### BICC35 — Central Chagos
**CRD page:** [Three Brothers & Central Chagos Coral Surveys (Mar 17-20, 2015)](https://coralreefdiagnostics.com/three-brotherscentral-chagos-20150317-to-0320)

**20 Mar 2015**
**GPS:** 6.18385°S, 71.64028°E
**Setting:** intermediate-exposure submerged lagoonal bank reef, forereef aspect outside the lagoon, 29.3 m
**Coral cover:** 13.8% (13 formal transects across 2 depth strata, D3–D4; **algal cover 53.9%**) · **published image bleaching:** 9.0%
**Live hard coral cover across methods:** formal transects 13.75% · CoralNet points 9.1% · Coralscapes masks 9.5% (19 frames). These products have different sampling footprints and denominators and should not be averaged.
**Temperature:** 29.5 °C · **Salinity:** 34.4
**Photo review:** all 19 of 19 frames examined (single dive, 20 Mar 2015, 12:22–13:04 UTC by file `DateTimeOriginal`; folder `BiTb35-finished` still carries the old Three Brothers code)
**Representative photos:** P3200420.jpg, P3200408.jpg, P3200409.jpg, P3200417.jpg, P3200387.jpg

BICC35 is a submerged lagoonal bank reef in Central Chagos at 29.3 m, forereef aspect but outside the lagoon, under intermediate exposure — and it is **one of the lower-cover sites in the archipelago: 13.8% coral against 53.9% algae, ranking 100th of 109 Chagos sites for coral cover and 16th of 109 for algal cover.** What makes it unusual is not the low cover but the *composition* of the ground it sits on. **Crustose coralline algae (PCCATS) is only 3.78% of total benthos — 106th of 109 sites, among the very lowest CCA readings anywhere in the Chagos dataset** — and fleshy turf (PTTS) is likewise near the bottom at 4.43% (97th of 109). Instead, the substrate is dominated by turf-plus-sediment (PTSTS) at 25.6%, bare substrate (PB) at 22.4%, and **erect algae (PETS) at a striking 18.3%** — the largest single algal component. Macroalgae (PMTS) adds 1.75% and non-coral invertebrates (PITS) 9.9%. Coral, algae, invertebrates and bare rock sum to exactly 100.0%. **This is a sediment-loaded, weakly cemented reef, not a hard coralline pavement:** the near-absence of CCA alongside a quarter of the benthos in turf-and-sediment and another fifth bare is the opposite of the pattern at the well-grazed forereef sites.

Genus richness is below average: **18 nonzero genera, 82nd of 109 Chagos sites**, and the 18 genus percentages sum to 13.7500% of the benthos, an exact match to the published cover figure. All values below are percent of total benthos, not percent of coral. **Unusually for Chagos, *Acropora* does not lead**: *Porites* is first at 3.78% of the benthos (28% of all the coral at the site), then *Pocillopora* 2.83% (21%), then *Acropora* only third at 2.23% (16%), with *Favia* 1.03% (8%) and *Montipora* 0.77% (6%) behind; those five hold 77% of the coral recorded. *Turbinaria* (0.67%), *Platygyra* (0.57%), *Favites* (0.52%) and *Echinopora* (0.30%) follow, then a tail of *Acanthastrea*, *Astreopora* and *Pavona* (0.20% each), *Leptastrea* (0.13%), *Galaxea* and *Goniopora* (0.10%), *Goniastrea* (0.05%), and *Cantharellus* and *Pachyseris* (0.03% each). A stress-tolerant, sediment-tolerant lead pair of *Porites* and *Pocillopora* over an *Acropora* third place is a coherent match to the sediment-heavy substrate described above — though the transect and substrate data are independent measurements and no causal link is tested here.

