Pith. sign in

REVIEW 4 major objections 5 minor 35 references

A conditional diffusion model reconstructs 1-meter Arctic sea ice roughness from 10-meter satellite images, reaching 9 cm error on unseen terrain.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 05:20 UTC pith:YUPLI5T3

load-bearing objection Worth a serious look: strong new application, candid self-assessment, but the headline 9 cm RMSE hides a trivial-baseline problem the authors should fix before publication. the 4 major comments →

arxiv 2607.13371 v1 pith:YUPLI5T3 submitted 2026-07-15 cs.CV

RoughNet: Mapping Arctic Sea Ice Roughness Using Diffusion-Based Super-Resolution of Satellite Imagery

Tessa Cannon , Michel Tsamados , Petru Manescu , Thomas Newman , Christian Haas , Veit Helm , Weibin Chen , Randall Scharien This is my paper
classification cs.CV
keywords sea ice roughnessconditional diffusionsuper-resolutionSentinel-2airborne LiDARArctic landfast icegenerative modeltopography reconstruction
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper claims that fine-scale (1 m) sea ice surface roughness—the elevation residuals that define how rough or smooth ice is—can be reconstructed directly from coarse (10 m) optical satellite images using a conditional diffusion model. RoughNet, a U-Net-based diffusion backbone conditioned on six multi-temporal Sentinel-2 views, maps those images to locally demeaned elevation-residual fields. Trained on airborne LiDAR from two Arctic regions and tested on an unseen third, the best configuration achieves an out-of-domain RMSE of about 9 cm, with low error in roughness amplitude and near-identical elevation distributions (JSD ≈ 0.08). The authors argue that preserving the statistical and spectral properties of the roughness field, rather than exact pixel alignment, is the right success criterion, since roughness statistics—not exact point heights—are what climate models and over-ice travel planning need. If correct, this provides a scalable, low-cost pathway to high-resolution roughness mapping from publicly available satellite data.

Core claim

The central discovery is that a conditional denoising diffusion model can learn a mapping G: X → Y_res from six 10 m Sentinel-2 multispectral patches to 1 m locally demeaned surface elevation residuals over winter landfast sea ice. On the unseen test region (Cambridge Bay), the best model (cosine noise schedule with PLMS sampling) produces residuals with RMSE ≈ 0.09 m, σ-error ≈ 18%, NAE ≈ 1.2°, and JSD ≈ 0.08, indicating close agreement between predicted and true elevation distributions and orientation statistics. However, zero-mean normalized cross-correlation drops to 0.11 out-of-domain, showing that the model captures amplitude and statistical realism better than exact spatial placement

What carries the argument

The key machinery is a conditional diffusion model based on a U-Net that denoises a noisy target LiDAR patch towards the clean residual field, conditioned on fused multi-temporal Sentinel-2 views and acquisition metadata. The conditioning is aggregated by a permutation-invariant, attribute-aware softmax weighting over the six views, and the reverse process uses a cosine noise schedule with the PLMS (Pseudo Linear Multi-Step) sampler. The target representation—locally demeaned elevation residuals, obtained by fitting a quadratic surface and subtracting it—isolates roughness from absolute elevation, which is crucial because absolute heights are region-dependent while roughness statistics gener

Load-bearing premise

The approach assumes that the 10 m optical reflectance of winter landfast ice carries enough information about 1 m surface-height residuals—and that the ice remains essentially unchanged during the ±14 day window between the satellite and LiDAR acquisitions—so that the learned mapping is physically meaningful rather than a statistical coincidence.

