Mapping rubber monoculture and jungle rubber across Kalimantan using multisource satellite data

Natural rubber is a strategic commodity and an important source of income for millions of smallholders in South-East Asia. In Indonesia, rubber systems vary widely, from high-yielding monocultures to complex agroforestry systems known as jungle rubber, making their accurate mapping challenging. The lack of reliable, high-resolution maps hampers efforts to monitor land-use change and comply with emerging deforestation-free commodity regulations. The objective of this study is to develop and evaluate a remote sensing-based methodology to map both rubber monocultures and jungle rubber across Kalimantan. We developed a method based on a Random Forest machine-learning classification algorithm calibrated on multi-source imagery from Sentinel-1 (radar) and Sentinel-2 (optical) satellites, applying a Principal Component Analysis for feature reduction. To distinguish jungle rubber from monocultures, we further integrated land-cover history obtained from the JRC Tropical Moist Forest (TMF) dataset, which showed good agreement with our training data: most jungle rubber points fell within TMF forest classes, while monocultures fell within non-forest classes. The final map, evaluated against extensive field data, gave high overall accuracy (87%) for total rubber area, estimated at 1.79 million hectares in 2020 for the three provinces of Central, South and West Kalimantan. The overall accuracy was 70% when including the distinction between monoculture and jungle rubber, with estimated areas of 0.59 Mha and 1.20 Mha, respectively. Spatial patterns revealed monocultures concentrated in accessible regions such as in South Kalimantan or along the Kapuas river, while jungle rubber dominated remote or less accessible zones, particularly in West Kalimantan. This classification approach is the first to produce reliable maps of monoculture and jungle rubber at large scale and provides a foundation for future monitoring and policy applications.