Answers are grounded in this paper and cite their sources.
To address the challenges of crop classification in China’s smallholder farming systems—characterized by complex cropping patterns and high spectral similarity—this study proposes a classification method based on remote sensing technology combined with phenological monitoring. Using the Google Earth Engine (GEE) platform, the method integrates long-term Landsat satellite imagery with phenological feature analysis. Initial training samples are extracted using NDVI time-series decision rules, followed by spatial filtering to enhance sample quality. A random forest model is then applied to achieve high-accuracy crop classification. Taking the Hetao Plain as a case study, we generated 30-meter resolution crop classification maps (wheat, maize, sunflower, and vegetables) for the period 2000–2024. Validation with field survey points shows an overall accuracy exceeding 90% and a Kappa coefficient greater than 0.88, confirming the approach’s effectiveness. Spatiotemporal analysis reveals a shift in cropping structure toward higher-value crops over the past 25 year, with a significant decline in wheat acreage and continuous expansion of maize and sunflower cultivation. By effectively integrating phenological time-series information with machine learning, this study provides a robust solution for crop classification in complex agricultural landscapes, supporting sustainable agricultural and water resource management.
No takes yet. Share an insight, caveat, or question.
Xing et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: