Mapping rice paddy fields in semiarid northeastern Thailand using Sentinel-1 and -2 time series data and phenological clustering

Abstract

Rice paddy fields in semiarid northeastern Thailand are a unique feature of traditional agriculture in the region, presenting unique agricultural practices and challenges, particularly in water management in semiarid conditions. This study aims to improve rice production efficiency and sustainability by understanding how these methods arose and how they could be enhanced to have an impact on agricultural practices in other regions with similar problems. This study developing a phenology-based method using the unsupervised classification of Sentinel-1 and 2 time-series data to identify rice paddy fields. Vegetation Index (VI) time series were used for identifying variations in the canopy of rice fields during growth stages. Monthly Sentinel-1 and 2 time series data between January 2020 and December 2022 were classified using k-means clustering to identify regions with similar phenological patterns. This methodology produced maps with a 10-meter resolution of rice field extent, intensity, and crop calendar. Validation was performed using the MS-700 Spectroradiometer and time-series image observed from the top of an automatic weather station. The results indicate that the proposed method based on phenology is cost-effective and capable of accurately mapping rice fields and growth stages across large areas. This study also highlights the importance of rice paddy fields in semiarid northeastern Thailand and how the developed methodology can help improve holistic water resources management and sustainability in the region and other similar areas.