Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Using neural networks to predict subterranean termite hazard in China Schmidt, Daniel 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_2007-0231a.pdf [ 10.17MB ]
JSON: 831-1.0074935.json
JSON-LD: 831-1.0074935-ld.json
RDF/XML (Pretty): 831-1.0074935-rdf.xml
RDF/JSON: 831-1.0074935-rdf.json
Turtle: 831-1.0074935-turtle.txt
N-Triples: 831-1.0074935-rdf-ntriples.txt
Original Record: 831-1.0074935-source.json
Full Text

Full Text

Using Neural Networks to Predict Subterranean Termite Hazard in Chi by  DANIEL SCHMIDT B.Arch., Universidad Mayor, Chile, 2003  A T H E S I S S U B M I T T E D IN PARTIAL F U L F I L M E N T O F THE REQUIREMENTS FOR THE D E G R E E OF  MASTER OF APPLIED SCIENCE in  THE FACULTY OF GRADUATE STUDIES  (Forestry)  T H E UNIVERSITY O F BRITISH C O L U M B I A December 2006  © Daniel Schmidt, 2006  Abstract  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.  TABLE OF CONTENTS Abstract  ii  Table of Contents  iii  List of Tables  v  List of Figures  vi  List of Illustrations  vii  Acknowledgements  viii  1. Introduction 1.1. Thesis objectives 2. Termites and their Control in China  1 5 6  2.1. Termites in China 2.1.1. Cryptotermes (Kalotermitidae) 2.1.2. Reticulitermes (Rhinotermitidae) 2.1.3. Coptotermes (Rhinotermitidae) 2.1.4. Macrotermes (Termitidae) 2.1.5. Odontotermes (Termitidae) 2.2. Comments on Termites in China 2.3. Termite Control in China  7 8 9 10 11 11 12 13  3. POP's and Integrated Pest Management (IPM)  15  3.1. Overview 3.2. Persistent Organic Pollutants (POPs) 3.3. China and the Stockholm Convention 3.4. Finding Alternatives to POPs 3.5. Integrated pest (termite) management (IPM) 3.6. The Australian Example 3.7. Facilitating the implementation of IPM 3.8. Comments on China, POP's and the Implementation of IPM 4. Developing a Termite Hazard Map 4.1. Background 4.2. Neural Networks 4.3. Materials  15 15 17 17 19 20 22 23 25 25 26 27  iii  4.4. Methodology 4.5. Finding samples 4.6. Training the Neural Network 4.7. Results  29 33 35 40  5. Conclusions  46  Bibliography  48  Appendices  52  Appendix I.  Original Japan Subterranean Termite Hazard Map  Appendix II.  Original North American Subterranean Termite  52  Hazard Map  53  Appendix III.  Original Australian Subterranean Termite Hazard Map  54  Appendix IV.  List of parameters used in figures 5 & 6  55  Appendix V.  Relationship between stakeholders involved in the implementation of IPM and their function Appendix VI. List of Japanese cities used as data locations in Training Networks  56 57  Appendix Vll. List of North American cities used as data locations in Training Networks  58  Appendix VIII. List of Australian cities used as data locations in Training Networks  59  Appendix IX.  List of Chinese cities used as production sample set for the final Network and their hazard classification  60  IV  LIST OF TABLES T a b l e 4.1  P a r a m e t e r s u s e d in c l i m a t e a s s e s s m e n t  LIST OF FIGURES Figure 2.1 Northern Limit for Termite activity in China and Population Density Figure 4.1 Japan Hazard Zones and data locations Figure 4.2 North America Hazard Zones and data locations Figure 4.3 Australia Hazard Zones and data locations Figure 4.4 Linear Correlation Coefficient (r) by exclusion.. Figure 4.5 Linear Correlation Coefficient (r) for each parameter tested individually Figure 4.6 Location of data point for China by colour according to hazard value Figure 4.7 Termite Hazard map for China Figure 4.8 China - Annual Precipitation and Termite Hazard Figure 4.9 China - Altitude and Termite Hazard  LIST O F ILLUSTRATIONS  Illustration 1.1 Villa in Woodland development, Shanghai  3  Illustration 1.2 Dongjiao Development under construction, Shanghai  4  Illustration 2.1 Coptotermes formosanus-eaten container in Guangdong Entomological Institute  10  Vll  ACKNOWLEDGEMENTS I want to thank all people that helped putting this work together. First of all, my loving wife Sibylle, for being the best company and support, second, my advisory committee, where Dr. Frank Lam and Dr. Helmut Prion accepted me as a graduate student working under their supervision and constantly advised and guided this lost architect in learning about wood and Dr. Paul Morris, who provided with fundamental information and precious time throughout the process and company in an intense journey to China. I also want to thank Dr. Kunio Tsunoda from Kyoto University, Dr. Robert Leicester of CSIRO, Dr. Zhong Junhong of Guandong Entomological Institute, Dr. Li Xiao-ying of Wuxi Institute for Termite Control and many others that contributed with valuable information. Dr. Wang Jieying and Dr. Ge Hua, for being an amazing company and help in visiting their country.  vni  1 INTRODUCTION On September 17 2001, China joined the World Trade Organization (WTO) tn  after more than 15 years of negotiations (Panitchpakdi & Clifford, 2002). The most populous country in the world opened new doors for international trade and agreed to undergo a process of modernization of the government administration and legal system to comply with international standards. China has since emerged as one of the biggest traders of products in the world and in 2004 became the third largest merchandise trader (WTO, 2005). Canada, as most of the countries in the world, sees this as an opportunity to expand its trading with China. Forest products are one of Canada's most important exports, accounting for 8% of the total export in 2005 (Statistics Canada, 2006). Therefore, it is not surprising to find Canadian initiatives promoting forest products in China and trying to find opportunities to introduce these products into the huge Chinese market. The emphasis is being put into those products that are more likely to have a large scale impact and most benefit to the forestry sector. In 2004, softwood lumber and wood panels combined accounted for close to 40% of the Canadian forest product's export (CFS, 2005), proving that construction is one of the industries that consumes the largest amount of forest products. However, 80% of all Canadian forest products' export goes to the US, meaning that Canada's forest industries rely heavily on the US and urgently need to diversify its market to mitigate this dependency. China accounts nowadays for only 3% of Canadian forest products' export (Statistics Canada, 2006). The government and the industry recognize that the promotion of light wood frame houses is a key point for the market of wood products in China to grow. Initiatives like the Canada Wood Export Program, that puts together the government and the industry to promote Canadian wood products overseas, is a response to this necessity. According to a report by Canada Wood, there was an increase in the consumption of wood products in China that coincided with the  1  housing policy reform in 1998. This policy reform absolved the state from the cost of providing its people with shelters (Canada Wood, 2003) and allowed employees to buy the apartments they were living in at heavy discounts {Economist, 2000). In 2004, the cycle of privatization was completed and despite the continuous endorsement of Marxist principles among communist Party officials, a constitutional amendment that described private property rights as inviolable was issued and approved by the Chinese communist party {Economist, 2004 (a)).  Additionally, a large amount of the unusually large Chinese rural population is migrating towards the cities. Many of the farmers' tiny land-plots are being merged to increase returns on agriculture and boost incomes. The surplus labor is moving into urban areas to work in the manufacturing facilities {Economist, 2003).  The housing market engine in China is moving, not only because of the millions of countryside people willing to step into the city, but also because of the millions of citizens wanting to leave overcrowded family homes. This creates a demand for wood products to be used in construction. Interior decoration products, including a variety of value-added products such as furniture, moldings and decorative floor paneling, are the most sought after. Also, the demand for softwood lumber and panels has increased due to the greater number of wood-frame houses that are being built (Canada Wood, 2003). Residential areas of luxury houses commonly known as "villas" have proliferated in and around many Chinese cities. The prices of these houses, characterized by its eclectic architecture, often copying extremes of western style, can go up to US$4 million (see Illustrations 1.1 & 1.2).  2  Illustration 1.1 Villa in Woodland development, Shanghai  Source: Picture taken by the author, Feb. 2006 The number of Chinese villas and luxury apartments built with concrete reached nearly one hundred thousand in 2002. This is about 35% more than in 2001 and much more than double the figure for 1999 (Economist, 2004(b)). Wood is slowly being accepted as a reliable construction material. Cost comparison analysis revealed that on-site construction and Chinese panelized wood-frame  single family  homes are cheaper than traditional  concrete  construction (Cartwright et al., 2002). A code for the design of timber structures was introduced in 2005 to facilitate the engineering design of the platform building system.  3  Illustration 1.2 D o n g j i a o d e v e l o p m e n t u n d e r construction, S h a n g h a i  an  S o u r c e : Picture t a k e n by the author, F e b . 2 0 0 6  However, buildings that use wood as their main structural material are still a rarity in China and building crews and architects are not very experienced in building with wood. Also, wooden homes are affordable to the upper and upper middle class segments of the population only (FM, 2003). Although this could be seen as a strategy to promote the use of wood appealing to the desire of general masses to live like the rich, this effect is not yet noticeable. Recommendations have been made to find solutions to the lack of experience in wood building in China. Many have stated that Canada should get involved in outlining the roles of the major market players such as industry, research and development institutes, academia and the provincial and federal governments.  4  Some also recommend involvement in technical training in both design and construction and also to participate in the development of a quality assurance program for wood-frame construction in China (Cartwright et al., 2002). The introduction of a new building system is not an easy task. It is not only a matter of translating the code and start building with wood. It requires a process of adaptation. There are many factors that differ between countries and there are also various aspects to consider within the wide range of geo-climatic zones of such a vast country as China. Even considering that light wood frame construction has a long record of excellent performance in Canada and many other regions of the world, this form of construction, as explained above, is very new to China and its performance is yet to be proven. Termite hazard is one of the factors that differentiate China from Canada. This one factor impacts directly on construction. 1.2 Thesis Objectives  This thesis intends to summarize the most important aspects concerning the termite situation in China today. With the possibility of light wood frame construction being introduced into China, there is an urgent need to acquire as much information as possible on all aspects that could have an influence on the performance of the system. This is the main reason why the issue of what the problems of introducing a new construction system could be is raised in this thesis. Termites need to be dealt with before wooden buildings are constructed in China. Current termite management practices in China are out of date and pose a serious threat to the environment.  5  While gathering background information, the issue was raised that in order to implement any termite management strategy, it is fundamental to define hazard areas so that any method used is targeted to the area where it is most needed. As termite hazard maps are already developed for other parts of the world, the idea of linking the already defined hazard areas in other regions to the Chinese territory became the main goal of this study. The use of a Neural Network platform seemed the best alternative to link any parameter found in those regions to their respective termite hazard, and climate seemed to be the factor that would most influence its spread. The hypothesis was made that, by training a Neural Network linking defined termite hazard in other regions of the world to the climate variables of those regions, it would be possible to forecast termite hazard in China by simply imputing the climate variables.  