Pith. sign in

REVIEW 4 major objections 6 minor 4 cited by

SatelliteCalculator: A Multi-Task Vision Foundation Model for Quantitative Remote Sensing Inversion

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read SatelliteCalculator, a frozen-backbone multi-task vision model, estimates eight ecological indicators from four-band Sentinel-2 imagery with canopy-height accuracy matching the best single-task models.

desk verdict The canopy-height result is plausible, but the 4-band input to a frozen 3-channel Swin is never specified, and the index/biomass tasks are partly circular – so the paper needs major revision before the multi-task claims can be taken at face value. read the letter →

arxiv 2504.13442 v1 pith:6E4YSBBE submitted 2025-04-18 cs.CV

Zhenyu Yu , Mohd. Yamani Idna Idris , Pei Wang This is my paper · ORCID
classification cs.CV
keywords quantitativeremotesensinginversionvisionfoundationmodelmulti-taskregressionSentinel-2canopyheightestimationabovegroundbiomasscarbonstockprompt-guidedcross-attentiveadapter
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

SatelliteCalculator is presented as the first vision foundation model built specifically for quantitative remote sensing inversion: given a four-band Sentinel-2 image and a task prompt, it regresses a dense map of any of eight ecological indicators. The paper's central claim is that a frozen Swin Transformer backbone, combined with prompt-guided cross-attentive adapters and lightweight MLP decoders, can match the accuracy of the best single-task canopy-height models (MAE 2.55 m, RMSE 4.02 m on Open-Canopy) while jointly estimating all eight variables at lower inference cost. To train this model, the authors synthesize roughly one million paired samples by applying standard physical index formulas and literature allometric equations to Open-Canopy imagery. A sympathetic reader would take away that multi-task foundation-model regression is feasible in remote sensing, and that parameter-efficient adapters can specialise shared visual features to heterogeneous physical targets.

What carries the argument

The load-bearing mechanism is the prompt-guided cross-attentive adapter: a task token (e.g., 'AGB') is mapped by a learnable embedding matrix to a query vector, and a cross-attention layer attends from that query to the visual tokens of the frozen Swin backbone, followed by a small task-specific MLP. This injects task semantics into shared features without updating the backbone; each task then gets a minimal MLP decoder, and training minimizes a weighted sum of per-task MAE losses. Equally load-bearing is the dataset-construction pipeline: five spectral indices are computed from physically defined formulas on the four Sentinel-2 bands, while canopy height comes directly from Open-Canopy LiDAR, and aboveground biomass and carbon stock are generated by the allometric equations $AGB = 0.067 H^{2.58}$ and $CS = AGB \times 0.47$.

What would settle it

Compare SatelliteCalculator's aboveground biomass and carbon-stock predictions against independent field inventory plots (or LiDAR-derived estimates from an independently calibrated allometric model) that were not used in label generation; if the error on such held-out field data is much larger than the reported RMSE of 26.04 t/ha for AGB and 21.84 Mg/ha for CS, the good numbers are an artifact of label circularity, while comparable errors confirm the allometric-labels premise.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that quantitative remote sensing inversion, predicting continuous biophysical quantities rather than class labels, can be cast as a prompt-guided multi-task regression problem and solved by a single foundation model. SatelliteCalculator takes a four-band (blue, green, red, near-infrared) Sentinel-2 image and a discrete task token, embeds the token into a query vector, and uses cross-attention between that query and the visual tokens of a frozen Swin Transformer to produce task-conditioned features, which a small MLP head decodes into a dense prediction map. Trained on a physically synthesized dataset of about one million samples over the Open-Canopy region, the model reports canopy height with MAE 2.55 m and RMSE 4.02 m, on par with the best single-task backbones, while also estimating NDVI, GNDVI, SAVI, EVI, NDWI, aboveground biomass, and carbon stock. The authors conclude that these results validate the feasibility of foundation models for physically interpretable regression and provide a scalable, modular template for task-adaptive remote sensing estimation.

Load-bearing premise

The reported biomass and carbon-stock accuracies assume that the literature allometric equation $AGB = 0.067 \times H^{2.58}$ and the carbon fraction $0.47$ are true for the French open-canopy forests in Open-Canopy; if those equations do not hold there, those two tasks' accuracies are circular artifacts of the same equations used to create the labels.

