the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 May 2025
IMPMCT: a dataset of Integrated Multi-source Polar Meso-Cyclone Tracks
Abstract. Polar Mesoscale Cyclones (PMCs), particularly their intense subset known as Polar Lows (PLs), characterized by short lifespans of 3-36 hours and horizontal scales below 1,000 km, pose significant hazards to polar maritime activities due to extreme winds exceeding 15 m s⁻¹ and wave heights surpassing 11 meters. These intense weather systems play a critical role in modulating sea-ice dynamics and ocean-atmosphere heat exchange. However, current understanding remains constrained by sparse observational records and overdependence on singular data sources (e.g., remote sensing or reanalysis). To address these gaps, this study presents the Integrated Multi-source Polar Meso-Cyclone Tracks (IMPMCT) dataset, a comprehensive 24-year (2001-2024) wintertime PMCs record for the Nordic Seas. IMPMCT combines vortices tracking algorithms from ERA5 reanalysis with deep learning-based detection of cyclonic cloud features in Advanced Very High-Resolution Radiometer (AVHRR) infrared imagery, while incorporating near-surface wind matching by Advanced Scatterometry (ASCAT) and Quick Scatterometry (QUIKSCAT) measurements. The dataset contains 1,184 vortex tracks, 16,630 cyclonic cloud features, and 4,373 wind speed records, with multi-dimensional attributes such as cloud morphology, core wind speed, and environmental advection wind speed. Validation demonstrates a 70–90 % match rate with existing PLs track datasets while providing more complete cyclone life cycle trajectories, more intuitive cloud imagery visualization, and a richer set of parameters compared to previous datasets. As the most comprehensive PMCs archive for the Nordic Seas, the IMPMCT dataset provides fundamental data for advancing our understanding of the genesis and intensification mechanisms, enables the development of enhanced monitoring and early warning systems, supports the validation and refinement of polar numerical weather prediction models, and facilitates improved risk assessment and safety protocols for maritime operations. The dataset is available at https://doi.org/10.5281/zenodo.15355602 (Fang et al., 2025).
- Preprint
(10205 KB) - Metadata XML
-
Supplement
(157 KB) - BibTeX
- EndNote
Status: open (until 09 Aug 2025)
-
RC1: 'Comment on essd-2025-186', Anonymous Referee #1, 12 Jul 2025
reply
Overview:
The manuscript presents the Integrated Multi-source Polar Meso-Cyclone Tracks (IMPMCT) dataset based on both ERA5 reanalysis and remote sensing data during winter in the Nordic Sea, demonstrates clearly the workflow of this method, and compares the results with existing manually identified and reanalysis-based track datasets. There remains a clear need for establishing a more comprehensive tracking dataset capable of capturing PMCs throughout their lifecycle due to their impacts on human activities and regional climate change. The manuscript is generally well-organized, and the figures effectively communicate the results while being concise. However, there are a few aspects where the presentation could be improved. The detailed comments are listed below, and I encourage the authors to make the necessary adjustments to improve the study.
Major comments:
- The present study utilized a series of datasets, including ERA5 reanalysis, AVHRR data andQuikSCAT/ASCAT data, which have different spatial and temporal resolutions, and these data are stored with different projections/grids. How are these multi-source datasets treated in the cyclone tracking algorithm to maintain consistency? Please clarify.
- Line 164: ERA5 data. How accurate are the ERA-5 fields used in the analysis of the Nordic Sea? What are the known biases? As the authors did not repeat their method with other reanalysis datasets to test the robustness of their results, I would suggest declaring the known biases of ERA5 in this part.
- Line 262: To maximize the inclusion of potential PMCs, we implement more lenient vortex detection criteria compared to Stoll et al. (2021). The selected criteria seem to be very subjective. Importantly, how sensitive are the results to subjective criteria such as the “vorticity peak threshold”, “isolated vortex threshold”? Have the authors conducted sensitivity tests, and what metrics were used to evaluate the robustness of the results? Please include this.
- It seems a YOLO (You Only Look Once) object detection algorithm is employed to detect and extract cyclonic cloud characteristics. This description of this procedure could be improved in my opinion. The authors start by generally describing the structure of the YOLOv8-obb model on line 377, with so many acronyms. However, the specific process by which this algorithm works to detect cloud features was oversimplified in the following paragraph.
