As climate change intensifies and global food security faces pressures, accurate monitoring of crop phenology—especially sowing dates—has become critical for optimizing agricultural management and improving climate resilience. Winter wheat, a staple crop supporting nearly 40% of the global population, relies heavily on timely sowing to maximize yield potential. However, traditional monitoring methods such as field surveys are labor-intensive and unscalable, while existing remote sensing approaches suffer from soil background interference and static environmental data limitations.
Addressing these challenges, a research team led by Prof. Jin Chen from Beijing Normal University, in collaboration with researchers from the Chinese Academy of Agricultural Sciences and Henan Academy of Agricultural Sciences, has developed a novel machine learning framework for early-season winter wheat sowing date estimation.
The study was made available online on April 02, 2026 in The Crop Journal .
"Our framework integrates high-resolution remote sensing, dynamic climate windows, and machine learning to solve long-standing bottlenecks in sowing date estimation," explains Jianlong Li. "We used the Normalized Difference Greenness Index (NDGI) from Sentinel-2 data to detect crop emergence—an earlier and more direct phenological marker than traditional green-up dates—while dynamic climate windows anchored to emergence timing capture site-specific environmental conditions before and after seedling emergence."
The innovation lies in three advancements:
Validated in Henan Province during the 2024 growing season using 335 field-collected sowing date samples, the framework achieved an R² of 0.82, with most estimates falling within ±5 days of observed values. The XGBoost model, identified as optimal, generated 10-meter resolution sowing date maps that clearly captured regional patterns—earlier sowing in northeastern Henan and delayed sowing in the south—and fine-scale variations within fields and villages.
"Feature importance analysis confirmed that emergence date and pre-emergence soil conditions are the most critical predictors, which aligns with crop physiological mechanisms," shares corresponding author Jin Chen. "Seeds and underground seedlings are highly sensitive to soil temperature and moisture during germination, and our dynamic windows effectively capture these key environmental drivers."
The method also demonstrates strong generalizability: extended testing on summer maize in Hebei Province yielded an R² of 0.51, outperforming benchmark methods despite limited samples. The NDGI threshold for emergence detection (0.04 above soil background) remained stable across crop types, while climate window lengths were adjustable for different species.
Notably, the framework enables early-season sowing date monitoring—weeks earlier than traditional remote sensing methods—providing actionable data for farmers and policymakers to adjust management strategies, reduce climate risks, and improve crop growth model accuracy. For global food security, this scalable tool supports large-scale phenology monitoring in diverse agricultural systems.
"In the next 3–5 years, we aim to integrate synthetic aperture radar (SAR) data to overcome cloud contamination limitations and refine the framework for more crop species," adds Chen. "This will further enhance its utility for climate-resilient agriculture and data-driven farming decisions worldwide."
###
Contact the author:
Jin Chen
Email address: chenjin@bnu.edu.cn
The publisher KeAi was established by Elsevier and China Science Publishing & Media Ltd to unfold quality research globally. In 2013, our focus shifted to open access publishing. We now proudly publish more than 200 world-class, open access, English language journals, spanning all scientific disciplines. Many of these are titles we publish in partnership with prestigious societies and academic institutions, such as the National Natural Science Foundation of China (NSFC).
The Crop Journal
Experimental study
Not applicable
Early-season estimation of winter wheat sowing date: Integration of dynamic climate windows and phenological indicators into machine learning models
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.