6  2 TERMITES AND THEIR CONTROL IN CHINA 2.1 Termites in China From the more than 2000 different species of termites in the world, 476 species assigned to 44 genera are found in China (Zhong & Liu, 2004). Not all of these are considered important pests. Only nine species assigned to five genera in three different families are considered economically important pests. Most sources agree in their description of the northern limit for termite activities in China (see Figure 2.1). This limit represents the accepted northern edge for all termite species in China. (Li, 2002; UNIDO, 2003; Zhong & Liu., 2004) As shown in Figure 2.1, termite activities are predominant in the area where population is concentrated.  Figure 2.1 Northern Limit for Termite activity in C h i n a a n d P o p u l a t i o n Density  Source: Li, 2002; U N I D O , 2003; Zhong & Liu., 2004 F i g u r e d e v e l o p e d by this author.  7  This coincidence of termite distribution  and demographics explains the  economical relevance of termites as pests in China. A detailed description of the relevant characteristics and behavior of the above listed termite genera and their respective species is given next. 2.1.1 Cryptotermes (Kalotermitidae)  All termites in the family Kalotermitidae are considered wood-dwellers, meaning that the colony is confined throughout its life to wood, which is usually above ground (Roonwall, 1970). The genus Cryptotermes is a dry-wood termite and it is also known as the "powderpost or furniture termite" in the United States (Mampe, 1990). According to Zhong (2004) there are eight described species of Cryptotermes found in China and all are restricted to the southern region of the country. Cryptotermes domesticus (Haviland) and C. declivis (Tsai & Chen) are the species among this family that cause the major damage to buildings. Light (1934) refers to these termites as "the true house termites", because of their nesting and feeding habits. Colonies are usually found in bookshelves or other hidden places inside the house. They are rarely found outside. Their distribution is mainly attributed to the transportation of infested furniture. In heavily infested areas, the evidence of their presence is noticeable in furniture, moldings, structural lumber or any other wood application in the house. Mampe (1990) describes the genus Cryptotermes as easily distinguishable from other Kalotermitidae species because of the small size of all castes of the colony and the small size of their fecal pellets. According to Zhong and Liu (2004) Cryptotermes are found in China in the Guizhou, Zhejiang, Sichuan, Yunnan, Fujian, Guangdong and Guangxi provinces and in the islands of Taiwan (C. declivis) and Hainan. Their northern limit is 28.4° Latitude north.  8  2.1.2  Reticulitermes  (Rhinotermitidae)  All termites belonging to the Rhinotermitidae family are considered subterranean. Their nests are normally found underground, but they usually feed on wood that is above ground (Roonwall, 1970) R. speratus (Kolbe), R. chinensis (Snyder) and R. flaviceps (Oshima) are the three species of termites belonging to this family that are considered serious pests in China. According to Mampe (1990), Reticulitermes are the termites that attack timber structures most commonly in the United States. In China, Reticulitermes is also important. Reticulitermes and Coptotermes belong to the Rhinotermitidae family and are considered the most economically important genera. The damage produced by these termites is not limited to structural timber. They are also responsible for damage to crops, living trees, underground cables and other materials (Zhong & Liu, 2004). Reticulitermes are classified among the subterranean termites, meaning they are ground dwelling termites that establish their nests in wood or vegetable material in contact with, or under-ground. This type of termite has the ability to build tubes from earth and their saliva, and reach their food source undercover. These tubes are rarely built in the open. Colonies are normally established in hidden places, where wood is in contact with the ground or is directly accessible through cracks or mortar joints (Mampe, 1990) Reticulitermes is by far the most widely distributed species in China and it is found everywhere from the 40°N latitude (Beijing) to Hainan Island at 18.5°N latitude (Zhong & Liu, 2004). Zhang (2000) attributes all attacks north of the Yangtze River to species belonging to the Reticulitermes genus, which is equivalent to around 3-10% of all timber structures. Along the Yangtze basin, the quantity of attacks is shared by Reticulitermes and Coptotermes. South of the  9  Yangtze River Coptotermes and Cryptotermes dominate while Reticulitermes is not as common.  2.1.3  Coptotermes (Rhinotermitidae)  C. formosanus (Shiraki), or the Formosan Subterranean Termite, is considered the most destructive termite species, not only in China, but in the world. A s with Reticulitermes, the damage produced by this termite is not limited to buildings. It also attacks wooden structures such as bridges, dams and also communication infrastructure (Zhong & Liu, 2004). Crops and forests are also targets for this species.  One of the characteristics that make this termite  so  important  aggressiveness.  It  is  as  a  pest  capable of  is  its  doing  serious damage to a structure in as little as  Illustration 2.2 C.formaosanus eaten plastic container in Guangdong Entomoloaical Institute  three months. The colonies can contain hundreds of thousands of termites and can be found as deep in the ground as ten feet (Mampe, 1990). Su and Sheffrahn (1988) found colonies containing about 6.8 million workers that foraged over an area of more than 30,000 square feet. Soldiers of this species can dissolve such materials as asphalt, lead, plastics (See Illustration 3) and  mortar,  by  secreting  an  acidic  substance produced by their frontal gland (Mampe, 1990).  Source: Picture taken by the author  This species is widely distributed in southern China. In Guangdong Province and Hainan Island, it attacks as much as 90% of all residential construction. Its  10  northern limit is Jianhu in the Jiangsu province, at 33.5°N latitude. It extends all the way to the southern Islands and is also present in Taiwan (Zhong & Liu, 2004). 2.1.4  Macrotermes  (Termitidae)  The Macrotermes genus is found from Africa, in the West, to the Philippines and Australia in the East. All species are large in size and they usually build mounds and grow fungus. M. Barneyi (Light) is, from the Macrotermes genus, the most destructive species in China. This species is known to attack buildings, but it is in public infrastructure, dams and tree nurseries where it produces the most damage. Together with Odontotermes, these genera account for 90% of river and reservoir dam damage in southern China (Li, 2002). The nest of these termites is underground, so it is classified under the subterranean termites and consists of a system of interconnected chambers containing fungus combs (Roonwall, 1970). Zhong & Liu (2004), state that the northern limit in China for this termite is at 22.8°N latitude. The same authors describe their presence as far north as Henan Province (32°N to 38°N latitude). 2.1.5 Odontotermes  (Termitidae)  Odontotermes is one of the most common genera found in the Oriental region. O. formosanus (Shiraki), O. hainanensis (Light) are, from this genus, the more important pests. It is not usual to find this termite attacking wooden buildings because it normally does not attack sound wood (Roonwall, 1970). Odontotermes formosanus is a major threat to sugarcane plantations and camphor trees in Taiwan (Roonwall, 1970). Odontotermes hainanensis is believed to feed only on dried leaves and rotten wood. In China, Odontotermes is  11  known as the number one termite to cause damage to dikes and dams (Zhang, 2000). Their northern limit is in Loyang in Henan pronvince, at 35°N latitude and they extend down the Yangtze River valley to Hainan Island (Zhong & Liu, 2004). 2.2  Comments on Termites in China  As previously described,, not all of the economically important termite species in China attack timber buildings; some of them attack living trees or other wooden structures such as dikes, electrical poles or bridges. Crops can also be seriously damaged by termites and reflect in the economical importance of the pest. Zhang (2000) describes the attacks on buildings in China as follows: To the north of the Yangtze river about 3-10% of buildings are attacked by Reticulitermes; along the Yangtze basin 40-50% of buildings are found to be attacked by Coptotermes and Reticulitermes; further south, 60-70% of buildings are attacked by Coptotermes and only a  few of them by Reticulitermes; and in the  Guangdong and Hainan provinces, over 80% of buildings are attacked by Coptotermes and Cryptotermes with few Reticulitermes cases (Zhang, 2000). By analyzing the above statistics, it is possible to conclude that Reticulitermes and Coptotermes are the most aggressive genera on buildings with a lower relevancy of Cryptotermes to the south. As expected Reticulitermes is more predominant to the north and Coptotermes becomes increasingly aggressive towards the south. Since 1980, termite damage in urban areas has decreased parallel to the increase of steel and concrete as construction materials. Paved environments with predominantly masonry, concrete and steel buildings provide little ground for subterranean termite spread. Nevertheless, in Guangdong province and Hainan,  12  about 90% of residential buildings are infested with Coptotermes formosanus and in Guanxi, Hunan, Fujian, Hubei, Zhejiang provinces about 60% are infested (Zhong and Liu, 2001). Presumably the termites are feeding on non-structural wood products and other cellulosic materials. 2.3 Termite Control in China  China is undergoing a process of modernization and adaptation to meet international standards. All practices are being revised and especially those that pose a threat to the environment. At this stage, it would be unrealistic to describe the regular practice for termite control in China and not fair to comment on its appropriateness. Thus, several facts described in a number of publications are going to be enumerated in this chapter. Jianhua Zhang (2000) from the Guangdong Entomological Institute in Guangzhou described three main ways of dealing with termites as a structural pest in China: 1) Mirex and Arsenic trioxide are sprayed as termiticide powder; 2) Mirex is also used as the main active ingredient in bait systems; and 3) Chlordane emulsions are the main active ingredient for soil treatment. The spraying of termiticide powder, installing baits, or pouring termiticide emulsions into the soil, depends on the targeted termite species and the geographic location. Construction types are also diverse in the different regions of China and will influence the treatment to be used. Construction types also differ from country to urban areas. In rural areas, most houses are made of brick and have wooden roofs. Termites are a big problem because of the likeliness of this construction type to be attacked. In the cities, reinforced concrete is predominant and termites are not a big concern. Since the 1980's, when concrete and steel became the main structural construction material in the cities, the use of Chlordane and Mirex has been  13  declining. The production or use of Mirex has never been legal in China, but studies report that small amounts, in the order of 1 Tonne per year, have been produced and used each year for termite control (Zhang, 2000). Also, the production and use of Chlordane has been increasingly restricted by the government since the 1980's, and its production was banned in 1999 after the cancellation of its registration for use in 1996. The Chinese Ministry of Construction in 2004 released the revised "Regulations on Termite Prevention and Control for Urban Buildings". This document mainly aims to reinforce the significance of termite control to ensure a healthy environment in urban dwellings and to mitigate the termite problem. It focuses on regulating termite prevention practices for newly-built, rebuilt, expanded, and renovated buildings, and to the administration of termite inspections and control in existing buildings. There is reference made to different termite hazard areas. These areas are determined by administrative authorities responsible for construction in provinces and autonomous regions and administrative authorities responsible for real estates in municipalities. The regulations encourage policy makers to focus on prevention and to implement integrated control methods. Also, research on more environmentally friendly chemicals, methods of control and technologies are encouraged. Quality assurance programs are enforced by fines to be paid by termite control firms in case of violation. Real estate developers are also subjected to fines if termite prevention contracts are not issued when selling commodity buildings. The existence of these regulations is a big step forward in implementing a sound termite management system nationwide. It's up to the regulatory agencies to put the rules in place and audit for compliance.  