What would settle it

Take the trained model and feed it Sentinel-2 patches whose spatial structure has been destroyed (e.g., the same power spectrum but random phases, or randomly permuted image tiles). If the output roughness statistics (JSD, variogram range, log-PSD) remain essentially unchanged compared to real inputs, the model is reproducing a learned roughness prior rather than conditioning on the optical signal, and the claimed reconstruction link is not doing the work.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the central claim holds, 1 m sea ice roughness maps can be produced from widely available, publicly funded satellite imagery, substantially increasing spatial and temporal coverage compared to airborne LiDAR campaigns.
  • The reconstructed fields preserve the log-normal elevation distribution and spatial continuity (variograms) of real sea ice, enabling synthetic topography generation for altimetry simulations and roughness parameterizations in climate models.
  • The model's stability in the presence of LiDAR collection artifacts suggests it could be used to detect or mitigate inconsistencies in airborne survey data.
  • For over-ice travel safety in Arctic communities, the approach could supply high-resolution roughness information for route planning, provided the stochastic variability between sampling runs is characterized and confidence maps are added.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because the conditioning is purely optical, the model may be implicitly learning a relationship between surface roughness and the texture/shadow patterns in visible/NIR reflectance; a targeted test would be to compare RoughNet outputs to independent measurements of snow grain size or surface faceting to see if roughness estimates correlate with these optical drivers.
  • The sharp drop in ZNCC on out-of-domain data suggests the model may rely on a learned prior over sea-ice texture as much as on the specific image content. A permutation test—feeding phase-scrambled or spatially permuted Sentinel-2 patches and checking whether output statistics change—would quantify how much of the prediction is truly image-conditional.
  • The fixed six-view conditioning and reliance on daylight optics are limiting for operational use; a natural extension would be to adapt the same conditional diffusion framework to radar backscatter inputs, which are illumination-independent and available year-round, though at coarser resolution.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes RoughNet, a conditional diffusion model that maps six 10 m Sentinel-2 multispectral images to 1 m locally demeaned LiDAR elevation residuals over landfast Arctic sea ice. The model is trained on two Arctic regions (Pond Inlet, Tuktoyaktuk) and tested on an unseen third region (Cambridge Bay). The authors report an out-of-domain RMSE of ~9 cm, σerror ~18%, JSD ~0.08, and claim that the model 'recovers physically meaningful surface structure from optical imagery alone.' However, the paper also reports out-of-domain ZNCC = 0.11 and nRMSEσ = 1.246 for the selected model, which raise serious concerns about whether the model achieves spatial reconstruction at all in the unseen region.

Significance. If the claims were substantiated, converting 10 m optical satellite imagery into 1 m sea ice roughness fields would be a novel and practically valuable capability for climate modeling, navigation, and operational mapping. The held-out test region design and the public code/data availability are strengths. However, the quantitative evidence reported in the paper undermines the central claim of spatial reconstruction: out-of-domain pixel-wise error is worse than a trivial zero predictor, and spatial correlation with ground truth is near zero. The paper's value may instead lie in distributionally realistic synthetic topography generation, but that framing is not what the title and abstract promise. The contribution as stated is therefore not yet established.

