Characterizing wheat and barley growth and phenology using multi-spectral remote sensing for site-specific precision agriculture

Abstract

Crop phenology informs in-season management practices such as fertilizer and pest and disease controls to optimize final yield. However, tracking crop growth stages across spatiotemporal domains remains challenging, particularly in rainfed broadacre systems subject to climatic variability. This study uses sequential high-resolution Sentinel-2 imagery to estimate phenological stages of wheat and barley across the Australian grain cropping region which comprises >20 million ha of production. An analysis pipeline was developed to estimate main crop growth stages using targeted vegetation indices (VI) and statistical model fitting approaches. Model accuracy was validated against biophysical simulated phenology and field observations from diverse environments. Both parametric models and non-parametric models were evaluated to fill data gaps and capture growth dynamics. The double logistic model was selected for its balance of performance and efficiency. Strong alignment was observed between VI-derived features and simulated phenology. Peak and right shoulder features showed high accuracy for estimating stages of flag leaf (R2 = 0.61, root mean square error (RMSE) = 8.67 days) and flowering (R2 = 0.7, RMSE = 7.66 days). Scalability was evaluated at 110 and 73 fields across Australia, for 2021 and 2022 seasons, respectively, showing moderate to high correlations with recorded phenology (flag leaf R2 = 0.57–0.58; flowering R2 = 0.7–0.85). The method also predicted tillering (R2 = 0.2–0.7), maturity (R2 = 0.73–0.85), and harvest (R2 = 0.55–0.66) where these observations were available. These results demonstrate the utility of high-resolution satellite data for estimating crop phenology and supporting zonal in-season agronomic management.