14  3 P E R S I S T A N T O R G A N I C P O L L U T T A N T S (POPs) A N D I N T E G R A T E D P E S T MANAGEMENT  3.1  (IPM)  Overview  The treatment of buildings for termite control in China has been historically done, either by pouring chemicals into the ground to reach the underground colonies of termites, or by spraying termiticide powder. Bait systems were not introduced until recently and are becoming the most accepted suppression method (Zhong et al. 2006). Pre-treated timber is found only in some cases and is mostly used for landscaping applications. Most of the systems still rely heavily on the use of Persistent Organic Pollutants (POPs) as their main active ingredient. POP's were used extensively in the past and as described in the previous chapter some organochlorine chemicals are still illegally in production and use in China despite international regulations (UNIDO, 2003).  3.2 Persistent Organic Pollutants (POPs)  POPs are mostly organochlorine chemical compounds that were widely used in the second half of the 20th century. They were used mainly, among other purposes, for combating vector-borne diseases and pests. POPs can also be released into the environment as unwanted byproducts of industrial production during incomplete combustion processes (IOMC, 2002). These compounds are highly stable and can last for years or decades in the environment before breaking down. They resist chemical, biological and photochemical degradation to varying degrees. A process known as the "grasshopper effect" allows POPs to become widely dispersed by wind and water, through a repeated process of evaporation and deposit (Sunden, 2002). POPs can accumulate, later, in terrestrial and aquatic ecosystems and build up in  15  the fatty tissue of living organisms. High concentrations of these substances are toxic to humans and wildlife. The fact that these chemicals released in one part of the world can be transported through the atmosphere, crossing international borders and threatening the health of people and the environment in regions far away from the original source, raised international concern causing some developed countries to start banning the production and use of POPs as early as the 1970's (Sunden, 2002). Developing and third world countries though, depended on the use of these substances to prevent the spread of disease and exterminate pests. Without an efficient alternative, POPs continued to be used throughout the world. It was not until 1998, when negotiations of a legally binding instrument to first, reduce, and in the long term eliminate the use of POPs were started under the auspices of the United Nations Environment Program (UNEP). In 2001, the Stockholm Convention was signed by over a 100 countries as a global treaty to protect human health and the environment from POPs. The Stockholm Convention entered into force May 17th 2004, becoming international law. Since then, governments seek action against the use of POPs. The goal set by the enforcement of the Convention though, may prove challenging for many countries, particularly for developing countries or countries with economies in transition. The required improvements in technologies and processes are expensive and require capacity building. According to the convention document (Article 7.1, paragraphs a & b), countries must develop a National Implementation Plan (NIP) to comply with their obligations under the convention and transmit it to the Conference of Parties of the Convention (COP) within 2 years from the day it enters into force. There is no single solution to replace the broad spectrum efficacy of POPs as pest control agents. Solutions must be tailored to the specific properties and uses of each chemical or the characteristic of each technique, as well as to each country's climatic and socio-economic conditions. This is why principles of  16  integrated pest management should be included in the development of strategies to replace POPs as termiticides in China.  3.3 China and the Stockholm Convention  In May 2001, China signed the international treaty on POPs. This means, that after its ratification on May 2004, China is under the obligation of complying with the agreement. A Project Brief, titled: "Building the Capacity of the Peoples Republic of China to Implement the Stockholm Convention on POPs and develop a National Implementation Plan", funded by the Global Environment Facility (GEF) and executed by the United Nations Industrial Development Agency, summarizes the current state of affairs in China regarding POPs. The project brief identifies two chemicals that are still in production and are used as termiticides in China, Chlordane and Mirex. These POPs have been used in centrally coordinated termite control strategies supported by the Chinese government since the 1980's. By the time the project brief was published, more than 800 control stations were in place and around 10.000 operators were involved in termite control in China. The current operations are targeted to protect wood buildings, dams, roads, communication facilities, forestry and orchard operations. It is estimated that US$200 million of economic loss per year would result from termite damage if effective control agents were not in place.  3.4 Finding Alternatives to P O P s  It won't be easy for China to eliminate the use of Mirex and Chlordane as termiticides. One of the major problems for the shift to more environmentally friendly control systems is the requirement by the Chinese government for termiticides to have a period of efficacy of at least 15 years. New, sounder  17  alternative control methods typically have a shorter period of efficacy. They also tend to be more expensive and typically require periodical inspection and skilled operators. This factors increase the cost of application of the alternative control methods in comparison with POPs. The soil composition in the different regions of China is also different from western countries. Successful alternatives currently in use in North America and Europe might not work in the Chinese environment. At present, there is not enough research data to prove these alternatives as suitable for use in China. Suitable chemical alternatives are not the only aspect to take into consideration to eliminate the use of POPs as termiticides. When considering the suitability of alternative chemicals to replace POPs as termiticides, many other factors have to be taken into account as part of an integrated termite management system. These factors are: education of the public, training of operators, risk assessment, legal frameworks for the application of chemicals and revision of out of date regulation, as well as infrastructure improvement and installation of new manufacturing plants, distribution and the integration of better design and building practices. Any system adopted to replace POPs as termiticides should take sustainability into account. Considering the replacement of a POP chemical with another potentially hazardous chemical that is not yet listed as a restricted substance is not a long-term solution. DDT (Dichlorodiphenyltrichloroethane) and DDE (Dichlorodiphenyldichloroethylene) were widely accepted as safe chemicals and extensively used before the discovery of their human, animal and environmental health threatening characteristics. It is time to learn the lesson and move over to sounder practices, where words like management prevail over extermination. However, with no alternative in place, the demand for Mirex and Chlordane remains strong in China and its production and use continues. China could  18  register for a relevant specific exemption as per Annex A of the Stockholm Convention, to continue the use of Mirex and Chlordane as termiticides. However, this exemption, unless an extension is granted, expires May 2009 and after that date the use of any POPs as termiticides is strictly prohibited by the Convention. 3.5 Integrated Pest (Termite) Management (IPM)  "Integrated Pest Management means a pest management system that, in the context of the associated environment and the population dynamics of the pest species, utilizes all suitable techniques and methods in as compatible a manner as possible, and maintains the pest populations at levels below those causing economically unacceptable damage or loss." (FAO, 1966) Based on this definition, implementation of IPM can be divided in three main aspects. The first aspect refers to the circumstances in which the system needs to be implemented. Conditions such as geography, climate and soil composition need to be fully understood before adopting a particular pest control technique or method. Cultural and social aspects and their interaction with the environment need to be addressed as well. Understanding the variables specific to each site makes knowledge-based decision-making possible and ensures long lasting results. A thorough survey of the technologies and methods available is also fundamental to the implementation of IPM. The search process for the right system to apply should consider not only the whole variety of possible solutions, but also its restrictions. A link needs to be established from this point to the previous, connecting treatment methods and technology with locality. The last aspect of the definition of IPM is the one that defines the philosophical approach to the system. Termites are beneficial to the environment in many ways  19  and play a fundamental roll as part of the ecosystem. Adopting an aggressive approach towards its control may result in unexpected consequences as proved by the extensive use of POPs in the past. As explained, IPM is an approach towards sustainability by applying preventive and remedial strategies based in knowledge. IPM should also consider strategies based on the evolution of the problem over time as termite control needs to be understood from a dynamic perspective. When developing a plan or working on a specific building, IPM considers the integration of termite control as a requirement throughout the whole lifecycle of the building, from the early stages of design, to the periodical monitoring after construction (Morris 2000). An administrative and regulatory framework needs to be implemented to achieve sustainable management of pests and to integrate the different aspects of control. To date, one of the only countries in the world that has experimented and succeeded in the creation of such framework is Australia.  3.6 The Australian Example Australia is a country with a long standing wood building tradition. From the many species of termites found in Australia, three are identified as the more economically important ones: 1) Cryptotermes brevis (Walker); 2) Mastotermes darwiniensis (Froggatt); and 3) Coptotermes acinaciformis (Froggatt).  Australia banned the use of all POP pesticides in 1997, with the exception of Mirex, which is still in use in northern Australia to protect orchards from the attack of Mastotermes darwiniensis. The conference of Parties of the Stockholm Convention granted Australia an exemption for the use of this POP pesticide, due to the aggressiveness of the termite species in question, until May 2009 or until a substitute is found.  20  When international concern about the use of POP's was raised in the late 1980, Australia took measures and shifted its approach towards pest control, from relying almost completely in the use of POP pesticides to something close to what we understand today as IPM. This shift culminated in the integration of the termite problem into the building code. Australia is one of few countries that have developed a national standard for the management of termites. Australian Standard 3660, issued in 1993 and revised in 2000, focuses mainly in two strategies for management: 1) Minimize access from the ground; and 2) Periodical monitoring for termite activities. Physical barriers were developed and incorporated in the standard to minimize the likeliness of access of termites from the ground. The code is concerned with the protection of the structural elements of the house rather than the whole of the house. Details are provided in the AS 3660 on how to properly install shields around wires and pipes accessing the house from the ground and along the wall cavity. Barriers incorporated in the design of the house are not aimed to fully protect the house from attack, but they can greatly reduce the likeliness of it and facilitate periodical inspection for an early discovery of termite presence. The responsibility for termite management no longer rests solely with the pest control operator, but rather with the builder, architect and inspector. Periodical control is encouraged, by specialized personal, but also by educated homeowners. The use of pesticides is limited to cases where termites are discovered. This way application of pesticides is reduced, because less treatment is needed and pesticides can be targeted to the affected area only, sparing the need to apply pesticides to the rest of the structure.  