### BITB33 — South Brother
**CRD page:** [Three Brothers & Central Chagos Coral Surveys (Mar 17-20, 2015)](https://coralreefdiagnostics.com/three-brotherscentral-chagos-20150317-to-0320)
**Date:** 2015-03-19 (dive window 10:25:38–11:12:31, 46 min 53 s, from `DateTimeOriginal` EXIF on all 27 files, Olympus TG-2. The counter runs P3190442→P3190490, so **the edited set keeps 27 of ~49 exposures** — a frame selection upstream of every photo-derived number below.)
**GPS:** 6.17930°S, 71.58126°E · **Depth:** 25.7 m (supplement site record)
**Frames reviewed:** 27 of 27 (100% of the site folder, `BiTb33-finished`; 0 evicted, 0 EPERM, and **all 27 carry CV data — complete coverage, nothing lost to the filename collision**)
**Coral cover:** 18.54% (formal transect surveys; algal cover 52.71%, bare substrate 20.48%, invertebrates 8.27%) · **% bleached:** 12.01% (`% bleaching (2nd training)`, Mayfield & Dempsey 2026 supplement) — the CRD site page independently gives "coral cover=19% (LOW), algal cover=52% (HIGH)".
**Reef type:** lagoonal bank reef, submerged, inside the lagoon, protected · **Temperature:** 29.5 °C · **Salinity:** 34.4
**Live hard coral cover across methods** (n=27 photos of 27): transect 18.54% · CoralNet points 12.72% · Coralscapes 11.11%. Deviation vs transect **−5.8 to −7.4 pp. Mean of 3: 14.1% ± 3.9.** **This is the SMALLEST deviation of the entire Three Brothers series, on complete coverage** — every written-up neighbour reads 9–28 pp low (BITB27 −11.8, BITB28 −9.0, BITB29 −9.9, BITB30 −24.3, BITB32 −17.0, BITB34 −28.4). Flags 1–3 explain why, and the answer is not the flattering one.
**Coral genera identified:** the transect recorded **27 scleractinian genera** across 10 transects in three depth strata — *Acropora* 4.43%, *Porites* 4.07%, *Pocillopora* 2.51%, *Platygyra* 2.14%, *Favia* 1.43%, *Favites* 0.93%, *Montipora* 0.90%, then a long tail to *Acanthastrea* and *Fungia* at 0.01% — **and *Seriatopora* at exactly zero**, matching the CRD page. From the photographs: ***Acropora*** in four habits (tabular plates to ~1.5 m, corymbose/digitate, gold-green arborescent staghorn, and **a maroon-purple branching form in P3190490**), ***Porites*** massive/domal, knobby-columnar and encrusting, ***Pocillopora***-type verrucose clumps, faviid submassives compatible with *Favia*/*Favites*/*Goniastrea*, meandroid brains compatible with *Platygyra*, sheets compatible with *Montipora*/*Pavona*, coarse honeycomb colonies compatible with *Astreopora*; plus **dark ruffled fleshy macroalgae compatible with *Lobophora*/*Dictyota*** (P3190487 is the best frame), *Halimeda*, and pervasive pink CCA. Growth-form calls from photographs are given as "compatible with".
**Representative photos:** P3190484.jpg, P3190490.jpg, P3190466.jpg, P3190454.jpg, P3190456.jpg

**South Brother is a coral garden on a sand-and-pavement plain, and the imagery makes 18.54% look exactly right.** Nothing here is continuous reef. **P3190470, P3190480, P3190485 and P3190489** are the defining frames — white sand and pale consolidated pavement running to the horizon, with tan corymbose *Acropora*, small *Porites* domes and *Pocillopora*-type clumps on it as **well-spaced, individually resolvable colonies**, none touching. **P3190443** is the emptiest: a low plain of small olive branching colonies over bare sand. **P3190481 and P3190469** are the same ground with dark maroon algal mats spreading across the pavement, and **P3190468** — pink-and-maroon CCA over old dead massive skeletons, threaded with green *Halimeda* — is the clearest statement of old, non-acute structural loss here. The transect agrees with the eye: **bare substrate is 20.48%, the highest of any Three Brothers site checked** (BITB29 12.91, BITB30 11.13, BITB32 11.57, BITB34 7.24), and the D4 transects are sand-dominated (233 and 210 sand points) while the shallower D3 transects carry none at all. This is a genuinely low-cover, high-heterogeneity site, not a badly photographed high-cover one.

