Pith. sign in

REVIEW 3 major objections 7 minor 44 references

Physics-Informed Feature Engineering 1D-CNN for Multilayer Cloud Detection from Geostationary Satellites

T0 review · 3 major / 7 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read A 1D-CNN guided by threshold-derived channel priors detects multilayer clouds with POD 0.620 and FAR 0.240, outperforming the conventional threshold algorithm.

desk verdict The paper's headline claim attributes the best CNN numbers to the physics-prior model, but the paper's own Table IV shows that model underperforms the threshold baseline on POD; the real gains come from gradient-selected channels. read the letter →

arxiv 2607.16270 v1 pith:M7VF36AZ submitted 2026-07-07 eess.SP cs.AIphysics.ao-ph

Fu Wang , Chi Yang , Qi-Feng Lu , Rui-Xia Liu , Xiao-Fei Yang , Xiao-Fang Liu , Bo Li , Lin Chen This is my paper · ORCID
classification eess.SPcs.AIphysics.ao-ph
keywords multilayerclouddetection1D-CNNphysics-informedfeatureengineeringthresholdalgorithmAGRIAHICALIPSOSat
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

The paper tries to establish that embedding physical priors—specifically, channel selections taken from a threshold-based algorithm—into a compact 1D-CNN improves multilayer cloud detection from geostationary imagery over the conventional threshold method. It reports that the CNN achieves a multilayer probability of detection of 0.620 and a false alarm rate of 0.240, whereas the threshold baseline yields 0.558 and 0.369. It further claims that gradient-based attribution identifies the 12.0 µm channel (AGRI C13) as the most informative, and that substituting it for the 10.8 µm channel (C12) improves the threshold algorithm itself, raising POD to 0.609. The paper also shows that the same substitution on a different sensor (AHI) yields negligible gains, which it attributes to spectral response function mismatches and on-orbit radiometric stability. A sympathetic reader cares because this demonstrates a two-way synergy between physical knowledge and machine learning that could lead to more accurate, operationally feasible cloud retrievals for numerical weather prediction.

What carries the argument

The central object is a one-dimensional convolutional neural network (1D-CNN) with three convolutional layers, two fully connected layers, and a softmax output, processing spectral channels as inputs. The key mechanism is feature routing: in the CNN-AuxConv variant, the core channels (those used by the threshold algorithm) bypass the convolutional layers and feed directly into the fully connected branch, while auxiliary channels pass through convolution. This design preserves the discriminative signals in the physics-selected channels and extracts higher-dimensional cues from auxiliary bands. The threshold algorithm's reflectance difference ratio (RDR) between 1.375 µm and 1.61 µm is the nam

What would settle it

Take the matched dataset, keep only pixels where CALIOP and CPR agree on the number of cloud layers and where the active footprint fully covers the imager pixel, and re-run the CNN and threshold comparisons; if the POD/FAR gap collapses to within statistical noise, the collocation labels are driving the result.

Watch

Extended reading notes

Core claim

The paper claims that a physics-informed feature-engineering prior—selecting channels according to the two-stage threshold algorithm's radiative-transfer logic—lets a compact 1D-CNN match or exceed the threshold method's detection rate while reducing false alarms. Using the full SWIR-to-TIR channel suite, the CNN reaches POD_mul = 0.620 and FAR_mul = 0.251; a gradient-selected five-channel version reaches POD_mul = 0.620 and FAR_mul = 0.240. Feature-gradient analysis ranks AGRI C13 (12.0 µm) as the most influential channel, and replacing C12 (10.8 µm) with C13 in the threshold algorithm raises POD_mul from 0.558 to 0.609 without materially changing FAR. For AHI, the analogous substitution (1

Load-bearing premise

The CPR-CALIOP joint product provides correct multilayer labels for each imager pixel after nearest-neighbor matching within ±5 minutes, so if those labels are wrong or misaligned in a way that correlates with the channels used, the reported gains could be artifacts.

Editorial extensions

If this is right

  • If the results hold, operational geostationary multilayer cloud detection can be improved by using machine-learning channel attribution to choose channels, rather than simply adding more spectral bands.
  • The 12.0 µm channel (C13) would be preferred over 10.8 µm (C12) for AGRI-based threshold algorithms, a direct substitution that could be adopted with minimal changes to existing operational workflows.
  • Sensor-specific spectral response functions and on-orbit radiometric stability must be accounted for when transferring channel-based algorithms between satellites; a channel that helps on AGRI may not help on AHI.
  • The compact CNN (~116k parameters, ~10 MB) combined with a physics-informed prior appears operationally feasible, with inference times in the range of a few minutes per full-disk dataset.
  • The gradient-based importance ranking identifies a manageable subset of five channels that performs as well as the full ten-channel CNN, suggesting that a simplified operational input set may suffice.

