### BITB28 — North Brother
**Date:** 2015-03-18 (dive window 08:21:49–09:18:02, from `DateTimeOriginal` EXIF on all 43 files, in four bursts: 08:21–08:23, 08:28–08:33, 08:42–08:48, 08:55–09:18. The morning dive of the 18th, the day after BITB27 at Middle Brother.)
**GPS:** 6.13840°S, 71.50910°E
**Frames reviewed:** 43 of 43 (100% of the site folder, `BiTb28-finished`; none evicted). **Only 22 of the 43 carry CV data.** Every photo-method number below rests on those 22, not on the 43.
**Coral cover:** 28.5% (formal transect surveys; algal cover 46.18%) · **% bleached:** 11.26% (Mayfield & Dempsey 2026, Oceans) — **thirty-two times BITB27's 0.35%, one day and ~2 km away, and by far the highest bleaching figure in Three Brothers so far.**
**Reef type:** channel reef, intermediate exposure — the CRD site page calls it a **leeward channel reef** · **Temperature:** 29.4 °C
**Live hard coral cover across methods** (n=22 photos of 43): transect 28.5% · CoralNet points 19.5% · Coralscapes 13.8%. Deviation vs transect **−9.0 to −14.7 pp** — both photo methods read LOW, the same direction as BITB27, BITB29, BITB30 and BIEA26. **Mean of 3: 20.6% ± 7.4.** **Unlike BITB27, that gap is NOT a sampling artefact** — see flag 2.
**Coral genera identified:** 8 scleractinian groups plus octocorals — ***Acropora*** in four growth forms (broad tabular plates, corymbose/digitate cushions including a golden *A. humilis*/*gemmifera*-type in P3180447 and P3180457, arborescent staghorn, and low encrusting-branching mats), ***Porites*** in three (massive/domal boulders, knobby-lobate heads, broad pale encrusting-plate sheets), ***Pocillopora***-type knobby clumps, ridged foliose plates compatible with *Echinopora*/*Merulina*/*Pavona*, cups compatible with *Turbinaria*, plocoid and meandroid faviids compatible with *Goniastrea*/*Platygyra*, sheets compatible with *Montipora*, and unresolvable submassives; plus **olive-yellow gorgonian sea fans over sand in P3180398** and pervasive crustose coralline algae. Growth-form calls are from photographs and given as "compatible with"; no species is asserted.
**Representative photos:** P3180408.jpg, P3180431.jpg, P3180462.jpg, P3180394.jpg, P3180440.jpg

**BITB28 is a mosaic site, and that is its defining property.** Where BITB27 one day earlier was a single continuous *Acropora* plate reef, North Brother's leeward channel is a patchwork: tiered gold tabular *Acropora* in the first minutes, then white sand aprons, bare CCA-crusted pavement, dense mauve-and-gold coral gardens, then more sand. **P3180392 is the best image of the good end** — tiered stacks of golden tabular *Acropora* running into the blue under fusiliers and a school of yellow snapper; **P3180390** and **P3180389** repeat it with a diver silhouetted over the plates. **P3180452 is the other end**: a bare pale carbonate plain carrying perhaps five small tables in the whole field. The 28.5% transect figure sits between them and reads as an honest average of a heterogeneous reef.

**The highest-cover ground is in the second half of the dive.** **P3180408 is the best benthic frame and the highest-cover scored one** (43.3% by points, 33.4% by Coralscapes) — grey-lilac digitate *Acropora*, brown *Pocillopora*-type knobs, sheets compatible with *Montipora*, a meandroid brain and a bivalve compatible with *Tridacna*, all under a broad tabular plate. **P3180431** is its equal (46.7% by points, the site maximum): pale plate-form *Porites* left, ridged foliose colonies centre, branching *Acropora* and *Porites* domes right. **P3180440 and P3180461, both unscored, are as good.** **P3180424** is the most atmospheric image — a gold plate and a pale *Porites* dome framing an overhang packed with orange anthias, a giant clam (*Tridacna*) wedged beneath.

