Integration of Multiscale Spectral Features as the Intermediate Variables for Improved Prediction of Grain Protein Concentration in Winter Wheat

Grain protein concentration (GPC) could be predicted with high accuracy using the reflectance of dried grain powder in the shortwave infrared (SWIR) region. However, the acquisition of grain-level hyperspectral data is time-consuming, and such a procedure must be conducted under laboratory conditions. This study proposed an efficient approach to integrating multiscale spectral features (SFs) as the intermediate variables for improved prediction of GPC in winter wheat by transporting the sensitive features from grain to canopy levels. The performance of the proposed method was compared in the form of “Dependent variable $\,\sim \,$ Independent variable $\,\sim \,$ Intermediate variable” with the traditional canopy-level “GPC $\,\sim \,$ SF $\,\sim \,$ leaf nitrogen concentration (LNC)” models. Our results demonstrated that the models of “GPC $\,\sim \,$ wavelet feature (WF) $\,\sim \,$ LNC” exhibited higher validation accuracies than the traditional “GPC $\,\sim \,$ vegetation index (VI) $\,\sim \,$ LNC” models. Specifically, the canopy-level sensitive $\mathrm{WFc}_{\mathrm {anopy-720,5}}$ (WF at 720 nm and the wavelet decomposition scale of $2^{5}$ ) yielded pronounced improvement for GPC prediction, with root-mean-square error (RMSE) values of 1.02% and 0.83% for the anthesis and the post-anthesis stages, respectively. Compared with models using the SF- or LNC-based intermediate variables, the model using the intermediate variable of $\mathrm{WF}_{\mathrm {grain-1510,4}}$ (grain-level sensitive WF) performed remarkably better, such as the single-feature model of “GPC $\,\sim \,$ $\mathrm{WF}_{\mathrm {canopy-1510,4}}\,\sim \,$ $\mathrm{WF}_{\mathrm {grain-1510,4}}$ ” at the post-anthesis stage (RMSE = 0.78%) and multifeature model of “GPC $\,\sim \,$ ( $\mathrm{WF}_{\mathrm {canopy-1510,4}}$ and $\mathrm{WF}_{\mathrm {canopy-1610,5}})\,\sim \,$ ( $\mathrm{WF}_{\mathrm {grain-1510,4}}$ and $\mathrm{WF}_{\mathrm {grain-1610,5}}$ )” at the anthesis stage (RMSE = 0.93%). The grain-level SFs are not expressed in the final model, and only canopy-level data are required for model applications. This study represents a first attempt to couple grain and canopy-level SFs to improve the spectroscopic prediction of GPC and provides new insight into the quantitative estimation of grain quality parameters with hyperspectral remote sensing.