Reading between the lines

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

  • The 'physics-informed' prior is only as good as the threshold algorithm's channel choices; using a prior derived from radiative transfer simulations rather than an existing algorithm might yield larger improvements than the paper reports.
  • The gradient-based importance ranking (C13 first, followed by C1, C12, C11, C4) is likely sensitive to the training data's seasonal and geographic sampling; testing on other seasons or regions could reveal whether the 12.0 µm dominance is universal or specific to the October–December daytime ocean scenes.
  • The CNN's lower false-alarm rate may stem from its ability to learn nonlinear combinations of brightness temperature differences that the threshold algorithm approximates with linear rules; if so, simpler nonlinear decision-tree models might achieve similar gains without deep learning.
  • The cross-sensor finding (AGRI C13 helps but AHI C15 does not) suggests that a similar channel-substitution study on GOES-16/17 ABI, which has different SRF shapes, would need to be done empirically rather than by analogy.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper proposes using channel selections derived from a threshold-based multilayer cloud detection algorithm as feature-engineering priors for a 1D convolutional neural network (CNN), with the goal of improving multilayer cloud detection from geostationary imagers (FY-4A/AGRI and Himawari-8/AHI). Evaluation is against the CPR-CALIOP 2B-CLDCLASS-LIDAR reference. The abstract and conclusion claim the CNN achieves POD_mul=0.620, FAR_mul=0.240, outperforming the threshold baseline (POD=0.558, FAR=0.369), and that replacing AGRI C12 (10.8 µm) with C13 (12.0 µm) improves the threshold algorithm's POD to 0.609. Additional experiments examine channel importance via gradient attribution, input-channel ablation, phase-conditioned diagnostics, and cross-sensor transfer.

Significance. If the headline results were correct, the paper would demonstrate a useful bidirectional synergy between physics-based threshold algorithms and deep learning for a difficult operational problem. The AGRI C12→C13 substitution result is concrete and testable, and the use of an active-sensor reference (CPR-CALIOP) is appropriate. The phase-conditioned metrics (MwI, MwW) are a useful diagnostic. However, the manuscript as written conflates the models behind the headline numbers and misreports the AHI comparison, so the significance of the claimed contributions cannot be assessed without substantial revision. The paper does not provide code or a fully specified train/test protocol, but it does provide detailed model configurations and a clear experimental structure.

major comments (3)
  1. [Abstract; Conclusion; Tables IV and V] The headline POD=0.620, FAR=0.240 is attributed to "channel selections derived from threshold-based algorithms" (Abstract) and to the framework that "embed[s] threshold-derived channel selections as physical priors" (Conclusion). However, the model that actually uses the threshold algorithm's channel set, CNN-PhysCore (Table IV), achieves POD_mul=0.549, FAR_mul=0.301—a POD lower than the threshold baseline's 0.558. The stated 0.620/0.240 values correspond to CNN-Gradient-Feat (Table V), whose inputs (R0.47, R1.375, BT8.5, BT10.8, BT12.0) are selected by gradient attribution, a purely data-driven procedure, or to CNN-SWIR2TIR (Table IV) using all ten channels. The central claim that physics-informed feature engineering outperforms the threshold algorithm with the reported margin is therefore not supported by the paper's own tables. The abstract and conclusion must be rewritten to attribut
  2. [Section III-C and Table VI] The abstract states that for AHI, substituting the 11.2 µm channel with the 12.3 µm channel yielded "negligible improvement." Table VI shows the opposite: AHI-SubC14 (11.2 µm) has POD_mul=0.621, FAR_mul=0.286, while AHI-SubC15 (12.3 µm) has POD_mul=0.600, FAR_mul=0.349. The 12.3 µm substitution degrades both POD and FAR. Furthermore, the row labels in Table VI are internally inconsistent with the definitions in Section II-C-3: AHI-SubC14 is defined as mapping AGRI 10.8 µm→AHI 11.2 µm, but the table row labeled AHI-SubC14 lists BT12.3. These errors invalidate the stated conclusion about sensor-specific effects and need correction.
  3. [Section II-C-3 and Section III-A] No train/validation/test split is described. The CNN models appear to be trained and evaluated on the same CAG and CAA datasets, with no held-out test set or uncertainty quantification. Without specifying which samples were used for training, validation, and testing (e.g., temporal or track-based split), and without confidence intervals or significance tests for the reported POD/FAR differences, the improvements attributed to the CNN (e.g., FAR reduction from 0.369 to 0.301) cannot be distinguished from overfitting or random initialization. The authors should report the data partition, training protocol, and ideally bootstrap confidence intervals for the key metrics.
minor comments (7)
  1. [Eqs. (8)-(11)] Equations (8)-(9) are repeated as Eqs. (10)-(11) in Section III-C with the same definitions. Use a single numbered set.
  2. [Table VI] The rows for AHI-SubC14 and AHI-SubC15 appear swapped relative to the textual definitions. Please verify the channel assignments and correct the table.
  3. [Section III-A] Typo: "CNN-SWIR2TTR" should be "CNN-SWIR2TIR" in the paragraph describing full-channel results.
  4. [Section II-C-2] The inference time of "about 242 seconds" should specify the number of samples and hardware; otherwise it is not interpretable.
  5. [Section III-B/Table V] The CNN-night model is labeled "nighttime" but is evaluated on daytime data. Clarify that it simply excludes solar channels, or provide genuine nighttime evaluation if the label is to be retained.
  6. [Section III-C] The claim that on-orbit radiometric stability is a primary factor for the AHI result is not directly evidenced in this paper; it relies on ref. [34]. Either add quantitative stability metrics for the specific channels or soften this causal claim.
  7. [Fig. 6] The gradient importance measure is not defined. State how the attribution is computed (e.g., mean absolute gradient of the softmax output with respect to input channels) and on which data segment.

