Pith. sign in

REVIEW 4 major objections 5 minor 118 references

Benchmarking Deep Learning Models for Dense Event Classification of Offshore Wind Infrastructure in Sentinel-1 Time Series

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A supervised BiLSTM outperforms the rule-based baseline at dense event classification of offshore wind infrastructure in Sentinel-1 time series.

desk verdict A useful, honestly written benchmark, but the headline comparison is not out-of-sample: test-set checkpoint selection inflates the BiLSTM's reported gain. read the letter →

arxiv 2608.04706 v1 pith:7MD3OARY submitted 2026-08-05 cs.LG

Thorsten Hoeser , Felix Bachofer , Claudia Kuenzer This is my paper
classification cs.LG
keywords offshorewindinfrastructureSentinel-1SARtimeseriesdenseeventclassificationBiLSTMdeeplearningbenchmarkingself-superviseddeploymentphasemonitoringeditsimilarityevaluation
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 asks whether deep learning can replace an expert rule-based classifier for dense, event-level labeling of offshore wind infrastructure in Sentinel-1 SAR time series. Using 661,732 newly hand-annotated events and a held-out test set of 328,657 events, it trains ten model-training variants and finds that a supervised bidirectional LSTM achieves the best agreement, raising the area under the collapsed edit-similarity curve from 0.7853 to 0.8509 and the perfect-match rate from 0.3508 to 0.5063. Combining the BiLSTM predictions with the baseline labels by minimizing label transitions recovers performance on rare classes such as platforms without harming the others. The improved labels let the authors isolate individual turbine deployment phases worldwide, yielding median deployment durations of 84 days for China, 242 days for the EU, and 258 days for the UK.

What carries the argument

The central object is the supervised bidirectional LSTM built on a shared event-profile encoder. Each Sentinel-1 acquisition is reduced to a 64-bin 1D profile of maximum backscatter in range direction, paired with a binary valid-bin mask, and standardized with global statistics (mean -22.67 dB, sigma 5.98 dB). The encoder is a three-block 1D convolutional network with masked average and max pooling; the BiLSTM then reads sequences of 64 events in both directions and outputs per-event class logits. Evaluation uses the collapsed edit-similarity curve (AUCEditSim), a Levenshtein-derived sequence agreement metric, and transition-count comparisons; the final ensemble selects, per sequence, the label set with fewer class transitions.

What would settle it

Take a sample of turbines across the EU, UK, and China with registry-verified construction and grid-connection dates, run the published BiLSTM and ensemble on their Sentinel-1 sequences, and measure the offset between predicted and official deployment-phase boundaries; systematic offsets exceeding the acquisition repeat cycle, or predicted phases with no registry counterpart, would falsify the claimed day-level monitoring capability.

Watch

Extended reading notes

Core claim

The central claim is that dense event classification of Sentinel-1 offshore wind time series can be advanced beyond rule-based heuristics by supervised deep learning with bidirectional temporal context. The supervised-only BiLSTM encodes each acquisition as a 1D SAR backscatter profile, processes windows of 64 events in both directions, and predicts one of six classes per event: water, vessel, foundation, mooring/construction, turbine, or platform. On the benchmark test set it reaches an AUCEditSim of 0.8509 and a perfect-match rate of 0.5063, both above the rule-based baseline's 0.7853 and 0.3508. A transition-minimizing ensemble of BiLSTM and baseline labels raises the sequence-level rank correlation with test-set transition counts to 0.924, repairs the platform-class F1 from 0.4677 to 0.7352, and yields the labels used for the global deployment-phase analysis.

Load-bearing premise

The 661,732 new hand labels and the 328,657-event test set are treated as reliable ground truth for the six event classes, yet no inter-annotator agreement, independent verification against construction records, or label-quality audit is reported.

Editorial extensions

If this is right

  • Dense event labels can now be produced automatically for all 14,840,637 events across 15,606 global sequences, replacing the rule-based classifier as the inference engine.
  • The deployment phase of individual turbines can be isolated at day-level precision, revealing regional differences in duration and seasonality across China, the EU, and the UK.
  • Bidirectional context is essential: all bidirectional variants beat the baseline, while causal and monotemporal models do not.
  • Self-supervised pretraining did not help bidirectional models but did improve causal transformers, so labeled data, not representation quality, is the current bottleneck for offline models.
  • The ensemble's recovery of platform and foundation classes shows that the deep model has not yet learned everything the rule-based classifier encodes, pointing to disagreement-based active learning as a next step.

Reading between the lines

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

  • Because the test labels were produced by the same group that designed the rule-based baseline, part of the measured gain may reflect systematic label bias shared by both human and rule-based labeling; an independent audit against construction records would quantify this.
  • The success of transition-minimizing ensembling suggests that a temporal-smoothness prior, applied directly inside the deep model, could recover some of the remaining gap without needing a second label source.
  • The fixed 64-event window covers variable real-time periods because revisit times differ by region, so extending the window or using acquisition-time-aware architectures could remove the temporally implausible foundation predictions the authors observe.
  • If the causal stream can be improved, a near-real-time global situational awareness service for offshore infrastructure becomes feasible; the paper's data and predictions provide the training basis for that.
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

4 major / 5 minor

Summary. The paper compares ten deep learning model-training variants for dense six-class event labeling of Sentinel-1 offshore wind infrastructure time series. It introduces 500 newly hand-annotated time series with 661,732 events, trains FCNN, LSTM, and transformer variants with monotemporal, causal, and bidirectional context, with and without self-supervised pretraining, and evaluates them on a strictly isolated 328,657-event test set against the rule-based baseline of DeepOWT v3.25.1. The supervised BiLSTM is reported as best, raising AUCEditSim from 0.7853 to 0.8509 and perfect match rate from 0.3508 to 0.5063. The authors then construct a transition-minimizing ensemble of BiLSTM and baseline labels, use the resulting labels to isolate deployment phases of individual turbines globally, and report deployment durations and regional patterns for China, the EU, and the UK.

