Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network
Abstract Straits are strategically and economically vital due to their role as maritime choke points, controlling access to regions and resources. This is particularly pertinent in the Arctic, where navigability along critical shipping routes relies on access through straits that are frequently ice...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-03-01
|
| Series: | Journal of Geophysical Research: Machine Learning and Computation |
| Subjects: | |
| Online Access: | https://doi.org/10.1029/2024JH000281 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849387755837063168 |
|---|---|
| author | M. L. Rocha A. H. Lynch K. J. Bergen |
| author_facet | M. L. Rocha A. H. Lynch K. J. Bergen |
| author_sort | M. L. Rocha |
| collection | DOAJ |
| description | Abstract Straits are strategically and economically vital due to their role as maritime choke points, controlling access to regions and resources. This is particularly pertinent in the Arctic, where navigability along critical shipping routes relies on access through straits that are frequently ice impacted. With the retreat of Arctic sea ice under anthropogenic climate change, scenarios using CMIP6 projections have the potential to provide valuable insights into future maritime accessibility regimes. However, typical climate model spatial resolutions limit the capacity to represent Arctic straits accurately. This study introduces a novel approach, the sea Ice Concentration Enhancement Generative Adversarial Network (ICE‐GAN), to enhance the spatial resolution of sea ice concentration (SIC) in Vilkitsky Strait, a passage along the Northern Sea Route (NSR). By employing the ICE‐GAN model, the spatial resolution is functionally increased from 1.00° to 0.25°. The approach is prototyped using ERA5 Reanalysis training data to predict ice cover for 2021 and 2022. The results indicate that the ICE‐GAN method outperforms, across multiple metrics, standard interpolation techniques such as Nearest Neighbor Interpolation and Bilinear Interpolation, both used in maritime accessibility models, as well as the super‐resolution convolutional neural network, the best practice method for super‐resolution in SIC. Importantly, the approach is robust to the non‐stationarity of the sea ice record. Moreover, by incorporating a physics‐informed approach into the ICE‐GAN, the model is able to further improve the accurate representation of sea ice cover in the studied Strait. |
| format | Article |
| id | doaj-art-df0fae6d748b4517aa78d193f711e25d |
| institution | Kabale University |
| issn | 2993-5210 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | Wiley |
| record_format | Article |
| series | Journal of Geophysical Research: Machine Learning and Computation |
| spelling | doaj-art-df0fae6d748b4517aa78d193f711e25d2025-08-20T03:44:24ZengWileyJournal of Geophysical Research: Machine Learning and Computation2993-52102025-03-0121n/an/a10.1029/2024JH000281Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial NetworkM. L. Rocha0A. H. Lynch1K. J. Bergen2Institute at Brown for Environment and Society Brown University Providence RI USAInstitute at Brown for Environment and Society Brown University Providence RI USAInstitute at Brown for Environment and Society Brown University Providence RI USAAbstract Straits are strategically and economically vital due to their role as maritime choke points, controlling access to regions and resources. This is particularly pertinent in the Arctic, where navigability along critical shipping routes relies on access through straits that are frequently ice impacted. With the retreat of Arctic sea ice under anthropogenic climate change, scenarios using CMIP6 projections have the potential to provide valuable insights into future maritime accessibility regimes. However, typical climate model spatial resolutions limit the capacity to represent Arctic straits accurately. This study introduces a novel approach, the sea Ice Concentration Enhancement Generative Adversarial Network (ICE‐GAN), to enhance the spatial resolution of sea ice concentration (SIC) in Vilkitsky Strait, a passage along the Northern Sea Route (NSR). By employing the ICE‐GAN model, the spatial resolution is functionally increased from 1.00° to 0.25°. The approach is prototyped using ERA5 Reanalysis training data to predict ice cover for 2021 and 2022. The results indicate that the ICE‐GAN method outperforms, across multiple metrics, standard interpolation techniques such as Nearest Neighbor Interpolation and Bilinear Interpolation, both used in maritime accessibility models, as well as the super‐resolution convolutional neural network, the best practice method for super‐resolution in SIC. Importantly, the approach is robust to the non‐stationarity of the sea ice record. Moreover, by incorporating a physics‐informed approach into the ICE‐GAN, the model is able to further improve the accurate representation of sea ice cover in the studied Strait.https://doi.org/10.1029/2024JH000281Northern Sea Routesea ice concentrationsuper‐resolutionmachine learning |
| spellingShingle | M. L. Rocha A. H. Lynch K. J. Bergen Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network Journal of Geophysical Research: Machine Learning and Computation Northern Sea Route sea ice concentration super‐resolution machine learning |
| title | Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network |
| title_full | Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network |
| title_fullStr | Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network |
| title_full_unstemmed | Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network |
| title_short | Enhancing Sea Ice Concentration Resolution in a Northern Sea Route Strait Using a Generative Adversarial Network |
| title_sort | enhancing sea ice concentration resolution in a northern sea route strait using a generative adversarial network |
| topic | Northern Sea Route sea ice concentration super‐resolution machine learning |
| url | https://doi.org/10.1029/2024JH000281 |
| work_keys_str_mv | AT mlrocha enhancingseaiceconcentrationresolutioninanorthernsearoutestraitusingagenerativeadversarialnetwork AT ahlynch enhancingseaiceconcentrationresolutioninanorthernsearoutestraitusingagenerativeadversarialnetwork AT kjbergen enhancingseaiceconcentrationresolutioninanorthernsearoutestraitusingagenerativeadversarialnetwork |