Review report on 'Effects of layered soil replacement on light-response parameters and model applicability of elm (Ulmus pumila) in an arid region of the Qinghai-Tibet Plateau' (Arid Zone Research, DOI: 10.13866/j.azr.2026.07.12)

Academic fraud report · Geng Detector

Summary

Verdict: 🟡 Questionable (存疑). This ecological-physiology paper by Zhan Dongshan et al. fits four light-response models to gas-exchange data from elm trees under two soil treatments (ALS vs FUS) across 2024–2025 growing seasons. Five concerns remain open: (1) Fig. 3 FUS curves lack a typical saturation plateau at high PAR, although the text itself states 'the FUS light-response curve rises slowly' with maximum Pn of 11.27 μmol·m⁻²·s⁻¹, providing a textual but not independent visual confirmation; (2) Fig. 4 diurnal stomatal-conductance waveforms show strong inter-year similarity in the same months, which cannot be distinguished between genuine eco-physiological rhythm, uniform smoothing, or potential data reuse without raw .csv files; (3) Fig. 6 ALS curves exhibit jagged kinks at 1400–1800 μmol·m⁻²·s⁻¹ despite triplicate averaging per PAR step; (4) Fig. 5 shows NRH-model AQY Cv >80% for FUS, which the authors themselves attribute to 'individual-month fitting outliers' consistent with Table 2 AQY spread (0.0075–0.0290); (5) experimental timing (planted May 2024, measured July 2024) raises design concerns but not fraud. Statistical (Benford, last-digit, inter-column) and image-forensic (copy-move, PRNU) flags could not be cross-validated against original numerical data and several tests have unmet preconditions for structurally constrained model parameters. Confidence: medium-low; original GFS-3000 raw data are required to resolve remaining questions.

Verdict

🟡 Questionable (存疑). The five flagged issues are individually explainable by stated methods or design, but in aggregate they require the authors to release raw data for independent verification. No finding rises to the level of confirmed misconduct.

Key findings

Fig. 3 (FUS light-response curves): Lack a saturation plateau at high PAR; the manuscript text itself supports this observation (FUS Pn max = 11.27 μmol·m⁻²·s⁻¹ vs ALS 17.999 and 19.75 μmol·m⁻²·s⁻¹ in July; ALS mean Pn 10.49 vs FUS 5.89, +78.1%). Visual 'unsmooth kinks' on ALS curves reported by the automated checker were only a weak match.
Fig. 4 (diurnal Gs): Inter-year same-month waveforms (especially 14:00–16:00 troughs) are highly similar. The manuscript states ALS Gs exceeded FUS by 9.31–31.94% (2024 Aug–Sep) and 11.80–22.84% (2025 Aug–Sep). Three hypotheses remain open (genuine eco-physiological rhythm, uniform smoothing, or data reuse).
Fig. 6 (model fits): Black measured lines show jagged kinks at 1400–1800 μmol·m⁻²·s⁻¹ despite the stated protocol of triplicate averaging per PAR step. The authors' method (Section 1.3) does not mention additional smoothing of plotted scatter points.
Fig. 5 / Table 2 (NRH AQY for FUS): Cv reportedly exceeds 80%, but the discussion acknowledges this and Table 2 shows NRH-FUS AQY spans 0.0075–0.0290 across 2024-08 (0.0290), 2024-09 (0.0095), 2025-08 (0.0075), consistent with the authors' explanation of 'individual-month fitting outliers'.
Experimental timing: Trees were planted May 2024 and photosynthesis measurements began July 2024 (~2 months later), which is tight for an arboreal species but not indicative of fraud.

Evidence highlights

DOI: 10.13866/j.azr.2026.07.12
Journal: 干旱区研究 / Arid Zone Research, Vol. 43, No. 7 (2026)
Quantitative anchors from the original text:
FUS max Pn = 11.27 μmol·m⁻²·s⁻¹; ALS July 2024 = 17.999, July 2025 = 19.75 μmol·m⁻²·s⁻¹.
Mean Pn: ALS 10.49 vs FUS 5.89 μmol·m⁻²·s⁻¹ (+78.1%).
Gs difference: ALS vs FUS = +9.31–31.94% (2024 Aug–Sep), +11.80–22.84% (2025 Aug–Sep).
NRH-FUS AQY: 0.0290 (2024-08), 0.0095 (2024-09), 0.0075 (2025-08).
Section 1.3: 'Each PAR gradient measured 3 times consecutively, mean used as observation.'
Image-forensic flags (img-000/001/004.jpg) — copy-move blocks at offsets (48,0) (42 pairs) and (0,32) (81 pairs); PRNU NCC = 0.006, PCE = 35.2 — could not be reconciled to specific paper elements and the report notes Fig. 6's 12 repeated sub-panels as a plausible benign source.
Statistical flags (Benford, last-digit χ² = 75.2, 6.8σ adjacent deviation, two constant inter-column differences): the report states the underlying numerical values were not located in the manuscript and the tests have unmet preconditions for structurally constrained parameters (Pnmax > 0, Rd ≥ 0, α > 0, θ ∈ (0,1]).

Notes

All five flagged items remain dependent on raw GFS-3000 .csv outputs (14 PAR steps × 6 months × 2 treatments × 3 replicates) and the four-model convergence logs to resolve.
The '88-page' copy-move match count must be interpreted in the context of repeated sub-panel layouts and identical journal typesetting; it does not by itself imply duplication of scientific content.
Benford-style and last-digit analyses are generally not applicable to non-negative, bound-constrained fitted parameters; their positive signals are informational at most.
The two-month interval between planting (May 2024) and first measurement (July 2024) is a legitimate design question, not a misconduct indicator.
Final determination of academic misconduct requires an institutional investigation with access to raw data and laboratory records; this report is AI-assisted and may contain false positives or miss genuine issues.