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation-level circularity; the abstract/conclusion attribution of POD=0.620/FAR=0.240 to the threshold-prior CNN is not supported by Table IV (CNN-PhysCore POD=0.549), but that is an internal-attribution/correctness problem, not a reduction of the result to its inputs.

full rationale

The paper's claimed chain is empirical rather than derivational. The labels come from the external CPR-CALIOP joint product (2B-CLDCLASS-LIDAR), not from the CNN or from the threshold algorithm's outputs. The threshold baseline is re-specified in Section II-C1 rather than treated as a black box, so the self-citation [16] is not load-bearing. CNN-PhysCore is a controlled comparison using exactly the baseline's channel set, and the C12-to-C13 substitution is a concrete algorithmic change whose outcomes are not predetermined: the AHI substitutions indeed behave differently. The only other self-citation, [34], is used as a post hoc explanation of on-orbit radiometric stability, not as the argument's foundation. The main defect is that the conclusion assigns the gradient-selected CNN-Gradient-Feat numbers (POD=0.620, FAR=0.240) to the method defined by threshold-derived channel selections, whereas Table IV reports CNN-PhysCore at POD=0.549, below Baseline-AGRI (0.558). That is a missing-support/attribution inconsistency and an in-sample selection-bias risk, but it does not make any equation reduce to itself or turn a fitted parameter into a prediction under the quote-and-reduction standard. Therefore no circular step can be established; the score reflects only the minor, non-load-bearing self-citations.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper does not introduce new physical entities. Its claims rest on benchmark labels, collocation assumptions, and threshold/model design choices; the most fragile are the reference-label fidelity and the undisclosed threshold-tuning procedure.

free parameters (2)
  • AGRI-SubC13 decision thresholds = BT12.0 = 250 K; BTD8.5−12.0 = 0.7 K
    Used for the claimed improved threshold algorithm; the paper does not state whether these were tuned on the CPR-CALIOP evaluation data, which would inflate the reported improvement.
  • 1D-CNN architecture/training hyperparameters
    Number of conv layers, kernel size, FC sizes, batch norm, and Adam settings were chosen via ablation on the same data; no independent validation set or full hyperparameter search details are given.
assumptions (5)
  • domain assumption 2B-CLDCLASS-LIDAR phase/layer labels are an accurate ground truth for the 'multilayer' class.
    All training labels and evaluation metrics come from this active-sensor product; no independent validation of the reference itself is provided.
  • domain assumption Nearest-neighbor collocation within ±5 minutes and mean/mode aggregation preserves correct pixel-level labels.
    The active-sensor footprint is a narrow curtain; a 4-km AGRI pixel may contain a mix of single- and multilayer profiles that aggregation cannot faithfully represent.
  • domain assumption Threshold parameters inherited from the prior AGRI algorithm (ref 16) transfer to the current dataset.
    The baseline threshold values (e.g., BT10.8 = 255 K, BTD8.5−10.8 = −0.5 K) are taken from a previous study without re-estimation on this matched dataset.
  • ad hoc to paper The new SubC13 thresholds (BT12.0 = 250 K, BTD8.5−12.0 = 0.7 K) are appropriate and were not selected on the evaluation data.
    The selection procedure for these thresholds is not described; if they were tuned to maximize performance on the same CPR-CALIOP data, the reported improvement is optimistic.
  • domain assumption Daytime ocean scenes with zenith angle <60° are sufficient to support operational-transfer conclusions for full-disk sensors.
    All results are restricted to daytime ocean pixels in October–December 2018; generalization to land, night, snow, and other seasons is not demonstrated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Physics-Informed Feature Engineering 1D-CNN for Multilayer Cloud Detection from Geostationary Satellites." pith.science (2026). https://pith.science/paper/M7VF36AZ

@misc{pith2026260716270,
  author       = {Pith},
  title        = {Pith review of: Physics-Informed Feature Engineering 1D-CNN for Multilayer Cloud Detection from Geostationary Satellites},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M7VF36AZ}},
  note         = {Machine review of arXiv:2607.16270}
}
abstract

Multilayer cloud detection from active--passive observation is vital for numerical weather prediction. In this study, channel selections derived from threshold-based algorithms are embedded as feature-engineering priors into a 1D-CNN, and machine learning (ML) is used to learn latent physical relationships to simplify physical retrievals for operational deployment. The results show that the 1D-CNN achieves a multilayer-cloud probability of detection ($\mathrm{POD}{\mathrm{mul}}$) of 0.620 and a false alarm rate ($\mathrm{FAR}{\mathrm{mul}}$) of 0.240, outperforming the conventional threshold algorithm ($\mathrm{POD}{\mathrm{mul}} = 0.558$, $\mathrm{FAR}{\mathrm{mul}} = 0.369$). These results demonstrate that prior physical knowledge derived from radiative transfer theory can serve as an effective feature-engineering prior. Further experiments show that ML-revealed physical mechanisms can also enhance traditional algorithms. Replacing AGRI channel 12 (C12, centered at $10.8~\mu\mathrm{m}$) with channel 13 (C13, centered at $12.0~\mu\mathrm{m}$) increased $\mathrm{POD}{\mathrm{mul}}$ from 0.558 to 0.609 without materially affecting $\mathrm{FAR}{\mathrm{mul}}$. However, for AHI, substituting the $11.2~\mu\mathrm{m}$ channel with the $12.3~\mu\mathrm{m}$ channel yielded negligible improvement. In addition to spectral response function (SRF) mismatches, a primary contributing factor is the channels' on-orbit radiometric stability. Hence, physics-informed machine-learning methods appear promising for advancing remote-sensing AI, while sensor-specific characteristics must be considered during operational transfer.

Figures

Figures reproduced from arXiv: 2607.16270 by the authors.

Figure 1
Figure 1. Spectral response functions (SRFs) of FY-4A/AGRI (orange) and Himawari-8/AHI (blue )channels. (a) Shortwave infrared (SWIR) channels. (b) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Flowchart of the multilayer cloud detection algorithm for FY [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. (a) The flowchart for generating the dataset. (b) Model configuration of the deep neural network (CNN). [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Variation of probability of detection (POD) and false alarm rate [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Scatter comparison of evaluation metrics for multilayer versus single-layer cloud detection algorithms based on the 2018 CPR–CALIOP merged cloud [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Feature gradient (mean absolute attribution) for all channels in the [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: The percentages of single-layer and multilayer clouds for products of 1D-CNN models with different parameters of the CPR-CALIOP reference labels. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Comparison of POD and FAR for threshold-based algorithms against the CPR-CALIOP merged cloud product. (a,b) POD and FAR for the baseline [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Joint distributions of brightness temperature (BT) and brightness-temperature difference (BTD) used by the two-stage threshold scheme. Each panel [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Comparison of the threshold-based algorithms’ top-layer cloud phase identification performance against the CPR-CALIOP merged cloud product. [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Threshold-based detection of top-layer cloud phase in multilayer clouds using the [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: Percentages of single-layer and multilayer clouds detected by threshold-based algorithms with varied parameter configurations against CPR–CALIOP [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 13
Figure 13. Figure 13: Case at 0700 UTC on 29 October 2018 over the East Asian continent. (a) True-color composite from Himawari-8 AHI channels (0.65 [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 1 canonical work pages

