IDENTIFICATION OF INITIAL STAGES OF IRRIGATED RICE-FIELDS UTILIZING SENTINEL-2 IMAGERY AND A MACHINE LEARNING ALGORITHM

Classifying different phenological stages of irrigated rice- fields is crucial for yield prediction and monitoring sowing progress. Remote sensing, combined with machine learning techniques, is increasingly used for agricultural mapping. This study utilizes spectral bands and indices from the Sentinel-2 satellite within an XGBoost algorithm to identify four initial stages of irrigated rice-fields. The combined approach of using both indices and bands achieved an overall classification accuracy of 83.9%, slightly outperforming the use of bands or indices alone.