. . . . . . "Open light+SST+bathymetry HMM replication of juvenile white-shark geolocation" . . . . . "- Open pangeo-fish HMM in place of the proprietary, non-recomputable GPE3 algorithm (same input ingredients per the paper: light + satellite SST + release/pop-up anchors + a movement prior), plus a bathymetry/land constraint GPE3 does not use.\n- Only the twilight TIMES are taken from the light series (anti-circularity): the tag's own seed latitude/longitude and SST columns are not used.\n- Baseline-free species-range grid (the GPE3 track is used for neither the fit nor the grid).\n- Argos referee cleaned of physically impossible fixes (out-of-region positions; implied speeds far above a shark's), as the paper's Technical Validation instructs; the median metric is in any case robust to them." . . "An open hidden Markov geolocation model (pangeo-fish) on a HEALPix (NESTED, level 9) state space spanning the documented NE-Pacific + Gulf-of-California species range (lon[-125,-106], lat[22,38]) — a baseline-free grid fixed by ecology, not by the GPE3/Argos tracks. The daily emission is the product of three independent factors: (1) an astronomical TWILIGHT emission that scores each cell by how well its modelled sunrise/sunset time (astropy) matches the tag's detected dawn/dusk times; (2) a TEMPERATURE emission matching the tag's external depth-temperature record against the GLORYS12V1 thetao field; and (3) a BATHYMETRY factor (GEBCO/ETOPO) that masks land and softly requires the seabed to be at least as deep as the animal's observed dives. Movement uses a land-barrier Gaussian transition (a custom kernel that forbids probability flow across land), so the model cannot move the animal over land. The Brownian diffusion sigma is fit per tag by maximum likelihood; daily positions are taken by maximum-a-posteriori decode and joined into a continuous, land-respecting track via a constrained shortest-path decode. Release and pop-up coordinates are the only spatial anchors. Each tag's track is compared to the held-out SPOT Argos fixes (quality classes 1/2/3) by median great-circle distance, with GPE3 scored identically for reference." . "The paper's computed daily PAT geolocations of juvenile white sharks. Specifically, whether the daily positions GPE3 produced from the light series can be recovered by an independent, fully open method, evaluated on the four sharks that carry a co-deployed SPOT tag, so that the held-out Argos satellite fixes serve as an independent positional referee. In scope: the daily geolocation track and its positional accuracy. Out of scope: the depth and temperature time series themselves, the tag hardware, and any non-geolocation data product. GPE3 is treated as a comparison baseline, never as ground truth; the Argos fixes are the referee and are never used to fit the model." . . . "Anne Fouilloux" . "2026-06-07T21:04:43.560Z"^^ . . . . . "Open light+SST+bathymetry HMM replication of juvenile white-shark geolocation" . . "RSA" . "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" . "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" . . . . . . . . . "Open temperature-at-depth replication of juvenile white shark geolocations using a HEALPix hidden Markov model" . . . . . . "This is a different-methodology replication, so the whole geolocation engine differs from the original — that difference is the point, not a deviation to apologise for. The deliberate configuration choices and honest exclusions that bound the comparison:\n\n1. Minimal configuration by design. The open method is run with a single emission signal (temperature-at-depth only) and only the release and pop-up endpoints as anchors. It does NOT use the light levels or satellite-SST that the original GPE3 method fuses, nor the acoustic-receiver detections available for many of these sharks, nor multi-variable (temperature-plus-salinity) or bathymetry constraints — all of which pangeo-fish supports. The result is therefore a floor for this bare configuration, not a ceiling for the method.\n\n2. Original method is proprietary and not recomputed. GPE3 (Wildlife Computers, a discretised hidden Markov model fusing light, SST and known endpoints, requiring a user-supplied mean-swim-speed prior) cannot be re-run; its already-released track outputs are used as the comparison baseline.\n\n3. Implementation note. The fit/decode uses pangeo-fish's low-level EagerEstimator + EagerBoundsSearch path rather than the high-level optimize_pdf helper, which trips a pint/xarray incompatibility in the installed version.\n\n4. Two honest exclusions from the analysed tag set. Tag 02_01 (a PAT2 unit) records only internal recorder temperature with no external ambient-water sensor, so the temperature-at-depth emission is invalid and the tag is dropped. Tag 06_10 is dropped because its basin-scale GLORYS reference subset repeatedly failed to download. One further recovered PAT tag has no co-deployed SPOT tag, so it has no Argos referee and contributes only a track-vs-GPE3 comparison, not an accuracy validation." . . "Input data: the released biologging archive (data DOI https://doi.org/10.24431/rw1k6c3). From the metadata table, the recovered PAT tags carrying a full depth-and-temperature time-series are selected, and each tag's depth/temperature record is read from its per-deployment archive entry.\n\nReference field: the GLORYS12V1 global ocean physical reanalysis (Copernicus Marine GLOBAL_MULTIYEAR_PHY_001_030), variable thetao (sea-water potential temperature), subset per tag to its deployment time window and a northeast-Pacific bounding box and accessed via the copernicusmarine client.\n\nGeolocation model: the open-source pangeo-fish hidden Markov model. For each timestep an emission likelihood is computed as the agreement between the tag's measured temperature-at-depth profile and the GLORYS thetao field; movement between timesteps is a Brownian-motion transition whose diffusion parameter (a spread in radians on the sphere, bounded by an assumed maximum swim speed) is fitted per tag by maximum likelihood. The state space is a HEALPix grid in NESTED ordering at refinement level 9 (about 6.4 km cells, matched to the GLORYS resolution; project convention is NESTED, never RING). Fitting and decoding use pangeo-fish's low-level EagerEstimator with an EagerBoundsSearch over the diffusion parameter, then the most-probable daily track is decoded. Track endpoints are anchored by the recorded release and pop-up positions.\n\nValidation/referee: for the tags that also carried a SPOT tag, the daily great-circle distance between the re-derived positions and the independent Argos satellite fixes is computed, and the same distance is computed for the released GPE3 track against those same fixes — so the open method and the original method are compared on a common, independent referee. The Argos fixes are the referee; GPE3 is a comparison baseline, not ground truth. Per-tag fitted diffusion and track-vs-GPE3 agreement are also recorded." . "Scope: the daily geolocation product released for the juvenile white sharks in this data descriptor — specifically the computed positions of the recovered pop-up archival (PAT) tags, whose full depth and temperature time-series are available in the archive. In scope: re-deriving those daily positions independently and judging their accuracy against the co-deployed Argos satellite fixes that serve as the referee. Out of scope: the dataset's tag/individual coverage counts (a separate, trivial check, not pursued here); the SPOT-only deployments, which already are direct Argos positions and need no geolocation; the depth records and any downstream ecological inference (migration, habitat) drawn from the tracks in other papers. The released GPE3 tracks are treated as the comparison baseline, not as ground truth." . . . "Anne Fouilloux" . "2026-06-06T08:28:36.984Z"^^ . . . . . "Open temperature-at-depth replication of juvenile white shark geolocations using a HEALPix hidden Markov model" . . "RSA" . "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAoDcOiD+jen8awiJ6DB2ewDw66PeG64hODmgNFwy7GrwQui4HKnHdvxd++1UhTgiOfycxyxBb7sXPSikLw/1TsSyPsEl0P3/+600szxpTGgLNzW+bZ2DVP3d8ERMV1aWpH0ci3B/5vmK+vXQZ4uCoq57NE0MiFg5c13Gy0gd6n7wZYEhYM4AjWSLL0QS/HY+TFZMYL9bCFeATennGrlB2UEjRlw21UB2Ah16ZZ6hxQlfctFJZE7TGnBJPB3ttTjfcOfamhjZVwQ0yV9mv7x6PGiSmkzpJTVLjn8hagoKT05YUwVQArFb+w7f6sXqvvljMigjd/Rbqgbye/lLUAZLfJSnFM58TubfpEJvXV4zNMDEoT3VQ7dokgoLgMrmjZCKATtQ7gomocoTJ1NhN2esRNtGzWaS2obL/mueUQlMlavssZnqL8WICkdAuDlwDVNbsbwEWKQ50kiPdAdduSigifxA4CM7TgvnxqZVoAResEGP6UhTTem3T4CsbEas1Caj9wa7M1jPjACu5LF5BwcVns3ZQHWLipjRjD+9/ur3G8QtuxbNhmXlDYQ6tXxB1lK+Oz7O519b3bA15ilzFl0SdvMBGTe46xaQ9DsJT18THKnPbUhNMy0dH0VtzpB+EEaXZ25Fp9VHMEUqo1lLS9e89eO3efiqkESKQ7wmB+/DlIRcCAwEAAQ==" . "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" . . . . . . . . . "Inter-product wave-exposure comparison at Natura 2000 marine sites across three European storms" . . . . . . "Loveland et al. 2024 ran a coupled wave-circulation model with full versus reduced-order source terms in the Gulf of Mexico and evaluated the internal wave and water-level fields. This study deviates in five ways. (1) Inter-product surrogate: rather than running a coupled model with full versus reduced-order source terms, it compares two operational production wave reanalyses, testing whether the product choice propagates to a downstream biodiversity-exposure metric rather than the source-term physics directly. (2) Domain: European coasts and three extratropical-and-Mediterranean storms, not Gulf of Mexico hurricanes. (3) Downstream metric: biodiversity exposure attribution at Natura 2000 marine sites, not the model's internal fields. (4) Native-grid comparison: each product is sampled on its OWN native grid per site polygon, NOT on a common regridded grid; the inter-product delta therefore conflates resolution and physics, which is intentional because the decision-relevant quantity for a user choosing a product is the bundle of resolution and physics that product delivers, not either factor in isolation. (5) Non-independence: the two products share ERA5 atmospheric forcing (explicitly for the North-West Shelf and Mediterranean products) and partial model lineage (MFWAM for WAVERYS, IBI, and Mediterranean; WAVEWATCH III for the North-West Shelf), so the comparison is conservative — the noise-floor threshold computed under an independence assumption over-estimates the true disagreement floor — and the inter-product signal is best framed as a resolution effect dominated by the fine regional resolving nearshore structure WAVERYS cannot." . . "For each storm, the WAVERYS global product and the basin-matched regional CMEMS product (IBI for Xynthia, North-West Shelf for Xaver, Mediterranean for Gloria) are downloaded over the storm window (with 24 hours of pre-storm padding) and a bounding box that fully contains the marine Natura 2000 sites of that region. Each product is kept on its own native grid — WAVERYS at 0.2 degrees, regionals at their native fine resolution — and trimmed to the storm window; the regional is NOT regridded onto the WAVERYS grid (regridding would smooth away the nearshore detail that is the regional product's value and would propagate coastal land-mask NaN onto the coarse grid, dropping about half the marine sites silently). EEA Natura 2000 (End-2024, revision 01) polygons are joined with the Standard Data Form; marine sites intersecting each storm's core region are selected and their Annex I habitat and Annex II species counts attached. For every site and each product, the per-product exposure is the spatial mean over the wet cells whose centres fall inside the polygon of the temporal-maximum significant wave height; sub-grid sites fall back to the single cell containing the polygon centroid, used only if that cell is wet (no nearest-wet-cell search outward — a 1-cell buffer on the 0.2-degree WAVERYS grid is ~22 km and would pull open-ocean values onto sheltered sites). A product that yields no wet value at the site does not resolve the site. Three tiers per site result: both products resolve (continuous inter-product delta), only one resolves (the product choice is categorically decisive — one product places the site in water, the other on land or sub-wave-field), neither resolves (out of scope; reported as a coverage caveat). Each site is weighted by log1p(number of Annex I habitats + number of Annex II species), additive so that bird Special Protection Areas with Annex II species but no Annex I habitat listing are not zero-weighted. The headline statistic is the biodiversity-weighted fraction of resolvable sites where the attribution differs — at sites both products resolve, the absolute peak-Hs delta exceeds the threshold X equal to the joint observational noise floor (root-sum-square of the two products' CMEMS QUID validation RMSE values, approximately 0.4 metres); at sites only one product resolves, the difference is counted as categorically decisive. The result is reported per storm and aggregated, with a threshold sweep over X = 0.2 to 0.8 metres to assess robustness, a per-site biodiversity-versus-delta diagnostic, and the three-way decomposition agree / magnitude-disagree / decisive. The pipeline is a four-notebook Snakemake workflow in a pixi-pinned environment." . "Tested: whether substituting the global WAVERYS wave reanalysis with the matched higher-resolution regional CMEMS reanalysis changes the storm-wave exposure attributed to coastal marine Natura 2000 sites. Scope is three named European storms spanning three physically distinct regimes — Xynthia (2010, French Atlantic / Bay of Biscay), Xaver (2013, southern North Sea), Gloria (2020, north-western Mediterranean) — and the marine Special Areas of Conservation and Special Protection Areas in each storm's landfall region. Exposure is operationalised at the level of the storm-window peak significant wave height attributed to each site polygon, and the claim is tested at the level of the biodiversity-weighted prevalence of inter-product disagreement across sites. Out of scope: storm surge, wave period and direction, duration, any ecological-impact or population modelling, and running a coupled wave-circulation model directly." . . . "Anne Fouilloux" . "2026-06-01T19:53:06.702Z"^^ . . . . . "Inter-product wave-exposure comparison at Natura 2000 marine sites across three European storms" . . "RSA" . "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" . "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" . . . . . . . . . "Computational replication of Arctic Amplification — Rantanen et al. 2022" . . . . "see draft for the full numbered list — 7 items covering extended time window, Python vs R, GISTEMP boundary, CDS API v2, HadCRUT5 version, boundary sensitivity scope, no CMIP comparison" . . "Annual mean temperature anomalies for the Arctic region (66.5°N–90°N) and the global mean are extracted from each of four observational datasets: ERA5 monthly 2-metre temperature (Copernicus CDS), GISTEMP v4 zonal means (NASA GISS), Berkeley Earth Land+Ocean monthly gridded data, and HadCRUT5. An ordinary least-squares (OLS) linear trend is fitted to the full available annual series starting from 1979, separately for the Arctic and global means. The Arctic Amplification ratio for each dataset is the Arctic OLS slope divided by the global OLS slope (dimensionless). The headline result is the unweighted mean of the four per-dataset AA ratios. All computations are implemented in Python using xarray, numpy, and scipy.stats.linregress. The full pipeline (data download → cleaning → analysis → figures) is automated via Snakemake and reproduced with pixi-managed dependencies." . "The observational component of the Arctic Amplification claim: whether the ratio of the Arctic (66.5°N–90°N) linear warming trend to the global mean warming trend, computed from the same four publicly available observational datasets, exceeds 3× for the period starting in 1979. The model comparison component of the original paper (showing that CMIP5/CMIP6 models underestimate the observed AA) is explicitly out of scope." . . . "Jean Iaquinta" . "2026-05-31T09:43:40.016Z"^^ . . . . . "Computational replication of Arctic Amplification — Rantanen et al. 2022" . . "RSA" . "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" . "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" . . . . . . . . . "Replication: does target-group-background SDM correction restore Iberian-bird hotspots against the EU Article 12 gold standard?" . . . . . "1. Different data: Iberian breeding-bird occurrences from GBIF, not Phillips' 226-species, six-region NCEAS benchmark collection (Elith et al. 2006).