### BITB29 — North Brother
**Date:** 2015-03-18 (dive window 10:30:58–11:24:23, 53 min, from `DateTimeOriginal` EXIF on all 49 files. The counter runs P3180467→P3180553, so **the edited set keeps 49 of ~87 exposures** — a frame selection upstream of every number below.)
**GPS:** 6.13459°S, 71.49796°E
**Frames reviewed:** 49 of 49 (complete CV coverage — no frames lost to the filename collision)
**Coral cover:** 30.2% (formal transect surveys; algal cover 50.66%) · **% bleached:** 7.02% (`% bleaching (2nd training)`, Mayfield & Dempsey 2026 supplement) — the CRD site page independently gives "coral cover=30.2%, algal cover=50.7%", confirming the transect figures.
**Reef type:** fringing, intermediate exposure · **Temperature:** 29.5 °C
**Live hard coral cover across methods** (n=49 photos of 49): transect 30.2% · CoralNet points 20.3% · Coralscapes 17.2%. Deviation vs transect **−9.9 to −13.0 pp** — both photo methods read LOW, the third Three Brothers site to do so. **Mean of 3: 22.6% ± 6.9.** **This is the control for the series: every frame is scored, so nothing here can be blamed on missing annotations.** See flag 1.
**Coral genera identified:** 7 scleractinian groups plus octocorals — ***Porites*** in three habits (massive/domal, knobby-lobed columnar, and **large pale grey-cream tiered plating sheets, the site's signature**), ***Acropora*** in four (**tabular plates to ~2–3 m**, corymbose/digitate, arborescent, and gold-green plating), ***Pocillopora***-type knobby clumps, meandroid brains compatible with *Platygyra*/*Leptoria*, faviid submassives compatible with *Goniastrea*/*Favites*, sheets compatible with *Montipora*/*Pavona*, and fungiids compatible with *Fungia*/*Lithophyllon* (P3180491); plus a **gorgonian sea fan** (P3180476, P3180500), ascidians and hydroids (P3180504), *Halimeda*, and pervasive CCA. Growth-form calls are from photographs, given as "compatible with".
**Representative photos:** P3180526.jpg, P3180493.jpg, P3180519.jpg, P3180474.jpg, P3180548.jpg

**BITB29 is a *Porites* reef, and that single fact drives most of what is odd about its numbers.** Where BITB27 was tiered *Acropora* plate on plate, North Brother is built from **large pale grey-cream plating and lobate *Porites*** laid across the slope in sheets metres wide. **P3180474 is the clearest image of it** — a continuous nodular pale-grey living pavement filling the frame, the site's second-highest-cover at 53.3% by points. **P3180469, P3180471, P3180500, P3180510 and P3180511** repeat it, whole slopes of tiered cream plates with fusiliers over them. The point set agrees with the eye: *Porites* is **52.3% of live coral points** against *Acropora*'s 36.6%, the reverse of BITB27's 87.4% *Acropora*. Pale, low-contrast, low-relief tissue is the hardest thing on a reef to separate from dead pavement, and it duly defeats one of the two models (flag 3).

**The *Acropora* is present, large and healthy where it occurs.** **P3180526 is the signature frame** — one olive-gold tabular plate two to three metres across standing alone over rubble pavement, margin unbroken, a small healed pale scar at the centre. **P3180537** is the best mixed frame: gold tables over tan lobed *Porites* with three large barred groupers behind; **P3180477, P3180478, P3180505 and P3180539** each carry single clean tables. Against these, **P3180512, P3180514 and P3180519 are the mosaic frames** — dense fields of small colonies packed shoulder to shoulder, mauve *Pocillopora*-type clumps, corymbose *Acropora*, tan *Porites* knobs, faviids and *Montipora*-type sheets, all fringed with pink CCA and turf. By eye those are 40–55% frames; the point set gives them 20.0%, 16.7% and 36.7%.

**A genuinely low-cover habitat also exists here, and it is not an artefact.** Roughly a third of the swim crosses **low-relief consolidated pavement, rubble and old dead tabular plates already resurfaced in maroon and pink CCA**. **P3180540 states it best** — a downward view of dead tables crusted maroon carrying only scattered tan *Acropora* recruits; **P3180516, P3180517, P3180527, P3180542 and P3180543** are the same ground at wide angle, and **P3180518** puts a **whitetip reef shark (*Triaenodon obesus*)** over a rubble plain with almost no coral in frame. A 53-minute swim samples this while a transect line need not. Structural loss is old rather than acute, and **there are no COTS, no disease lesions and no predation scars in any of the 49 frames.** The one thing the imagery cannot settle is bleaching: the cream colour of the big *Porites* plates in P3180469, P3180471, P3180474 and P3180510 is the dive's dominant visual impression and is not separable photographically from paling — Chagos *Porites* is often this colour at depth — while **P3180539 holds one distinctly cream-white massive dome** among normally-pigmented neighbours.

