Estimation of citrus yield from canopy spectral features determined by airborne hyperspectral imagery

Hyperspectral imagery has become increasingly available in recent years and this has necessitated the evaluation of its potential for crop monitoring and precision agriculture applications. The potential of using airborne hyperspectral imagery to develop yield prediction models for citrus fruits was examined in this paper. Hyperspectral images in 72 visible and near-infrared (NIR) wavelengths (407-898 nm) were acquired over a citrus orchard in Japan by an Airborne Imaging Spectrometer for Applications (AISA) Eagle system. The canopy spectral features of individual trees were identified using pixel-based average spectral reflectance values at various wavelengths from the acquired images, which were then used to develop yield prediction models. Yield prediction models were developed using three different techniques: (i) three commonly employed vegetation indices, i. e. the normalized difference vegetation index (NDVI), simple ratio (SR) and photochemical reflectance index (PRI); (ii) a few significant wavelengths; and (iii) partial least squares (PLS) regression factors. Greater prediction accuracy was obtained with PLS models than with the models based on NDVI, SR or PRI, or the significant wavelengths. PLS models showed a significant correlation between hyperspectral imagery data and actual citrus yield for data acquired in 2003 and 2004. These results confirmed the hypothesized correlation between canopy spectral features and citrus yield. This information is valuable for forecasting yields, planning harvest schedules and generating prescription maps for the application of tree-specific alternate bearing control measures and management practices.