21  The Australian experience proves that it is possible to move from the reliance on POP pesticides for termite control to a more integrated termite management system. However, the shift requires a great range of measures. The education of homeowners, architects, builders and inspectors is necessary. More important, the development and implementation of national standards is essential for the success of such an initiative. 3.7 Facilitating the implementation of IPM  The United Nation environmental Program (UNEP), FAO and The World Health Organization (WHO), have concentrated efforts on providing guidance for countries willing to shift to IPM systems. A document titled "Reducing and eliminating the use of Persistent Organic Pollutants (guidance on alternative strategies for sustainable pest and vector management)" was published in  2002  with this purpose. The first chapter of the document is a roadmap that leads readers interested in learning about measures that need to be taken to implement IPM, by a system of questions and answers. It is possible to surf the document and reach conclusions if the official information required to answer the questions is at hand. In the second chapter of the document, IPM and Integrated Vector Management IVM are explained in detail and a set of recommendations to implement IPM is suggested. IPM, as presented in the document is targeted to agricultural pests rather than structural pests, but the principles for implementation are the same. The table found in Appendix V shows an abstract of the relationship between parties or stakeholders involved in the implementation of IPM and their function. The table is an adaptation from an agricultural based approach to a theoretical relationship between builders, designers, homeowners and monitoring agencies.  22  These parties have specific functions that need to be integrated in the system and need to be addressed for implementation. The implementation of IPM is not a process that can be done overnight. Many different steps and measures need to be taken to develop a long lasting and operational plan. It is necessary that all parties involved in the process perform their function. The interaction and information transfer between parties needs to be expedite and well organized. 3.8 Comments on China, P O P ' s and the Implementation of IPM  China, unlike Australia or North America, is a country where light wood frame construction is not established. Masonry is the main construction type in the countryside and reinforced concrete dominates the multifamily apartment buildings market in the cities. Light wood frame is in its early stages of adoption. The building code for light wood frame structures is in place, but there is still the need for guidelines and training of engineers, designers, contractors and the labor force. Developers also still need to recognize light wood frame construction as a possible solution for housing. An example of IPM implementation for termite control like the Australian one is not directly applicable in China, where there is no tradition of wood frame construction and wide public acceptance of the system is not yet in place. Two questions arise when thinking about the implementation of IPM in China. First, is it possible to successfully implement an IPM program for timber frame housing in a country where concrete and masonry are more popular, considering all the different parties that need to interact and respond for the development of such a plan? Second, will the cost of implementing IPM out-weigh the benefits of timber frame construction, giving advantage to masonry, concrete and steel and making the effort useless?  23  If IPM were to be implemented in China for termite control, there would be an urgent need to acquire as much information as possible on the reality of the termite problem. One key piece of information is the identification of hazard zones for the different economically important termite species. As mentioned in chapter one, the Genus Reticulitermes and the highly aggressive species Coptotermes formosanus are of most concern to buildings. The mapping of the hazard for Reticulitermes and C. formosanus would result in a really useful tool to shape the implementation of IPM for termite control in buildings.  24  4 DEVELOPING A TERMITE HAZARD MAP 4.1 Background Mapping of termite hazard normally requires extensive surveys and a good amount of educated guessing. This is especially difficult in such an extensive country as China. Termite behavior is a discipline of its own. Many researchers have attempted to forecast termite spread and have mapped termite distribution in the world based on records of infested areas. Suggestions of identifying simple climatic boundaries (Morris 2000) have not always been successful possibly due to human introductions beyond the natural range and the heat island effect of large cities. The only termite maps available for China show the northern limits of all termite activity (Li, 2002; UNIDO, 2003; Zhong and Liu., 2004). Descriptions of locations where Coptotermes formosanus are found (Zhong and Liu 1994, 2002, 2004, Zhang 2000) suggest that it may be possible to define a northern boundary for this species. But considering that the effort of mapping termite distribution and assigning termite hazard zones has already been done in Japan, North America and Australia, the opportunity arises to use this previous experience to create a forecasting tool that would plot hazard values for any location in the world. Matching the myriad characteristics of locations with and without specific termites in one part of the world to the characteristics of comparable locations in other parts of the world, considering that many parameters had to be used in the assessment of climate, would have been a daunting task if only a formal regression based approach was attempted. This is because no sufficient prior knowledge about the functional relationship between the various parameters was at hand. Alternatively, the Neural Network approach offers a powerful method to address this issue without needing a prior knowledge of the functional relationship between the various factors.  25  4.2 Neural Networks  The earliest work in artificial neural computing goes back to the 1940's. However the technology available at that time did not allow for further development of the technique. Technological limitations resulted in the decline of the field but renewed interest was shown in the early 1980's (Russel, 1996). Artificial Neural Networks (ANN) are inspired to mimic biological nervous systems such as the brain. Brains, however are in an order of magnitude that far exceeds any ANN considered to date (Smith, 1996). The primordial idea of ANN's is to generate an artificial multiprocessor computer system that is able to learn from example. Networks are composed of a large number of highly interconnected simple processing elements working in parallel to solve a specific problem. These elements that form the basic structure of the network are called neurons. The first artificial neuron was produced in 1943 by the neurophysiologist Warren McCulloch and the logician Walter Pits (Stergiou & Siganos, 1997). ANN's have a remarkable ability to extract meaning from complex data and they can be used to read patterns and detect trends that are too complex to be noticed by simple observation or other regular computer techniques (Stergiou & Siganos, 1997). They can be used when traditional algorithmic solutions can't be formulated or when the variability of the input information is unintelligible. Training is the fundamental step to achieving valuable results from an ANN. The training process is performed with information that is known to provide a specific result. The network will analyze the information to find the optimal weights for each input parameter. An ANN can always be retrained. Its adaptability to new environments is one of the advantages of using ANN over traditional algorithmic techniques.  26  Once the ANN is trained, it can be used to classify information that has an unknown output or to provide projections given a change in the weight of the input parameters. Basically, it becomes an answering tool to "what if "questions. One of the principal uses of Neural Networks is pattern recognition. ANN have been used for this purpose in investment analysis, monitoring, process control, and complex pattern recognition techniques such as hand written digit recognition and signature analysis (Jain and Fanelli, 2000). ANN'S are being applied in many other tasks involving pattern recognition of complex input. The commercially available Neural Network software NeuroSolutions® was used as the platform for network training and classification production in this study. Its user friendly interface allowed performing network training and production without the need of thorough understanding of the underlying network architecture. This process is called unsupervised learning.  4 . 3 Materials  As mentioned before, the only way to train a Neural Network is to use information that is known to provide a certain result. The only way to use Neural Networks to forecast termite hazard is to have information on hazard levels related to classifiable information. Termite behavior and spread is influenced by different parameters including, among others, soil properties, climate, geographical location in respect to the equator, presence of other species of termites and insects and the movement of people.. Mapping of termite hazard has been done in many regions where termites are a problem. The most accurate map found is the one for Japan. This subterranean termite hazard map is the 2004 version of a distribution map for the two major subterranean termite species: Reticulitermes speratus and Coptotermes  27  formosanus. The map was generated with the information gathered by the Japan Termite Control Association (JTCA) from a recent questionnaire survey collected from Pest Control Organizations (PCOs) all over the country. There are only two hazard areas shown in these maps for both termite species. Either there has been termite activity reported or not. Termite activity has been tracked in Japan for decades; therefore, a very reliable source of information is available. North America has also developed its own termite hazard map. This map can be found in the "termites and wood" section of the wood durability website of Forintek Canada Corp (  and  has been generated using different published sources and expert opinions. The map shows four different hazard zones for native subterranean termites such as Reticulitermes flavipes and Reticulitermes hesperus. The four classes are heavy,  moderate, light and none. The occurrence of Coptotermes formosanus is shown only in the specific spots where there has been reported activity and are graphically symbolized as yellow dots. The spots where Coptotermes formosanus have occurred in North America are all within the heavy hazard zone for Reticulitermes. Thus, for the purpose of this study, the assumption is made that Coptotermes formosanus hazard area is coincident with the high hazard zone for the native subterranean termite. The Australian termite Hazard map was developed by CSIRO Australia ( iaRelease). Six hazard zones are found in this map : very high, high, moderate, low, very low and negligible. The termite species found in Australia are different than the ones found in Japan and North America. Therefore, only the Japanese and North American maps are used in the training of the final Network. The Australian map is only used prior to the development of the final network to assist in the acquisition of enough data to consolidate the network training sample-set.  28  The information that the maps provide is the hazard level for different locations within each of the above-mentioned regions. Therefore, these maps give information on the desired output of the network. The input information consists of the climatic data for the specific locations. Climate is used as the input information due to the ready availability and relative ease with which this information can be acquired. Still this information has to be normalized for the purpose of comparison. The Weatherbase® is a free source of climatological information for more than sixteen thousand locations in the world. The information is gathered from several different public domain sources and normalized for ease of access. The information gathered in this database has undergone the same process of normalization for all different locations in all four different countries included in this research; therefore, it is considered the best source of climatic information available. 4.4 Methods  The maps found for Japan, North America and Australia differ in the criteria used to classify termite hazard zones. A process of normalization is required to produce comparable information. Therefore, the hazard zones in all maps were normalized and reduced to three levels: 1. High 2. Moderate-Low 3. Negligible High is the area where Coptotermes formosanus occurs based on the Japan and North American Hazard maps. The three levels provide the northern hazard limits for both Reticulitermes and Coptotermes formosanus. These limits are of most importance for the purpose of hazard zoning. A lower order of classification could  29  be useful for the Coptotermes  formosanus  hazard zone, but no detailed  information of such kind is available to provide input data to a Neural Network. The map used in this study for Japan (Figure 4.1) is the result of overlapping both termite species hazard maps.  Figure 4.1 Japan Hazard Zones and data locations  » «Nishinoyama  N a  ^»,  | High  l Amami Islands  Moderate - Low Swo  Jima  urasoe  j Negligible  30  The map for North America (Figure 4.2)  is the result of joining the moderate and  low hazard zones into one large zone.  Figure 4.2 North America Hazard Zones and data locations  | High j Moderate - Low j Negligible  31  At first, the intention was to include only North America and Japan in this study because the same termite genera are mapped for both countries. The Australian map shown in Figure 4.3 is a simplified version of the one developed by CSIRO. The simplification is justified, as the map served only as a reference; the data derived from this map were not included in the final network. Figure 4.3 Australia H a z a r d Z o n e s a n d data locations  Lockhart  Cairns "ownsville Mackay lockhampton  Gladstone gara ndaberg Brisbane Forrdst  Woomera  9  Leigh Creek Broken Hill  fCoffs Harbour Narrabri .TarrAorth  Cedilla perance Mbany  Cobar oranqe •Wriliamtown .Mildura^ Newcastle . X ^ i , * Cowra* fRichmond Adelaide Nhill Wagga* ./Sydney 1 Canberra* £™owra Broadmeadows *Merimbula Mount Gambler*. ••Melbourne Portland * •sale •Flinders Island unceston ambridge l u  u a n D e r r a  5693  High Moderate - Low j Negligible  32  4.5 Finding samples  The Weatherbase® provides information for about 16000 different locations in the world and up to 46 different information fields. Most locations include 10 to 25 information fields; however, not all locations include the same fields. A criterion had to be set in order to find sufficient information to provide a real assessment of the climate properly distributed in all four countries including China. The criteria used for the selection of the locations included in this study are: 1. Relevance of location (major cities) 2. Sufficient information to assess climate 3. Climatic information format (for comparability) Major cities were selected at the beginning to define the formatting of the climatic information sought. As mentioned before, not all information found in the Weatherbase® database is displayed in the same format and also many locations lack sufficient information to match the requirements set to assess climate. Often the information was presented in a different format than that of the major cities and could not be compared with the data already collected. The selection of the variables needed to assess climate in all four countries was based on a literature review on climate in relation to termite behavior (Marais, 1939; Snyder, 1948, 1949, 1956; Skaife, 1955; Roonwall, 1970; Lee etal., 1971; Ebeling, 1975; Wood, 1977; Mathews era/., 1978; Edwards etal., 1986; Mampe, 1990; Lenz, 1994; Pearce, 1997; Li, 2002; Tsunoda et al., 2002; Shelton et al., 2003). The variables found were then adapted to fit the database format (Table 1)A total of 91 locations were found in North America (see Appendix Vll). This number is considered sufficient to create a robust sample-set for assessing  33  climate in this region. However, it was not possible to find enough locations that meet the same criteria for Japan. Table 1 Parameters used in climate assessment  1*'  Elevation Latitude  Unit in D°  ivr  Temperature  Precipitation  cm  Relative Humidity  %  Dew Point  Variable  Elevation Latitude Year average temperature Monthly average temperature standard deviation over a year Maximum monthly average temperature Minimum monthly average temperature Number of days over 32° C over a year Number of days under 6° C over a year Year average precipitation Monthly average precipitation standard deviation over a year Maximum monthly average precipitation over a year Minimum monthly average precipitation over a year Yearly average morning relative humidity (MRH) Monthly average MRH standard deviation over a year Maximum monthly MRH over a year Minimum monthly MRH over a year Year average evening relative humidity (ERH) Monthly average ERH standard deviation over a year Maximum monthly ERH over a year Minimum monthly ERH over a yenr Year average dew point Monthly average dew point standard deviation over a year Maximum monthly dew point over a year Minimum monthly dew point over a year Note: rows in gray identify parameters eliminated alter assessment (see point 4.6)  34  Most locations in Japan did not provide enough information, thus making it impossible to produce a realistic sample-set to train the network. Therefore, data from Australia were added to the sample data. It is known that the subterranean termite species found in Australia are different from the ones found in Japan and North America; however, as mentioned before and explained later, the data gathered from the Australian locations was only used as a reference for the consolidation of the final network. The assumption was made that by producing a robust enough sample-set by considering a wide range of different geographical locations to test the network accuracy, it would be possible to decrease the number of parameters necessary to assess climate by analyzing their relevance. Sixty locations were found in Australia to meet the criteria to be included in the network sample-set (see Appendix VIII). The inclusion of the Australian locations permitted the generation of a consistent network with a total of 151 sample locations.  4.6 T r a i n i n g the N e u r a l N e t w o r k  Backpropagation Neural Networks are one of the most commonly used learning processes because they are simple and effective. The one used for this study in particular is composed of a network of nodes arranged in three layers: 1) the input layer, 2) the hidden layer, and 3) the output layer. The input layer processes the information given as the training input data-set and passes it to the hidden layer. The nodes in the hidden layer multiply each value using different weights and the result is passed on to the output layer which compares the resultant values, in this case, to the classification values given in the testing sample-set. The relationship between the classification value and the resultant  35  value are backpropagated through the network which causes adjustments to the weights in the input and hidden layer nodes. The number of iterations defines how many times the loop is repeated before the network stops. After testing many times with random numbers, 20,000 iterations were defined as sufficient for this hazard mapping problem, in fact the linear correlation coefficients between the classification value and the resultant value remained stable after more than 2,000 iterations. All networks run in this exercise considered 75% of the samples as the training sample-set and 25% as the testing sample-set. In order to assess the accuracy of the network, the samples were randomized each time a network was created. Thus the 25% testing sample-set always consisted of different observations. The randomization of the samples for each network permitted the confirmation of the model during the whole training process. Since all three maps used to collect the samples are hand drawn and the limits represent an estimated graphical representation of the hazard areas, many locations fell in an uncertainty area where the hazard factor is not well defined. Considering that the Neural Network software plot values within either one of the three hazard zones, it is necessary to eliminate the locations that confuse the network results due to their proximity to the virtual border. These locations were considered to have a "border condition". The process of eliminating from the sample-set locations with a border condition, that were recurrently misclassified in the testing of the network, was named "cleaning" of the sample-set. An assumption was made that for the purpose of classification, using locations that lay well within the different hazard zones contributes to the training process by providing the network with more reliable information. This process also allows for easier identification of trends.  36  Using the training method described before and eliminating the locations that met the border condition and were recurrently misclassified, the samples were randomized 20 times, creating a new classification network each time. The data was cleaned and run 20 times again randomizing the data each time. After this second pass, the cleaning process was repeated and the resulting samples constituted the final classification network data.  After the two cleaning  processes, the classification achieved an average linear correlation coefficient (r) of 0.96 between the classification and the resultant values. As explained previously, there were not enough locations found for Japan with a complete set of variables. Locations from Australia were included in this part of the exercise in order to build a larger sample-set and to identify which parameters had a higher relevancy in the network outcome. The assumption was made that by doing this, some variables might not be needed, and then more locations from Japan could be included. The previously cleaned samples-set was used to run networks first by including and then by excluding only one variable (or variable set) at a time (see Appendix IV for the list of variable-sets). The training and testing sample-sets used for all tests were always the same in order to get comparable results. None of the variables (or variable sets) showed a particularly low degree of relevance when excluded from the network sample-set. As shown in Figure 4.4 (see Appendix IV for variable sets), most linear correlation coefficient values continued to be over 0.9 at least for the high hazard rating.  However, the average linear correlation coefficient of the network  improved when excluding elevation (variable 2) and all relative humidity variables together (variable set 25).  37  Figure 4.4 Linear Correlation Coefficient ( r) between classification and resultant value by excluding one parameter at a time • high m Ivbderate/Lcw • Negligible  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19 20  21  22  23 24 25  26  27  28 29  30  Variable-sets  Considering each variable or variable set individually showed that relative humidity variables, and elevation were surprisingly not particularly relevant, as the linear correlation coefficients were close to 0 (Figure 4.5; see Appendix IV for variable sets). All variables related to relative humidity and elevation were, therefore, eliminated subsequently from the network sample-set. Tests were performed to confirm that the network still had a high average linear correlation coefficient. The low relevance of altitude can be associated with the changing climate conditions in altitude throughout the globe. As one gets further away from the Equator, temperature decreases at the same altitude. Thus, elevation will not represent a consistent behavioral measurement throughout  the region  considered for this study. The little relevance that the set of variables for relative humidity shows can be associated with the low influence that this factor poses upon underground conditions. All termites included in this study are subterranean and normally find the moisture they need to survive from sources like rain  38  infiltration and underground water tables. They also facilitate movement of moist air through their shelter tubes thereby increasing the moisture content of the wood products they are attacking. Figure 4.5  Linear Correlation Coeficient ( r) between classification and resultant value for each parameter tested individualy • high B iVbderate/Lcw • Negligible  0,8 0,6 0,4  0,2  111 1  21  3  4  5  6  7  8  9  10 11 12 13 14 15 16 17 18 19 20 21  J ab  23  25 26 27 28 29 30 fc4  -0,2  Variable-sets  The assumption made earlier that variables (or variable-sets) could be eliminated from the network after testing their relevancy was confirmed. The elimination of these variables permitted more locations from Japan into the network sampleset.  A total of 62 locations in Japan were included in the network sample set,  and the locations in Australia were eliminated, coming to a total of 153 samples between North America and Japan (see appendix VI & Vll).  