Editorial extensions

If this is right

  • Adding a new ecological variable costs only a new prompt token, cross-attentive adapter, and MLP decoder, since the Swin backbone stays frozen; this makes expansion to new inversion targets cheap.
  • The model can be trained on physically synthesized labels, avoiding the annotation bottleneck for spectral indices, while structural variables still need reliable ground truth, here LiDAR plus allometry.
  • Multi-task training does not hurt per-task accuracy: canopy height matches the best single-task backbones on Open-Canopy, so a shared feature space suffices.
  • Spectral index tasks (NDVI, GNDVI, SAVI, EVI, NDWI) are close to solved (RMSE 0.22–0.51), whereas structural tasks (H, AGB, CS) show lower R² values, locating the remaining difficulty in the reflectance-to-structure link rather than in the architecture.
  • The MLP decoder keeps inference cheap (0.48 s/image and 1566 MB GPU memory), making country-scale wall-to-wall mapping feasible.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the prompt-guided adapter design transfers beyond France, the same frozen backbone could support operational global biomass and carbon mapping from Sentinel-2, since per-task adaptation requires only lightweight modules.
  • Because the biomass and carbon labels are deterministic functions of the LiDAR height, the R² values of 0.55 and 0.52 on those tasks may partly measure how well the model recovers height rather than true biophysical accuracy; an independent field-data test would separate the two.
  • The four-band input choice, with no shortwave infrared, is a testable bottleneck for structural variables; feeding the same adapters eight or twelve bands would reveal whether cross-attention can absorb additional spectral information without unfreezing the backbone.
  • The same dataset-synthesis recipe, physical formulas plus an empirical allometric chain, could be reused to generate training labels for other continuous quantities such as leaf area index, soil moisture, or water quality where closed-form forward models exist.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper introduces SatelliteCalculator, a multi-task vision foundation model for quantitative remote sensing inversion. The model takes four-band Sentinel-2 imagery (B2, B3, B4, B8) and a task prompt as input, and jointly predicts five spectral indices (NDVI, GNDVI, SAVI, EVI, NDWI) and three structural variables (canopy height H, aboveground biomass AGB, and carbon stock CS). The architecture combines a frozen Swin Transformer backbone with prompt embeddings, cross-attentive adapters, and lightweight task-specific MLP decoders. The training dataset is constructed from the Open-Canopy benchmark by computing the spectral indices via closed-form formulas and deriving AGB and CS from LiDAR-derived canopy height through allometric equations. Experiments on Open-Canopy report canopy-height accuracy comparable to strong single-task baselines (MAE 2.55 m, RMSE 4.02 m) and include ablations over decoder architectures and MLP depth. The paper claims to be the first vision foundation model tailored to quantitative remote sensing inversion and releases code and data.

Significance. If the technical gaps identified below are resolved, the work could contribute a scalable, parameter-efficient recipe for multi-task regression in remote sensing: the synthetic one-million-sample dataset construction is a practical idea, the prompt-guided cross-attentive adapter is a sensible mechanism, and the efficiency analysis is useful. However, the manuscript currently contains a load-bearing architectural omission and missing baselines that substantially weaken the evidence for the central claims. The work is of interest to the remote sensing and vision foundation model communities, but the claims of being the first quantitative-inversion foundation model and of achieving competitive accuracy across all eight tasks are not yet supported by the presented experiments.

major comments (4)
  1. [§III-C, Eq. (2); §III-E] The paper states that a four-band input X∈R^{H×W×4} is fed to a Swin Transformer initialized from ImageNet-21k and kept entirely frozen. A standard Swin-T patch embedding is a 3-channel convolutional layer; the manuscript does not specify how the fourth band is handled. If a newly initialized projection is inserted before the frozen backbone, then the claim in §III-E that only the prompt embedding layer, cross-attentive adapters, and task-specific decoders are updated is false. If a band is dropped or handled by padding, the four-band and multispectral claims are inaccurate. This is a load-bearing omission because every reported result depends on the actual input pipeline; as written, the architecture is not instantiable or reproducible.
  2. [§IV-D3, Table III; Appendix A, Eqs. (A1)–(A5)] The five spectral index targets are deterministic closed-form functions of the four input bands. The paper reports R² values of 0.66–0.86 for these tasks but never compares against the trivial baseline that computes the index directly from the input using the same formulas. Such a baseline would achieve essentially zero error, so the reported RMSE values (e.g., 0.22 for NDVI) indicate that the model is not learning the exact relationship. Without this comparison, the index results do not support the claim that SatelliteCalculator performs quantitative inversion; they also raise the question of whether the multi-task loss is interfering with the model's ability to learn these simple, physically defined mappings.
  3. [§IV-A; Appendix A, Eqs. (A6)–(A8)] The AGB and CS labels are generated from LiDAR-derived canopy height H via fixed allometric equations with literature coefficients a, b, and CF. The reported AGB RMSE of 26.04 t/ha and CS RMSE of 21.84 Mg/ha therefore measure the model's ability to reproduce a deterministic transform of the height labels, not directly measured biophysical quantities. The manuscript should explicitly discuss this label-derivation chain, justify the transfer of the allometric coefficients to the Open-Canopy forests in France, and ideally report a sensitivity analysis over a, b, and CF. This issue affects two of the three structural tasks and the overall multi-task claim.
  4. [Tables I, III, IV, V] All results are reported as single point estimates without error bars or statistical significance tests. Several comparisons hinge on small margins (e.g., MAE 2.55 vs 2.52 for PVTv2 in Table I; MLP vs ResNet in Table IV), and the ablation conclusions (e.g., that a 4-layer MLP is best) are based on differences that may be within run-to-run variation. The paper should report means and standard deviations over multiple seeds or otherwise establish that the observed differences are statistically meaningful.
minor comments (6)
  1. [Figure 3] The two "Groundtruth" labels in each row appear to be a labeling error; the second row should likely be labeled with the model name or a scene identifier.
  2. [Table A5] The R² values quoted in the text (0.68 for 4 layers) do not match the table values (0.6816); please make the reported precision consistent.
  3. [§IV-B, Eq. (6)] The text says "The model training employs the ℓ1 norm as the loss function" while Eq. (6) defines a weighted MAE; please unify the terminology.
  4. [§IV-C] The definition of nMAE is incomplete: it says the error is normalized by the target height, but the exact normalization formula is not given.
  5. [Appendix A, §A vs §C] The appendix introduction says structural variables "are estimated from regression models using NDVI, SAVI, and LiDAR-derived height information," but §C states that H is obtained directly from Open-Canopy and that AGB/CS are derived from H alone; please reconcile these descriptions.
  6. [Abstract and Introduction] The claim of being the "first vision foundation model tailored for quantitative remote sensing inversion" should be qualified with respect to existing regression-oriented remote sensing foundation models or accompanied by a clearer definition of "tailored."

Circularity Check

2 steps flagged · score 7.0 of 10

Seven of the eight 'inversion' targets are deterministic transforms of the model's own input bands or of another target, so the reported multi-task accuracies mostly measure consistency with the label-generation formulas rather than independent geophysical inversion.

  1. self definitional [Section IV-A and Appendix A-B, Eqs. (A1)-(A5); Table III]
    "we use spectral reflectance from four Sentinel-2 bands (B2–B4, B8) to compute five vegetation and water indices—NDVI, GNDVI, EVI, SAVI, and NDWI—using standard physical formulas. ... spectral indices are physically well-defined and can be directly computed from the input bands"

    Each of the five spectral-index targets (Eqs. A1-A5) is a closed-form function of exactly the same four bands that form the model input X in R^{H x W x 4}. Learning to predict NDVI/GNDVI/SAVI/EVI/NDWI from those bands is therefore function approximation of the very formula used to generate the labels; a trivial analytic baseline that evaluates Eq. (A1) directly on the test bands would achieve zero RMSE by construction. The paper reports no such baseline for the index tasks, so the R2/RMSE values for these five tasks validate how well the network memorized the arithmetic formulas, not an independent quantitative inversion.

  2. renaming known result [Appendix A-C, Eqs. (A6)-(A8); Section IV-A; Table III]
    "AGB = a x H^b (A6) ... CS = AGB x CF (A8) ... Structural and biomass-related labels are derived by directly using the provided canopy height (H) from LiDAR data, while aboveground biomass (AGB) and carbon stock (CS) are estimated through regionally calibrated empirical models."

    Canopy height H is itself one of the eight targets. With the stated coefficients (a=0.067, b=2.58, CF=0.47), AGB and CS are deterministic monotone functions of H, so their ground-truth maps contain no information beyond the H map. The reported AGB RMSE (26.04 t/ha) and CS RMSE (21.84 Mg/ha) therefore measure self-consistency with the assumed allometric equations applied to the model's H estimates, not agreement with independently measured biomass or carbon. The three structural 'tasks' reduce to one measured quantity renamed through fixed empirical coefficients.

full rationale

The only target with genuinely independent ground truth is canopy height H, which comes from LiDAR in the external Open-Canopy benchmark; for H, the comparison against single-task models is a legitimate external evaluation. The five vegetation/water indices, however, are by definition closed-form functions of the input bands, and AGB/CS are by definition closed-form functions of H. Thus seven of the eight tasks reduce by the paper's own equations to the label-generation formulas, making the multi-task quantitative-inversion claim substantially circular. The four-band input to a frozen three-channel Swin backbone is a serious reproducibility and instantiation defect, but it is not a derivation that reduces to its own input, so it is not counted toward the circularity score. I find no load-bearing self-citation: the allometric coefficients are attributed to external literature (refs [44]-[46]) and the benchmark to an external dataset (ref [32]). Overall score 7, reflecting that the central quantitative demonstration is largely forced by definition while the H task retains independent content.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim rests on synthetic labels generated from the input bands and from the height target. The allometric coefficients are chosen by hand, and the loss weights are fit to initial losses. No new physical entities are introduced, but the validity of the AGB and CS targets is an unverified domain assumption.

