Significant wavelengths for prediction of winter wheat growth status and grain yield using multivariate analysis

In order to select significant wavelengths related to winter wheat growth characteristics, field experiments were conducted in three consecutive years. Diffuse reflectance of crop leaves was recorded with other crop variables during growth stages. Multivariate analysis including partial least squares regression (PLSR) and stepwise multiple linear regression (SMLR) procedures were used to determine important wavelengths. The results showed strong relationships between predicted and actual crop variables. The best prediction model built on wavelengths selected by SMLR so that R2, root mean square error (RMSR) and relative error (RE) for the validation dataset were 0.85, 1.56 and 3.64% for SPAD, 0.89, 413 and 6.21% for grain yield, and 0.84, 0.56 and 4.85% for protein content.