ABSTRACT Net photosynthetic rate (Pn) serves as a key indicator for evaluating plant growth and yield. In order to explore an effective method for monitoring winter rapeseed Pn using unmanned aerial vehicle (UAV) multispectral technology, the winter rapeseed was taken as the research object and the multispectral images were obtained through UAVs in this study, combined with field measured Pn data in different growth periods, and Pearson correlation analysis was used to screen vegetation indices (VIs) and texture features (TFs) that were strongly correlated with Pn, and then screened again through recursive feature elimination (RFE), least absolute shrinkage and selection operator (LASSO), and maximum relevance minimum redundancy (MRMR). Pn inversion models of winter rapeseed were constructed based on back propagation neural network (BPNN), random forest (RF), support vector regression (SVR), and multiple linear regression (MLR), and SHapley Additive exPlanations (SHAP) was used to reveal the importance of features. The analysis results showed that the RF model had the highest accuracy in inverting Pn in different growth stages, with the coefficient of determination ( R 2 ) of the test sets in the bolting stage and flowering stage being 0.911 and 0.881, respectively, while the root mean square error (RMSE) was 0.727 μmol·m −2 ·s −1 and 0.917 μmol·m −2 ·s −1 , respectively. Moreover, the R 2 of the test set in the whole stage was 0.954, and the RMSE was 0.715 μmol·m −2 ·s −1 . SHAP analysis showed that the red‐edge chlorophyll index 1 (CI rededge1 ) and the red‐edge chlorophyll index 2 (CI rededge2 ) played an important role in the RF model inversion of VIs+TFs based on RFE. The research results can provide a theoretical basis and technical support for the inversion of winter rapeseed Pn using multispectral remote sensing by UAV.
Ni et al. (2026) studied this question.