### BIPB64 — Peros Banhos
**CRD page:** [Peros Banhos Coral Surveys (Apr–May 2015)](https://coralreefdiagnostics.com/peros-banhos-)

**15 Apr 2015**
**GPS:** 5.46244°S, 71.82438°E
**Setting:** intermediate-exposure emergent fringing forereef, 27 m
**Coral cover:** 31.2% (3 formal transects, D2–D4; **algal cover 42.7%**) · **published image bleaching:** 9.5%
**Live hard coral cover across methods:** formal transects 31.2% · CoralNet points 14.6% · Coralscapes masks 16.8% (21 frames). These products have different sampling footprints and denominators and should not be averaged.
**Temperature:** 30 °C · **Salinity:** 34.3
**Photo review:** all 24 frames examined — 3 in the parent `BiPb64-finished` folder plus 21 in the nested `BiPb64-CoralNet` subfolder (see Flag 1)
**Representative photos:** P4150667.jpg, P4150688.jpg, P4150690.jpg, P4150674.jpg, P4150691.jpg

BIPB64 sits below the Chagos median for coral cover — **31.2% ranks 74th of 109 sites** — against a higher 42.7% algal cover, and its 9.5% published bleaching figure ranks 66th of 98 sites with a value, placing it in the upper third of the archipelago for recorded bleaching. Non-coral invertebrate cover is comparatively high at 20.3% (19th of 109), which the photographs help explain: several frames are dominated by carpets of anemone tentacles rather than coral. CCA cover is 19.3% (29th of 109) and turf 13.0% (30th of 109), with bare substrate low at 5.9% (88th of 109) — this is a well-covered, structurally busy reef even where coral itself is not dominant.

Twenty-five genera are non-zero, more than at some higher-cover Chagos sites, but no single genus dominates. **Porites leads at 8.95% of the benthos, followed by Acropora at 6.04%, Pocillopora at 5.00% and Montipora at 3.74%** — together just under 70% of all coral recorded. *Platygyra* (1.44%), *Favites* (0.84%) and *Pavona* (0.76%) form a second tier. Notably, ***Ctenella* is recorded at 0.72%** — this is *Ctenella chagius*, a monotypic brain-coral genus endemic to the Chagos Archipelago, and its presence here (fourth-highest reading among the "rare" genera at this site) is a small but genuine biogeographic note (Flag 2). *Leptastrea* (0.71%), *Favia* (0.68%) and *Heliopora* (0.47%, the blue coral, re-labelled per the HELI code note in the benthic CSV header) round out the top ten, with *Goniastrea*, *Stylophora*, *Astreopora*, *Psammocora* and a long tail of thirteen further genera each under 0.25%.

**Depth structure is the clearest signal in the transect data, and it runs in the opposite direction from BIPB62's zonation: coral cover falls and algae rises steadily with depth.** D2 (shallowest) carries 35.6% coral against 37.0% algae, led by *Porites* at 12.63% and *Pocillopora* at 7.83% — together 58% of the coral on that transect — with *Acropora* a distant third at 3.87% and *Ctenella* recorded at 2.17%, its highest reading on any transect at this site. D3 inverts the lead genus: *Acropora* rises to 11.16% and edges out *Porites* (10.32%) for the first time, while cover holds nearly steady at 33.3% and algae climbs to 41.4%. D4, the deepest transect, drops to 24.6% coral against a full 49.6% algae — coral has essentially been halved from D2 while algae has grown by a third — and the assemblage flattens out into a more even spread led by *Montipora* (5.40%), *Pocillopora* (4.00%), *Porites* (3.90%), *Acropora* (3.10%), *Favites* and *Pavona* (1.70% each). No genus holds more than 22% of the coral present at D4, versus 35% for *Porites* at D2. **The site-level figures of 8.95% *Porites* and 6.04% *Acropora* therefore blur a real gradient**: a *Porites*/*Pocillopora* shallow zone giving way through an *Acropora*-*Porites* mid zone to a diffuse, algae-dominated deep zone.

The photographs largely confirm a moderate-to-good-cover, structurally complex forereef with patches of open sand and rubble. P4150658, P4150659, P4150666, P4150674, P4150691, P4150692, P4150693, P4150694 and P4150695 show dense, high-relief assemblages: tan-brown massive Porites-type domes, encrusting Montipora-type plates, digitate Pocillopora/Acropora-type thickets, foliose colonies and brain-coral heads (consistent with *Platygyra*/*Favites*), several draped in schooling anthias. P4150674 in particular shows a large tan massive colony (consistent with *Porites*) with a pink-purple digitate cluster at its base, matching the *Porites*-led D2 signature. Against this, P4150672, P4150682, P4150683, P4150684, P4150685, P4150686 and P4150687 show a contrasting low-cover sand-and-rubble pavement zone — scattered small coral heads on bleached-white carbonate rubble, home to a black triggerfish (consistent with *Melichthys niger* or *Odonus niger*) and a dark filefish (consistent with *Aluterus scriptus*) rather than dense coral growth. This two-texture pattern in the imagery — dense coral thickets alternating with open rubble flats — is a plausible photographic counterpart to the algae-heavy, lower-cover D4 transect, though the two cannot be linked frame-to-transect with the data at hand.

