Spatiotemporal analysis of oil palm plantation expansion using a cloud computing-based random forest algorithm

Oil palm expansion has been widespread in tropical peatland ecosystems in Southeast Asia. This study assesses land-cover change and associated fire risk in peat hydrological units in Ogan Komering Ilir District, South Sumatra, Indonesia, from 2013 to 2025. A cloud-based Random Forest framework integrating multi-sensor optical and radar imagery on the Google Earth Engine platform produced annual land-cover maps with high accuracy for 2025 (overall accuracy: 91.46%; κ: 0.87). Oil palm plantations expanded from 6,707.07 ha in 2013 to 30,332.59 ha in 2025 (net increase: 23,625.51 ha). The event-based trajectory analysis developed in this study showed that expansion was dominated by the Forest → Bareland → Oil Palm pathway (73.11% of the expansion area over 8.8 years). Most expansion occurs on shallow peat, but 3126.47 ha are located on deep peat (≥3 m). Fire risk is linked to El Niño and oil palm expansion. This suggests that enforcement needs to be strengthened to protect the peatland.