free parameters (2)
  • Task loss weights lambda_t = NDVI 0.0386, GNDVI 0.0440, SAVI 0.0501, EVI 0.1700, NDWI 0.0418, H 0.2052, AGB 0.2121, CS 0.2381
    The loss weights in Table II are derived from initial training loss magnitudes, so they are fit to the data at hand. They influence the balancing of the multi-task objective and are not justified by a principled criterion beyond the heuristic described in Section IV.D.2.
  • Allometric coefficients a, b for AGB and carbon fraction CF = General model a=0.067, b=2.58, CF=0.47 (the paper also lists coniferous, broadleaf, and mixed coefficients)
    The paper cites typical values from the literature but does not specify which set of coefficients was used to generate the dataset or how forest type was assigned per pixel. The choice of coefficients directly determines the AGB and CS ground truth, so it is a hand-chosen parameter that affects the labels.
assumptions (3)
  • domain assumption The allometric equations AGB = a H^b and CS = AGB * CF are valid representations of true AGB and carbon stock in the Open-Canopy forests.
    This is invoked in Appendix A, Eqs. A6-A8, where the paper generates AGB and CS labels from LiDAR height. No ground truth from field plots is used to validate these equations in the study area.
  • domain assumption The frozen Swin Transformer, pretrained on ImageNet, can accept and process four-band Sentinel-2 imagery as effectively as three-band RGB imagery.
    The paper specifies a four-band input in Section III.A and uses a frozen Swin backbone in Section III.C, but does not explain how the four channels are mapped to the three-channel pretrained patch embedding. This is an unsupported implementation assumption.
  • domain assumption The Open-Canopy LiDAR-derived canopy height maps are accurate and aligned with the Sentinel-2 VHR imagery.
    The paper relies on this alignment in Section IV.A to create paired samples. Errors in geo-registration or LiDAR processing would propagate to all derived labels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SatelliteCalculator: A Multi-Task Vision Foundation Model for Quantitative Remote Sensing Inversion." pith.science (2026). https://pith.science/paper/6E4YSBBE

@misc{pith2026250413442,
  author       = {Pith},
  title        = {Pith review of: SatelliteCalculator: A Multi-Task Vision Foundation Model for Quantitative Remote Sensing Inversion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6E4YSBBE}},
  note         = {Machine review of arXiv:2504.13442}
}
read the original abstract

Quantitative remote sensing inversion plays a critical role in environmental monitoring, enabling the estimation of key ecological variables such as vegetation indices, canopy structure, and carbon stock. Although vision foundation models have achieved remarkable progress in classification and segmentation tasks, their application to physically interpretable regression remains largely unexplored. Furthermore, the multi-spectral nature and geospatial heterogeneity of remote sensing data pose significant challenges for generalization and transferability. To address these issues, we introduce SatelliteCalculator, the first vision foundation model tailored for quantitative remote sensing inversion. By leveraging physically defined index formulas, we automatically construct a large-scale dataset of over one million paired samples across eight core ecological indicators. The model integrates a frozen Swin Transformer backbone with a prompt-guided architecture, featuring cross-attentive adapters and lightweight task-specific MLP decoders. Experiments on the Open-Canopy benchmark demonstrate that SatelliteCalculator achieves competitive accuracy across all tasks while significantly reducing inference cost. Our results validate the feasibility of applying foundation models to quantitative inversion, and provide a scalable framework for task-adaptive remote sensing estimation.

Figures

Figures reproduced from arXiv: 2504.13442 by the authors.

Figure 1
Figure 1. Overview of the SatelliteCalculator framework. The model takes a multi-band Sentinel-2 VHR image and a task prompt as input, and performs quantitative inversion through four components: (1) prompt embedding module, (2) Swin Transformer feature extractor, (3) Cross-attentive adapter, and (4) Task-specific decoder. under a unified multi-task loss. This formulation provides a flexible interface for expanding the model … view at source ↗
Figure 2
Figure 2. Comparison of canopy height inversion results across different models. From left to right: Sentinel-2 VHR input, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Multi-task inversion results on four Sentinel-2 scenes. We show groundtruth (top) and [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Effect of MLP layer depth on decoder performance. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. SatelliteFormula: Multi-Modal Symbolic Regression from Remote Sensing Imagery for Physics Discovery

    cs.CV 2025-06 reject novelty 4.0 of 10

    SatelliteFormula couples a Swin Transformer image encoder with a symbolic regression decoder to generate expressions for indices such as NDVI and biomass from satellite imagery.

  2. A Multimodal Deep Learning Framework for Early Diagnosis of Liver Cancer via Optimized BiLSTM-AM-VMD Architecture

    cs.LG 2025-09 reject novelty 3.0 of 10

    The paper claims a BiLSTM-AM-VMD model achieves AUC 0.963 for early HCC diagnosis, but the evidence is undermined by contradictory dataset descriptions and missing artifacts.

