Methods to compare the spatial variability of UAV-based spectral and geometric information with ground autocorrelated data: a case of study for precision viticulture

One of the key steps that would lead winegrowers to implement precision viticulture as a management tool would be the clear demonstration of the agronomic and oenological significance of the zones delineated within a vineyard based, totally or partially, on remote-acquired information. To perform this analysis, it is necessary to compare image-derived variables to crop characteristics. Classical ordinary least square (OLS) regression is not well suit for spatially structured data, while Moran’s index (MI) and local indicators of spatial autocorrelation (LISA) take autocorrelation into account. The aim of this work was to evaluate the performance of statistical methods to compare different maps of a vineyard, some including variables derived from UAV acquired imagery, and some from in situ ground characterization. The study was conducted during 2015 and 2016 seasons in an adult 7.5 ha cv. ‘Tempranillo’ vineyard located in Traibuenas, Navarra, Spain. The maps obtained out of UAV-imagery, volume index (VI) and normalized difference vegetation index (NDVI) were compared to the maps obtained for the agronomic variables measured (yield, berry weight and total soluble solids). The bivariate MI and the bivariate LISA cluster map obtained using Geoda software indicate depict the spatial cluster association between variables in 2015 and 2016 with different types of local spatial autocorrelation. The use of these methods that take into account data spatial structure, to compare ground autocorrelated data and spectral and geometric information derived from UAV-acquired imagery has been proved to be highly necessary and advisable.