Arctic shoreline displacement with open satellite imagery and data fusion: A pilot study 1984–2022
The Arctic coast is facing rapid changes due to thawing permafrost and melting glaciers and sea ice. Communities all around the Arctic urgently need local-scale information on coastal change. This study aimed at developing a scalable and transferable procedure for mapping shoreline displacement in A...
| Main Authors: | , , |
|---|---|
| Format: | Text |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://doi.org/10.5194/egusphere-2023-1399 https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1399/ |
| Summary: | The Arctic coast is facing rapid changes due to thawing permafrost and melting glaciers and sea ice. Communities all around the Arctic urgently need local-scale information on coastal change. This study aimed at developing a scalable and transferable procedure for mapping shoreline displacement in Arctic conditions by using an archive of satellite images. Our approach utilizes cloud computing in Google Earth Engine to process a large number of open satellite images for large areas and a long period of time (here 39 years). The procedure was iteratively developed in two contrasting study areas in Arctic Norway. It applies data fusion (including sensor fusion, algorithm fusion, and decision fusion) to improve classification accuracy and processing efficiency. For one 2 500 km 2 area of interest, the procedure utilizes c. 600 satellite images to create coastal land cover and shoreline time series in less than one hour. Data fusion reduces problems related to the low availability and quality of satellite data in the Arctic before 2013 and reduces the impacts of noise and short-term changes. However, low data availability tends to create local gaps in the time series. Validation in the Tanafjorden and north-western Svalbard coasts indicates an overall classification accuracy of more than 99 % (against an independent sample of 2000 coastal points) and a median shoreline error distance of less than 15 m (against manually digitized shoreline) in 2019–2022. We exemplify how the method produces new information for identifying coastal change hotspots and examining long-term trends and local scale processes. We give examples of glacier retreat, spit migration, and delta development. This procedure is scalable and transferable to any coastal area demonstrating potential for producing the first circumpolar dataset of shoreline displacement. |
|---|