Using neural networks to predict subterranean termite hazard in China

• Main Author: Schmidt, Daniel
• Language: English
• Published: University of British Columbia 2011
• Online Access: http://hdl.handle.net/2429/32351

In this thesis a neural network is used to create a termite hazard map for China. First, a brief overview of the current situation in China regarding the efforts Canadians are making to introduce light wood frame construction and the challenges they are facing. Amongst these challenges, one of the most important is the hazard of termites in China. The most economically important termite species found in China are then described along with methods commonly used to control the spread of these pests. It serves to identify which species are of more relevance to light timber frame structures in order to concentrate the efforts only on these species in creating the hazard map.. Following this, more information on termite control is given and the issue of Persistent Organic Pollutants (POP's) and the alternative integrated pest management (IPM) system are introduced as possible solutions to deal with the termite problem and, at the same time, comply with international environmental agreements such as the Stockholm Convention. The rationale for methods used to predict the hazard of termite attack in the different geo-climatic zones of China using Neural Network technology is then presented. Existing geo-climatic information for different locations in Japan, the United States and Australia, was linked with previously developed survey-based hazard maps for the three countries. This matrix was used to train a Neural Network, to accurately predict the hazard of the two most economically important subterranean termite genera, Reticulitermes and Coptotermes. The development of a subterranean termite hazard map using verified Neural Network techniques reduces the effort of performing extensive surveys and provides important information for designers, developers and researchers. This work is the first attempt to apply Neural Networks in the forecasting of termite hazard. Further exercises could be carried out to improve the methodology used and expand its field of application. === Forestry, Faculty of === Graduate