Date of Award
Doctor of Philosophy
School of Computer and Security Science
Health, Engineering and Science
Dr Jitian Xiao
Professor Craig Valli
Dr Mike Johnstone
Climate change and the reduction of available agricultural land are two of the most important factors that affect global food production especially in terms of wheat stores. An ever increasing world population places a huge demand on these resources. Consequently, there is a dire need to optimise food production.
Estimations of crop yield for the South West agricultural region of Western Australia have usually been based on statistical analyses by the Department of Agriculture and Food in Western Australia. Their estimations involve a system of crop planting recommendations and yield prediction tools based on crop variety trials. However, many crop failures arise from adherence to these crop recommendations by farmers that were contrary to the reported estimations. Consequently, the Department has sought to investigate new avenues for analyses that improve their estimations and recommendations.
This thesis explores a new approach in the way analyses are carried out. This is done through the introduction of new methods of analyses such as data mining and online analytical processing in the strategy. Additionally, this research attempts to provide a better understanding of the effects of both gradual variation parameters such as soil type, and continuous variation parameters such as rainfall and temperature, on the wheat yields.
The ultimate aim of the research is to enhance the prediction efficiency of wheat yields. The task was formidable due to the complex and dichotomous mixture of gradual and continuous variability data that required successive information transformations. It necessitated the progressive moulding of the data into useful information, practical knowledge and effective industry practices. Ultimately, this new direction is to improve the crop predictions and to thereby reduce crop failures.
The research journey involved data exploration, grappling with the complexity of Geographic Information System (GIS), discovering and learning data compatible software tools, and forging an effective processing method through an iterative cycle of action research experimentation. A series of trials was conducted to determine the combined effects of rainfall and temperature variations on wheat crop yields. These experiments specifically related to the South Western Agricultural region of Western Australia. The study focused on wheat producing shires within the study area. The investigations involved a combination of macro and micro analyses techniques for visual data mining and data mining classification techniques, respectively.
The research activities revealed that wheat yield was most dependent upon rainfall and temperature. In addition, it showed that rainfall cyclically affected the temperature and soil type due to the moisture retention of crop growing locations. Results from the regression analyses, showed that the statistical prediction of wheat yields from historical data, may be enhanced by data mining techniques including classification.
The main contribution to knowledge as a consequence of this research was the provision of an alternate and supplementary method of wheat crop prediction within the study area. Another contribution was the division of the study area into a GIS surface grid of 100 hectare cells upon which the interpolated data was projected. Furthermore, the proposed framework within this thesis offers other researchers, with similarly structured complex data, the benefits of a general processing pathway to enable them to navigate their own investigations through variegated analytical exploration spaces. In addition, it offers insights and suggestions for future directions in other contextual research explorations.
Vagh, Y. (2013). Mining climate data for shire level wheat yield predictions in Western Australia. Retrieved from http://ro.ecu.edu.au/theses/695