Significance. The data contribution is substantial: 500 hand-labeled time series, open predictions for 14.8 million events, and a systematic comparison of ten training variants are exactly the kind of resource the remote-sensing time-series community needs. If the headline comparison were clean, the result would be practically valuable for global offshore wind monitoring and for benchmarking future models. However, the evaluation protocol currently makes the central quantitative claim an optimistic upper bound rather than a fair out-of-sample comparison. The manuscript therefore needs a revision of the evaluation before the benchmark claim can be accepted.

major comments (4)
  1. [§4.1; §3.1.3] The evaluation is not fully out-of-sample because the reported numbers come from "the checkpoint with the peak AUCEditSim" (§4.1), while §3.1.3 states that early stopping is based on validation macro F1. Since AUCEditSim is computed on the 328,657-event test set, selecting the checkpoint by peak test-set score and then comparing that score with the rule-based baseline on the same test set converts training fluctuation into apparent gains. Please report the validation-selected checkpoint as the primary result, or use a nested or separate hold-out for checkpoint selection, and quantify the difference between validation-selected and test-selected numbers.
  2. [§4.1; §4.2] Each experiment appears to be a single training run, and no confidence intervals, standard deviations, or seed repetitions are reported anywhere in §4. With run-to-run variability unknown, the headline AUCEditSim gap of 0.8509 versus 0.7853 cannot be assessed for statistical reliability. Please provide multiple seeds, at least for the supervised BiLSTM and the best competitive variants, and report paired differences with confidence intervals, especially because checkpoint selection by peak test-set score amplifies the effect of random variation.
  3. [§3.1.6; §4.1; §4.2] The ensemble rule in §3.1.6 (choose the label source with fewer transitions, with BiLSTM on ties) is introduced after §4.1 reports that the BiLSTM's transition counts correlate better with the test set, and the ensemble is then evaluated on the same test set in §4.2. This is test-set-based method selection for the ensemble, so the reported AUCEditSim of 0.86 and the per-class F1 recoveries are optimistic. Please specify the ensemble rule before looking at the test set, tune it on a validation split, or show via sensitivity analysis that the conclusion is robust to the rule choice. Similarly, the guarding threshold of "ten or more consecutive water/vessel observations" used in §3.2 to define deployment onset was introduced after observing BiLSTM misclassifications (§5.3); its choice needs a sensitivity analysis or pre-specification.
  4. [§2.1; §3.1.5] The hand-annotated training labels and the test set come from the same group, and no inter-annotator agreement, label-quality audit, or independent verification against construction records is reported. Since the benchmark claims are measured against these labels, label noise or systematic annotation conventions shared with the rule-based baseline could inflate the apparent improvements. Please report annotation quality checks or at least discuss the potential direction and magnitude of this bias, and state clearly whether the test set is publicly available for independent re-labeling.
minor comments (5)
  1. [§2.1] The phrase "V-polarized (vertical sent - horizontal received)" is awkward and incomplete; it should read "vertical transmit, horizontal receive" with the hyphen and spacing fixed.
  2. [References] The Flanders Marine Institute reference has an incomplete year: "last access: 12 February 202." should be a complete year.
  3. [§3.1.6] The text contains "theunclearlabel" without a space; it should read "the unclear label."
  4. [§3.2] The phrase "V oronoi tessellation" has an unintended space; it should read "Voronoi tessellation."
  5. [§3.1.5] The abbreviation AUCEditSim is used in the abstract and throughout but is only defined in §3.1.5; please define it at first use in the abstract or introduction.

Circularity Check

2 steps flagged · score 6.0 of 10

Reported test-set results are selected by maximizing the test metric itself: Section 4.1 chooses the checkpoint with peak AUCEditSim, and the transition-minimizing ensemble is motivated by test-set correlations and then evaluated on the same test set.

  1. fitted input called prediction [Section 4.1 (Model performances), with Sections 3.1.4 and 3.1.5]
    "For each, the following figures report the evaluation results of the checkpoint with the peak AUCEditSim. — All experiments produce trained model checkpoints, which are used to predict event labels on the benchmark test set."

    AUCEditSim (Eq. 2) is defined on the M test sequences, and Section 3.1.4 says checkpoints 'are used to predict event labels on the benchmark test set.' Early stopping (Section 3.1.3) monitors validation macro F1, so the model selected for deployment would come from the validation criterion, not the test set. By reporting the checkpoint with peak AUCEditSim, the paper selects an epoch whose test-set score is the maximum over all evaluated epochs and then presents that maximum as the model's test-set performance. The headline comparison — BiLSTM 0.8509 vs. rule-based 0.7853 — is thus an optimistic upper bound obtained by optimizing the reported metric on the very data used for the comparison. The perfect match rate 0.5063 has the same status.

  2. fitted input called prediction [Section 4.1 (last paragraph), Sections 3.1.6 and 4.2]
    "We correlate the transition counts of each prediction set, the BiLSTM and the rule-based baseline, with the transition counts of the test set. The BiLSTM reaches a Spearman’s ρ of 0.89, compared to 0.843 for the rule-based baseline. ... This motivated us to go further into minimizing label transition counts globally for the predictions of the 15,606 sequences."

    The ensemble rule (Section 3.1.6) chooses, for each sequence, the labels with fewer transitions. The choice of transition-count minimization as the combination criterion is explicitly motivated in Section 4.1 by correlations computed against the test set ('We correlate the transition counts ... with the transition counts of the test set ... These statistical findings ... motivated us'). Section 4.2 then evaluates the resulting ensemble on that same test set and reports improved agreement (Spearman ρ 0.924; platform F1 0.7352).

full rationale

The paper is an empirical benchmark rather than a formal derivation chain, and most of its self-citations are legitimate data/comparator citations: DeepOWT v3.25.1, the rule-based baseline, and the hand-labeled test set come from Hoeser et al. (2026), and the newly hand-labeled 661,732 events are external to the model optimization. Those do not, by themselves, make the central comparison circular. The two concrete reductions are test-set-driven model selection steps. First, Section 4.1 reports, for every variant, the checkpoint with peak AUCEditSim. Since AUCEditSim (Eq. 2) is computed on the 328,657-event held-out test set and checkpoints are evaluated on that set (Section 3.1.4), the headline BiLSTM scores (0.8509 AUC, 0.5063 perfect match) are maxima over test-set epoch scores, not independent out-of-sample estimates; early stopping (Section 3.1.3) uses validation macro F1, so the reported checkpoint differs from the validation-selected deployable model. This is a fitted-input-called-prediction issue affecting the central claim. Second, the ensemble's transition-minimization objective is chosen after correlating transition counts with test-set labels (Section 4.1) and is then evaluated on that same test set (Section 4.2); the reported ensemble gains (ρ 0.924, platform F1 recovery) are in part a restatement of the criterion used to construct it. However, the model itself is trained only on human labels, the SSL/causal comparisons are not selected on the headline metric in the same way, and the deployment-duration medians are descriptive rather than derived from the test-set comparison. The circularity is therefore partial: score 6, not 8-10.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. Its central results rest on hand-label quality, the prior data set's profile construction, and several hand-chosen thresholds.

