Hosted by Utah State University Libraries
Analysis of Irrigation Decision Behavior and Forecasting Future Irrigation Decisions
Abstract
dc:description.abstractFarmers play a pivotal role in food production. To be economically successful, farmers must make many decisions during the course of a growing season about the allocation of inputs to production. For farmers in arid regions, one of these decisions on any given day is whether to irrigate. This research is the first of its kind to investigate the probable reasons that lead a farmer to make irrigation decisions and use those reasons/factors to forecast future irrigation decisions. This study can help water managers and canal operators to estimate short-term irrigation demands, thereby gaining information that might be useful to more efficiently manage irrigation supply systems. This work presents three approaches to study farmer irrigation decision behavior: Bayesian belief networks, decision trees, and hidden Markov models. All three models are in the class of evolutionary algorithms, which are often used to analyze problems in dynamic and uncertain environments. These algorithms learn the connections between measured input and output data and can make predictions about future events. The models were used to study irrigation decisions of a set of farmers in the Canal B command area, located in the Lower Sevier River Basin, Delta, Utah. Alfalfa, barley, and corn are the major crops in this area. Biophysical variables (plant, soil, canal flow, and weather conditions) that are measured during the growing seasons were used as inputs to build the models. Information about crop phenology (growth stages), soil moisture, and weather variables were compiled. Information about timing of irrigation events was available from soil moisture probes (which measure soil moisture content) installed on some irrigated fields at the site. The models were capable of identifying the variables that are important in forecasting an irrigation decision, classes of farmers, and decisions with single and multi-factor effect regarding farmer behavior. The models did this across years and crops. The advantage of using these models to study a complex problem like behavior is that they do not require exact information, which can never be completely obtained, given the complexity of the problem. This study uses biophysical inputs to forecast decisions about water use. Such forecasts cannot be obtained satisfactorily or in a cost-effective manner using survey methodologies. The study reveals irrigation behavior characteristics. These conform to previous beliefs that a farmer might look at crop conditions, consult a neighbor, or irrigate on a weekend if he has a job during the week. When presented with new data, these models gave good estimates for probable days of irrigation, given the past behavior. All three models can be adequately used to explore farmer irrigation decision behavior for a given site. They are capable of answering questions related to the likely driving forces behind irrigation decisions and the classes of subjects involved in a complex process.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy (PhD)
- Level thesis:degree_level
- Dissertation
- Discipline thesis:degree_discipline
- Civil and Environmental Engineering
- Grantor dc:publisher
- Hosted by Utah State University Libraries
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Andriyas, Sanyogita
- Contributors dc:contributor
-
- Mac McKee
- Wynn R. Walker
- Christopher M. U. Neale