Effect of multi-level and multi-scale spectral data source on vineyard state assessment

Leaf water potential (LWP) is widely used to assess plant water status and it is commonly used by growers to make immediate crop and water management decisions. However, LWP measurement via direct method presents challenges as it is labour, time intensive and represents leaf-level conditions for sampling of small vineyard block. An alternative approach is using pigment concentration as a proxy for the canopy’s water status. Spectral data methods have been applied to monitor and evaluate crops’ biophysical variables. In this study, a model to predict LWP using via UAS equipped with a VIS-NIR multispectral camera and trained machine learning algorithm, is developed and tested.  The model was tested on three dates in 2020 in a commercial vineyard in the Tufo Wine Region. Three modelling approaches (partial least square regression PLSR, support vector machine SVM, artificial neural network ANN) and two input datasets (combining spectral data and spectral vegetation indices) were used to estimate LWP. All approaches predicted LWP-based on spectral data classified from high to low; the results were consistent in direct proportion to the laboratory results and performed the best results. This research shows the potential for estimating LWP at a vineyard scale based on UAS information, represents a good and relatively cheap solution to assess plant status spatial distribution and therefore it could provide a direct way to achieve precise agricultural vineyard.