### BIPB66 — Peros Banhos
**CRD page:** [Peros Banhos Coral Surveys (16–17 Apr 2015)](https://coralreefdiagnostics.com/peros-banhos-20150416-to-0417)

**16 Apr 2015**
**GPS:** 5.3813°S, 71.7518°E
**Setting:** intermediate-exposure emergent channel reef (forereef, channel/other lagoon), 23 m
**Coral cover:** 7.0% (2 formal transects, D3–D4; **algal cover 52.1%**) · **published image bleaching:** 6.3%
**Live hard coral cover across methods:** formal transects 7.0% · CoralNet points 9.4% · Coralscapes masks 23.1% (6 frames — see Flag 2). These products have different sampling footprints and denominators and should not be averaged.
**Temperature:** 29.9 °C · **Salinity:** 34.5
**Photo review:** all 13 frames examined — 7 at the top level of `BiPb66-Manta-finished` plus 6 in the nested `BiPb66-CoralNet` subfolder (see Flag 1)
**Representative photos:** P4160559.jpg, P4160583.jpg, P4160576.jpg, P4160574.jpg, P4160584.jpg

BIPB66 is one of the most coral-depleted reefs in the Chagos dataset: **at 7.0% transect coral cover it ranks 5th-lowest of the 109 surveyed sites**, ahead of only BIDI12 (4.2%), BICA01 (4.6%), BIEA17 (4.8%) and BICC36 (5.8%). Only ten genera are recorded at all, and the assemblage is a near-monoculture — ***Porites* holds 4.72% of the benthos, 67.6% of all the coral present**. Everything else is minor: *Leptastrea* 0.75%, *Acropora* 0.40%, *Astreopora* 0.33%, *Pavona* 0.33%, *Montipora* 0.15%, *Millepora* 0.13%, *Pocillopora* 0.10%, and *Favites* and *Goniastrea* at 0.03% each. The non-coral substrate tells the rest of the story: crustose coralline algae is the single largest category at 25.8%, non-coral invertebrates hold 20.6%, bare substrate 20.3%, turf 13.5%, turf+sediment 7.4%, macroalgae 3.6%, erect algae 1.6% and cyanobacteria 0.2% — the "algal cover" figure of 52.1% quoted above is therefore led by CCA and turf rather than fleshy macroalgae.

**Depth structure reinforces the low-cover reading rather than complicating it.** The two transects run D3 (shallower) and D4 (deeper). D3 carries 11.27% coral across six non-zero genera — *Porites* 7.53%, *Leptastrea* 1.00%, *Acropora* 0.80%, *Astreopora* 0.67%, *Pavona* 0.67%, *Millepora* 0.27% — against 63.5% algae, the highest algal reading recorded for this site. D4 drops to just 2.70% coral across only four genera (*Porites* 1.90%, *Leptastrea* 0.50%, *Pocillopora* 0.20%, *Montipora* 0.10%), and — notably — its algal cover also falls, to 40.7%, rather than rising to fill the space vacated by coral. Both coral and algae decline with depth here; whatever occupies the difference at D4 is bare pavement or non-coral invertebrate cover, not algal turf. *Porites* alone drops nearly fourfold between the two transects (7.53% to 1.90%) while remaining the leading genus at both depths.

The photographs corroborate this reading directly rather than revealing a "hidden" high-cover reef the transects missed. P4160559, in the `BiPb66-CoralNet` subfolder, is the clearest wide view: a pale sand-and-rubble plain scattered with small-to-medium massive coral colonies (domal forms consistent with *Porites*), a patch of gorgonian sea fans at upper left, and a dense field of small dark cone-shaped structures peppering the rubble across the mid-ground, consistent with echinoid (sea urchin) tests or spine clusters — not identified to species. P4160565 and P4160566 are dominated less by scleractinian coral than by these same gorgonian sea fans, standing in dense stands across low-relief rubble and pavement, with divers visible only as faint silhouettes in the water column of P4160565; the sea fans (Alcyonacea/gorgonian octocorals, not scleractinian) are not part of the coral-cover accounting even though they visually dominate these two frames. P4160583 is the clearest methodology frame in the set: two divers, one holding a camera rig and the other trailing a PVC quadrat frame laid directly on the substrate, working over exactly this low-relief rubble plain — a direct visual match to the transect methodology rather than a contradiction of the low cover figure. P4160584 is a close frame of a single caespitose/digitate colony (branch form consistent with *Acropora* or *Pocillopora*), dark grey-brown and fully pigmented with no sign of stress.

