• No se han encontrado resultados

DIAGRAMA DE BLOQUES

5. MODULADOR - EXCITADOR FM

5.2. DIAGRAMA DE BLOQUES

Independently from whether optimal irrigation-controlling parameters or optimal soil and plant parameters are sought, the general procedure for solving optimization problems with EA includes the following steps (Maier et al. 2014):

1. Formulating the optimization problem, which includes the selection of decision variables, objective functions, and constraints.

2. Initialization of the optimization through the selection of sets of random decision variable values by the EA.

3. Evaluation of the objective functions and constraints through simulation runs with the SVAT model.

4. Selection of new and updated sets of decision variable values based on the feedback from the evaluation process using some sort of search method.

5. Evaluation and selection of new sets of decision variable values are repeated (points 3 and 4) until a stopping criterion is met and little to none improvement in the objective functions achieved anymore.

6. The final sets of decision variable values are the solution to the optimization problem.

5 Conceptual Outline

For investigating and improving sensor-based DI systems in this work, a joined approach in the combination of the three components: (i) SVAT-modeling/simulation, (ii) optimization, and (iii) irrigation experiment is used (Fig. 5.1). Through their combination, limitations that arise from only using the individual parts alone for the investigation are mitigated and the wider spectrum of methods for investigating WP in sensor-based DI systems used, thus getting closer to unlocking their full potential.

Irrigation Experiment Sensor

SVAT-Modeling/Simulation

Optimal Irrigation Control soil-water potential

optimal threshold and water amount calibration and

validation pr

ovides obser

vation da

ta model-based

design

applica tion in

realiza tion thr

ough

Publication 1

Publication 3 Publication 2

Optimization

Figure 5.1: Conceptual outline of this work with the three main components (blue boxes) and their combined application for investigating and improving sensor-based DI systems, along with their relationship to each other and their coverage in the individual publications.

The SVAT-modeling/simulation components serves as the basis for a model-based design of the irrigation experiment by employing a SVAT model. The design involves all aspects that describe the irrigation system (e.g., plant and row spacing, plant density) as well as parameters that determine the behavior of the irrigation system. For sensor-based systems, this includes the position or depth of the irrigation-controlling sensor, the number of irrigation thresholds throughout the growing period of the crop, the minimum irrigation interval between individual irrigation events, and others.

A suitable SVAT model can reliably simulate water flow through soils and predict crop production under water-limiting conditions and climate variability. The investigation and selection of suitable candidates of SVAT models that are employed in the framework for investigating DI systems was dealt with in publication 1 – “Evaluation of Crop Models for Simulating and Optimizing Deficit Irrigation Systems in Arid and Semi-arid Countries under Climate Variability” (section 6.1).

The SVAT-modeling/simulation and means of optimization are used to find optimal values of parameters for irrigation control. For sensor-based DI systems, parameters comprise the irrigation threshold at which point an irrigation is triggered and the amount of irrigation water per irrigation event for that particular irrigation threshold. Values are sought according to a certain objective that generally aims at achieving a high WP.

The optimal irrigation control finds its application in the irrigation experiment and realization through a special sensor capable of performing the intended task. Other sensors are used to continuously collect data on weather, soil-water status, plant development, and management during the experiment as it was laid out in section 4.1.1.

A model-based design of the irrigation system and the determination of optimal param-eters for irrigation control of different sensor-based irrigation strategies, their test and verification in an intensively monitored irrigation experiment, the evaluation of influences of those strategies on WP and water consumption along with other parameter that determine the behavior of irrigation systems (depth of irrigation-controlling sensor, irrigation interval) is covered in publication 2 – “Investigation of Deficit Irrigation Strategies Combining SVAT-modeling, Optimization and Experiment” (section 6.2).

Subsequently, collected data is used for model (re-) calibration and validation and thus model improvements of the previously applied SVAT model, mainly by improving parameters on soil and crop. A SVAT model improved in such a way allows for better model accuracy and model predictions and enables the systematical investigation of DI systems. This includes the range of operations of the irrigation system in order to achieve highest WP and thus explore the full potential of DI systems.

A systematical investigation of sensor-based DI systems for parameters that control irrigation (sensor threshold, irrigation amount, irrigation interval) after recalibrating the model with observed data from an irrigation experiment was conducted in publication 3 – “Evaluation of Very High Soil-Water Tension Threshold Values in Sensor-Based Deficit

Irrigation Systems” (section 6.3).

Model-based design of the irrigation system and calibration and validation is an iterative process. Therefore, each new “round” of SVAT-modeling/simulation and irrigation experi-ment will further improve model predictions. Essentially, not only the irrigation control but all aspects regarding the design of the irrigation systems may be included into the optimization.

6 Overview of Publications

6.1 Evaluation of Crop Models for Simulating and Optimizing Deficit Irrigation Systems in Arid and Semi-arid Countries under Climate Variability

This paper investigated the suitability of the four SVAT-models CropWat, PILOTE, Daisy and APSIM for modeling and simulating DI systems. The SVAT models were evaluated with respect to their performance when being part of a stochastic simulation-based approach to investigate and improve WP under limiting water conditions. Focus was hereby primarily put on the effects of climate variability.

The stochastic approach OCCASION was used, which consists of three components: (i) a weather generator for providing regional climate series and thereby adding statistical variability to the investigated objective, (ii) an optimization algorithm that calculates optimal irrigation schedules for any given but limited total amount of irrigation water and respective climate series, and (iii) the SVAT-model for determining water consumption and crop production. Resulting SCWPF were evaluated and discussed for their plausibility since they serve as a tool for assessing the effects and risks on potential yields due to drought stress and climate variability. The study was conducted in conjunction with data from field experiments from India, Malawi, France, and Oman.

The framework OCCASION was successfully applied for all before-mentioned SVAT-models and proved to be a valuable tool when investigation DI systems and their potential to improve WP. The SVAT models CropWat and PILOTE performed unsatisfactorily with respect to obtained CWPF. The former showed no variation in yields for the different climate series once FI was reached due to an inadequate representation of crop production within the model. The latter model lacked a robust crop parameterization, which prevents the model from being transferred and applied to other geographical locations.

The SVAT-model Daisy performed well but showed problems when dealing with severe drought stress. However, this could be disregarded since that range of the CWPF is not of scientific or economic interest. APSIM as well showed plausible results and is a promising candidate for the investigation of DI systems and improvement of WP.

This investigation was limited to climate variability and regarded global radiation, temperature, and precipitation only. However, other yields-affecting factors could find their consideration within the presented approach without increasing the complexity of the optimization problem. OCCASION is a robust tool and viable approach to investigate and improve WP in many environments and for various crop models.

6.2 Investigation of Deficit Irrigation Strategies Combining

Documento similar