Understanding crop phenology—the timing of key developmental stages such as germination, growth, and senescence—is essential for agricultural management. Accurate knowledge of crop calendars helps optimize irrigation, fertilization, disease monitoring, and yield prediction. Traditional approaches rely on field observations or ground-based monitoring systems, but these methods are often limited in spatial coverage and require intensive labor. Satellite remote sensing provides large-scale monitoring capabilities, yet detecting early crop stages such as sowing and emergence remains difficult because satellite pixels capture mixed signals from soil and sparse vegetation. Cloud cover and data gaps further complicate time-series analysis. Due to these challenges, there is a need to develop new methods capable of accurately estimating sowing and emergence dates using satellite observations.
Researchers from Mississippi State University and collaborating institutions reported a new framework for estimating crop sowing and emergence dates using satellite observations. The study, published (DOI: 10.34133/remotesensing.0878) on March 11, 2026 in the Journal of Remote Sensing, integrates daily synthetic Harmonized Landsat Sentinel-2 (HLS) imagery with machine-learning models to reconstruct vegetation dynamics across agricultural fields. By analyzing vegetation index time series, the framework infers early crop development stages that are typically difficult to detect from space. The technology addresses a major challenge in remote sensing–based agricultural monitoring: identifying the start of the crop growth cycle accurately at large spatial scales.
The research introduces an operational pipeline that combines satellite time-series reconstruction with phenological modeling to estimate crop planting timelines. The method first reconstructs continuous vegetation index data from Landsat and Sentinel-2 imagery, filling gaps caused by cloud cover. From the reconstructed time series, six phenological stages—greenup, mid-greenup, maturity, senescence, mid-greendown, and dormancy—are extracted using an asymmetric double-sigmoid model. Machine-learning algorithms then infer sowing and emergence dates based on the relationships between these phenological stages.
Among the tested models, elastic net regression achieved the best performance, predicting sowing and emergence dates with an average error of about ±10 days. The approach successfully estimated crop calendar events across large corn and soybean fields in the United States, demonstrating strong agreement with ground observations from PhenoCam monitoring sites.
The researchers built the framework using daily synthetic time series derived from Harmonized Landsat Sentinel-2 (HLS) data, which combine observations from Landsat 8/9 and Sentinel-2 satellites at a spatial resolution of 30 meters. Because satellite observations are frequently affected by cloud contamination, the team tested four gap-filling approaches—median interpolation, polynomial regression, harmonic modeling, and Light Gradient Boosting Machine (LightGBM)—to reconstruct continuous vegetation index signals.
The polynomial method produced the most accurate reconstructions, preserving seasonal vegetation dynamics while minimizing noise in the time series. These reconstructed data were then used to calculate the Enhanced Vegetation Index (EVI) and derive phenological curves representing crop growth cycles.
To validate the approach, the researchers compared satellite-derived phenological stages with ground observations from 20 PhenoCam monitoring sites across 13 U.S. states. The comparison showed strong agreement between satellite and ground measurements, with a coefficient of determination (R²) of 0.94 and a bias of approximately 12 days.
Using the phenological stages as predictors, machine-learning models were trained to estimate sowing and emergence dates. The best model was then applied to thousands of agricultural fields, successfully mapping planting timelines across large agricultural regions.
"Our framework demonstrates that early crop development stages can be inferred indirectly from later phenological signals," the researchers noted. "Even though sowing and emergence are difficult to observe directly from satellite imagery, the seasonal growth trajectory contains enough information to reconstruct these dates." The team emphasized that the approach enables scalable monitoring of crop calendars across large regions, which could significantly improve agricultural forecasting and management strategies.
The study used daily synthetic vegetation index time series derived from Harmonized Landsat Sentinel-2 imagery collected between 2021 and 2023. After cloud removal, four gap-filling algorithms were tested to reconstruct missing data. Crop phenological stages were extracted using an asymmetric double-sigmoid model applied to Enhanced Vegetation Index curves. Ground observations from 20 PhenoCam sites were used to validate satellite-derived phenology. Three machine-learning models—multiple linear regression, elastic net regression, and support vector machines—were trained using leave-one-out cross-validation to estimate sowing and emergence dates.
The proposed framework could significantly enhance large-scale agricultural monitoring and decision-making. Accurate knowledge of sowing and emergence dates enables better crop growth modeling, yield forecasting, and climate risk assessment. The system could also support early detection of crop stress, disease outbreaks, and extreme weather impacts. With further refinement, the method could be integrated into global agricultural monitoring platforms and precision agriculture systems. As satellite observations and artificial intelligence technologies continue to advance, such data-driven tools may become essential for ensuring food security and improving agricultural sustainability worldwide.
###
References
DOI
10.34133/remotesensing.0878
Original Source URL
https://spj.science.org/doi/10.34133/remotesensing.0878
Funding information
This work is supported by the Data Science for Food and Agricultural Systems program, project award no. 2024-67021-42530, and Soil Health program, project award 2023-67019-39169, from the USDA's National Institute of Food and Agriculture. U.R.V.A., L.B.F., and V.S.M. were partially supported by the USDA Agricultural Research Service (nos. 58-0200-0-002 and 58-6060-3-005). Y.Y. is partially supported by the NASA Acres project (80NSSC23M0034). I.D.A.S. was grateful to the Brazilian National Council of Scientific and Technological Development (CNPq) for the Research Productivity Fellowship (310042/2021-6).
About Journal of Remote Sensing
Journal of Remote Sensing, an online-only Open Access journal published in association with AIR-CAS, promotes the theory, science, and technology of remote sensing, as well as interdisciplinary research within earth and information science.