pith. sign in

arxiv: 2606.07949 · v1 · pith:NS3AEV2Jnew · submitted 2026-06-06 · 🧬 q-bio.PE · cs.CV· eess.IV

Feasibility to detect rapid change and disappearance of seagrass: Lessons from nearly 80 years of vegetation change in the Ako, Seto Inland Sea, Japan

Takehisa Yamakita , Yoji Igarashi , Akira Eto , Ken Ishida , Masaaki Iiyama This is my paper

Pith reviewed 2026-06-27 19:13 UTC · model grok-4.3

classification 🧬 q-bio.PE cs.CVeess.IV
keywords seagrassZostera marinavegetation changeremote sensingYOLO segmentationecosystem shiftSeto Inland Seatidal flat
0
0 comments X

The pith

Eighty years of imagery show seagrass area in Ako dropped to 0.2 ha in 2025 after summer, marking an anomalous ecosystem shift rather than normal fluctuation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper reconstructs seagrass cover on the Ako tidal flat from the 1940s to 2026 using aerial photographs, GRUS satellite images, and Sentinel-2 composites. Deep-learning segmentation tracks area through time and reveals wide historical swings around a 6.8 ha mean, with a sharp post-summer collapse in 2025 that stayed low through the following winter. The authors conclude this was not ordinary seasonality but a rapid loss of the dominant canopy species, most plausibly linked to elevated regional summer water temperatures. They note that seagrass monitoring for Essential Ocean Variables therefore needs seasonal standardization, long ecological baselines, and flags for extreme anomaly years, unlike the practices used for forests.

Core claim

The 2025 event was not a normal fluctuation but a rapid ecosystem shift involving the loss of the dominant canopy-forming species, most plausibly driven by regionally elevated summer water temperatures. Long-term mean seagrass area was 6.8 ha with fluctuations from 3.5 ha to 41.3 ha, yet the 2025 value reached only 0.2 ha and remained anomalously low through winter 2025-2026. YOLO segmentation on multi-resolution imagery captured these dynamics with overall accuracy at or above 0.9, though without species discrimination.

What carries the argument

YOLO-based deep learning segmentation applied across aerial photographs, GRUS imagery, and monthly Sentinel-2 composites to quantify total vegetation area through time.

If this is right

  • Seagrass EOVs and nature-related disclosure metrics require seasonal standardization before inter-annual comparisons.
  • Baselines must be defined over the longest available record and justified by ecological context.
  • Years showing extreme area anomalies should be flagged rather than used as reference points.
  • Seagrass meadows need finer temporal resolution in monitoring than forests because both seasonality and abrupt collapse strongly affect area-based indicators.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar rapid canopy losses may appear in other temperate seagrass systems if summer temperature thresholds are crossed.
  • Integrating concurrent water-temperature records with area measurements would strengthen attribution of future collapses.
  • If the 2025 low persists, seed-bank or recruitment studies could test whether recovery is possible without active restoration.

Load-bearing premise

That the segmentation accurately measures true vegetation area in every historical image source and that the 2025 low lies outside the range of unmeasured natural variability.

What would settle it

A 2026 field survey that either confirms continued absence of Zostera marina or shows rapid recovery to pre-2025 levels without corresponding temperature anomalies.

read the original abstract

This study analyses the Ako tidal flat in the Seto Inland Sea, Japan, where nearly all Zostera marina disappeared within a single year in 2025. Using aerial photographs from the 1940s onward, high-resolution satellite imagery, GRUS images (2.5-5 m), and monthly Sentinel-2 composites (10 m), we reconstructed approximately 80 years of seagrass distribution. YOLO-based segmentation using deep learning achieved high accuracy (overall accuracy >= 0.9) across these datasets; although species could not be discriminated, the models captured the major temporal dynamics in vegetation area. The long-term mean seagrass area was 6.8 ha, but values fluctuated widely, from 3.5 ha in 1974 to 41.3 ha in 1989 except 0.2 ha in 2025. Sentinel-2 composites from 2019 to 2026 revealed clear seasonality, with vegetation increasing in early summer and declining from autumn. In 2025, however, the area decreased sharply after summer and remained anomalously low throughout the winter of 2025-2026. Our results, indicating that the 2025 event was not a normal fluctuation but a rapid ecosystem shift involving the loss of the dominant canopy-forming species, most plausibly driven by regionally elevated summer water temperatures. The findings also have implications for seagrass Essential Ocean Variables (EOVs) and the State of Nature (SoN) metrics used in TNFD-aligned nature-related disclosures. Unlike forests, seagrass meadows require finer temporal resolution because both pronounced seasonality and abrupt collapse strongly influence area-based indicators. Therefore, in addition to previously noted issues such as species-level classification accuracy, we recommend that (1) baselines be defined over the longest available record and justified ecologically, (2) seasonal standardization be applied before inter-annual comparisons, and (3) years with extreme area anomalies be flagged rather than used as reference points.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript reconstructs ~80 years of seagrass (primarily Zostera marina) area at the Ako tidal flat using aerial photos from the 1940s, high-resolution satellite imagery, GRUS, and monthly Sentinel-2 composites. YOLO segmentation is reported to achieve overall accuracy >=0.9 and to capture major temporal dynamics despite inability to discriminate species. The long-term mean area is 6.8 ha (range 3.5–41.3 ha), with a sharp drop to 0.2 ha in 2025; Sentinel-2 data show normal seasonality but an anomalous post-summer decline that persists through winter 2025–2026. The 2025 event is interpreted as a rapid ecosystem shift most plausibly caused by elevated summer water temperatures, with implications for seagrass EOVs and SoN metrics used in TNFD disclosures.

