Pickens et al. (2025) Rapid monitoring of global land change

Identification

• Journal: Nature Communications
• Year: 2025
• Date: 2025-10-08
• Authors: Amy Pickens, Matthew C. Hansen, Zhen Song, Andrew Poulson, Anna Komarova, Antoine Baggett, Theodore Kerr, Aleksandra Mikus, Caroline Domínguez, Alexandra Tyukavina, Annamaria Lima
• DOI: 10.1038/s41467-025-64014-9

Research Groups

• Department of Geographical Sciences, University of Maryland, College Park, MD, USA.

Short Summary

This paper introduces DIST-ALERT, a global land change monitoring system that rapidly tracks vegetation loss and generic land cover anomalies at 30 m resolution using Harmonized Landsat Sentinel-2 imagery. In 2023, it identified 28.6 ± 7.6 Mha of anthropogenic land use conversions and 14.9 ± 4.3 Mha of conversion fires, demonstrating high accuracy and low latency for detecting diverse land changes across all ecosystems.

Objective

• To present and validate DIST-ALERT, an operational global land change monitoring system that rapidly tracks vegetation loss anomalies and generic land cover changes at 30 m resolution across all ecosystems.

Study Configuration

• Spatial Scale: Global, 30 m pixel resolution.
• Temporal Scale: Near-real-time monitoring with a 1–4 day revisit rate (average 2–4 days in tropics, every other day in temperate, daily in boreal during sunlit periods). Alerts generated since January 2023, with operational production since March 2024. Anomaly detection uses a rolling baseline period of the previous three years within a 31-day calendar window (±15 days).

Methodology and Data

• Models used:
  • DIST-ALERT system: Comprises two primary algorithms:
    1. Vegetation Loss Algorithm: Maps vegetation cover for every observation and flags vegetation cover loss relative to a rolling baseline minimum. Employs a K-Nearest Neighbor (KNN) model (K=100) trained on unoccupied aerial vehicle (UAV) data for vegetation fraction estimation.
    2. Generic Spectral Anomaly Algorithm: Identifies land cover changes with reflectance outside the typical spectral reflectance envelope using Mahalanobis distance from the mean of baseline observations.
  • DIST-ANN: An annual summary product derived from DIST-ALERT.
• Data sources:
  • Satellite Imagery: Harmonized Landsat Sentinel-2 (HLS) dataset (Landsat 8/9 and Sentinel-2A/B/C) for primary monitoring. HLS data includes per-pixel quality flags (Fmask 4.2) to filter out cloud, cloud buffer, cloud shadow, and snow/ice contaminated observations. Red, NIR, SWIR1.6, and SWIR2.1 bands were used.
  • Training Data: 232 UAV flights (8–10 cm resolution, MicaSense RedEdge-MX camera) collected across 6 biomes and 24 ecoregions for training the vegetation fraction model.
  • Reference Data: A probability sample of 30 m pixels labeled using HLS, monthly true color PlanetScope composites, and very high-resolution Google Earth images for validation and area estimation.

Main Results

• 2023 Land Cover Change:
  • Anthropogenic land use conversions totaled 28.6 ± 7.6 Mha (± standard error), with 15.7 ± 6.0 Mha replacing long-lived or secondary natural vegetation.
  • Conversion fires resulted in 14.9 ± 4.3 Mha of land cover change.
  • Combined, these dynamics represent 0.3% of the global land surface.
  • Dominant drivers of land use expansion into natural vegetation included agriculture expansion, logging, shifting cultivation, mining, and built-up expansion.
  • Natural variations (e.g., climate-driven drought, green-up events) accounted for 1371 ± 280 Mha, representing over two-thirds of all identified change.
  • Human land management, primarily crop cycle changes, contributed 280 ± 27 Mha.
• Accuracy of Vegetation Loss Alerts (high confidence, high loss ≥50%):
  • User's accuracy (precision): 94.2 ± 2.5%.
  • Producer's accuracy (recall): 93.9 ± 5.6%.
  • For stable tree cover, these figures were 100.0% precision and 94.3 ± 5.2% recall.
• Accuracy of all confirmed alerts with vegetation cover loss ≥10%:
  • Precision: 90.6 ± 3.4%.
  • Recall: 62.1 ± 5.4% (for provisional and confirmed alerts).
• Accuracy of Generic Change Alerts (provisional and confirmed):
  • Precision: 66.6 ± 5.2%.
  • Recall: 23.8 ± 6.0%.
  • For high magnitude change, precision was 69.5 ± 9.2% and recall was 61.0 ± 9.1%.
  • The generic change algorithm captured most change conversion without vegetation loss (e.g., building demolition) with a recall of 96.9 ± 3.8%.
• Latency:
  • Average lag from satellite imaging to public release: 2–3 days.
  • Mean detection lag for high loss events (≥50%): 9.9 days (±0.01) with reference date adjustment, and 6.1 days (±0.01) without adjustment.
  • Precision and recall for high magnitude events reached 90% by 15 days and stabilized at approximately 94% after 30 days.

Contributions

• First Operational Global High-Resolution Monitoring System: Introduces DIST-ALERT, the first operational global land change monitoring system providing rapid (1–4 day revisit) and high-resolution (30 m) tracking of vegetation loss and generic land cover anomalies across all ecosystems.
• Comprehensive and Actionable Change Detection: Offers a broad spectrum of detected land dynamics, including anthropogenic conversions (agriculture, urbanization, logging, mining) and natural events (fire, drought, landslides), providing actionable information for conservation, land management, and enforcement efforts related to international agreements.
• Foundation for Long-Term Environmental Data Record: The annual summaries (DIST-ANN) serve as a valuable long-term global environmental data record, supporting climate modeling, trend analysis, and impact assessment.
• Robust Methodology: Employs a novel methodology combining a vegetation fraction loss algorithm and a spectral anomaly algorithm, utilizing a rolling baseline and time-series tracking to enhance confidence and contextual information for detected changes.

Funding

• NASA Jet Propulsion Laboratory contract #1669907
• Bezos Earth Fund through the World Resources Institute Global LCLU Monitoring Program (#G2436)
• NASA Observational Products for End-Users from Remote Sensing Analysis (OPERA) project, managed by the NASA Jet Propulsion Laboratory and funded by the Satellite Needs Working Group.

Citation

@article{Pickens2025Rapid,
  author = {Pickens, Amy and Hansen, Matthew C. and Song, Zhen and Poulson, Andrew and Komarova, Anna and Baggett, Antoine and Kerr, Theodore and Mikus, Aleksandra and Domínguez, Caroline and Tyukavina, Alexandra and Lima, Annamaria},
  title = {Rapid monitoring of global land change},
  journal = {Nature Communications},
  year = {2025},
  doi = {10.1038/s41467-025-64014-9},
  url = {https://doi.org/10.1038/s41467-025-64014-9}
}