Evaluation of within-season grapevine evapotranspiration patterns and drivers using generalized additive models

Evapotranspiration (ET_c) levels are influenced by the area of plant canopy, atmospheric conditions, plant physiology, and soil-water relations, which are all subjected to changes throughout the growing season. Understanding temporal trends, variability, and interactions between ET_c and its governing factors is valuable for modeling, predictions and vineyard water management. Our research objective was to quantify temporal patterns of ET_c of ‘Cabernet Sauvignon’ grapevines and affecting meteorological (temperature, relative humidity, radiation, wind speed) and vegetative (leaf area) factors during the growing season and within five phenological growth stages (0–4). Temporal variability of ET_c was modeled using five consecutive seasons of lysimeter time-series data, capturing the non-stationary nature of the data in terms of seasonality, trends and within-seasonal patterns. The temporal dependence of ET_c and its drivers throughout growing seasons was evaluated using the Box-Pierce test, autocorrelation function (ACF) and partial ACF. Patterns of the relations between ET_c and its covariates were quantified using multiple nonlinear regression, the generalized additive model (GAM), at the full growing season scale and for each phenological stage. Further examination on the effect of leaf area on ET_c was conducted using area under curve calculations and ET_c-leaf area ratio. The results demonstrate temporal autocorrelation structure of the data, supporting the incorporation of time variables in the GAM. Each phenological stage had a unique composition of relative importance of the covariates, with variation in ET_c being largely explained by time variables. Ordinarily, ET_c in early season (Stage 0) and at late season (Stage 3, approaching harvest) was mostly influenced by leaf area dynamics, while in mid-season it was highly affected by temperature. The GAM enabled quantification of within-seasonal patterns of interrelations between covariates and their effects on ET_c, and revealed inter-seasonal variability due to dissimilar meteorological conditions. Agro-technical management affects canopy dimensions and structure, thus influencing ET_c-leaf area relations.