Automatic assessment of agro-environmental indicators from remotely sensed images of tree orchards and its evaluation using olive plantations

Key agronomic and environmental characteristics of tree orchards can be automatically assessed from remote sensing images by a computer program named Clustering Assessment® (CLUAS). The aim of this paper is to describe the CLUAS software development and the information generated by CLUAS for selected olive orchards and its validation with ground-truth data. CLUAS works as an add-on of ENVI®, and operates integrating the digital values (DV) of the neighbouring pixels within a defined range of DV. In the orchards plots trees, other vegetation cover and bare soil were the land uses considered and the range of digital values (BDV) which best define each of them determined. CLUAS provides parameters of each tree, such as the geometric centre, the number of pixels or area, and the integrated digital values or relative potential yield. CLUAS also characterizes key parameters of tree groves, such as the total area and the number, area and the relative potential productivity of the whole trees; and similarly for the other land uses such as vegetation cover and bare soil. Remote images with spatial resolution from 0.25 to 1.5m were suitable for olive grove characterization. CLUAS can contribute to the site-specific management of tree groves, providing quantitative information on each tree, small areas of an orchard, or whole orchards. Ground-truth data taken in an olive orchard of about 2ha at the Experimental Station of Cabra (Cordoba, Spain) in 2004 and 2005 and remote images of the same zone were studied for validation purposes. The wavebands green, NIR (near-infrared), panchromatic and the vegetation indexes normalized difference vegetation index (NDVI) and ratio vegetation index (RVI) were selected for olive grove image assessment by CLUAS. The average size of olive trees was similar regardless of the waveband or vegetation index used, varying from about 23.5-26.0m², and significantly interrelated between each other at 99%. Olive tree area and potential yield estimated by remote sensing were also highly related to the olive tree area estimated on the ground, with significant correlation coefficients at 99% varying from 0.62 to 0.82 and 0.52 to 0.74 in 2004 and 2005, respectively. On the other hand, olive tree size and potential yield estimated in green, panchromatic, NDVI and RVI images were significantly related to the ground-truth yield with correlation coefficients of around 0.50 in 2004, an “on” year, and of 0.30-0.40 in 2005, an “off” year, respectively.