  3. Multi-Modal Machine Learning Framework for Predicting Early Recurrence of Brain Tumors Using MRI and Clinical Biomarkers

    cs.LG 2025-09 reject novelty 3.0 of 10

    XGBoost combining MRI radiomics and clinical biomarkers reportedly reaches C-index 0.782 for early brain tumor recurrence, but the paper's methods describe a liver-cancer cohort and no evaluation of its claimed tempor...

  4. From Physics to Foundation Models: A Review of AI-Driven Quantitative Remote Sensing Inversion

    cs.CV 2025-07 reject novelty 1.0 of 10

    A review of quantitative remote sensing inversion that traces the shift from physics-based models through machine learning to foundation models, but with incomplete coverage and citation problems.

Reference graph

Works this paper leans on

54 extracted references · 45 canonical work pages · cited by 4 Pith papers

  1. [1]

    Remote sensing of environment: State of the science and new directions,

    S. L. Ustin, “Remote sensing of environment: State of the science and new directions,” Remote sensing of natural resources management and environmental monitoring, 2004. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, APRIL 2025 9

    work page 2004

  2. [2]

    The enduring world forest carbon sink,

    Y . Pan, R. A. Birdsey, O. L. Phillips, R. A. Houghton, J. Fang, P. E. Kauppi, H. Keith, W. A. Kurz, A. Ito, S. L. Lewis et al., “The enduring world forest carbon sink,” Nature, vol. 631, no. 8021, pp. 563–569, 2024

    work page 2024

  3. [3]

    Impact of climate change on sars-cov-2 epidemic in china,

    Z. Yu, J. Wang, Z. Tan, and Y . Luo, “Impact of climate change on sars-cov-2 epidemic in china,” Plos one , vol. 18, no. 7, p. e0285179, 2023

    work page 2023

  4. [4]

    Quantitative remote sensing of land surfaces,

    J. Hogg, “Quantitative remote sensing of land surfaces,” The Photogram- metric Record, vol. 19, no. 108, pp. 413–415, 2004

    work page 2004

  5. [5]

    Spatiotemporal analysis of xco2 and its relationship to urban and green areas of china’s major southern cities from remote sensing and wrf-chem modeling data from 2010 to 2019,

    Z. Tan, J. Wang, Z. Yu, and Y . Luo, “Spatiotemporal analysis of xco2 and its relationship to urban and green areas of china’s major southern cities from remote sensing and wrf-chem modeling data from 2010 to 2019,” Geographies, vol. 3, no. 2, pp. 246–267, 2023

    work page 2010

  6. [6]

    Deep learning in remote sensing: A comprehensive review and list of resources,

    X. X. Zhu, D. Tuia, L. Mou, G. S. Xia, L. Zhang, F. Xu, and F. Fraundorfer, “Deep learning in remote sensing: A comprehensive review and list of resources,” IEEE Geoscience & Remote Sensing Magazine, vol. 5, no. 4, pp. 8–36, 2018

    work page 2018

  7. [7]

    Capan: Class-aware prototypical adversarial net- works for unsupervised domain adaptation,

    Z. Yu and P. Wang, “Capan: Class-aware prototypical adversarial net- works for unsupervised domain adaptation,” in 2024 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2024, pp. 1–6

    2024

  8. [8]

    Remote sensing object detection in the deep learning era—a review,

    S. Gui, S. Song, R. Qin, and Y . Tang, “Remote sensing object detection in the deep learning era—a review,” Remote Sensing, vol. 16, no. 2, p. 327, 2024

    work page 2024

Show all 54 references
  1. [9]

    Multi-batch nuclear-norm adversarial net- work for unsupervised domain adaptation,

    P. Wang, Y . Yang, and Z. Yu, “Multi-batch nuclear-norm adversarial net- work for unsupervised domain adaptation,” in 2024 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2024, pp. 1–6

    2024

  2. [10]

    Geography- aware self-supervised learning,

    K. Ayush, B. Uzkent, C. Meng, M. Burke, and S. Ermon, “Geography- aware self-supervised learning,” 2020

    2020

  3. [11]

    Yuan: Yielding unblemished aesthetics through a unified network for visual imperfections removal in generated images,

    Z. Yu and C. S. Chan, “Yuan: Yielding unblemished aesthetics through a unified network for visual imperfections removal in generated images,” arXiv preprint arXiv:2501.08505 , 2025

    2025 arXiv

  4. [12]

    Remoteclip: A vision language foundation model for remote sensing,

    F. Liu, D. Chen, Z. Guan, X. Zhou, J. Zhu, Q. Ye, L. Fu, and J. Zhou, “Remoteclip: A vision language foundation model for remote sensing,” IEEE Transactions on Geoscience and Remote Sensing , 2023