The network was cleaned twice again, following the same method described previously and also randomizing the data 20 times for each cleaning process. Twenty-nine locations were eliminated from the network due to proximity to the borders and recurrent  misclassification. A final average linear correlation  coefficient of 0.9 was achieved for this network after testing it 20 times with 39  random training and testing sample-set arrangements, and using all variable sets except for elevation and relative humidity. This sample set was used then as the training sample set for classifying 274 locations in China (see Appendix IX).  4.7 Results The 274 locations in China formed the production sample-set. A three layer backpropagation classification network was classify the locations as high, moderate-low or negligible hazard zone. For each location considered, the Neural Network software provides resultant values in each one of the three hazard zones. The training process used considered outcome values of either 1, meaning that the sample belonged to that particular hazard zone or 0, meaning that it did not. The criteria used to classify the results was that all locations that plotted a value for any of the three hazard zones higher than 0.75 were considered to have a strong relationship with the classification value and thus fall within that specific hazard zone. All locations that didn't meet this criteria were declared unclassified (i.e., no clear class was identified). From the 274 locations, 76 fell within the high hazard zone, 99 within moderatelow, 68 within negligible and 31 locations were declared unclassified. This result showed that the sample locations were well distributed and that the sample-set constitutes a good representation of the region.  40  The map shown in Figure 4.6 is the result of the plotted network outcome. All locations were assigned a number colored accordingly to its hazard zone. The numbers were then placed in their corresponding geographical location. This way zones are recognized by the clustering of numbers of the same color. The zones are then extrapolated from these clusters. The result is the map shown in Figure 4.7.  Figure 4.6 Location of data point for China by color according to hazard value  62  W~f 132  120 1*1  3 ,100 P 109, 95  v  ^  85  1  1  3  6  103 8  12o  117114 78  2-^34  25  9 4<??3  56  38 , 244 g?^ -  110.  5  1  §  Z  1 >4  267 4= 265 223 2 » ? M  0  1  2 d 7  9  i s 2  ,,182 High  5  *U8 Moderate - Low  2lM  ^ 1 J8?4 4  8  146 204  S& 2402  2  42BT9SfW  201  Negligible Unclassified  41  Figure 4.7 Termite hazard map for China  J Negligible  The northern boundary for the moderate-low hazard zone partially coincides with previous maps and descriptions of termite distribution in China (Li, 2002; UNIDO, 2003; Zhong and Liu., 2004) Although found to be slightly north of previous representations, the line that winds down from the southern Jilin region passing just north of Beijing and then curving south to pass just north of Xi'an and after curving west into Tibet (Xizang) has been described and drawn many times before.  Something new observed in this map, though, is the extension of the moderatelow limit into the Xinjiang region, considered to be outside of the recognized termite hazard zone outlined by Li (2002), the UNIDO report (2003), and Zhong 42  and Liu (2004). This may suggest that conditions only need to change slightly for termites to spread into that region or that their absence is due to a natural barrier that could be breached by human transport of infested material. The northern limit for Coptotermes formosanus also coincides with previous descriptions (Zhong and Liu., 1994 - 02 - 04; Zhang, 2000). After the termite hazard map was produced, it was compared with precipitation and altitude maps for China. Figure 4.8 shows a comparison of the average annual precipitation and the limits between hazard zones defined by this study.  Limit f o r termite Hazard Z o n e ~ ~  High - Moderate/Low  ' ' " " M o d e r a t e / L o w - Negligible  43  There is no exact coincidence between precipitation and termite hazard zones, but it is observable that most precipitation is concentrated within the high hazard zone. It is also observable that the north-western region gets some precipitation and is separated by a dry belt from the rest of the rainy region. This confirms, to some extent, that it is possible that termites are being stopped from crossing to the north-western region by this dry belt. The comparison with the altitude map shown in Figure 4.9, also shows some degree of coincidence. Altitude was eliminated from the Neural Network variables, nonetheless climate conditions are influenced by altitude and thus it was considered relevant to see if altitude could be a factor influencing termite hazard.  —  Limit for termite Hazard Z o n e High - Moderate/Low  • • • • M o d e r a t e / L o w - Negligible  44  No direct relationship was observed between elevation and termite hazard. It was noted that the curve, which describes the limit between high hazard and moderate/low hazard, follows, to some extent, the difference in elevation from east to west.. The higher regions of western China also show some coincidence with the line that describes the limit between moderate-lorn and negligible hazard. This termite hazard map developed in this study was shown to termite experts in China, many of whom had participated in the development of the previous maps. Most of their views coincided in that there was no report of termites in the Xinijiang region and that the surveys made did not find any evidence of termites in that corner of China. Some, though, mentioned rumors of localized termite activity as far as Xinijiang but did not have any evidence of it. In general the map was well received. It is important to remember that this map is a representation of hazard areas, which should not be interpreted as risk zones. The first condition for this map to apply is the observed presence of termites in the region, which could have been prevented by geographical or microclimatic barriers. The map represents a rough approximation of the areas where it is more or less likely that Reticulitermes as a Genus and Coptotermes formosanus may occur. The intention of this study is to prove the effectiveness of the method used. The method and the outcome map are subject to improvements. Also, the accuracy of the outcome will depend on the quality of the input information; therefore, further studies, including new parameters such as soil properties, degree of urbanization, forested areas and others, could be performed later to generate a more precise forecasting tool. Also, similar exercises could be carried out for other species of termites.  45  5  CONCLUSIONS  Canada, as in many other countries, recognizes in China an opportunity for increasing its trade. The housing market is especially attractive as Canadians are major producers and exporters of wood products. However, China is not prepared yet for receiving light wood frame as a massively used building model. Training and regulation still need to be put in place to assure a sustainable incorporation of the model. The villas and luxury apartments constructed lately, mirrors of western style, in response to the desire of Chinese people to live like their western counterparts, are just exceptions to the general rule of the concrete and masonry constructions. It is fundamental that these few examples perform properly, not only to assure future acceptance of the model, but also to assure quality to homeowners. The adaptation of the light wood frame building model to China needs to be thorough, including China's commitment to meet international trade and environmental standards. A Neural Network was used in this study to innovatively develop a termite hazard map for China. The hazard map for Reticulitermes and Coptotermes formosanus  generated by the trained Neural Network may help in the implementation of a successful termite management plan for China. The hazard zones shown in the map could be incorporated in such a plan as a frame structure for the enforcement of strategies within each zone. Developers who plan to use wood as the main structural material in their projects can also benefit from this map. Strategies could be implemented in accordance with the different hazard areas to address the termite issue. Current termite management in China has to change because of the international regulations that obligate China to adapt in order to compete in the global market. The hazard areas defined in the resulting map will help to define sounder  46  strategies that are targeted to the specific needs of each region, rather than evenly applied throughout the country. Persistent Organic Pollutants are probably going to be used as the main suppression method for a while, but if hazard areas were recognized as such and chemicals were applied depending on the hazard level, a reduction in its use could occur. The use of the Neural Network technique to map termite hazard provides little information to evaluate which climatic parameters or their interaction might influence most termite distribution. It was observed, though, that relative humidity is not a significant factor in the behavior of subterranean termites. The neural network results could be further analyzed to be used as reference in future studies or to create a more accurate forecasting tool. Finally, this Neural Network approach to mapping termite hazard can also be used to generate termite hazard maps for other regions in the world. Perhaps even the North American map, which is based on experts' opinions, could be improved by using the same method based on a hazard map for China, if one could be developed.  47  BIBLIOGRAPHY Anonymous. 2000. Housing's Great Leap Forward. Retrieved April 2004 from: The Economist Anonymous. 2003. Rich Man, Poor Man. Retrieved April 2004 from: The Economist Anonymous. 2004a. Castles in the Sky. Retrieved April 2004 from: The Economist Anonymous. 2004b. Ideological Gymnastics; China. Retrieved April 2004 from: The Economist Canada Wood. 2003. Opportunities for Canadian Wood Products in Selected Pacific Rim Markets. Retrieved January 2006 from: Wood/english/view.asp?x=15&mid=21 Cartwright, D., Gaston, C. Ni, Ch. & Dai, Ch. 2002. Peoples Republic of China: Market opportunities for Wood-Frame Construction. Canada Wood, CWB, CAF, FN & Forintek Canada Corp, Vancouver. 67pp. CFS, Canadian Forest Service. 2005. The State of Canada's Forest 2004 -2005. Natural Resources Canada, Ottawa. 94pp. Ebeling, W. 1975. Urban Entomology. University of California Division of Agricultural Sciences, Berkeley, California. 659 pp. Edwards, R. & Mill, A.E. 1986. Termites in Buildings. Rentokil Limited, East Grinstead, 261 pp. FAO. 1966. .Proceedings FAO Symposium on Integrated Pest Control. Rome, Oct. 11-15, 1965. Rome. Part 1, 91 pp.; Part2, 186 pp.; Part 3, 129 pp. FN (Forestry Innovation Investment). 2003. Market Opportunities in Asia for BC Forest Products. Earnscliff Research & Communication, Vancouver. 68pp. IOMC (Inter - Organization Programme for the Sound Management of Chemicals). 2002. Reducing and Eliminating the Use of Persistent Organic Pesticides. Geneva. 89pp. Retrieved December, 2004 from:  48  Jain, L.C. & Fanelli, A.M. 2000. Recent Advances in Artificial Neural Networks: Design and Applications. CRC Press, Boca Raton. 350pp Lee, K. E. & Wood, T. G. 1971. Termites and Soils. Academic Press, London. 251pp. Lenz, M. 1994. Food Resources, Colony Growth and Caste Development in Wood Feeding Termites. In: Hunt, J.H. and Nalepa, C A . (eds) Nourishment and Evolution in Insect Societies. Westview Press, Boulder, pp 159-202 Li, G. 2002. Termites and their Control in China (in Mandarin). Science Press, Guangzhou. 280 pp.. Light, S.F. 1934. Drywood Termites, their classification and distribution. In Termites and Termite Control. Univ. of California Press, Berkeley, pp 312-316 Mampe, C. D. 1990. Termites. In: Mallis, A. (ed.) Handbook of Pest Control. The Behaviour, Life, History and Control of Haushold Pests. Franyak and Foster, Cleveland, Ohio, pp. 201-263 Marais, E. N. 1939. The Soul of the White Ant. Methuen & Co., Ltd.. London. 184pp. Matthews, R. W. & Matthews, J. R. 1978. Insect Behaviour. John Wiley & Sons, New York. 507 pp. Morris, P.I. 2000.Integrated Control of Subterranean Termites: The 6S approach. Proceedings AWPA 96 (American Wood Preservers' Association). Birmingham, AL. pp 93-106 Panitchpakdi, S. & Clifford, M. L.2002. China and the WTO. John Wiley & Sons, Singapore. 251 pp. Pearce, M.J. - 1997. Termites: Biology and Pest Management. CAB International, Wallingford. 172 pp. Roonwall M. L. 1970. Termites of the Oriental Region. In: Krishna, Kumar & Weesner Frances M. Biology of Termites. Academic Press, New York, pp 315-391. Russel, Igrid. 1996. Brief History of Neural Networks. Retrieved March, 2006 from: Shelton, T. G. & Grace, J. K. 2003. Termite Physiology in Relation to Wood Degradation and Termite Control. In: Goodell, Barry; Nicholas, Darrel D. and  49  Shultz, Tor P. (eds) Wood Deterioration and Preservation. American Chemical Society, Washington D.C., pp. 242-252. Skaife, S.H. 1955. Dwellers in Darkness: An Introduction to the Study of Termites. Longmans, Green & Co., Ltd., London. 