The optimal soil water content models based on crop-LAI and hyperspectral data of winter wheat

The use of spectral data to predict soil water content has gained wide application in agricultural science. However, it is difficult to guarantee crop water status prediction accuracy based on spectral parameters because the physiological indices and crop water status change daily. Therefore, screening representative crop growth indicators could improve the accuracy of the crop water prediction model. In this study, winter wheat was used as the crop of interest. Initially, spectral characteristics proposed by previous studies were selected and screened. Subsequently, soil water content prediction models were constructed based on a combination of crop leaf area index (LAI) and its spectral characteristics and crop growth physiological indices to predict the field soil water content. These models were constructed using three types of parameters, including single spectral characteristics of canopy water content, single spectral characteristics of canopy water content and measured LAI, as well as spectral characteristics of both canopy water content and LAI. The coefficient of determination ( R 2 ) that reflects the reliability of the models was 0.31–0.36, 0.57–0.62, and 0.45–0.54, respectively. The model constructed based on measured LAI and spectral characteristics was the most accurate in each growth period and the whole growth period of winter wheat, followed by that based on dual-spectral characteristics, whereas the single spectral characteristics model was the least accurate. The R 2 of the model constructed based on measured LAI and characteristic spectral parameters of canopy water content increased by 0.47, 0.16, 0.69, and 0.37 during the entire growth period, respectively, implying that combining LAI with spectral characteristics can improve the accuracy of soil water content prediction models. However, it was difficult to obtain the relevant measured index for the models constructed using measured LAI. Therefore, the dual-spectral characteristics model was recommended as the most appropriate for practical application.