**The two depth strata are the interesting result at this site, because coral cover is essentially flat between them while everything around the coral changes completely.** D3 (10 transects, the shallower stratum) carries 13.3% coral, and D4 (3 transects) carries 14.2% — a difference of under one percentage point. **But algal cover falls from 67.8% at D3 to 40.0% at D4, and the whole algal budget restructures with it:** turf-plus-sediment drops from 36.4% to 14.9%, erect algae from 21.2% to 15.4%, macroalgae from 3.3% to 0.2%, while crustose coralline algae — negligible at both — more than doubles from 2.1% to 5.5%. Bare and unaccounted ground rises correspondingly: coral plus algae account for 81.1% of the benthos at D3 but only 54.2% at D4, leaving roughly 46% of the deeper stratum as bare pavement, sand or non-coral invertebrate cover that this depth-level summary does not itemise. **The genus order also inverts between strata.** At D3 *Acropora* leads at 3.80% of the benthos (29% of that transect's coral), ahead of *Porites* 2.70%, *Pocillopora* 2.60%, *Montipora* 1.20%, *Echinopora* 0.60% and *Favites* 0.50%. At D4 ***Acropora* collapses to 0.67% and drops to sixth place**, while *Porites* takes the lead at 4.87% (34% of that transect's coral), followed by *Pocillopora* 3.07%, *Favia* 1.67%, *Turbinaria* 1.33% and *Platygyra* 1.13%. **The site-level figure of 2.23% *Acropora* therefore describes neither stratum**: it averages a shallower station where *Acropora* is the leading genus against a deeper one where it is a trace taxon and the reef has handed over to *Porites*, *Pocillopora* and faviid genera. The D4 result rests on 3 transects and is less well replicated than D3's 10.

The photographs — all 19 frames — corroborate a low-cover, sediment-dusted bank reef with high fish biomass, and they are consistent with the transect picture rather than flattering to it. The wide shots (P3200388, P3200389, P3200416, P3200417, P3200419) show broken, pale limestone pavement and rubble carrying a discontinuous scatter of small colonies — knobby clumps consistent with *Pocillopora*, low mounds and columnar forms consistent with *Porites*, small cup and funnel forms consistent with *Turbinaria*, and occasional plate and digitate colonies consistent with *Acropora* — with most of the surface reading as green-brown turf, fine sediment and pale bare rock rather than as living tissue. **P3200420 is the structural exception and the site's most impressive single coral: a tabular colony well over a metre across, olive-tan and apparently healthy, dominating an otherwise sparse pavement** — a good illustration of why the shallower stratum's *Acropora* figure is what it is despite the low site cover. P3200387 and P3200407 are close-range frames of the substrate, showing exactly the mixture the numbers describe: crimson and orange encrusting sponges, dark green fleshy and calcified erect algal tufts, a fine sediment dusting over everything, and only small isolated coral colonies — with **very little of the pink-purple coralline crust that dominates most Chagos forereefs**, which is precisely what the 106th-of-109 CCA rank predicts. Water clarity is good and the light is dark blue, consistent with the 29.3 m survey depth.

**Condition is mixed and the photographs record two clear pallor observations.** The published image-based bleaching value is **9.0%, ranking 36th of 98 Chagos sites carrying a value — above the median but outside the worst quartile.** Two frames carry unambiguous, colony-scale evidence. **P3200387 shows a small branching colony that is uniformly chalk-white across every branch, sitting on turf-and-sediment ground with fully pigmented colonies in the same frame** — within-frame contrast that exposure or strobe artefact cannot explain. **P3200418 shows a second, larger *Pocillopora*-type clump in which roughly half the branch tips are bright white against olive-brown tissue on the same colony** — the partial, tip-first pattern typical of an early or recovering bleaching response, and directly adjacent to a large, evenly pale-grey massive colony consistent with *Porites*. Because a single still cannot separate a freshly bleached colony from a very recently dead one, both are recorded as **frame-level bleaching-or-recent-mortality observations, not as a recalculated site percentage**. No disease lesion or tissue-loss margin, no *Drupella* aggregation, no crown-of-thorns starfish and no predation scarring was identified in any of the 19 frames.

