GLOBAL MESSAGE PASSING IN NETWORKS VIA TASK-DRIVEN RANDOM WALKS FOR SEMANTIC SEGMENTATION OF REMOTE SENSING IMAGES

We introduce GlobalBuildingAtlas, a publicly available dataset offering global and complete coverage of building polygons (GBA.Polygon), heights (GBA.Height) and Level of Detail 1 3D models (GBA.LoD1). This is the first open dataset to offer high quality, consistent, and complete building data in 2D and 3D at the individual building level on a global scale. With more than 2.75 billion buildings worldwide, it surpasses the most comprehensive database to date by more than 1 billion buildings.

ChatEarthNet is an image–text dataset that provides high-quality, detailed natural language descriptions for global-scale satellite data. It consists of 163 488 image-text pairs with captions generated by ChatGPT-3.5 and an additional 10 000 image-text pairs with captions generated by ChatGPT-4V(ision). This dataset has significant potential for training and evaluating vision–language geo-foundation models in remote sensing.

Glacier ice thickness is difficult to measure directly but is essential for glacier evolution modelling. In this work, we employ a novel approach combining physical knowledge and data-driven machine learning to estimate the ice thickness of multiple glaciers in Spitsbergen, Barentsøya, and Edgeøya in Svalbard. We identify challenges for the physics-aware machine learning model and opportunities for improving the accuracy and physical consistency that would also apply to other geophysical tasks.

Comprehensive datasets of calving-front changes are essential for studying and modeling outlet glaciers. Current records are limited in temporal resolution due to manual delineation. We use deep learning to automatically delineate calving fronts for 23 glaciers in Greenland. Resulting time series resolve long-term, seasonal, and subseasonal patterns. We discuss the implications of our results and provide the cryosphere community with a data product and an implementation of our processing system.