**The last two minutes of the dive are not a reef survey at all, and that matters more here than anywhere else in the redo.** From 11:22:15 the camera goes under a plate overhang and stays: **P3180544–P3180553 are eight consecutive macros of a spiny lobster den** — at least three and probably four *Panulirus*-type animals with boldly black-and-white banded legs and striped carapaces, **compatible with *P. versicolor* or *P. longipes* and not callable from these frames** — under a red-sponge-crusted ledge over coarse rubble. **They are the best invertebrate record in the redo and also 16% of the CV sample**, composed on animals in shadow rather than substrate: mean Unknown 46.7% against 29.4% elsewhere, mask cover near zero, and **9 of the site's 20 bleached points** (flag 4).

**The macrofauna is the best of the Three Brothers series, and the survey apparatus makes one appearance.** **P3180493 is the standout** — a **giant moray (*Gymnothorax javanicus*-type)** beneath a broad pale *Porites* plate, in the highest-cover frame at the site. **P3180509** holds a large silver jack in open blue water, **compatible with *Caranx ignobilis* or *C. melampygus*, not callable**; **P3180470** a heavy dark grouper with pale saddle bars, **compatible with a dark-phase blacksaddled coralgrouper (*Plectropomus laevis*)** — the same unresolvable animal as BITB27's P3170460. Beyond those: **bluestripe snapper (*Lutjanus kasmira*)** in P3180500, **powder-blue surgeonfish (*Acanthurus leucosternon*)** in P3180508, P3180511 and P3180515, **yellowback fusiliers** compatible with *Caesio xanthonota* in P3180470, P3180503 and P3180506, a **rudderfish school (*Kyphosus* sp.)** in P3180517, and dense **anthias** in P3180503 and P3180504. **P3180508 carries a survey line across the lower-left of the frame** — the only apparatus at this site; **P3180541 is a close "spot" macro** composed for settlement or size rather than cover.

**One taxonomic note.** The CRD site page records that ***Seriatopora* spp. WERE present here and three colonies (ChSe06-08) were sampled**, plus other pocilloporids for a collaborator — one of the few Chagos sites where it was found. **Nothing in the 49 frames unambiguously shows it**: the fine-branched pocilloporid clumps in P3180467, P3180512, P3180514 and P3180519 are compatible in habit, but at this resolution *Seriatopora* is not separable from small *Pocillopora* or *Stylophora*. The field record stands; the photographs neither confirm nor contradict it.

<b>Assemblage composition.</b> **This paragraph rests on the FULL 49-frame set, unlike its Three Brothers neighbours.** Across all 49 the Coralscapes masks put live hard coral at 17.19% of image area, of which massive and meandroid is much the largest component at 53.4%; the rest breaks down as tabular <i>Acropora</i> 21.6%, other hard coral 12.7%, <i>Acropora</i> (non-tabular) 6.0% of live cover. The independent 30-point annotation of the same frames (1470 points) gives live hard coral as 20.3% of all points and apportions it <i>Porites</i> 52.3%, <i>Acropora</i> 36.6%, other hard coral 10.1%, <i>Pocillopora</i> 1.0% of live coral — **the only <i>Porites</i>-dominated assemblage of the redo so far.** Non-coral cover is dominated by turf 14.5% and crustose coralline algae 6.8% of all points. Structural loss registers as dead coral 2.48% of image area and rubble 2.53%. Bleaching stands at 1.36% of annotated points — **20 points**, mostly <i>Acropora</i> 0.95% — against a 7.02% transect figure. Frame-to-frame live cover runs 0% to 60.0% (P3180493 highest-cover, P3180544 poorest). Conspicuous fauna and sessile invertebrates appear as sponges in 29 frames, giant clams in 2 frames, urchins in 2 frames, sea cucumbers in 2 frames.