**Where coral does concentrate it is healthy and Acroporid.** **P3190484 is the highest-cover frame** (36.7% by points, 30.5% by mask — the closest two-method agreement at the site): two gold-green tabular *Acropora* a metre and more across, over tan lobed *Porites* and small faviids, a moorish idol (*Zanclus cornutus*) behind. **P3190486** is the best mixed frame — gold digitate *Acropora*, a large brown knobby-columnar *Porites*, and a pinkish-grey branching colony that is **the site's clearest partial-mortality candidate**, branch tips already carrying CCA. **P3190482** is a staghorn-and-corymbose *Acropora* thicket with a **powder-blue surgeonfish (*Acanthurus leucosternon*)** in it, and **P3190490, the last frame of the dive, is the signature image**: gold-green and maroon-purple branching *Acropora* **with a yellow transect tape running clean across the upper third of the frame** (flag 5). **P3190487 and P3190488** are the two close benthic frames and carry the site's highest mask cover (34.9%, 31.6%) on encrusting and plating colonies the wide shots never resolve.

**The fauna is dominated by one enormous trevally aggregation.** **P3190463, P3190464 and P3190465 are among the best schooling-fish frames of the redo** — several hundred silvery, black-spotted, yellow-finned jacks filling the water column, **compatible with *Carangoides fulvoguttatus* but not callable to species**. **P3190466 is the same school with a diver in it** (flag 5). Earlier, **P3190444, P3190445 and P3190446** hold dense fusilier schools compatible with *Caesio*/*Pterocaesio*, the last banked over a broad lobate *Porites* dome on white sand. **P3190442, the opening frame, is a requiem shark in blue water — and it is NOT a whitetip**: broad dark caudal margin, no white dorsal tip, so **grey reef (*Carcharhinus amblyrhynchos*) or a dusky-phase relative; not callable** (flag 8). Beyond those: moorish idols in P3190470, P3190484 and P3190485, a sergeant-major-type damsel and dark surgeonfish in P3190485, blue damsels and *Chromis* in P3190489.

**And, as at BITB29, the dive goes under a ledge.** **P3190453, P3190454 and P3190455 are three consecutive macros of a painted spiny lobster** — boldly black-and-white banded legs, striped carapace, pale antennae — beneath an overhang crusted in **yellow and red encrusting sponge**, over coarse rubble. **Compatible with *Panulirus versicolor* or *P. longipes*; not callable.** At least one animal, with a second set of antennae and legs entering the top-left of P3190454. They are 11% of the CV sample, composed on an animal in shadow rather than substrate, and carry **3 of the site's 13 bleached points** (flag 4).

<b>Assemblage composition.</b> **This paragraph rests on the FULL 27-frame set.** Across all 27 the Coralscapes masks put live hard coral at 11.11% of image area, of which <i>Acropora</i> (non-tabular) is the largest component at 30.0%; the rest breaks down as massive and meandroid 28.1%, other hard coral 19.7%, tabular <i>Acropora</i> 15.2% of live cover. The independent 30-point annotation of the same frames (810 points) gives live hard coral as 12.72% of all points and apportions it <i>Acropora</i> 44.7%, <i>Porites</i> 29.1%, other hard coral 23.3%, <i>Pocillopora</i> 2.9% of live coral — **an Acropora-led assemblage, the reverse of BITB29's <i>Porites</i> dominance, and consistent with the transect's *Acropora* 4.43% > *Porites* 4.07%.** Non-coral cover is dominated by turf 23.0% and crustose coralline algae 6.7% of all points. Structural loss registers as dead coral 3.45% of image area and rubble 11.23%. Bleaching stands at 1.60% of annotated points — **13 points** — and Coralscapes puts bleached area at **0.01% of the entire site**, essentially zero. Frame-to-frame live cover runs 0% (P3190442, P3190443, P3190456, P3190465, P3190466) to 36.7% (P3190484); per-frame point-vs-mask correlation is r = 0.590, below the Chagos-wide 0.7155. Conspicuous fauna and sessile invertebrates appear as sponges in 15 frames, sea cucumbers in 7 frames, giant clams in 2 frames, urchins in 2 frames. **The 52.71% "algal cover" needs unpacking before it is read as degradation: it is CCA 15.28% + turf 12.02% + turf-with-sediment 11.67% + macroalgae 7.94% + erect algae 5.74% + cyanobacteria 0.06%. Only 13.68 pp of it is fleshy algae.**

