Multi-stage prediction of winter wheat grain protein content: integrating mechanism model and hyperspectral inversion of agronomic parameters

Predicting grain protein content (GPC) of winter wheat using hyperspectral remote sensing technology is important for the cultivation and production management of high-quality wheat. However, traditional destructive sampling methods are time-consuming, costly, and challenging to implement for rapid monitoring. Based on the integrating of mechanism model and hyperspectral inversion agronomic parameters (APs), a multi-stage model construction method for predicting winter wheat grain quality was proposed in this study. By comparing the GPC prediction models constructed by different APs, the optimal prediction model parameter selection under multi-stages is determined. And the accurate inversion framework of hyperspectral APs was constructed through feature screening combined with machine learning. In the hyperspectral prediction model of multi-stage grain quality obtained after coupling, partial least squares regression with leaf nitrogen content at water ripe stage had the best prediction accuracy, R2 = 0.74, nMAE = 6.80%, and nRMSE = 5.12%. By comparing error sources in multi-stage wheat GPC prediction models, we should focus on improving mechanistic model accuracy in early growth stages and enhancing hyperspectral inversion precision of APs in later growth stages. This study has provided a solid theoretical foundation and empirical support for the accurate multi-stage prediction of winter wheat quality.