<![CDATA[Indonesia's national sugar demand continues to increase, while domestic production is still limited due to the lack of accurate data on the area and type of sugarcane cultivation. To address this issue, this research utilized the Random Forest (RF) algorithm on PlanetScope satellite imagery to classify the types of sugarcane cultivation, i.e. early planted sugarcane and pressed sugarcane. This study offers a more detailed approach than previous studies by dividing sugarcane based on its growth cycle into two main classes: < 3 times and ≥ 3 times. The results showed that the Random Forest (RF) method was able to increase the mapping accuracy to 94.52%, higher than conventional methods which are generally in the range of 85-91%. Of the three classification schemes tested, Scheme 3 produced the best performance with an accuracy of 94.52% and a Kappa coefficient of 43.87%. The mapping results also revealed that sugarcane cultivation of ≥ 3 squeezes dominated the study area with 93.08% coverage of the total sugarcane land, while sugarcane of < 3 squeezes only covered 7.55%. The difference between the classification results and the field data shows that the imbalance in the number of samples and spectral similarity between classes are the main challenges in mapping sugarcane cultivation. This finding proves that the Random Forest algorithm with PlanetScope images can significantly improve the accuracy of sugarcane cultivation type mapping compared to previous methods. The results of this study make an important contribution in providing more accurate spatial data to support sugarcane production estimation, optimization of plantation management, and strategic planning in achieving national sugar self-sufficiency.]]>
Copyrights © 2025