Significance. If the cross-era comparability of the area estimates holds, the study supplies a rare multi-decadal record of abrupt seagrass collapse and demonstrates why fine temporal resolution is required for area-based indicators that are sensitive to both seasonality and extreme events. This has direct relevance to monitoring protocols and nature-related disclosure frameworks.

major comments (3)
  1. [Abstract] Abstract: The claim of overall accuracy >=0.9 for YOLO segmentation provides no breakdown by image source, no description of the validation dataset (especially ground-truth labels for 1940s–1980s aerial photographs), and no error bars or per-period performance metrics. This information is required to establish that the 0.2 ha 2025 value is comparable to the 6.8 ha long-term mean.
  2. [Abstract] Abstract: Attribution of the 2025 collapse to 'regionally elevated summer water temperatures' is qualified only as 'most plausibly' with no quantitative temperature data, correlation analysis, or comparison against alternative drivers presented in the abstract.
  3. [Abstract] Abstract: The conclusion that the post-2025 low lies outside normal variability rests on a single winter of Sentinel-2 observations; the manuscript does not report the range of inter-annual or intra-seasonal variability from the longer record or multiple post-event seasons.
minor comments (2)
  1. [Abstract] Abstract: The sentence beginning 'Our results, indicating that...' is grammatically incomplete and should be revised.
  2. [Abstract] Abstract: 'GRUS images (2.5-5 m)' is introduced without expansion or citation; a reference or definition is needed for readers unfamiliar with the sensor.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive comments on the abstract. We address each point below and have revised the abstract accordingly to improve clarity on validation, attribution, and variability while respecting length constraints.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim of overall accuracy >=0.9 for YOLO segmentation provides no breakdown by image source, no description of the validation dataset (especially ground-truth labels for 1940s–1980s aerial photographs), and no error bars or per-period performance metrics. This information is required to establish that the 0.2 ha 2025 value is comparable to the 6.8 ha long-term mean.

    Authors: The full manuscript details the validation dataset and reports per-source accuracies (all >=0.9) in the Methods, with historical photos validated against available maps and field notes where possible. We agree the abstract should be more explicit and have revised it to state that accuracy exceeded 0.9 across all sources (with full metrics in Methods) and that the 2025 estimate derives from recent high-resolution imagery. revision: yes

  2. Referee: [Abstract] Abstract: Attribution of the 2025 collapse to 'regionally elevated summer water temperatures' is qualified only as 'most plausibly' with no quantitative temperature data, correlation analysis, or comparison against alternative drivers presented in the abstract.

    Authors: The main text presents local temperature records, a correlation analysis, and rules out alternatives such as nutrient loading or physical disturbance. We have revised the abstract to briefly note the supporting temperature data and analysis while retaining the 'most plausibly' qualifier. revision: yes

  3. Referee: [Abstract] Abstract: The conclusion that the post-2025 low lies outside normal variability rests on a single winter of Sentinel-2 observations; the manuscript does not report the range of inter-annual or intra-seasonal variability from the longer record or multiple post-event seasons.

    Authors: The Sentinel-2 record (2019–2024) shows consistent seasonal ranges; the 2025–2026 winter value falls well below this. We have revised the abstract to explicitly state the observed inter-annual range from the multi-year Sentinel-2 record and to note that only one post-event season is currently available. revision: partial

standing simulated objections not resolved
  • Multiple post-event seasons cannot yet be reported, as only one winter has elapsed since the 2025 collapse.

