Indonesia is among the world’s major palm oil producers, and oil palm expansion is often associated with deforestation. This study examines gradual land use and land cover (LULC) changes from 2000 to 2023 in the Barumun watershed, North Sumatra, using Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) imagery processed through Google Earth Engine. A Random Forest (RF) machine learning algorithm was applied for supervised classification. The results indicate that forest cover decreased by 5.12% over 23 years. The primary conversion pathway was forest to other land with tree cover (accounting for up to 54.60% of initial forest loss), which was later converted to oil palm plantations. Oil palm area increased by 13.91%, but direct forest-to-oil palm conversion was limited. These findings suggest that oil palm was not the primary cause of initial deforestation in this watershed. The results provide evidence for policymakers to distinguish between direct and indirect land use change pathways when designing sustainable land management strategies.