\n2. Different instrument and outcome metric: Phillips evaluates the correction by per-species AUC / point-biserial correlation on independent presence-absence test sites; this study evaluates it by top-5% hotspot misidentification against the EU Article 12 expert rangemaps as the gold standard. We test Phillips' correction mechanism on a new instrument (hotspot identity recovery), not on his original metric.\n3. Different substrate: an equal-area HEALPix-NESTED grid (geographic, WGS84-aware), rather than Phillips' region-specific projected raster grids; this is the same substrate as the parent and sibling chains, chosen so results compose across the family.\n4. Aggregation step Phillips does not perform: per-species suitabilities are summed to a per-cell predicted-richness surface and then hotspot-ranked. Phillips stops at per-species evaluation; the stacking-to-richness step is added to address the biodiversity-hotspot question.\n5. MaxEnt engine: elapid/maxnet (pure-Python), not Phillips' original Maxent software; the companion reproduction sdm-phillips-reproduction shows this engine reproduces the direction and magnitude of his AUC result on his own data, so the engine is not a confound for the hotspot question.\n6. Predictors: CHELSA v2 bioclimatic variables only (no land cover or topography), an analogue of Phillips' 11-13 region-specific layers." . . "We fit a separate presence-background species-distribution model (MaxEnt, via the pure-Python elapid library, CPU) for every breeding-bird species with at least 20 occupied grid cells, using CHELSA v2 bioclimatic variables (standardised) as environmental predictors over an equal-area HEALPix-NESTED grid (geographic, WGS84-aware, via healpix-geo). Each species is fitted twice: once with target-group background — background cells drawn from the pooled all-bird occurrence cells, weighted by per-cell record count, so the background carries the same spatial sampling bias as the occurrences (Phillips' proposed correction) — and once with random (uniform) background as the contrast that fits rather than cancels the bias. MaxEnt features are linear + quadratic + hinge (the simpler linear + quadratic set is run as a sensitivity check). MaxEnt with presence + background is the maximum-entropy estimate of the relative occurrence-rate surface and is equivalent to an inhomogeneous-Poisson-point-process intensity fit (Warton & Shepherd 2010; Renner et al. 2015), so changing the background reference measure is what cancels the shared bias. Per-cell predicted suitabilities are summed across species to give a per-cell predicted-richness surface (one for each background type). For each surface we take the top-5% cells as hotspots and measure misidentification as the symmetric set non-overlap of those hotspots against the EU Article 12 expert-rangemap hotspots (matched to the species intersection of each GBIF strategy). We compare the two SDM surfaces against an uncorrected raw occurrence-count richness baseline; the uncorrected baseline is required to reproduce the sibling study's published misidentification before the corrected numbers are trusted. We report the full resolution ladder (Nside 16 to 512) and adopt Nside 256 (about 25 km) as the pre-registered headline scale. GBIF inputs reuse the sibling download DOIs identically (museum 10.15468/dl.r8pcat; all-observations 10.15468/dl.e9xv7p). The per-cell occurrence-frequency tables, the Article 12 GeoPackage gold standard, and the equal-area-grid substrate are reused from the sibling study." . "Scope: whether Phillips et al. (2009)'s target-group-background sampling-bias correction can restore the IDENTITY of the top-5% biodiversity hotspots — i.e. recover which grid cells are genuine richness hotspots — when those hotspots are derived from spatially biased occurrence data and benchmarked against an independent expert-rangemap gold standard (the EU Article 12 breeding-bird distributions for Iberia). In scope: both basis-of-record strategies (museum-grade records and all observations), so the test is not contingent on one data-quality regime; a ladder of equal-area grid resolutions, with a single pre-registered headline scale; and a comparison of the corrected hotspots against both an uncorrected occurrence-count baseline and a random-background SDM baseline. Out of scope: Phillips' own claim that target-group background improves per-species predictive discrimination (AUC) — that is the faithfulness question, established separately on Phillips' own data in the companion reproduction sdm-phillips-reproduction, and is NOT disputed here. This study tests only whether the per-species correction propagates up to recovering true hotspot identity." . . . "Anne Fouilloux" . "2026-05-31T09:11:17.715Z"^^ . . . . . "Replication: does target-group-background SDM correction restore Iberian-bird hotspots against the EU Article 12 gold standard?" . . "RSA" . "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" . "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" . . . . . . . . . "Reproduction of Phillips et al. 2009 Table 2 — MaxEnt random vs target-group background AUC" . . . . "(1) MaxEnt engine: the open-source elapid/maxnet implementation is used rather than Phillips' original Java Maxent, so exact AUC decimals are expected to differ even where direction and magnitude agree. \n(2) Only MaxEnt is run — Phillips' broader Table 2 also covered BRT, MARS, GAM and other methods, which are not reproduced here. \n(3) One species fewer is modelled than Phillips' 226 (225 here) because a species whose presence-absence evaluation column has no presence/absence variation gives an undefined AUC and is dropped. \n(4) A minimum-presence threshold of 5 occurrences is applied before a species is fit." . . "The Elith et al. 2006 NCEAS presence-only / presence-absence benchmark — the same data Phillips used — is obtained from the rspatial/disdat R data package (data paper doi 10.17161/bi.v15i2.13384) by downloading its .rds tables and reading them in Python with pyreadr. For each species across the six regions (AWT, CAN, NSW, NZ, SA, SWI), a MaxEnt model (elapid engine, linear + quadratic + hinge features) is fit twice: once against the region's random background sites supplied by disdat, and once against a target-group background formed from the pooled presence localities of all species in the same biological target group. Both models predict at the independent presence-absence evaluation sites and AUC is computed with scikit-learn. Per-species AUC for the two background types is aggregated to region, group and overall means, and the paired difference is tested with a Wilcoxon signed-rank test. 225 species across 6 regions are modelled." . "This study reproduces the Maxent row of Phillips et al. 2009 Table 2: the comparison of mean predictive AUC for presence-only species distribution models trained with random background versus target-group background. In scope: the direction and magnitude of the AUC gain from target-group background, aggregated across species, and the stronger gain in regions with greater sampling bias. Out of scope: the other modelling methods Phillips also tested (BRT, MARS, GAM and others) and the absolute predicted-distribution maps — only the MaxEnt AUC comparison on this benchmark is tested here." . . . "Anne Fouilloux" . "2026-05-31T08:08:45.811Z"^^ . . . . . "Reproduction of Phillips et al. 2009 Table 2 — MaxEnt random vs target-group background AUC" . . "RSA" . "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" . "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" . . . . . . . . . "Coverage-based effort correction of Iberian-bird richness hotspots (HEALPix-NESTED)" . . . . . . . "Chao & Jost (2012) introduce coverage-based rarefaction for comparing whole-community samples by completeness. This study deviates by applying their estimator at a new unit of analysis — the individual HEALPix-NESTED grid cell as a \"sample\" — to correct a spatial biodiversity-hotspot comparison, an application the methods paper does not itself perform. Relative to the prior sibling chain (Hurlbert & Jetz 2007 scale-dependence replication), this study reuses that chain's GBIF download DOIs, Iberian study system, HEALPix-NESTED substrate, year-split, top-5% hotspot definition, and Article 12 comparator unchanged, and adds the coverage-correction layer plus the target-coverage sweep as the new method. The estimator can only be evaluated on cells with enough records, so effort-poor cells are necessarily censored; the consequence of that censoring for the comparison is reported in the linked Outcome's limitations." . . "Iberian bird occurrences (Spain, Portugal, Andorra, Gibraltar) are taken from GBIF under two basis-of-record strategies with the prior chain's existing download DOIs (museum: PRESERVED_SPECIMEN + MACHINE_OBSERVATION, 10.15468/dl.r8pcat; all-observations: additionally HUMAN_OBSERVATION, 10.15468/dl.e9xv7p), and binned onto a HEALPix-NESTED ladder (Nside 16, 32, 64, 128, 256, 512) using the geographic, WGS84-aware healpix-geo library, NESTED ordering throughout. Records are year-split at 2000: post-2000 form the occurrence/atlas layer, pre-2000 the per-species convex-hull rangemap layer.\n\nFor each cell, the per-species record-count vector is retained. Coverage-based rarefaction (Chao & Jost 2012) is applied per cell: sample coverage is estimated with the bias-corrected Good–Turing estimator; per-cell richness is then standardised to a common target sample coverage C* by size-based rarefaction (inverting the rarefied-sample coverage curve) where C* is at or below the cell's observed coverage, and by coverage-based extrapolation (Chao1 undetected-species term) where C* exceeds it. Cells with too few records for a stable coverage estimate are censored. The target coverage C* is swept over a grid (0.80, 0.90, 0.95, 0.99) plus the minimum common coverage Cmin, rather than fixed at one value.\n\nAt each (strategy, Nside, C*), the top-5% richest cells of the coverage-standardised surface are recomputed and their symmetric set non-overlap (\"misidentification\") is measured against (a) the EU Birds Directive Article 12 expert distribution polygons (EEA 2013–2018, 10 km grid, reprojected to WGS84, matched to the species intersection) and (b) the pre-2000 occurrence-hull rangemap. The uncorrected raw-richness surface is computed identically as the baseline. The full pipeline is a Snakemake workflow (download → clean → analysis → figures) reproducible from a fresh checkout; the coverage estimators are validated standalone." . "Scope: whether standardising per-cell sample completeness removes observer-effort distortion of biodiversity-hotspot identity. The study tests whether coverage-standardising modern occurrence-derived per-cell bird richness makes the top-richest-cells (\"hotspot\") set agree with the hotspot set derived from an expert rangemap, in the same Iberian-bird, equal-area HEALPix system as the prior scale-dependence chain.\n\nIn scope: the top-5% richest-cells hotspot definition; the symmetric set non-overlap between the occurrence-derived hotspot set and the expert-rangemap-derived hotspot set; how that non-overlap changes when occurrence richness is coverage-standardised rather than left raw; both basis-of-record strategies (museum-grade specimens/sensors vs all observations including citizen science); the EU Article 12 expert polygons as the gold-standard rangemap comparator, with the historical occurrence-hull rangemap as a secondary comparator.\n\nOut of scope: the scale-dependence-across-resolutions question itself (that is the prior chain's claim, which this study takes as given); non-bird taxa; regions outside the Iberian peninsula; and any correction of the rangemap layer (only the occurrence/atlas layer is corrected)." . . . "Anne Fouilloux" . "2026-05-29T21:00:17.925Z"^^ . . . . . "Coverage-based effort correction of Iberian-bird richness hotspots (HEALPix-NESTED)" . . "RSA" . "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" . "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" . . . . . . . . . "Replication of Hurlbert & Jetz 2007 hotspot scale-dependence in Iberian birds on a HEALPix-NESTED substrate" . . . . . "Hurlbert & Jetz 2007 overlaid expert-drawn bird range maps (Handbook of the Birds of the World; regional atlases) on two structured regional\nbird-atlas surveys (Australia, southern Africa) on lat-lon grids at 0.25-8 degrees, circa 2007. This replication deviates on four axes:\n\n- Range-map layer: convex hulls of GBIF occurrences (a surrogate), cross-checked against EU Article 12 expert distribution polygons — not WWF / Handbook expert range maps.