Deciphering maize vertical leaf area profiles by fusing spectral imagery data and a bell-shaped function

Abstract

Leaf area profiles (LAP) represent the green leaf area per unit ground area distributed with the vertical leaf layer, which is a key trait for guiding nutrition diagnosis, crop management and crop breeding. However, passive mono-angle optical sensors don't have direction information on vertical LAP, which makes spectral remote sensing can't capture the canopy-scale vertical leaf area information from top to bottom. To meet this challenge, we present a modeling framework to decipher maize vertical LAP from spectral imagery data. It first employed a hybrid method to derive LAI from the spectral imagery, and then configured a bell-shaped function to decipher the vertical LAP. We conducted a five-year field experiment in critical growth stages to test the ability of the proposed method. Results showed great disagreements between vegetative and reproductive stages. Such differences impacted the leaf area development and the largest leaf layer for LAP modeling. The proposed method considered these two phenological stage to improve the LAP estimation. The performance of this method was assessed by comparing the derived vertical LAP with measurements over different planting years and maize grain production fields. Results showed robust canopy-level modeling for vertical LAP (RMSE = 0.083–0.094 m2/m2). This study highlights that this method extends the ability of passive optical remote sensing to derive vertical information. This method is a valuable and effective remote-sensing approach for deriving vertical LAP over maize canopy scale, also has potential reference value for other vegetation with similar vertical structure.