  1. [1]

    A determination of the cloud feedback from climate variations over the past decade,

    A. E. Dessler, “A determination of the cloud feedback from climate variations over the past decade,”Science, vol. 330, no. 6010, pp. 1523– 1527, 2010

    2010

  2. [2]

    Earth’s global energy budget,

    K. E. Trenberth, J. T. Fasullo, and J. Kiehl, “Earth’s global energy budget,”Bulletin of the american meteorological society, vol. 90, no. 3, pp. 311–324, 2009

    2009

  3. [3]

    Global distributions of multi-layer and multi- phase clouds and their cloud radiative effects,

    B. Yi, R. Li, and Z. Yang, “Global distributions of multi-layer and multi- phase clouds and their cloud radiative effects,”Journal of Quantitative Spectroscopy and Radiative Transfer, vol. 333, p. 109334, 2025

    2025

  4. [4]

    Diurnal cloud cycle biases in climate models,

    J. Yin and A. Porporato, “Diurnal cloud cycle biases in climate models,” Nature Communications, vol. 8, no. 1, p. 2269, 2017, publisher: Nature Publishing Group

    2017

  5. [5]

    Building unified global 3d cloud data from multiple satellites for advancing weather and climate research,

    Y . J. Noh, J. M. Haynes, S. D. Miller, C. White, B. Daub, I. Ebert-Uphoff, H. Yu, J. Apke, R. Chase, K. Haynes, J. Solbrig, E. Rose, M. King, and L. Cheatwood, “Building unified global 3d cloud data from multiple satellites for advancing weather and climate research,” inIGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, 2024, ...

    2024

  6. [6]

    G. G. Mace, Q. Zhang, M. Vaughan, R. Marchand, G. Stephens, C. Trepte, and D. Winker, “A description of hydrometeor layer oc- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12 Fig. 11. Threshold-based detection of top-layer cloud phase in multilayer clouds using the10.8µm baseline and12.0µm SubC13 threshold algorithms (T/F: correct/incorrect). ...

    2021

  7. [7]

    Detection of multi-layer and vertically-extended clouds using a-train sensors,

    J. Joiner, A. Vasilkov, P. Bhartia, G. Wind, S. Platnick, and W. Menzel, “Detection of multi-layer and vertically-extended clouds using a-train sensors,”Atmospheric Measurement Techniques, vol. 3, no. 1, pp. 233– 247, 2010

    2010

  8. [8]

    Discriminating clear sky from clouds with modis,

    S. A. Ackerman, K. I. Strabala, W. P. Menzel, R. A. Frey, C. C. Moeller, and L. E. Gumley, “Discriminating clear sky from clouds with modis,” Journal of Geophysical Research: Atmospheres, vol. 103, no. D24, pp. 32 141–32 157, 1998

    1998

Show all 44 references
  1. [9]

    The modis cloud products: Algorithms and examples from terra,

    S. Platnick, M. D. King, S. A. Ackerman, W. P. Menzel, B. A. Baum, J. C. Ri ´edi, and R. A. Frey, “The modis cloud products: Algorithms and examples from terra,”IEEE Transactions on geoscience and Remote Sensing, vol. 41, no. 2, pp. 459–473, 2003

    2003

  2. [10]

    Validation of MODIS cloud mask and multilayer flag using CloudSat-CALIPSO cloud profiles and a cross-reference of their cloud classifications,

    T. Wang, E. J. Fetzer, S. Wong, B. H. Kahn, and Q. Yue, “Validation of MODIS cloud mask and multilayer flag using CloudSat-CALIPSO cloud profiles and a cross-reference of their cloud classifications,”Journal of Geophysical Research: Atmospheres, vol. 121, no. 19, pp. 11,620 – ...

    2016

  3. [11]

    A multilayer cloud detection algorithm for the suomi-npp visible infrared imager radiometer suite (viirs),

    J. Wang, C. Liu, B. Yao, M. Min, H. Letu, Y . Yin, and Y . L. Yung, “A multilayer cloud detection algorithm for the suomi-npp visible infrared imager radiometer suite (viirs),”Remote Sensing of Environment, vol. 227, pp. 1–11, 2019

    2019

  4. [12]

    A Global Multilayer Cloud Identification with POLDER/PARASOL,

    M. Desmons, N. Ferlay, F. Parol, J. Riedi, and F. Thieuleux, “A Global Multilayer Cloud Identification with POLDER/PARASOL,” Journal of Applied Meteorology and Climatology, vol. 56, pp. 1121– 1139, 2017. [Online]. Available: https://www.semanticscholar.org/paper/ cd388ad56952a...

