@prefix this: . @prefix sub: . @prefix np: . @prefix rdf: . @prefix prov: . @prefix npx: . @prefix dc: . @prefix xsd: . sub:Head { this: a np:Nanopublication; np:hasAssertion sub:assertion; np:hasProvenance sub:provenance; np:hasPublicationInfo sub:pubinfo . } sub:assertion { sub:wgs84-vs-sphere-sst-outcome a ; "2026-04-22"^^xsd:date; "WGS84 ellipsoid HEALPix geometry reduces FOSCAT SST gap-filling RMSE by 167.7 mK (12.6%) over sphere at operational resolution (nside=128)"; "The PCC research question is answered affirmatively at operational resolution: using WGS84 ellipsoid geometry instead of perfect-sphere HEALPix when resampling lat/lon SST observations to HEALPix cells reduces FOSCAT scattering-transform gap-filling RMSE by 167.7 mK (12.6%) at nside=128, on a single day of Copernicus Marine data. The geometry choice has a quantifiable effect on accuracy at the resolutions actually used in operational SST analysis pipelines, where 0.25° native data is binned into HEALPix cells comparable in size to the underlying observations. The benefit is resolution-dependent: at coarser resolution (nside=32), prior testing showed the WGS84 vs sphere difference collapses to ~1.5 mK (0.14%), well within run-to-run variability — i.e. at coarse resolution the geometry choice doesn't matter. The practical recommendation is to migrate to WGS84 geometry for operational SST gap-filling pipelines."; ; """Single-day comparison on Copernicus Marine SST for 2026-04-01 at HEALPix nside=128 (12 × 128² = 196,608 total pixels). Inputs: L3S PMW (cmems_obs-sst_glo_phy_l3s_pmw_P1D-m) with cloud gaps, L4 analysis (cmems_obs-sst_glo_phy-temp_nrt_P1D-m) as gap-free reference. Pipeline parameters: NORIENT=4, KERNELSZ=3, lmax=60 (harmonic baseline), 300 L-BFGS iterations, CPU. RMSE measured against L4 in cloudy ocean pixels only. Sphere geometry: 134,587 ocean / 69,547 observed / 65,040 gap pixels — RMSE = 1333.0 mK, runtime 2081.8 s.\\n\\nWGS84 geometry: 134,472 ocean / 69,626 observed / 64,846 gap pixels — RMSE = 1165.3 mK, runtime 2083.2 s. Difference: WGS84 lower by 167.7 mK (12.6% relative reduction), with similar gap-pixel counts and identical runtime so the comparison is balanced. Full results JSON archived at Zenodo 10.5281/zenodo.19693573."""; "Three caveats limit the strength of the conclusion. (1) Single-date evaluation. Only one day (2026-04-01) of L3S/L4 SST was tested at nside=128; cloud distribution, ocean regimes, and the gap pattern vary across days and seasons, so a multi-day study would strengthen the result. (2) L4 as ground truth. The 1333.0 / 1165.3 mK numbers are computed against the L4 operational analysis, not against in-situ measurements; L4 itself is an interpolation that may have its own systematic biases. (3) Single comparison resolution. This study's canonical run is at nside=128 only. The complementary nside=32 result (sphere 1061.6 mK, WGS84 1060.1 mK, 0.14% difference) is documented in the project's planning notes but is not part of this canonical archive — a future study should systematically scan resolution to map exactly how the WGS84 advantage scales."; ; ; . } sub:provenance { sub:assertion prov:wasAttributedTo . } sub:pubinfo { "Anne Fouilloux" . this: dc:created "2026-04-26T19:00:25.366Z"^^xsd:dateTime; dc:creator ; dc:license ; npx:introduces sub:wgs84-vs-sphere-sst-outcome; npx:wasCreatedAt ; "NP created using Declaring a replication study outcome according to FORRT"; . sub:sig npx:hasAlgorithm "RSA"; npx:hasPublicKey "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"; npx:hasSignature "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"; npx:hasSignatureTarget this:; npx:signedBy . }