Deficit irrigation model-based control of plant growth

Abstract

Listen

Translate article

In the field, irrigation is commonly implemented by the control of water availability in the soil, or in more recent approaches also by measuring the joint water consumption of plant and soil (evapotranspiration) to derive irrigation time and quantity. Here, plant dynamic behavior is not (thoroughly) considered, neglecting the plants’ ability to cope with hostile conditions (deficit irrigation). Plant-based approaches have the potential to reflect plant dynamics but are still subject of scientific studies. Here, the target is mostly to detect stress incipience indicating the moment of irrigation initiation, irrigation quantification is is not (yet) possible based on these approaches. Applying control engineering perception to the irrigation task, the introduction of a suitable model of plant dynamics into irrigation approaches allows directly for plant growth control by means of water input variation. Existing (crop) models (input: agricultural inputs (soil, fertilizer etc.), output: growth/yield) consider plant dynamics due to water deficits very restrictively in form of linear yield reduction functions. A model of plant adaptive behavior including the ability to memorize stress is not available. Hence, a deliberately planned deficit irrigation is not feasible. In this thesis a novel approach to control plant growth based on water stress training is presented. Here, training is denoted as intended sequencing of water deficit events to control growth behavior. The presented approach is based on deficit irrigation control experiments resulting in a state machine model. In this model, plant adaptive behavior is described in terms of stress level (water deficit detected by the plant), stress memory level (adaptive response aimed at future stresses), and damage level (upper boundary of productive water stress). A quantitative distinction between stress levels is introduced by two new plant-based thresholds: response and recuperation threshold. Open-loop control options are based on the sequencing of two experimentally detected growth performance ranges, i.e. ’hydrological time’-based growth in not memorized states, and ’usage-bound growth’ in states with memory. Growth performance in ’usage-bound’ range is 47 % higher than in ’hydrological time’-based range. The memorization is proved to be reversible after three days. Statistical hypothesis testing resulted in a proof of concept for the proposed approach. Additionally, the potential of leaf temperature oscillation analysis for remote water stress state identification is presented. The application of explorative frequency analysis methods on leaf temperature data from the same control experiments resulted in a detection of different superimposed oscillation frequencies. An assumed inference to stomatal oscillation behavior due to water stress could not be proved. However, new insights into leaf temperature behavior based on standard frequency analysis tools are given and discussed regarding application for irrigation purposes.