    2017

  5. [13]

    The modis cloud optical and microphysical products: Collection 6 updates and examples from terra and aqua,

    S. Platnick, K. G. Meyer, M. D. King, G. Wind, N. Amarasinghe, B. Marchant, G. T. Arnold, Z. Zhang, P. A. Hubanks, R. E. Holz et al., “The modis cloud optical and microphysical products: Collection 6 updates and examples from terra and aqua,”IEEE Transactions on Geoscience and...

    2016

  6. [14]

    Daytime cloud overlap detection from avhrr and viirs,

    M. J. Pavolonis and A. K. Heidinger, “Daytime cloud overlap detection from avhrr and viirs,”Journal of Applied Meteorology, vol. 43, no. 5, pp. 762–778, 2004

    2004

  7. [15]

    Multilayer Cloud Detection with the MODIS Near-Infrared Water Vapor Absorption Band,

    G. Wind, S. Platnick, M. D. King, P. A. Hubanks, M. J. Pavolonis, A. K. Heidinger, P. Yang, and B. A. Baum, “Multilayer Cloud Detection with the MODIS Near-Infrared Water Vapor Absorption Band,”Journal of Applied Meteorology and Climatology, vol. 49, no. 11, pp. 2315–2333, Nov...

    2010

  8. [16]

    Algorithm for detecting ice overlaying water multilayer clouds using the infrared bands of FY-4a/AGRI,

    K. Wang, F. Wang, Q. Lu, R. Liu, Z. Zheng, Z. Wang, C. Wu, Z. Ni, and X. Liu, “Algorithm for detecting ice overlaying water multilayer clouds using the infrared bands of FY-4a/AGRI,”IEEE Geoscience and Remote Sensing Letters, vol. 21, p. 1001205, 2024, publisher: IEEE

    2024

  9. [17]

    Multi- layer cloud detection and distributions over the asia–pacific region based on geostationary satellite imagers,

    J. Wang, C. Liu, B. Yao, Y . Qian, X. Gu, Y . Kong, and S. Fan, “Multi- layer cloud detection and distributions over the asia–pacific region based on geostationary satellite imagers,”Journal of Geophysical Research: Atmospheres, vol. 129, no. 19, p. e2023JD040287, 2024

    2024

  10. [18]

    7.05 - satellite remote sensing of cloud vertical structure,

    A. K. Heidinger, Y . Li, S. Wanzong, Y . J. Noh, A. Walther, S. Tushaus, and S. Miller, “7.05 - satellite remote sensing of cloud vertical structure,” inComprehensive Remote Sensing, S. Liang, Ed. Elsevier, 2018, pp. 97–136

    2018

  11. [19]

    Evaluation of the MODIS Collection 6 multilayer cloud detection algorithm through comparisons with CloudSat Cloud Profiling Radar and CALIPSO CALIOP products,

    B. Marchant, S. Platnick, K. Meyer, and G. Wind, “Evaluation of the MODIS Collection 6 multilayer cloud detection algorithm through comparisons with CloudSat Cloud Profiling Radar and CALIPSO CALIOP products,”Atmospheric Measurement Techniques, vol. null, p. JOURNAL OF LATEX C...

    2021

  12. [20]

    Object-based cloud and cloud shadow detection in landsat imagery,

    Z. Zhu and C. E. Woodcock, “Object-based cloud and cloud shadow detection in landsat imagery,”Remote sensing of environment, vol. 118, pp. 83–94, 2012

    2012

  13. [21]

    Cloud detection methodologies: Variants and development—a review,

    S. Mahajan and B. Fataniya, “Cloud detection methodologies: Variants and development—a review,”Complex & Intelligent Systems, vol. 6, no. 2, pp. 251–261, 2020

    2020

  14. [22]

    Detection of single and multilayer clouds in an artificial neural network approach,

    S. Sun-Mack, P. Minnis, W. Smith, G. Hong, and Y . Chen, “Detection of single and multilayer clouds in an artificial neural network approach,” inRemote Sensing of Clouds and the Atmosphere XXII, 2017. [Online]. Available: https://www.semanticscholar.org/paper/ 70c372418610b9de...

