This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.
Downloads
Authors
Abstract
Forests are in decline worldwide due to human activities such as agricultural expansion, urbanization, and mineral extraction. Forest loss due to generally temporary causes, such as wildfire and forest management, is important to distinguish from permanent land use conversion due to the differing ecological and climate impacts of these disturbances and for the purposes of developing effective policies and management strategies. Existing global maps of the drivers of forest loss that are widely used are not spatially or thematically detailed enough for decision makers at local-to-regional scales, such as governments, land managers, or companies. Using publicly available satellite observations (Landsat, Sentinel) and ancillary biophysical and population data, we developed a 1 km resolution, global map of the dominant drivers of forest loss from 2001 to 2022 with seven classes: permanent agriculture (e.g., commodity crops or pasture), hard commodities (e.g., mining), shifting cultivation, forest management (e.g., logging or wood fiber plantations), settlements and infrastructure, wildfire, and other natural disturbances. We interpreted nearly 7,000 reference samples to train a global neural network model that classifies the driver of forest loss with an overall accuracy of 90.5%. Our results show that permanent agriculture was the leading driver of forest loss globally, representing 35% of loss from 2001 to 2022. The drivers of forest loss vary by region, with the leading driver identified as forest management in Europe, permanent agriculture across the tropics, and wildfire in Russia, the Asian mainland, North America, and Oceania. Our results enable assessment of forest disturbance dynamics from local to global scales and can support tracking progress towards corporate and governmental zero-deforestation commitments, monitoring deforestation risks within jurisdictions and supply chains, and assessment of global biodiversity targets.
DOI
https://doi.org/10.31223/X5HQ6K
Subjects
Computer Sciences, Environmental Sciences, Geographic Information Sciences, Remote Sensing
Keywords
Drivers of forest loss, Deep learning, remote sensing, machine learning, deforestation, agriculture, Forest disturbance attribution
Dates
Published: 2024-12-20 06:07
Last Updated: 2024-12-20 14:07
License
CC BY Attribution 4.0 International
Additional Metadata
Data Availability (Reason not available):
Data will be available upon publication.
There are no comments or no comments have been made public for this article.