Integrity Review: Improving atmospheric CO2 retrieval based on the collaborative use of Greenhouse gases Monitoring Instrument and Directional Polarimetric Camera sensors on Chinese hyperspectral satellite GF5-02

Academic fraud report · Geng Detector

Summary

This review examines a 2024 Geo-spatial Information Science paper (DOI: 10.1080/10095020.2023.2238773) by Ye et al. on joint CO2 retrieval using the GMI and DPC sensors aboard China's GF5-02 satellite. The overall verdict is "questionable but not indicative of fraud." Internal textual inconsistencies were identified: the DPC spatial resolution is reported as 3.5 km in Section 2.2 and Table 2, but as 3.3 km in a later passage of Section 2.2 and in Section 3.2. Section 2.1 also omits physical units (μm) for GMI spectral band centers. Statistical back-calculations (filtering fractions of 75%, 88%, and 61%) reproduce the reported counts within rounding tolerance, showing realistic non-integer ratios. The satellite-to-submission timeline (launch 2021-09-07, GMI operations 2021-10, data 2021-11 to 2022-01/03, TCCON download 2022-10-01, submission 2023-02-01) is internally consistent. No data fabrication, image manipulation, or temporal anomalies were detected.

Verdict

Questionable (🟡). The paper contains clear editorial and methodological inconsistencies, but no evidence of data fabrication or scientific misconduct. An Erratum is recommended.

Key findings

DPC spatial resolution is internally inconsistent: 3.5 km in Section 2.2 prose and Table 2, but 3.3 km in another passage of Section 2.2 and in Section 3.2.
Section 2.1 (GMI) lists spectral band centers (0.76, 1.58, 1.65, 2.0) without physical units (μm should appear).
Statistical filtering fractions back-calculate cleanly: 9600/38000 ≈ 25.26%; 8400/9600 = 87.5% (reported 88%); 8070/13200 ≈ 61.13% (reported 61%).
Table 4 aggregate statistics (mean bias −0.9, SD 3.3) are internally consistent across subgroups.
No last-digit regularity or rounded fake patterns were observed in reported numbers.
Satellite timeline is closed and plausible: launch (2021-09-07) → GMI operations (2021-10) → data window (2021-11 to 2022-01/03) → TCCON access (2022-10-01) → submission (2023-02-01).

Evidence highlights

Spatial resolution conflict: Section 2.2 states "The observed spatial resolution is 3.5 km… shown in Table 2"; Table 2 lists 3.5; later in Section 2.2 the value becomes "3.3 km"; Section 3.2 opens with "The spatial resolution of the DPC pixel is 3.3 km."
Unit omission: Section 2.1 lists NIR center 0.76 and SWIR centers 1.58, 1.65, 2.0 without μm, despite Table 1 reporting nm ranges (e.g., 759–769 nm).
Counts verified: 38000 L1 → 75% filtered → ~9600 → 88% converge → ~8400; parallel branch: 13200 → 61% → 8070.
DOI: 10.1080/10095020.2023.2238773 (published online 02 Aug 2023; journal year 2024).

Notes

Recommended action: Erratum to reconcile the DPC spatial resolution and add missing units; no formal editorial complaint warranted on current evidence.
Confidence in fraud-negative findings is high for statistics and timeline; confidence in the textual-error classification is moderate, as the 3.5 vs 3.3 km discrepancy could also reflect a sensor specification update not propagated across the manuscript.
Limitations: this review is text-based; underlying L1/L2 data and code were not independently re-run.