    2017

  15. [23]

    Detecting multilayer clouds from the geostationary advanced himawari imager using machine learningtechniques,

    Z. Tan, C. Liu, S. Ma, X. Wang, J. Shang, J. Wang, W. Ai, and W. Yan, “Detecting multilayer clouds from the geostationary advanced himawari imager using machine learningtechniques,”IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1– 12, 2022. [Online]. Availabl...

    2022

  16. [24]

    Low cloud detection in multilayer scenes using satellite imagery with machine learning methods,

    J. M. Haynes, Y .-J. Noh, S. D. Miller, K. D. Haynes, I. Ebert- Uphoff, and A. Heidinger, “Low cloud detection in multilayer scenes using satellite imagery with machine learning methods,”Journal of Atmospheric and Oceanic Technology, 2022, section: Journal of Atmospheric and O...

    2022

  17. [25]

    Cloud Identification and Properties Retrieval of the Fengyun-4A Satellite Using a ResUnet Model,

    Z. Zhao, F. Zhang, Q. Wu, Z. Li, X. Tong, J. Li, and W. Han, “Cloud Identification and Properties Retrieval of the Fengyun-4A Satellite Using a ResUnet Model,”IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–18, 2023

    2023

  18. [26]

    Detection of single and multilayer clouds in an artificial neural network approach,

    S. Sun-Mack, P. Minnis, W. L. Smith Jr, G. Hong, and Y . Chen, “Detection of single and multilayer clouds in an artificial neural network approach,” inRemote Sensing of Clouds and the Atmosphere XXII, vol. 10424. SPIE, 2017, pp. 29–40

    2017

  19. [27]

    Cloud detection and classification algorithms for himawari-8 imager measurements based on deep learning,

    W. Li, F. Zhang, H. Lin, X. Chen, J. Li, and W. Han, “Cloud detection and classification algorithms for himawari-8 imager measurements based on deep learning,”IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–17, 2022

    2022

  20. [28]

    Mcdnet: Multilevel cloud detection network for remote sensing images based on dual- perspective change-guided and multi-scale feature fusion,

    J. Dong, Y . Wang, Y . Yang, M. Yang, and J. Chen, “Mcdnet: Multilevel cloud detection network for remote sensing images based on dual- perspective change-guided and multi-scale feature fusion,”International Journal of Applied Earth Observation and Geoinformation, vol. 129, p....

    2024

  21. [29]

    Cdunet: Cloud detection unet for remote sensing imagery,

    K. Hu, D. Zhang, and M. Xia, “Cdunet: Cloud detection unet for remote sensing imagery,”Remote Sensing, vol. 13, no. 22, p. 4533, 2021

    2021

  22. [30]

    Cnn-transnet: A hybrid cnn-transformer network with differential feature enhancement for cloud detection,

    N. Ma, L. Sun, Y . He, C. Zhou, and C. Dong, “Cnn-transnet: A hybrid cnn-transformer network with differential feature enhancement for cloud detection,”IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1–5, 2023

    2023

  23. [31]

    A naive bayesian cloud-detection scheme derived from calipso and applied within patmos-x,

    A. K. Heidinger, A. T. Evan, M. J. Foster, and A. Walther, “A naive bayesian cloud-detection scheme derived from calipso and applied within patmos-x,”Journal of Applied Meteorology and Climatology, vol. 51, no. 6, pp. 1129–1144, 2012

    2012

  24. [32]

    Physics-driven machine learning algorithm facilitates multilayer cloud property retrievals from geostationary passive imager measurements,

    W. Li, F. Zhang, B. Guo, H. Fu, and H. Letu, “Physics-driven machine learning algorithm facilitates multilayer cloud property retrievals from geostationary passive imager measurements,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–18, 2024

    2024

  25. [33]

    Physics-Driven Machine Learning Algorithm Facili- tates Multilayer Cloud Property Retrievals From Geosta- tionary Passive Imager Measurements,

    ——, “Physics-Driven Machine Learning Algorithm Facili- tates Multilayer Cloud Property Retrievals From Geosta- tionary Passive Imager Measurements,”IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–18,

  26. [34]

    Effects of Linear Calibration Errors at Low Temperature End of Thermal Infrared Band: Lesson from Failures in Cloud Top Property Retrieval of FengYun-4A Geostationary Satellite,

    F. Wang, M. Min, N. Xu, C. Liu, Z. Wang, and L. Zhu, “Effects of Linear Calibration Errors at Low Temperature End of Thermal Infrared Band: Lesson from Failures in Cloud Top Property Retrieval of FengYun-4A Geostationary Satellite,”IEEE Transactions on Geoscience and Remote Se...

