ARCtic Sea-ice and CleAN monitoring -

You are here

The Arctic is one of the most vulnerable regions to climate change effects, as has become apparent with the considerable reduction of sea ice coverage in last recent years. In the other hand, the reduction of sea ice brings interest in commercial activities such as oil/gas exploration and maritime transport. However, the observation of sea ice variability and the detection of oil spills in a spread zone like Arctic is not an easy work. One of the most effective solutions for this issue is the use of satellite imagery, especially Synthetic Aperture Radar (SAR) data due to their high spatial resolution, wide swath, and availability (0.5 day of revisit frequency). Although there are many methods for the detection of sea ice and oil spill from SAR images, Machine Learning (ML) is selected for this project, since it can process thousands of images with high accuracy. Among the ML methods, Support Vector Machines are used for the automatic detection of sea ice and oil spills, since they are still efficient within high dimensional vector spaces and still work when number of dimensions is greater than the number of samples. Additionally, they do not request too much computing memory.

Users and their needs

The user communities are:

  • Energy (oil and Gas)   
  • Environmental watching
  • Marine Security
  • Shipping
  • Defence

End-User Needs:

  • High resolution product
  • Quality - associated to Academic competencies (with related scientific paper)
  • Near real Time access
  • Easy access via Web Services
  • Cost effective (reduced compared to current services)

Location of the targeted users: Worldwide

Service/ system concept

The ARC-SCAN provide a supply solution for the automatic detection of sea ice and oil spill from Sentinel-1 images with high accuracy. It permits to reduce significantly the processing time of the users which is used for doing many steps (pre-processing, manual classification, validation). Based on the proposed solution, ARC-SCAN provides, over a region of interest and acquisition time, the detection of sea ice and oil spill with additional information such as coordinates (lon, lat), surface wind speed, sea surface roughness, etc.

In detail, the data/images acquired by Sentinel-1A/B (Level-1) are downloaded via the ESA-Copernicus platform. They are pre-processed and then applied automatic classification based on Machine Learning in the EXWEXs servers. The classified images are coordinated and stored in the tile servers. The tiles maybe visualized by any Web browser and especially those specialized in earth data treatment.

Once the users select a region of interest and acquisition time (via web interface or code lines), the stored data are called and displayed. The users also may zoom in or out in the zone (by using the tile display in the browser) where sea ice or oil spill is detected.

Space Added Value

The ARC-SCAN provide a supply solution for the automatic detection of sea ice and oil spill from Sentinel-1 images with high accuracy. It permits to reduce significantly the processing time of the users which is used for doing many steps (pre-processing, manual classification, validation). Based on the proposed solution, ARC-SCAN provides, over a region of interest and acquisition time, the detection of sea ice and oil spill with additional information such as coordinates (lon, lat), surface wind speed, sea surface roughness, etc.

In detail, the data/images acquired by Sentinel-1A/B (Level-1) are downloaded via the ESA-Copernicus platform. They are pre-processed and then applied automatic classification based on Machine Learning in the EXWEXs servers. The classified images are coordinated and stored in the tile servers. The tiles maybe visualized by any Web browser and especially those specialized in earth data treatment.

Current Status

This feasibility project was performed in six months. The obtained results showed that the proposed methodology corresponded to the automatic detection of sea ice and oil spill. The next step is to improve the accuracy of the classification based on Machine Learning by reducing the impact of the noise on Sentinel-1 (speckle noise and thermal noise) for the detection of sea ice, and by removing the look-alike (low wind, seaweed) for the detection of oil spill.

All the process developed in ARC-SCAN are currently moving towards a Demonstration project (funded by private sector). The aim is to automate the processes, to access data quickly and efficiently, to have a large computing power (via ESA-DIAS) and to develop an end-user web interface to let the end-users select the region of interest, acquisition time and treatment to trigger. This should lead to a fully operational service.

Prime Contractor

Project Managers

Contractor Project Manager

M. Messager
2 Rue de Keraliou
29200 Brest
France
+33 645651343

ESA Project Manager

Piera Di Vito
ESA / ESTEC
Keplerlaan 1
2200 AG Noordwijk ZH
Netherlands

Status Date

Updated: 17 January 2019 - Created: 17 January 2019