**Flags:** (1) **VERDICT — THE SMALLEST GAP IN THE SERIES IS NOT A BETTER RESULT. IT IS THE SAME BIAS MEASURED AT THE LOWEST-COVER SITE, WHICH IS ITSELF PROOF THE BIAS IS MULTIPLICATIVE.** BITB33's raw point cover is **0.687× its transect**; the other three complete-coverage Three Brothers sites give **0.686 (BITB28), 0.677 (BITB27), 0.671 (BITB29)** — four sites spanning 18.5–36.6% transect cover, ratio constant to ±0.008. A multiplicative bias yields a small *absolute* pp gap wherever cover is low. Across the 24 Three Brothers sites with a transect value and ≥8 frames, **r(transect %, gap pp) = −0.724**: a site's own cover, not anything about its imagery, dominates how large its deviation looks. **Do not cite BITB33 as a site where the photo methods "did better".** (2) **THE UNKNOWN-DENOMINATOR TEST — CONFIRMED, BUT AS THE SECOND-ORDER TERM. BITB33's Unknown fraction is 25.06% (203 of 810 points), lower than every written-up neighbour** (BITB27 27.92, BITB30 29.49, BITB29 32.24, BITB28 32.27) — the direction the artefact predicts — and **dropping `Unk` from the denominator lifts the point set 12.72% → 16.97% against an 18.54% transect, a residual of −1.57 pp**, squarely in the LOW-reading pattern (BIEA26 −0.75, BITB29 −0.30, BITB27 −2.21). **The mechanism is explicit: 1/(1−0.2506) = 1.334, and 12.72 × 1.334 = 16.97 — the Unknown pool IS the multiplier of flag 1.** Two cautions: **25.06% is only modestly low** (Chagos median 26.99% across 102 folders; Three Brothers median 26.63%), and across those 24 sites **r(Unknown %, gap pp) is only −0.391 against −0.724 for the transect value**. **A smaller Unknown pool is real and contributes, but the reason BITB33's gap is small is that BITB33's reef is poor. It is nonetheless the cleanest natural control in the series — low Unknown, low cover, complete coverage, corrected value 1.6 pp off the transect.** (3) **DROPPING NON-SURVEY FRAMES *AND* `Unk` LANDS ON THE TRANSECT TO WITHIN 0.15 pp — the closest landing of the redo.** Eight of 27 frames are not benthic survey imagery: three lobster macros (P3190453/454/455), one close "spot" macro (P3190456), four open-water fish/shark frames (P3190442, P3190463, P3190464, P3190465). On the remaining 19, excluding Unknown gives **18.69% vs 18.54%**; adding the two methodology frames gives 19.25% (+0.71). **30% of this site's frame set is not a cover measurement.** (4) **THE 12.01% PUBLISHED BLEACHING FIGURE IS A METRIC ARTEFACT, AND BITB33 IS THE SITE THAT PROVES IT.** Anderson recorded **no bleaching in March or April 2015 — only May** (half-month means 6.1 / 8.2 / 8.4 / 10.0% through March–April against **31.8% in early May**), and **the photographic record does not corroborate 12.01%**: 13 bleached points in 810, with the Coralscapes mask reading 0.01% bleached area site-wide. **BITB33 is among the highest March values in the archipelago, which sharpens the discrepancy — and the cause is now identifiable.** `% bleaching (2nd training)` is **the unweighted mean, across a site's images, of bleached/(live+bleached) points *per image***. Recomputed that way it reproduces **12.006% against the published 12.00617%** here and **7.021% against BITB29's published 7.02%**; across 101 sites the median absolute difference is **0.0002 pp, 67/101 inside 0.5 pp** (the misses are the known duplicate-code and May folders, where folder→code mapping is ambiguous). **A frame with 2 bleached points and ZERO live-coral points therefore scores 100% and is weighted equally with a frame carrying 30 coral points.** That is exactly what happens: **P3190456 — a close downward macro of a pink-CCA-crusted mound, scored 0.0% live despite plainly showing four small gold and grey colonies — scores 100% and contributes 3.70 pp of the site's 12.01%; P3190444 (53.3% open water, 3.3% live) scores 66.7% and contributes 2.47 pp. Those two frames alone are 51% of the published figure**, and three more of the eight contributing frames are ≥16% water. **Recommendation for Anderson: this is a corpus-wide finding, not a BITB33 one — the March/April column should be recomputed as a points-weighted ratio (here 13/116 = 11.2% of coral points, or 1.60% of all points) before these figures are cited again.