Fauna is the standout at this site and is dominated by large, mobile animals rather than by the benthos. **Two grey reef sharks cruise the pavement together in P3200408**, well lit and unambiguous. **A hawksbill-type sea turtle rests wedged under a ledge in P3200409 and P3200410**, photographed from two angles with the carapace scute pattern and beaked head clearly visible (species hedged from stills). Five consecutive frames — P3200392, P3200394, P3200395, P3200396 and P3200406 — document a **large, dense aggregation of surgeonfish and unicornfish-type acanthurids**, hundreds of individuals silhouetted against the surface over sand and low reef, with gorgonian sea fans, sea whips and antipatharian bushes visible on the substrate beneath them; a distant shark-shaped silhouette also appears at the right edge of P3200394 (hedged, not confirmed). P3200391 is a close pass of a fusilier school, P3200421 a single deep-bodied fish silhouetted in open blue water near the surface, and P3200389, P3200416 and P3200417 show dense clouds of small blue-and-yellow damsels and anthias over the reef. **A reef supporting sharks, turtles and an acanthurid aggregation of this size on 13.8% coral cover is a useful reminder that fish biomass and benthic cover are not the same measurement.**

**BICC35 is a flat, algae-dominated carbonate plain with a very large fish population and a handful of outsized corals on it.** **P3200416 states the site better than any number does** — a pale rubble-and-pavement flat running out to a hazy horizon with a cloud of small blue anthias hanging over it and coral reduced to widely spaced knuckles of brown *Pocillopora*-type and small domes. **P3200388, P3200389 and P3200419** are the same ground at other angles. Algal cover of 53.9% against 13.75% coral gives a coral:algae ratio of **0.26**, and the algae is mostly not the conspicuous kind: **turf-with-sediment is 25.6% and erect algae 18.3%**, against turf 4.4%, CCA 3.8%, macroalgae 1.8% and cyanobacteria zero. **P3200407 is the frame that makes that visible** — the dive's one flash macro, a cryptic ledge of crimson encrusting sponge, ascidian buttons and a pale *Pocillopora*-type knob, everything under a grey film of carbonate sediment. That film, not fleshy weed, is what "54% algal cover" means here.

**Two corals carry a disproportionate share of the site.** **P3200420 is the signature frame**: one **tabular *Acropora* three to four metres across**, olive-gold, laid flat over the pavement with old holes through the middle and a pale abraded streak, being worked by a dark parrotfish compatible with *Chlorurus*, a *Chaetodon*-type butterflyfish, a maroon-and-white wrasse compatible with *Bodianus*, and a thicklip wrasse compatible with *Hemigymnus*. It is the highest-cover frame at the site by both methods (points 26.7%, mask 33.1%). **P3200419** holds the site's only cluster of smaller tables and staghorn with a meandroid brain beside them; **P3200417** is the densest small-colony mosaic, cream corymbose *Acropora*, lobed pocilloporids, a *Turbinaria* cup and honeycomb faviids under a dense violet-and-yellow anthias cloud compatible with *Pseudanthias evansi*. Outside those three, coral is genuinely scarce, and **the transect and the eye agree about that**.

**The fish and the megafauna are the reason this dive was worth 42 minutes.** **P3200408 is the best frame in the redo for sharks** — **two requiem sharks in one well-lit frame**, both showing a broad black caudal margin and no white fin tips, **consistent with grey reef shark (*Carcharhinus amblyrhynchos*)**; the BIEG04 silvertip lesson argues for leaving that as "consistent with" rather than asserted. A third, distant and hazy, cruises the background of **P3200394**. **P3200409 and P3200410 hold a sea turtle wedged under a ledge** — mottled tortoiseshell scutes and a serrated posterior carapace margin, **compatible with hawksbill (*Eretmochelys imbricata*)**, green turtle not excluded from these angles; it is the first turtle of the redo and the mask registers it as 4.63% "other animal" on P3200409, the only such reading at the site. **P3200392, P3200394, P3200395 and P3200396 are massed dark surgeonfish and unicornfish**, with rostral horns clearly visible on several *Naso*, and **P3200391 is a close, beautifully lit fusilier school** compatible with *Pterocaesio*/*Caesio*. **P3200387** adds two brilliant blue-and-yellow damselfish compatible with *Chrysiptera*, **P3200421** a lone mid-water emperor compatible with *Lethrinus* in pure blue water.

