The Farmland Irrigation Research Institute of Chinese Academy of Agricultural Sciences (CAAS) used the UAV-based multispectral and thermal infrared image features as inputs for the framework of elastic network regression and entropy weight ensemble algorithm to predict the wheat yield under water stress conditions, providing a method for accurate monitoring of wheat growth and precise irrigation management under water deficit conditions. The research was published online in Frontiers in Plant Science.

Timely and accurate pre-harvest prediction of yield under different water stress conditions enables rapid and non-destructive assessment of the impact on wheat growth. The researchers used time-series multispectral and thermal infrared image features collected by the UAV as predictors of elastic network regression to build yield prediction models. In addition, the entropy weight ensemble algorithm was used to weight average the yield prediction values of multiple growth periods. Among multiple development stages, thermal infrared and multispectral data achieved yield prediction accuracy at the filling and flowering stages, and the combination of yield prediction values from multiple development stages resulted in higher prediction accuracy than any single stage. This study provides a reference for high-throughput analysis of wheat yield traits and introduces a scientific method for precision irrigation management in agricultural production.

The research was carried out jointly with Xiao Yonggui's group at the Institute of Crop Science of CAAS. The research was funded by the technology innovation program of CAAS (CAAS-ZDXT-2019002) and key grant technology project of Xinxiang City of Henan Province (ZD2020009).

Link:

https://doi.org/10.3389/fpls.2021.730181

By Chen Zhen (chenzhen@caas.cn)