Evaluation of neural network modeing to calculate well-watered leaf temperature of wine grape

Mild to moderate water stress is desirable in wine grape for controlling vine vigor and optimizing fruit yield and quality, but precision irrigation management is hindered by the lack of a reliable method to easily quantify and monitor vine water status. The crop water stress index (CWSI) that effectively monitors plant water status has not been widely adopted in wine grape because of the need to measure well-watered and non-transpiring leaf temperature under identical environmental conditions. In this study, a daily CWSI for the wine grape cultivar Syrah was calculated by estimating well-watered leaf temperature with an artificial neural network (NN) model and non-transpiring leaf temperature based on the cumulative probability of the measured difference between ambient air and deficit-irrigated grapevine leaf temperature. The reliability of this methodology was evaluated by comparing the calculated CWSI with irrigation amounts in replicated plots of vines provided with 30, 70 or 100% of their estimated evapotranspiration demand. The input variables for the NN model were 15-minute average values for air temperature, relative humidity, solar radiation and wind speed collected between 13:00 and 15:00 MDT. Model efficiency of predicted well-watered leaf temperature was 0.91 in 2013 and 0.78 in 2014. Daily CWSI consistently differentiated between deficit irrigation amounts and irrigation events. The methodology used to calculate a daily CWSI for wine grape in this study provided a real-time indicator of vine water status that could potentially be automated for use as a decision-support tool in a precision irrigation system