**The octocorals are a structural component this site's hard-coral numbers cannot see.** **P3200406 is the frame — a foreground forest of pale gorgonian sea fans and bushy whips over sand**, with unicornfish above; fans and whips also appear in **P3200395, P3200396, P3200408** and as dark antipatharian bushes in **P3200391 and P3200394**. The transect's `% non-coral invertebrates` of 9.93% is the only place they register. Notably, Coralscapes reads **21.7% "live hard coral"** on P3200406 against 13.3% by points — on a frame whose dominant benthic feature is octocoral, not scleractinian (flag 5).

**There is no COTS, no disease and no predation scarring in any of the 19 frames.** Structural loss is old and unspectacular: mask dead-coral 2.21% of image area, rubble 4.32%, and the plain itself is long-consolidated pavement rather than fresh rubble.

**BLEACHING — and BICC35 breaks the pattern the redo has been building.** The published figure is **9.01%**. Anderson recorded **no bleaching on any March or April dive, only in May**, and the corpus agrees: half-month means run 6.1 / 8.2 / 8.4 / 10.0% through March–April and then **31.8% in early May**, which is the real 2015 event. At every site checked so far the March/April figure has dissolved on inspection — bleached points sitting on bare sand, on lobster macros, on 73%-water frames. **Here it does not.** The point set puts **9 bleached points of 570 (1.58%)**, and **6 of those 9 — 67% — sit on a single frame, P3200418**. But P3200418 is **not** a low-coral frame: it is the site's second-highest-cover by mask (29.9% live), and **it is a close macro composed deliberately on a bleaching colony**. A small corymbose colony compatible with *Acropora* has **stark bone-white branch tips over tan-brown tissue at the branch bases** — polyps intact, no algal fouling, a textbook partial/tip bleaching pattern — set against a large pale grey-lavender massive colony compatible with *Porites*. **Coralscapes independently puts 7.78% bleached area on that frame, 77.6% of the site's entire mask-bleached signal, and the photograph corroborates both.** **P3200387 is a second, harder case**: a stark bone-white branching colony sitting in the open on the pavement, which the point set scored with **zero** bleached points while the mask read 2.09%; white skeleton with green film starting on the lower branches makes recent death as likely as bleaching. So the honest statement for BICC35 is neither "the 9.01% is corroborated" nor "it is an artefact": **individual bleached and freshly-white colonies were present and were being photographed on 2015-03-20, but they are individuals, not a site-scale event, and 9.01% of coral is far more than these frames show.** The remaining 3 bleached points sit on P3200395, P3200396 and P3200408 — three ≥46%-water fish and shark frames with no bleached coral visible in any of them, and those three are ordinary provenance artefacts. Drop P3200418 and the point signal falls 1.58% → 0.56%; drop it from the mask and the site mean falls 0.527% → 0.125%. **Anderson's judgement is needed here (flag 8).**

<b>Assemblage composition.</b> Formal transects give 13.8% coral against 53.9% algae across 18 nonzero genera, led — unusually for Chagos — by *Porites* (3.78% of benthos, 28% of all coral) and *Pocillopora* (2.83%) rather than *Acropora* (2.23%), over a substrate that is 25.6% turf-plus-sediment, 22.4% bare and 18.3% erect algae, with crustose coralline algae almost absent at 3.78% (106th of 109). The two depth strata carry near-identical cover (13.3% and 14.2%) but completely different surroundings — algae 67.8% versus 40.0% — and invert the genus order, *Acropora* leading at D3 and falling to sixth at D4 where *Porites* takes over. The photographs match: sediment-dusted pavement with scattered small colonies and one large tabular *Acropora* (P3200420), two colony-scale bleaching records (P3200387, P3200418), and an unusually strong vertebrate fauna — two grey reef sharks, a hawksbill-type turtle and a large acanthurid aggregation.