**Massive *Porites* is the structural backbone, and the point set says so.** *Porites* takes **45.0% of live coral points** against *Acropora*'s 37.2% — the reverse of BITB27, where *Acropora* held 87.4%. Domes anchor P3180448, P3180454, P3180457, P3180459 and P3180461; pale encrusting plates fill the foreground of P3180395. The difference is not a scoring artefact: *Porites* is just as dominant in the unscored imagery.

**One taxonomic note carries over from the CRD site page**, which records that **neither *Seriatopora* spp. nor *P. acuta*/*P. damicornis* were present at this leeward channel reef, and that other pocilloporids were collected here for a collaborator.** The imagery is consistent: pocilloporid material in P3180408, P3180440 and P3180446 is blunt and knobby, compatible with a *P. verrucosa*-type habit, nothing reads as *Seriatopora*, and **the point set puts *Pocillopora* at only 2.3% of live coral.** A hedged photographic impression, not a survey result.

**Mortality is present, old, and concentrated in the tables.** **P3180445 is the clearest frame** — two broad *Acropora* tables reduced to bare pink-white CCA-plated slabs, no live tissue, living tables and *Porites* domes alongside; Coralscapes reads 8.92% dead and 23.3% of its points are CCA. **P3180455 carries the site's highest dead reading, 11.09%**, over pale turf- and CCA-covered pavement with a pair of yellow rabbitfish (*Siganus* sp.) above it. **P3180440 holds the only stark-white colony in the collection**, unscored and not resolvable as bleaching versus recent death; **P3180395's cream-pale encrusting *Porites*** could as easily be a pale morph.

**The fauna is the best in Three Brothers so far.** **P3180394 and P3180396 hold a whitetip reef shark (*Triaenodon obesus*)** under a ledge, the second close and sharply lit. **P3180462 is the single best fauna frame of the redo so far** — an adult **emperor angelfish (*Pomacanthus imperator*)**, two **powder-blue surgeonfish (*Acanthurus leucosternon*)**, a **palette surgeonfish (*Paracanthurus hepatus*)**, an anemonefish (*Amphiprion* sp.) and a parrotfish in one field. **P3180443 and P3180444 are macro portraits of a moray (*Gymnothorax* sp.)**; **P3180430** is a cave mouth filled with golden sweepers or cardinalfish. Fusiliers compatible with *Caesio*/*Pterocaesio* work the water column in eight frames (P3180389, P3180390, P3180392, P3180397, P3180420, P3180421, P3180422, P3180445), and orange anthias swarm P3180424, P3180459 and P3180460.

<b>Assemblage composition.</b> **This paragraph rests on 22 of the site's 43 frames.** Across those 22 the Coralscapes masks put live hard coral at 13.8% of image area, of which massive and meandroid is much the largest component at 47.4%; the rest breaks down as <i>Acropora</i> (non-tabular) 16.5%, other hard coral 16.5% and tabular <i>Acropora</i> 12.7% of live cover. The independent 30-point annotation of the same frames (660 points) gives live hard coral as 19.5% of all points and apportions it <i>Porites</i> 45.0%, <i>Acropora</i> 37.2%, other hard coral 15.5%, <i>Pocillopora</i> 2.3% of live coral — **a <i>Porites</i>-led assemblage, the inverse of BITB27.** Non-coral cover is dominated by turf 20.5% and crustose coralline algae 3.5% of all points. Structural loss registers as dead coral 2.80% of image area and rubble 5.55%. Bleaching stands at 2.58% of annotated points — **17 points, of which 11 sit on a single bare-sand frame** — against the 11.26% transect figure; Coralscapes reads 0.00% bleached across all 22 frames. Frame-to-frame live cover runs 0% to 46.7% by points (P3180431 highest-cover, P3180430 poorest) and 0.0% to 33.4% by Coralscapes (P3180408 highest-cover, P3180420 poorest). Conspicuous fauna and sessile invertebrates appear as sponges in 17 frames, sea cucumbers in 4 frames, giant clams in 1 frame, urchins in 1 frame, anemones in 1 frame.