**Condition: the published 9.5% bleaching figure is corroborated, and in places exceeded, by the imagery.** P4150667 shows two conspicuous white tabular colonies (table-forming, consistent with *Acropora*) growing directly beside a pigmented tan massive colony — intact-tissue pallor, not a lighting artefact. P4150688 is more dramatic still: a large table coral photographed from above is entirely white to pale lavender-blue across its whole surface, with no pigmented tissue visible — one of the clearest single-colony bleaching records in this photo set. Both frames sit within the CoralNet subfolder, i.e. within the imagery scored by the multi-method comparison, so the bleaching seen here is plausibly part of what depresses the CoralNet (14.6%) and Coralscapes (16.8%) live-cover readings well below the 31.2% transect figure — dead white coral skeleton reads as non-living substrate to point- and mask-based classifiers even when transect protocols may still log the colony as coral. This is consistent with the site's above-median bleaching rank and is flagged rather than asserted as the explanation (Flag 3).

Fauna recorded in the imagery include a pair of large anemones (consistent with *Heteractis magnifica*) hosting anemonefish at the top-level frames P4150669 and P4150670, an oriental sweetlips (*Plectorhinchus vittatus*-consistent) sheltering under a coral ledge beside an orange-lined triggerfish (*Balistapus undulatus*-consistent) in P4150690, powder-blue tangs, butterflyfish and abundant anthias schools throughout the higher-cover frames, and the black triggerfish/filefish pair on the sand-rubble flat described above.

<b>Assemblage composition.</b> Formal transects give 31.2% coral against a higher 42.7% algae, with coral led by *Porites* (8.95%), *Acropora* (6.04%), *Pocillopora* (5.00%) and *Montipora* (3.74%) across 25 genera, and a real depth gradient from a *Porites*/*Pocillopora* shallow zone to a diffuse, algae-dominated deep zone. The photographs document both a dense, high-relief coral community and an alternating open rubble-and-sand pavement, plus two unambiguous colony-scale bleaching records that track the site's above-median published bleaching figure.

**Flags:**
1. **Nesting trap confirmed and resolved:** the photo folder holds 24 frames total, not the 21 recorded in `chagos_multimethod_cover_by_site.csv`'s `n_photos` field — 3 additional frames (P4150669, P4150670, P4150690) sit in the parent `BiPb64-finished` folder rather than the nested `BiPb64-CoralNet` subfolder that the multi-method pipeline scored. All 24 were reviewed for this shard; the multi-method percentages (CoralNet 14.6%, Coralscapes 16.8%) reflect only the 21 subfolder frames. The 3 excluded frames are close-up anemone/fish shots not well suited to benthic point-counting, which may explain their omission from the CV pipeline, but this has not been confirmed.
2. **CNEL genus code:** `sitefacts.py`'s internal lookup table maps CNEL to *Cynarina*, but the canonical `benthic_data/BIPB64_benthic.csv` (and the equivalent file for every other Chagos site checked) labels it *Ctenella*. This shard follows the benthic CSV: *Ctenella chagius* is a Chagos-endemic monotypic brain-coral genus, so its 0.72% site-level reading (2.17% at D2) is a genuine, notable biogeographic record, not a data-entry artefact. Do not use "Cynarina" for this code at BIPB64 or other Chagos sites — check the CSV, not the script's internal dictionary.
3. **Bleaching in the imagery is at least as prominent as the published 9.5% figure suggests.** P4150667 and especially P4150688 show unambiguous, colony-scale white/pale tissue on tabular corals within the CoralNet-scored subfolder. Given that dead/bleached coral skeleton would be scored as non-coral substrate by CoralNet points and Coralscapes masks, this is a plausible (not confirmed) partial explanation for why both CV methods read roughly half the transect-based cover figure. The transect cover value is taken as published; the relationship between the visible bleaching and the CV cover gap has not been formally tested.
4. **Depth-transect coverage is shallow relative to the site's listed depth.** The site-level record lists a 27 m depth, but the three formal transects run D2–D4 only, all evidently well above 27 m based on their cover/algae profile continuity with typical Chagos forereef shallow-to-mid zonation; no deep transect is available to characterize the reef near the listed maximum depth.
5. **Two structurally distinct habitats co-occur in the imagery** — dense coral-thicket frames and open sand/rubble-pavement frames with reef fish but little live coral — and cannot be assigned to specific transect depths from the data available. Treat any single site-level photo as representative of only part of this reef, not the whole.
6. **The published bleaching figure is a COMPUTER-VISION product, not an in-situ observation.** It comes from CoralNet annotation of the informal survey images (Mayfield & Dempsey 2026, *Oceans*). Per-image human-vs-model agreement is substantially weaker for bleaching than for cover: r = 0.40 for bleached against r = 0.72 for live hard coral (`CHAGOS_MULTIMETHOD_README.md`). Treat the site bleaching percentage as the weaker of the two published numbers, and never read it as a diver's field estimate.
7. **NON-CANONICAL BLEACHING RANK — added 2026-08-26.** Any "of 98" bleaching rank above was computed off the site table's `Mean(%bleaching)` column, which equals the **1st** CoralNet training run. Anderson's decision of 2026-08-26 (HOW_TO rule 4e) makes **`% bleaching (2nd training)` canonical**, The file carries TWO pools for it, and they are not interchangeable: a **110-row pool** ranking every dive (revisit dives included, so BIPB70B/72B/69C compete) and a **99-site pool** ranking one value per site. The rank quoted below is on the 110-row dive pool. For this site the canonical value is **9.52%, rank 41 of 110** (1st training: 9.52%, rank 40 of 110). The two runs agree to <0.05 pp on 100 of 110 rows and diverge only on the bleaching-onset dives, so for most sites the distinction changes the rank but not the number.

**Benthic data:** [Download BIPB64 benthic survey data (CSV)](../benthic_data/BIPB64_benthic.csv) — scleractinian cover by genus with full names, plus substrate and algal categories, from the LOF BIOT-1 survey (Mayfield & Dempsey 2026, *Oceans*, supplemental file S1).