Comparison of different methods for corn LAI estimation over northeastern China

Abstract

Leaf area index (LAI) is a crucial variable in all kinds of ecosystem, climate and crop yield models, describing the fluxes of energy, mass and momentum between the surface and the planetary boundary layer. To accurately determine the corn LAI, several methods of LAI estimation have been evaluated in this investigation, including vegetation indices, principal component analysis (PCA), the neural network method (NN), the look-up table (LUT) inversion from PROSAIL model and the Hybrid model. Comparisons were conducted based on field-measured corn canopy hyperspectral reflectance and LAI data over northeastern China. In order to fairly compare the LAI estimation performance of different methods, the ground-measured data were separated into two sets (modeling data and validation data), except the LUT and hybrid methods of PROSAIL-based. The results indicated that the PCA method delivered the best performance for corn LAI estimation (with maximum R2=0.814 and minimum RMSE=0.501) in this study. The hybrid model and EVI provided moderate results. Comparatively, the LUT and NN methods were less successful and NDVI provided the worst corn LAI estimation performance in this study. The PCA method shows great potential for performing well on corn LAI estimation from hyperspectral information. PCA can avoid the reflectance saturation defect of dense canopy in a certain extent, can utilize hyperspectral reflectance data much more effectively than other methods, and is not limited by the band numbers, it can also reduce noise and provide an great correlation with LAI from the hyper- bands or the multi- bands reflectance.