free parameters (4)
  • Window length L=64
    Fixed sequence length for all multitemporal models; chosen by hand. The paper notes in §5.3 that this context may be too short to learn long-range temporal plausibility.
  • Deployment-phase guarding threshold = 10 consecutive water/vessel events
    In §3.2, deployment start is anchored to the latest long water/vessel sub-sequence of ten or more events; this hand-chosen threshold suppresses BiLSTM misclassifications and directly affects deployment duration estimates.
  • Ensemble transition rule = Choose label source with fewer transitions; ties go to BiLSTM
    In §3.1.6, the rule is motivated by test-set transition-count correlations (§4.1) and is then evaluated on the same test set, making its reported benefit optimistic.
  • Transition window jitter and padding = jitter -3 to -8, 8 to 16; overshoot 16
    In §2.1, these data-augmentation choices shape the training distribution of transition windows.
assumptions (4)
  • domain assumption The DeepOWT v3.25.1 rule-based labels are sufficiently accurate to serve as pseudo-labels for SSL sampling.
    In §2.1, SSL training sets are sampled using the rule-based classifier's transitions, so errors in those labels propagate into pretraining data.
  • domain assumption Column-wise maximum VH-polarized backscatter profiles are sufficient representations to distinguish the six event classes.
    The entire data set is built on this reduction from 2D SAR patches, as described in §2 from Hoeser et al. [2026].
  • domain assumption The held-out test set labels from Hoeser et al. [2026] are ground truth.
    All benchmark numbers are computed against this test set (§2.1 and §3.1.5); no external validation is provided.
  • domain assumption Collapsed edit similarity at quality threshold is the most meaningful evaluation metric.
    The paper adopts this metric on the recommendation of Hoeser et al. [2026] (§3.1.5); it rewards homogeneous sequences and may mask certain error types.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Benchmarking Deep Learning Models for Dense Event Classification of Offshore Wind Infrastructure in Sentinel-1 Time Series." pith.science (2026). https://pith.science/paper/7MD3OARY

@misc{pith2026260804706,
  author       = {Pith},
  title        = {Pith review of: Benchmarking Deep Learning Models for Dense Event Classification of Offshore Wind Infrastructure in Sentinel-1 Time Series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7MD3OARY}},
  note         = {Machine review of arXiv:2608.04706}
}
read the original abstract

Monitoring of offshore wind energy infrastructure life cycles, especially during the deployment phase, is an important contribution for stakeholders to make informed decisions in a phase of increasing deployment activities. ESA's Sentinel-1 Synthetic Aperture Radar (SAR) mission produces large data archives that enable the global monitoring of offshore wind infrastructure. Turning these high-volume archives into information requires algorithms that automatically extract single event labels from dense time series at a global scale. In this study, we present a structured comparison of ten deep learning model-training variants for the dense classification of Sentinel-1 based offshore wind infrastructure time series, aiming to advance rule-based event classification of this task. We trained LSTM, Transformer, and fully connected model variants with monotemporal, unidirectional, and bidirectional context awareness, each with and without self-supervised pretraining. Among these, the supervised BiLSTM performs best, raising the target AUC score from 0.7853 for the rule-based baseline to 0.8509, and the perfect match rate from 0.3508 to 0.5063. Combining the BiLSTM predictions with the existing baseline labels in a label-transition-minimising ensemble further improves agreement with the test data. Using these improved labels, we isolate the deployment phase of individual turbines at a global scale and conduct a regional and subregional analysis covering 2016-01-01 to 2025-03-31, reporting median deployment durations of 84 d (China), 242 d (EU), and 258 d (UK). Deployment-related drivers, including legal regulations such as subsidies, and environmental conditions, emerge clearly from the analysed results across multiple spatial scales.

Figures

Figures reproduced from arXiv: 2608.04706 by the authors.

Figure 1
Figure 1. The upper rows show panels of example Sentinel-1 SAR backscatter profiles taken at offshore wind [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Overview of data set sampling and restructuring to create different deep learning ready data sets from the [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. The proposed experimental setup is driven by three major design choices: • Compare baseline models without sequential awareness to models with sequential awareness — monotemporal stream. • Train models with access to data around a potential class transition, giving them the capability to observe events before and after potential label transitions — bidirectional stream, useful for offline inference, such as historic… view at source ↗
Figures from the paper (13 more)
Figure 3
Figure 3. Figure 3: Experiment design for the structured investigation of different data, model and training combinations, centered [PITH_FULL_IMAGE:figures/full_fig_p008_3.png]
Figure 4
Figure 4. Figure 4: Visualization of the three different self supervised learning data preparation processes. Masking is applied [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Training curves across all ten model–training variants. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Performance metrics comparison across all model–training variants and comparison with the rule-based [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Class-wise performance metrics comparison across all model–training variants and comparison with the [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 10
Figure 10. Figure 10: Panels a+b) illustrate cases where the BiLSTM produces more homogeneous and more accurate predictions, [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 8
Figure 8. Figure 8: Edit similarity curves for all model–training variants and comparison with the rule-based baseline classifier. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Divergence expressed as the number of label transitions in a sequence between the BiLSTM and rule-based [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Comparison of four sequences with a+b) noisy labels, predicted by the rule-based baseline classifier and [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: Correlations of transition rates, where transition rates are the number of label transitions in a sequence [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Consolidated performance metric overview to compare improvements of the ensemble results with the two [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Evolution of deployment phases over time grouped by the three major markets China (CN), European Union [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Temporal evolution of the deployment dynamics as derived from the predictions in this study, grouped by [PITH_FULL_IMAGE:figures/full_fig_p021_14.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

118 extracted references · 39 canonical work pages

  1. [1]