- When comparing the results from the IMPMCT to existing identified PL lists from previous studies, the authors give the difference in parameters and plot them. It is more appropriate to conduct a significance test between two samples in order to statistically validate the accuracy.
- Figure issues:
- Specify what is plotted in Figure 1 in the name of the colorbar, same comments for Figure 3b, and Figure 7.
- The green star symbols denoting the local vorticity maxima are hard to read when overlaid on the AVHRR infrared imagery. Please change the color or enlarge the symbols. Same comments for stars in Figure 10b and wind vectors in Figure 11.
- The unit of the colobar in Figure 7a should be 1e-4s-1
Minor comments:
Lines 41-42: Add references about this statement.
Lines 59-61: Add references about this statement or remove it as it seems irrelevant to the core points of this paragraph.
Lines 129-131: Moreover, fundamental questions persist regarding the differences in formation mechanisms between PMCs and PLs, and whether PMCs can transition into PLs under specific meteorological conditions. This question seems not to be addressed.
Line 138: Winter should be defined here rather than in the Data part.
Line 140: “multi-dimensional” to “multiple”
Line 161: “sourced” to “obtained”
Line 169: delete “for atmospheric, land, and ocean variables”
Lines 191- 192: Notably, QuikSCAT data spans only 1999–2009, while ASCAT has remained operational since 2010. Rephrase to: QuikSCAT operated from 1999 to 2009, whereas ASCAT has continued operations since 2010.
Lines 281-284: “Specifically, for a vortex at a given time step, its ideal point after experiencing a time step under the steering wind influence is first calculated A search radius of 180 km is then applied around this estimated location to facilitate vortex tracking in subsequent time steps..” Should be two separate sentences.
Lines 293-294: Rephrase to: If no spatially connectable vortices are identified in adjacent time steps, the vortex is classified as being terminated.
Lines 316-319: Rephrase to: Building upon the lenient vorticity identification criteria established in prior analysis, a substantial population of vortex tracks has been identified within the reanalysis dataset. This collection encompasses not only cyclonic systems but also terrain-induced shear flows, low-pressure troughs, and small-scale atmospheric disturbances.
Line 373: Delete “deliberately”
Lines 391-393: Rephrase to: To ensure prediction stability, particular emphasis is placed on maintaining consistent oriented bounding box annotations and center point positions across similar evolutionary phases of cyclonic cloud morphologies.
Linee 409-413: Rephrase to: To remove duplicate records, we implement a selection criterion: for any cluster of detections from the same AVHRR infrared scan (with cyclone centers <50 km apart), only the detection whose center is nearest to the VCI image center is retained.
Line 436: Delete “frequently”
Lines 453-455: Rephrase to: To reduce the influence of strong winds in the cyclone core, we use the 75th percentile of wind speeds within the extended search radius as the environmental advection speed (reference value).
Lines 484-485: Rephrase to: All reference datasets are spatially and temporally co-located with our derived tracks, retaining only those persisting for ≥3 hours.
Line 526: “extraneous” to “irrelevant”
Line 545: Rephrase to: Additionally, since the dataset includes remote sensing images of cyclones, users can easily verify the accuracy of cyclone properties and make necessary adjustments based on their specific use cases.
Line 568: “these categories” to “them”
Citation: https://doi.org/10.5194/essd-2025-186-RC1
Data sets
IMPMCT: a dataset of Integrated Multi-source Polar Meso-Cyclone Tracks Runzhuo Fang and Jinfeng Ding https://doi.org/10.5281/zenodo.15355602
validation dataset for yolov8-obb-pose cyclone-detect-model [Data set] Runzhuo Fang https://doi.org/10.5281/zenodo.15119534
Model code and software
IMPMCT Runzhuo Fang https://github.com/thebluewind/IMPMCT
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 230 | 26 | 13 | 269 | 12 | 12 | 20 |
- HTML: 230
- PDF: 26
- XML: 13
- Total: 269
- Supplement: 12
- BibTeX: 12
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1