Analysis of forest-plantation dynamics using multi-temporal optical and SAR satellite data in West Kalimantan, Indonesia.

This study aims to explore the dynamics of forest and oil palm plantations in West Kalimantan to improve sustainable management practices. Oil palm plantations play an important role in global food security and are expanding continuously in South-East Asia. They often have a significant environmental impact due to habitat fragmentation, biodiversity loss, and deforestation. Understanding the spatial and temporal patterns of these plantations is crucial for better management and reducing environmental impact. To assess forest and plantation dynamics, land cover maps were created using a Random Forest classifier on data from Sentinel-1, Sentinel-2, ALOS PALSAR-2, and Landsat 8 satellites, incorporating elevation and slope data from SRTM. The model achieved an overall accuracy of 80.7%±1% and a Kappa coefficient of 0.76. Time series analysis was performed on LAI and FAPAR derived from MODIS and Sentinel-2. Seasonal FAPAR anomalies revealed distinct wet and dry season patterns. Forests showed greater variability compared to plantations, attributed to the unmanaged nature of forests. Conversely, plantations demonstrated uniform structure and productivity due to consistent management and fertilization practices. Oil palm productivity was analysed in relation to FAPAR, LAI, precipitation, and temperature using an ARIMA model. FAPAR and precipitation showed a weak linear relationship with fresh fruit bunch yield, while LAI had a stronger positive correlation. Maximum temperature exhibited a strong positive relationship with productivity (R² = 0.62). The ARIMA model demonstrated the potential for forecasting productivity fluctuations using satellite and weather data. Overall, this study highlights the complex interplay between human activities and environmental factors in shaping forest and plantation landscapes, providing insights for sustainable management practices. Earth Observation imagery and machine learning models can improve oil palm productivity, resource efficiency, and forest protection.