⚙
AI-generated deep summary
by claude@2026-06, 2026-06-24
· read from full text
ⓘ
This study analysed five years of standardized coral bleaching monitoring in Japan (Monitoring Site 1000 program) using 2,288 observations from 585 survey points across 26 sites in the Ryukyu Archipelago, asking how bleaching prevalence is distributed over time. Across all five years, bleaching prevalence was consistently bimodally distributed: the intermediate 20–80% range stayed stable at about 21–27% of observations, while year-to-year change reflected redistribution between low and high domains, supported by Hartigan’s dip test and beta mixture modelling. The authors also found that the 2022 and 2024 bleaching events differed qualitatively (selective partial mass bleaching vs more comprehensive symmetric bleaching), and that a threshold metric of days above 30°C discriminated bleaching severity better than Degree Heating Weeks when assessed with GEE-corrected AUC. The paper’s main caveat is that its conclusions rely on observational monitoring data and on how ecological severity thresholds and thermal-stress metrics are defined, which the authors note are inseparable design considerations. The paper does not explicitly discuss endometriosis or adenomyosis; it was included in the corpus via a keyword match in the upstream search index.
Abstract
Coral bleaching is conventionally modelled as a continuous response to cumulative thermal stress, yet the distributional structure of site-level bleaching prevalence has rarely been examined. Here we analyse five years (fiscal years 2020–2024) of standardised bleaching surveys from Japan’s Monitoring Site 1000 program, encompassing 2,288 observations across 585 survey points at 26 sites in the Ryukyu Archipelago and adjacent waters. We document three principal findings. First, bleaching prevalence is bimodally distributed in all five years: the intermediate range (20–80%) remains stable at 21–27% of observations, while inter-annual variation is driven by redistribution between low and high domains — confirmed by Hartigan’s dip test (all p < 0.001) and beta mixture modelling (ΔBIC = 9.0–113.9). Second, the 2022 and 2024 bleaching events are qualitatively distinct: 2022 was a partial mass bleaching (positively skewed, selective), while 2024 was comprehensive (symmetric, median 60.0%). Third, a simple threshold metric (days above 30°C) outperformed Degree Heating Weeks in discriminating bleaching across all severity levels (GEE-based AUC: 0.877 vs 0.624 at ≥50% prevalence, p 0%, p = 0.007), indicating that metric structure and the ecological severity threshold defining the outcome are inseparable design considerations.
Full text
2,630 characters
· extracted from
oa-html
· click to expand
Abstract
Coral bleaching is conventionally modelled as a continuous response to cumulative thermal stress, yet the distributional structure of site-level bleaching prevalence has rarely been examined. Here we analyse five years (fiscal years 2020–2024) of standardised bleaching surveys from Japan’s Monitoring Site 1000 program, encompassing 2,288 observations across 585 survey points at 26 sites in the Ryukyu Archipelago and adjacent waters. We document three principal findings. First, bleaching prevalence is bimodally distributed in all five years: the intermediate range (20–80%) remains stable at 21–27% of observations, while inter-annual variation is driven by redistribution between low and high domains — confirmed by Hartigan’s dip test (all p < 0.001) and beta mixture modelling (ΔBIC = 9.0–113.9). Second, the 2022 and 2024 bleaching events are qualitatively distinct: 2022 was a partial mass bleaching (positively skewed, selective), while 2024 was comprehensive (symmetric, median 60.0%). Third, a simple threshold metric (days above 30°C) outperformed Degree Heating Weeks in discriminating bleaching across all severity levels (GEE-based AUC: 0.877 vs 0.624 at ≥50% prevalence, p 0%, p = 0.007), indicating that metric structure and the ecological severity threshold defining the outcome are inseparable design considerations.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
Major revisions from v1: 1. Statistical correction: All AUC comparisons now use Generalised Estimating Equations (GEE) with exchangeable correlation structure, clustering by site (n=21), to correct for spatial pseudoreplication. Bootstrap cluster Wald tests (10,000 iterations) replace DeLong tests. 2. Updated AUC values: GEE-corrected AUCs show larger separation between days above 30 degrees C and DHW than uncorrected values (e.g., 0.877 vs 0.624 at >=50% threshold, compared to 0.846 vs 0.758 in v1). 3. Expanded beta mixture model: EM algorithm now uses 50 random initial parameter sets (vs 3 in v1); convergence verified by identical BIC values. 4. ICC recalculated on balanced panel: All ICC values now computed on the 21 sites present in all 5 fiscal years. 5. Enhanced Discussion: Added sections on structural inadequacy of DHW for state transitions, distributional parameter vocabulary for event characterisation, and parallel with psychiatric epidemiology. 6. Data count correction: Total observations corrected from 2,287 to 2,288 (FY2024: 470 to 471). 7. Title revised to emphasise consistency with state transition dynamics. 8. Japanese abstract added.
Text is read by the "Ask this paper" AI Q&A widget below.
Extraction quality varies by source — PMC NXML preserves structure
cleanly, OA-HTML may include some navigation residue, and OA-PDF can
have broken hyphenation. The publisher copy
(via DOI)
is the canonical version.