How to Sample to Calibrate a Spatial Model: Application for Modelling Vine Water Status

Several studies have shown that vineyards are spatially variable at within field scale. Characterizing this variability could be used to improve grape and wine production, by adapting management practices and harvesting to the vineyard variability. The goal of this work is to calibrate a spatial model by coupling high quality, high cost punctual vine measurements with low cost medium-high density ancillary data sources. For operational constraints, the number of measurements required to calibrate the model have to be as small as possible. The approach we propose is tested on a particular but significant example: the spatial variability of vine water status. Exposed leaf area (ELA), trunk circumference (TC) and Pre-dawn leaf water potential (PLWP) were measured in 49 sites in a non-irrigated Syrah block. A sampling method was applied to rank these sites according to their suitability to represent the statistical variability on ELA and TC. The first selected sites were used for the calibration of a model relating ELA and TC with PLWP. The model was validated with the remaining sites. The proposed method allowed the model calibration using only 3 sites. From 3 to 9 sites, the model performance improved for each additional site used for calibration from a Standard Error of Prediction (SEP) of 0.12 MPa to 0.10 MPa. The block statistical variability was well represented with the first 9 chosen sites. This study proves the suitability of the approach to calibrate a spatial model describing plant water status spatial variability. The obtained site selection showed to be well suited to represent the within field variability and optimize site number and location for the calibration of the model. The work deals with an important operational constraint for the commercial application of spatial models: reducing the number of measurements needed for their calibration. This is a first, essential step for vine growers to be able to spatially optimize vine management and harvest.