Sentinel-2 for Crop Yield Estimation: A Systematic Review

A team of researchers has published a comprehensive 29-page review, 'Sentinel-2 for Crop Yield Estimation: A Systematic Review,' analyzing the transformative impact of the Copernicus Sentinel-2 satellite constellation on agricultural monitoring. The review synthesizes recent advances, highlighting a key industry shift from broad regional yield models to high-resolution, field- and sub-field-scale analysis. This granularity is critical for modern precision agriculture, enabling targeted farm management, resource allocation, and policy decisions based on highly localized data.

The paper identifies three primary technical approaches driving this revolution. The first involves empirical models that combine traditional vegetation indices with advanced machine learning (ML) and deep learning (DL) methods like Random Forest and Convolutional Neural Networks (CNNs). The second approach integrates process-based crop growth models, such as WOFOST and SAFY, by assimilating Sentinel-2-derived variables like Leaf Area Index (LAI). The third strategy employs data fusion, combining Sentinel-2's optical data with Sentinel-1's Synthetic Aperture Radar (SAR) to overcome the persistent challenge of cloud cover, ensuring more consistent data streams.

While these AI and hybrid modeling frameworks can explain substantial within-field yield variability, the review notes significant constraints. Performance is still limited by a scarcity of high-quality ground-truth data for model training, gaps caused by cloud cover, and difficulties in transferring models across different growing seasons and geographic locations. The authors point to future directions that include tighter integration of multi-modal data sources and improved methods for in-season observation, which are essential for building robust, operational decision-support systems for global food security and sustainable agricultural intensification.