    2023

  5. [13]

    Ast: Adaptive self-supervised transformer for optical remote sensing representation,

    Q. He, X. Sun, Z. Yan, B. Wang, Z. Zhu, W. Diao, and M. Y . Yang, “Ast: Adaptive self-supervised transformer for optical remote sensing representation,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 200, pp. 41–54, 2023

    2023

  6. [14]

    Qrs-trs: Style transfer-based image-to-image translation for carbon stock estimation in quantitative remote sensing,

    Z. Yu, J. Wang, H. Chen, and M. Y . I. Idris, “Qrs-trs: Style transfer-based image-to-image translation for carbon stock estimation in quantitative remote sensing,” IEEE Access, 2025

    2025

  7. [15]

    Red and photographic infrared linear combinations for monitoring vegetation,

    C. J. Tucker, “Red and photographic infrared linear combinations for monitoring vegetation,” Remote sensing of Environment , vol. 8, no. 2, pp. 127–150, 1979

    1979

  8. [16]

    Ai for science: A comprehensive review on innovations, challenges, and future directions,

    Z. Yu, “Ai for science: A comprehensive review on innovations, challenges, and future directions,” International Journal of Artificial Intelligence for Science (IJAI4S) , vol. 1, no. 1, 2025

    2025

  9. [17]

    Improved implicit diffusion model with knowledge distillation to estimate the spatial distribution density of carbon stock in remote sensing imagery,

    ——, “Improved implicit diffusion model with knowledge distillation to estimate the spatial distribution density of carbon stock in remote sensing imagery,” arXiv preprint arXiv:2411.17973 , 2024

    2024 arXiv

  10. [18]

    Deep-learning-based automatic extraction of aquatic veg- etation from sentinel-2 images—a case study of lake honghu,

    H. Gao, R. Li, Q. Shen, Y . Yao, Y . Shao, Y . Zhou, W. Li, J. Li, Y . Zhang, and M. Liu, “Deep-learning-based automatic extraction of aquatic veg- etation from sentinel-2 images—a case study of lake honghu,” Remote Sensing, vol. 16, no. 5, p. 867, 2024

    2024

  11. [19]

    When geoscience meets foundation models: Toward a general geoscience artificial intelligence system,

    H. Zhang, J.-J. Xu, H.-W. Cui, L. Li, Y . Yang, C.-S. Tang, and N. Boers, “When geoscience meets foundation models: Toward a general geoscience artificial intelligence system,” IEEE Geoscience and Remote Sensing Magazine, 2024

    2024

  12. [20]

    A machine learning scheme for estimating fine-resolution grassland aboveground biomass over china with sentinel-1/2 satellite images,

    H. Li, F. Li, J. Xiao, J. Chen, K. Lin, G. Bao, A. Liu, and G. Wei, “A machine learning scheme for estimating fine-resolution grassland aboveground biomass over china with sentinel-1/2 satellite images,” Remote Sensing of Environment , vol. 311, p. 114317, 2024

    2024

  13. [21]

    Forest aboveground biomass estimation using multisource remote sensing data and deep learning algorithms: a case study over hangzhou area in china,

    X. Tian, J. Li, F. Zhang, H. Zhang, and M. Jiang, “Forest aboveground biomass estimation using multisource remote sensing data and deep learning algorithms: a case study over hangzhou area in china,” Remote Sensing, vol. 16, no. 6, p. 1074, 2024

    2024

  14. [22]

    Land cover classification in high-resolution remote sensing: using swin transformer deep learning with texture features,

    Y . Zhang, M. Huang, Y . Chen, X. Xiao, and H. Li, “Land cover classification in high-resolution remote sensing: using swin transformer deep learning with texture features,” Journal of Spatial Science , pp. 1– 25, 2024

    2024

  15. [23]

    Estimating forest aboveground biomass in trop- ical forests with landsat time-series data and recurrent neural network,

    Q. Zhang and Q. Zhuo, “Estimating forest aboveground biomass in trop- ical forests with landsat time-series data and recurrent neural network,” International Journal of Remote Sensing, vol. 45, no. 14, pp. 4764–4787, 2024

    2024

  16. [24]

    Forecasting vegetation indices from spatio-temporal remotely sensed data using deep learning-based approaches: A systematic literature review,

    A. Ferchichi, A. B. Abbes, V . Barra, and I. R. Farah, “Forecasting vegetation indices from spatio-temporal remotely sensed data using deep learning-based approaches: A systematic literature review,” Ecological Informatics, vol. 68, p. 101552, 2022

    2022

  17. [25]

    Spatial-aware sar-optical time- series deep integration for crop phenology tracking,

    W. Zhao, Y . Qu, L. Zhang, and K. Li, “Spatial-aware sar-optical time- series deep integration for crop phenology tracking,” Remote Sensing of Environment, vol. 276, p. 113046, 2022

