. . . . "2026-07-25"^^ . . "Open-source Sentinel-2 chlorophyll-a retrieval in the Westerschelde qualifies the Sado weak-Chl-a finding" . "This replication qualifies the original claim. Sent et al. (2021) found that\nSentinel-2 MSI retrieves water-quality parameters with parameter-dependent\naccuracy — strong for turbidity but weak for chlorophyll-a, concluding that \"for\nthe key parameter Chl-a further research is needed, with a more complete set of\nmatch-ups.\" Testing that chlorophyll-a limb in an independent, more turbid estuary\n(the Westerschelde, 2016–2026) with a fully open-source atmospheric-correction\nchain (Acolite + Gons et al. 2005, the aGS chain), we obtained no useful agreement\nwith in situ chlorophyll-a: R² = 0.10 over the full record (N = 33) and R² = 0.13\nfor the 2018–2020 subset (N = 14), with a negative regression slope and a positive\nbias. This does not confirm that an operationally reproducible, open-source chain\ncan retrieve estuarine chlorophyll-a at the accuracy of the paper's selected cGS\nchain (R² = 0.63, C2RCC + Gons). Rather than contradicting the paper, the result\nsupports and extends its central caution: chlorophyll-a is the parameter for which\nMSI retrieval is not yet reliable, and that weakness persists — indeed deepens —\nwhen the proprietary C2RCC step is replaced by an open-source processor and the\nmethod is transferred to a different optical regime." . . "From results/chla_satellite_acolite.parquet, agreement computed in log10 space\n(scripts/matchup_stats.py), matching the paper's log-scale treatment of Chl-a:\n\n Full record (2016–2026): N = 33, R² = 0.10, slope = −0.20,\n RMSE = 0.63 (log10), BIAS = +0.40\n Original-window subset (2018–2020): N = 14, R² = 0.13, slope = −0.14,\n RMSE = 0.59 (log10), BIAS = +0.30\n\nSatellite chlorophyll-a (median 8.7 µg/L, range 1.9–27.7) systematically exceeds\nthe coincident Rijkswaterstaat in situ chlorophyll-a (median 3.1 µg/L, range\n0.9–26.0) across six along-axis stations. The negative slope is the decisive\nfeature: the retrieval does not track in situ chlorophyll-a at all, consistent with\nthe Gons red-edge algorithm responding to backscatter/turbidity rather than pigment\nin this high-SPM estuary. For comparison, the original study's selected cGS chain\nreported R² = 0.63. Match-up figure: figures/main_result.png." . "Three factors bound this conclusion and must be read together.\n\n1. Processor change. The paper's headline Chl-a number (R² = 0.63) is the cGS\n chain (C2RCC + Gons). We ran the aGS chain (Acolite + Gons) because C2RCC exists\n only inside ESA SNAP and crashes natively in the container\n (docs/atmospheric-correction-choice.md). We therefore did not test cGS and\n cannot dispute its number. The paper itself reports Acolite as its worst-\n performing processor (mean BIAS 2.78, mean APD 254 percent), so a weaker aGS\n result is partly anticipated by the original study — a reason this is a\n qualification, not a contradiction.\n\n2. Site and period change. This is a Replication, not a Reproduction: the Sado\n in situ data are not public (paper Data Availability Statement), so we validated\n in the Westerschelde (2016–2026), a different and more turbid mesotidal estuary.\n A different optical regime tests generalisability, not the correctness of the\n original analysis. We report both the full record and the 2018–2020 subset to\n separate the period change from the site change as far as the data allow.\n\n3. Sample size. N = 33 (14 in the original window) over six stations is small — as\n in the original (N = 19–21) — so every R² carries wide uncertainty. The\n direction of the result (no useful agreement) is robust; its precise magnitude\n is not.\n\nA grid-sensitivity check on a HEALPix / DGGS (GRID4EARTH) representation of the\ncorrected fields is deliberately out of scope here and deferred to a separate\nfollow-up chain that will extend this one, so the primary validation stays on the\nnative Sentinel-2 grid." . . . . . "Anne Fouilloux" . "2026-07-25T16:23:44.374Z"^^ . . . . . "Open-source Sentinel-2 chlorophyll-a retrieval in the Westerschelde qualifies the Sado weak-Chl-a finding" . . "RSA" . "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" . "dDesqRQfCp2aMXfLc0PLyoVfO3SFNnYoqy0+z5CNqhvoeYSRU1X5kHvkMd/66c/JQLvolzVjc8xi0AnLPgjwE9SkYtLA+7YN9YIGXhzrSmE2pG1OMCVbucYkC82jSqZ304Jqdn869ZTgNhY5O18KS2YJq6+Lm8Ask3dR3FD3VlUgbiLq9huARnaCSxGnrD+JiXUWlddQf46oUfSz1IdQ0UHftM7Gx733b/jj/B1Lvbkn4+ZKmXlJI6PMW3kbdjgWf+PmqT7RKMA1TbxqXGmzKEO4Pvai/6vSiktPmdMlA42DLS0E3qBG4w6NbNKKsOLd3Rc9q+G8T9VKmwb2LSXkwhdOrx2lt6sGWzLSRwuYbvoabYJoBrVJ9NJpr98QxTOV/atTw+TAmpBQoIDNr56B3JgFho7zjOtPgHYVXaaP7ImvHygBn6QG0UBFvUoEseBqEZgSBD5D8Plri9hazEvfgxZSgl9aYA3SXxAvT9zFvdXZorwkDCVK29O+/xFiVc4m5ZNUTll4veq692aOCptnapoYMaev/YW3WvLOj6hHfX1cANa3Pk0yHVzpOR21BTEaaixxjrNgGEjhrAleZSLUD8PD5/MrwxFViqP+kfUusJ4Um1Hey9L8/ymYoq9MM4EV5rlPdnsPsxZKxd5Ip86KOPGZx/qDlC+JsmAyycNXQvM=" . . .