Remote Sensing-Based Mapping of Paddy Rice Cultivation Areas: Progress and Prospects in the Era of Artificial Intelligence

Accurate and timely mapping of paddy rice cultivation areas underpins food security assessment, agricultural resource management, and climate change adaptation. Because remote sensing provides repeated, large-scale observations, it has become indispensable for monitoring rice cultivation. This review synthesizes studies published since 2000, emphasizing methodological and technological advances since 2016, and critically evaluates the role of artificial intelligence (AI) in rice mapping. We examine the observational characteristics and inherent limitations of optical imagery, synthetic aperture radar (SAR), hyperspectral data, unmanned aerial vehicle (UAV) observations, and multi-source data fusion for rice identification. We then trace the evolution of mapping methods, from threshold-based classification and phenology-driven time-series analysis to machine learning, deep learning, and cloud-computing frameworks, comparing their accuracy, scalability, computational efficiency, transferability, and interpretability. Applications at global, regional, and field scales, each characterized by distinct objectives and spatial requirements, are reviewed alongside prevailing validation strategies and sources of uncertainty. Persistent challenges include cloud contamination, spectral and phenological confusion among crop types, ambiguous target definitions, limited cross-regional generalizability, difficulty in distinguishing multiple cropping cycles, insufficient evidence of model transferability, and geographical concentration of the existing evidence base. Priority directions include adaptive multi-source data fusion, self-supervised and physics-guided learning, Earth observation foundation models, explainable and uncertainty-aware AI, parcel-level product generation, and operational monitoring systems coupled with crop growth models.