The observation of Earth by satellites has demonstrated the feasibility of establishing differences between plant species, from their spectral features. The reflectance spectrum of vine plants follows this trend, being possible to identify vineyards in satellite images, among other species. However, identification at grape variety level is still to be investigated. This was presently addressed, using satellite multi-spectral images of two terroirs at Rio Grande do Sul State, Brazil. Spectral informations for 13 grape varieties (Cabernet-Sauvignon, Merlot, Semillon and others) were extracted from images collected by the ASTER sensor aboard Terra satellite, at 9 bands, with resolutions of 15 m at visible and 30 m at infrared. Field, radiometric measurements provided additional spectra. For one terroir, with vines in rows, 9-points spectra were constructed, each being the average of three plots of a given variety. These spectra are either polynomials, or sets of normalized intensities for the 9 bands. The other terroir, 500 km apart, has smaller plots in the traditional pergola style. Results point that: a) field measurements are compatible with orbital data; b) spectra for one variety, taken from three different plots, are mutually consistent; c) it is possible, from satellite images, to identify varieties, from their respective equations; d) the spectral information is coherent between both terroirs. It is concluded that middle resolution satellite images (pixel 15-30m), especially at infrared, are a valuable tool for surface measurements and grape variety identification, leading to multiple applications, including precision viticulture.
Authors: Jorge Ricardo DUCATI and Patrícia RODRIGUES DA SILVA
Centro Estadual de Pesquisas em Sensoriamento Remoto e Meteorologia
Universidade Federal do Rio Grande do Sul, Av. Bento Gonçalves 9500, Porto Alegre, Brazil
Emails: ducati@if.ufrgs.br; patricia.rodrigues@ufrgs.br
Keywords: remote sensing, ASTER images, image classification, radiometry, vineyard monitoring