Tea (Camellia sinensis) plays a crucial role in the Indonesian economy, but its production has witnessed a decline over the last two years. The majority of tea plantations in Indonesia have historical roots from the Dutch East Indies colonial government, underscoring the importance of monitoring pruning practices in tea plantation management. However, manual monitoring methods prove ineffective, prompting the exploration of precision farming principles using satellite imagery and machine learning to address this challenge. This study was conducted on a 231-hectare tea plantation south of the Tangkuban Perahu Volcano in West Bandung Regency, Indonesia. Sentinel-2B imagery from June to October 2019-2023 was utilized to calculate Soil Adjusted Vegetation Index (SAVI) values and assess tea productivity. Employing the K-Means algorithm, productivity values were grouped, and pruning intensity was categorized into three classes, revealing spatial dynamics influencing tea tree productivity. Our results illustrate distinct spatial patterns of pruning intensity across the 231-hectare tea plantation, identifying three classes: less pruned, moderately intensive pruned, and intensively pruned. We show that proposed model of pruning intensity from K-Means algorithm has overall accuracy of 0.57, with highest precision in less pruned class, while the lowest in moderately intensive pruned class. Notably, consistently higher productivity was observed in less pruned tea trees, while an increase in productivity in intensively pruned trees occurred after the second pruning in 2022. These findings highlight the potential of satellite imagery and machine learning for enhancing precision monitoring in tea plantations, offering a practical approach for long-term plantation management. Emphasizing the significance of pruning strategies, our study suggests that optimizing tea productivity amidst environmental and management challenges is achievable through informed monitoring and strategic pruning practices.
Read full abstract