A STAC Extension for discovering and cataloguing 3D city models

Abstract. National-scale 3D city model datasets are growing rapidly—Japan’s PLATEAU project alone publishes over 170 000 files, and the Netherlands’ 3D BAG covers all 10+ million buildings in the country—yet no standardised mechanism exists for discovering and cataloguing these datasets across repositories and data infrastructures. Essential properties such as coordinate reference systems, levels of detail, and city object types remain embedded inside data files in diverse formats (e.g. CityGML, CityJSON, or FlatCityBuf), invisible to search engines and catalogue services. This paper presents three contributions to address this gap. First, we define the STAC 3D City Models Extension, a formal extension to the SpatioTemporal Asset Catalog (STAC) specification that adds metadata fields for levels of detail, city object types, semantic surfaces, textures, materials, and attribute schemas. Second, we develop city3dstac, an open-source command-line tool written in Rust that automatically extracts metadata from 3D city models in different formats and generates standards-conformant STAC metadata. Third, we construct a prototype registry that currently contains 53 STAC Collections, of which 31 are fully indexed, totalling 14 766 STAC Items. The indexed Collections include the national-scale 3D BAG, American Cities, Estonia, and PLATEAU datasets. Existing STAC clients can be used for basic browsing of the catalogue without modification, while extension-aware clients can support richer 3D-specific filtering.