### BIEA18 — Eagle Island
**Date:** 2015-03-15 (dive window 08:11–09:01, from frame EXIF — the first of three Eagle Island dives that day)
**GPS:** 6.2024°S, 71.4042°E
**Frames reviewed:** 43 of 43 (100% of the site folder, `BiEa18-finished`; the manifest's 44th entry, P3150281, is a filename collision with BIEA20 and does not belong to this site — see the flags)
**Coral cover:** 20.7% (formal transect surveys) · **% bleached:** 11.5% (CoralNet analysis of informal survey images; Mayfield & Dempsey 2026, Oceans)
**Reef type:** bank, exposed · **Temperature:** 29.3 °C
**Live hard coral cover across methods** (n=24 photos): transect 20.7% · CoralNet points 18.2% · Coralscapes 14.7%. Deviation vs transect -6.0 to -2.5 pp. **Mean of 3: 17.9% ± 3.0.** **But the photo columns rest on only 24 of the 43 frames, and on 15 of them once blue water is excluded — at which point the human read RISES to 24.9% and overtakes the transect by +4.2 pp** (Coralscapes 18.4%, −2.3 pp). See the flags.
**Coral genera identified:** 8 scleractinian groups + octocorals — *Acropora* (corymbose, digitate, branching **and broad tabular plates**), *Porites* (massive/domal and digitate-columnar), *Pocillopora* (knobby/verrucose), meandroid brains compatible with *Platygyra*/*Leptoria*, ridged foliose plates compatible with *Pachyseris* (or *Merulina*/*Echinopora*), vase-and-plate colonies compatible with *Turbinaria* or *Pavona*, encrusting sheets compatible with *Montipora*, and a solitary free-living **mushroom coral (*Fungia*-type)**; plus **yellow gorgonian sea fans in at least seven frames** and one lobed colony compatible with a *Sarcophyton*/*Lobophytum* soft coral. Growth-form calls from photographs, given as "compatible with".
**Representative photos:** P3150293.jpg, P3150313.jpg, P3150340.jpg, P3150310.jpg, P3150341.jpg

BIEA18 is an **exposed bank with a real drop-off**, and it is the most three-dimensional site in the redo so far. The dive divides cleanly in two. The **shallower pavement half** (P3150279–P3150286, P3150322–P3150328, P3150342–P3150348) is a low-relief plain of consolidated carbonate, rubble and coralline crust across which coral sits as discrete, well-spaced colonies: corymbose and digitate *Acropora*, knobby *Pocillopora*, domal *Porites*, tan plating tables, and recurring **dark purple-black encrusting patches** — dark turf or cyanobacteria — over broad stretches of bare pavement. P3150347 shows how far that plain runs: widely spaced small colonies to the limit of visibility. The **slope half** (P3150301–P3150317) is a different reef altogether — a wall carrying broad overlapping tabular *Acropora*, ridged foliose plates, digitate thickets and, repeatedly, **yellow gorgonian sea fans**, the first octocorals to feature this prominently at any Chagos site in this redo. **P3150313 is the site's best reef frame**: several large pale tables stacked down the slope, a gorgonian in the foreground, and a haze of orange anthias over all of it.

**The fish assemblage is the strongest of the redo.** Large mixed schools of dark surgeonfish and unicornfish (*Naso*-type, white-margined tails and visible keels) work the slope in P3150285, P3150286, P3150301–P3150308, P3150324, P3150325, P3150326 and P3150348; orange anthias carpet the wall in P3150307–P3150313; **bluefin trevally (*Caranx melampygus*, unmistakable in P3150315 and P3150323)** patrol the blue in fours and fives; and there is an adult **emperor angelfish (*Pomacanthus imperator*)** in P3150282, a yellow-tailed sweetlips compatible with *Plectorhinchus* in P3150326, Moorish idol, butterflyfishes, a large hogfish-type wrasse compatible with *Bodianus*, cleaner wrasse, damselfishes and sergeant majors. **P3150310 carries a pale, blunt-snouted shark, most consistent with a whitetip reef shark (*Triaenodon obesus*)** — the first shark of the Eagle Island block — and P3150347 holds a second, more distant elongate silhouette that is **not callable** from the frame.

**Two invertebrate finds are worth pulling out.** P3150283 is a **tight aggregation of six or more long-spined *Diadema*-type urchins** on open pavement, the second such aggregation in the redo after BIDI10 and, given how much dead consolidated framework this site carries, a bioerosion signal worth recording. And P3150334–P3150340 are **seven consecutive frames of one painted spiny lobster** sheltering under a ledge — ornate black-and-white marbled carapace, banded walking legs, long pale antennae, compatible with ***Panulirus versicolor***. P3150340 is a frame-filling head study and the best invertebrate portrait in the collection. The benthos around the lobster is the site's highest-cover: a large lavender-grey ridged plate compatible with *Pachyseris* directly overhead, brown digitate *Porites* clusters, bright green calcified alga compatible with *Halimeda*, red and orange encrusting sponges, and a brown massive *Porites* carrying **pale, sharply-edged patches of recent partial mortality** — which is where Coralscapes reads its highest dead-coral values at this site (P3150335 6.6%, P3150339 6.5%).

