Abstract
<jats:p>Open Earth-observation missions have expanded the spectral, spatial, and temporal information available for environmental screening, but candidate remote-sensing methods remain dispersed across application domains and vary widely in implementation and validation maturity. We present the Global Spectral Index Atlas (GSIA), an open registry organized into 24 capability families containing 91 governed environmental remote-sensing method-specification records across twelve domains. The record count is an inventory measure, not a claim of 91 unique band combinations, scientifically unprecedented equations, or validated detectors. Each record documents an intended screening use, proposed and implemented formulas where applicable, required sensors and operators, physical rationale, contribution class, method role, maturity, calibration and validation status, and an intended-use/inference-limit field whose specificity varies by record. Confounders are addressed mainly at the domain and study-design levels rather than exhaustively itemized for every record. The version 3 registry snapshot contains 37 live M3 screening proxies, 16 M2 executable but non-live formulas, and 38 M1 specified concepts or retired formulas. Method roles identify 15 primary representatives, 10 variants, 12 components, one reference product, 51 research models, and two retired records; these roles organize the registry and do not establish novelty or performance. All 37 renderable evalscripts passed static band and output checks. A public-service display audit returned nonblank overlays for all 37; the one record whose script was subsequently corrected was re-audited and retained its verdict. All 42 evidence packs associated with the 37 live records and five separate Sentinel-1 or Sentinel-5P demonstrations met the registry's source-coverage rule. These are software, display, and provenance results, not measurements of environmental detection performance. No registry record has completed the independent, held-out accuracy assessment defined in this paper. Targeted prior-art review identified related methods in several showcased domains, so catalog-wide priority and "first formula" claims are not made. GSIA is therefore presented as a transparent hypothesis and method-specification registry rather than a collection of validated detectors. Version 3 additionally reports a structured review that found seven records overstating their required sensors, four whose implemented formula did not match their executing code, one mechanistically incorrect physical rationale plus three additional rationale or implementation-description errors, and a scene-dependent radiometric offset defect in one rendering path; all are corrected and documented here as a demonstration of the registry's intended correction pathway. A separate structural audit enumerated six bounded formula families over Sentinel-2 MSI: the full 13-band instrument produced 15,054 role-specific expressions from 1,079 unordered two- to four-band sets, while a ten-band reflected-surface core produced 4,770 expressions from 375 sets. Symmetry and monotonic-equivalence rules reduced the surface-core result to 2,340 information classes. Exact comparison with a pinned established-index catalog found 71 named matches representing 48 distinct equations. The structural audit measures formula-space organization and prior documented use, not environmental usefulness, novelty, or accuracy. We define a validation protocol using explicit labels, hard negative controls, locked formulas, geographic and temporal holdouts, established baselines, uncertainty estimates, and external replication. The registry is intended to make environmental remote-sensing hypotheses inspectable, testable, and correctable, not to replace field measurement or regulatory assessment.</jats:p>