**Flags:** (1) **21 OF 43 FRAMES WERE NEVER ANNOTATED, AND THE ANNOTATIONS DO NOT EXIST TO RECOVER.** All 43 are on disk, none evicted; only 22 carry CV data. The CSVs key on the bare lowercase filename with no site qualifier, and **exactly 21 of BITB28's filenames also occur in `BiTb30-finished` — precisely the unscored ones.** Verified: all 21 collided rows carry `model_folder = BiTb30-finished`, so they describe **BITB30's** pixels; the 22 scored are unique to this folder, so nothing is cross-contaminated. **A re-key recovers nothing.** Same failure mode as BITB27, different partner folder. (2) **VERDICT ON THE SUBSET QUESTION — OPPOSITE TO BITB27: THE 21 UNSCORED FRAMES ARE NEITHER HIGHER- NOR LOWER-COVER, AND THE −9 TO −15 pp GAP IS REAL.** They span the full range both ways: highest-cover **P3180440, P3180461, P3180424, P3180394, P3180396, P3180462, P3180447** read 28–38% by eye, matching the best scored (P3180408 43.3%, P3180431 46.7%); poorest **P3180398**, **P3180452**, **P3180421** and the **P3180443**/**P3180444** moray macros read 3–8%, matching the worst scored (P3180407, P3180430, P3180420). **Both halves centre near 20%.** The arithmetic agrees: dropping the two non-benthic frames lifts Coralscapes only 13.8→15.0 and the points 19.5→20.5; dropping the five lowest-scoring frames reaches only 17.4 and 22.4 against 28.5. **BITB28 IS usable as evidence that photo methods under-read cover, where BITB27 is not.** (3) **BLUE-WATER DILUTION IS NOT THE EXPLANATION.** Only **2 of 22 scored frames are ≥40% water** (P3180395 40.0%, P3180420 43.3%; site mean 20.3% vs BITB27's 26.0%). Under BIEA20's flag-(1) filter **the gap narrows by under a point and does not flip**: Coralscapes 13.77→13.89 (−14.7 → −14.6 pp), points 19.55→20.17 (−9.0 → −8.3 pp). At <25% water it reaches 16.7 and 24.6, still 11.8 and 3.9 pp low. (4) **THE 11.26% TRANSECT BLEACHING FIGURE IS CORROBORATED BY NO CV OUTPUT, AND THE POINT SET'S 2.58% IS ITSELF ONE BAD FRAME.** Of 17 bleached points in 660, **11 sit on P3180407 — essentially bare white sand, no coral in the foreground.** Coralscapes reads 0.03% live and 0.00% bleached there; the point set reads 20% live and 36.7% bleached *Acropora*. **The clearest annotation-provenance artefact in the Chagos set so far** (~76% of these annotations are CoralNet robot output). Removing it drops site bleaching to 0.91%; **P3180420** and **P3180430** are weaker versions. **Anderson's judgement needed on whether the published 11.26% and the photo record can be reconciled at all.** (5) **METHODOLOGY IMAGERY, THREE FRAMES, ALL SCORED — BUT UNLIKE BITB27 IT BIASES NOTHING.** **P3180406** carries a yellow **transect tape** with a second line at left, **P3180408** a tape down the right edge, **P3180409** a tape plus a **diver and camera housing**: 14% of the CV sample against BITB27's 38%, and **P3180408 is the second-highest-cover scored frame** — dropping all three makes the gap WORSE (13.8→12.8, 19.5→18.4). (6) **32.3% OF ALL POINTS ARE "UNKNOWN"** (6.7–56.7%). Renormalising live coral over labelled benthic points gives **42.2%, far above the transect** — the Unknown pool, not water, is the dominant lever here. (7) **P3180440's stark-white colony (unscored) and P3180395's pale cream encrusting *Porites*** are the only bleaching-relevant images, and neither resolves bleaching versus pale morph; **P3180445 (8.92% dead) and P3180455 (11.09%, site maximum)** read as old turnover. (8) **Whitetip reef shark in P3180394 and P3180396, both unscored** — P3180396 flagged for illustrative use with **P3180462**, **P3180408** and **P3180424**. (9) **Gorgonian sea fans in P3180398** — octocorals are structurally present and invisible to a hard-coral point scheme. (10) **No COTS, no disease lesions and no predation scars in any of the 43 frames.** (11) No per-frame depth data.

**Benthic data:** [Download BITB28 benthic survey data (CSV)](../benthic_data/BITB28_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).