    Wong and Christian Thomas and Patrick Halpin , keywords =

    Brian A. Wong and Christian Thomas and Patrick Halpin , keywords =. Automating offshore infrastructure extractions using synthetic aperture radar & Google Earth Engine , journal =. 2019 , issn =. doi:https://doi.org/10.1016/j.rse.2019.111412 , url =

    work page arXiv 2019

  2. [3]

    Soviet Physics Doklady , volume=

    Binary codes capable of correcting deletions, insertions and reversals , author=. Soviet Physics Doklady , volume=

  3. [4]

    Heavy maintenance campaign completed on Hywind Scotland floating offshore wind farm , year =

  4. [5]

    and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and

    Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and. Nature Methods , year =

  5. [6]

    Earth System Science Data , VOLUME =

    Hoeser, Thorsten and Feuerstein, Stefanie and Kuenzer, Claudia , TITLE =. Earth System Science Data , VOLUME =. 2022 , NUMBER =

    2022

  6. [7]

    2017 , note =

    Google Earth Engine: Planetary-scale geospatial analysis for everyone , journal =. 2017 , note =. doi:10.1016/j.rse.2017.06.031 , url =

    doi:10.1016/j.rse.2017.06.031 2017

  7. [8]

    Sensors , VOLUME =

    Zhang, Jing and Wang, Qi and Su, Fenzhen , TITLE =. Sensors , VOLUME =. 2019 , NUMBER =

    2019

  8. [9]

    Sentinel-1 Algorithms: Sentinel-1 Preprocessing , year =

Show all 118 references
  1. [10]

    Entropy , VOLUME =

    Wang, Qi and Zhang, Jing and Su, Fenzhen , TITLE =. Entropy , VOLUME =. 2019 , NUMBER =

    2019

  2. [13]

    International Journal of Remote Sensing , volume =

    Shuai Zhang and Fangxiong Wang and Yingzi Hou and Junfu Wang and Jianke Guo , title =. International Journal of Remote Sensing , volume =. 2024 , publisher =

    2024

  3. [14]

    GIScience & Remote Sensing , volume =

    Qiannan Ding and Bo Tian and Chunpeng Chen and Yuekai Hu and Xue Li , title =. GIScience & Remote Sensing , volume =. 2024 , publisher =

    2024

  4. [16]

    Remote Sensing , VOLUME =

    Song, Xike and Li, Ziyang , TITLE =. Remote Sensing , VOLUME =. 2025 , NUMBER =

    2025

  5. [17]

    Extraction of Offshore Wind Turbines in China by Combining Multispectral and SAR Image Data , year=

    Wang, Fangxiong and Zhang, Shuai and Hou, Yingzi and Wang, Junfu , journal=. Extraction of Offshore Wind Turbines in China by Combining Multispectral and SAR Image Data , year=

  6. [19]

    LandTrendr — Temporal segmentation algorithms , journal =

    Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr — Temporal segmentation algorithms , journal =. 2010 , issn =. doi:10.1016/j.rse.2010.07.008 , url =

    2010 doi

  7. [21]

    Geo-spatial Information Science , volume =

    Qiannan Ding and Chunpeng Chen and Bo Tian and Wenxiang Ji and Bin Zhang and Xue Li and Qing Yuan , title =. Geo-spatial Information Science , volume =. 2025 , publisher =

    2025

  8. [22]

    and Matos-Carvalho, João P

    Mota, Tiago and Vieira, Mario and Filipe, Leonardo and Fernandes, Daniel and Garcia, Nuno C. and Matos-Carvalho, João P. , booktitle=. Applying Machine Learning to SENTINEL Satellite Images to Predict the Operational State of Floating Offshore Wind Turbines , year=

  9. [23]

    Energies , VOLUME =

    de Carvalho, Osmar Luiz Ferreira and de Carvalho Junior, Osmar Abílio and de Albuquerque, Anesmar Olino and Silva, Daniel Guerreiro e , TITLE =. Energies , VOLUME =. 2025 , NUMBER =

    2025

  10. [24]

    Remote Sensing , VOLUME =

    Fei, Yasen and Gao, Yongnian and Gu, Hongyuan and Sun, Yongqi and Tian, Yanjun , TITLE =. Remote Sensing , VOLUME =. 2025 , NUMBER =

    2025

  11. [27]

    Automatic extraction of offshore platforms using time-series Landsat-8 Operational Land Imager data , journal =

    Yongxue Liu and Chao Sun and Yuhao Yang and Minxi Zhou and Wenfeng Zhan and Wangyu Cheng , keywords =. Automatic extraction of offshore platforms using time-series Landsat-8 Operational Land Imager data , journal =. 2016 , issn =. doi:https://doi.org/10.1016/j.rse.2015.12.047 , url =

    2016 doi

  12. [29]

    2023 , issn =

    A review of geospatial technologies for improving Marine Spatial Planning: Challenges and opportunities , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.ocecoaman.2022.106280 , url =

    2023

  13. [30]

    Susan Gourvenec and Fraser Sturt and Emily Reid and Federico Trigos , keywords =. Global assessment of historical, current and forecast ocean energy infrastructure: Implications for marine space planning, sustainable design and end-of-engineered-life management , journal =. 20...

    2022

  14. [31]

    2023 , issn =

    The properties of the global offshore wind turbine fleet , journal =. 2023 , issn =. doi:10.1016/j.rser.2023.113667 , url =

    2023

  15. [32]

    2024 , issn =

    Future global offshore wind energy under climate change and advanced wind turbine technology , journal =. 2024 , issn =. doi:10.1016/j.enconman.2024.119075 , url =

    2024

  16. [33]

    2025 , issn =

    From macro to micro: A multi-scale method for assessing coastal wind energy potential in China , journal =. 2025 , issn =. doi:10.1016/j.apenergy.2025.125729 , url =

    2025

  17. [34]

    GMES Sentinel-1 mission , journal =

    Ramon Torres and Paul Snoeij and Dirk Geudtner and David Bibby and Malcolm Davidson and Evert Attema and Pierre Potin and BjÖrn Rommen and Nicolas Floury and Mike Brown and Ignacio Navas Traver and Patrick Deghaye and Berthyl Duesmann and Betlem Rosich and Nuno Miranda and Cla...