<b>Assemblage composition.</b> Across all 19 analysed frames — complete coverage, nothing lost to the filename collision — the Coralscapes masks put live hard coral at 9.45% of image area, of which other hard coral is the largest component at 32.9%; the rest breaks down as massive and meandroid 30.0%, tabular <i>Acropora</i> 22.4% and <i>Pocillopora</i> 6.7% of live cover. The independent 30-point annotation of the same frames (570 points) gives live hard coral as 9.1% of all points and apportions it <i>Porites</i> 42.3%, <i>Acropora</i> 36.5%, other hard coral 21.2% of live coral — a <i>Porites</i>-led assemblage, matching the transect's PORI 3.78% > POCI 2.83% > ACRO 2.23% ordering. Non-coral cover is dominated by turf 22.3% and crustose coralline algae 5.4% of all points. Structural loss registers as dead coral 2.21% of image area and rubble 4.32%. Bleaching stands at 1.58% of annotated points — <b>9 points</b>, apportioned <i>Acropora</i> 1.05% and other hard coral 0.53% — against a 9.01% transect figure. Frame-to-frame live cover runs 0% to 26.7% by points and 0% to 33.1% by mask (P3200420 highest-cover by both; P3200421 and P3200389 poorest). Per-frame agreement between the two methods is <b>r = 0.778</b>, above the Chagos-wide 0.7155. Conspicuous fauna and sessile invertebrates appear as sponges in 14 frames, sea cucumbers in 2 frames, giant clams in 1 frame.

