Meningkatkan Efisiensi Pemetaan Kelapa Sawit Dengan Google Earth Engine dan Algoritma Machine Learning Untuk Mendukung Sertifikasi RSPO

Abstract

Accurate and efficient palm oil plantation mapping is crucial to support the RSPO (Roundtable on Sustainable Palm Oil) Certification program. This study aims to develop an efficient palm oil plantation mapping method using Google Earth Engine and machine learning algorithms. Two machine learning algorithms, Deep Learning and SVM (Support Vector Machine), were used to classify palm oil plantations from Landsat 8, Landsat 9, and SRTM satellite images. The results showed that the SVM algorithm produced higher classification accuracy (98.1% on training and 96% on testing) compared to Deep Learning (90.53% and 93.46%). The mapping results were compared with actual mapping obtained from Serawak Oil Palm Concession data on land owned by Sunbest Mill Sdn Bhd, yielding SVM accuracy of 94.7% and Deep Learning accuracy of 93%. This mapping meets 5 points in the RSPO Formatting Requirements for Map Data Submission document, but does not yet meet points 4 and 6 related to concession boundary mapping and palm oil plantation attribute information. This study demonstrates that using Google Earth Engine and machine learning algorithms can improve the efficiency of palm oil plantation mapping with high accuracy. However, further development is needed to meet all points in the RSPO document.