Crop growth and yield are determined by multiple factors, including genotype, environment, and their interactions. The assimilation of remote sensing data with crop growth modeling represents a significant trend for crop monitoring and yield estimation. This study aims to explore an effective data fusion method for estimating soybean yield by utilizing canopy remote sensing data and crop growth models. Based on field experiment data, remote sensing retrieval models for the leaf area index (LAI) and leaf nitrogen accumulation (LNA) were developed using the Principal Component Analysis–Ridge Regression (PCA–Ridge) algorithm. Using remotely sensed estimates as state variables in the DSSAT model, the results indicated that, compared with using only the LAI (VLAI) or only LNA (VLNA), the accuracy of soybean yield estimation was superior when both the LAI and LNA (VLAI+LNA) were used as state variables. Additionally, the Nash–Sutcliffe efficiency (NSE) coefficient was a viable optimization function in optimizing the state variables. In conclusion, these results indicate that assimilating two key physiological and biochemical parameters for soybean, derived from hyperspectral data, with crop growth models provides a viable approach for enhancing the precision of estimating the LAI, LNA, and yield.
Han et al. (Sun,) studied this question.