\n- Atlas layer: modern GBIF occurrence richness (post-2000, citizen-science-dominated for the all-observations strategy) — not a dedicated structured atlas survey with controlled effort.\n- Substrate: HEALPix-NESTED equal-area cells — not a lat-lon graticule, removing the cell-area distortion intrinsic to lat-lon grids.\n- Region and era: the Iberian peninsula, 2013-2024 GBIF / 2013-2018 Article 12 — not Australia or southern Africa, circa 2007.\n\nThe two-basis-of-record design (museum vs all-observations) is an addition to the original: it probes observer-effort sensitivity, a dimension unavailable to Hurlbert & Jetz before citizen-science data dominated occurrence records. The consequences of these deviations for the headline number are quantified in the linked Replication Outcome's Evidence and Limitations fields." . . "Iberian bird occurrence records (Spain, Portugal, Andorra, Gibraltar) were obtained from GBIF as two basis-of-record strategies, each with a citable download DOI: a \"museum\" strategy (PRESERVED_SPECIMEN + MACHINE_OBSERVATION; GBIF download DOI 10.15468/dl.r8pcat) and an \"all-observations\" strategy (additionally HUMAN_OBSERVATION, i.e. citizen-science records; GBIF download DOI 10.15468/dl.e9xv7p).\n\nRecords were year-split at 2000. Post-2000 occurrences form the atlas-equivalent layer (per-cell richness = number of distinct species observed in the cell). Pre-2000 occurrences form the range-map- equivalent layer (per species, the convex hull of its occurrences, whose cell coverage contributes to per-cell range-map richness). Both layers are binned onto a HEALPix-NESTED ladder at Nside 16, 32, 64, 128, 256 and 512 (cell side approximately 407 km down to 13 km, bracketing Hurlbert & Jetz's 0.25-8 degree range) using the geographic, WGS84-aware healpix-geo library, NESTED ordering throughout.\n\nAt each resolution, hotspots are the top-5% richest cells (Hurlbert & Jetz's definition); the headline metric is the symmetric non-overlap of the atlas-derived and range-map-derived hotspot sets. A Wilcoxon signed-rank test on the paired per-cell richness provides the dissolution criterion (statistical indistinguishability at P > 0.10, matching the original's >= 4 degree threshold).\n\nA verification battery tests robustness: a peninsula-only land mask (NaturalEarth 10m); a top-K hotspot-threshold sweep (1-25%); a per-species pre/post convex-hull drift measure (Jaccard distance); an atlas-richness-versus-observer-effort (per-cell record count) correlation; and two tighter range-map surrogates — a concave hull, and EU Birds Directive Article 12 expert distribution polygons (EEA, 2013-2018, 10 km grid, CC-BY 4.0; 260 Iberian breeding species) used as a gold-standard cross-check on the convex-hull surrogate.\n\nThe full pipeline is a Snakemake workflow (download -> clean -> analysis -> figures) reproducible from a fresh checkout; the verification battery lives in companion diagnostic notebooks." . "Scope: the scale-dependence of bird species-richness hotspot identification. The replication tests whether Hurlbert & Jetz's central finding — that range-map-derived richness hotspots disagree with finer-resolution survey-derived hotspots, with the disagreement growing as the grid is refined and dissolving only at coarse grain — re-emerges for a different region (the Iberian peninsula) under modern occurrence data and an equal-area sphere-aware substrate.\n\nIn scope: the top-5%-richest-cells hotspot definition; the symmetric non-overlap between range-map-derived and atlas-derived hotspot sets; how that non-overlap varies across a spatial-resolution ladder; and the coarse-grain resolution at which the two richness surfaces become statistically indistinguishable.\n\nOut of scope: Hurlbert & Jetz's secondary findings (per-species range occupancy / commission rates, and protected-area coverage of hotspots); and all non-bird taxa — the original claim and this replication are bird-only." . . . "Anne Fouilloux" . "2026-05-24T07:27:28.283Z"^^ . . . . . "Replication of Hurlbert & Jetz 2007 hotspot scale-dependence in Iberian birds on a HEALPix-NESTED substrate" . . "RSA" . "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" . "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" . . . . . . . . . "Stat-level reproduction of the Loveland et al. (2024) coupled ADCIRC+SWAN reduced-order source-term trade-off" . . . . . . "1. No ADCIRC+SWAN model re-runs. Loveland's runs were on 1064 Intel Xeon Platinum 8280 cores (19 nodes of TACC Frontera, \"Cascade Lake\"), with unstructured meshes of 6,675,517 elements / 3,352,598 nodes for Ike and 3,102,441 elements / 1,593,485 nodes for Ida. This compute scale is out of scope for a laptop / GitHub Actions / Docker reproduction. The replication therefore transcribes Loveland's published model-side outputs rather than regenerating them. The trade-off ratios reported in the Outcome's Evidence field are derived from Loveland's Table 4; the WSE-RMSE values are from Tables 5-6; the wave-statistics RMSE values are from Tables 5-7 and the §5.2 prose.\n\n2. DesignSafe deposit not retrievable. Loveland deposited their model inputs (meshes, fort.26 source-term configs, OWI Ike winds, HURDAT2-derived Ida GAHM winds, NOAA gauge / buoy time series, model output files) at DOI 10.17603/DS2-7HBT-EF65 (DesignSafe-CI project PRJ-4678). Both /api/projects/v2/PRJ-4678/ and /api/datafiles/listing/public/designsafe.storage.published/PRJ-4678/ return HTTP 401 to anonymous requests, so the deposit cannot be re-fetched by an unauthenticated reproducer. The fort.26 files for the Ida run are reprinted in the paper's Appendix on pp. 12-13.\n\n3. NDBC wave-buoy 42007 absent from the 2021 historical archive. Loveland's Table 3 lists 10 buoys; only 9 are retrievable for Ida from NDBC's URL pattern. The missing buoy is recorded in data/raw/sources.json.\n\n4. Per-storm wall-clock-cost framing granularity. Loveland's §5.1 prose (\"around 1.5 times longer\") and §6 prose (\"about a 40 percent increase\") cover Hurricane Ike (Gen3/Gen1 = 1.44×, Gen3/Gen2 = 1.52×) but understate Hurricane Ida (Gen3/Gen1 = 1.70×, Gen3/Gen2 = 1.76×) by approximately 20 percentage points relative to Loveland's own Table 4. This is not a deviation from method — the replication uses the same per-storm cells from Table 4 — but is surfaced by the Outcome because Loveland's prose summary lacks this per-storm distinction. See Outcome Limitations item 5." . . "This study is a stat-level (table-and-figure-level) reproduction, deliberately distinguishing what was independently re-derived from what was transcribed from the source paper. The downstream reader should not infer model-level verification.\n\nINDEPENDENTLY RE-DERIVED (notebooks/01_data_download.py + 02_data_clean.py):\n- Observational baseline. NOAA CO-OPS water-level gauge time series for the 14 Ike stations (Table 1 of the paper) over 5-14 September 2008 and the 13 Ida stations (Table 2) over 26 August - 4 September 2021, downloaded fresh from the public CO-OPS API. NDBC wave-buoy data (significant wave height, peak period, mean wave direction) for 10 buoys (Table 3), 9 of which were retrievable; buoy 42007 is absent from NDBC's 2021 historical archive at the URL pattern. Cleaned into tidy per-storm xarray Datasets per-variable, with per-variable resampling onto a common time grid (a naive nearest-reindex collapses NDBC's 10-min wind interleave into NaN-filled wave records; per-variable resampling avoids this).\n- Storm-peak consistency check. Per-gauge peak water levels were extracted (Galveston Pier 21 = 3.20 m on 13 September 2008; Grand Isle = 1.65 m on 29 August 2021) and verified against NHC historical reports as a sanity check on the observational baseline Loveland modelled against.\n\nTRANSCRIBED FROM THE SOURCE PAPER (notebooks/03_analysis.py, data/published_baselines/):\n- All model-side outputs. Run times (Table 4) for each (storm, config) cell, WSE-RMSE per gauge (Tables 5-6 and the prose summaries in §5.3 of the paper), Hs / Tp / mean-wave-direction RMSE per buoy (Tables 5-7 transcribed from §5.2). These values come from Loveland's Frontera-scale ADCIRC+SWAN runs and were not re-computed.\n- Source-term configuration files. The fort.26 files printed in the paper's Appendix (pp. 12-13) were inspected for documentation but not used to drive any new SWAN runs.\n- Spatial figures. The paper's Fig. 9 (spatial Hs differences near hurricane tracks) and Fig. 12-13 (spatial WSE differences) were inspected for qualitative context but not regenerated.\n\nWhat the comparison therefore tests: the internal consistency of Loveland's published model-vs-observation Δ values with the publicly retrievable observational baseline, and the per-storm trade-off ratios visible in Loveland's own Table 4. Not the reproducibility of the model runs themselves.\n\nThe headline statistic (Gen3 / (Gen1 or Gen2) wall-clock ratio per storm, and maximum WSE-RMSE Δ across source-term configurations per storm) is consolidated in results/headline_comparison.csv and visualised in figures/main_result.png. Orchestrated via Snakemake (pipeline rules in Snakefile); reproducible environment via pixi (pixi.toml + pixi.lock); container build via Dockerfile to ghcr.io/annefou/coastal-rom-replication." . "In scope: the conditional trade-off claim from Loveland's §6 Conclusions as carried verbatim by the Quote — that reduced-order SWAN source terms (Gen1 first-generation, Gen2 second-generation) save computation relative to the third-generation ST6 Gen3 package without compromising water-surface-elevation (WSE) accuracy at NOAA gauges when WSE is of primary interest, with Loveland's own quantified caveat that large source-term sensitivities are observed in significant-wave-height fields near hurricane tracks. The two storm scenarios (Hurricane Ike 2008, Hurricane Ida 2021) and the four model configurations (No SWAN, Gen1, Gen2, Gen3) are tested in full.\n\nOut of scope: (1) operational forecasting where wind-field uncertainty dominates the error budget — a second conditioning Loveland states in §6 (\"if the meteorological forcing is not sufficiently accurate ... the additional computational cost associated with the detailed Gen3 source terms may not improve accuracy of the model\") that the chosen Quote does not carry; the Outcome's Validated label therefore applies to the hindcasting regime only. (2) Model-level reproducibility of the ADCIRC+SWAN runs themselves — see Methodology field for what was and was not independently re-derived. (3) Generalisation beyond the Gulf of Mexico or beyond Loveland's two test storms." . . . "Anne Fouilloux" . "2026-05-22T21:06:19.667Z"^^ . . . . . "Stat-level reproduction of the Loveland et al. (2024) coupled ADCIRC+SWAN reduced-order source-term trade-off" . . "RSA" . "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAoDcOiD+jen8awiJ6DB2ewDw66PeG64hODmgNFwy7GrwQui4HKnHdvxd++1UhTgiOfycxyxBb7sXPSikLw/1TsSyPsEl0P3/+600szxpTGgLNzW+bZ2DVP3d8ERMV1aWpH0ci3B/5vmK+vXQZ4uCoq57NE0MiFg5c13Gy0gd6n7wZYEhYM4AjWSLL0QS/HY+TFZMYL9bCFeATennGrlB2UEjRlw21UB2Ah16ZZ6hxQlfctFJZE7TGnBJPB3ttTjfcOfamhjZVwQ0yV9mv7x6PGiSmkzpJTVLjn8hagoKT05YUwVQArFb+w7f6sXqvvljMigjd/Rbqgbye/lLUAZLfJSnFM58TubfpEJvXV4zNMDEoT3VQ7dokgoLgMrmjZCKATtQ7gomocoTJ1NhN2esRNtGzWaS2obL/mueUQlMlavssZnqL8WICkdAuDlwDVNbsbwEWKQ50kiPdAdduSigifxA4CM7TgvnxqZVoAResEGP6UhTTem3T4CsbEas1Caj9wa7M1jPjACu5LF5BwcVns3ZQHWLipjRjD+9/ur3G8QtuxbNhmXlDYQ6tXxB1lK+Oz7O519b3bA15ilzFl0SdvMBGTe46xaQ9DsJT18THKnPbUhNMy0dH0VtzpB+EEaXZ25Fp9VHMEUqo1lLS9e89eO3efiqkESKQ7wmB+/DlIRcCAwEAAQ==" . "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" . . . . . . . . . "Iberian Lacertidae h_r mechanism replication under Destination Earth Climate DT SSP3-7.0 (2020-2039)" . . . . . . . "Six explicit deviations from Sinervo 2010, ordered by interpretive significance:\n\n(1) Climate forcing. DestinE Climate Digital Twin SSP3-7.0 IFS-NEMO at HEALPix nside=128 (~46 km), 2020-2039. Sinervo used WorldClim 1.4 at 10 arc-minute resolution interpolated between 1975 and 2020 baselines and projected to 2050 / 2080 under IPCC AR3 scenarios (CCCMA / HADCM3 / CSIRO a2a). The DestinE archive's near-term-only horizons mean Sinervo's headline 24 % / 46 % Lacertidae projections for 2050 / 2080 cannot be tested directly — only the mechanism's behaviour at 2020s and 2030s horizons.\n\n(2) Taxon. Iberian Lacertidae (47 species across 264 presence cells in the Iberian Peninsula). Sinervo's empirical calibration used 48 Mexican Sceloporus (Phrynosomatidae) at 200 sites with a 1975 baseline census; the global projection used 587 species across 34 families with the Lacertidae row of Table 1 (n=89 records, n_spp=279) as the family parameterisation we inherit.\n\n(3) Thermal-physiology priors. Used the family-level Lacertidae Table 1 row (T_b = 35.4 +/- 0.31 °C, h_r threshold = 3.1 h, heliothermic mode) as the baseline. Iberian-specific T_b records from the Spanish herpetological literature (Carretero, Monasterio et al.) were not incorporated in v0.1.0 — explicitly deferred to v0.2.0. A sensitivity branch substitutes Iberolacerta-style cool-adapted T_b = 31 °C to test prior-conditional behaviour.\n\n(4) Reproductive-window choice. Default April-May (the conservative single-point estimate for Iberian temperate-zone Lacertidae). Sinervo's Sceloporus calibration used March-April; SOM page 4 specifies \"2 months in the temperate zone\" generically without an Iberian-specific source. Sensitivity branches test May-June (closer to actual Iberian Podarcis / Iberolacerta breeding chronology) and April-June (extended-window dilution).\n\n(5) Diurnal-cycle reconstruction. Not applied. SOM Equation S2 is parameterised in daily T_max; DestinE 6-hourly snapshots are aggregated to daily max during the data-clean step, then fed directly to the h_r formula. A sub-daily Polytope-access probe in the data-download step of the cited Research Software returned \"credentials absent\" (recorded in the results-registry artefact); the sinusoidal Tmin-Tmax diurnal-cycle reconstruction scoped as a fallback was not needed because the mechanism is daily-Tmax-driven, not hourly-integration-driven.