**Flags:** (1) **SITE-CODE ALIAS — this site is filed under two codes.** The formal dataset carries it as **BICC35**, island and region "Central Chagos"; the photo folder (`BiTb35-finished`) and the multi-method table both key it as **BITB35**, a Three Brothers code. The same alias affects BICC36/50/51 and BIGC80 (= BINI80). All numbers in this shard are read from the BICC35 rows of the published tables and the multi-method row is read from BITB35. (2) **THE MULTI-METHOD TABLE'S TRANSECT CELL IS WRONG FOR THIS SITE — flagged for action.** `chagos_multimethod_cover_by_site.csv` records BITB35 with a **blank `transect_pct`, the note "no transect cover", and `n_methods = 2`** — but a transect cover figure of **13.75%** does exist, in `Chagos benthic data by site (n=109).csv` under BICC35, backed by 13 transects across two depth strata. The blank is an artefact of the alias in flag 1, not a real absence, and it means the file's `mean_all_pct` (9.3%) and `sd_all` (0.3) for this site are computed from the two CV products only and **exclude ground truth entirely**. That row should be corrected. (3) **A pre-existing shard for this site already exists as `shards/desc_BITB35.md` (20.8 KB).** It is a detailed, identity-corrected description under the TB alias and it was NOT modified by this pass. The two files must be reconciled before publishing so that only one description reaches `crd_dive_sites.json`; note also that `desc_BITB35.md` begins with a non-standard header shape, which is the exact pattern that blanked three live entries via the `paragraphs[2:]` parser trap (HOW_TO §4c). (4) **Multi-method deviation is small here but the direction is consistent.** CoralNet points (9.1%) and Coralscapes masks (9.5%) sit 4.3–4.7 percentage points below the transect figure (13.75%). The deviation column is blank in the source file because of flag 2; the values above were computed against the published transect cover. Do not average the three figures; the transect value is treated as ground truth for cover. (5) **The CV figures rest on a 19-frame scored subset**, which happens to equal the folder count here — that coincidence does not make the products comparable, and the denominators still differ. (6) **The CoralNet points figure is a CV product, not verified ground truth**; point-annotation data corpus-wide is roughly 76% unconfirmed machine output. (7) **`chagos_pct_bleached_recomputed.csv` carries a recomputed value for this site (14.75% from 9 bleached of 61 coral points across 570 points in 19 images) with `status = no_value_in_master`** — again a keying artefact of the alias in flag 1, not a missing value. The recomputed figure runs **5.7 pp above** the published 9.0%. The published column is authoritative and is what the header quotes; bleaching is never recomputed from point counts for publication. The gap is logged because it is the same direction as the photographic impression. (8) **Exposure is recorded two ways** — `exposure-original` = leeward, `exposure-revised` = intermediate. The header uses the revised value; both are recorded here. (9) **D4 rests on 3 transects** against 10 at D3, so the deeper stratum's *Porites*-led genus order and its 40.0% algal cover are less well replicated than the shallower stratum's. (10) **Roughly 46% of the D4 benthos is unaccounted for by coral or algae** in the site-depth summary (against ~19% at D3), likely bare pavement, sand or non-coral invertebrate cover not itemised by depth in this dataset. (11) **Depth codes D3–D4 carry no metre boundaries** in `Chagos benthic data by site-depth (n=336 rows).csv`; they are ordered shallow-to-deep strata, and the 29.3 m in the header is the site-level survey depth, not the depth of any one transect. **No D1 or D2 stratum exists for this site**, so the shallowest reef zone is unsampled. (12) **Two colony-scale bleaching records: P3200387** (a small branching colony uniformly white, pigmented colonies in the same frame) and **P3200418** (a *Pocillopora*-type clump with roughly half its branch tips white against olive tissue on the same colony). Recorded as bleached-or-recently-dead; no site percentage is derived from them. (13) **No COTS, disease, *Drupella* or predation scarring was observed** across the full 19-frame review — an explicit absence-of-evidence finding. (14) **Vertebrate fauna records worth keeping:** two grey reef sharks (P3200408), a hawksbill-type turtle under a ledge (P3200409, P3200410), a large acanthurid aggregation over five consecutive frames (P3200392, P3200394–396, P3200406), a fusilier school (P3200391), plus gorgonian sea fans, sea whips and antipatharians on the substrate beneath the aggregation. A distant shark-shaped silhouette at the right edge of P3200394 is hedged, not confirmed. (15) **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. (16) **Ranks off the `Mean(%bleaching)` column are approximate** — that column carries at least one known bad value (BITB58 listed at 22.22% against a corrected 0.77%), so the 36th-of-98 placement has not been independently audited. (17) **IDENTITY CORRECTION — THIS SITE IS CENTRAL CHAGOS, NOT THREE BROTHERS.