A Six-year circum-Antarctic icebergs dataset (2018-2023)

What's this? This is a dataset providing annual iceberg vector distributions from October 2018 to October 2023, extracted from Sentinel-1 SAR imagery covering the region south of 55°S. It spans a continuous six-year time series (2018–2023) with a minimum detectable iceberg area of approximately 0.04 km². The dataset includes vector outlines of icebergs, geometric parameters (area, perimeter, major and minor axes), geographic coordinates, mass estimates based on average density, and associated uncertainties in both area and mass. A semi-automated machine learning method (incremental random forest classification) was used to achieve high-precision iceberg detection on large-scale SAR imagery, supplemented by manual interactive correction steps to minimize misclassifications and omissions. Compared to existing low-resolution or regionally focused iceberg products, this dataset offers broader spatial coverage (encompassing the entire Southern Ocean) and an annual temporal resolution, thereby providing essential support for research on the impacts of Antarctic icebergs on the global climate system and the Southern Ocean environment. This is a "living" dataset, meaning it will be continuously updated and expanded as new data becomes available. Data Processing Methods Summary Data Acquisition:Sentinel-1 SAR (Extra Wide mode, HH polarization) images from October of each year between 2018 and 2023 are acquired on the Google Earth Engine (GEE) platform. The region south of 55°S is divided into several 5°×5° tiles, which are then mosaicked Image Segmentation:After denoising with the Total Variation (TV) algorithm, the SAR images are segmented using the SLIC superpixel method to obtain superpixels with relatively consistent backscatter characteristics. A superpixel is defined as a small, contiguous cluster of adjacent pixels that share similar backscatter characteristics, effectively representing a meaningful image region rather than individual pixels. Iceberg Detection:For each superpixel, statistical features, histogram features, and texture features (a total of 24 features) are extracted. An incremental random forest classifier is then used for initial classification, and further post-processing and correction are performed interactively in ArcMap 10.8. Attribute Calculation:For the final detected iceberg vector outlines, we calculate their areas, perimeters, and the lengths of the major and minor axes, and quantify the uncertainty in iceberg area. In addition, assuming a fixed ice density, we apply the Iceberg Classes Model to estimate iceberg mass together with its associated uncertainty bounds. Iceberg Vector Outline Attributes Attributes Description Unit lon Longitude of iceberg centroid (WGS84) degrees (°E) lat Latitude of iceberg centroid (WGS84) degrees (°S) area_km2 Planimetric area of the iceberg polygon (EPSG:3031) km² area_uncertainty_km2 The uncertainty of iceberg area is estimated as follows: for individual icebergs, it is derived from the product of perimeter and pixel size, whereas for the total area, it is obtained as the square root of the sum of squared terms from the total perimeter × pixel size and the proportional error of the total area. km² perimeter_km Perimeter length of the iceberg polygon (EPSG:3031) km long_axis_km Length of the major axis of the minimum bounding rectangle km short_axis_km Length of the minor axis of the minimum bounding rectangle km mass_gt Estimated iceberg mass using a two-segment area–volume scheme under a fixed density assumption Gt mass_uncertainty_gt Half-width of the 95% confidence interval for iceberg mass Gt Data Applications and Significance Provides critical foundational data for studying the distribution patterns of Antarctic icebergs, their impact on the Southern Ocean environment, and their relationship with global climate change. This dataset can serve as initial conditions for iceberg dynamics models or coupled ocean-iceberg models, and be used to investigate the effects of iceberg melting on ocean circulation, temperature-salinity structures, and ecosystems.