### BITB27 — Middle Brother
**Date:** 2015-03-17 (dive window 15:18:59–16:10:48. **The manifest carries no date for this site** — the date comes from the CRD website page "Three Brothers & Central Chagos Coral Surveys (Mar 17–20, 2015)", the window from `DateTimeOriginal` EXIF on all 45 files, in three bursts: 15:18–15:27, 15:53–16:01, 16:01–16:10. Same day as BIEA25 and BIEA26 — which is exactly why the frames collide with theirs. See flag 1.)
**GPS:** 6.15622°S, 71.50993°E
**Frames reviewed:** 45 of 45 (100% of the site folder, `BiTb27-meter spot-finished`; none evicted). **Only 16 of the 45 carry CV data.** Every photo-method number below rests on those 16, not on the 45.
**Coral cover:** 36.6% (formal transect surveys; algal cover 46.4%) · **% bleached:** 0.35% (Mayfield & Dempsey 2026, Oceans) — **effectively an unbleached site, the lowest of the redo.**
**Reef type:** fringing, intermediate exposure · **Temperature:** 29.5 °C
**Live hard coral cover across methods** (n=16 photos of 45): transect 36.6% · CoralNet points 24.8% · Coralscapes 22.3%. Deviation vs transect **−11.8 to −14.3 pp** — both photo methods read LOW, the same direction as BIEA26 and the opposite of BIEA22/23/25. **Mean of 3: 27.9% ± 7.6.** **Most of that gap is an artefact of which 16 frames happened to be scored** — see flags 2 and 3.
**Coral genera identified:** 6 scleractinian groups — ***Acropora*** in four growth forms (**tabular plates up to ~2 m across**, corymbose/digitate cushions, arborescent staghorn, and a **purple-mauve branching form** in P3170444), ***Porites*** (massive/domal and knobby-lobed), ***Pocillopora***-type knobby clumps, meandroid brains compatible with *Platygyra*/*Leptoria*, sheets compatible with *Montipora*, and faviid submassives compatible with *Goniastrea*; plus pervasive pink and mauve **crustose coralline algae**, sponges in 12 frames, one giant clam and one sea cucumber. Growth-form calls are from photographs and given as "compatible with"; no species is asserted.
**Representative photos:** P3170440.jpg, P3170458.jpg, P3170461.jpg, P3170462.jpg, P3170460.jpg

**BITB27 is an *Acropora* plate reef, and the healthiest site written in this redo so far.** At 36.6% cover with 0.35% bleaching it is nearly three times BIEA24 (12.7%) and well ahead of BIEA22 (21.6%) and BIEA26 (21.0%) — though within Three Brothers it is unremarkable, sitting between BITB29 (30.2%) and BITB30 (42.5%) and far below BITB34 (61.1%). The structure is consistent across the whole hour: **tiered, overlapping tabular *Acropora* plates laid over pale carbonate pavement**, with staghorn and corymbose *Acropora* filling the gaps and massive *Porites* domes anchoring the low ground. **P3170461 is the best single image of it** — one enormous olive-gold plate, easily two metres across, standing alone over white sand with a healed pale patch at its centre. **P3170458 and P3170456** show the other mode: plates packed so closely they form a continuous gold-green pavement of living tissue across the whole foreground, broken only by pink CCA-crusted skeleton. **P3170455, P3170457 and P3170459** repeat it at wider angle, plate stacked on plate to the limit of visibility.

**Where the reef is not plated it is branched, and densely so.** **P3170440 is the highest-cover branching frame** — golden-tan corymbose and digitate *Acropora* filling the entire field with a single *Porites* dome pushing through. **P3170438** holds arborescent staghorn thickets with bright pink encrusting sponges and dark gaping bivalves set into the framework; **P3170444** carries the site's most distinctive colony, a **large purple-mauve branching *Acropora*** beside a tan tabular plate; **P3170462** is a mixed stand of at least four *Acropora* forms with **knobby brown *Pocillopora*-type clumps** at its base; and **P3170433, P3170436, P3170439, P3170445 and P3170459** all carry **meandroid brains compatible with *Platygyra*** tucked at the foot of the plates. Massive *Porites* is constant but never dominant — domes in P3170435, P3170445, P3170449 and P3170459, and a knobby-lobed sheet across the foreground of P3170453.