    2022

  18. [26]

    Self-supervised spatio-temporal representation learning of satellite image time series,

    I. Dumeur, S. Valero, and J. Inglada, “Self-supervised spatio-temporal representation learning of satellite image time series,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , vol. 17, pp. 4350–4367, 2024

    2024

  19. [27]

    Masked autoencoders for microscopy are scalable learners of cellular biology,

    O. Kraus, K. Kenyon-Dean, S. Saberian, M. Fallah, P. McLean, J. Le- ung, V . Sharma, A. Khan, J. Balakrishnan, S. Celik et al. , “Masked autoencoders for microscopy are scalable learners of cellular biology,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pat...

    2024

  20. [28]

    Sd- dit: Unleashing the power of self-supervised discrimination in diffusion transformer,

    R. Zhu, Y . Pan, Y . Li, T. Yao, Z. Sun, T. Mei, and C. W. Chen, “Sd- dit: Unleashing the power of self-supervised discrimination in diffusion transformer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 8435–8445

    2024

  21. [29]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” in Proceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4015–4026

    2023

  22. [30]

    Satmae: Pre-training transformers for temporal and multi-spectral satellite imagery,

    Y . Cong, S. Khanna, C. Meng, P. Liu, E. Rozi, Y . He, M. Burke, D. Lo- bell, and S. Ermon, “Satmae: Pre-training transformers for temporal and multi-spectral satellite imagery,” Advances in Neural Information Processing Systems, vol. 35, pp. 197–211, 2022

    2022

  23. [31]

    Geogpt: Understanding and processing geospatial tasks through an autonomous gpt,

    Y . Zhang, C. Wei, S. Wu, Z. He, and W. Yu, “Geogpt: Understanding and processing geospatial tasks through an autonomous gpt,” 2023

    2023

  24. [32]

    Open-canopy: A country-scale benchmark for canopy height estimation at very high resolution,

    F. Fogel, Y . Perron, N. Besic, L. Saint-Andr ´e, A. Pellissier-Tanon, M. Schwartz, T. Boudras, I. Fayad, A. d’Aspremont, L. Landrieu et al., “Open-canopy: A country-scale benchmark for canopy height estimation at very high resolution,” arXiv preprint arXiv:2407.09392 , 2024

    2024 arXiv

  25. [33]

    Root-mean-square error (rmse) or mean absolute error (mae): when to use them or not,

    T. Hodson, “Root-mean-square error (rmse) or mean absolute error (mae): when to use them or not,” Geoscientific Model Development , vol. 15, no. 14, 2022

    2022

  26. [34]

    Na ¨ıve filterbots for robust cold-start recommendations,

    S. T. Park, D. Pennock, O. Madani, N. Good, and D. Decoste, “Na ¨ıve filterbots for robust cold-start recommendations,” ACM, 2006

    2006

  27. [35]

    J. A. C. Sterne and M. Egger, Regression Methods to Detect Publication and Other Bias in Meta-Analysis . Publication Bias in Meta-Analysis: Prevention, Assessment and Adjustments, 2006

    2006

  28. [36]

    Urban tree cover changes using object-based convolution neural network (ob-cnn),

    S. Timilsina, J. Aryal, and J. B. Kirkpatrick, “Urban tree cover changes using object-based convolution neural network (ob-cnn),” Remote Sens- ing, vol. 12, no. 18, pp. 1–27, 2020

    2020

  29. [37]

    A note on a general definition of the coefficient of determination,

    N. J. Nagelkerke et al., “A note on a general definition of the coefficient of determination,” biometrika, vol. 78, no. 3, pp. 691–692, 1991

    1991

  30. [38]

    Image quality metrics: Psnr vs. ssim,

    A. Hore and D. Ziou, “Image quality metrics: Psnr vs. ssim,” in 2010 20th international conference on pattern recognition . IEEE, 2010, pp. 2366–2369

    2010

  31. [39]

    Satellite remote sensing of vegetation phenology: Progress, challenges, and opportunities,

    Z. Gong, W. Ge, J. Guo, and J. Liu, “Satellite remote sensing of vegetation phenology: Progress, challenges, and opportunities,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 217, pp. 149–164, 2024

    2024

  32. [40]

    Signature analysis of leaf reflectance spectra: algorithm development for remote sensing of chloro- phyll,

    A. A. Gitelson and M. N. Merzlyak, “Signature analysis of leaf reflectance spectra: algorithm development for remote sensing of chloro- phyll,” Journal of plant physiology, vol. 148, no. 3-4, pp. 494–500, 1996

    1996

  33. [41]

    A soil-adjusted vegetation index (savi),

    A. R. Huete, “A soil-adjusted vegetation index (savi),” Remote sensing of environment, vol. 25, no. 3, pp. 295–309, 1988

    1988

  34. [42]

