IDENTIFICATION OF RICE FIELDS USING THE RANDOM FOREST METHOD BASE ON SPECTRAL AND PHYSICAL PARAMETERS (CASE STUDY: BANDUNG REGENCY)

• 主要作者: Sutia Rahayu, Jania
• 格式: Theses
• 語言: Indonesia
• 在線閱讀: https://digilib.itb.ac.id/gdl/view/84931

This research aims to identify paddy fields in Bandung Regency using the Random Forest method by incorporating spectral and physical parameters. Typically, mapping paddy fields requires several satellite images over different periods due to rice's planting cycle. To reduce time and cost, this study optimizes Random Forest classification with single-time satellite images, adding physical characteristics of paddy fields. Data includes NDVI, NDWI, LST, SMI, BSI, SAVI, slope, elevation, soil pH, total nitrogen, and clay content. Results show that combining these parameters enhances classification accuracy, achieving 91% overall accuracy and a Kappa of 0.89. This approach is both effective and efficient for mapping paddy fields, crucial for agricultural management and food security, and supports sustainable monitoring of paddy field distribution in Bandung Regency. By integrating spectral parameters (NDVI, NDWI, LST, SMI, BSI, SAVI) and physical characteristics (slope, elevation, soil pH, total nitrogen, and clay content), the Random Forest method significantly improves the classification accuracy of paddy fields compared to using only spectral data. The classification results indicated a substantial improvement, with the overall accuracy reaching 91% and a Kappa coefficient of 0.89. This methodological approach not only demonstrates its effectiveness and efficiency but also plays a vital role in agricultural management and food security. It provides a sustainable solution for monitoring paddy field distribution in Bandung Regency.