Evaluating photosynthetic pigment contents of maize using UVE-PLS based on continuous wavelet transform

Abstract

Estimation of photosynthetic pigments status using hyperspectral remote sensing technology is highly significant for the effective evaluation of maize (Zea mays L.) productivity. Effective mathematical statistics and modeling methods can improve the accuracy and reliability of pigment contents. In this study, field experiments were conducted with different water and nitrogen levels to develop a quantitative assessment model that determines chlorophyll (Chl) and carotenoids (Car) contents in maize. A pigment prediction model was established by vegetation indices (VIs), wavelet feature, uninformative variable elimination (UVE), and UVE based on continuous wavelet transform (CWT-UVE) and then compared with various methods. The wavelet feature db1 of maize leaf reflectance had a better correlation with Chl and Car with correlation coefficients (r) of −0.9247 and −0.8705, respectively. By using a calibration set, the best results for monitoring the Chl and Car status were observed using the rbio3.3-UVE-PLS, which indicated fewer variables. The R2 and RMSE were respectively 0.9435 and 0.0937 for Chl, 0.8807, and 0.0130 for Car. In addition, tests conducted on the independent validation set showed that rbio3.3-UVE-PLS could be used to estimate the Chl (R2 = 0.8534; RMSE = 0.1773; RPD = 2.6458), and bior3.3-UVE-PLS can be used to estimate the Car (R2 = 0.8299; RMSE = 0.0187; RPD = 2.3832). Thus, Chl and Car are accurately monitored with the CWT-UVE-PLS method. This study provides an important support method for the practical application of photosynthetic pigment contents analysis in maize.