A comprehensive review of rice mapping from satellite data: Algorithms, product characteristics and consistency assessment

With a growing global population and intensifying regional conflicts, the need for food is more urgent than ever. Rice, as one of the world's major staple crops especially in Asia, sustains over 50 percent of the global population. Accurate rice mapping is fundamental to ensuring global food security and sustainable agricultural development. Remote sensing has become an essential tool for mapping rice cultivation due to its ability to cover large areas and provide timely observation. Existing reviews mainly focus on the paddy rice mapping methods. However, it lacks a comprehensive understanding on the quality of different paddy rice maps from regional to global scales. This paper provides a comprehensive review of existing satellite-based rice mapping methods and products. Firstly, we categorized all previous methods into four classes: 1) spatial statistical method; 2) traditional machine learning method; 3) phenology-based method; and 4) deep learning method. Secondly, we summarized 25 products, including 3 global products and 22 regional products. Furthermore, we examined the consistency and discrepancy among different products in China, Heilongjiang China and Vietnam respectively and explored the underlying reasons. We found that 1) rice fields with simple cropping patterns and intensive cultivation can be correctly recognized using various algorithms; 2) different products share low consistency in fragmented rice fields 3) the prevalence of clouds and complicated rice cropping patterns or diverse growing environments in subtropical and tropical regions poses challenges to accurate rice mapping. Due to these challenges, currently it still lacks paddy rice maps with both large spatial coverage, high spatial resolution, and long time series. Moreover, deficiency of ground-truth samples impedes product development and validation. For improved paddy rice mapping at large scale, we suggest to apply sample-free rice mapping techniques and remote sensing foundation models to leverage the strengths of phenology-based methods and deep learning methods.