** (5) **METHODOLOGY IMAGERY, TWO EXCELLENT FRAMES. P3190490 carries a YELLOW TRANSECT TAPE across the frame** — one of the clearest direct images of the apparatus behind a Chagos cover figure — and **P3190466 shows a diver in yellow fins with a white rectangular open frame, apparently a photoquadrat, on the reef at lower left. Both need verifying at full resolution**, alongside BIEA20/P3150311, BITB27/P3170391 and BITB29/P3180508. **P3190456 is a third, "spot"-framed measurement macro.** (6) **BLUE-WATER DILUTION IS REAL HERE, UNLIKE AT BITB29.** Eight frames are ≥40% water (P3190442, P3190444, P3190445, P3190463, P3190464, P3190465, P3190466, P3190482); site mean 28.64%. Dropping them moves the point set **12.72 → 15.26 (+2.54 pp)** and Coralscapes **11.11 → 14.25 (+3.14 pp)** — ~43% of the gap, where the same test at BITB29 moved them 1.15 and 0.19 pp. **Do not stack the corrections: excluding Unknown on the <40%-water subset overshoots to 21.64% (+3.10 pp).** The ≥40%-water frames also carry a *lower* Unknown fraction (14.58%) than the rest (29.47%), so here dilution and the Unknown pool are largely independent. (7) **SUBSTRATE RENORMALISATION OVER-CORRECTS, AS AT BITB27/29** — excluding water, Unknown and fish gives **28.22% against 18.54%**; excluding Unknown alone is the correction that behaves. **Water scoring is again internally inconsistent**: P3190490 is WATER 0.0% with 43.3% Unknown while Coralscapes reads 31.2% dark background on the same pixels; P3190488 is 0.0% water / 30.0% Unknown / 17.3% mask background. **Part of the Unknown pool is misfiled water.** (8) **THE POINT SET UNDER-READS THE ACROPORA-THICKET FRAMES BADLY.** P3190482 scores 6.7% live by points against 25.9% by mask on a plain staghorn thicket; **P3190443 scores 0.0% live with 60.0% Unknown** on a frame full of small live colonies; P3190463 is 23.3% by points against 0.09% by mask. **With ~76% of Chagos point annotations being CoralNet robot predictions these are annotation-provenance artefacts, not observer error.** (9) **THE OLD DESCRIPTION'S CALLS NEED CORRECTING.** P3190442 is **not** "a whitetip reef shark" — no white fin tips, broad dark caudal margin; leave it uncallable (the BIEG04 lesson). The P3190453–455 ledge holds **encrusting sponge, not "fluorescent-green encrusting coral"**. And "pale-to-white patches appear on several isolated coral clusters (P3190469, P3190481)" **is not supported** — neither frame carries a paling colony. (10) **THE ONE GENUINE PALING CANDIDATE IS P3190489** — a cream-yellow lobed massive against a dark maroon algal mat, among normally pigmented neighbours, carrying one bleached point. **Not callable from the frame**; it may equally be a naturally pale *Porites*. **Anderson's judgement needed**, and it is the only colony here worth asking about. (11) **NO CROWN-OF-THORNS IN ANY OF THE 27 FRAMES; no disease lesions, no corallivore predation scars.** Mortality is old — CCA-plated dead massives (P3190468), maroon-crusted pavement (P3190481, P3190489), Coralscapes dead-coral site mean 3.45%, maximum 24.35% on P3190456. (12) **The edited set is 27 of ~49 exposures** — the frame-selection confound, upstream of every photo-derived number above. (13) **Survey effort: 10 transects across three depth strata (D2–D4)**, below the 15-transect Chagos median; the 27-genus richness is nonetheless solid and is not a richness floor. (14) No per-frame depth data.

**Benthic data:** [Download BITB33 benthic survey data (CSV)](../benthic_data/BITB33_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).