\n\n(6) Spatial substrate. HEALPix-NESTED nside=128 throughout, matching Bombus chain-2 substrate. A substrate-sensitivity diagnostic mirroring Bombus chain 3 runs the same h_r computation at HEALPix nside=64 (downsampled via NESTED bit-shift) under the S3a config — the only sensitivity-matrix config that produces non-zero rates and is therefore the only config where the substrate comparison is non-degenerate. Sinervo's spatial substrate is the WorldClim 10 arc-minute lat-lon grid; this is not a like-for-like substrate comparison but is documented to enable interpretation of the substrate-sensitivity diagnostic across the wider FORRT constellation." . . "A four-step Snakemake pipeline (data download → data clean → analysis → figures) implemented in the cited Research Software nanopub. Inputs: (a) Destination Earth Climate DT t2m field, daily-aggregated from 4-times-daily IFS-NEMO snapshots over 2020-2039 SSP3-7.0 at HEALPix nside=128 NESTED ordering (DestinE-native, no re-projection; ~25 GB total, retrieved via Polytope); (b) CRU TS 3.24.01 historical Tmax / Tmin NetCDF (Soroye 2020 Figshare deposit 10.6084/m9.figshare.9956471; reused as a diurnal-cycle baseline); (c) GBIF Lacertidae × Iberia occurrence download (DOI 10.15468/dl.rh4rfn; 136,210 records). The Iberian HEALPix cell mask is derived from the global nside=128 grid by spatial subset to bbox (-10°, 35°, 4°, 44°) yielding 479 cells. Occurrences are assigned to cells via the geographic-HEALPix `lonlat_to_healpix` operation with NESTED ordering. h_r is computed per cell × per day under Sinervo SOM Equation S2 with conditional clamping: `h_r = (0.74 * (T_max − T_b) + 6.1) if T_max > T_b else 0` — the bracket notation `h_r[T_e > T_b_preferred]` in the SOM is a conditional, not a max-zero clamp, so daily T_max ≤ T_b means no refuge needed and h_r = 0. Daily-mean h_r is averaged over the reproductive window per year and compared against the family-calibrated threshold of 3.1 h. A cell-year is flagged as locally extinct when daily-mean h_r > 3.1 h. Per-species local-extinction rate is the fraction of presence cells flagged in each decade (2020-2029, 2030-2039). A six-config sensitivity matrix (2 T_b values × 3 windows) and a substrate-sensitivity diagnostic comparing per-species rankings at HEALPix nside=128 vs nside=64 (downsampled via the NESTED bit-shift parent = pix >> 2) accompany the headline result. Intermediate artefacts are persisted in self-describing formats (NetCDF for arrays, Parquet for tabular). Headline numbers and the full sensitivity matrix are recorded in the cited Research Software's results-registry artefact for downstream consumption by the Outcome nanopub." . "The transferability of the Sinervo 2010 h_r mechanism (cumulative daily hours of activity restriction during the critical reproductive period predicting local extinction in heliothermic lizards) to the Iberian Lacertidae assemblage at near-term horizons reachable under the current Destination Earth Climate Digital Twin archive (2020-2039 under SSP3-7.0). In scope: all 47 Iberian Lacertidae species with at least one georeferenced occurrence record in the GBIF download 10.15468/dl.rh4rfn (Spain, Portugal, Andorra, Gibraltar; year >= 1900; hasCoordinate = TRUE; hasGeospatialIssue = FALSE; basisOfRecord IN HUMAN_OBSERVATION/PRESERVED_SPECIMEN/MACHINE_OBSERVATION); the family-calibrated h_r threshold from Sinervo Table 1 (3.1 h for Lacertidae); two T_b prior choices (family mean 35.4 °C from Table 1, Iberolacerta-style cool-adapted 31 °C); three reproductive-window choices (April-May, May-June, April-June); two HEALPix substrates (nside=128 DestinE-native, nside=64 substrate-sensitivity diagnostic). Out of scope: Sinervo's headline 2050 / 2080 projections (not reachable under current DestinE archive); non-heliothermic Iberian lizard families (Gekkonidae, Anguidae, Scincidae — the basking-window mechanism is calibrated for diurnal heliotherms only); species-specific T_b refinement using Iberian-specific records from Carretero / Monasterio / Spanish herpetological literature (deferred to a v0.2.0 iteration)." . . . "Anne Fouilloux" . "2026-05-18T07:21:47.050Z"^^ . . . . . "Iberian Lacertidae h_r mechanism replication under Destination Earth Climate DT SSP3-7.0 (2020-2039)" . . "RSA" . "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" . "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" . . . . . . . . . "Cross-substrate substrate-sensitivity diagnostic for TEI-based extirpation projection (Iberian Bombus, HEALPix nside=64 vs nside=128, SSP3-7.0)" . . . . . "1. Different methodological character. Soroye et al. 2020 fit one GLMM at the CEA grid and reported a continental-scale extirpation summary. This Study does NOT refit the GLMM — it uses the substrate-fit GLMMs from the two upstream replications and asks how their per-species PROJECTION rankings concord across substrates. The substrate-coupling diagnostic is downstream of Soroye's mechanism, not a direct replication of it.\n\n2. Variant (c) tested as approximation. The most rigorous test of \"shared reference standardisation alone fixes substrate coupling\" is to refit the GLMM with shared (μ, σ) and re-project. Variant (c) here uses substrate-local β with shared-σ predictors at projection time — mathematically equivalent to a constant scaling of η, refuting the hypothesis as a standalone fix BUT not the same as testing a full refit. Documented in the Outcome's Limitations as a deferred follow-up.\n\n3. Substrate-invariant physical metrics added. Variants (d) and (d2) compute per-species mean future TEI and fraction TEI_future>0.5 — physical-unit metrics that bypass the GLMM entirely. These are NOT in Soroye's original analysis; they are introduced here as a substrate-invariant cross-check." . . "METHOD\n\nThe diagnostic reads input artefacts from two upstream substrate replications (the canonical nside=64 sibling and the nside=128 substrate-extension sibling) and computes five candidate per-species rankings at each substrate, plus a mechanistic per-species η decomposition.\n\nInputs (per substrate; symlinked from sibling repos in development, fetched from Zenodo in v0.2.0):\n - GLMM coefficient posterior (variational-Bayes summary): healpix_port/outputs_iberia/posterior_vb_summary.csv\n - Species random intercepts (full NUTS posterior): results/posterior_bambi_healpix.nc\n - Substrate-local predictor scaling (μ, σ): healpix_port/outputs_iberia/dataGLMM_extinction.parquet\n - Per-species niche limits T_min_spp, T_max_spp, P_min_spp, P_max_spp: healpix_port/outputs_iberia/climate_tei_pei_healpix.nc\n - Future-period predictors (per species per cell): climate_tei_pei_future__healpix.nc\n - Per-species observation mask: presence_absence_healpix.nc\n - Per-cell sampling effort: sampling_continent_healpix.nc\n\nFive projection variants (notebooks/02_decompose.py and 03_variants.py):\n (a) Full GLMM η, within-substrate predictor standardisation.\n (b) Main-effects-only η, within-substrate standardisation (drop the four interaction terms at projection only; keep them in the fit).\n (c) Full GLMM η, shared CEA reference standardisation (substrate-fit β + shared (μ, σ) computed from the original Soroye CEA pool; tested as a refit-equivalent approximation).\n (d) Mean future TEI per species (substrate-invariant physical metric, no GLMM).\n (d2) Fraction of cells with future TEI > 0.5 per species (substrate-invariant threshold metric).\n\nFor each variant, compute per-species community-mean η (or fraction) across the species' currently-occupied + active cells, at both substrates and both horizons. Apply four n_cells filters (≥1, ≥5, ≥10, ≥20). Compute Spearman rank correlation between nside=64 and nside=128 rankings under each (variant, filter) combination.\n\nMechanistic decomposition (scripts/compare_substrates.py): for a diagnostic set of 12 species spanning the n_cells distribution, decompose per-species η into its 10 GLMM-term contributions plus the species random intercept at each substrate. Identify which term(s) are responsible for substrate-coupling.\n\nCode: scripts/compare_substrates.py, scripts/compare_variants.py, scripts/plot_variant_concordance.py. Notebooks: 01_inputs_fetch.py, 02_decompose.py, 03_variants.py, 04_figures.py." . "SCOPE: the cross-substrate concordance of per-species ranking from Soroye et al. 2020's TEI-based GLMM, when projected to SSP3-7.0 future climate from two single-substrate Iberian Bombus replications.\n\nIN SCOPE\n - The TWO substrates that the upstream replications already published as GitHub releases + Zenodo deposits: HEALPix nside=64 (~92 km, weatherxbiodiversity-projection v0.1.0) and HEALPix nside=128 (~46 km, weatherxbiodiversity-projection-nside128 v0.1.0).\n - The TWO horizons that DestinE Climate DT SSP3-7.0 has populated: 2020–2029 and 2030–2039.\n - FIVE projection variants (full GLMM η, main-effects-only η, shared-CEA-reference η, mean future TEI, fraction TEI_future > 0.5).\n - FOUR per-species n_cells filters (≥1, ≥5, ≥10, ≥20).\n - Mechanistic decomposition of per-species η into 10 GLMM-term contributions plus the species random intercept, at both substrates, for a diagnostic set of 12 species spanning the n_cells distribution.\n\nOUT OF SCOPE\n - Other substrates (CEA, EASE-Grid 2.0, S2, additional HEALPix levels) — flagged in the Outcome's Limitations.\n - Other species or regions — Iberian Bombus only.\n - Other future climate forcings — SSP3-7.0 only.\n - End-of-century horizons (2046–2055, 2076–2085) — DestinE Climate DT archive unavailable past 2039 at time of analysis." . . . "Anne Fouilloux" . "2026-05-11T20:20:36.215Z"^^ . . . . . "Cross-substrate substrate-sensitivity diagnostic for TEI-based extirpation projection (Iberian Bombus, HEALPix nside=64 vs nside=128, SSP3-7.0)" . . "RSA" . "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" . "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" . . . . . . . . . "TEI mechanism — HEALPix nside=128 substrate extension on Iberian Bombus (full GLMM refit at native DestinE Climate DT pixelisation)" . . . . "1. Pixelisation. Soroye 2020 fit at the CEA grid (~100 km, equal-area cylindrical). This Study fits at HEALPix-NESTED nside=128 on the WGS84 ellipsoid (~46 km). Cell area is ~4× smaller than Soroye's; cell shape and topology differ. This is the SAME deviation kind as the canonical nside=64 sibling, just at a finer resolution.\n\n2. Region. Iberian peninsula only — same as the canonical sibling.\n\n3. Inference engine. Two independent Python implementations (statsmodels VB, bambi/PyMC NUTS) — same as the canonical sibling.\n\n4. Prior choices. Bambi defaults (Half-StudentT on group SDs) rather than Soroye's MCMCglmm informative inverse-Wishart — same deviation as the canonical sibling.\n\n5. Climate inputs. Soroye's bundled CRU TS 3.24.01 from his Figshare deposit, used unchanged — NOT a deviation.\n\n6. Tier 2 projection grid alignment. Unlike the canonical nside=64 sibling, this Study fits AND projects at the same native HEALPix nside=128 substrate — no parent-cell aggregation between fit and projection grids. This eliminates one source of cross-substrate aggregation noise but introduces a finer-grained per-cell extrapolation tail (more cells lie far outside the training distribution per species)." . . "METHOD\n\nThe methodology mirrors the canonical nside=64 sibling exactly except for the spatial substrate. Cell coverage and per-species niche limits are computed on HEALPix-NESTED nside=128 cells of the WGS84 ellipsoid (using the healpix-geo Python library); per-species niche limits and the GLMM are refit at this substrate.\n\nGLMM specification (identical to Soroye 2020 and to the canonical sibling):\n extinction ~ continent + sc_sampling + sc_TEI_bs + sc_TEI_delta + sc_TEI_bs:sc_TEI_delta + sc_PEI_bs + sc_PEI_delta + sc_PEI_bs:sc_PEI_delta + sc_TEI_bs:sc_PEI_bs + sc_TEI_delta:sc_PEI_delta + (1|species)\n\nInference: variational-Bayes via statsmodels.BinomialBayesMixedGLM (fast first pass) and full-posterior NUTS via bambi/PyMC, 4 chains × 2000 samples (authoritative HDIs).\n\nTier 2 — SSP3-7.0 future projection. DestinE Climate DT IFS-NEMO standard SSP3-7.0 GRIB files retrieved via polytope on LUMI for the 2020–2029 and 2030–2039 horizons, decoded with eccodes (Python API, NESTED-aware), subset to pre-computed Iberian HEALPix nside=128 cells. The future-period TEI_delta and PEI_delta are computed on the SAME substrate the GLMM was fit on (no cross-substrate aggregation step). Per-species ranking is reported following the protocol established in the methodological sibling chain (n_cells ≥ 10, main-effects-only η at projection time)." . "SCOPE: the GLMM coefficient on standardised TEI_delta at HEALPix nside=128 (the native pixelisation of DestinE Climate DT IFS-NEMO standard), Iberian Bombus only.\n\nIN SCOPE\n - Soroye 2020's GLMM specification (identical to the canonical nside=64 sibling).\n - Soroye 2020's CRU TS 3.24.01 climate inputs (identical, kept unchanged).\n - HEALPix-NESTED nside=128 on the WGS84 ellipsoid (~46 km cells; the native DestinE Climate DT pixelisation).\n - Tier 1 historical fit on the 1901–1974 baseline period and 2000–2014 recent period.\n - Tier 2 SSP3-7.0 future projection at substrate-matched nside=128 (no parent-aggregation deviation between fit and projection grids).\n\nOUT OF SCOPE for this Replication Study (handled by separate chains)\n - The canonical CEA + nside=64 substrate-comparison at coarser resolution — see weatherxbiodiversity-projection.\n - Cross-substrate methodological diagnostic — see weatherxbiodiversity-substrate-sensitivity.\n - Bombus species outside the Iberian peninsula." . . . "Anne Fouilloux" . "2026-05-11T19:28:01.754Z"^^ . . . . . "TEI mechanism — HEALPix nside=128 substrate extension on Iberian Bombus (full GLMM refit at native DestinE Climate DT pixelisation)" . . "RSA" . "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" . "fC6jI5oFeW5OCJ/TcgTWnMi31rpsDVp6NbbjNY147s7xMf10on9R0BwiAUjPTPUb0kipCvr5IginpfRarBFEIuF9aEVKlklwup3LSwX7apTUQavH0MXY7sh3ARVNry36aOGH9IuDDGieOrQzobU8TloDPa7YOYXJI4IJL0wigzfvRCryRCklr3oaJHep2OWW3if01wVne+Jo8mRDN4J9MIevxI0grn8gHRJ9BitaIaB0dFKsvPQdUinAUcbPeY+pfoF9AHls3dLrMrEn6HtPFta5DfciNLarQCBnoXcJFWMKWlS4U3ZWUXs8lQ594LSH9EUpUE4FOqQoRO6v5coGVM7nEfXM6N/xVKyWKqTT06TBUi5a+CzOhGoWc526Qhdq9TCQk5988ZDYsRKLeElvKZwXOswgsjgPJ1OT7ASdXNbBTo/re8cScRdivPvl5Nnjo/RkRtWyFp58YhLyFniUIx975RWBRqMHZi9JvQYPCSj7ASAxg3zNUoUdnIlMdjzCCzAbqqnE2NVgrXv+HtsB0gAdOZ1saU7JJ+iddl9nEtbwJ27StsfUKltRV/2DwR+FSLptIf9oKAzMVLs6DpFftTohmoOlOxHNC8ylsUhuMYYHHuHSi5t+W7VralYHXkmwHjnWB8RPvCc9oiQOSCLddqP2yO6kudwOTpO2bqAnA1g=" . . . . . . . . . "TEI mechanism replicated on Iberian Bombus — CEA grid + HEALPix nside=64 substrate extension" . . . . . "1. Pixelisation. Soroye 2020 fit at the CEA grid (~100 km, equal-area cylindrical projection at the equator). This Study additionally fits at HEALPix-NESTED nside=64 (~92 km cells on the WGS84 ellipsoid). Cell area is comparable (within ~10%); cell shape and topology differ.\n\n2. Region. Soroye 2020 fit on North America + whole Europe (all Western Palearctic Bombus). This Study fits on Iberian peninsula only. Iberia is a subset of Soroye's European fitting domain; the same GLMM specification applies.\n\n3. Inference engine. Soroye 2020 used MCMCglmm (R). This Study uses two independent Python implementations (statsmodels variational-Bayes for the fast first pass; bambi/PyMC NUTS for the authoritative posterior HDIs reported in the Outcome). Both target the same Bayesian posterior.