### BIEA23 — Eagle Island
**Date:** 2015-03-16 (dive window 14:31–15:19, from frame EXIF — one continuous 48-minute sequence, blue water first, benthos last)
**GPS:** 6.24221°S, 71.29063°E
**Frames reviewed:** 26 of 26 (100% of the site folder, `BiEa23-finished`)
**Coral cover:** 24.6% (formal transect surveys) · **% bleached:** 5.8% (CoralNet analysis of informal survey images; Mayfield & Dempsey 2026, Oceans)
**Reef type:** fringing, intermediate exposure · **Temperature:** 29.2 °C
**Live hard coral cover across methods** (n=26 photos): transect 24.6% · CoralNet points 24.7% · Coralscapes 24.8%. Deviation vs transect +0.1 to +0.2 pp. **Mean of 3: 24.7% ± 0.1 — the tightest three-method agreement in the redo, tighter even than BIEA20's ±1.1. It does not survive the benthic-frame test.** Seven of the 26 frames are ≥40% water; **drop them and the photo methods rise to 31.4% and 32.3% against the same 24.6% transect — +6.8 and +7.7 pp, both now reading HIGH.** See flag (1): what is real here is that the two photo methods agree with *each other*, not that all three converge.
**Coral genera identified:** 7 scleractinian groups plus an octocoral and an antipatharian order — *Acropora* in four growth forms (tables in P3160424 and P3160420, corymbose/digitate clusters in P3160419 and P3160421, open staghorn-like branches in P3160421, fine-corallite plates in P3160405), *Porites* (massive domes throughout, plus orange-brown nodular-encrusting sheets in P3160417 and P3160422), *Pocillopora* (blunt maroon-brown branching in P3160390, P3160419, P3160421), meandroid brains compatible with *Platygyra*/*Leptoria* (P3160384, P3160420, P3160421, P3160424), ridged and vermiculate foliose plates compatible with *Echinopora*/*Pachyseris* (P3160397, P3160407, P3160417), laminar and encrusting sheets compatible with *Montipora*/*Turbinaria* (P3160416, P3160422, P3160423), and small submassives not resolvable to genus; plus **a gorgonian sea fan in P3160377** and **antipatharian wire corals in P3160375–P3160377**, with olive algal tufts compatible with *Halimeda* in P3160417 and P3160424. Growth-form calls are from photographs, given as "compatible with".
**Representative photos:** P3160375.jpg, P3160387.jpg, P3160405.jpg, P3160407.jpg, P3160421.jpg

**BIEA23 is a dive of two halves, and the split is what makes the numbers behave.** The first fifteen minutes (P3160373–P3160390) are largely water column: schools of dark unicornfish and surgeonfish compatible with *Naso* banked over the reef in P3160373 and P3160374, then wave after wave of **yellowtail and yellow-striped fusiliers compatible with *Caesio* and *Pterocaesio*** in P3160376, P3160377, P3160384–P3160386 and P3160388–P3160390. From 14:57 the camera drops onto the substrate and stays there — P3160405, P3160407 and P3160416–P3160424 are all benthic. **All seven water-heavy frames fall in the first half, and they are what pull the photo estimates down onto the transect figure.**

**Where the camera is actually on the reef, this is a genuinely good Chagos site.** P3160406 is the highest-cover wide frame: cream and tan massive *Porites* domes with pitted surfaces, lavender-purple digitate *Acropora*, laminar plates and two **Moorish idols (*Zanclus cornutus*)** working the tops. P3160416 holds a brown corymbose *Acropora* thicket against cream *Porites* plates; P3160421 sets three open *Acropora* tables and a maroon *Pocillopora*-type colony against a tan meandroid brain; P3160424 has two clean tables with blue-green damselfish over them. Point-count live cover in these frames runs 53–80% (P3160405 highest at 80.0%) against 0% in the blue-water frames — **a 0–80% range that is itself the argument against averaging an unfiltered frame set.**

**The macro sequence is the best coral portraiture at Eagle Island.** **P3160407 is the finest frame** — a near-circular plating colony, deep chocolate-brown with dense fine white-tipped corallites and a clean pale growing margin on bare white pavement, compatible with *Echinopora* or *Montipora*. **P3160422 and P3160423** are close studies of tan-and-cream nodular-foliose colonies with dark maroon tissue in the valleys and pink coralline algae at their edges. **P3160405** carries the old entry's bleaching claim: an olive knobby colony left, beside it a large fine-corallite plate grading from tan to broad pale grey-white — see flag (3).

**Macrofauna is modest but includes one clean record.** **P3160375 is the shark frame** — a **whitetip reef shark (*Triaenodon obesus*)** over a rubble-strewn sand plain, the white first-dorsal tip plainly visible and the animal unambiguous; a dark parrotfish compatible with *Chlorurus*/*Scarus* forages on rubble in the same frame. **P3160398** holds a large triggerfish in mid-blue compatible with *Balistoides viridescens*, and **P3160388 a giant clam (*Tridacna*)**, fluted shell open beneath the fusilier school. Beyond those: a yellow tang compatible with *Zebrasoma* in P3160417, bream- or snapper-type fish in P3160387 and P3160419, blue-green chromis in three frames, and Moorish idols in three.

