching than for cover: r = 0.40 for bleached against r = 0.72 for live hard coral (`CHAGOS_MULTIMETHOD_README.md`). Treat it as the weaker of the two published numbers and never as a diver's field estimate. (3) **Bleaching percentages here use three different denominators and do not reconcile.** The published `Mean(%bleaching)` is 6.0%; `chagos_pct_bleached_recomputed.csv` gives 16 bleached points, which is 1.48% of all 1,080 points but 4.92% of the 325 coral points including bleached, and that file marks the site `MISMATCH` at −1.08 pp against the published value; Coralscapes masks average 0.22% of image area. None of these is wrong — they measure different things — but they must never be quoted interchangeably. (4) **Multi-method deviation is −12.4 to −11.9 pp** (transects 40.5%, CoralNet 28.6%, Coralscapes 28.1%, `sd_all` 7.0). That is *smaller* than the corpus expectation for a 40%+ cover site, where the documented photo-vs-transect gap averages −23.3 pp (`CHAGOS_MULTIMETHOD_README.md`). Do not average the three; the transect value is ground truth for cover. (5) **`coralnet_points_pct` is a human 30-point read, not a model output** — do not describe it as a CV product. (6) **`n_photos` = 36 for the CV comparison and the folder also holds 36 frames**, which is unusual for this corpus; the two happening to match here does not license printing a folder count beside a CV figure at other sites. Only 25 of the 36 were examined in this review. (7) **P3230450 carries a strong cyan/blue-green colour cast** relative to the other four frames of the same scorpionfish and should be excluded from any colour-based condition scoring. (8) **The transect file contains two rows both keyed `BITB44_d3_T1`** with different values (26% and 36% coral, 6% and 16% bare substrate). One of the fifteen D3 transect IDs is duplicated in the source; the depth mean is computed over both rows regardless, but the ID collision should be resolved upstream. (9) **Formal data mix 100- and 500-point transects** (11 at 100 points, 4 at 500) and are represented here by percentage fields only. (10) **No D1 or D2 stratum exists at this site** and no per-frame depths are recorded, so no photograph can be tied to a depth transect. (11) **Genus richness collapses with depth, 25 genera at D3 to 9 at D4** — the deep stratum is close to a *Porites* monoculture and should not be described using site-level composition. (12) **No COTS, disease or predation scarring was observed** across the 25 frames reviewed — an explicit absence-of-evidence finding. The COTS mask signal previously reported in P3230422 is confirmed false. (13) **Fish and invertebrate identifications are kept conservative:** the lobsters in P3230421–422 are painted-spiny-lobster-type, the ambush predator in P3230447–453 is scorpionfish-type and is one individual across five frames, and the pufferfish in P3230463 is identified to type only. (14) **NON-CANONICAL BLEACHING RANK — added 2026-08-26.** Any "of 98" bleaching rank above was computed off the site table's `Mean(%bleaching)` column, which equals the **1st** CoralNet training run. Anderson's decision of 2026-08-26 (HOW_TO rule 4e) makes **`% bleaching (2nd training)` canonical**, The file carries TWO pools for it, and they are not interchangeable: a **110-row pool** ranking every dive (revisit dives included, so BIPB70B/72B/69C compete) and a **99-site pool** ranking one value per site. The rank quoted below is on the 110-row dive pool. For this site the canonical value is **5.98%, rank 69 of 110** (1st training: 5.98%, rank 68 of 110). The two runs agree to <0.05 pp on 100 of 110 rows and diverge only on the bleaching-onset dives, so for most sites the distinction changes the rank but not the number.

**Benthic data:** [Download BITB44 benthic survey data (CSV)](../benthic_data/BITB44_benthic.csv) — scleractinian cover by genus with full names, plus substrate and algal categories (bare substrate, CCA, turf, turf+sediment, macroalgae, erect algae, cyanobacteria, invertebrates), from the LOF BIOT-1 survey (Mayfield & Dempsey 2026, *Oceans*, supplemental file S1).