**Condition is mixed but localized, and the photographic evidence sits between two competing numbers.** The published site mean bleaching is 6.25% (rounded 6.3%, rank 42nd of 109, squarely mid-table), but a separate CoralNet-point recomputation over the 6-frame `BiPb66-CoralNet` subfolder gives 15.0% bleached of 20 coral points scored (3 of 20) — an 8.8-percentage-point gap flagged as a `MISMATCH` in the site's recomputation audit (see Flag 4; per project convention the published figure is treated as authoritative and quoted above, not the recomputation). The imagery is consistent with the higher, patchier reading in at least one instance: P4160576 shows a cauliflower-form colony (growth form consistent with *Pocillopora*) with a distinct white patch across its center against otherwise pigmented brown-purple tissue on the same colony — an unambiguous, colony-scale partial-bleaching signal. Set against that, P4160584's digitate colony from the same subfolder is fully pigmented, so the imagery itself supports a patchy, localized bleaching pattern rather than either a site-wide event or the complete absence implied by treating 6.3% as evenly distributed.

At the top level of the folder, P4160574 and P4160575 show a large, dense school of small planktivorous fish (translucent olive-gold bodies, likely cardinalfish or a related planktivore) sheltering beside a large massive coral colony (consistent with *Porites*) — good fish biomass directly over one of the *Porites* colonies the transects record as this site's dominant genus. The remaining five top-level frames, P4160586 through P4160590, are all the same reef manta ray (*Mobula alfredi*) photographed in open blue water with no benthos in frame at all — the source of the folder's "Manta" name — leaving genuine reef substrate documented in only 8 of the site's 13 photographs.

<b>Assemblage composition.</b> Formal transects give 7.0% coral cover — 5th-lowest of 109 Chagos sites — carried almost entirely by *Porites* (67.6% of the coral present) across just ten genera, against 52.1% "algal" cover that is itself led by CCA (25.8%) rather than fleshy algae, plus 20.6% non-coral invertebrates and 20.3% bare substrate. The photographs document exactly this reef: a rubble-and-sand plain with isolated massive coral colonies, extensive gorgonian sea fan cover, dense echinoid fields, one clearly partially bleached and one fully pigmented digitate colony, a resident fish school over a *Porites* head, a survey team working the plain with a camera and quadrat frame, and — separately — a reef manta ray in the water column above.

**Flags:** (1) **Photo set is manta-heavy.** 5 of 13 site frames (P4160586–P4160590) show only a reef manta ray in open water with no benthic content; genuine reef substrate is documented in just 8 of 13 frames. (2) **Multi-method `n_photos` (6) covers only the `BiPb66-CoralNet` subfolder**, not the full 13-frame site; the CoralNet-point (9.4%) and Coralscapes-mask (23.1%) cover figures rest on fewer than half the site's available images and specifically exclude the top-level "Manta" folder (which is mostly non-benthic imagery anyway). (3) **Multi-method comparison, not to be averaged:** formal transects 7.0% vs CoralNet points 9.4% vs Coralscapes masks 23.1% (n=6 frames) — a pattern consistent with the known tendency of CoralNet/Coralscapes to over-read cover on low-cover Indo-Pacific reefs. (4) **Bleaching data conflict:** the published site mean is 6.25% (used above as the authoritative figure), but a CoralNet-point recomputation over the same 6-frame subfolder gives 15.0% (3 of 20 coral points scored bleached), an 8.8-point gap logged as `MISMATCH` in `chagos_pct_bleached_recomputed.csv`. The published figure is quoted per project convention; the gap is plausible given the very small point count (n=20) behind the recomputation and is corroborated qualitatively by the distinct partial bleaching visible in P4160576. (5) **Localized partial bleaching** visible in P4160576 (cauliflower-form colony, white center patch vs. pigmented margins); P4160584 shows an equivalent-form colony fully pigmented, supporting a patchy rather than site-wide read. (6) **Depth pattern:** both coral and algal cover decline from D3 (11.27% coral / 63.5% algae) to D4 (2.70% coral / 40.7% algae) — the deeper transect is not simply an algae-dominated version of the shallower one; the space vacated by coral at D4 goes to bare pavement / invertebrate substrate rather than algal turf. (7) **Dense echinoid fields and extensive gorgonian sea-fan stands** visible across the rubble in P4160559, P4160565 and P4160566 — the sea fans are soft coral (Alcyonacea), not scleractinian, and are excluded from the coral-cover accounting despite visually dominating two of the six CoralNet-subfolder frames; the echinoid fields are consistent with the site's comparatively high 20.6% non-coral-invertebrate substrate reading. Neither is identified to species from the imagery. (8) **CoralNet point annotations are pipeline-wide ~76% unconfirmed robot output** (only ~600 of ~2,500 corpus images are Anderson-confirmed); the 9.4% CoralNet-points cover figure and the 15.0% recomputed-bleaching figure quoted here should not be read as human-verified. (9) **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.

**Benthic data:** [Download BIPB66 benthic survey data (CSV)](../benthic_data/BIPB66_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).