ALOS-Sensor data integration for the detection of smallholder's oil palm plantation in Southern Sumatra, Indonesia

Abstract

The high demand for palm oil has triggered the expansion of oil palm plantation across the world, and especially in Indonesia. This phenomenon is indeed contributing to the national income and represents an alternative energy source, but is concurrently leading to numerous environmental and land management problems. The emergence of private smallholder's oil palm plantations is complicating the quantification of the environmental impacts of oil palm expansion, since its sparse distribution in small patches is difficult to be identified. In Mesuji area of Southern Sumatra, Indonesia, this type of plantations has been rapidly growing and potentially affecting the land management especially for protected areas. This study aimed to explore the ability of the multi-sensor and multi-scale images obtained by Advanced Land Observing Satellite (ALOS) to detect the oil palms and to develop the effective methodology to accurately classify smallholder's oil palm plantation in Mesuji area. By utilizing the texture analysis, the triangular planting pattern was recognized as the unique characteristic of oil palm plantation reflected on Synthetic Aperture Radar (SAR) data to identify the small scale smallholder's plantation. Finally, the study also revealed the integration of mean-variance texture features extracted from ALOS PALSAR data in 11 x 11 moving window sizes with the use of all ALOS AVNIR-2 bands resulting in the best accuracy to detect smallholder's oil palm plantation with 92.45% of producer's accuracy and 66.67% of user's accuracy for the mature oil palm. For the young stages, the accuracy was 64.44% and 63.04% for producer's and user's accuracy, respectively.