    2022

  27. [35]

    Introducing the new generation of chinese geostationary weather satellites, fengyun-4,

    J. Yang, Z. Zhang, C. Wei, F. Lu, and Q. Guo, “Introducing the new generation of chinese geostationary weather satellites, fengyun-4,” Bulletin of the American Meteorological Society, vol. 98, no. 8, pp. 1637–1658, 2017. [Online]. Available: https://journals.ametsoc.org/ view/...

    2017

  28. [36]

    An introduction to himawari-8/9—japan’s new-generation geostationary meteorological satellites,

    K. Bessho, K. Date, M. Hayashi, A. Ikeda, T. Imai, H. Inoue, Y . Ku- magai, T. Miyakawa, H. Murata, T. Ohnoet al., “An introduction to himawari-8/9—japan’s new-generation geostationary meteorological satellites,”Journal of the Meteorological Society of Japan. Ser . II, vol. 94...

    2016

  29. [37]

    Characterization of bias of ad- vanced himawari imager infrared observations from nwp background simulations using crtm and rttov,

    X. Zou, X. Zhuge, and F. Weng, “Characterization of bias of ad- vanced himawari imager infrared observations from nwp background simulations using crtm and rttov,”Journal of Atmospheric and Oceanic Technology, vol. 33, no. 12, pp. 2553–2567, 2016

    2016

  30. [38]

    CALIPSO mission: spaceborne lidar for observation of aerosols and clouds,

    D. M. Winker, J. R. Pelon, and M. P. McCormick, “CALIPSO mission: spaceborne lidar for observation of aerosols and clouds,” inLidar Remote Sensing for Industry and Environment Monitoring III, vol. 4893. International Society for Optics and Photonics, 2003, pp. 1–11

    2003

  31. [39]

    The cloudsat mission and the a-train.-a new dimension of space-based observations of clouds and precipitation,

    G. Stephen, “The cloudsat mission and the a-train.-a new dimension of space-based observations of clouds and precipitation,”Bull. Amer . Meteor . Soc., vol. 83, pp. 1771–1790, 2002

    2002

  32. [40]

    Construction of Nighttime Cloud Layer Height and Classification of Cloud Types,

    S. Chen, C. Cheng, X. Zhang, L. Su, B. Tong, C. Dong, F. Wang, B. Chen, W. Chen, and D. Liu, “Construction of Nighttime Cloud Layer Height and Classification of Cloud Types,”Remote. Sens., vol. 12, p. 668, 2020. [Online]. Available: https://www.semanticscholar.org/paper/ 15afe...

    2020

  33. [41]

    New insights about cloud vertical structure from CloudSat and CALIPSO observations,

    L. Oreopoulos, N. Cho, and D. Lee, “New insights about cloud vertical structure from CloudSat and CALIPSO observations,”Journal of Geophysical Research: Atmospheres, vol. 122, no. 17, pp. 9280– 9300, 2017, publisher: Wiley Online Library. [Online]. Available: https://agupubs.o...

    2017 doi

  34. [42]

    Global distribution of cirrus clouds from cloudsat/cloud-aerosol lidar and infrared pathfinder satellite ob- servations (calipso) measurements,

    K. Sassen, Z. Wang, and D. Liu, “Global distribution of cirrus clouds from cloudsat/cloud-aerosol lidar and infrared pathfinder satellite ob- servations (calipso) measurements,”Journal of Geophysical Research: Atmospheres, vol. 113, no. D8, 2008

    2008

  35. [43]

    Identifica- tion of ice-over-water multilayer clouds using multispectral satellite data in an artificial neural network,

    S. Sun-Mack, P. Minnis, Y . Chen, G. Hong, and W. L. S. Jr, “Identifica- tion of ice-over-water multilayer clouds using multispectral satellite data in an artificial neural network,”Atmospheric Measurement Techniques, vol. 17, p. 3323–3346, 2024

    2024

  36. [2024]

    Available: https://www.semanticscholar.org/paper/ 7205260816f322689fe1bd52245aaeb1aec0d32c

    [Online]. Available: https://www.semanticscholar.org/paper/ 7205260816f322689fe1bd52245aaeb1aec0d32c

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.