Development of a plant-based strategy for water status monitoring and stress detection in grapevine

Water shortage has become a major problem, leading to a growing interest for efficient and precise irrigation scheduling even in areas that were completely rain-fed so far. Appropriate irrigation for grapevines (Vitis vinifera L.) is not exclusively a story of fulfilling water demand, but rather of defining the optimum level and timing and having a good knowledge of the grapevine water status. Specific levels of soil water deficit at specific times in the growing season are known to play a key role in the production of high quality grapes and resulting wines, but both severe and no drought stress are not desired as they negatively influence the grape’s and wine’s potential. Innovative techniques for monitoring the plant water status and for applying an adequate irrigation scheduling are required to achieve this crucial water balance for a grapevine. It is internationally recognised that such tools should rely on plant measurements, as they provide information on the actual plant water status, rather than be based on soil or microclimatic measurements. The aim of this thesis was to develop and evaluate a strategy for water status monitoring and stress detection in grapevine based on automated plant measurements. To this end, both experimental and modelling work was carried out on potted grapevines that were subjected to conditions ranging from fully irrigated to severe drought. Two different plant-based monitoring approaches were tested and compared. In a first approach, an accurate monitoring of the grapevine water status and a fast detection of drought stress (i.e. several days before the first clear visible symptoms appeared) were accomplished using two data-driven models: Unfold Principle Component Analysis (UPCA) and Functional Unfold Principle Component Analysis (FUPCA). These models were originally developed for statistical process monitoring of multivariate data sets where accurate mechanistic knowledge is lacking or difficult to achieve. In this study, the multivariate data set consisted of measured microclimatic variables and a plant measurement that served as indicator for plant water status, either sap flow rate or stem diameter variations. Using a large amount of data from well-watered conditions, the models extracted the information and patterns underlying these measured variables and made a profile of normal, expected data behaviour under sufficient water availability. Monitoring new data then implied checking these data against this pattern. When a discrepancy between new data and this normal pattern was observed, the models indicated abnormality, which was in this study related to a deviating water status or drought stress. Unlike the data-driven approach in which a priori information on underlying plant mechanisms was not crucial, the second approach focused on developing a comprehensive mechanistic water transport and storage model for grapevine. This mechanistic model mathematically describes the axial and radial water transport and stem diameter dynamics of grapevine. The basic principles originated from an existing tree water transport and storage model, which enabled among others accurate simulations of the stem water potential (Ψstem) under well-watered conditions, which is one of the best indicators for plant water status. To obtain better drought response simulations with the model, the constant hydraulic plant resistances were replaced by equations in this PhD study. Both the integrated hydraulic resistance experienced during upward water transport through the soil-to-stem segment (RX) and the hydraulic resistance encountered during radial water transport between xylem and elastic living tissues (RS) were dependent on soil water potential. Modelled and measured data were compared to verify the implemented mechanisms. The mechanistic model was applied twofold. First, the model contributed to our understanding of grapevine functioning during drought conditions, as it revealed new insights. Despite the generally assumed constant RX and RS behaviour in several other plant models, the improved model demonstrated that both RX and RS showed daily fluctuations and, superimposed on these fluctuations, exponentially increased when drought progressed. Furthermore, it was shown that mean turgor in the elastic storage tissues rapidly decreased with drought. Finally, an in situ soil-to-stem vulnerability curve that integrated the hydraulic conductance in soil and plant (KX = 1/RX) was generated using the model. Such a curve depicts the loss in KX as a function of declining Ψstem and is often applied in the literature to assess vulnerability of species to drought. Second, the mechanistic model was elaborated as a tool to monitor grapevine water status in real-time. Except under most severe drought stress conditions, which are not favourable for grape and wine quality and should be avoided in practice, the model simulated Ψstem well and kept a tight supervision over the grapevine water status, as Ψstem could be continuously compared against expected plant behaviour defined under well-watered conditions. Simulated Ψstem, representing the actual water status of the grapevine, were then compared with a dynamic threshold beyond which the grapevine is considered to experience drought stress. In this study, the uncertainty band on the dynamic threshold estimation was used to represent the range within which Ψstem was expected to occur under well-watered conditions. Two different dynamic Ψstem thresholds were tested: an approach using vapour pressure deficit (VPD) as input, and a more elaborate approach using potential evapotranspiration (λEp). The latter includes VPD and radiation, both known as key driving variables for plant transpiration. The use of both the VPD- or the λEp-based dynamic threshold and uncertainty band allowed a fast detection of drought stress and tight supervision over the plant water status during a drought experiment on grapevines. To conclude, both the data-driven and the mechanistic modelling approach were judged promising as plant-based strategy for monitoring the grapevine water status. To apply these strategies for optimising grape and wine quality in practice, some challenges remain. As all experiments in this study were conducted on potted grapevines, future experiments should test the performance of the models under field conditions. In addition, the exact impact on the grape berries of different drought levels at specific times during the growing season should be investigated, in order to be able to steer grape and wine quality in the future.