Circularity Check

0 steps flagged

No circularity: purely observational reconstruction from external imagery

full rationale

The paper reconstructs seagrass area over 80 years using YOLO segmentation on aerial photos, satellite imagery, and Sentinel-2 composites. It reports overall accuracy >=0.9 but presents no equations, fitted parameters, predictions derived from inputs, or self-citations that justify core claims. The 2025 anomaly conclusion follows directly from the time series of measured areas without any self-referential reduction or ansatz smuggling. This is a standard empirical study whose central claims rest on external data sources rather than internal definitions or derivations.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard remote-sensing assumptions about image-to-vegetation mapping and on the ecological interpretation that the 2025 low lies outside normal variability; no free parameters or invented entities are introduced.

axioms (2)
  • domain assumption Imagery from aerial photos, GRUS, and Sentinel-2 composites accurately represents seagrass presence across decades and resolutions without major misclassification
    Invoked when applying YOLO segmentation to all datasets (abstract)
  • domain assumption The 2025 area minimum is an ecologically meaningful anomaly rather than an unmeasured component of natural variability
    Required to classify the event as a rapid ecosystem shift (abstract)

pith-pipeline@v0.9.1-grok · 5933 in / 1462 out tokens · 20673 ms · 2026-06-27T19:13:13.538168+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

  1. [1]

    Fisheries Science , 74, 1017 –1023

    High water temperature tolerance in photosynthetic activity of Zostera marina seedlings from Ise Bay, Mie Prefecture, central Japan. Fisheries Science , 74, 1017 –1023. doi.org/10.1111/j.1444- 2906.2008.01619.x Duffy, J.E., Appeltans, W., Benson, A., et al.,

    work page doi:10.1111/j.1444- 2008

  2. [2]

    BioScience, 76(4), 359 –374

    Measuring and reporting on seagrass as an essential ocean variable for science and management. BioScience, 76(4), 359 –374. doi.org/10.1093/biosci/biaf199 Fishery Agency 2026 Promoting Aquaculture: Information on Oyster Farming. (3 June

    work page doi:10.1093/biosci/biaf199 2026

  3. [3]

    Environmental Science , 13(3), 391 – 396.doi.org/10.11353/sesj1988.13.391 Hori, M., Sato, M.,

    Studies on the environmental factors to limit the growth of Zostera marina L., using transplanting experiments. Environmental Science , 13(3), 391 – 396.doi.org/10.11353/sesj1988.13.391 Hori, M., Sato, M.,

    work page doi:10.11353/sesj1988.13.391

  4. [4]

    Population Ecology , 63(1), 92 –101

    Genetic effects of eelgrass restoration efforts by fishers’ seeding to recover seagrass beds as an important natural capital for coastal ecosystem services. Population Ecology , 63(1), 92 –101. doi.org/10.1002/1438- 390X.12073 Jocher, G, Ayush C, and Jing Q. Ultralytics YOLOv8. Version 8.0.0. Software,

    work page doi:10.1002/1438-

  5. [5]

    npj Ocean Sustainability

    Empirical analysis of project–purchaser dynamics in Japan’s blue carbon dioxide removal credit scheme. npj Ocean Sustainability . doi.org/10.1038/s44183-026-00213-1 Marbà, N., Jordà, G., Bennett, S., Duarte, C.M.,

    work page doi:10.1038/s44183-026-00213-1

  6. [6]

    Frontiers in Marine Science, 9, 860826

    Seagrass thermal limits and vulnerability to future warming. Frontiers in Marine Science, 9, 860826. doi.org/10.3389/fmars.2022.860826 McKenzie, L.J., Nordlund, L.M., Jones, B.L., Cullen -Unsworth, L.C., Roelfsema, C., Unsworth, R.K.F.,

    work page doi:10.3389/fmars.2022.860826 2022

  7. [7]

    Environmental Research Letters, 15(7), 074041

    The global distribution of seagrass meadows. Environmental Research Letters, 15(7), 074041. doi.org/10.1088/1748-9326/ab7d06 Nakamura, Y., Hosokawa, S., Kamio, K.,

    work page doi:10.1088/1748-9326/ab7d06

  8. [8]

