Estimating the winter wheat harvest index with canopy hyperspectral remote sensing data based on the dynamic fraction of post-anthesis phase biomass accumulation

Remote sensing-based crop harvest index (HI) information is of great significance for crop yield estimation, crop variety breeding and evaluation of crop cultivation effectiveness. The method for estimating HI using fG (fraction of post-anthesis phase biomass accumulation) has been widely used and shows good performance at the field scale, but the upscaled regional application of this method had not been achieved using remote sensing information. In this paper, a remote sensing method for estimating the dynamic harvest index (D-HI) based on the remote sensing-based dynamic fG (D-fG) was proposed to solve this problem and was verified based on D-fG and D-HI measurements. This approach was based on accurate D-fG parameters estimated by using the NDSI constructed from hyperspectral sensitive band centres. The results showed that the D-fG estimation, the overall verification accuracy of the D-HI estimation at different growth stages and the accuracy of the D-HI estimation in a single growth stage were highly accurate. In the overall D-HI estimation verification, the normalized root square mean error (NRMSE) was between 10.83% and 14.45%, and the mean relative error (MRE) was between 9.62% and 13.99%. At maturity, the D-HI estimation accuracy based on band centre λ (732 nm, 834 nm) was the highest, and the NRMSE and MRE were 9.62% and 9.27%, respectively. Based on these results, the proposed method is feasible and effective at accurately estimating the D-HI, thus providing a technical reference for the use of satellite remote sensing data to obtain regional crop HI information based on fG.