Why does spatial extrapolation of the vine water status make sense? Insights from a modelling approach

Abstract

This work is devoted to precision agriculture and more precisely to the spatial monitoring of water status in viticulture. An empirical approach was introduced in 2008 based on the extrapolation across a domain (vineyard block, vineyard, region) of vine water status observations from a reference site using a simple statistical model, called SPIDER, and proved efficient in many studies. Once the extrapolation model is calibrated, this approach leads to a concentration of measurements for one site only (reference site) while providing an estimate of the grapevine water status at a larger spatial scale. It is a promising hybrid approach based both on regular (but targeted) measurements and on modelling. However, so far only empirical guidelines for its practical use have been provided. Moreover, the limits of validity (spatial, temporal, etc.) of such an approach are not known. This work intends to use a mechanistic model based on grapevine water balance modelling to study to what extent a simulated water status can be spatially extrapolated at the field scale. The water balance model was calibrated on two datasets (different cultivars and weather data) and used to analyse the performances of SPIDER. The results confirmed the relevance of the empirical approach (SPIDER) based on water status spatial extrapolation with a low error level on the two datasets studied. The use of the water balance model also helped define the validity domain of SPIDER: it confirmed the importance of having dominantly dry conditions and revealed the possibility of recovering good prediction quality after strong rainfall or irrigation. This study globally demonstrates the relevance of spatial extrapolation of the vine water status from a reference site with a linear regression model and provides new insights on the properties of the predictions for application in viticulture either at the within-field level or at larger scale.