Methodology:

The floe-scale fragmentation data were derived from high-resolution satellite imagery provided by the USGS Global Fiducials Library (GFL). The original images were cropped to the individual floe scale so that each floe could be clearly identified. These GFL images were acquired by tracking drifting buoys, allowing us to identify the same floe across a sequence of images. In this way, we could observe the floe before and after the fragmentation event. Identifying the same floe was done manually by searching through the images based on floe and melt pond patterns.

Once the floe-scale images were created, we used OpenCV drawing tool in CVAT AI (www.cvat.ai) to delineate the boundaries of the floes. The results were exported as segmentation masks, which were analysed by a MATLAB code to calculate floe parameters including the size, area, perimeter, shape factor and circularity of the floes.

The data are organised according to the buoys that the images were tracking - IABP123530, MIZ Cluster 3, MIZ Cluster 4, and MIZ Cluster 2.

IABP123530 is the buoy ID used in the USGS GFL data portal. This corresponds to IABP Buoy ID 300025010123530 on the IABP website. This buoy is co-located with CRREL IMB ID 2012L.

MIZ Cluster 3 is the second northernmost cluster of drifting buoys deployed during the ONR MIZ program. The co-located buoys are WHOI ITP (ID: ITP78), AWS 3, and IMB 4 (see ONR MIZ website).

MIZ Cluster 4 is the northernmost cluster of drifting buoys deployed during the ONR MIZ program. The co-located buoys are WHOI ITP (ID: ITP79), AWS 4, and IMB 9 (see ONR MIZ website).

MIZ Cluster 2 is the second southernmost cluster of drifting buoys deployed during the ONR MIZ program. The co-located buoys are WHOI ITP (ID: ITP77), AWS 2, and IMB 17 (see ONR MIZ website).

The spatial coverage of the dataset covers the Beaufort Sea. Latitude and longitude coordinates for each floe are provided in CSV files in the floe_parameter folder.

Data collection:

Characteristics of satellite imagery data used to derive the FSD data are described below.

USGS GGFL: 1-m resolution visible (panchromatic) satellite imagery from the USGS Global Fiducials Library (GFL) (www.glf.usgs.gov), declassified by the MEDEA group.

Data quality:

The data quality was checked by manually inspecting each image. It should be noted that some of the fragments in the post-fragmentation images could not be identified. Some were too small, while others were difficult to distinguish based on floe and melt pond patterns. The delineation of the floe boundaries was performed using the OpenCV tool (CVAT AI). The delineation was relatively accurate for large floes, but the accuracy decreased for smaller floes.