Integrity Assessment Report: Optimization and accuracy validation of near-space 1.27 μm O2(a1Δg) dayglow forward model based on limb observation mode

Academic fraud report · Geng Detector

Summary

This report assesses the 2026 Geo-spatial Information Science article by Zeng et al. (DOI: 10.1080/10095020.2026.2633870), which presents a forward model for 1.27 μm O2(a1Δg) dayglow under limb observation. The overall verdict is SUSPICIOUS (yellow). The strongest finding is a substantive technical error: the authors claim to use SABER ozone profiles retrieved from a '960 nm absorption band,' but SABER on TIMED retrieves ozone via the 9.6 μm thermal infrared emission band (~1042 cm⁻¹). This misidentification directly affects the credibility of the local accuracy validation (Section 4.2.2). Two additional concerns remain unverified: Table 1 sensitivity values appear unusually clean (multiple 0.00%/0.01% responses inconsistent with realistic non-LTE limb radiative transfer), and the choice of MODTRAN5 rather than MODTRAN6 for 2025–2026 high-precision benchmarking is dated. Image manipulation analysis could not be performed due to lack of source figures. Confidence is high on Finding 1; Findings 2–4 carry lower confidence and require raw data for resolution.

Verdict

🟡 SUSPICIOUS — One confirmed substantive technical error compromising the validation pipeline; other concerns remain unverified.

Key findings

Confirmed methodological error (Finding 1): Authors state SABER retrieved ozone from a "960 nm absorption band"; SABER actually uses the 9.6 μm thermal infrared emission band. This misidentification undermines the local accuracy validation in Section 4.2.2.
Unverified — suspicious data regularity (Finding 2): Table 1 (Section 5.2) shows unusually clean sensitivity values (e.g., +2% solar irradiance perturbation producing 0.00% or 0.01% responses at multiple wavelengths/altitudes). Residual values in Section 5.3 (0.8%, 0.99%) appear overly tidy for non-LTE radiative transfer.
Unverified — dated reference code (Finding 3): Use of MODTRAN5 rather than MODTRAN6 for 2026-era "high-precision" validation is anachronistic; however, using older software is not by itself misconduct.
Not assessable (Finding 4): Image-level forensics on Figures 1–19 not possible without source files.

Evidence highlights

Direct quote, Section 4.2.2 (Page 10): "we use the O3 concentration profile data retrieved from the 960 nm absorption band by the SABER payload."
Factual correction: SABER on TIMED observes ozone via the 9.6 μm CO2–O3 thermal emission band, not 960 nm. The 960 nm range falls outside SABER's spectral channels and is unrelated to its O3 retrieval algorithm.
Table 1 anomaly: Solar irradiance (+2%) sensitivity entries reported as 0.00% or 0.01% at 1264.6 nm, 1269.2 nm, and 1274.5 nm across 35 km and 45 km.
Section 5.3 residuals: Complete-model relative residuals reported as ~0.8% and ~0.99%.
DOI: 10.1080/10095020.2026.2633870
Journal/Year: Geo-spatial Information Science, 2026 (published online 09 Apr 2026).

Notes

The SABER band error may originate from a unit confusion (possibly 9.6 μm ↔ 960 nm) or a citation/AI-generation artifact. Authors should be asked to provide the actual data source and processing code for the SABER O3 profiles.
Findings 2 and 3 should be treated as hypotheses pending access to the authors' simulation code, raw MODTRAN outputs, and source figure files.
Recommended actions: (1) request raw data and code from authors, (2) post structured queries on PubPeer regarding the 960 nm attribution and Table 1 patterns, (3) monitor for corrigenda. Formal institutional referral is not currently warranted based on available evidence.
Confidence: HIGH on Finding 1; LOW–MEDIUM on Findings 2–3; N/A on Finding 4.
This report is AI-assisted and intended for academic discussion only.