Estimating green LAI in four crops: Potential of determining optimal spectral bands for a universal algorithm

Abstract

Vegetation indices (VIs) have been used previously for estimating green leaf area index (green LAI). However, it has not been verified how characteristics of the relationships between these indices and green LAI (i.e., slope, intercept, standard error) vary for different crops and whether one universal algorithm may be applied for accurate estimation of green LAI. By analyzing the data from four different crops (maize, soybean, wheat, and potato) this study aimed at: (1) determining if the previously used VIs for estimating green LAI in maize and soybean may be applicable for potato and wheat and vice versa; and (2) finding a robust algorithm for green LAI estimation that does not require re-parameterization for each crop. Spectral measurements of wheat and potato were obtained in Israel from 2004 to 2007 and of maize and soybean in the USA from 2001 to 2008, and various VIs calculated using measured reflectance were compared with green LAI measured in the field. For all four crops, ten different VIs were examined. Similarities in relationships between VIs and green LAI were found. Among the examined VIs, two variants of the chlorophyll index and wide dynamic range vegetation index with the green and red edge bands were the most accurate in estimating green LAI in all four crops. Hyperspectral reflectance data were used to determine optimal diagnostic bands for estimating green LAI in four crops using a universal algorithm. The green (530–570nm) and red edge (700–725nm) regions were identified for both the wide dynamic range vegetation index and chlorophyll index as having the lowest errors estimating green LAI. Since the Landsat 8 – OLI has a green spectral band and the forthcoming Sentinel-2, Sentinel-3 and VENμS have both green and red edge bands, it is expected that these VIs can be used to monitor green LAI in multiple crops using a single algorithm by means of near future satellite missions.