134pp. Smith, L. 1996. An Introduction to Neural Networks. Retrieved March, 2006 from: Snyder, T. E. 1948. Our Enemy the Termites. Comstock Pub Co, Ithaca, N.Y..196pp. Snyder, T. E. 1949. Catalog of the Termites (Isoptera) of the World. Smithsonian Institution, Washington. 490 pp. Snyder, T. E. 1956. Annotated, Subject-Heading Bibliography of Termites, 1350 B.C. to A.D. 1954..Smithsonian Institute, Washington. 305 pp. Statistics Canada. 2006. Exports of Goods on a Balance-of-Payments Basis, by Product. Retrieved January 2006 from: http ://www40. statca n. ca/I01 /cst01 /q b Iec04. htm. Stergiou Ch. & Siganos D. 1997. Neural Networks. Retrieved March, 2006 from:, html. Su, N. Y. & Scheffrahn, R.H. 1988. The Formosean Subterranean Termite. In: Pest Management 7(7), pp. 16-25. Retreived August, 2004 from: web exclusive. Sunden, A. 2002. Persistent Organic Pesticides. In: The Encyclopedia of Pest Management. Retrieved August, 2004 from: Tsunoda, K., Grace, J.K., Byrne, A. & Morris, P.l. 2002. Effectiveness of disodium octaborate tetrahydrate (Timbor®) in controlling subterranean termite attack and decay of house sill plates. Mokuzai Gakkaishi 48:107-114. UNEP (United Nations Environment Programme) 2001. Stockholm Convention on Persistent Organic Pollutants. Retrieved November 2004 from: en.pdf UNIDO (United Nations Industrial development Organization) 2003. Building the Capacity of the People's Republic of China to Implement the Stockholm Convention on POPs and Develop a National Implementation Plan. Retrieved November 2004 from:  50 Documents/GEF C21/P0P China - Proiect Document.pdf UN - DESA (United Nations Division for Sustainable Development). 2001. Agenda 21, Retrieved November 2004 from: m Wood, T.G. 1977. Food and Feeding Habits of Termites. In: Brian, M.V. (ed.) Production Ecology of Ants and Termites. International Biological Programme 13. Cambridge University Press, Cambridge, pp. 55-80. WTO. 2005. Trade Growth in 2005 to slow from Record 2004 Pace. Retrieved January 2006 from: e/pres05 e/pr417 e.htm Zhang, J. 2000. Major termite problems and termiticides in China. In: Report of the UNEP/FAO/Global IPM Facility Termite Biology and Management Workshop. IOMC-(lnter-Orginazation Programme for the sound Management of Chemicals).A cooperative agreement among UNEP, ILO, FAO, WHO, UNIDO, UNITAR & OECD. Retrieved November 2004 from: Zhong J. & Liu, L. 1994. Control of Formosean Subterranean Termite in China. The International Research Group on Wood Preservation Document No. IRG/WP/10088. 12 pp. Zhong,J. & Liu, L. 2002. Experience with Coptotermes formosanus Shiraki in China. In: Proceedings of the 2 International Symposium on Coptotermes formosanus, May 13th-15th, 2002, New Orleans, 10 pp. nd  Zhong, J & Liu, B. 2004. Current termite Management In China. 1st Conf. In: Proceedings from Pacific Rim Termite Research Group, Penang, Malaysia, 89 March 2004, pp. 11-16. Zhong, J., Liu, B., Huang, Z. & Huang, J. 2006. Evaluation of Sentricon termite elimination system against Coptotrmes formosanus Shiraki. In: Proceedings of the Third Conference of Pacific Rim Termite Research Group. March 6th-7th, 2006, Guangzhou, P.R. China, pp. 61-66.  51  APPENDICES  Appendix I Japan Subterranean Termite Hazard Map It was not possible to get authorization to publish this image. The map shown in the image was developed by the Japan Termite Control Association (JTCA). It contains two separate maps of Japan, the map on the left side shows in black the area where Coptotermes Formosanus is found, the map on the right side shows in gray the area where termites form the Reticulitermes genera are found. The map used for this study is the representation of both maps put together.  52  Appendix II  North American Subterranean Termite Hazard Map  Subterranean Termite Zones of North America  Source: Forintek Canada Corp  53  Appendix III  Australian Subterranean Termite Hazard Map It was not possible to get authorization to publish this image. The map shown in the image was developed by CSIRO Australia. It contains one map of Australia showing six different termite hazard areas; 1) Very high, 2) high, 3) Moderate, 4) Low, 5) Very Low and 6) Negligible. The hazard areas refer to all termite species found in Australia. For the purpose of this study, the very high and high hazard areas were joint into one big area called high, the moderate, low and very low hazard areas were joint into the moderate-low hazard area and the negligible area was left the same.  54  Source: CSIRO Australia asp?type=mediaRelease&id=Termite&stylesheet=mediaRelease  55  Appendix IV L i s t o f v a r i a b l e s e t s u s e d in F i g u r e s 4.4  &  4.5.  1.  All p a r a m e t e r s together (reference)  2.  Elevation  3.  Latitude  4.  A v e r a g e temperature over year  5.  A v e r a g e monthly t e m p e r a t u r e s t a n d a r d deviation o v e r y e a r  6.  M a x i m u m a v e r a g e monthly temperature o v e r y e a r  7.  Minimum average temperature over year  8.  All t e m p e r a t u r e p a r a m e t e r s together  9.  D a y s o v e r 32° C o v e r y e a r  10. D a y s u n d e r 0° C o v e r y e a r 11. N u m b e r of d a y s with e x t r e m e t e m p e r a t u r e s (9+10) 12. A v e r a g e precipitations o v e r y e a r 13. A v e r a g e monthly precipitation s t a n d a r d deviation o v e r y e a r 14. M a x i m u m a v e r a g e monthly precipitation o v e r y e a r 15. M i n i m u m a v e r a g e precipitation over y e a r 16. All precipitation p a r a m e t e r s together 17. A v e r a g e morning relative humidity o v e r y e a r 18. A v e r a g e monthly morning relative humidity s t a n d a r d deviation o v e r y e a r 19. M a x i m u m a v e r a g e monthly morning relative humidity o v e r y e a r 20. M i n i m u m a v e r a g e morning relative humidity o v e r y e a r 21. A v e r a g e e v e n i n g relative humidity o v e r y e a r 22. A v e r a g e monthly e v e n i n g relative humidity s t a n d a r d deviation over y e a r 23. M a x i m u m a v e r a g e monthly e v e n i n g relative humidity o v e r y e a r 24. M i n i m u m a v e r a g e e v e n i n g relative humidity o v e r y e a r 25. All relative humidity p a r a m e t e r s together 26. A v e r a g e d e w point over y e a r 27. A v e r a g e monthly d e w point s t a n d a r d deviation o v e r y e a r 28. M a x i m u m a v e r a g e monthly d e w point o v e r y e a r 29. M i n i m u m a v e r a g e d e w point o v e r y e a r 30. All d e w point p a r a m e t e r s together  Appendix V R e l a t i o n s h i p b e t w e e n s t a k e h o l d e r s i n v o l v e d in t h e i m p l e m e n t a t i o n o f I P M a n d their function. Stakeholder  Function  Homeowners and Local Communities:  Learn by doing. Inform themselves about alternative control methods. Engage in pilot Projects Stimulate their communities  Unions and construction Workers:  Push for safer pest control methods Report continued or illegal use of POP  Pesticide Companies:  Develop pesticides compatible with IPM Make users aware of risks and precautions to be taken  The Various Public Sectors of Government:  Revise policies, regulations and legislations on pesticides Develop new Standards Actively enable and support community and household efforts by providing technical backstopping, information, training and financial assistance Upgrade facilities for chemical analysis  Multilateral Organizations and NGO:  Influence and facilitate policy reform Carry out independent assessment and evaluation Disseminate information and set up Pilot projects  Multi and Bi-lateral external Support Agencies:  Financial support for implementation and research  The National and International Research Community:  Carry out studies on development and implementation of IPM Research on new pesticides and their effect on the environment  Consumers and Consumer Groups:  Put pressure on demanding safe products  Schools and Universities  Introduce IPM in their curricula Innovative research to strengthen the concept  57  Appendix VI List of Japanese cities used as data locations in Training Networks. Included 1  2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22  Abashiri Aikawa Aizu-Wakamatsu-shi Akita Amami Island Aomori Choshi Fukuoka Gojikkoku Haki Hakodate Hiroshima Hofu Ishigaki Ishinomaki Iwaki Iwakuni Iwo Jima Kagoshima Kanasawa Komatsu Kumagaya  23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44  Kumamoto Matsushima Miyazaki Nagano Nagasaki Naha Nakashibetsu Nara Naze Nemuro Niigata Nishinoyama Oita Osaka Oshima Shimonoseki Tagajo Urasoe Utsunomiya Wakkanai Yao Yokosuka  10 11 12 13 14 15 16 17 18  Miho Okayama Otsu Suigasaki Tokushima Tokyo Tsu Usa Wajima  Cleaned  1 2 3 4 5 6 7 8 9  Akeno Asahikawa Ashiya Atsugi Chiba Gifu Hamamatsu Kudatchi Matsuyama  Appendix Vll  List of North American cities used as data locations in Training Networks. Included  1 Albany, NY 2 Albuquerque, NM 3 Alpena, Ml 4 Atlanta, GA 5 Augusta, GA 6 Austin, TX 7 Boise, ID 8 Boston, MA 9 Calgary, AL 10 Cape Dyer, NNV 11 Casper, OR 12 Charleston, WV 13 Cheyenne, WY 14 Chicago, IL 15 Churchill, MB 16 Cleaveland, OH 17 Columbia, SC 18 Columbus, GA 19 I Comox, BC 20 | Dallas, TX 21 j Denver, CO 22 Des Moines, IA 23 Detroit, Ml .24 Duluth, MN 25 Edmonton, AL 26 El Paso, TX 27 Eureka, NNV 28 Flint, Ml  29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56  Gander, NF Goose Bay, NF Grand Forks, ND Green Bay, Wl Helena, MO Honolulu, HW Houston, TX Indianapolis, IN Inuvik, NT Iqualit, NNV Jackson, MS Kansas City, MO Lake Charles, LA Lansing, Ml Las Vegas, NV Louisville, KY Madison, Wl Miami, FL Minneapolis, MN Missoula, MO Mobile, AL Montgomery, AL Montreal, QC New Orleans, LA New York, NY North Bay, ON Olympia, WA Orlando, FL  57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83  Ottawa, ON Philadelphia, PA Phoenix, AZ Portland, OR Quebec City, QC Raleigh, NC Resolute, NNV Richmond, VA Sable Island, NS Sacramento, CA Saint John, NB Salem, OR Salt Lake City, UH Sault Ste. Marie, Ml Savannah, GA Shearwater, NS Spokane, WA St. John's, NF Sydney, NS Tallahassee, FL Tampa, FL Toronto, ON Tulsa, OK Tupelo, MS Washington, DC Yellowknife, NWT Yuma, AZ  4 5 6  Bismarck, ID Burlington, VE Memphis, TN  7 8  Nashville, TN Williston, ND  Cleaned  1 | Vancouver, BC 2 Victoria, BC 3 Winnipeg, MB  59  Appendix VIII List of Australian cities used as data locations in Training Networks. Included 1 Adelaide, SA 2 Albany, WA 3 Alice Springs, NT 4 Bagara, QL 5 Bega, NSW 6 Broadmeadows, VI 7 Bracken Hill, NSW 8 Broome, WA 9 Bundaberg, QL 10 Cairns, QL 11 Canberra, CT 12 Ceduna, SA 13 Charleville, QL 14 Cobar, NSW 15 Cowra, NSW 16 Darwin, NT 17 Derby, WA 18 Esperance, WA 19 Forrest, WA 20 Geraldton, WA 21 Gladstone, QL 22 Kalgoorlie. WA 23 Leigh Creek, SA 24 Lockhart, QL Cleaned 1 I Brisbane, QL 2 | Cambridge, TA 3 I Coffs Harbour, NSW 4 Daly Water, NT 5 i Flinders Island, TA 6 | Halls Creek, WA 7 | Launceston, TA  25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48  Longreach, QL Mackay, QL Meekatharra, WA Melbourne, VI Merimbula, NSW Mildura, NSW Moree, NSW Mount Gambier, SA Mount Isa, QL Narrabri, NSW Newcastle, NSW Nhill, VI Nowra, NSW Oodonattta Orange, NSW Port Hedland Richmond, NSW Rockhampton, QL Sydney, NSW Tamworth. NSW Townsville, QL Wagga, NSW Williamstown, NSW Woomera, SA  8 9 10 11 12 13  Normanton, QL Onslow, WA Perth, WA Portland, VI Sale, VI Warrnambool, VI  Appendix IX  List of Chinese cities used as production sample set for the final Network and their hazard classification. H= High; ML= Moderate- Low; N= Negligible; UC= Unclassified  133  Aksu  High -0,0478820  55  Altay  0,0081563  54  Anda  53  Anqog  198  Anquing  -0,0101976 0,0520491 1,0526654  52  Arxan  132  Bachu  131  Bag Gol  51  Bagou  Nr.  City  269  Baicheng  130  Baoding  129  Baoji  50  Baoqing  128  Baotou  127  Bawolung  49  Bayan Ul Hot  0,0589963 -0,0543622 -0,0527888 0,0609588 -0,0321178 -0,0554898 -0,0465718 0,0542742 -0,0536071 -0,0316333 0,0461892 -0,0070335 -0,0549091 0,5961718  Mod.-Low 1,0516681 -0,0431462  Negligible -0,0534276  Class.  1,0492380  N  -0,0269937  1,0419633 1,0470228 -0,0545284  N  -0,0455148 -0,0538089 -0,0508202  N  0,0287494  205  Changsha  125  Changzhi  196  Changzhou  124  Chaoyang  Beijing  267  Bengbu  266  Bijie  265  Bose  48  Bugt  126  Cangzhou  264  Changcheng  273  Changchun  197  Changde  47  N  0,2194713 0,8832170 1,0555379 -0,0553538 1,0519843 -0,0543141 -0,0503470 1,0519587 1,0544369 -0,0538968 1,0187258 -0,0418349 -0,0499546 1,0519213  Changling  141  H ML  0,7024650 0,2729147 0,0244289 -0,0535081 0,7023858 -0,0346452 1,0431536  Baykurt  N  1,0551597 1,0527984 -0,0516699 -0,0501840 1,0515403  0,1223628 1,0553509 0,4352295 0,3483897 0,7223043 -0,0482474 1,0548483 0,2648498 0,3367075 -0,0472911  268  1,0522354 -0,0548747  ML  ML N ML ML N ML ML N  0,8762362  N  -0,0550268 -0,0550867  ML  -0,0537080 -0,0554967 1,0515883 -0,0547248 -0,0554415 0,7849371 -0,0550342  UC  ML  -0,0427364  1,0465471  N  1,0533446  -0,0542597  -0,0544948  H  -0,0484776  1,0524789  -0,0538680  ML  1,0472542  -0,0497377  -0,0544649  H  -0,0548323  1,0553209  -0,0549888  ML  UC UC N UC N H  61  Appendix IX (con't) List of C h i n e s e cities u s e d a s production s a m p l e set for the final Network a n d their h a z a r d classification.  