** BITB35 = **BICC35**, island/region "Central Chagos", sub-region **GCB-West** (western Great Chagos Bank), 6.18385°S 71.64028°E, 2015-03-20, lagoonal bank reef. **A lookup on "BITB35" returns nothing in the formal dataset**, which is exactly why this section sat unresolved with the hedge "South Brother area (exact atoll unconfirmed)". Everything downstream of the code — GPS, cover, algal cover, % bleached, genus breakdown, depth strata, salinity — was present under BICC35 all along. **Same alias pattern: BITB36/50/51 = BICC36/50/51, BIGC80 = BINI80.** The photo folder is still named `BiTb35-finished`; do not rename it without re-checking the manifest paths. Minor sub-point: the CRD page calls this a "leeward lagoonal bank reef" while the formal record gives exposure as **"intermediate"**, reef location **fore reef**, lagoon **outside** — the two are not contradictory but the formal wording should win in the dataset. (18) **DATA-INTEGRITY ERROR IN `chagos_multimethod_cover_by_site.csv`.** The row for BITB35 carries an **empty `transect_pct`, an empty `method_deviation_range_pp`, `n_methods = 2` and the note "no transect cover"**. That is false. The supplement's `formal vs. informal comparison` sheet holds **13.75%** for this site, filed under BICC35 — the table looked up the wrong code. **Corrected values: transect 13.75%, deviations −4.65 pp (CoralNet 9.1%) and −4.25 pp (Coralscapes 9.5%), n_methods = 3, mean 10.8% ± 2.6.** **BITB36 has the identical fault** (BICC36 = 5.8% transect, 60.44% algae; its 13.1/7.7 photo readings are +7.3 and +1.9 pp HIGH). Both rows need rebuilding, and the whole table should be re-run against the alias map before its corpus-wide offset is quoted again. (19) **THE UNKNOWN-DENOMINATOR CORRECTION CONFORMS — BITB35 IS THE FIFTH LOW-READING SITE IN A ROW TO LAND ON ITS TRANSECT.** 147 of 570 points (**25.79%**) are labelled Unknown. Dropping them lifts the point set **9.12% → 12.29% against a 13.75% transect, residual −1.46 pp** — closing 68% of the raw −4.63 pp gap and landing inside the band already set by BIEA26 (−0.75), BITB27 (−2.21) and BITB29 (−0.28). **Counting bleached coral as coral, which is what the transect does, gives 14.42%, +0.67 pp — an effectively exact land-on.** Restricting to the 11 genuine survey frames (dropping the two close macros, the blue-water frame and the five fish-school frames) gives **13.28%, −0.47 pp**. Excluding water instead of Unknown gives **13.65%, −0.10 pp**, which is suggestive that much of the Unknown pool here is misfiled water — the BITB29 flag-7 finding again. **Substrate renormalisation over-corrects as always**: excluding water, Unknown and fish gives 22.91%, +9.2 pp. (20) **THE BLUE-WATER DILUTION TEST IS POSITIVE HERE, UNLIKE BITB29 — but it does not act alone.** **8 of 19 frames (42%) are ≥40% water** — P3200391 (53.3%), P3200394 (66.7%), P3200395 (50.0%), P3200396 (50.0%), P3200406 (63.3%), P3200408 (46.7%), P3200416 (60.0%) and P3200421 (100%) — site mean water 33.2%. Dropping them moves the point set 9.12 → **10.61** and the mask 9.45 → **12.22**; with the Unknown correction on top, 12.29 → **14.89**, i.e. **+1.14 pp past the transect**. This is the BIEA20 sign-flip in miniature: at a low-cover site the two biases run in opposite directions and partially cancel. **Do not apply both corrections together without saying so.** (21) **THE MASK IS READING GORGONIANS AS HARD CORAL.** P3200406 — a foreground forest of gorgonian fans and whips over sand — is scored **21.66% live hard coral** by Coralscapes against 13.3% by points, its third-highest reading at the site. Fans also occur in P3200395, P3200396 and P3200408. Octocorals are a real structural component here (transect non-coral invertebrates 9.93%) and are invisible to the hard-coral point scheme while being visible-but-misclassified to the mask. Same class of error as BIEA18's gorgonians and BIEA20's antipatharians. (22) **THE TWO CLOSE MACROS ARE WHERE THE MASK FAILS WORST, IN OPPOSITE DIRECTIONS.** **P3200407** (sediment-caked cryptic ledge, ISO 160, the only flash frame) reads 13.3% by points and **2.06%** by mask, −11.2 pp; **P3200418** reads 16.7% by points and **29.87%** by mask, +13.2 pp. Both are composed on a patch rather than for cover and bias photo-derived cover by construction. Together with the pure blue-water frame **P3200421** (100% water, 0.0% live by both methods, a safety-stop exposure at 09:04) they are 3 of 19 frames — **16% non-survey imagery**, a smaller share than BITB27's 38% or BITB29's 20%. (23) **ANNOTATION-PROVENANCE ARTEFACTS, FOUR CLEAR CASES.