major comments (4)
  1. [Results, Table 2 / Evaluation Methods] The headline '9 cm RMSE' is not evidence of reconstruction because all metrics are computed in locally demeaned space. In that space, a constant-zero predictor yields RMSE equal to the patch standard deviation, i.e., nRMSEσ = 1.0. Table 2 reports nRMSEσ = 1.246 for the selected Cosine PLMS model on the unseen Cambridge Bay region, meaning its pixel-wise RMSE is ~25% worse than simply predicting a flat surface. The paper never reports this trivial baseline, and its statement that 'nRMSEσ values are close to unity' understates the problem: the model is not at the noise floor; it is worse than the noise floor. This is load-bearing for the claim of recovering surface structure and must be addressed with an explicit zero-prediction baseline and a discussion of what the RMSE figure actually demonstrates.
  2. [Results (paragraph on generalization) / Fig. 3] The reported test ZNCC of 0.1129 and log-PSD RMSE of 1.073 for Cosine PLMS indicate that predicted and true fields are essentially spatially uncorrelated in the unseen region, and high-frequency spectral power is substantially misallocated. The paper's own admission that the model 'over-smooths surfaces and misallocates high-frequency spectral power' contradicts the abstract's claim that it 'partially reproduces small-scale topographic structure.' To support the causal role of the optical conditioning, the authors must include control experiments: (a) an unconditional diffusion model with no Sentinel-2 input, and (b) a model with shuffled or corrupted conditioning input. Without such controls, the good JSD (0.08) and σerror (18.36%) could be produced by the diffusion prior alone, independent of the satellite imagery. This is essential for the claim of mapping 'from optical imagery alone.
  3. [Methods, Data Collection / Data Preprocessing] The ±14-day temporal co-location window between Sentinel-2 and LiDAR is a key assumption that is stated but not validated. The paper says this assumes 'landfast winter ice conditions remain stable over this window,' but provides no evidence for that stability or for a direct relationship between 10 m optical reflectance and 1 m elevation residuals. Given the near-zero out-of-domain ZNCC, the optical-to-microtopography link is questionable. The authors should quantify sensitivity to the temporal window and report a direct correlation analysis between Sentinel-2 features and residual elevation. Without this, the physical plausibility of the learned mapping is unsupported.
  4. [Discussion (comparison to prior work)] The comparison of the 9 cm RMSE to prior works reporting absolute elevation RMSE of 2.5–4 m (Refs. [11,12,18]) is misleading because the target here is a locally demeaned residual with much smaller amplitude. The paper acknowledges that 'direct comparison remains imperfect' but then claims 'sub-decimeter accuracy' and 'substantial improvement in spatial granularity.' The comparison should either be removed or reframed to avoid overstating the contribution, especially since the absolute elevation range is not predicted.
minor comments (5)
  1. [Introduction] Typo: 'UA Vs' should be 'UAVs'.
  2. [Tables 1 and 2] Column header 'σError (%)' should be 'σ error (%)'; several entries lack a space before the value (e.g., 'Cosine PLMS0.087' in Table 1).
  3. [Data Preprocessing] The train/validation split is described as 'along region-based spatial zones to prevent leakage from overlapping patches.' It is unclear whether validation patches share the same Sentinel-2 scenes as training patches; if they do, validation metrics may be optimistic because the conditioning imagery is not fully held out. Please clarify.
  4. [Results, Fig. 3] The statement that 'there is no evidence of the model attempting to hallucinate unseen large-scale structures' is based on visual inspection; this is not a quantitative test and should be softened or supported with a structural metric.
  5. [Abstract / Results] The abstract says 'best-performing model achieves an out-of-domain RMSE of 9 cm,' but Table 2 shows Linear PLMS achieves lower RMSE (0.083 m) than Cosine PLMS (0.089 m). Clarify that Cosine PLMS is selected on the basis of σerror and JSD, not RMSE.

Circularity Check

0 steps flagged

No significant circularity: supervised empirical learning with a genuinely held-out test region; self-citations are contextual, not load-bearing.

full rationale

RoughNet is an empirical supervised-learning pipeline, not a derivation from first principles. The claimed prediction (1 m locally demeaned LiDAR residual fields from 10 m Sentinel-2 patches) is trained on LiDAR and optical pairs from Pond Inlet and Tuktoyaktuk and evaluated on Cambridge Bay, a region withheld from training and model selection. The target variable is defined by the authors' preprocessing choices (RANSAC quadratic detrending, ±10 m residual cutoff, local demeaning), but this does not make the optical-to-topography mapping circular: the LiDAR residuals are measured independently of the Sentinel-2 inputs, and the out-of-domain evaluation is genuinely unseen. The paper does not fit a parameter to the test region and then report it as a prediction; the best model is selected on validation metrics and then applied to Cambridge Bay. The self-citations that appear (e.g., [4], [21], [26], [27]) support contextual statements about prior roughness mapping, super-resolution methods, or statistical properties of sea ice; they are not the load-bearing justification for the central reconstruction claim. The nRMSEσ values near unity and the low out-of-domain ZNCC (0.11) are substantive validity concerns—they suggest the pixel-wise reconstruction may be no better than a trivial flat prediction in the test region—but that is a correctness/benchmarking limitation, not circularity. The paper explicitly acknowledges that normalized errors are close to the intrinsic variability and that fine-scale structure is imperfectly reconstructed, so it does not hide this issue. No equation is shown to be equal to its inputs by construction, and no fitted parameter is renamed as an independent prediction. Therefore the circularity score is 0.

