By combining Sentinel-1 data, GF-3 satellite SAR, and SDGSAT-1 thermal infrared data, multiple time interval observations of the sea ice edge zone were obtained. The time intervals ranged from a few minutes to several tens of hours, covering key areas of the Arctic Kara Sea, Beaufort Sea, and Greenland Sea from August 2021 to August 2022 and from October to December 2023. The grid size was 10km * 10km. Using object-oriented segmentation and thresholding methods to obtain ice water classification maps for Sentinel-1 and Sentinel-1 image pairs, Sentinel-1 and GF-3 image pairs, and Sentinel-1 and SDGSAT-1 TIR image pairs, with time intervals ranging from 1 minute to 72 hours. The spatial resolution of GF-3 SAR data is 25m, SDGSAT-1 data is 30m, and Sentinel-1 data is 40m on a 10km grid sample.
Data naming convention: region_time interval time_time grid numbering satellite resolution. tif
Attribute information:
0: Water
1: Ice
The Sentinel-1 data is sourced from publicly available satellite imagery data, and the download link is as follows: Sentinel-1：Alaska Satellite Facility (ASF)（ https://search.asf.alaska.edu/# ）
The data from GF-3 and SDGSAT-1 satellites are domestic satellite data (the relevant data has been received through the Chinese remote sensing satellite ground station)
GF-3： https://logindataservices.ceode.ac.cn/cas/login?service=http://ids.ceode.ac.cn/gfds/gflogin
SDGSAT-1： http://124.16.184.48:6008/query
1. Data preprocessing: Perform radiometric calibration, in orbit stitching, and other preprocessing on GF-3 L2 SAR data; The processing of Sentinel-1 EW mode data includes radiometric calibration, incident angle correction, and reprojection to generate geographic data that matches GF-3 data; The preprocessing of SDGSAT-1 TIS data includes radiometric calibration, temperature conversion, and reprojection.
2. Image matching cropping: Crop the overlapping areas of Sentinel-1, GF-3 SAR, and SDGSAT-1 TIS preprocessed data to form sample pairs of Sentinel-1 and GF-3, Sentinel-1 and Sentinel-1, and Sentinel-1 and SDGSAT-1 with the same coverage range within a certain time interval.
3. Image classification: Use object-oriented and threshold segmentation methods to classify the ice water images in these areas, assign water value to 0 and ice value to 1, generate 10 km × 10 km grid binary classification sample pairs of ice water, and obtain the final dataset.
The overall accuracy of Sentinel-1 classification results is 98.20%, with a kappa coefficient of 0.96; The overall accuracy of GF-3 classification results is 95.58%, with a kappa coefficient of 0.89; The overall accuracy of SDGSAT-1 classification results is 84%, with a kappa coefficient of 0.68.
| # | number | name | type |
| 1 | 2022YFF0711700 | National key R & D plan |
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.NIEER.DB6496.2024
10.12072/ncdc.nieer.db6496.2024