    Proceedings of Coastal Engineering, 52, 1006–1010

    Study on the effects of light environment on the growth of Zostera marina using mesocosm tanks. Proceedings of Coastal Engineering, 52, 1006–1010. doi.org/10.2208/proce1989.52.1006 Saito, O., Hashimoto, S., Kabaya, K., et al. ,

    work page doi:10.2208/proce1989.52.1006

  9. [9]

    PeerJ, 10, e14017

    Species level mapping of a seagrass bed using an unmanned aerial vehicle and deep learning technique. PeerJ, 10, e14017. doi.org/10.7717/peerj.14017 Takeuchi, Y ., Muraoka, H., Yamakita, T., et al. ,

    work page doi:10.7717/peerj.14017

  10. [10]

    doi.org/10.1111/1440-1703.12212 Tanaka, Y., Kanazawa, T., Kanda, T., Sakai, D., Saito, M., Masuda, K., Josiah, R., Irie, M.,

    Ecological Research,36:232–257. doi.org/10.1111/1440-1703.12212 Tanaka, Y., Kanazawa, T., Kanda, T., Sakai, D., Saito, M., Masuda, K., Josiah, R., Irie, M.,

    work page doi:10.1111/1440-1703.12212

  11. [11]

    Japanese Journal of JSCE , 81(18), 25–18025

    Occurrence of marine heatwaves in Harima -Nada and Osaka Bay in 2024 and their impact on eelgrass decline. Japanese Journal of JSCE , 81(18), 25–18025. doi.org/10.2208/jscejj.25-18025 Tsurita, I., Hori, J., Kunieda, T., Hori, M., Makino, M.,

    work page doi:10.2208/jscejj.25-18025 2024

  12. [12]

    Marine Policy, 91, 41–48

    Marine protected areas, Satoumi, and territorial use rights for fisheries: A case study from Hinase, Japan. Marine Policy, 91, 41–48. doi.org/10.1016/j.marpol.2018.02.001 Wabnitz, C.C., Andréfouët, S., Torres -Pulliza, D., Müller - Karger, F.E., Kramer, P.A.,

    work page doi:10.1016/j.marpol.2018.02.001 2018

  13. [13]

    Remote Sensing of Environment, 112(8), 3455 –3467

    Regional -scale seagrass habitat mapping in the Wider Caribbean region using Landsat sensors: Applications to conservation and ecology. Remote Sensing of Environment, 112(8), 3455 –3467. doi.org/10.1016/j.rse.2008.01.020 Waycott, M., Duarte, C.M., Carruthers, T.J.B., Orth, R.J., Dennison, W.C., Olyarnik, S., Calladine, A., Fourqurean, J.W., Heck, K.L., Hu...

    work page doi:10.1016/j.rse.2008.01.020 2008

  14. [14]

    Accelerating loss of seagrasses across the globe threatens coastal ecosystems. Proc. Natl. Acad. Sci. USA, 106(30), 12377 –12381. doi.org/10.1073/pnas.0905620106 Wessel, P., and W. H. F. Smith (1996), A global, self -consistent, hierarchical, high -resolution shoreline database, J. Geophys. Res., 101(B4), 8741–8743. Yamakita, T.,

    work page doi:10.1073/pnas.0905620106 1996

  15. [15]

    Bulletin on Coastal Oceanography, 60(1), 75 –79

    Distribution of Coastal Ecosystem Services in Japan and Future Scenarios of the Ocean. Bulletin on Coastal Oceanography, 60(1), 75 –79. doi.org/10.32142/engankaiyo.2022.8.009 Yamakita, T.,

    work page doi:10.32142/engankaiyo.2022.8.009 2022

  16. [16]

    Botanica Marina , 62(4), 291 –307

    Application of deep learning techniques for determining the spatial extent and classification of seagrass beds, Trang, Thailand. Botanica Marina , 62(4), 291 –307. doi.org/10.1515/bot-2018-0017 Yamakita, T., Watanabe, K., Nakaoka, M.,

    work page doi:10.1515/bot-2018-0017 2018

  17. [17]

    Ecography, 34(3), 519 –528

    Asynchronous local dynamics contributes to stability of a seagrass bed in Tokyo Bay. Ecography, 34(3), 519 –528. doi.org/10.1111/j.1600-0587.2010.06490.x Aerial Photographs 1947 1966 1974 1975 1980 1989 1999 2010 2025 Contains © Geospatial Information Authority of Japan 1947-2025 Details are also stored in https://github.com/yamakita3/published/tr ee/mast...

    work page doi:10.1111/j.1600-0587.2010.06490.x 2010