    2012 doi

  18. [35]

    Sentinel-1B In-Flight Anomaly Summary Report , year =

  19. [36]

    and Porchetta, S

    Ivanova, T. and Porchetta, S. and Buckingham, S. and Glabeke, G. and van Beeck, J. and Munters, W. , TITLE =. Wind Energy Science , VOLUME =. 2025 , NUMBER =

    2025

  20. [37]

    2024 , issn =

    Offshore wind energy potential in Shandong Sea of China revealed by ERA5 reanalysis data and remote sensing , journal =. 2024 , issn =. doi:10.1016/j.jclepro.2024.142745 , url =

    2024

  21. [38]

    2025 , issn =

    Quantification of turbid wakes in offshore wind farms using satellite remote sensing , journal =. 2025 , issn =. doi:10.1016/j.scitotenv.2025.178814 , url =

    2025

  22. [39]

    2022 , issn =

    SyntEO: Synthetic dataset generation for earth observation and deep learning – Demonstrated for offshore wind farm detection , journal =. 2022 , issn =. doi:10.1016/j.isprsjprs.2022.04.029 , url =

    2022 doi

  23. [40]

    Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , year=

    Ren, Shaoqing and He, Kaiming and Girshick, Ross and Sun, Jian , journal=. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , year=

  24. [41]

    arXiv preprint arXiv:2405.14458 , year=

    YOLOv10: Real-Time End-to-End Object Detection , author=. arXiv preprint arXiv:2405.14458 , year=

    arXiv

  25. [42]

    Journal of Renewable and Sustainable Energy , volume =

    Roddier, Dominique and Cermelli, Christian and Aubault, Alexia and Weinstein, Alla , title =. Journal of Renewable and Sustainable Energy , volume =. 2010 , month =. doi:10.1063/1.3435339 , url =

    2010 doi

  26. [43]

    Emergence of floating offshore wind energy: Technology and industry , journal =

    Nuno Bento and Margarida Fontes , keywords =. Emergence of floating offshore wind energy: Technology and industry , journal =. 2019 , issn =. doi:https://doi.org/10.1016/j.rser.2018.09.035 , url =

    2019 doi

  27. [44]

    Union of the ESRI Country shapefile and the Exclusive Economic Zones (version 4) , howpublished =

  28. [45]

    Net Zero Strategy: Build Back Greener , howpublished =

  29. [46]

    An EU Strategy to harness the potential of offshore renewable energy for a climate neutral future , howpublished =

  30. [47]

    Gesetz zur Entwicklung und Förderung der Windenergie auf See (WindSeeG) , year =

  31. [48]

    UN Climate Summit 2025, 24 September 2025 , year =

    2025

  32. [49]

    An innovative platform for exploring the Earth , howpublished =

    terrabyte. An innovative platform for exploring the Earth , howpublished =

  33. [50]

    and Bachofer, F

    Hoeser, T. and Bachofer, F. and Kuenzer, C. , title =. 2025 , doi =

    2025

  34. [51]

    and Kuenzer, C

    Hoeser, T. and Kuenzer, C. , title =. 2022 , doi =

    2022

  35. [52]

    doi:10.5281/zenodo.18735421 , url =

    Hoeser, Thorsten and Bachofer, Felix and Kuenzer, Claudia , title =. doi:10.5281/zenodo.18735421 , url =

    doi

  36. [53]

    Frontiers in Handwriting Recognition (ICFHR), 2014 14th International Conference on , pages=

    Real-time segmentation of on-line handwritten arabic script , author=. Frontiers in Handwriting Recognition (ICFHR), 2014 14th International Conference on , pages=. 2014 , organization=

    2014

  37. [54]

    Soft Computing and Pattern Recognition (SoCPaR), 2014 6th International Conference of , pages=

    Fast classification of handwritten on-line Arabic characters , author=. Soft Computing and Pattern Recognition (SoCPaR), 2014 6th International Conference of , pages=. 2014 , organization=

    2014

  38. [55]

    Advanced Data Mining and Applications: 12th International Conference, ADMA 2016, Gold Coast, QLD, Australia, December 12-15, 2016, Proceedings 12 , pages=

    Prediction-Based, Prioritized Market-Share Insight Extraction , author=. Advanced Data Mining and Applications: 12th International Conference, ADMA 2016, Gold Coast, QLD, Australia, December 12-15, 2016, Proceedings 12 , pages=. 2016 , organization=

    2016

  39. [56]

    Download Data | Global Energy Monitor , year =

  40. [58]

    2025 , howpublished =

    2025

  41. [59]

    2025 , howpublished =

    China Installs 26. 2025 , howpublished =

    2025

  42. [62]

    The Journal of Supercomputing , year =

    Khan, Mehak and Wang, Hongzhi and Riaz, Adnan and Elfatyany, Aya and Karim, Sajida , title =. The Journal of Supercomputing , year =. doi:10.1007/s11227-020-03560-z , url =

    doi

  43. [64]

    A Review of Deep Learning Models for Time Series Prediction , year=

    Han, Zhongyang and Zhao, Jun and Leung, Henry and Ma, King Fai and Wang, Wei , journal=. A Review of Deep Learning Models for Time Series Prediction , year=

  44. [67]

    LSTM can Solve Hard Long Time Lag Problems , url =

    Hochreiter, Sepp and Schmidhuber, J\". LSTM can Solve Hard Long Time Lag Problems , url =. Advances in Neural Information Processing Systems , editor =

  45. [68]

    , journal=

    Amari, S.-I. , journal=. Learning Patterns and Pattern Sequences by Self-Organizing Nets of Threshold Elements , year=

  46. [69]

    , journal=

    Werbos, P.J. , journal=. Backpropagation through time: what it does and how to do it , year=

  47. [70]

    1990 , issn =

    Finding structure in time , journal =. 1990 , issn =. doi:https://doi.org/10.1016/0364-0213(90)90002-E , url =

    1990 doi

  48. [71]

    and Kaiser, Lukasz and Polosukhin, Illia , title =

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N. and Kaiser, Lukasz and Polosukhin, Illia , title =. Advances in Neural Information Processing Systems 30 (NeurIPS 2017) , year =

    2017

  49. [73]

    and Paliwal, K.K

    Schuster, M. and Paliwal, K.K. , journal=. Bidirectional recurrent neural networks , year=

  50. [74]

    Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) , year =

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina , title =. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) , year =. doi:10.18653/v1/N19-1423 , url =

    2019 doi

  51. [75]

    2018 , howpublished =

    Radford, Alec and Narasimhan, Karthik and Salimans, Tim and Sutskever, Ilya , title =. 2018 , howpublished =

    2018

  52. [76]

    Language Models are Few-Shot Learners , url =

    Brown, Tom and Mann, Benjamin and Ryder, Nick and Subbiah, Melanie and Kaplan, Jared D and Dhariwal, Prafulla and Neelakantan, Arvind and Shyam, Pranav and Sastry, Girish and Askell, Amanda and Agarwal, Sandhini and Herbert-Voss, Ariel and Krueger, Gretchen and Henighan, Tom a...

