Tracking temporal variations in the soil-plant-atmosphere continuum in wheat using multisensor data

• Multi-sensor phenotyping links soil, canopy, and atmosphere in real time • PLSR with VIP retrieves photosynthetic rate (A) and stomatal conductance (Gs) • Temporal dynamics captured with GAMs under contrasting water regimes • Targeted blue and red bands, together with a wide NIR spectrum, dominate trait prediction beyond NDVI proxies • Scalable for breeding and on-farm monitoring with minimal ground truthing Understanding the soil-plant-atmosphere continuum (SPAC) is essential for breeding and advancing precision agriculture. Despite advances in hyperspectral monitoring, few studies have captured dynamic photosynthetic traits, such as net photosynthetic rate (A) and stomatal conductance (Gs), limiting insight into their temporal fluctuations and utility in breeding for stress resilience. This study integrates plant, soil and atmosphere sensor data, with statistical modelling to monitor season-long, fine-scale physiological and environmental variables, including A, Gs, vapor pressure deficit, soil moisture and crop water stress. A multi-sensor high-throughput phenotyping platform (HTPP) with a novel soil moisture system enabled high-resolution monitoring. Partial least squares regression (PLSR) models were used to predict photosynthetic traits from hyperspectral bands (∼400-1000 nm) and selected 20 vegetation indices (VIs). Temporal dynamics of both observed and predicted values were fitted using generalized additive models (GAMs) to describe the seasonal trajectories of photosynthetic traits, crop stress status and soil moisture across genotypes and water regimes. In wheat field trials, hyperspectral data predicted A and Gs with high accuracy (Root mean square error of prediction 3.71 and 58.93, respectively; R-squared 0.72 and 0.70, respectively) and the predicted temporal dynamics closely matched ground-truth measurements. Additionally, soil moisture and crop water status were monitored throughout the season, along with physiological traits. This approach provides scalable, data-driven solutions to support breeding for resilient cultivars and improvements in crop management, as the predicted data can be integrated into mechanistic crop models to establish empirical relationships with parameters that vary throughout the growing season.