** **P3200389: 0.0% live by points** on a wide frame carrying visible corymbose *Acropora*, brown *Pocillopora*-type clumps and small faviid domes — the mask reads 4.86%. **P3200419: 1 live point of 30 (3.3%)** on a frame holding several distinct tabular *Acropora*, a meandroid brain and a *Porites* dome; 43.3% Unknown, the site's highest. **P3200392: 1 FISH point of 30** on a frame roughly half filled with a massed surgeonfish school, which the mask scores at 33.7% fish. **P3200417: 36.7% Unknown** on a plainly benthic-dominated frame with only 30% water. With ~76% of Chagos point annotations being CoralNet robot predictions, these are provenance artefacts, not scoring mistakes by a person. (24) **BLEACHING — ANDERSON'S JUDGEMENT NEEDED, AND THIS SITE IS THE FIRST GENUINE COUNTEREXAMPLE IN THE SERIES.** Published 9.01%; the diver recorded none before May. But **P3200418 is a close macro of a partially bleached corymbose colony compatible with *Acropora*, white tips over pigmented bases, tissue intact — photographed at 08:52 on 2015-03-20** — and both independent methods flag it (6 of 9 bleached points, 7.78% mask bleached area, 77.6% of the site's mask signal). **P3200387** shows a second stark-white branching colony that the point set scored at **zero** bleached points (mask 2.09%) and that may be recent mortality rather than bleaching. The three remaining bleached points (P3200395, P3200396, P3200408) are ordinary artefacts on ≥46%-water frames with no bleached coral visible. **The question for Anderson: do these two colonies revise the "no bleaching before May" recollection, or are they isolated individuals — paling, predation or recent death — that do not amount to a bleaching signal?** Note that the supplement's `PB` (percent barren substrate) at this site is 22.42%, and the D4 stratum's 35.5% barren substrate is the highest component of its benthos. Either way **9.01% of coral bleached is far more than 19 frames show**, and both figures cannot stand as written. (25) **THE PAPER'S OWN IMAGE-BASED ESTIMATE READS HIGH WHILE THE MULTIMETHOD PHOTO METHODS READ LOW.** The supplement gives `coral cover (%; 2nd training)` = **17.20%** for BICC35 against the 13.75% transect, a published `% coral cover difference` of **−22.32%**. The multimethod table's CoralNet points give 9.1%. **The same annotations, differently normalised, straddle the transect by ±4 pp in opposite directions** — a direct demonstration that the denominator, not the imagery, is what the corpus-wide photo/transect offset is measuring. This belongs with open question 4 in RESUME_CHAGOS §6. (26) **THE TWO DEPTH STRATA DISAGREE ABOUT THE SUBSTRATE, NOT THE CORAL.** D3 (10 transects): coral 13.3%, algae 67.8%, barren 9.3%, turf-sediment 36.4%. D4 (3 transects): coral 14.2%, algae 40.0%, barren 35.5%, turf-sediment 14.9%, and roughly double D3's *Porites* and *Pocillopora*. Cover is flat with depth; the algal/barren balance is not. Any single-number characterisation of this site averages two quite different benthic surfaces, and D4 rests on only 3 transects. (27) **REGIONAL CONTEXT — CENTRAL CHAGOS IS CORAL-POOR AND ALGAE-DOMINATED.** Same-day neighbour **BICC36 reads 5.8% coral against 60.44% algae**, a coral:algae ratio of 0.096 against BICC35's 0.26. Against Three Brothers proper (BITB27 36.6, BITB29 30.2, BITB30 42.5, BITB34 61.1) these two are in a different regime entirely — which is a second, independent reason the old "South Brother area" placement should never have stood. (28) **COMPOSITION JSONs ARE KEYED ON THE OLD CODE.** Both `chagos_site_composition.json` and `chagos_site_coralnet_composition.json` hold this site under **`BITB35`**; **`BICC35` returns nothing**. The mapping is inverted relative to the supplement, so any join between the two must go through the alias table in both directions. (29) **No COTS, no disease lesions, no corallivore predation scars in any of the 19 frames.** (30) **P3200408 (two grey reef sharks), P3200409/P3200410 (resting turtle) and P3200420 (3–4 m tabular *Acropora*) are flagged for illustrative use.** The turtle and the shark pair are the strongest megafauna images of the Central Chagos group. (31) **The edited set is 19 of ~35 exposures** — the frame-selection confound sits upstream of every photo-derived number here, as everywhere. (32) No per-frame depth data. (33) **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 **9.01%, rank 44 of 110** (1st training: 9.01%, rank 43 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 BICC35 benthic survey data (CSV)](../benthic_data/BICC35_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). *(The superseded BITB35-coded shard stated that no per-site workbook existed for this site; it does, under the BICC35 code.)*