Axiom & Free-Parameter Ledger

9 free parameters · 5 axioms · 0 invented entities

The central result rests on supervised fitting; the only 'free parameters' that matter are the preprocessing definitions of the target and the model-selection choices. No new physics is postulated.

free parameters (9)
  • LiDAR residual outlier cutoff = ±10 m
    Points with residual magnitude exceeding ±10 m relative to RANSAC quadratic surface removed; hand-set threshold changes target roughness distribution and evaluation domain.
  • Quadratic surface fit for residualization = second-order polynomial via RANSAC + linear regression
    Local elevation residual is defined relative to this fit; choice absorbs/removes large-scale topography before training, so the model never sees absolute elevation.
  • Sentinel-2 normalization percentile clipping = 2nd–98th percentile, min range 1e-3
    Hand-chosen per-patch normalization of optical inputs; affects conditioning features.
  • Invalid-patch discard threshold = 30% invalid LiDAR pixels
    Patches with >30% invalid pixels excluded; influences training set and biases toward dense-coverage areas.
  • Temporal co-location window = ±14 days
    Sentinel-2 scenes within ±14 days of LiDAR campaigns; assumes stable landfast ice over this period (unvalidated).
  • Number of Sentinel-2 conditioning views = 6 scenes per region
    Fixed-length conditioning; real-world cloud cover/polar night will require variable-length handling, acknowledged.
  • Diffusion configuration = cosine schedule + PLMS sampler
    Selected from validation ablation by prioritizing σerror and JSD; not derived from first principles.
  • Patch extraction geometry = 256×256 px LiDAR patch, 128-px stride, 26×26 px Sentinel-2 crop resized to 256
    Balances conditioning context and training set size; receptive-field limitation can miss elongated ridges.
  • U-Net architecture and training loss weights = unspecified in text
    Channels, attention, gradient-regularization weight, timesteps (0-999) are hand-designed; not ablated exhaustively.
axioms (5)
  • domain assumption 10 m multispectral reflectance (Sentinel-2 bands B2/B3/B4/B8) carries recoverable information about 1 m surface-height residuals on winter landfast ice.
    No physical forward model links optical intensity/shadows to microtopography; the supervised fit is the only evidence. Central generalization depends on this.
  • domain assumption Landfast ice surface is stable between LiDAR and Sentinel-2 acquisitions (within ±14 days).
    Stated in Data Collection; if snow redistribution, ridging, or melt occurs in the window, label-conditioning pairs are misaligned.
  • domain assumption Airborne LiDAR-derived 1 m DEM residuals are a sufficiently accurate ground truth for roughness.
    Authors acknowledge point-density banding and artifacts in LiDAR (Supplementary Fig. S1); assumes these do not dominate roughness statistics and that predictions can be compared.
  • domain assumption Patch-level metrics averaged by region support regional-scale conclusions.
    Evaluation pools patch-level RMSE/JSD; mosaics assessed qualitatively; no geostatistical account of overlapping patches or autocorrelation in significance.
  • standard math Diffusion model training objective (MSE + gradient regularization) with x0 parameterization yields physically meaningful elevation samples.
    Standard diffusion theory from [30,31] assumed; not re-derived.

pith-pipeline@v1.3.0-alltime-deepseek · 11465 in / 15819 out tokens · 153726 ms · 2026-08-02T05:20:56.714723+00:00 · methodology

0 comments
read the original abstract

Accurate estimation of landfast sea ice roughness is critical for climate modeling and safe Arctic over-ice travel, yet existing approaches rely on costly airborne surveys or sparse in-situ measurements, limiting spatial coverage and operational scalability. Here we show that high-resolution sea ice topography can be reconstructed directly from optical satellite imagery using a conditional diffusion framework. Our approach, RoughNet, learns to map 10 m Sentinel-2 multispectral images to locally normalized 1 m surface elevation residual fields, enabling fine-scale roughness characterization from widely available satellite data. Trained on airborne LiDAR data from two Arctic regions and evaluated on an unseen third Arctic region, the model generalizes across diverse ice conditions and partially reproduces small-scale topographic structure. The best-performing model achieves an out-of-domain root mean squared error of 9 cm while preserving the statistical and spectral properties of the underlying roughness field. These results demonstrate that generative diffusion models can recover physically meaningful surface structure from optical imagery alone, providing a scalable pathway for high-resolution sea ice mapping and roughness estimation in data-sparse environments.

Figures

Figures reproduced from arXiv: 2607.13371 by Christian Haas, Michel Tsamados, Petru Manescu, Randall Scharien, Tessa Cannon, Thomas Newman, Veit Helm, Weibin Chen.