\n\n4. Prior choices. Soroye 2020's MCMCglmm priors (default informative inverse-Wishart on variance components) are not exactly reproducible across to bambi (Half-StudentT default on group SDs). This is the largest non-trivial methodological deviation; it should not affect the headline coefficient sign or order of magnitude (verified: it does not). The full prior table is in notebooks/03h_analysis_healpix.py.\n\n5. Climate inputs. Soroye 2020's CRU TS 3.24.01 from his Figshare deposit is used unchanged — NOT a deviation, this is the SAME climate dataset.\n\n6. Sampling-effort proxy. Same as Soroye (cells_per_decade × records_per_cell from the cleaned occurrence table) — NOT a deviation.\n" . . "METHOD\n\nTier 1 — historical fit.\n\n(1) Climate inputs. Soroye's bundled CRU TS 3.24.01 NetCDFs (mean monthly temperature and total monthly precipitation, 1901–2014, 0.5° lat-lon grid) are downloaded from Soroye's Figshare deposit and used unchanged. No substitution with cdsapi or other reanalysis sources at fit time — by design, to keep the climate forcing identical to the original.\n\n(2) Occurrence inputs. GBIF Iberian Bombus occurrences are issued via own-DOI download (cited in the Outcome step 05 \"evidence\" field). Filtering follows Soroye's species list (genus Bombus, valid taxonomy, occurrences geo-referenced to the Iberian peninsula bounding box).\n\n(3) GLMM specification. Replicates Soroye's exact formula:\n extinction ~ continent + sc_sampling + sc_TEI_bs + sc_TEI_delta + sc_TEI_bs:sc_TEI_delta + sc_PEI_bs + sc_PEI_delta + sc_PEI_bs:sc_PEI_delta + sc_TEI_bs:sc_PEI_bs + sc_TEI_delta:sc_PEI_delta + (1|species)\nwhere TEI = (T_obs − T_min_spp) / (T_max_spp − T_min_spp), PEI is the analogous precipitation index, and sc_ is within-substrate z-scoring.\n\n(4) Two spatial substrates.\n Pass 1 (CEA): cell coverage and per-species niche limits computed on Soroye's original CEA grid (~100 km cells); GLMM fit on this grid.\n Pass 2 (HEALPix nside=64): cell coverage re-computed on HEALPix-NESTED nside=64 cells of the WGS84 ellipsoid (using the healpix-geo Python library); per-species niche limits and the GLMM are refit at this substrate.\n\n(5) Inference. Two independent fitting strategies on the same Pass-2 design matrix:\n (a) variational-Bayes mean-field via statsmodels.BinomialBayesMixedGLM (fast first pass);\n (b) full-posterior NUTS via bambi/PyMC, 4 chains × 2000 samples (authoritative HDIs).\n\nTier 2 — SSP3-7.0 future projection — runs on the DestinE Jupyter platform and is part of this same chain's Outcome (step 05), not a separate Study. Briefly: DestinE Climate DT IFS-NEMO standard SSP3-7.0 GRIB files are retrieved via polytope on LUMI for the 2020–2029 and 2030–2039 horizons, decoded with eccodes (Python API, NESTED-aware), subset to pre-computed Iberian HEALPix cells, aggregated to monthly means matching the CRU TS units, and used to compute future TEI_delta and PEI_delta per species per cell. Per-species ranking is reported following the protocol established in the methodological sibling chain.\n\nCode: notebooks/01_data_download.py (CRU TS + GBIF), notebooks/02_data_clean.py (CEA Pass 1 cleaning), notebooks/02h_data_clean_healpix.py (HEALPix Pass 2 cleaning), notebooks/03_analysis.py (CEA fit), notebooks/03h_analysis_healpix.py (HEALPix fit), notebooks/04_figures.py + 04h_figures_healpix.py (figures)." . "SCOPE: the GLMM coefficient on standardised TEI_delta (Soroye et al. 2020's \"climatic position index change\" between baseline 1901–1974 and recent 2000–2014), restricted to Bombus species observed on the Iberian peninsula in GBIF.\n\nIN SCOPE\n - Soroye 2020's GLMM specification: extinction ~ continent + sc_sampling + sc_TEI_bs + sc_TEI_delta + sc_PEI_bs + sc_PEI_delta + (TEI:PEI interactions) + (1|species).\n - Soroye 2020's CRU TS 3.24.01 climate inputs (Figshare bundle, identical NetCDFs).\n - Two spatial substrates: original CEA grid (~100 km cells, Pass 1) and HEALPix-NESTED nside=64 on the WGS84 ellipsoid (~92 km cells, Pass 2).\n - Tier 1 historical fit on the 1901–1974 baseline period and 2000–2014 recent period.\n\nOUT OF SCOPE for this Replication Study (handled by separate chains)\n - Higher-resolution substrates (HEALPix nside=128) — see weatherxbiodiversity-projection-nside128.\n - Cross-substrate methodological diagnostic (substrate-coupling at projection time) — see weatherxbiodiversity-substrate-sensitivity.\n - Tier 2 SSP3-7.0 future projection — reported as a separate Outcome contribution within this same chain (see step 05), not as a separate Study.\n - Bombus species outside the Iberian peninsula (Soroye's original ranges in North America + whole-Europe)." . . . "Anne Fouilloux" . "2026-05-11T18:27:02.619Z"^^ . . . . . "TEI mechanism replicated on Iberian Bombus — CEA grid + HEALPix nside=64 substrate extension" . . "RSA" . "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" . "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" . . . . . . . . . "Substrate-aware multi-taxon biodiversity-exposure overlay for the 2011 Western Australian Ningaloo Niño event on HEALPix-NESTED nside=128" . . . . . . . . . . "Wernberg et al. 2016 sampled the kelp-forest community via diver-transect surveys at 65 reefs across a ~2,000 km tropical-to-temperate transition zone in Western Australia, between 2001 and 2015. This study uses GBIF marine occurrence records as the data source (different sampling — opportunistic occurrence vs structured transect; different scope — broader marine taxa, not just kelp) and a shared HEALPix-NESTED substrate for spatial overlap with the MHW footprint (different aggregation method). The MHW threshold uses a simplified Hobday-style rule (fixed anomaly threshold relative to a 3-year per-DOY climatology + minimum-duration persistence) rather than the canonical Hobday 2016 90th-percentile-per-DOY-over-30-years threshold; the qualitative spatial footprint matches the documented Ningaloo Niño event." . . "NOAA OISST v2.1 daily SST for the Western Australian region in 2011 is aggregated onto a HEALPix-NESTED substrate at WGS84 ellipsoidal projection via healpix-geo and healpix-plot. Marine-heatwave detection follows a Hobday-style rule: anomaly above a per-day-of-year reference climatology, fixed temperature threshold, and minimum-duration persistence. The reference climatology is built from yearly OISST files across a multi-year baseline preceding the event year. The HEALPix aggregation weights only OISST sea cells (using OISST's native sea/land mask) so coastal cells are not diluted by land contributions. Marine biodiversity occurrences are pulled from GBIF for the same region and year, restricted to taxonKeys for unambiguously-marine groups (Elasmobranchii, Cephalopoda, Cnidaria, Echinodermata, Porifera) and to HEALPix cells with OISST sea coverage. Each occurrence is mapped to its HEALPix cell and tagged as MHW-exposed if the cell has at least one MHW-day. The full pipeline is reproducible via the repository's environment.yml and Snakefile." . "We test the multi-taxon spatial-temporal correspondence between Wernberg et al.'s documented 2011 Western Australian marine-heatwave regime-shift event and contemporaneous marine biodiversity occurrences in the same region. Scope: extending the original work's kelp-forest evidence to all marine taxa available in GBIF for the same region and year, on a shared HEALPix-NESTED substrate that the climate-event field and the biodiversity field both live on. The original work's mechanistic claims about kelp regime-shift dynamics (range contraction, turf dominance, recovery suppression) are out of scope here — we are not making a new claim about kelp dynamics, only on the broader marine-biodiversity-exposure footprint of the same MHW event. " . . . "Anne Fouilloux" . "2026-05-08T19:36:58.651Z"^^ . . . . . "Substrate-aware multi-taxon biodiversity-exposure overlay for the 2011 Western Australian Ningaloo Niño event on HEALPix-NESTED nside=128" . . "RSA" . "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" . "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" . . . . . . . . . "Cross-discipline transfer of sphere-harmonic vs lat-lon-flat matched filters between cosmology-like and climate-like synthetic domains on HEALPix-NESTED " . . . . . . . "First in-repository demonstration of cross-discipline transfer with this matched-filter pair; no prior single-paper precedent for this specific experiment. The two synthetic regimes are constructed to share feature physics across different background spectra so the substrate effect is isolated from the model class." . . "Two synthetic discipline regimes are generated as Gaussian random fields directly on HEALPix-NESTED, distinguished by their angular power spectra and an optional latitudinal baseline. The first regime represents a cosmology-like domain (steeper power spectrum, no latitudinal baseline, features placed at uniformly-random sphere locations); the second represents a climate-like domain (shallower power spectrum, cosine-of-latitude baseline approximating an SST equator-pole gradient, features confined to high latitudes). The same physical feature physics — same angular radius and same amplitude — is used in both. The sphere-aware pipeline applies a sphere-harmonic band-pass matched filter (high-pass to suppress the latitudinal baseline plus a Gaussian beam matched to the feature scale) and reads out the maximum, mean, and standard deviation of the response field. The lat-lon-flat baseline cross-correlates the equirectangular projection of the same data with a feature-shape template at the equator and reads out the same three statistics. Both feed identical logistic-regression heads. Each pipeline is trained on the first regime only and evaluated three ways: in-domain on the first regime (sanity), cross-domain on the second regime without retraining (the headline test), and in-domain on the second regime when trained directly on it (upper bound). The full pipeline is reproducible via the repository's environment.yml and Snakefile." . "We test the cross-discipline transfer aspect of the substrate-dependence claim. Scope: two synthetic discipline regimes that share the same physical feature physics (same compact-feature size and amplitude) but differ in their stochastic background and in the latitudinal distribution of where features appear. The classifier is trained on the first discipline only and applied to the second discipline without retraining; the comparator is a lat-lon-flat baseline trained identically on the same first-discipline data. The within-discipline regime (same feature distribution at training and test) is out of scope here." . . . "Anne Fouilloux" . "2026-05-07T21:54:59.055Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Sphere vs flat matched-filter pipelines for marine-heatwave detection on synthetic global SST on HEALPix-NESTED" . . . "First end-to-end demonstration in this repository of the substrate-dependence with this matched-filter pair; no prior single-paper implementation. The within-discipline failure mode is documented qualitatively in DeepSphere and DLWP-HEALPix follow-ups but not quantified at this granularity with this minimal feature pair. " . . "Synthetic global SST samples are generated on a 1° lat-lon raster as a cosine-of-latitude baseline plus Gaussian noise, with optional marine-heatwave events injected as fixed-radius spherical caps with a positive temperature anomaly at uniform-random longitude and a chosen latitude band. The sphere-aware pipeline aggregates each sample onto a HEALPix-NESTED grid and applies a sphere-harmonic band-pass matched filter — a high-pass that suppresses the cosine-of-latitude SST baseline followed by a Gaussian beam matched to the cap diameter — then reads out the maximum, mean, and standard deviation of the inverse-spherical-harmonic-transform response field. The lat-lon-flat baseline cross-correlates the same SST anomaly with an equator-shape cap template and reads out the same three statistics. Both feed identical logistic-regression classifier heads. The classifiers are trained on samples with events at low latitudes only and evaluated at four progressively higher test latitude bands. The full pipeline is reproducible via the repository's environment.yml and Snakefile. " . "We test the latitude-invariance aspect of the claim. Scope: the within-discipline regime — same task (binary marine-heatwave detection), same data distribution (synthetic global SST), same training set (events confined to low latitudes only). The variable being swept is the test latitude band: we evaluate the trained classifier at progressively higher latitudes to see whether the substrate-aware pipeline preserves detection accuracy where the lat-lon-flat baseline loses it." . . . "Anne Fouilloux" . "2026-05-06T20:23:13.819Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Multi-criteria evaluation of DGGS substrates for the integration of biodiversity occurrence data with Copernicus EO and Destination Earth climate-model output" . . . . . "No prior implementation of this multi-criteria DGGS-substrate comparison for biodiversity × Copernicus × Destination Earth integration exists in the literature. The foundational references — Górski et al. 2005 (HEALPix), Sahr et al. 2003 (DGGS framework), Hauffe et al. 2023 (Eco-ISEA3H), Birch et al. 2007 (equal-area for ecology), Kmoch et al. 2022 (DGGS area distortions) — advocate equal-area or DGGS but without quantitative multi-grid comparison and without attention to ML-ecosystem compatibility or ellipsoidal correctness for the integrated regime. This study operationalises the comparison and establishes the baseline." . . "Eight reproducible Jupyter notebooks aggregate 20,100 Quercus suber GBIF occurrences and synthetic uniform sphere data onto nine grid systems: lat-lon (cautionary), Behrmann, Mollweide, EEA reference grid (LAEA Europe / EPSG:3035), HEALPix-on-sphere, HEALPix-on-WGS84 via authalic-sphere mapping (the GRID4EARTH approach, healpix-geo), rHEALPix, H3, and ISEA3H (DGGRID v8.41 via dggrid4py). Each notebook isolates one fitness criterion: count-bias measurement (notebooks 01–02, 07), cell-shape anisotropy across latitudes (03), 3×3 ML-kernel locality at 65°N, 15°E (04), 3-grid synthetic comparison (05), HEALPix NESTED hierarchical refinement via bit-shift (06), and HEALPix-specific properties — sphere-vs-WGS84 systematic area error via pyproj.Geod, NESTED bit-shift verification, iso-latitude pixelization vs H3 hex tessellation (08). \nAll notebooks executed end-to-end via Snakemake on each commit; deployed as a Jupyter