  53. [78]

    Learning Complex, Extended Sequences Using the Principle of History Compression , year=

    Schmidhuber, Jürgen , journal=. Learning Complex, Extended Sequences Using the Principle of History Compression , year=

  54. [82]

    2024 , eprint =

    Tseng, Gabriel and Cartuyvels, Ruben and Zvonkov, Ivan and Purohit, Mirali and Rolnick, David and Kerner, Hannah , title =. 2024 , eprint =

    2024

  55. [95]

    Akhtar, B

    N. Akhtar, B. Geyer, and C. Schrum. Impacts of accelerating deployment of offshore windfarms on near-surface climate. Scientific Reports, 12 0 (1): 0 18307, 2022. ISSN 2045-2322. doi:10.1038/s41598-022-22868-9. URL https://doi.org/10.1038/s41598-022-22868-9

    2022 doi

  56. [96]

    Brown, B

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-Voss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S. Gray, B. Che...

    1901

  57. [97]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova. BERT : Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologie...

    2019 arXiv

  58. [98]

    Q. Ding, B. Tian, C. Chen, Y. Hu, and X. Li. Identifying the spatio-temporal distribution characteristics of offshore wind turbines in china from sentinel-1 imagery using deep learning. GIScience & Remote Sensing, 61 0 (1): 0 2407389, 2024. doi:10.1080/15481603.2024.2407389

    2024

  59. [99]

    Djath and J

    B. Djath and J. Schulz-Stellenfleth. Characterizing atmospheric wakes of offshore wind farms with sentinel-1a sar observations. Journal of Physics: Conference Series, 3224 0 (3): 0 032110, may 2026. doi:10.1088/1742-6596/3224/3/032110. URL https://doi.org/10.1088/1742-6596/322...

    2026 doi

  60. [100]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2021. UR...

    2010 arXiv

  61. [101]

    J. L. Elman. Finding structure in time. Cognitive Science, 14 0 (2): 0 179--211, 1990. ISSN 0364-0213. doi:https://doi.org/10.1016/0364-0213(90)90002-E. URL https://www.sciencedirect.com/science/article/pii/036402139090002E

    1990

  62. [102]

    EnBW he dreiht: First wind turbine installed

    EnBW Energie Baden-Württemberg AG . EnBW he dreiht: First wind turbine installed. https://www.enbw.com/press/enbw-he-dreiht-offshore-wind-farm.html, 2025. Press release. Accessed: 2026-06-29

    2025

  63. [103]

    An eu strategy to harness the potential of offshore renewable energy for a climate neutral future

    European Commission . An eu strategy to harness the potential of offshore renewable energy for a climate neutral future. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:52020DC0741, 2020. last access: 12 February 2026

    2020

  64. [104]

    Union of the esri country shapefile and the exclusive economic zones (version 4)

    Flanders Marine Institute . Union of the esri country shapefile and the exclusive economic zones (version 4). https://www.marineregions.org/, 2024. last access: 12 February 202

    2024

  65. [105]

    F. A. Gers, J. Schmidhuber, and F. Cummins. Learning to forget: Continual prediction with lstm. Neural Computation, 12 0 (10): 0 2451--2471, 10 2000. ISSN 0899-7667. doi:10.1162/089976600300015015. URL https://doi.org/10.1162/089976600300015015

    doi

  66. [106]

    Download data | global energy monitor

    Global Energy Monitor . Download data | global energy monitor. https://globalenergymonitor.org/download-data, 2026. Accessed: 2026-06-29

    2026

  67. [107]

    Graves and J

    A. Graves and J. Schmidhuber. Framewise phoneme classification with bidirectional lstm and other neural network architectures. Neural Networks, 18 0 (5): 0 602--610, 2005. ISSN 0893-6080. doi:https://doi.org/10.1016/j.neunet.2005.06.042. URL https://www.sciencedirect.com/scien...

    2005 doi

  68. [108]

    Z. Han, J. Zhao, H. Leung, K. F. Ma, and W. Wang. A review of deep learning models for time series prediction. IEEE Sensors Journal, 21 0 (6): 0 7833--7848, 2021. doi:10.1109/JSEN.2019.2923982

    2021

  69. [109]

    T. He, Y. Hu, F. Li, Y. Chen, M. Zhang, Q. Zheng, Y. Jin, and H. Ren. Mapping land- and offshore-based wind turbines in china in 2023 with sentinel-2 satellite data. Renewable and Sustainable Energy Reviews, 214: 0 115566, 2025. ISSN 1364-0321. doi:10.1016/j.rser.2025.115566. ...

    2023

  70. [110]

    Z. He, S. Xu, W. Shen, R. Long, and H. Yang. Overview of the development of the chinese jiangsu coastal wind-power industry cluster. Renewable and Sustainable Energy Reviews, 57: 0 59--71, 2016. ISSN 1364-0321. doi:https://doi.org/10.1016/j.rser.2015.12.187. URL https://www.sc...

    2016 doi

  71. [111]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber. Lstm can solve hard long time lag problems. In M. Mozer, M. Jordan, and T. Petsche, editors, Advances in Neural Information Processing Systems, volume 9. MIT Press, 1996. URL https://proceedings.neurips.cc/paper_files/paper/1996/file/a4d2f0d23...

    1996

  72. [112]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber. Long short-term memory. Neural Comput., 9 0 (8): 0 1735–1780, Nov. 1997. ISSN 0899-7667. doi:10.1162/neco.1997.9.8.1735. URL https://doi.org/10.1162/neco.1997.9.8.1735

    1997 doi

  73. [113]

    Hoeser and C

    T. Hoeser and C. Kuenzer. Global dynamics of the offshore wind energy sector monitored with sentinel-1: Turbine count, installed capacity and site specifications. International Journal of Applied Earth Observation and Geoinformation, 112: 0 102957, 2022. ISSN 1569-8432. doi:10...

