Mapping of temperate upland habitats using high-resolution satellite imagery and machine learning

• DOI: 10.1007/s10661-024-12998-0
• Publication Date: 2024-08-31T08:02:22Z
• Authors: Charmaine Cruz; Philip M. Perrin; James R. Martin; Jerome O’Connell; Kevin McGuinness; John Connolly
• Keywords: Satellite Imagery; Machine Learning; Conservation of Natural Resources; Research; Remote Sensing Technology; 0401 agriculture, forestry, and fisheries; 04 agricultural and veterinary sciences; Ireland; Ecosystem; Environmental Monitoring

Abstract Upland habitats provide vital ecological services, yet they are highly threatened by natural and anthropogenic stressors. Monitoring these vulnerable is fundamental for conservation involves determining information about their spatial locations conditions. Remote sensing has evolved as a promising tool to map the distribution of upland in space time. However, resolutions most freely available satellite images (e.g., 10-m resolution Sentinel-2) may not be sufficient mapping relatively small features, especially heterogeneous landscape—in terms habitat composition—of uplands. Moreover, use traditional remote methods, imposing discrete boundaries between habitats, accurately represent often occur mosaics merge with each other. In this context, we used high-resolution (2 m) Pleiades imagery Random Forest (RF) machine learning at two Irish sites. Specifically, investigated impact varying on classification accuracy proposed complementary approach methods complex habitats. Results showed that generally improved finer data, highest values (80.34% 79.64%) achieved both sites using 2-m datasets. The probability maps derived from RF-based fuzzy technique can gradual transitions occurring presented potentially enhance our understanding spatiotemporal dynamics over large areas.