**Flags:** (1) **VERDICT — THE THREE BROTHERS LOW BIAS IS NOT A REAL COVER DEFICIT, AND IT IS NOT A SAMPLING ARTEFACT EITHER. IT IS A DENOMINATOR ARTEFACT IN THE POINT SET, AND BITB29'S COMPLETE COVERAGE IS WHAT MAKES THAT PROVABLE.** All 49 frames are scored and the point set still reads −9.9 pp low — **but 32.24% of its points are labelled "Unknown", the highest of any site checked, and dropping those from the denominator alone lifts the point set from 20.27% to 29.92% against a 30.2% transect** (30.88% on the 41 non-lobster frames). **The correction is diagnostic, not prescriptive**: across sites it lands on the transect at exactly the three LOW-reading sites (BIEA26 20.3 vs 21.0; BITB27 34.4 vs 36.6; BITB29 29.9 vs 30.2) and blows past it at the HIGH-reading ones (BIEA22 56.1 vs 21.6; BIEA24 24.4 vs 12.7; BIEA20 16.5 vs 10.8). **That asymmetry is the finding: where the point methods read low, an inflated Unknown pool accounts for the whole of it. Do not cite Three Brothers as evidence that photo methods under-read cover.** With ~76% of Chagos point annotations being robot predictions, an Unknown pool this size is an annotation-provenance problem, not a reef property. (2) **BLUE-WATER DILUTION FAILS HERE — the BIEA20 flag-(1) test comes back nearly null.** Dropping the 9 frames ≥40% water moves the point set 20.27→**21.42** and Coralscapes 17.19→**17.38**: the gap narrows by a tenth and does not flip. (3) **SEGMENTATION FAILURE IS REAL BUT INVERTS THE BIEA26 RULE — Coralscapes collapses on LARGE PALE *PORITES*, not on fine mosaics.** Worst cases **P3180474** (points 53.3%, mask 27.6%, 12.97% called "dead") and **P3180493** (60.0% vs 38.0%); on the dense small-colony mosaics the mask is fine or better (P3180514 +9.8, P3180537 +7.8, P3180468 −0.3). **The controlling variable is tissue contrast, not colony size** — and with *Porites* at 52.3% of live points that explains most of the extra 3.1 pp by which the mask trails the point set. (4) **45% OF THE SITE'S BLEACHED POINTS SIT ON EIGHT LOBSTER-DEN MACROS** (9 of 20, on P3180544/545/547/548); **P3180545 carries 3 bleached-*Acropora* points on a frame with no visible *Acropora* at all**, and Coralscapes puts 4.5% bleached area on P3180548 — **the mask is calling the lobster banding bleached coral.** Treat the 1.36% point figure as unusable; 7.02% is this site's only quantitative bleaching statement. (5) **NON-SURVEY IMAGERY IS 20% OF THE FRAME SET (10 of 49) AND IS WORTH ~2 pp.** Dropping the eight lobster macros: 21.79% / 20.17%; all ten (adding P3180541, P3180509): 22.48% / 20.83%; those plus the ≥40%-water frames: 23.65% / 21.24% — **still −6.6 and −9.0 pp.** A smaller lever than at BITB27. (6) **GENUINE LOW-COVER GROUND EXISTS AND SHOULD NOT BE CORRECTED AWAY** — P3180516, P3180517, P3180518, P3180527, P3180540, P3180542, P3180543. Compatible with flag 1: real heterogeneity sets the spread, the Unknown pool sets the offset. (7) **SUBSTRATE RENORMALISATION OVER-CORRECTS BADLY** — excluding water, Unknown and fish gives **46.20%** against a 30.2% transect, as at BITB27 (54.6% vs 36.6%). Excluding Unknown alone is the correction that works, because **much of the Unknown pool IS misfiled water**: 11 frames are scored WATER 0.0% while Coralscapes reads 9–47% dark background on the same pixels (P3180550/551 are 63.3% Unknown with 0% water). BIEA20's flag-(3) water-scoring inconsistency is present and worse here. (8) **P3180504 is the clearest single annotation-provenance artefact** — 14 Unknown and 13 WATER points, 6.7% live, on a frame whose lower half is a dense ledge of knobby *Porites*, tabular *Acropora*, ascidians and hydroids. (9) **P3180508 carries a transect line**; verify at full resolution alongside BIEA20/P3150311 and BITB27/P3170391. (10) **P3180544–P3180553: *Panulirus*-type spiny lobster aggregation, ≥3 animals**, not callable to species; flagged for illustrative use with **P3180548**. (11) ***Seriatopora* was sampled here (ChSe06-08) but is not identifiable in the frames** — **Anderson's judgement needed** on whether the sampled colonies are traceable to a frame, and on whether the pale *Porites* of P3180469/471/474/510/511 includes genuinely paling colonies given the 7.02% figure (not assertable from the imagery). (12) **The edited set is 49 of ~87 exposures** — the CoralSCOP frame-selection confound again. (13) **P3180470's grouper and P3180509's jack are NOT callable to species**; the BIEG04 lesson applies. (14) No per-frame depth data.

**Benthic data:** [Download BITB29 benthic survey data (CSV)](../benthic_data/BITB29_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).