Efficient large-scale vegetation mapping at the formation level using multi-source data: A case study in Beijing, China

Formation-level vegetation mapping is pivotal for understanding ecological processes and mechanisms, as it reveals the distribution of dominant species that shape ecosystem structure and dynamics. However, fast and accurate formation-level mapping over large geographic areas is often hindered by the lack of robust mapping frameworks, limited field survey data, and unscientific or inefficient division of vegetation patches. To address these challenges, we proposed an automated mapping framework that integrates multi-source data for formation-level vegetation mapping. Our approach introduced an innovative strategy for automatically delineating vegetation patches based on slope units, improving mapping efficiency and ensuring results align more closely with actual vegetation distribution. Additionally, we developed a crowdsource-based vegetation survey system that aggregates data from diverse sensors, significantly increasing the sample size and diversity of vegetation formations. Using this framework, we successfully mapped 16 formations in Beijing with an overall accuracy of 65.7%, achieving F-scores exceeding 60% for major formations. The result indicates that Beijing’s vegetation is dominated by forests and shrublands, with the largest vegetation formation being Vitex negundo (deciduous broadleaf shrubland), covering 20% of the city in the southwestern mountains, followed by Quercus mongolica (deciduous broadleaf forest), occupying 10% in the northwestern mountains. This study provides a solid foundation for understanding Beijing’s vegetation distribution and its ecological functions. By integrating remote sensing and crowdsourced data, it demonstrates an effective approach for precise, large-scale formation-level vegetation mapping, offering valuable support for refined ecological management and interdisciplinary research.