Figure 1
Figure 1. Figure 1: Geographic locations of data collection regions in Northern Canada: Tuktoyaktuk, Cam [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic of the conditional diffusion framework, illustrating the mapping from multi [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Patch-wise reconstructions from the conditional diffusion model across validation regions [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Regional reconstruction for Tuktoyaktuk. Patches are represented in locally demeaned [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Morphological and statistical patch analysis for a single Tuktoyaktuk patch highlighted [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Example patch set from the Pond Inlet region used as training input into the conditional [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

35 extracted references · 12 linked inside Pith

  1. [1]

    Nolin and Eugene Mar

    Anne W. Nolin and Eugene Mar. Arctic Sea Ice Surface Roughness Estimated from Multi- Angular Reflectance Satellite Imagery.Remote Sensing, 11(1):50, 2019

    2019

  2. [2]

    Segal, Randall K

    Rebecca A. Segal, Randall K. Scharien, Silvie Cafarella, and Andrew Tedstone. Characterizing winter landfast sea-ice surface roughness in the Canadian Arctic Archipelago using Sentinel-1 synthetic aperture radar and the Multi-angle Imaging SpectroRadiometer.Annals of Glaciol- ogy, 61(83):284–298, 2020. 15

    2020

  3. [3]

    S. L. Farrell, K. Duncan, E. M. Buckley, J. Richter-Menge, and R. Li. Mapping Sea Ice Surface Topography in High Fidelity With ICESat-2.Geophysical Research Letters, 47(21):e2020GL090708, 2020

    2020

  4. [4]

    Mapping Arctic Sea-Ice Surface Roughness with Multi-Angle Imaging SpectroRadiometer.Remote Sensing, 14(24):6249, 2022

    Thomas Johnson, Michel Tsamados, Jan-Peter Muller, and Julienne Stroeve. Mapping Arctic Sea-Ice Surface Roughness with Multi-Angle Imaging SpectroRadiometer.Remote Sensing, 14(24):6249, 2022

    2022

  5. [5]

    Macdonald, Randall K

    Grant J. Macdonald, Randall K. Scharien, Kyle Duncan, Sinead L. Farrell, Parnian Rezania, and Aikaterini Tavri. Arctic Sea Ice Topography Information From RADARSAT Constellation Mission (RCM) Synthetic Aperture Radar (SAR) Backscatter.Geophysical Research Letters, 51(4):e2023GL107261, 2024

    2024

  6. [6]

    ResDepth: A deep residual prior for 3D reconstruction from high-resolution satellite images.ISPRS Journal of Photogrammetry and Remote Sensing, 183:560–580, 2022

    Corinne Stucker and Konrad Schindler. ResDepth: A deep residual prior for 3D reconstruction from high-resolution satellite images.ISPRS Journal of Photogrammetry and Remote Sensing, 183:560–580, 2022

    2022

  7. [7]

    Y. Cai, Z. Yao, S. Lang, X. Cui, F. Wan, and Y. He. Generative Elevation Inpainting: An Efficient Completion Method for Generating High-Resolution Antarctic Bed Topography. IEEE Transactions on Geoscience and Remote Sensing, 61:1–17, 2023

    2023

  8. [8]

    High Accuracy Interpolation of DEM Using Generative Adversarial Network.Remote Sensing, 13(4):676, 2021

    Li Yan, Xingfen Tang, and Yi Zhang. High Accuracy Interpolation of DEM Using Generative Adversarial Network.Remote Sensing, 13(4):676, 2021

    2021

  9. [9]

    DeepBedMap: a deep neural network for resolving the bed topography of Antarctica.The Cryosphere, 14(11):3687–3705, 2020

    Wei Ji Leong and Huw Joseph Horgan. DeepBedMap: a deep neural network for resolving the bed topography of Antarctica.The Cryosphere, 14(11):3687–3705, 2020

    2020

  10. [10]

    Demiray, Muhammed Sit, and Ibrahim Demir

    Bekir Z. Demiray, Muhammed Sit, and Ibrahim Demir. D-SRGAN: DEM Super-Resolution with Generative Adversarial Networks, 2020. arXiv:2004.04788 [cs]

    Pith/arXiv arXiv 2020

  11. [11]

    IMG2DSM: Height Simulation From Single Imagery Using Conditional Generative Adversarial Net.IEEE Geoscience and Remote Sensing Letters, 15(5):794–798, 2018

    Pedram Ghamisi and Naoto Yokoya. IMG2DSM: Height Simulation From Single Imagery Using Conditional Generative Adversarial Net.IEEE Geoscience and Remote Sensing Letters, 15(5):794–798, 2018

