Leaf area index estimation under wheat powdery mildew stress by integrating UAV‑based spectral, textural and structural features

Abstract

The infection of powdery mildew changes the internal physiological activity and external morphology of winter wheat, resulting in damage to crop health. Leaf area index (LAI) is an important agronomic parameter to assess crop growth. However, how to accurately and quickly interpret the spatial variation of LAI under powdery mildew stress has become a difficult problem. To detect the spatial variation and improve the estimation accuracy of winter wheat LAI under powdery mildew stress, unmanned aerial vehicle (UAV) multispectral remote sensing technique was used to continuously observe winter wheat infected with diseases in the field. The artificial inoculation experiment of powdery mildew was carried out in the experimental base of the Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Xinxiang, China in 2022. The occurrence of powdery mildew changed the canopy spectral reflectance and morphology of winter wheat. Parameters of vegetation indices (VIs), structure (Str) and texture (Tex) features were used to quantify the spectral reflectance and morphology of winter wheat canopy. Sensitivity analysis of different types of features and LAI and crop disease index (DI) was conducted to obtain specific features responding to LAI and DI. Partial least squares method (PLSR) was used to construct LAI estimation model under powdery mildew stress. The results of this study showed that (i) in contrast to crop height, crop coverage and canopy volume maintained consistency with the time change of LAI, which could be utilized to assess LAI variation under powdery mildew stress. (ii) In general, Tex features were more sensitive to LAI and DI than the Str and VIs. (iii) The R2 and RMSE of LAI estimated under powdery mildew stress by Tex, Str and VIs were 0.56, 0.50, 0.45 and 0.32, 0.35, and 0.36, respectively. The combination of VIs, Str, and Tex provided the highest estimation accuracy (R2 = 0.71, RMSE = 0.26). Compared with VIs, Str, Tex, VIs + Str, VIs + Tex, Str + Tex, the R2 of winter wheat LAI estimation under powdery mildew stress was increased by 58%, 42%, 27%, 22%, 15%, and 8%, respectively. This study analyzed the spatial variation of LAI of winter wheat under powdery mildew stress using remote sensing technique, which provided an important theoretical basis for crop disease control in the field.