    Overview of the radiometric and biophysical performance of the modis vegetation indices,

    A. Huete, K. Didan, T. Miura, E. P. Rodriguez, X. Gao, and L. G. Ferreira, “Overview of the radiometric and biophysical performance of the modis vegetation indices,” Remote sensing of environment , vol. 83, no. 1-2, pp. 195–213, 2002

    2002

  35. [43]

    The use of the normalized difference water index (ndwi) in the delineation of open water features,

    S. K. McFeeters, “The use of the normalized difference water index (ndwi) in the delineation of open water features,” International journal of remote sensing , vol. 17, no. 7, pp. 1425–1432, 1996

    1996

  36. [44]

    Improved allometric models to estimate the aboveground biomass of tropical trees,

    J. Chave, M. R ´ejou-M´echain, A. B ´urquez, E. Chidumayo, M. S. Colgan, W. B. Delitti, A. Duque, T. Eid, P. M. Fearnside, R. C. Goodman et al., “Improved allometric models to estimate the aboveground biomass of tropical trees,” Global change biology , vol. 20, no. 10, pp. 317...

    2014

  37. [45]

    Role of woody biomass in carbon capture, circular bioeconomy, and biomanufacturing,

    T. Raj, S. Verma, N. Kumar, and R. Agrawal, “Role of woody biomass in carbon capture, circular bioeconomy, and biomanufacturing,” in Sustainable Biorefining of Woody Biomass to Biofuels and Biochemicals. Elsevier, 2024, pp. 291–318

    2024

  38. [46]

    Including methane emissions from agricultural ponds in national greenhouse gas inventories,

    M. E. Malerba, T. de Kluyver, N. Wright, O. Omosalewa, and P. I. Macreadie, “Including methane emissions from agricultural ponds in national greenhouse gas inventories,” Environmental Science & Tech- nology, vol. 58, no. 19, pp. 8349–8359, 2024. JOURNAL OF LATEX CLASS FILES, V...

    2024

  39. [47]

    It utilizes the difference between near-infrared (NIR, B8) and red (B4) reflectance to assess chlorophyll activity in plants

    Normalized Difference Vegetation Index (NDVI) [39] NDVI = B8−B4 B8 +B4 (A1) NDVI is widely used to measure vegetation health and biomass productivity. It utilizes the difference between near-infrared (NIR, B8) and red (B4) reflectance to assess chlorophyll activity in plants

  40. [48]

    Green Normalized Difference Vegetation Index (GNDVI) [40] GNDVI = B8−B3 B8 +B3 (A2) GNDVI is a modification of NDVI that enhances sensi- tivity to vegetation chlorophyll content by incorporating the green band (B3) instead of the red band (B4)

  41. [49]

    Soil-Adjusted Vegetation Index (SA VI) [41] SAVI = (B8−B4)× (1 +L) B8 +B4 +L (A3) whereL = 0.5 is the soil adjustment factor, which min- imizes soil brightness effects in low-vegetation areas

  42. [50]

    EVI improves upon NDVI by reducing atmospheric and background noise, making it more suitable for dense vegetation monitoring

    Enhanced Vegetation Index (EVI) [42] EVI =G× (B8−B4) B8 +C1×B4−C2×B2 +L (A4) where G = 2 .5, C1 = 6 , C2 = 7 .5, and L = 1 . EVI improves upon NDVI by reducing atmospheric and background noise, making it more suitable for dense vegetation monitoring

  43. [51]

    It leverages the high reflectance of water in the green band (B3) and the strong absorption in the NIR band (B8)

    Normalized Difference Water Index (NDWI) [43] NDWI = B3−B8 B3 +B8 (A5) NDWI is used to monitor water bodies, distinguishing open water from land features. It leverages the high reflectance of water in the green band (B3) and the strong absorption in the NIR band (B8). C. Ecolo...

  44. [52]

    Canopy Height (H) The canopy height data is directly obtained from the Open-Canopy dataset [32] and is not derived from NDVI or any regression model

  45. [53]

    Aboveground Biomass (AGB) [44] AGB =a×H b (A6) where H represents canopy height obtained from the Open-Canopy dataset, and a,b are empirical constants derived from literature or site-specific calibration. For temperate forests in France, typical values from lit- erature sugges...

  46. [54]

    CF = 0.47 is the carbon fraction factor, indicating that approximately 47% of the total biomass is composed of carbon, as suggested by IPCC guidelines

    Carbon Stock (CS) [46] CS =AGB×CF (A8) where CS represents the carbon stock (tC/ha), which is the total amount of carbon stored in aboveground biomass.AGB refers to the aboveground biomass (t/ha), estimated from canopy height using empirical models. CF = 0.47 is the carbon fra...

    2025

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.