    2018

  12. [12]

    Generating Elevation Surface from a Single RGB Remotely Sensed Image Using Deep Learning

    Emmanouil Panagiotou, Georgios Chochlakis, Lazaros Grammatikopoulos, and Eleni Charou. Generating Elevation Surface from a Single RGB Remotely Sensed Image Using Deep Learning. Remote Sensing, 12(12):2002, 2020

    2002

  13. [13]

    Repurposing Diffusion-Based Image Generators for Monocular Depth Estima- tion

    Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Metzger, Rodrigo Caye Daudt, and Kon- rad Schindler. Repurposing Diffusion-Based Image Generators for Monocular Depth Estima- tion. In2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9492–9502, 2024

    2024

  14. [14]

    DiffusionDepth: Diffusion Denoising Approach for Monocular Depth Estimation, 2023

    Yiqun Duan, Xianda Guo, and Zheng Zhu. DiffusionDepth: Diffusion Denoising Approach for Monocular Depth Estimation, 2023. arXiv:2303.05021 [cs]

    Pith/arXiv arXiv 2023

  15. [15]

    ECoDepth: Effective Conditioning of Diffusion Models for Monocular Depth Estimation, 2024

    Suraj Patni, Aradhye Agarwal, and Chetan Arora. ECoDepth: Effective Conditioning of Diffusion Models for Monocular Depth Estimation, 2024. arXiv:2403.18807 [cs]

    Pith/arXiv arXiv 2024

  16. [16]

    Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data, 2024

    Lihe Yang, Bingyi Kang, Zilong Huang, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data, 2024. arXiv:2401.10891 [cs]. 16

    Pith/arXiv arXiv 2024

  17. [17]

    Arce, and Mark Stephen

    Andres Ramirez-Jaime, Nestor Porras-Diaz, Gonzalo R. Arce, and Mark Stephen. Super- Resolved 3-D Satellite Lidar Imaging of Earth via Generative Diffusion Models.IEEE Trans- actions on Geoscience and Remote Sensing, 63:1–19, 2025

    2025

  18. [18]

    Simultaneous Super-resolution and Depth Estimation for Satellite Images Based on Diffusion Model

    Yuwei Zhou and Yangming Lee. Simultaneous Super-resolution and Depth Estimation for Satellite Images Based on Diffusion Model. In2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 411–418, 2024

    2024

  19. [19]

    SAR deep learning sea ice retrieval trained with airborne laser scanner measurements from the MOSAiC expedition.The Cryosphere, 18(5):2207–2222, May 2024

    Karl Kortum, Suman Singha, Gunnar Spreen, Nils Hutter, Arttu Jutila, and Christian Haas. SAR deep learning sea ice retrieval trained with airborne laser scanner measurements from the MOSAiC expedition.The Cryosphere, 18(5):2207–2222, May 2024

    2024

  20. [20]

    W. Hong, Z. Huang, A. Wang, Y. Liu, J. Cai, and H. Su. SeaIceNet: Sea Ice Recognition via Global-Local Transformer in Optical Remote Sensing Images.IEEE Transactions on Geo- science and Remote Sensing, 62:1–16, 2024

    2024

  21. [21]

    ARISGAN: Extreme super- resolution of arctic surface imagery using generative adversarial networks.Frontiers in Remote Sensing, 5, August 2024

    Christian Au, Michel Tsamados, Petru Manescu, and So Takao. ARISGAN: Extreme super- resolution of arctic surface imagery using generative adversarial networks.Frontiers in Remote Sensing, 5, August 2024

    2024

  22. [22]

    Farrugia, Gianluca Valentino, Mang Chen, Johann A

    Tomasz Tarasiewicz, Jakub Nalepa, Reuben A. Farrugia, Gianluca Valentino, Mang Chen, Johann A. Briffa, and Michal Kawulok. Multitemporal and multispectral data fusion for super-resolution of Sentinel-2 images.IEEE Transactions on Geoscience and Remote Sensing, 61:1–19, 2023. arXiv:2301.11154 [cs]

    Pith/arXiv arXiv 2023

  23. [23]

    Super Resolution Reconstruction Technique in Passive Microwave Images of Arctic Sea Ice