**One taxonomic note needs recording.** The CRD site page states that **neither *Seriatopora* spp. nor *P. acuta*/*P. damicornis* were present here, and that other pocilloporids were collected instead for a collaborator.** The frames are consistent with that — the pocilloporid material visible in **P3170447, P3170449 and P3170462** is blunt and knobby, compatible with a *P. verrucosa*-type habit rather than the fine compact *damicornis*/*acuta* form, and nothing reads as *Seriatopora*. **This is a hedged photographic impression, not a survey result.**

**Health is genuinely good, and the little mortality present is old.** The dominant structural loss is **long-dead skeleton already resurfaced in pink and mauve CCA** — maroon-crusted branching frameworks in P3170442 and P3170444, a CCA-plated dead table in P3170441, algal-toned plates in P3170446 and P3170455. That is ordinary plate-reef turnover, not an event. **Two frames need separate attention and both are unscored.** **P3170443 is a macro of a massive *Porites*-type surface caught mid-paling** — pigmented brown polyps still standing out against a coenosarc gone stark white across the right two-thirds of the frame, with normal tan tissue at the left edge. **P3170454 shows a cream-white massive dome** among normally-pigmented neighbours. Neither is resolvable photographically as bleaching versus a pale morph versus recent partial mortality, and **neither carries a CV score**, so the 0.35% transect figure remains this site's only quantitative bleaching statement.

**The fauna is good but not the story.** The water column is worked continuously by **fusiliers** compatible with *Caesio*/*Pterocaesio* (P3170450, P3170452, P3170453, P3170454, P3170457, P3170458) and by **schools of dark surgeonfish and rudderfish** (P3170446, P3170448, P3170449, P3170451). On the reef itself: **powder-blue surgeonfish (*Acanthurus leucosternon*)** in P3170434, P3170435, P3170441, P3170460 and P3170462; **moorish idols (*Zanclus cornutus*)** in P3170433 and P3170435; red soldierfish compatible with *Myripristis* under plates in P3170439 and P3170447; blue chromis, cleaner wrasse and butterflyfish throughout. **The best macrofauna frame is P3170460** — a large dark grouper, white-spotted with pale saddle bars, resting on pavement beside a tabular plate, **compatible with a dark-phase blacksaddled coralgrouper (*Plectropomus laevis*)** but not callable from this frame.

**The folder name "meter spot" is earned, but by only six of the 45 frames.** **P3170391** carries a **transect line running diagonally across a tabular plate**; **P3170392** a **small white rectangular frame on a plate in the mid-background**; **P3170394** a **large white PVC quadrat frame in the immediate foreground** over sparse pavement; **P3170403** is the collection frame — a diver with a **punch/hammer** and a **yellow mesh collecting bag**, a slate diver behind and a vertical downline; and **P3170395 and P3170405** are close-up "spot" frames, pavement patches of pink CCA and small recruits composed for settlement or size measurement rather than cover. The other 39 frames are ordinary wide-angle survey imagery. **The problem is not that the folder is measurement-oriented — it is that all six measurement frames fell inside the 16 that got scored.**

<b>Assemblage composition.</b> **This paragraph rests on 16 of the site's 45 frames.** Across those 16 the Coralscapes masks put live hard coral at 22.3% of image area, of which tabular <i>Acropora</i> is overwhelmingly the largest component at 55.0%; the rest breaks down as <i>Acropora</i> (non-tabular) 27.0% and massive and meandroid 7.4% of live cover. The independent 30-point annotation of the same frames (480 points) gives live hard coral as 24.8% of all points and apportions it <i>Acropora</i> 87.4%, <i>Porites</i> 5.0%, other hard coral 5.0%, <i>Pocillopora</i> 2.5% of live coral — **the most <i>Acropora</i>-dominated assemblage of any site in the redo.** Non-coral cover is dominated by turf 15.0% and crustose coralline algae 5.2% of all points. Structural loss registers as dead coral 2.61% of image area and rubble 4.96%. Bleaching stands at 0.21% of annotated points — **a single point, on <i>Acropora</i>** — against the 0.35% transect figure. Frame-to-frame live cover runs 3.3% to 56.7% by point count (P3170391 highest-cover; P3170394, P3170395, P3170396 and P3170451 poorest). Conspicuous fauna and sessile invertebrates appear as sponges in 12 frames, giant clams in 1 frame, sea cucumbers in 1 frame, survey apparatus/debris in 2 frames.