**Structural loss is present but low-key** — no single catastrophic frame of the kind BIEA20's P3150281 provided. The pattern is broad pale carbonate pavement between coral patches: P3160417, P3160418, P3160419 and P3160424 are largely bare, turf-filmed rock, and turf is the largest non-coral category (17.2% of all points, peaking at 46.7% in P3160424). Coralscapes puts dead coral highest in P3160397 (8.81%) and P3160421 (6.59%), both showing bare pavement rather than freshly killed colonies.

<b>Assemblage composition.</b> Across 26 analysed frames the Coralscapes masks put live hard coral at 24.8% of image area, of which other hard coral is the largest component at 50%; the rest breaks down as massive and meandroid 25.4%, tabular <i>Acropora</i> 10.2%, <i>Acropora</i> (non-tabular) 9.3% of live cover. The independent 30-point human annotation of the same frames (780 points) gives live hard coral as 24.7% of all points and apportions it <i>Acropora</i> 46.1%, <i>Porites</i> 31.1%, other hard coral 21.2%, <i>Pocillopora</i> 1.6% of live coral. Non-coral cover is dominated by crustose coralline algae 4.5%, turf 17.2% of all points. Structural loss registers as dead coral 1.9% of image area and rubble 4.9%. Bleaching stands at 1.4% of annotated points — **11 points in total** — concentrated in other hard coral 0.8%, <i>Acropora</i> 0.4%, <i>Pocillopora</i> 0.3% of points. Frame-to-frame live cover runs 0% to 80.0%. Conspicuous fauna and sessile invertebrates appear as sponges in 13 frames, sea cucumbers in 3 frames, giant clams in 2 frames, anemones in 1 frame, seagrass in 1 frame, crown-of-thorns in 1 frame.

**Flags:** (1) **THE HEADLINE ±0.1 pp AGREEMENT IS AN ARTEFACT — AND THE CLEANEST DEMONSTRATION OF THE MECHANISM IN THE REDO.** 7 of 26 frames are ≥40% water, all in the first half of the dive. Strip water progressively and both photo methods climb monotonically away from the transect together: **<60% water (n=24) 26.8 / 26.9 · <50% (n=22) 28.3 / 28.7 · <40% (n=19) 31.4 / 32.3 · <30% (n=14) 37.1 / 38.6**, against a fixed 24.6% transect. At every threshold CoralNet and Coralscapes stay within 1.5 pp of each other (per-frame r = 0.952, mean absolute per-frame difference 5.8 pp). **So the defensible positive result at BIEA23 is convergence between the two PHOTO methods — not three-way convergence with the transect.** The full-set 24.7 ± 0.1 is blue-water dilution cancelling a genuine high bias — the BIEA20/BIEA18 mechanism, with a larger sign-flip. **Do not cite BIEA23 as evidence that photo and transect methods agree.** (2) **THE TWO WATER SCORES DISAGREE, SO THE FILTER ITSELF IS UNSTABLE.** Human WATER points and Coralscapes `background_dark` correlate only r = 0.73 and differ by 12 pp on average (29.6% vs 41.9%); **P3160373 is 23.3% water by points but 56.7% by model.** Switching columns moves the benthic set from n=19 to n=12 and the estimates to 35.3/38.9. Any published filter must state which water score it used. (3) **P3160405 — THE OLD ENTRY'S BLEACHING CLAIM IS NOT SUPPORTED BY THE ANNOTATION.** The pale grey-white surface is real, but the annotator placed 16 live *Acropora* points and only 1 bleached point on that colony; the paling is equally consistent with a pale morph, strobe overexposure, or sediment/algal film. **The actual bleaching-signal frame is P3160423, carrying 7 of the site's 11 bleached points (23.3% of its own).** Substitute it for P3160405, and hedge both. (4) **CORALSCAPES SEES ESSENTIALLY NO BLEACHING ANYWHERE (max 0.016% of any frame) WHILE THE POINT COUNT SEES 1.4%** — the two methods that agree so well on live cover do not agree at all on bleaching (region-wide r = 0.40), and neither reproduces the 5.8% reef-wide figure. (5) **THE COTS DETECTION IN P3160419 IS ALMOST CERTAINLY A FALSE POSITIVE** — 0.01% of frame, a few pixels; the only candidate reads as an urchin, not *Acanthaster*. **Log as model artefact, not a sighting.** (6) **METHODOLOGY IMAGERY, FIVE FRAMES. P3160387 is the important one: a black-and-white banded line crosses the reef with a long dark bar beside it — almost certainly the transect tape and a marker/scale bar.** Divers with housings appear in P3160386, P3160389, P3160397 and P3160398. **Verify P3160387 at full resolution before calling it a tape.** (7) **THE OLD ENTRY'S "GREY REEF SHARK" AND "MUSHROOM CORAL COMPATIBLE WITH *FUNGIA*" ARE BOTH UNSUPPORTED** — the only shark in 26 frames is a whitetip (P3160375), and no fungiid appears anywhere. (8) **HUMAN POINTS RECORD ZERO SPONGE IN ALL 26 FRAMES WHILE CORALSCAPES REPORTS SPONGES IN 13**; the "seagrass" frame is most likely *Halimeda*. (9) **"Unknown" runs 20–37% of points in 12 frames** (≥30% in seven), diluting live-cover-as-%-of-all-points — an independent reason not to read 24.7% as precise. (10) **Antipatharians and a gorgonian (P3160375–P3160377) are invisible to a hard-coral point scheme** — the BIEA18/BIEA20 blind spot again. (11) No per-frame depth data.

**Benthic data:** [Download BIEA23 benthic survey data (CSV)](../benthic_data/BIEA23_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).