    2022

  74. [114]

    Hoeser, S

    T. Hoeser, S. Feuerstein, and C. Kuenzer. Deepowt: a global offshore wind turbine data set derived with deep learning from sentinel-1 data. Earth System Science Data, 14 0 (9): 0 4251--4270, 2022. doi:10.5194/essd-14-4251-2022. URL https://essd.copernicus.org/articles/14/4251/2022/

    2022 doi

  75. [115]

    Hoeser, F

    T. Hoeser, F. Bachofer, and C. Kuenzer. Global offshore wind infrastructure: Deployment and operational dynamics from dense sentinel-1 time series. Science of Remote Sensing, 14: 0 100451, 2026. ISSN 2666-0172. doi:https://doi.org/10.1016/j.srs.2026.100451. URL https://www.sci...

    2026

  76. [116]

    Howard and S

    J. Howard and S. Ruder. Universal language model fine-tuning for text classification. In I. Gurevych and Y. Miyao, editors, Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 328--339, Melbourne, Australia, Ju...

    2018 doi

  77. [117]

    Ismail Fawaz, G

    H. Ismail Fawaz, G. Forestier, J. Weber, L. Idoumghar, and P.-A. Muller. Deep learning for time series classification: a review. Data Mining and Knowledge Discovery, 33 0 (4): 0 917--963, July 2019. ISSN 1573-756X. doi:10.1007/s10618-019-00619-1. URL https://doi.org/10.1007/s1...

    2019 doi

  78. [118]

    Lai, C.-Y

    Y.-C. Lai, C.-Y. Choi, K. Lee, I.-K. Kwon, C.-H. Lin, L. Gibson, and W.-Y. Chen. Endangered black-faced spoonbills alter migration across the yellow sea due to offshore wind farms. Ecology, 106 0 (1): 0 e4485, 2025. doi:10.1002/ecy.4485. URL https://esajournals.onlinelibrary.w...

    2025 doi

  79. [119]

    V. I. Levenshtein. Binary codes capable of correcting deletions, insertions and reversals. In Soviet Physics Doklady, volume 10, pages 707--710, 1966

    1966

  80. [120]

    Lin and Y

    S. Lin and Y. Xu. Were subsidy phase-outs premature for offshore wind power in china? Energy for Sustainable Development, 92: 0 101967, 2026. ISSN 0973-0826. doi:https://doi.org/10.1016/j.esd.2026.101967. URL https://www.sciencedirect.com/science/article/pii/S097308262600044X

    2026

  81. [121]

    L. Liu, M. Wu, J. Zhao, L. Bing, L. Zheng, S. Luan, Y. Mao, M. Xue, J. Liu, and B. Liu. Deep learning-based monitoring of offshore wind turbines in shandong sea of china and their location analysis. Journal of Cleaner Production, 434: 0 140415, 2024. ISSN 0959-6526. doi:10.101...

    2024

  82. [122]

    L. Liu, M. Wu, G. Chen, Y. Tang, X. Song, M. Zou, Y. Xu, X. Zhang, S. Wang, J. Lv, and L. Zheng. Satellite observations and deep learning unveil the rapid expansion of offshore wind turbines in china. Resources, Conservation and Recycling, 227: 0 108706, 2026. ISSN 0921-3449. ...

    2026

  83. [123]

    Loshchilov and F

    I. Loshchilov and F. Hutter. SGDR : Stochastic gradient descent with warm restarts. arXiv preprint arXiv:1608.03983, 2017. URL https://arxiv.org/abs/1608.03983

    2017 arXiv

  84. [124]

    Loshchilov and F

    I. Loshchilov and F. Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2019. URL https://arxiv.org/abs/1711.05101

    2019 arXiv

  85. [125]

    A. Miao, Y. Yuan, H. Wu, X. Ma, C. Shao, and S. Xiang. Pathway for china's provincial carbon emission peak: A case study of the jiangsu province. Energy, 298: 0 131417, 2024. ISSN 0360-5442. doi:https://doi.org/10.1016/j.energy.2024.131417. URL https://www.sciencedirect.com/sc...

    2024

  86. [126]

    P. Ouro, R. Fernandez, A. Armstrong, B. Brooks, R. R. Burton, A. Folkard, S. Ilic, B. Parkes, D. M. Schultz, T. Stallard, and F. M. Watson. Environmental impacts from large-scale offshore renewable-energy deployment. Environmental Research Letters, 19 0 (6): 0 063001, may 2024...

    2024 doi

  87. [127]

    F. S. Paolo, D. Kroodsma, J. Raynor, T. Hochberg, P. Davis, J. Cleary, L. Marsaglia, S. Orofino, C. Thomas, and P. Halpin. Satellite mapping reveals extensive industrial activity at sea. Nature, 625 0 (7993): 0 85--91, 2024. ISSN 1476-4687. doi:10.1038/s41586-023-06825-8. URL ...

    2024 doi

  88. [128]

    Y. Qu, J. K. Swales, T. Hooper, M. C. Austen, X. Wang, E. Papathanasopoulou, J. Huang, and X. Yan. Economic trade-offs in marine resource use between offshore wind farms and fisheries in scottish waters. Energy Economics, 125: 0 106811, 2023. ISSN 0140-9883. doi:https://doi.or...

    2023

  89. [129]

    Radford, K

    A. Radford, K. Narasimhan, T. Salimans, and I. Sutskever. Improving language understanding by generative pre-training. https://api.semanticscholar.org/CorpusID:49313245, 2018. Accessed: 2026-06-30

    2018

  90. [130]

    Rußwurm and M

    M. Rußwurm and M. Körner. Self-attention for raw optical satellite time series classification. ISPRS Journal of Photogrammetry and Remote Sensing, 169: 0 421--435, 2020. ISSN 0924-2716. doi:https://doi.org/10.1016/j.isprsjprs.2020.06.006. URL https://www.sciencedirect.com/scie...