123  Chengde  -0,0546459  1,0552558  -0,0549344  ML  214  Chengdu  1,0536986  -0,0544519  -0,0541697  H  263  Chengzitan  0,0013157  0,1591037  0,7916559  N  195  Chenzhou  1,0310950  -0,0383913  -0,0551660  H  122  Chifeng  -0,0529146  1,0541240  -0,0538710  ML  194  Chin-men  1,0542552  -0,0548247  -0,0542665  H  262  Chishui  -0,0218059  0,9908237  -0,0374058  ML  121  Chongcheng  -0,0289405  1,0210656  -0,0451154  ML  213  Chongging  1,0537974  -0,0545445  -0,0544081  H  193  Chongqing  1,0537974  -0,0545445  -0,0544081  H  46  Da Qaidam  0,0506408  -0,0498341  1,0516268  N  -0,0363138  0,3660836  0,7713114  N  261  Daban  192  Dadong  1,0545194  -0,0549829  -0,0541422  H  120  Dalaoba  -0,0554237  1,0555183  -0,0552767  ML  119  Dalian  -0,0435706  1,0494030  -0,0530783  ML  260  Datong  -0,0384767  0,9729596  0,0038265  ML  118  Dezhou  -0,0547529  1,0552957  -0,0550072  ML  191  Dongsha  1,0524032  -0,0535478  -0,0541235  H  190  Dongtai  1,0493746  -0,0514477  -0,0545669  H  45  Dongxiaoshan  0,0648803  -0,0510022  1,0522178  N  189  Dongzhen  1,0086864  -0,0208553  -0,0552589  H  44  Dund Hot  0,0018658  -0,0401693  1,0480062  N  117  Dunhuang  -0,0554281  1,0555197  -0,0552809  43  Dut Nur  0,0342765  -0,0490371  1,0516828  N  42  Emei  0,0613269  -0,0486002  1,0498407  N  188  Enshi  1,0443926  -0,0481767  -0,0550113  H  259  Erdene  -0,0376666  0,1892097  0,9434824  N  116  Fengcheng  -0,0498492  1,0529325  -0,0538508  ML  115  Fengyi  -0,0356908  1,0463989  -0,0533510  ML  41  Fujin  0,0063822  -0,0427455  1,0491611  258  Fuyang  0,1187629  0,9405760  -0,0546311  ML  207  Fuzhou  1,0524951  -0,0537233  -0,0546445  H  114  Gaotai  -0,0476053  1,0504924  -0,0526244  ML  257  Garze  -0,0042049  0,3069252  0,5902554  UC  0,0208559  -0,0353551  1,0405978  N  56  Golmud  ML  N  62  Appendix IX (con't) List of C h i n e s e cities u s e d a s production s a m p l e set for the final N e t w o r k a n d their h a z a r d classification.  187  Gongguoqiao  40  Gonghe  186  Guangnan  202  Guangzhou  256  Guilin  204  Guiyang  255  Gushi  185  Guzhou  39  Gyamotang  38  Gyumai  199  Haikou  57  Hailar  37  Hailun  36  Hailut  113  Hami  208  Hangzhou  112  Har Orbog  274  Harbin  211  Hefei, Anhui  111  Hejing  254  Hengshan  184  Hengshi  253  Heyuan  110  Heze  35  Hezou  34  Hoboksar  272  Hohhot  183  Hong K o n g  33  Honggor  109  Hotan  108  Hua Shan  32  Huade  107  Huailai  31  Huangheyan  106  Huangshan  1,0520517 0,0569493 1,0479699 1,0527731 0,7708365 1,0535977 0,6383880 1,0542914 0,0439980 0,0680215 1,0544515 0,0295402 0,0027064 0,0025101 -0,0555414 1,0517331 -0,0554826 -0,0413767 1,0053170 -0,0427977 0,4738025 1,0529034 0,9813558 -0,0451757 0,0627609 0,0217580 -0,0357264 1,0480183 0,0449566 -0,0534297 -0,0374591 0,0313924 -0,0541933 0,0627419 -0,0359123  -0,0532934 -0,0491553 -0,0500804 -0,0539016 0,1962754 -0,0543791 0,4317336 -0,0548443 -0,0401543 -0,0510389 -0,0549425 -0,0488285 -0,0437474 -0,0201060 1,0555522 -0,0531768 1,0555358 0,6224316 -0,0099421 1,0234351 0,6079580 -0,0539368 0,0016407 1,0514506 -0,0500658 -0,0467711 0,8812821 -0,0507285 -0,0498710 1,0548204 1,0283720 -0,0473919 1,0550878 -0,0509043 1,0375119  -0,0539998 1,0506694 -0,0539669 -0,0546088 -0,0554267 -0,0541169 -0,0540349 -0,0542227 1,0416309 1,0521533 -0,0541782 1,0517538 1,0502023 1,0317340 -0,0554431 -0,0545820 -0,0553354 0,5339415 -0,0543299 -0,0336291 -0,0537082 -0,0543639 -0,0553119 -0,0540754 1,0513459 1,0506553 0,0891900 -0,0549243 1,0518991 -0,0546962 -0,0434293 1,0504937 -0,0548125 1,0521914 -0,0492346  H N H H H H UC H N N H N N N ML H ML UC H ML UC H H ML N N ML H N ML ML N ML N ML  63  Appendix IX (con't)  List of Chinese cities used as production sample set for the final Network and their hazard classification. 182  Huili  105  Huimin  30  Huma  252  Huoshan  103  Ikanbujmal  104  Ikanbujmal  29  Jiamusi  102  Ji'An  101  Jiangjunmiao  181  Jiangxi  180  Jiaodongshan  28  Jiaohe  179  Jiaojiang  100  Jiashi  251  Jiayuguan  99  Jiexiu  250  Jilin  137  Jinan  178  Jingdezhen  98  Jinghe  27  Jingjiadian  97  Jinzhou  177  Jiujiang  26  Jixi  96  Karhan  95  Kuandian  203  Kunming  135  Lahsa  94  Langmai  176  Lengshuitan  134  Lianyungang  249  Lianzhou  93  Lijiang  175  Lincang  248  Lindong  1,0537380 -0,0549916 0,0494053 0,9427582 -0,0555477 -0,0555477 -0,0046570 -0,0495585 -0,0478644 1,0545206 1,0546030 -0,0100069 1,0507431 -0,0525498 -0,0377703 -0,0491244 -0,0315806 -0,0533400 1,0481396 -0,0515534 0,0556996 -0,0485872 1,0402594 -0,0165204 -0,0539407 -0,0546031 1,0517533 -0,0362549 -0,0395271 1,0314214 -0,0313864 0,9804840 -0,0281789 1,0486052 -0,0109683  -0,0544719 1,0553781 -0,0503022 0,0608849 1,0555537 1,0555537 -0,0372614 1,0503185 1,0168765 -0,0549835 -0,0550325 -0,0191499 -0,0525738 1,0544701 0,9684724 1,0531070 0,3523263 1,0547759 -0,0508067 1,0531542 -0,0475309 1,0512955 -0,0451573 0,0209843 1,0550087 1,0551328 -0,0530603 1,0387436 1,0488912 -0,0386371 1,0447467 0,0023750 1,0430042 -0,0506882 0,0420468  -0,0540807 -0,0550716 1,0521018 -0,0541205 -0,0554668 -0,0554668 1,0471763 -0,0514349 -0,0082799 -0,0541337 -0,0540774 1,0365166 -0,0547941 -0,0545430 0,0057877 -0,0541853 0,7355955 -0,0547755 -0,0549151 -0,0533619 1,0490068 -0,0528915 -0,0550645 1,0104276 -0,0547875 -0,0546139 -0,0540102 -0,0496781 -0,0535237 -0,0551636 -0,0540954 -0,0553131 -0,0534193 -0,0539229 0,9830558  H ML N H ML ML N ML ML H H N H ML ML ML UC ML H ML N ML H N ML ML H ML ML H ML H ML H N  64  Appendix IX (con't)  List of Chinese cities used as production sample set for the final Network and their hazard classification. 92  Linfen  247  Lishui  91  Litang  246  Liuzhou  174  Longping  173  Longyan  245  Longzhou  25  Mangnai  90  Matang  244  Mawu  243  Meihekou  89  Meixing  242  Mengshan  172  Mengzi  171  Mexian  24  Moqiang  23  Mudanjiang  206 241  Nanchang Nanchangshan Dao  240  Nancheng  88  Nanjian  210  Nanjing  201  Nanning  170  Neijiang  22  Nenjiang  239 87  Ondor S u m Pingliang  238  Pingnan  237  Qamdo  21  Qelin Ul  20  Qiang  86  Qiemo  19  Qigzhi  85  Qiitaojing  -0,0545572 0,5475677 -0,0388500 0,7174322 1,0297845 1,0155861 0,9996339 0,0303518 -0,0326188 -0,0173636 -0,0164093 -0,0390048 0,8030757 1,0544675 1,0538808 0,0598982 0,0099702 1,0494369  1,0552289 0,4259904 1,0403630 0,2496020 -0,0373965 -0,0263590 -0,0135182 -0,0474001 1,0222368 0,6276115 0,0735873 1,0482012 0,1647067 -0,0549519 -0,0545966 -0,0508124 -0,0357035 -0,0516957  -0,0549335 -0,0554657 -0,0495941 -0,0554385 -0,0551745 -0,0552383 -0,0552803 1,0505686 -0,0430597 0,2518486 0,9616420 -0,0533122 -0,0554181 -0,0541664 -0,0543919 1,0521976 1,0432205 -0,0548652  ML  -0,0245892 0,7862541 -0,0415114 1,0023135 1,0522258 1,0538348 0,0075126 -0,0451026 -0,0382217 0,9460685 -0,0295012 0,0469643 0,0310262 -0,0549160 0,0689124 -0,0554904  0,8955488 0,1811282 1,0352002 -0,0064280 -0,0535497 -0,0545653 -0,0459146 0,9230675 1,0404405 0,0319669 0,9915723 -0,0500698 -0,0447049 1,0553530 -0,0512125 1,0555378  0,0279219 -0,0554227 -0,0449013 -0,0542491 -0,0546754 -0,0543797 1,0511201 0,1233554 -0,0498616 -0,0553503 -0,0285469 1,0519912 1,0481499 -0,0550271 1,0523015 -0,0553406  ML  UC ML UC H H H N ML UC N ML H H H N N H  H ML H H H N ML ML H ML N N ML N ML  65  Appendix IX (con't) List of C h i n e s e cities u s e d a s production s a m p l e set f o r the final N e t w o r k a n d their h a z a r d classification.  138  Qingdao  84  Qingshui  83  Qinhuangdao  169  Qionghu  236  Qiqihar  235  Qiuqiao  168  Quzhou  5  Sagamagsumdo  82  Sangejing  81  Shache  234  Shandan  209  Shanghai  18  Shangkuli  167  Shangnan  17  Shangzhi  166  Shantou  165  Shanwei  164  Shaoguan  233  Shengping  142  Shenyang  80  Shidao  79  Shidongsi  232  Shihezi  139  Shijiazhuang  78  Shuangchengzi  77  Shuiji  231  Siping  16  Suiyang  15  Sunwu  230  Taicheng  270  Taipei  76  Taiyuan  163  Tancheng  229  Tangjiang  75  Tangshan  -0,0404085 -0,0423067 -0,0495322 1,0479880 -0,0379989 0,3571627 1,0475477 0,0657021 -0,0539762 -0,0540193 -0,0235437 1,0510232 0,0326231 1,0541863 0,0566275 1,0543319 1,0540773 1,0389118 0,0224181 -0,0490600 -0,0407751 -0,0419475 -0,0380031 -0,0555169 -0,0551898 -0,0429766 -0,0420579 0,0269313 0,0188017 0,8081509 0,9890318 -0,0432711 1,0544442 0,9997117 -0,0513567  1,0487141 1,0484332 1,0529763 -0,0500769 0,2404601 0,7273378 -0,0503990 -0,0504967 1,0550111 1,0550370 0,5673663 -0,0525648 -0,0491116 -0,0547830 -0,0505482 -0,0548710 -0,0547169 -0,0442277 0,0067193 1,0498890 1,0482916 1,0485466 0,9230892 1,0555456 1,0553978 1,0502461 0,9551041 -0,0458745 -0,0476509 0,1597841 -0,0047772 1,0490831 -0,0549379 -0,0135812 1,0540009  -0,0533989 -0,0528347 -0,0539809 -0,0540588 0,9069886 -0,0545132 -0,0549359 1,0516768 -0,0547740 -0,0548028 0,3722106 -0,0542936 1,0518218 -0,0542920 1,0520607 -0,0542344 -0,0543319 -0,0550991 0,9854905 -0,0513388 -0,0531209 -0,0529677 0,0554386 -0,0553922 -0,0548800 -0,0536364 0,0423002 1,0495457 1,0514888 -0,0554166 -0,0553001 -0,0529681 -0,0541766 -0,0552802 -0,0543982  ML ML ML H N UC H N ML ML UC H N H N H H H N ML ML ML ML ML ML ML ML N N H H ML H H ML  66  Appendix IX (con't)  List of Chinese cities used as production sample set for the final Network and their hazard classification. 14  Taxkorgan  13  Temeke  162  Tengchong  140  Tianjin  161  Tingzhou  12  Tonghe  74  Tongliao  228  Tongmu  160  Tongshan  73  Tongyuanpu  72  Tuanjie  11  Tulihe  159  Tunchang  71  Turpan  227  Ulan Balgas  226  Ulan Hot  10  U s a n Turu  70  Usu  69  Weifang  158  Wenzhou  68  Wudu  212  Wuhan  157  Wuhu  156  Wuzhou  9  Xarag  155  Xiaguan  136  Xi'an  154  Xichang  225  Xifengzhen  224  Xigaze  8  Xilangzi  7  Xingquanbu  153 6 271  Xingren Xingxingxia Xining  0,0540243 0,0417489 1,0541255 -0,0517810 1,0239309 -0,0167690 -0,0457866 0,5242935 1,0543559 -0,0489598 -0,0450532 0,0542440 1,0542821 -0,0555544 -0,0368865 -0,0390753 0,0356060 -0,0449917 -0,0539995 1,0539686 -0,0228423 1,0521259 1,0512152 1,0245402 0,0383858 1,0526366 -0,0348656 1,0533811 -0,0152171 -0,0022251 0,0216563 0,0364016 1,0540693 -0,0082979 -0,0139105  -0,0492205 -0,0497846 -0,0547310 1,0541880 -0,0328975 -0,0233401 1,0236612 0,4507406 -0,0548820 1,0487806 1,0430793 -0,0505773 -0,0548410 1,0555553 0,9457180 0,4015754 -0,0486311 1,0175596 1,0550256 -0,0546501 1,0406431 -0,0534437 -0,0527620 -0,0333705 -0,0427551 -0,0537004 1,0466097 -0,0542434 0,6089731 0,2690435 -0,0436766 -0,0443402 -0,0546942 0,0082832 0,4881033  1,0508809 1,0519593 -0,0541227 -0,0544717 -0,0552055 1,0419986 -0,0269654 -0,0554688 -0,0541755 -0,0502721 -0,0481293 1,0521730 0,0542556 -0,0555147 0,0251012 0,7678722 1,0512824 -0,0221016 -0,0548466 -0,0543623 -0,0538004 -0,0545561 -0,0544581 -0,0552028 1,0454015 -0,0540770 -0,0539885 -0,0542320 0,2598718 0,6366042 1,0481902 1,0472672 -0,0541022 1,0127076 0,3996146 :  N N H ML H N ML UC H ML ML N H ML ML N N ML ML H ML H H H N H ML H UC UC N N H N UC  67  Appendix IX (con't)  List of Chinese cities used as production sample set for the final Network and their hazard classification. 223  Xinju  152  Xinqiao  151  Xinshan  222  Xinzhou  67  Xuzhou  150  Ya'an  66  Yan'an  221  Yancheng  220  Yanchi  149  Yangjiang  219  Yanji  65  Yantai  148  Yibin  147  Yichang  64  Yichun  63  Yingkou  62  Yining  215  Yong'an  218  Yong'an  146  Yuanling  145  Yuanma  217  Yumen  61  Yuxian  4  Zaindainxol  3  Zalantun  2  Zamar  60  Zhaidian  59  Zhangwu  200  Zhanjiang  58 216 1  Zhanyi Zhaotong Zhidamsumdo  144  Zhijiang  143  Zhongduo  0,9045531 1,0509321 1,0544285 -0,0057773 -0,0346531 1,0533008 -0,0421490 0,1301780 -0,0244506 1,0545547 -0,0284849 -0,0454819 1,0538772 1,0184307 -0,0550901 -0,0454708 -0,0442550 0,8604762 0,8604762 1,0374701 1,0478307 -0,0296011 -0,0482609 0,0580507 -0,0028499 0,0695261 -0,0544508 -0,0541883 1,0533745 -0,0081165 0,0448950 0,0701263 1,0537609 1,0531211  0,0690855 -0,0526999 -0,0549268 0,4054780 1,0460815 -0,0542192 1,0490479 0,9364054 0,8903629 -0,0550038 0,6079072 1,0515178 -0,0545938 -0,0285479 1,0553915 1,0475332 1,0388614 0,1098485 0,1098485 -0,0431419 -0,0505972 0,5921669 1,0486516 -0,0480866 -0,0411599 -0,0511157 1,0551800 1,0550500 -0,0542822  -0,0553783 -0,0547825  H  -0,0541583 0,4554438 -0,0544323 -0,0544279 -0,0531786 -0,0542655 0,0315501 -0,0541221 0,3663099 -0,0539044 -0,0543878 -0,0552234 -0,0550014 -0,0512973 -0,0453425 -0,0553989 -0,0553989 -0,0551083  H  -0,0549289 0,3992282 -0,0506767  H UC ML H ML ML ML H UC ML H H ML ML ML H H H H UC ML  1,0494671 1,0492748  N  1,0521838 -0,0550341  N  N ML ML  1,0268731 0,9984928 -0,0512986 -0,0544944  -0,0547337 -0,0545104 -0,0526926 -0,0535181 1,0523522 -0,0541868  -0,0540588  -0,0541903  H  H ML ML N H  68  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items