https://w3id.org/np/RAmR-xqMgO.../conceptconcepthttp://purl.org/dc/terms/descriptiondescription"The methodological choice of HEALPix geometry — perfect-sphere versus WGS84-ellipsoid — when resampling lat/lon-gridded SST observations to a HEALPix cell tessellation, prior to applying scattering-transform synthesis. The two geometries differ in how lat/lon points are binned to cells."
.
https://w3id.org/np/RAmR-xqMgO.../contextcontexthttp://purl.org/dc/terms/descriptiondescription"FOSCAT scattering-transform gap-filling pipelines for satellite-derived SST imagery, run at HEALPix resolutions ranging from coarse (nside=32, ~6° angular) to operational (nside=128, ~0.5° angular). The gap-filling task is to reconstruct cloud-obscured SST values using the L4 gap-free product as a statistical reference, evaluating the result via RMSE against L4 in cloudy pixels."
.
https://w3id.org/np/RAmR-xqMgO.../populationpopulationhttp://purl.org/dc/terms/descriptiondescription"Daily Copernicus Marine sea surface temperature observations over the global ocean — specifically the L3S Passive Microwave product (cmems_obs-sst_glo_phy_l3s_pmw_P1D-m, with cloud gaps) and the L4 gap-free analysis (cmems_obs-sst_glo_phy-temp_nrt_P1D-m) used as ground truth, both at 0.25° native resolution. The 'population' here is the set of ocean pixels in a single day's observation."
.