Book; archived on Zenodo (concept DOI 10.5281/zenodo.19848749)." . "The full claim is evaluated end-to-end across all six fitness criteria for the integration substrate: equal-area correctness, cell-shape preservation across latitudes, hierarchical-refinement efficiency, ellipsoidal correctness, iso-latitude pixelization, and compatibility with the spherical-CNN / scattering-network / sphere-harmonic-transform ML ecosystem." . . . "Anne Fouilloux" . "2026-05-04T21:11:08.723Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "CNN + scattering stacking on Decrop 2025's plankton splits" . . . . . . . "This is an extension, not a reproduction. Differences from Decrop et al. 2025: (1) we add a stacking meta-classifier on top of the CNN's softmax outputs; the original paper reports CNN-alone metrics. (2) We compute scattering features as additional input — the paper does not use scattering. (3) Reported metrics include mean rare-class recall (computed on the 13 classes with train-set size <200), which the paper does not report; aggregate top-1, top-5, micro/macro/weighted F1 are all from the same test.txt evaluation pipeline. Differences from Delouis et al. 2022 (FOSCAT method paper): we use FOSCAT's 2D scattering operator (scat_cov.funct) on flat RGB images, not the spherical HEALPix variant the paper develops; same library, different operator. As with chains #1–#4, FOSCAT was used via the annefou/FOSCAT@v0.1.0-cpu fork; that CPU patch has since been merged upstream and is included in foscat>=2026.4.1 on PyPI." . . "Three-stage pipeline. \nStage 1 (01_scattering_features.py): compute multi-scale scattering features on RGB phytoplankton images via FOSCAT scat_cov.funct(NORIENT=8, KERNELSZ=3) at 64×64 pixels, producing 246-dimensional feature vectors per image. Features are computed on (a) a balanced training subset, (b) Decrop's released val.txt (33,829 images), and (c) Decrop's released test.txt (33,718 images). \n\nStage 2 (02_cnn_predictions.py): obtain CNN softmax probabilities for the same val and test images by running Decrop's pretrained EfficientNetV2-B0 via the authors' planktonclas package with 10-crop test-time augmentation. These predictions can equivalently be sourced from the upstream fiesta-decrop-reproduction repository (Zenodo 10.5281/zenodo.19701133), which is the FAIR-archived form of the same computation. \n\nStage 3 (03_stacking.py): fit a class-weighted scikit-learn LogisticRegression on the 246-dim scattering features (StandardScaler + class_weight='balanced') to obtain scattering-derived softmax probabilities; concatenate with CNN softmax probabilities to form 190-dim meta-features; train a second class-weighted LogisticRegression on Decrop's val split as the stacking meta-classifier; evaluate on Decrop's test split. Compute top-1, top-5, and per-class recall against the integer labels in test.txt. Mean rare-class recall is averaged over the 13 'rare' classes defined as those with fewer than 200 training-set examples in Decrop's train.txt. Comparator results: CNN alone, scattering+LR alone, naive 50/50 probability ensemble, and a hard-switch oracle ceiling (perfect choice between CNN and scattering predictions per image) for context." . "This study tests whether multi-scale scattering features computed independently of the CNN can be combined with the CNN's softmax probabilities via a class-weighted stacking meta-classifier to lift mean rare-class recall on Decrop et al. 2025's exact 95-class phytoplankton classification benchmark, at acceptable cost to aggregate top-1 accuracy. The chain reuses Decrop's released dataset, pretrained CNN weights, and train.txt/val.txt/test.txt splits, asking whether the rare-species class-imbalance limitation acknowledged in the paper can be partially addressed by complementary feature representations." . . . "Anne Fouilloux" . "2026-04-26T19:36:55.873Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAoDcOiD+jen8awiJ6DB2ewDw66PeG64hODmgNFwy7GrwQui4HKnHdvxd++1UhTgiOfycxyxBb7sXPSikLw/1TsSyPsEl0P3/+600szxpTGgLNzW+bZ2DVP3d8ERMV1aWpH0ci3B/5vmK+vXQZ4uCoq57NE0MiFg5c13Gy0gd6n7wZYEhYM4AjWSLL0QS/HY+TFZMYL9bCFeATennGrlB2UEjRlw21UB2Ah16ZZ6hxQlfctFJZE7TGnBJPB3ttTjfcOfamhjZVwQ0yV9mv7x6PGiSmkzpJTVLjn8hagoKT05YUwVQArFb+w7f6sXqvvljMigjd/Rbqgbye/lLUAZLfJSnFM58TubfpEJvXV4zNMDEoT3VQ7dokgoLgMrmjZCKATtQ7gomocoTJ1NhN2esRNtGzWaS2obL/mueUQlMlavssZnqL8WICkdAuDlwDVNbsbwEWKQ50kiPdAdduSigifxA4CM7TgvnxqZVoAResEGP6UhTTem3T4CsbEas1Caj9wa7M1jPjACu5LF5BwcVns3ZQHWLipjRjD+9/ur3G8QtuxbNhmXlDYQ6tXxB1lK+Oz7O519b3bA15ilzFl0SdvMBGTe46xaQ9DsJT18THKnPbUhNMy0dH0VtzpB+EEaXZ25Fp9VHMEUqo1lLS9e89eO3efiqkESKQ7wmB+/DlIRcCAwEAAQ==" . "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" . . . . . . . . . "WGS84 vs sphere HEALPix geometry for FOSCAT SST gap-filling at operational resolution" . . . . . . "This study extends the chain #3 fiesta-scattering-sst workflow (Zenodo 10.5281/zenodo.19686691), which used standard perfect-sphere HEALPix at nside=32, to higher operational resolution (nside=128) with paired sphere-vs-WGS84 comparison. The FOSCAT method, the Copernicus Marine data sources, and the FOSCAT hyperparameters (NORIENT, KERNELSZ, optimiser, iterations) are identical to chain #3; the new variables are (a) the resolution (nside=128 vs nside=32), and (b) the explicit two-arm geometry comparison enabled by the healpix-resample package's ellipsoid parameter. As with chain #3, FOSCAT was used via the annefou/FOSCAT@v0.1.0-cpu fork for CPU execution; that CPU patch has since been merged upstream and is included in foscat>=2026.4.1 on PyPI." . . "The pipeline uses Copernicus Marine L3S PMW SST (cmems_obs-sst_glo_phy_l3s_pmw_P1D-m, with cloud gaps) and L4 SST analysis (cmems_obs-sst_glo_phy-temp_nrt_P1D-m, gap-free reference) for 2026-04-01, both at 0.25° native resolution. The L3S product is regridded onto the L4 grid via nearest-neighbour interpolation. The two products are then independently resampled to HEALPix nside=128 (level=7) under each geometry: standard perfect-sphere (the healpy default) and WGS84 oblate ellipsoid. WGS84 resampling is performed via healpix_resample.GroupByResampler(level=7, reduce='mean', ellipsoid='WGS84'); sphere resampling uses the same package with ellipsoid='sphere' to ensure the implementation difference between the two runs is exactly the geometry argument. Cloud-gap pixels are identified as ocean cells (defined by the L4 mask) without L3S observations. A spherical-harmonic baseline (lmax=60) on the L4 product provides an initial guess for the gaps. FOSCAT scattering synthesis (scat_cov.funct(NORIENT=4, KERNELSZ=3), 300 L-BFGS iterations, CPU) is then run separately under each geometry, using L4's scattering coefficients as the statistical reference and a cloud-only mask for gradient updates. Each geometry's gap-filled output is compared against L4 in cloudy pixels via RMSE in millikelvin." . "This study tests the research question posed by the parent PCC nanopublication: whether using WGS84 ellipsoid geometry instead of the perfect-sphere assumption when binning lat/lon SST observations to HEALPix cells improves the accuracy of FOSCAT scattering-transform gap-filling at operational resolution. We compare the two geometries on a single day of Copernicus Marine SST data (2026-04-01) at HEALPix nside=128, using the L4 gap-free analysis as ground truth in cloudy pixels." . . . "Anne Fouilloux" . "2026-04-26T18:52:46.842Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Scattering-transform SST gap-filling on Copernicus Marine data (Delouis 2022 + IGARSS 2024 tutorial replication)" . . . . . "Differences from Delouis et al. 2022 (paper): (1) input data is Copernicus Marine SST in oceanography, not Planck dust polarisation in astrophysics; (2) the application is gap-filling clouds in remote-sensing imagery, not separating dust signal from instrument noise; (3) the resolution is nside=32 rather than the paper's nside=256, so the workflow runs on a CPU rather than requiring GPU. Differences from Jean-Marc Delouis's IGARSS 2024 tutorial notebook (which is the more direct source): we use the current PMW L3S product (0.25°) matched to the L4 product's grid; the original notebook may have used a different L3S variant. Same FOSCAT software — at the time of the experiment we used the annefou/FOSCAT@v0.1.0-cpu fork to enable CPU execution. The CPU patch (jmdelouis/FOSCAT#40) has since been merged upstream and is included in FOSCAT 2026.4.1 on PyPI (released 2026-04-24), so future runs of this workflow can use the standard PyPI release." . . "We follow the workflow established by Jean-Marc Delouis in the IGARSS 2024 Pangeo tutorial (10.5281/zenodo.19793350). Inputs are Copernicus Marine L3S PMW SST (cmems_obs-sst_glo_phy_l3s_pmw_P1D-m, with cloud gaps) and L4 SST analysis (cmems_obs-sst_glo_phy-temp_nrt_P1D-m, gap-free reference) for 2026-04-01, both at 0.25° native resolution. Pipeline: (a) quality-filter L3S, regrid to L4 grid, build ocean mask from L4 NaN; (b) convert both to HEALPix at nside=32 via averaging; (c) fit a spherical-harmonic baseline (lmax=30) on observed pixels for an initial guess; (d) run FOSCAT scattering synthesis with scat_cov.funct(NORIENT=4, KERNELSZ=3) and L-BFGS over 300 epochs, using the L4 map's scattering coefficients as the statistical target and a cloud-only mask for gradient updates; (e) validate FOSCAT's filled values against L4 in cloudy regions via RMSE, comparing to the harmonic baseline. Inference runs on CPU (Apple M1 Pro, ~139 seconds)." . "This study tests Delouis et al. 2022's generalisation claim on a different domain than the paper's Planck dust polarisation: the framework is applied to operational Copernicus Marine sea surface temperature observations to fill cloud gaps. We evaluate whether scattering-transform synthesis with a gap-free L4 reference product as the statistical target produces gap-filled L3S maps whose values approach the L4 ground truth in cloudy regions, and whether this outperforms a standard spherical-harmonic interpolation baseline." . . . "Anne Fouilloux" . "2026-04-26T16:16:03.211Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Scattering synthesis on a cosmological LSS map (Delouis 2022 replication)" . . . . "Three substantive differences from Delouis et al. 2022:\n \n1. Input data: cosmological large-scale-structure map (LSS_map_nside128.npy from the FOSCAT_DEMO repository) rather than Planck SRoll2 353 GHz dust polarisation Stokes Q and U maps. \n2. Application: synthesis of a new map from random noise such that its scattering statistics match a target — rather than the FoCUS component-separation algorithm the paper used to denoise Planck observations.\n3. Resolution: nside=32 (12,288 pixels), downgraded from the input map's native nside=128. The paper uses nside=256. The lower resolution allows execution on CPU rather than requiring GPU, which keeps the replication runnable on commodity hardware.\n \nThe library, the scattering-transform implementation, and the underlying methodology are identical to the paper's. The deviations are in the data, the target task (synthesis vs denoising), and the resolution." . . "We use the FOSCAT Python package (github.com/jmdelouis/FOSCAT) authored by Delouis — the same software used in the original paper. Scattering coefficients are computed via FOSCAT's scat_cov.funct(NORIENT=4, KERNELSZ=3) on a HEALPix sphere; synthesis is run from random noise using foscat.Synthesis with an L-BFGS optimiser over 300 epochs. The loss is the mean squared difference between the target's scattering coefficients and those of the current synthesised map, normalised by the variance of the target coefficients. We evaluate the synthesised map against the target via three metrics: power spectrum ratio (mean of synthesised / target across multipoles), scattering coefficient improvement percentage, and pixel-level correlation. Inference runs on CPU." . "This study tests the generalisation claim of Delouis et al. 2022 — that the scattering-transform framework developed for Planck dust polarisation can be applied to other processes defined on the sphere. We test that claim on a cosmological large-scale-structure (LSS) map rather than the paper's Planck dust polarisation data. Specifically, we evaluate whether scattering-transform synthesis from random noise can produce a new astrophysical map whose multi-scale statistics (power spectrum and scattering coefficients) match those of a target input." . . . "Anne Fouilloux" . "2026-04-26T14:52:17.764Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Independent reproduction of Decrop 2025 phytoplankton CNN metrics" . . . . . . . . "None of substance. The reproduction uses the same code package (planktonclas), the same released model weights, and the same test.txt partition that the original evaluation reports against. One minor naming note: Table 1 of the paper lists test=33,829 and val=33,718, whereas the released split files have those numbers swapped (test.txt=33,718, val.txt=33,829). This study uses the released test.txt, which is the file the model was actually evaluated on. Hardware differs (CPU instead of GPU); CPU inference is exact arithmetic and does not change metric values. " . . "The package planktonclas==0.2.0 can be installed from PyPI — the same code package authored by Decrop and colleagues and used to train the original model. The pretrained EfficientNetV2-B0 weights and the dataset split files are downloaded from Zenodo 10.5281/zenodo.15269453, and the FlowCam phytoplankton image dataset from Zenodo 10.5281/zenodo.10554845. Inference is run on the 33,718 images listed in the released test.txt using planktonclas.test_utils.predict, which applies 10-crop test-time augmentation. Top-1, top-5, micro F1, macro F1, and weighted F1 were computed via scikit-learn against the integer labels in test.txt. " . "This study reproduces the headline classification metrics reported by Decrop et al. 2025 for the EfficientNetV2-B0 phytoplankton classifier on the held-out test partition: top-1 accuracy, top-5 accuracy, and the macro/micro/weighted F1 scores. We use the authors' publicly released pretrained weights, the exact train/val/test split files distributed alongside the model, and the same 95-class taxonomy." . . . "Anne Fouilloux" . "2026-04-26T13:50:38.174Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Iberian regional replication of Soroye et al. 2020 using open GBIF Bombus occurrence data" . . . . . . . . . " Three intentional deviations from the original lme4/MCMC implementation: \n (i) the species random effect is fit by variational Bayes (statsmodels.BinomialBayesMixedGLM) rather than full MCMC, a fast approximation that typically underestimates posterior credible-interval width by approximately 10–20%; \n (ii) the climateLimits[[1]]/[[2]] precomputation step is reconstructed from the available R helper scripts because the precomputation script itself is not in the public Figshare release of the original code; \n (iii) the input occurrence dataset is replaced with an open GBIF Iberian Bombus query (https://doi.org/10.15468/dl.3frmsq) rather than the authors' curated continental dataset; this is the deliberate experimental change that makes the study a regional replication. " . . " The Python re-implementation of Soroye et al. 2020's pipeline (weatherxbiodiversity, https://doi.org/10.5281/zenodo.19756173) was applied to a regional dataset of Bombus occurrences for Spain and Portugal retrieved from the GBIF Occurrence Download API (https://doi.org/10.15468/dl.3frmsq, 36,560 records, georeferenced, no geospatial issues, accessed 2026-04-25). The cleaning pipeline applies Soroye's species filter, IUCN exclusion list, the same baseline-vs-recent presence/absence construction on a 100 km cylindrical equal-area grid, and per-species Thermal Exposure Index (TEI) and \n Precipitation Exposure Index (PEI) computed from CRU TS 3.24.01 monthly climate. The mixed-effects logistic GLMM is fit with the species random effect and the same fixed-effect interactions as the original (thermal-position baseline + delta + their interaction, precipitation analogues, baseline × baseline interaction, delta × delta interaction). The GBIF download was minted specifically for this study via \n notebooks/01b_gbif_download_doi.py in the same repository. " . " The complete claim — that local extirpation rate in bumble bee species rises with the frequency at which local temperatures exceed species-specific historically observed thermal tolerances — is tested on an independent regional dataset of bumble bee occurrences for the Iberian Peninsula (Spain and Portugal). The regression specification, response variable, grid resolution, and baseline-vs-recent period definitions all match those used in the Phase 2 Python port; only the input occurrence dataset is replaced." . . . "Anne Fouilloux" . "2026-04-25T12:59:22.004Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Python re-implementation of Soroye et al. 2020 pipeline applied to the original Bombus dataset " . . . . . " Two intentional deviations from the original lme4/MCMC implementation:\n (i) the species random effect is fit by variational Bayes (statsmodels.BinomialBayesMixedGLM) rather than full MCMC, a fast approximation that typically underestimates posterior credible-interval width by approximately 10–20%; \n (ii) the climateLimits[[1]]/[[2]] precomputation step is reconstructed from the available R helper scripts because the precomputation script itself is not in the public Figshare release. " . . "All five preprocessing and modelling steps in the original R pipeline (Cleandata_and_makeMCPs.R, CalcSpeciesPr_Rich.R, CalcSamplingEffort_Cont.R, the climate-position computation, and binomialGLMM4Presence.R) were re-implemented in Python 3.12 using xarray, pandas, statsmodels, and scipy. The mixed-effects logistic GLMM was fit with the species random effect and the same fixed-effect interactions as the original (thermal-position baseline + delta + their interaction, precipitation analogues, baseline × baseline interaction, delta × delta interaction). The datasets are the authors' bundled Bombus + CRU TS 3.24.01 monthly climate data (Figshare 10.6084/m9.figshare.9956471) on a 100 km cylindrical equal-area grid for the 1901–1974 baseline and 2000–2014 recent periods." . "The complete claim — that local extirpation rate in bumble bee species rises with the frequency at which local temperatures exceed species-specific historically observed thermal tolerances — is tested on the authors' original dataset across all 66 bumble bee species and the entire North America + Europe study area. Both the regression specification (mixed-effects logistic with thermal-position baseline + delta + their interaction, precipitation analogues, and species random effect) and the response variable (binary local extirpation between the 1901–1974 baseline and 2000–2014 recent\n period) match the original. " . . . "Anne Fouilloux" . "2026-04-25T10:32:21.205Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "EarthCARE Level-2 to HEALPix DGGS conversion pipeline (MSI, ATLID, CPR)" . . . . . . "No prior implementation of this claim exists. This is the first end-to-end pipeline demonstrating EarthCARE Level-2 → HEALPix DGGS conversion on the WGS84 ellipsoid. We therefore classify the study as a Replication (new evidence produced) rather than a Reproduction (re-running prior code), following FORRT terminology. " . . "Python package with 7 Jupyter notebooks: (1) data download via earthcarekit; (2) structural exploration of Level-2 products; (3) 2D Multi-Spectral Imager (MSI) swath → HEALPix cells with per-pixel nearest-cell assignment and type-aware aggregation (mode for classification variables, mean for continuous variables, quadrature-in-variance for uncertainty variables); (4) 1D Atmospheric Lidar (ATLID) and Cloud Profiling Radar (CPR) profiles → HEALPix cell identifier with preserved vertical axis; (5) DGGS-Zarr persistence for downstream analysis. WGS84 geodetic cell placement provided by healpix-geo (10.5281/zenodo.19337734); xarray DGGS integration via xdggs (10.5281/zenodo.14216728); reproducible Pixi environment. Released as v0.1.0 on GitHub and archived on Zenodo (10.5281/zenodo.19709327). " . "Conversion of EarthCARE Level-2A products (MSI_AOT_2A, MSI_COP_2A, ATL_AER_2A, ATL_ALD_2A, ATL_CTH_2A, ATL_EBD_2A, ATL_ICE_2A, CPR_CLD_2A, CPR_FMR_2A) to a HEALPix Discrete Global Grid System representation on the WGS84 ellipsoid, covering both 2D swath imagery and 1D vertical atmospheric profiles." . . . "Anne Fouilloux" . "2026-04-23T14:47:09.474Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Supervised pretraining as alternative to episodic meta-learning for satellite imagery classification" . . . . . "Guo et al. trained Prototypical Networks episodically (40,000 episodes, approximately 3 hours). We replaced episodic training with standard supervised classification (10 epochs, approximately 15 minutes). Same backbone and image resolution. This tests whether the training method matters for cross-domain transfer, or whether the backbone features alone are sufficient." . . "We trained a ResNet-10 backbone (4.9 million parameters, 224×224 pixel images) using standard supervised classification on mini-ImageNet's 64 object categories for 10 epochs with data augmentation. At test time, we froze the backbone and used Prototypical Network-style nearest-prototype classification on EuroSAT (27,000 real Sentinel-2 satellite patches). This approach requires no meta-learning framework — only standard PyTorch classification training. Evaluation over 100 random 5-way tasks with 5, 20, and 50 labeled examples." . "Testing whether standard supervised pretraining — training a model to classify everyday objects using conventional classification — achieves comparable cross-domain few-shot accuracy on satellite imagery to the episodic meta-learning approach used by Guo et al. (2020). " . . . "Anne Fouilloux" . "2026-04-18T16:01:09.256Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Reproducing cross-domain few-shot benchmark on Sentinel-2 with matching architecture" . . . . . . "We reimplemented the ResNet-10 architecture from Guo et al.'s published code rather than running their original codebase (which requires Python 3.5 and PyTorch 0.4). Our implementation uses PyTorch 2.11. Minor differences in random seed and data loading may exist." . . "We built a ResNet-10 backbone (4.9 million parameters) matching Guo et al.'s custom implementation: SimpleBlock residual blocks, custom weight initialization, 7×7 average pooling for 224×224 pixel input images. Training used 40,000 episodic steps on mini-ImageNet (38,400 photographs of everyday objects) with data augmentation (random resized crop, colour jitter, horizontal flip). Evaluation on EuroSAT (27,000 real Sentinel-2 satellite patches, 10 land cover types) over 200 random 5-way tasks with 5, 20, and 50 labeled examples." . "Reproducing the exact Prototypical Networks results from Guo et al. (2020) Table 1 on EuroSAT satellite imagery, matching their architecture, image resolution, and training procedure. " . . . "Anne Fouilloux" . "2026-04-18T15:57:37.040Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Cross-domain few-shot Sentinel-2 classification with simple Prototypical Networks" . . . . . . "Guo et al. used a deeper ResNet-10 backbone (4.9 million parameters) with 224×224 pixel images and 40,000 training episodes. We used a simpler 4-block CNN with 84×84 images and 5,000 episodes to test whether the cross-domain transfer works even with a minimal architecture." . . "We trained a Prototypical Network with a simple 4-block convolutional backbone (110,000 parameters) on 38,400 mini-ImageNet photographs of everyday objects using 5,000 episodic training steps. We then evaluated the trained model on EuroSAT, a dataset of 27,000 real Sentinel-2 satellite image patches covering 10 land cover types across Europe. Classification was evaluated over 600 random 5-way tasks with 5, 20, and 50 labeled examples per class." . "Testing whether Prototypical Networks — an AI method that classifies new categories from a few labeled examples by computing distances to class prototypes — can transfer from everyday photographs to Sentinel-2 satellite imagery for land cover classification." . . . "Anne Fouilloux" . "2026-04-18T15:53:29.280Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "MIICIjANBgkqhkiG9w0BAQEFAAOCAg8AMIICCgKCAgEAoDcOiD+jen8awiJ6DB2ewDw66PeG64hODmgNFwy7GrwQui4HKnHdvxd++1UhTgiOfycxyxBb7sXPSikLw/1TsSyPsEl0P3/+600szxpTGgLNzW+bZ2DVP3d8ERMV1aWpH0ci3B/5vmK+vXQZ4uCoq57NE0MiFg5c13Gy0gd6n7wZYEhYM4AjWSLL0QS/HY+TFZMYL9bCFeATennGrlB2UEjRlw21UB2Ah16ZZ6hxQlfctFJZE7TGnBJPB3ttTjfcOfamhjZVwQ0yV9mv7x6PGiSmkzpJTVLjn8hagoKT05YUwVQArFb+w7f6sXqvvljMigjd/Rbqgbye/lLUAZLfJSnFM58TubfpEJvXV4zNMDEoT3VQ7dokgoLgMrmjZCKATtQ7gomocoTJ1NhN2esRNtGzWaS2obL/mueUQlMlavssZnqL8WICkdAuDlwDVNbsbwEWKQ50kiPdAdduSigifxA4CM7TgvnxqZVoAResEGP6UhTTem3T4CsbEas1Caj9wa7M1jPjACu5LF5BwcVns3ZQHWLipjRjD+9/ur3G8QtuxbNhmXlDYQ6tXxB1lK+Oz7O519b3bA15ilzFl0SdvMBGTe46xaQ9DsJT18THKnPbUhNMy0dH0VtzpB+EEaXZ25Fp9VHMEUqo1lLS9e89eO3efiqkESKQ7wmB+/DlIRcCAwEAAQ==" . "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" . . . . . . . . . "Testing few-shot land cover classification on Sentinel-2 satellite imagery" . . . . . "The original paper tested on mini-ImageNet, a benchmark of 100 everyday object categories (dogs, cars, furniture) photographed with regular cameras. We replace this with Sentinel-2 satellite imagery — a fundamentally different visual domain with overhead perspective, multispectral sensors, and land cover semantics instead of object recognition. We also use a within-domain split (common vs. rare land cover from the same dataset) rather than the original same-domain setup, and we use only the RGB bands (3 out of 13 available Sentinel-2 bands)." . . "We use the EuroSAT dataset containing 27,000 real Sentinel-2 satellite image patches covering 10 land cover types across Europe. We split these into 7 common types (forest, cropland, urban, etc.) used for training, and 3 types (herbaceous vegetation, permanent crop, river) held back as \"rare\" classes. The model learns from the common types, then must classify the rare types using only 1, 5, or 20 labeled examples. We evaluate over 600 random classification tasks and report mean accuracy with 95% confidence intervals." . "We test whether Prototypical Networks — an AI method designed to learn new categories from very few examples — works for classifying land cover types in real satellite imagery. The original method was only tested on everyday photographs (animals, objects, vehicles). We apply it to Sentinel-2 Earth observation data. " . . . "Anne Fouilloux" . "2026-04-17T12:43:55.738Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Replication of SST-fish community claim in the Mediterranean Sea" . . . . "Different region (Mediterranean vs NE Atlantic), different data sources (visual census + demersal trawl vs ICES standardised trawl surveys), different species sets (15 + 230 vs 198), narrower SST range (5.6°C vs 10°C). MEDITS data filtered to shelf depth ≤200m to match the original study's continental shelf scope." . . "Fish abundance data from two Mediterranean sources were matched to sea surface temperature and salinity from Bio-ORACLE. ClimateFish provides visual census counts of 15 climate indicator species at 28 coastal sites (2009-2021). MEDITS provides standardised trawl catches of 230 species across 115 grid cells on the continental shelf (depth ≤200m, 2005-2024). For each dataset, community structure was analysed using Principal Coordinates Analysis on Bray-Curtis dissimilarity, and the relative importance of SST, salinity, and depth was assessed by correlation with the main community gradients." . "The core claim is tested in a different region: does sea surface temperature also drive fish community structure in the Mediterranean Sea? Two independent datasets are used: one from visual census surveys of shallow coastal fish, and one from bottom trawl surveys on the continental shelf." . . . "Anne Fouilloux" . "2026-04-16T16:44:40.890Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Reproduction of SST-fish community claim using ICES DATRAS" . . . "Filters applied: shelf depth ≤200m, fish species only (via WoRMS classification), Baltic surveys (BITS, SE-SOUND) excluded. 258 species and 247 grid cells vs Rutterford's 198 species and 193 grid cells due to different filtering thresholds." . . "Fish abundance data (catch per unit effort) from standardised bottom trawl surveys were matched to sea surface temperature (SST) and salinity from Bio-ORACLE environmental layers. Sampling locations were aggregated into 1x1 degree grid cells. Community structure was analysed using Principal Coordinates Analysis (PCoA) on Bray-Curtis dissimilarity between grid cells. The relative importance of SST, salinity, and depth as drivers of community structure was assessed by correlating each environmental variable with the main axes of community variation." . "The full claim is tested: whether SST is the primary environmental driver of fish community structure on the NE Atlantic continental shelf, using the same ICES DATRAS trawl survey data source as Rutterford et al." . . . "Anne Fouilloux" . "2026-04-16T16:02:35.367Z"^^ . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "ZKP compliance verification applied to biodiversity monitoring" . . . . . "Different domain (water quality monitoring vs carbon emissions), single-party proof (no supply chain), no EdDSA signature verification inside circuit. Tests the core claim that zk-SNARKs can verify environmental compliance without revealing private data." . "Circom 2.1.9 circuit with snarkjs 0.7.6 using Groth16 protocol. Synthetic hydromet data modelled on LifeWatch ERIC Donana monitoring stations. Circuit enforces that each reading is non-negative and strictly below a public threshold. Proof generation and verification measured on commodity hardware. " . "Applied zk-SNARK-based compliance verification to water quality monitoring at Donana National Park, a Natura 2000 site monitored by LifeWatch ERIC. Implemented a Groth16 circuit in Circom that proves 24 hourly conductivity readings are below the EU Water Framework Directive threshold without revealing individual values. " . . . " Anne Fouilloux" . "2026-04-12T21:05:03.548Z"^^ . . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Replication of FAIR2Adapt ODRL access control on synthetic biodiversity dataset" . . . . . . "Uses synthetic biodiversity data instead of real sensitive research data. Consumer DID is a walkthrough identity (did:web:fair2adapt.github.io:fair-data-access:example-consumer) rather than a real researcher's DID. Access request is simulated locally rather than through the GitHub Actions automated workflow." . . "Using the fair-data-access Python framework: \n(1) generate AES-256-GCM dataset key and encrypt a synthetic biodiversity CSV, \n(2) publish an ODRL Offer policy as a signed nanopub with Public Benefit purpose constraint, \n(3) simulate a consumer access request with a did:web identity,\n(4) evaluate the ODRL policy against the declared purpose, (5) wrap the dataset key using ECDH key agreement with the consumer's public key,\n(6) unwrap and decrypt on the consumer side, (7) verify data integrity by comparing decrypted output to the original." . "End-to-end test: encrypt dataset, publish ODRL policy as nanopub, evaluate automated access request, wrap key for consumer DID, decrypt, verify integrity." . . . " Anne Fouilloux" . "2026-04-12T09:23:54.494Z"^^ . . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "TYn4dWDXiXRzPTmCtDTBDwinF4M0QAFjTYXHID2YFay5MuX29JmhETUWXj7GjZY3LbmnRzKmyl0Q0nUD38lgNOVl1i+TlXnuoPPaOFX8hCVyXlHnMaTh0zpW6xMvNzDjAuv0OivVTl1HTPiu3Kfdz9UifX5KBNi95G2hGW80SuGVJqIBe4xzb1rqP+05g9mf2Wuc3J7tvAkgHchPZD1OFskAQpdy7zuhBesaOG7CWoFoCJMZKlEQOeLtn3dzBCGgU6180KrMqYIhh5y51D8OA/245Q5dLk8GE0j48ZzC/rGr8f12ufJsC1FcfBqt22eWH1rqgdWVTFJ87QiGk9If3jgHq+fcSeTPONzSWmTTsbBCXtg/NLd1GJmDzz1vW23jKpgAxiGj4ZgweLA+8bBv8P4sIezl5EX0mbcxcbso+U4KZumaz+Zi7/HF9rDxNXFt02oFbLegqrjQn11Ouar6FrEupCG9hbEjpe42nv57MbaEAdPOLXn1vNrBw/DwN4cuNQQRwpsLk5rO3sczjQDP10f5DAhFlKv68dG3BYzCyS8Lj17qYlLLwIxfS8/QhCPdr+YNsPBJLm44FMW+qcCSRGGcpZTjHYtkstvjJMRB2l/jzI7hFXFSLo4KgEF2SGKH6tSdABLsSyuNHRUnohCYwa5HOjEqefJLLV1Vy73+APM=" . . . . . . . . . "Analysis of Sphere vs WGS84 Ellipsoid Impact on HEALPix Cell Assignment" . . . . " - The original paper uses H3 (sphere-only) and does not consider ellipsoid effects -- this is a novel extension\n - Analysis uses synthetic raster grids, not real Sentinel/Copernicus data \n - Four latitude bands chosen to represent typical European EO coverage" . . "For each region, a 50x50 pixel raster grid is indexed to HEALPix depth 9 cells twice -- once with ellipsoid='sphere' and once with ellipsoid='WGS84' -- using healpix_geo.nested.lonlat_to_healpix(). Cell assignment differences are quantified as percentage of pixels in a different cell. Polygon coverage similarity is measured via Jaccard index using healpix_geo.nested.polygon_coverage(). Environment: Python 3.11, healpix-geo 0.1.1, Docker containerized, seeded RNG (seed=42)." . "Extends the replication of Law & Ardo (2024) by comparing HEALPix cell assignments using sphere-based vs WGS84 ellipsoid-based indexing via healpix-geo. Measures the percentage of raster pixels assigned to different cells depending on the reference surface, across four latitude bands: equatorial (0°), mid-latitude (+48°), high-latitude (+62°), and arctic (+78°)." . . . " Anne Fouilloux" . "2026-04-09T20:56:50.129Z"^^ . . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Replication of DGGS Benchmark Using HEALPix-geo with Sphere and WGS84 Ellipsoid" . . . . " - Uses HEALPix (healpix-geo) instead of H3 -- this is the replication variable\n - Tests both sphere and WGS84 ellipsoid (extension beyond original) \n - Default testing up to 50 vector layers and 500 raster layers \n - Comparison against H3 uses pre-computed reference results from the companion H3 replication study (DOI: 10.5281/zenodo.18904498) " . . "Implements polygon polyfill via healpix_geo.nested.polygon_coverage() and raster indexing via healpix_geo.nested.lonlat_to_healpix() at HEALPix depth 9 (~1 km² cells, matching H3 resolution 9). Vector benchmark: Voronoi polygons, dissolve by value, polyfill to HEALPix cells, join on cell ID, 7-bit classification. Raster benchmark: Gaussian-smoothed layers, pre-indexed pixel-to-cell mapping, aggregate and classify. Ellipsoid analysis: compares sphere vs WGS84 cell assignments across equatorial (0°), Mediterranean (+48°), Scandinavian (+62°), and Arctic (+78°) regions. Environment: Python 3.11, healpix-geo 0.1.1, Docker containerized, seeded RNG (seed=42)." . "Replicates both vector and raster benchmarks from Law & Ardo (2024) using HEALPix indexing via the healpix-geo library (Rust-based) instead of H3. Extends the analysis with a geodetic comparison: all benchmarks are run twice -- once with sphere-based indexing, once with WGS84 ellipsoid indexing -- across four latitude bands (equatorial, mid-latitude, high-latitude, arctic). " . . . " Anne Fouilloux" . "2026-04-09T20:38:45.956Z"^^ . . . . . . "NP created using Declaring a replication study design according to FORRT" . . "RSA" . "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" . "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" . . . . . . . . . "Replication study: Determinants of Bactrocera oleae abundance in olive groves" . . "Analysis types: Regression analysis. Machine-readable descriptions generated using dtreg and published in the TIB Knowledge Loom." . "Reproduction of analyses from Knowledge Loom record: Determinants of Bactrocera oleae abundance in olive groves. In a study of 25 olive groves in the Beira Interior region of Portugal, landscape simplification (i.e. the share of the area surrounding the olive groves covered by other olive groves) and landscape diversity (i.e. landscape Shannon diversity index) displayed the most notable effects on Bactrocera oleae abundance in olive groves. The bivariate generalized linear models show that B. oleae abundance increases with decreasing landscape complexity and diversity." . . . . "Anne Fouilloux" . "2026-04-02T18:50:03.000Z"^^ . . . . . "Replication study: Determinants of Bactrocera oleae abundance in olive groves" . . . . . "RSA" . "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAosxbitQQzLXi1949Zd9JmSkGfYHHlj/CZZ7iiYs1TrZ5/Jk/wGA7kHEv7f9NtsinOdBo9EtHj/jgHE5W2Vv404JbOAY280PvH5Jns5ObWdVZmtHeCw0ZIdPEqNrurrEweKhzcTJW/YRpYWPwVPo47XyIW6IAcmx6gfdtmdPddMpplqExrP6G99ksXfXlZI0InQtZJRSGK5lYLLNzaofFtupPI5OAAGjooDyHijp0Ap2HIXH6WpO4S44cFPKU34pH2xhIY4/XT5DG1X5UoiVHs2Yoo30BHFudj/kAFwdzcy6Yh4tMDaB3ox6p7pi267d7n0y7kypC0Nt+hfgHQ1FpgwIDAQAB" . "AgfgCAvTeFCgntou/Awp2vkIdaR9rSf1/5MGxa5WwlfMyV/jBi6EUOykH+H9IdL0P/m8MXJ/UhDyf9uz7sBtF5OKCITahEJ1wxOjfNXd6dRC26srahEXWBvrWNh1UG6S/AxHvWvGeT6gn2SVEatGTRel4mBvZ1YxObH9uERXCLSVirjsZv77fDYVM1vB+5M1ebjN6v4o/+vLbuES6PdbAr3HRQ87vm1A7UVKEnGIh6P+VwQn5UPmrm1UfxPurpHqwYEv3A/8Zh6f2BkKIz1Lk2Q6cbjrMmLO7UI1npRJ2pAVPIiJQidRZv2ZLx7YXnamDA6JrD7eSFK0C7o60R5Z4w==" . . . . . . . . . "Replication study: Determinants of Bactrocera oleae abundance in olive groves" . . "Analysis types: Regression analysis. Machine-readable descriptions generated using dtreg and published in the TIB Knowledge Loom." . "Reproduction of analyses from Knowledge Loom record: Determinants of Bactrocera oleae abundance in olive groves. In a study of 25 olive groves in the Beira Interior region of Portugal, landscape simplification (i.e. the share of the area surrounding the olive groves covered by other olive groves) and landscape diversity (i.e. landscape Shannon diversity index) displayed the most notable effects on Bactrocera oleae abundance in olive groves. The bivariate generalized linear models show that B. oleae abundance increases with decreasing landscape complexity and diversity." . . . . "Anne Fouilloux" . "2026-04-02T18:30:27.000Z"^^ . . . . . "Replication study: Determinants of Bactrocera oleae abundance in olive groves" . . . . . "RSA" . "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAosxbitQQzLXi1949Zd9JmSkGfYHHlj/CZZ7iiYs1TrZ5/Jk/wGA7kHEv7f9NtsinOdBo9EtHj/jgHE5W2Vv404JbOAY280PvH5Jns5ObWdVZmtHeCw0ZIdPEqNrurrEweKhzcTJW/YRpYWPwVPo47XyIW6IAcmx6gfdtmdPddMpplqExrP6G99ksXfXlZI0InQtZJRSGK5lYLLNzaofFtupPI5OAAGjooDyHijp0Ap2HIXH6WpO4S44cFPKU34pH2xhIY4/XT5DG1X5UoiVHs2Yoo30BHFudj/kAFwdzcy6Yh4tMDaB3ox6p7pi267d7n0y7kypC0Nt+hfgHQ1FpgwIDAQAB" . "Hw5fZ+meeRP2lmgLJeXHF3YJvPQo92P3NmZ8L9lhyD0eAk+cghSKKR35F4ZUa7MKRsPKI8Q6BHfEKtw9tG9v/kKJMHSR34gilwNGfe1xh9MpyKjbKFoSIMMQAXXz+HE730QHSpby2rdvSQeTOtCXvTS5ATAIb85NkxbrHdjy0xsF5N59Yl962D8lDCIgysw5oKMYhQbxWjbORqTQxWhEL2m9yN8D7GXAeGwtabP4IdRm4GkKdoqnUznIe4QxSvywoZFpZnoo++bhiMXjgK6O8tWu+YBvm9wqQ9TnT+ksoL8faLkK0vUq0Z3OBYHkOW5Eyz4kXW9nKDkFFivXfGTa2w==" . . . . . . . . "Reproduction and Replication of DGGS Benchmark" . . . . . . "1. SCALE: The original paper tested up to 500 vector layers; our default configuration tests [5, 10, 20, 50, 100] layers but supports scaling to 500.\r\n\r\n2. RASTER GENERATION: The paper used NLMpy mid-point displacement algorithm. Our implementation uses NLMpy when available, with Gaussian filter fallback.\r\n\r\n3. RANDOM MISALIGNMENT: The paper mentions \"jittering the origin point by up to one pixel\" for raster alignment - this feature is not implemented in our reproduction.\r\n\r\n4. ADDITIONAL COMPARISON: We added xdggs as an alternative DGGS implementation not present in the original study, extending the work from pure reproduction to include replication with different tools.\r\n\r\n5. PRE-INDEXED SCENARIO: The paper's raster benchmark used pre-indexed data in Apache Parquet queried with Polars. Our benchmark includes both on-the-fly indexing and pre-indexed scenarios to enable direct comparison." . . "REPRODUCTION METHODOLOGY:\r\n- Vector benchmark: Implemented H3 polyfilling algorithm via h3-py library to convert Voronoi polygons to H3 cells at resolution 14, matching the paper's approach\r\n- Raster benchmark: Used H3 Python loop (h3.latlng_to_cell) to index raster pixels to H3 cells, replicating the paper's indexing method\r\n- Classification: Implemented all 7 number-theoretic classification functions (prime, perfect, triangular, square, pentagonal, hexagonal, Fibonacci) as described in the paper\r\n- Data generation: Created synthetic Voronoi polygons and NLM raster landscapes following the paper's specifications\r\n\r\nREPLICATION METHODOLOGY:\r\n- Raster benchmark: Replaced H3 Python loop with xdggs library (xdggs.H3Info.geographic2cell_ids) for vectorized coordinate-to-cell conversion\r\n- This tests whether alternative DGGS implementations affect the benchmark conclusions\r\n\r\nCOMPUTATIONAL ENVIRONMENT:\r\n- Python 3.11 with h3 4.x, xdggs, NumPy, GeoPandas, Polars\r\n- Docker container for reproducibility\r\n- Benchmarks run on standardized hardware with multiple iterations" . "This study aims to reproduce and replicate the computational benchmark experiments from Law & Ardo (2024) \"Using a discrete global grid system for a scalable, interoperable, and reproducible system of landuse mapping\" (DOI: 10.1080/20964471.2024.2429847).\r\n\r\nSpecifically:\r\n1. VECTOR BENCHMARK (Figure 6): Reproduces the comparison between traditional vector overlay operations and DGGS-based methods using H3 polyfilling, testing scalability across 5-500 input layers.\r\n\r\n2. RASTER BENCHMARK (Figure 7): \r\n - REPRODUCTION: Recreates the paper's comparison using H3 Python bindings for coordinate-to-cell conversion\r\n - REPLICATION: Implements an alternative approach using xdggs for vectorized H3 indexing\r\n\r\nThe study aims to validate the paper's claims that (1) DGGS provides orders of magnitude performance improvement for vector operations, and (2) DGGS and raster methods show roughly equivalent performance for raster operations when using pre-indexed data." . . . "h3 - Hexagonal Hierarchical Geospatial Indexing System" . "discrete global grid system - DGGS as described by ISO 19170-1:2021" . "geospatial analysis - type of spatial analysis" . "reproducibility - agreement with previous measurements using the same methodology in the same context" . "Earth science - fields of science dealing with planet Earth and its nearby planets in space" . "benchmark - test to measure the performance of a computer system or component" . "Anne Fouilloux" . "2026-02-28T14:07:08.443Z"^^ . . . . . . . . . "RSA" . "MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQDWv2pJnmDsBOq8OlT1aSvYXSuWT34WOp4FYqEzdnn2F0kqzcFevBqWGZDxJWC0lqCrDEuNfp2QFyPe/+nES9dlHGYIhqPi68fwK6ZiNUotRFxXou+rjFznVvUxtCL8Ede79EBHwWN61QtwSIcU12bLoZsNPFlqQASQ93BJuKlihwIDAQAB" . "eFedl8xRjFaxdcW08KxGasxANPXTQbDF/JjEmlmaL71uPKEIppZh/00QMczy1uQ/dKyy70YsWVIFcJSZ6TzRWJkz3Aez/X058qYj4jsHTAzScB1QlPRj+tWEop7VfaNGAXrTC6CWSmX49Vq6N9YI0LIUk0P57vHv+WLsSAbZ55M=" . . . "Effect of DGGS Indexing on Associating Vector and Raster Geospatial Data" . . . . . "2026-01-22"^^ . . "DGGS Benchmark Replication Study" . . "Anne Fouilloux" . "2026-02-18T21:00:22.992Z"^^ . . . . . . . . . "RSA" . "MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQDWv2pJnmDsBOq8OlT1aSvYXSuWT34WOp4FYqEzdnn2F0kqzcFevBqWGZDxJWC0lqCrDEuNfp2QFyPe/+nES9dlHGYIhqPi68fwK6ZiNUotRFxXou+rjFznVvUxtCL8Ede79EBHwWN61QtwSIcU12bLoZsNPFlqQASQ93BJuKlihwIDAQAB" . "pbCkmJCVspPSzzOuCOouTfuVDzX8HXiEjCzMn+S/Hn6jRfj1qCKOyrecI7I/xOxtxugJfyEmY+W38SFhV1gnirR7ZMslF/C2dOUtSxNubDbSaJt8lqNUVBmfvRR4w5KYS8+WpJadKgI4qZ/kcRacT3L+rewNvo5g7x0PdALIpmI=" . . . . . . . "2026-01-05"^^ . . "Quantum Computing for Biodiversity: Technology Readiness Assessment" . . "Anne Fouilloux" . "2026-02-17T10:09:57.504Z"^^ . . . . . . . . . "RSA" . "MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQDWv2pJnmDsBOq8OlT1aSvYXSuWT34WOp4FYqEzdnn2F0kqzcFevBqWGZDxJWC0lqCrDEuNfp2QFyPe/+nES9dlHGYIhqPi68fwK6ZiNUotRFxXou+rjFznVvUxtCL8Ede79EBHwWN61QtwSIcU12bLoZsNPFlqQASQ93BJuKlihwIDAQAB" . "z2UQkG92dmKekUYlZqcqEa6NM1WTHyD7p/MnJc28KmHsuMOMYyXWUxhNTDCp7M2JlgvmOSiDN3NxlP6Cw7JW7gwmLx/qJaSmnaTOlFiZDCeHapFvySRWzYhDLqdQEUzLCUnu1HI81PXd6RkE60WT+ZlfDXIU+d9DXnvFQx2iUnA=" . . .