**Flags:** (1) **29 OF 45 FRAMES WERE NEVER ANNOTATED, AND THE ANNOTATIONS DO NOT EXIST TO RECOVER.** All 45 are on disk, none evicted, but only 16 carry CV data. The comparison CSVs key on the bare lowercase filename with no site qualifier, and **29 of BITB27's 45 filenames are also used in other folders — 18 shared with `BiEa26-finished` and 11 with `BiEa25-finished`**, both dived the same day on the same continuous camera counter; those 29 rows read the **Eagle Island** folders' pixels. The 16 scored are exactly the filenames unique to this folder. **Verified: none of the 29 collided filenames is among the 16, so nothing is cross-contaminated and BIEA25/BIEA26 are unaffected — but BITB27's 29 frames were never uploaded to CoralNet and a re-key recovers nothing.** The visual review of all 45 is the stronger evidence here. (2) **VERDICT ON THE SUBSET QUESTION: THE 29 UNSCORED FRAMES ARE VISIBLY HIGHER-COVER, AND THE −12 TO −14 pp GAP IS LARGELY A SAMPLING ARTEFACT.** The densest imagery is almost entirely unscored — **P3170389, P3170390, P3170433, P3170436, P3170438, P3170439, P3170440, P3170444, P3170447, P3170456, P3170457, P3170458, P3170459, P3170461, P3170462** all read 35–55% by eye and **not one carries a CV number**; the scored 16 hold only two comparable frames (P3170391 at 56.7%, P3170449 at 50.0%). **The arithmetic agrees: drop the five lowest scored frames — P3170394, P3170395, P3170396, P3170451, P3170452 — and CoralNet rises from 24.8% to 33.6% against a 36.6% transect, −3.0 pp instead of −11.8.** Those five are precisely the blue-water and close-up "spot" frames. **Do not cite BITB27 as evidence that photo methods under-read cover.** (3) **ALL SIX METHODOLOGY FRAMES FELL INSIDE THE SCORED 16 — 38% of the CV sample is non-survey imagery** (P3170391, P3170392, P3170394, P3170395, P3170403, P3170405). Measurement frames are composed on a patch or on apparatus, not on the reef, so including them biases cover downward by construction. **Recommend excluding measurement imagery from photo-derived cover corpus-wide.** (4) **BLUE-WATER DILUTION IS PRESENT BUT MINOR — NOT THE EXPLANATION.** 5 of 16 scored frames are ≥40% water (P3170394 40.0%, P3170448 40.0%, P3170451 43.3%, **P3170452 66.7%**, P3170453 40.0%; mean 26.0%). Dropping them lifts CoralNet 24.8→28.2 and Coralscapes 22.3→24.6 — **still 8.4 and 12.0 pp below transect.** At most a quarter of the gap. (5) **P3170443 — PALING MACRO, UNSCORED**: the clearest bleaching-relevant image at the site, contributing nothing to the 0.21% point figure; paling versus recent partial mortality not resolvable. **P3170454** is a weaker second case. (6) **The point set puts one bleached point here (0.21%, *Acropora*) against a 0.35% transect figure — both say "unbleached".** With ~76% of Chagos point annotations being robot predictions, treat that as consistent, not as independent confirmation. (7) **27.9% OF ALL POINTS ARE "UNKNOWN"** (10.0–43.3%; P3170396 worst). Chagos-typical (corpus mean 27.25%; BITB27 53rd of 120 folders) but still over a quarter of the point budget unlabelled — **renormalising live coral over labelled benthic points only gives 53.4%, far above the transect**, so the Unknown pool is a bigger lever here than the water fraction. (8) **P3170396 is an annotation-provenance candidate**: 43.3% Unknown and 3.3% live by points, yet the foreground plainly carries corymbose *Acropora*; Coralscapes reads 12.7% on the same frame. (9) **P3170460's grouper is NOT callable to species** (dark-phase *Plectropomus laevis* is only compatible); the BIEG04 lesson applies. Flagged for illustrative use with **P3170461**, **P3170458** and **P3170440**. (10) **No COTS, no disease lesions and no predation scars in any of the 45 frames.** (11) **The *Seriatopora*/*P. acuta* absence on the CRD site page is consistent with the imagery but not testable from it — Anderson's judgement needed.** (12) No per-frame depth data. (13) **Recommend backfilling the missing BITB27 date into `chagos_work_manifest.json`.**

**Benthic data:** [Download BITB27 benthic survey data (CSV)](../benthic_data/BITB27_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).