the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Feb 2026
Building a dataset of offshore oil and gas extraction platforms from satellite data (2017–2023)
Abstract. Accurate information on the location and operational status of offshore oil and gas platforms (OOGPs) is important to inform decision-making by various stakeholders and to evaluate the environmental impacts of OOGPs. However, existing OOGP databases are often incomplete or outdated data. In this work, we use satellite data and the Google Earth Engine (GEE) platform to construct a new database of OOGPs for six major offshore oil and gas basins in the world between 2017 and 2023. We use synthetic aperture radar (SAR) images from the Sentinel-1 satellite mission to detect OOGP candidates due to its high sensitivity to OOGPs, dense spatio-temporal sampling, and global coverage. Our main processing steps comprise the detection of OOGP candidates using monthly averages of SAR images and the removal of noise and false positive objects from annual image composites. With the resulting dataset of OOGPs, we map the spatiotemporal distribution of OOGPs in the study regions and analyze platform status after the post-processing of the platform targets. Using these methods, we identified a total of 5,358 OOGPs distributed in six offshore basins: the Gulf of Mexico (GoM) (1,593), Persian Gulf (PG) (1,437), North Sea (440), Caspian Sea (CS) (794), Gulf of Guinea (460), and Gulf of Thailand (634). An independent validation dataset was used to evaluate the performance of the detection algorithm, which achieved an extraction accuracy of 98 %. This OOGPs dataset substantially enhances and complements the existing offshore platform database in terms of spatial and temporal coverage. From our analysis of this OOGP dataset, we observed that offshore platform activity has declined in regions like the GoM due to infrastructure aging and policy shifts, while it has expanded in the PG and CS, reflecting ongoing offshore development. These different regional trends highlight the need for targeted environmental oversight and region-specific mitigation strategies.
- Preprint
(18141 KB) - Metadata XML
-
Supplement
(49343 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2026-63', Anonymous Referee #1, 07 Apr 2026
-
RC2: 'Comment on essd-2026-63', Anonymous Referee #2, 14 Apr 2026
This manuscript presents a satellite-based framework for detecting offshore oil and gas platforms (OOGPs) using multi-source remote sensing data, primarily Sentinel-1 SAR imagery, and constructs a global dataset covering six major offshore basins from 2017 to 2023. The study addresses an important gap in existing infrastructure inventories by providing a more temporally consistent and spatially comprehensive dataset, with potential applications in environmental monitoring and methane emission attribution. However, in its current form, the manuscript still suffers from methodological ambiguity, insufficient validation rigor, and limited discussion of uncertainty and generalizability. Recommendation: Major RevisionMajor Comments:1. The manuscript claims to develop a “robust” and “automated” framework, yet the methodological components (e.g., thresholding, morphological filtering, occurrence frequency filtering) largely rely on established techniques. What is the clear methodological advancement beyond existing SAR-based OOGP detection studies? How does this approach quantitatively outperform prior methods (e.g., deep learning-based or multi-sensor fusion approaches)?2. The thresholding strategy, including the use of a fixed backscatter cutoff and a percentile-based adaptive threshold, lacks sufficient theoretical justification. It also remains unclear how sensitive the results are to these parameter choices.3. The regional variability of Occurrence Frequency (OF) thresholds (e.g., OF=12 for the English Channel vs. OF ≥ 2 for the GoG) is a pragmatic solution to maritime traffic noise. However, the manuscript lacks a quantitative justification for these specific values. Providing sensitivity analysis results for these thresholds would strengthen the methodological rigor.4. The use of the Mann–Kendall test to infer installation dates from SAR backscatter time series is an interesting approach, but its reliability is uncertain given the limited temporal coverage of Sentinel-1 data. Abrupt changes in backscatter may not uniquely correspond to construction events. The manuscript would benefit from additional validation or uncertainty analysis for these inferred temporal attributes.Minor Comments:
1. In the Abstract, "existing OOGP databases are often incomplete or outdated data" should be corrected to "incomplete or contain outdated data".2. The abstract and conclusions report slightly different accuracy figures (“98 %” vs. “0.99”). Aligning all statements with the precise metrics in Table 3 (precision = 0.98, recall = 0.92, F1 = 0.95) would eliminate any confusion.3. P2, lines 25–30: “these aging platforms typically have three potential decommissioning options…” is wordy; tighten.4. Terminology is generally consistent, but “OOGP(s)” and “offshore platforms” are occasionally mixed within the same paragraph. Using the acronym uniformly after its first definition would enhance precision.5. Figure 9’s color legend (red points) could be slightly deepened for better readability in print or PDF format.6. Table 2 would benefit from a brief note on the exact date format (YYYYMM) and how null values for installation/removal dates are handled.Citation: https://doi.org/10.5194/essd-2026-63-RC2
Data sets
The Offshore Oil and Gas Platforms (OOGPs) dataset based on satellite data spanning 2017 to 2023 Lulu Si, Shanyu Zhou, Itziar Irakulis-Loitxate, Javier Roger, Luis Guanter https://doi.org/10.5281/zenodo.18350974
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 434 | 1,351 | 40 | 1,825 | 237 | 34 | 47 |
- HTML: 434
- PDF: 1,351
- XML: 40
- Total: 1,825
- Supplement: 237
- BibTeX: 34
- EndNote: 47
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Thanks for this work, try to develop a robust algorithm to generate global offshore oil and gas datasets from satellite observations. The methodology and results look reliable and reasonable. I have several comments for your reference.
1. Fig. 3, why do you only use Sentinel-2 satellite imagers during Jan-Mar for OOGP generation?
2. L155-160, the FIRMS with 1-km spatial resolution, Near-real time frequency was used to identify platform activity status. This is an interesting point and skillful solution. The key challenge is how to address the discrepancy in spatiotemporal resolutions between this fire data and Sentinel-1/2 data (with 10-60 m resolution). Please add more details on this point. For example, how is proximity defined in the proposed approach?
3. L160, as mentioned above, FIRMS has a different spatiotemporal resolution from the primary satellite data (Sentinel-1/2). Please clarify this point.
4. Eq.(1), although two local examples are given to demonstrate the rationale regarding the BCmax threshold of -20db, an adaptive threshold like Eq. (2) is more reliable.
5. Fig. 10, how can we control the effect of the uncertainty in satellite data and algorithm (like threshold selection) on these temporal variations?