    Xiaomin Liu, Tiantian Feng, Junqiao Zhao, and Rongxing Li. Super Resolution Reconstruction Technique in Passive Microwave Images of Arctic Sea Ice. InIGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium, pages 4234–4237, July 2019

    2019

  24. [24]

    Super-resolution restoration of MISR images using the UCL MAGiGAN system.Remote Sensing, 11(1):52, 2019

    Yu Tao and Jan-Peter Muller. Super-resolution restoration of MISR images using the UCL MAGiGAN system.Remote Sensing, 11(1):52, 2019

    2019

  25. [25]

    Improved Denoising Diffusion Probabilistic Models, 2021

    Alex Nichol and Prafulla Dhariwal. Improved Denoising Diffusion Probabilistic Models, 2021. arXiv:2102.09672 [cs]

    Pith/arXiv arXiv 2021

  26. [26]

    Landy, Michel Tsamados, and Randall K

    Jack C. Landy, Michel Tsamados, and Randall K. Scharien. A Facet-Based Numerical Model for Simulating SAR Altimeter Echoes From Heterogeneous Sea Ice Surfaces.IEEE Transac- tions on Geoscience and Remote Sensing, 57(7):4164–4180, July 2019

    2019

  27. [27]

    Sub-kilometer Scale Snow Depth Distribution on Sea Ice of Different Ages and Thickness.EGUsphere, November 2025

    Lanqing Huang, Julienne Stroeve, Thomas Newman, Robbie Mallett, Rosemary Willatt, Lu Zhou, Malin Johansson, Carmen Nab, and Alicia Fallows. Sub-kilometer Scale Snow Depth Distribution on Sea Ice of Different Ages and Thickness.EGUsphere, November 2025. preprint

    2025

  28. [28]

    C. Haas, R. Kelly, J. Welch, A. Jutila, N. Hutter, and M. Gehrmann. Icebird can24 campaign report. Internal report, Alfred Wegener Institute, 2024

    2024

  29. [29]

    Iceberg influence on snow distribution and slush forma- tion on Antarctic landfast sea ice from airborne multi-sensor observations.The Cryosphere, 19(11):6319–6339, November 2025

    Steven Franke, Mara Neudert, Veit Helm, Arttu Jutila, Oc´ eane Hames, Niklas Neckel, Ste- fanie Arndt, and Christian Haas. Iceberg influence on snow distribution and slush forma- tion on Antarctic landfast sea ice from airborne multi-sensor observations.The Cryosphere, 19(11):6319–6339, November 2025

    2025

  30. [30]

    Weiss, Niru Maheswaranathan, and Surya Ganguli

    Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep Un- supervised Learning using Nonequilibrium Thermodynamics, 2015. arXiv:1503.03585 [cs]. 17

    Pith/arXiv arXiv 2015

  31. [31]

    Denoising Diffusion Probabilistic Models, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising Diffusion Probabilistic Models, 2020. arXiv:2006.11239 [cs]

    Pith/arXiv arXiv 2020

  32. [32]

    Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction, 2025

    Jing He, Haodong Li, Wei Yin, Yixun Liang, Leheng Li, Kaiqiang Zhou, Hongbo Zhang, Bingbing Liu, and Ying-Cong Chen. Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction, 2025. arXiv:2409.18124 [cs]

    Pith/arXiv arXiv 2025

  33. [33]

    Denoising Diffusion Implicit Models, 2022

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising Diffusion Implicit Models, 2022. arXiv:2010.02502 [cs]

    Pith/arXiv arXiv 2022

  34. [34]

    Pseudo Numerical Methods for Diffusion Models on Manifolds, 2022

    Luping Liu, Yi Ren, Zhijie Lin, and Zhou Zhao. Pseudo Numerical Methods for Diffusion Models on Manifolds, 2022. arXiv:2202.09778 [cs]

    Pith/arXiv arXiv 2022

  35. [35]

    U-Net: Convolutional Networks for Biomedical Image Segmentation, 2015

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-Net: Convolutional Networks for Biomedical Image Segmentation, 2015. arXiv:1505.04597 [cs]. 18 Supplementary Figures Figure S1: Regional reconstruction for Pond Inlet. Patches are represented in locally demeaned space and mosaicked into a regional representation. 19 Figure S2: Regional reconstruction f...

    Pith/arXiv arXiv 2015