**P3150341 is a methodology frame** and belongs with BIDI08/P3120298 and BIEG04/P3100401 in the record of how the mission's numbers and samples were made: a **gloved hand holding a stainless surgical rongeur** at the substrate, beside two or three smooth pink globular masses that the annotator scored as sponge. Cross-check it against the BIOT *Pocillopora damicornis* and *Seriatopora hystrix* collection permits.

<b>Assemblage composition.</b> Across the 24 analysed frames the human 30-point annotation (720 points) puts live hard coral at 18.2% of all points and bleached hard coral at 1.25% — **9 bleached points in total**, a small enough number that a single misplaced point moves the site figure by more than a percentage point. Coralscapes reads live coral at 14.7% of image area and dead coral highest in the lobster frames. Frame-to-frame live cover runs the full range, 0% on the open-water frames to 40% on P3150293 and P3150340, and the two halves of the dive differ systematically: the 15 frames with less than 40% water average 24.9% live coral by point count, against a site figure of 18.2% that is dragged down by nine water-heavy frames. **The recorded 11.5% bleached cannot be reconstructed from these annotations under any denominator** — 9 bleached points is 1.25% of all points and 6.9% of live-coral points, neither of which is 11.5%; the recorded figure evidently comes from a different or larger image set.

**Flags:** (1) **THE COVER NUMBERS USE 24 OF 43 FRAMES — CAUSE FOUND: A FILENAME COLLISION.** The 20 missing frames are exactly **P3150279–P3150285 and P3150301–P3150313**, the first half of the dive, where the slope, every gorgonian frame, the *Diadema* aggregation, the emperor angelfish and the shark are. **`BiEa20-finished` contains 20 files with those same names**, and the `match_key` in `chagos_human_vs_coralscapes_side_by_side.csv` is the bare lowercase filename with no site qualifier — so the two sets collided and all 20 rows were credited to BIEA20. **EXIF settles the ownership beyond doubt: BIEA18's frames run 08:11:30–09:01:28 and BIEA20's run 14:04:40–14:54:49 on 2015-03-15 — three separate dives that day (BIEA19 sits between them at 10:41–11:11), with no temporal overlap.** The folder is authoritative and correct; the camera simply reused frame numbers. **BIEA18's 20 morning frames were never scored at all** — they are absent from the dataset, not misfiled — so this site's photo columns rest on its afternoon half only. BIEA20 is *not* contaminated: all 41 of its rows read pixels from `BiEa20-finished`. **Fix `match_key` to include the site folder before any re-run**, and always qualify Chagos frame IDs with the site code (same trap as BIDI08/BIDI10). (2) **4 OF THE 9 BLEACHED POINTS SIT ON WATER-HEAVY FRAMES (≥40% water) — 44% of the site's bleached signal.** The worst case is **P3150315: 29 water points and one "bleached *Acropora*" point in a frame that is nothing but blue water and four trevally** — no substrate at all. **P3150324** repeats it: 0% live coral, a dark backlit silhouette where 16 points were scored "unknown", and one bleached-*Acropora* point. That the annotator scored **P3150323 correctly as 30/30 water** shows these are slips, not policy. Same mechanism as BIDI16.5 but four times the impact. (3) **Excluding blue-water frames flips the sign of the photo-vs-transect gap here**: human 18.2% → **24.9%**, against a 20.7% transect. At low-cover sites the water-frame dilution is larger than the tabular-foreshortening bias that dominates BIDI08. (4) **P3150326 and P3150340 are each credited with 2 bleached points and nothing in either frame reads as bleached to me** — 326 is a turf-and-surgeonfish frame whose only pale colony is a normally pigmented yellow *Porites*; 340 is the lobster close-up. **Query these 4 points**; they are 44% of the remaining bleached signal. (5) **P3150281 is in the manifest but absent from BIEA18's folder** — and the P3150281 that does exist, in `BiEa20-finished`, is a **different photograph** (14:35:41, the afternoon dive). The manifest entry is the same collision leaking in. **BIEA18 is a 43-frame site; correct the manifest rather than hunting for a lost file.** (6) ***Diadema* aggregation at P3150283** — log with the BIDI10 aggregation for the bioerosion question. (7) **P3150334–P3150340: painted spiny lobster, 7 frames** — flagged for illustrative use; P3150340 is the portrait. (8) **P3150341 is methodology imagery** (gloved hand + surgical rongeur). (9) **Yellow gorgonians are a defining feature of this site** and are absent from the annotated subset entirely — any octocoral tally built from the point data will read zero here, wrongly. (10) **Whitetip reef shark in P3150310** (hedged); a second elongate silhouette in P3150347 is not callable. (11) No per-frame depth data.

**Benthic data:** [Download BIEA18 benthic survey data (CSV)](../benthic_data/BIEA18_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).