Estimasi Produksi Padi Ramah Lingkungan Menggunakan Kombinasi Indeks Vegetasi dari Citra Sentinel-2

Abstract

The increasing adoption of environmental friendly rice cultivation methods has highlighted the need for efficient and accurate yield estimation models to support sustainable agricultural practices. This study aimed to develop a predictive model for estimating rice yield using vegetation indices derived from Sentinel-2 satellite imagery, processed through the Google Earth Engine (GEE) platform. The research was conducted in Tanggung Village, Padang Subdistrict, Lumajang Regency, East Java, where sustainable farming has been consistently implemented. Five vegetation indices NDVI, GNDVI, NDII, NDRE, and SAVI were analyzed, with the maximum value of each index extracted to reflect peak crop conditions during the growing season. Both simple and multiple linear regression models were employed to examine the relationship between vegetation indices and actual rice yields. Model performance was evaluated using the R² and Root Mean Square Error (RMSE). The correlation analysis showed that GNDVI had the highest correlation coefficient 0.770 with actual rice yield, indicating a strong association with crop performance. This high correlation was attributed to GNDVI’s sensitivity to chlorophyll content, which plays a crucial role in photosynthesis and yield formation. In the regression analysis, GNDVI also demonstrated the best performance among single-index models, with an R² of 0.637 and an RMSE of 0.352 tons/ha. The highest-performing model overall was obtained through the combination of all five indices, yielding an R² of 0.727 and an RMSE of 0.442 tons/ha. These findings confirmed that integrating multiple vegetation indices improved model accuracy and offered a more robust approach for estimating rice yields in environmentally friendly farming systems using remote sensing data.