Exploring the Optimal Wavelet Function and Wavelet Feature for Estimating Maize Leaf Chlorophyll Content

Ensuring global food security depends heavily on attaining and sustaining high maize yields. Effective N fertilization management and precise predictions of maize yields, however, require accurate and timely estimation of leaf chlorophyll content (LCC). In this study, we determined the optimal spectral features for predicting LCC in maize by comparing spectral indices and wavelet features. The robustness of the wavelet functions in estimating maize LCC was evaluated, and the results showed that LCC was strongly correlated with the wavelet coefficient between 400 and 800 nm, located at higher scales (9 and 10). The best wavelet function for estimating LCC was the Mexican hat (Mexh) continuous wavelet transform (CWT) (W718, S9). Compared with the currently accepted best spectral index model (mND705, R2 = 0.80 -0.95), the LCC estimation model based on the CWT wavelet function (Mexh, R2 = 0.90 -0.98) was more accurate. The newly developed model was validated using two independent datasets, from 2017 to 2018, yielding root mean squared errors of 2.35 and 2.39~μ g/cm2, respectively. The relative errors of LCC estimation obtained by the new model were 3.70% and 3.62%, respectively. Validations based on the PROSPECT model confirmed the robustness and stability of the CWT Mexh function compared to the best-performing spectral indices. In conclusion, the higher estimation accuracy of the Mexh function-based wavelet transform across growth stages, leaf layers, locations, and varieties demonstrated the universality and stability of the wavelet transform approach in estimating maize LCC.