Answers are grounded in this paper and cite their sources.
Leaf area index (LAI) is a key determinant of canopy structure, photosynthetic capacity, and biomass accumulation in wheat, thereby playing a critical role in regulating final grain yield (GY). Accurate and rapid estimation of LAI and GY is essential for precision crop management and high-yield breeding. Leveraging the capacity of multi-task learning (MTL) to exploit intrinsic trait correlations for improved prediction accuracy. This study developed an integrated framework based on canopy hyperspectral remote sensing to simultaneously forecast winter wheat LAI and GY across multiple growth stages and environments. To establish a benchmark, four single-task algorithms, including ridge regression (RR), random forest, support vector regression, and partial least squares regression, were initially evaluated. Among these algorithms, RR exhibited the most robust performance, achieving independent testing coefficients of determination ( R² ) of 0.57 for GY and 0.70 for LAI. Subsequently, the MTL framework integrated three optimization strategies: task-weight balancing, NDVI fusion, and using LAI as an auxiliary variable. While initial weight optimization yielded moderate gains ( R² = 0.61 for GY, incorporating NDVI boosted LAI prediction accuracy to an R² of 0.86). Furthermore, integrating LAI as an auxiliary trait notably enhanced GY prediction, achieving an R² of 0.75 and RMSE of 0.25 kg. Compared to the single-task RR baseline, the fully integrated MTL framework improved predictive performance, increasing the R² for GY from 0.57 to 0.75 and for LAI from 0.70 to 0.86. These results demonstrate that the proposed MTL approach effectively captures canopy-yield associations, offering a robust tool for rapid field-scale phenotyping and precision yield forecasting.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: