Crunching the numbers. Responsibly.
Reliability concern: This paper was retracted on 2026-06-01. See notice: 10.1016/j.rse.2026.115381.
the HBiM2 model reduces the overall RMSE by approximately 25% and 20%, respectively
direct model comparison supports the direction and magnitude, but absolute gains are modest and fitting-based
we propose an unified soil RTM framework capable of modeling soil reflectance across both dry and wet conditions
model equations instantiate the unified framework, with validation bounded to controlled datasets
The HB model demonstrates high accuracy across diverse conditions (R2 > 0.98, RMSE <0.02)
dry-soil tables and figures consistently report high fitting accuracy across textures, particles, and geometries
the HBiM2 model exhibits consistently robust performance under varying moisture and angular conditions (R2 > 0.95, RMSE <0.02)
wet-soil and BRDF results are consistent, though validation remains controlled rather than operational
highlighting its strong potential for soil parameter retrieval
retrieval relevance is plausible, but no independent retrieval experiment or deployment workflow is shown
Derived from the full evaluation — not a separate score.
Strengths
The article develops HB for dry soils and HBiM2 for wet soils by coupling Hapke, BSM, and an angularly extended MARMIT-2 structure. This directly targets the stated gap in models that separately handle spectral, directional, and moisture-driven processes.
↳ Introduction; Sections 3.1–3.3
Validation spans 10 soil textures, 14 particle sizes, multiple moisture levels, and multi-angle geometries over the 0.4–2.4 μm range. The results are reported with R², RMSE, NRMSE, MRE, and bias rather than a single aggregate statistic.
↳ Sections 4.1–4.4; Tables 3–6; Figs. 8–15
The BM2, BiM2, and HBiM2 comparison isolates the value of adding angular treatment to the wet layer alone versus both dry and wet components. Figure 11 and related analyses show HBiM2 reducing overall RMSE relative to both simpler configurations.
↳ Section 5.3; Figs. 11–15
Limitations
Section 3.2 describes a three-path wet-soil decomposition in which Eq. (31) includes direct Fresnel reflection R1, but Eq. (32) presents the complete model without clearly carrying that term through. This is a localized internal-coherence and clarity concern for a derivation-heavy paper.
↳ Section 3.2, Eqs. (28)–(32)
The evaluation reports close agreement between simulated and measured reflectance after constrained inversion, but it does not demonstrate independent soil-parameter retrieval or operational satellite/airborne workflows. The reported accuracy therefore supports model fit more strongly than deployment performance.
↳ Section 4.4; Sections 5.1–5.4
The paper gives equations and parameter ranges, but no code or independently available compiled dataset is identified, and grid-search step sizes are referenced rather than fully specified in the supplied text. This limits independent replication of the computational workflow.
↳ Section 4.4; Table 2
Withdrawn — shown for reference only
The quality score is driven by a real methodological advance: HBiM2 links spectral, angular, and moisture-related soil reflectance processes in a single RTM framework. The validation is broad for a controlled modeling paper, with results across texture classes, particle sizes, moisture levels, and angular configurations. The score is held below the highest band because Section 3.2 contains a derivational ambiguity around the full wet-soil model, and Section 4.4 frames performance through constrained inversion rather than independent prediction. Impact potential is credible but mid-stage because the article establishes a useful model component rather than an operational retrieval system.
Strong3.9/5.0
The score reflects a clear methodological contribution: coupling Hapke, BSM, and an angularly extended MARMIT-2 into a unified dry/wet soil RTM. The advance is meaningful within soil radiative transfer, but it remains bounded by controlled validation and prospective retrieval claims.
a structurally unified framework that consistently integrates spectral, directional, and moisture driven processes remains absent
The study uses physically grounded derivations, ablation-style model comparisons, and multiple fit metrics across texture, particle-size, moisture, and angular conditions. The score is limited by the Eq. (31) to Eq. (32) derivation ambiguity, absence of code, incomplete grid-search detail, and evaluation that mainly reports inversion/fitting accuracy rather than independent prediction.
the optimal parameter set was identified by minimizing the root mean square error
The paper’s sequence from gap, model construction, sensitivity analysis, and validation is easy to follow, and the comparison among HB, BM2, BiM2, and HBiM2 aids interpretation. Local clarity is reduced by the wet-soil closed-form expression not being fully reconciled with the preceding path decomposition and by occasional editing issues.
Construction of the full wet soil model
The paper engages directly with Hapke, BSM, MARMIT, and MARMIT-2 and explains why existing models only partially address spectral, directional, and moisture processes. It also states several modeling assumptions, though the bridge from laboratory/ground validation to operational remote-sensing retrieval remains more aspirational than demonstrated.
similar to the widely used PROSPECT + SAIL (PROSAIL) model
Concerns4 of 4 checks
The paper is broadly consistent in its reported performance metrics and model-comparison narrative, but the wet-soil derivation contains a localized equation-to-path mismatch that reduces confidence in the transparency of the improved MARMIT-2 formulation. The issue does not rise to a red concern because the main empirical comparisons remain interpretable, but it is a substantive internal-coherence note.
This is a computational and laboratory/previous-dataset modeling paper for which human or animal ethics approval is not evidently required. No visible conflicts, image-integrity problems, preregistration issues, or suspicious data-availability claims are present in the supplied text.
Flags: 0 declared / 5 total
42 references verified via PubMed, Crossref, or OpenAlex.
This paper was retracted on 2026-06-01. See notice: 10.1016/j.rse.2026.115381.
Sources: Retraction Watch ✓PubPeer (coming soon)
Medium3.3/5.0
The work addresses a recognized need in soil parameter retrieval, vegetation background correction, and land-surface reflectance modeling. Its traction is visible at the research-tool level, but the article does not identify specific operational adopters or decision contexts.
improving the retrieval of soil parameters
The model is derived, sensitivity-tested, and validated under controlled laboratory and ground multi-angle settings, placing it beyond a purely conceptual proposal. Readiness remains preliminary because no field/satellite retrieval workflow, uncertainty propagation, or deployment-oriented implementation is demonstrated.
strong agreement between simulated and measured reflectance under controlled conditions
Testing spans 10 soil textures, 14 particle sizes, multiple moisture levels, and a wide angular/spectral range, supporting moderate transferability within the sampled domain. Transfer to heterogeneous field scenes, mixed pixels, atmospheric effects, and broader mineralogical conditions remains untested.
covering 10 soil textures, 14 particle sizes, multiple moisture levels
The article sets out a coherent trajectory toward a PROSAIL-like soil RTM component for cumulative remote-sensing research. The next steps are identifiable, especially coupling to retrieval and vegetation-background workflows, but operational readiness is not yet shown.
highlighting its strong potential for soil parameter retrieval
Nabu’s assessment, alongside the field’s view.
Evaluation Metadata
AI-powered, human-governed. Something look off? Contact us to request a review.