Rice Field Detection by Dual Polarization SAR Images

Paddy rice plays a significant role in Asian agriculture, particularly in Taiwan. However, monitoring parcel-level activities and quantifying potential yield during the two crop cycles present challenges. The application of remote sensing to track paddy phenology emerges as a valuable strategy for improving crop management and ensuring food security. Synthetic Aperture Radar (SAR) satellites, among various spaceborne sensors, provide timely and extensive information unaffected by cloud cover. This study aims to extract time series data on paddy-specific phenology using dual-polarized SAR data and subsequently map paddy rice parcels in Taiwan. The process involves three primary steps: (1) Identifying phenological curves in training sites based on the temporal behavior of SAR backscattering coefficients; (2) Utilizing signal decomposition to analyze periodic patterns; (3) Recognizing rice fields by identifying the start and end of each crop cycle in the time series; (4) Validating the results with in situ data. In our preliminary findings, the accuracy in certain townships in western Taiwan achieves a kappa value of >0.6, with an overall accuracy exceeding 0.8. Additionally, we aim to unveil potential connections among crop cycles, groundwater changes, and land subsidence.