Estimating wheat spike-leaf composite indicator (SLI) dynamics by coupling spectral indices and machine learning

Abstract

The contribution of spike photosynthesis to wheat yield grain yield (GY) has been overlooked in crop models. Our objectives were to (i) construct a yield-related phenotypic trait, spike–leaf composite indicator (SLI), accounting for the contribution of the spike to photosynthesis, (ii) develop an enhanced triangle vegetation index 3 (ETVI3) based on the SLI response to canopy spectra, and (iii) establish and evaluate SLI estimation models by integrating spectral indices and eXtreme Gradient Boosting with feature selection (XGB-FS) algorithm and four regression models, using field and spectral reflectance data from 19 wheat cultivars under two nitrogen fertilization conditions in two years. The results showed that spike photosynthesis accounted for 28 % of GY. SLI was sensitive to nitrogen fertilizer and wheat cultivar variation as well as a better predictor of yield than leaf area index. ETVI3 displayed SLI differences through wheat canopy tissue abundance, chlorophyll content, and plant water content, was strongly correlated with SLI, whereas other spectral indices were more poorly correlated. Spectral features selected from XGB-FS algorithm enhanced the interpretability of the regression models and improved the estimation accuracy of SLI, wherein XGB-FS-LASSO was validated with higher accuracy for SLI estimation at the post-heading stage with coefficient of determination, root mean square error (RMSE), and relative RMSE values of 0.71 %, 0.047 %, and 26.93 %, respectively. These results provide new insights into the role of fruiting organs for yield prediction. This high-throughput SLI estimation approach can be applied for accurate spectral prediction of GY at whole growth stages and may be assisted with agronomical practices and variety selection.