A dataset of 512x512 tundra lakes imagery and binary masks from Sentinel-1 in the Yamal and Alaska areas, summer, 2015-2022

Data are available at: arcticdata.io/data/10.18739/A2N29P78F Permafrost tundra contains more than twice as much carbon as is currently in the atmosphere and is warming six times as fast as the global mean. Tundra lakes dynamics is a robust indicator of Global climate processes and still not well understood. Satellite data, particularly, from synthetic aperture radar (SAR) are a great source for tundra lakes recognition and their changes monitoring. However, manual analysis of their boundaries can be slow and inefficient, therefore reliable automated algorithms are required. This dataset aimed to fill the gap of the ground truth satellite images for algorithms training and validation and contains synthetic aperture radar imagery of tundra lakes from Sentonel-1 complemented with manually labeled masks of the lakes. The dataset covers two test sites in Yamal and Alaska areas for the summer months of 2015-2022. The images are generated for machine learning algorithms with a spatial resolution of 512x512 pixels.

数据可获取于：arcticdata.io/data/10.18739/A2N29P78F。永久冻土苔原（Permafrost tundra）储存的碳量是当前大气碳储量的两倍以上，且其升温速率为全球平均水平的六倍。苔原湖泊动态变化是全球气候过程的可靠指示因子，但目前相关机制仍未得到充分阐释。卫星数据，尤其是合成孔径雷达（Synthetic Aperture Radar, SAR）数据，是识别苔原湖泊并监测其变化的优质数据源。然而，人工解析湖泊边界的过程既耗时且效率低下，因此亟需可靠的自动化算法。本数据集旨在填补算法训练与验证所需的真实标注（Ground Truth）卫星图像空白，包含源自哨兵一号（Sentinel-1）的苔原湖泊合成孔径雷达影像，并附带人工标注的湖泊掩膜。数据集涵盖亚马尔（Yamal）与阿拉斯加（Alaska）两处试验站点，包含2015年至2022年夏季时段的影像数据。所有影像均适配机器学习算法，空间分辨率为512×512像素。