A cyber-age approach for global management of barley yellow dwarf virus in winter wheat

Open Access
Walls, Joseph Thaddeus
Graduate Program:
Master of Science
Document Type:
Master Thesis
Date of Defense:
October 14, 2014
Committee Members:
  • Cristina Rosa, Thesis Advisor
  • John Frazier Tooker, Thesis Advisor
  • Edwin George Rajotte, Thesis Advisor
  • Joseph Russo, Thesis Advisor
  • Barley yellow dwarf virus
  • decision support systems
  • winter wheat
  • aphid
  • iPIPE
  • dependency network
  • integrated pest management
Precision agriculture often uses computer-based decision-support systems (DSS’s) to disseminate pest and disease information to growers to more efficiently manage agricultural productions. In this thesis, a DSS is developed to be accessed by growers for management of barley yellow dwarf disease caused by Barley yellow dwarf virus. The disease devastates grain growing regions around the world in epidemic patterns. It is a well-studied disease, but management can be greatly improved with determination of necessity and optimal timing of insecticide-treated seed, planting date, pest scouting, and foliar insecticide spray treatments, in chronological order. Using published literature and interviews with experts in BYDV epidemiology and agricultural decision-making, dependency networks were used to model field conditions that would logically warrant these management actions. The networks represented nine possible outputs: use insecticide-treated seed, use untreated seed, plant crop immediately, delay planting, scout for aphid vectors of BYDV, do not scout, full foliar insecticide spray, ½ (diluted) insecticide spray, and no insecticide spray. There were a total of 243 total combinations of conditions to reach the seed treatment recommendations, 9,720 to reach the planting date recommendations, 62,208 to reach the scouting recommendations, and 216 to reach the insecticide spray recommendations. In this work, I consider and strive to improve the mechanism for inferring output recommendations even when using only partial data sets. Inference mechanisms are necessary components of DSS’s to extrapolate outputs from input data to give users recommendations. The dependency networks represent inference mechanisms that require all input information be present before a management recommendation can be made. This thesis proposes a novel secondary inference mechanism structure to be overlaid onto the dependency networks that uses a numerical, rather than categorical or ordinal, calculation system to handle partial input information. This inference mechanism used the dependency networks as a template to make a prototype numerical representation of importance of field conditions in making management decisions. Secondly it calculates a likely success of these management decisions when executed, and penalizes the grower if he or she executes an incorrect management tactic. The success or penalties are measured in terms of optimum yield. It is also proposed that this secondary inference mechanism can allow a BYD management decision forecast based on pest and disease statuses, as well as real-time recommendations. The purpose of the DSS developed in this thesis is to show the applicability of implementing DSS’s based on expert knowledge into a platform (iPIPE) that is capable of gathering data from users in a two-way feedback loop. Since future management decisions rely on previous ones the feedback loop allows management practices conducted by the grower to alter future management recommendations given by the system. It also enhances large scale (regional) pest monitoring with input of individual field data. The DSS reported in this thesis will serve as a basis for the evolution of precision management of crop diseases. It will aid in reducing input cost and increasing sustainability and cereal grain yield in BYD management and eventually it will serve as model to better manage other crop pests and diseases.