    2020 doi

  91. [131]

    Schmidhuber

    J. Schmidhuber. Learning complex, extended sequences using the principle of history compression. Neural Computation, 4 0 (2): 0 234--242, 1992. doi:10.1162/neco.1992.4.2.234

    1992 doi

  92. [132]

    Schuster and K

    M. Schuster and K. Paliwal. Bidirectional recurrent neural networks. IEEE Transactions on Signal Processing, 45 0 (11): 0 2673--2681, 1997. doi:10.1109/78.650093

    1997 doi

  93. [133]

    H. S. Seung, M. Opper, and H. Sompolinsky. Query by committee. In Proceedings of the Fifth Annual Workshop on Computational Learning Theory, COLT '92, page 287–294, New York, NY, USA, 1992. Association for Computing Machinery. ISBN 089791497X. doi:10.1145/130385.130417. URL ht...

    1992

  94. [134]

    Song and Z

    X. Song and Z. Li. Seasonally robust offshore wind turbine detection in sentinel-2 imagery using imaging geometry-aware deep learning. Remote Sensing, 17 0 (14), 2025. ISSN 2072-4292. doi:10.3390/rs17142482. URL https://www.mdpi.com/2072-4292/17/14/2482

    2025 doi

  95. [135]

    Spanier, T

    R. Spanier, T. Hoeser, J. Truckenbrodt, F. Bachofer, and C. Kuenzer. Offshore oil and gas platform dynamics in the north sea, gulf of mexico, and persian gulf: exploiting the sentinel-1 archive. Big Earth Data, 0 0 (0): 0 1--27, 2026. doi:10.1080/20964471.2026.2679328. URL htt...

    2026

  96. [136]

    China installs 26 MW offshore wind turbine, a world record

    State-owned Assets Supervision and Administration Commission of the State Council . China installs 26 MW offshore wind turbine, a world record. http://en.sasac.gov.cn/2025/09/04/c_19791.htm, 2025. Accessed: 2026-06-29

    2025

  97. [137]

    J. Su, M. Ahmed, Y. Lu, S. Pan, B. Wen, and Y. Liu. RoFormer : Enhanced transformer with rotary position embedding. Neurocomputing, 568: 0 127063, 2024. doi:10.1016/j.neucom.2023.127063. URL https://arxiv.org/abs/2104.09864

    2024

  98. [138]

    Szostek, S

    C. Szostek, S. Watson, N. Trifonova, N. Beaumont, and B. Scott. Spatial conflict in offshore wind farms: Challenges and solutions for the commercial fishing industry. Energy Policy, 200: 0 114555, 2025. ISSN 0301-4215. doi:https://doi.org/10.1016/j.enpol.2025.114555. URL https...

    2025

  99. [139]

    Tseng, R

    G. Tseng, R. Cartuyvels, I. Zvonkov, M. Purohit, D. Rolnick, and H. Kerner. Lightweight, pre-trained transformers for remote sensing timeseries, 2024. URL https://arxiv.org/abs/2304.14065

    2024 arXiv

  100. [140]

    Net zero strategy: Build back greener

    UK Government . Net zero strategy: Build back greener. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1033990/net-zero-strategy-beis.pdf, 2021. last access: 12 February 202

    2021

  101. [141]

    Un climate summit 2025, 24 september 2025

    United Nations, Department of Global Communications . Un climate summit 2025, 24 september 2025. Video, UN Web TV, 2025. URL https://webtv.un.org/en/asset/k12/k12ebey0bu. Broadcast by the Department of Global Communications; Subjects: Climate Change; Sustainable Development Go...

    2025

  102. [142]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention is all you need. In Advances in Neural Information Processing Systems 30 (NeurIPS 2017), pages 5998--6008, 2017. URL https://arxiv.org/abs/1706.03762

    2017 arXiv

  103. [143]

    F. Wang, S. Zhang, Y. Hou, and J. Wang. Extraction of offshore wind turbines in china by combining multispectral and sar image data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17: 0 9266--9281, 2024 a . doi:10.1109/JSTARS.2024.3392695

    2024

  104. [144]

    K. Wang, W. Xiao, T. He, and M. Zhang. Remote sensing unveils the explosive growth of global offshore wind turbines. Renewable and Sustainable Energy Reviews, 191: 0 114186, 2024 b . ISSN 1364-0321. doi:10.1016/j.rser.2023.114186. URL https://www.sciencedirect.com/science/arti...

    2024

  105. [145]

    Wei, Q.-P

    Y. Wei, Q.-P. Zou, and X. Lin. Evolution of price policy for offshore wind energy in china: Trilemma of capacity, price and subsidy. Renewable and Sustainable Energy Reviews, 136: 0 110366, 2021. ISSN 1364-0321. doi:https://doi.org/10.1016/j.rser.2020.110366. URL https://www.s...

    2021

  106. [146]

    P. Werbos. Backpropagation through time: what it does and how to do it. Proceedings of the IEEE, 78 0 (10): 0 1550--1560, 1990. doi:10.1109/5.58337

    1990 doi

  107. [147]

    F. Xing, Y. P. Wang, W. Ni, S. Gao, J. Jia, and J. Gao. Modeling multi-decadal morphological evolution of the radial-shaped sand ridges in the southern yellow sea, china. CATENA, 238: 0 107884, 2024. ISSN 0341-8162. doi:https://doi.org/10.1016/j.catena.2024.107884. URL https:/...

    2024

  108. [148]

    W. Xu, Y. Liu, W. Wu, Y. Dong, W. Lu, Y. Liu, B. Zhao, H. Li, and R. Yang. Proliferation of offshore wind farms in the north sea and surrounding waters revealed by satellite image time series. Renewable and Sustainable Energy Reviews, 133: 0 110167, 2020. ISSN 1364-0321. doi:h...

    2020

  109. [149]

    Zhang, F

    S. Zhang, F. Wang, Y. Hou, J. Wang, and J. Guo. Global offshore wind turbine detection: a combined application of deep learning and google earth engine. International Journal of Remote Sensing, 45 0 (18): 0 6601--6623, 2024. doi:10.1080/01431161.2024.2391587

    2024

  110. [150]

    Zhang, B

    T. Zhang, B. Tian, D. Sengupta, L. Zhang, and Y. Si. Global offshore wind turbine dataset. Scientific Data, 8 0 (1): 0 191, 2021. ISSN 2052-4463. doi:10.1038/s41597-021-00982-z. URL 10.1038/s41597-021-00982-z

    2021 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.