Use of vegetation indices from Sentinel2 and UAV in precision viticulture applications

Precision viticulture is based on information related to the within field variability of the crop soil sys- tem; this information is used to optimize input supply and management practices, to improve grape quality and yield. Technological tools are available to accurately investigate this variability, through remotely or proximal sensed data. Geophysical surveys are usually employed to investigate soil variability, while canopy variability is monitored through multispectral imagery. In the latter case, vegetation indices are computed from measurements acquired with sensors mounted on different platforms, from sat ellite to unmanned aerial vehicle ( and quad. Sentinel 2 data are free and readily available, while UAV data hav e a cost for both acquisition and processing. Conversely, spatial resolution increases from Sentinel 2 (10 m) to UAV (few centimetres) data. The accuracy in crop row monitoring provided by UAV is required for a site specific and variable rate management of agronomic inputs, or to assess whether a certain intervention has reached the goal However a variable rate irrigation management is often operated not looking at the crop variability in a certain date but mainly considering the soil variability, since soil is the reservoir from which the roots get their water supply. Even though g eophysical methods are the ref erence tools to delineat e the agronomic management zones ( from soil maps also multispectral images time series can be used As multispectral imag es describe crop variability both in space and time, multi temporal images can be used to extract time stationary components, mainly related to soil variability rather than to time vari able factors as plant diseases or agro-meteorological conditions. This research focuses primarily on compari ng Sentinel 2 and UAV data acquired in drip irrig ated vineyards to assess their ability to describe the vine vigour along the rows. Secondly, after demon- strating how Sentinel 2 data are more correlated with the grass inter row vigour strongly affected by soil properties, as grass is not irrigated than with the vine condition the use of Sentinel 2 multi temporal images to delineate MZs is explored. Both objectives were pursued on datasets acquired in the Colli Morenici region (south of Lake Garda, Italy).