Show simple item record

dc.contributor.advisorBuckeridge, John
dc.contributor.advisorSallis, Philip
dc.contributor.authorShanmuganathan, S
dc.date.accessioned2008-04-18T01:12:23Z
dc.date.available2008-04-18T01:12:23Zen_US
dc.date.copyright2006-09-20
dc.date.issued2006-09-20
dc.identifier.urihttp://hdl.handle.net/10292/327
dc.description.abstractThis research is a case study evaluation of the use of self-organising map (SOM) techniques for ecosystem modelling to overcome the perceived inadequacies with conventional ecological data analysis methods. SOMs provide an analytical method within the connectionist paradigms of artificial neural networks (ANNs), developed from concepts that evolved from late twentieth century neuro-physiological experiments on the cortex cells of the human brain. The rate and extent at which humans influence environmental deterioration with commensurate biodiversity loss is a cause for major concern and to prevent further degradation by human impact, parsimonious models are urgently needed. Indeed, the need for better modelling techniques has never been so great. Ecologists and many national and international bodies see the situation as 'significantly critical' for the conservation of our global ecosystem to foster the continued wellbeing of humanity on this earth.The thesis investigates and further refines SOM based exploratory data analysis methods for modelling naturally evolving, highly diverse and extremely complex ecosystems. Earlier studies provide evidence on SOM ability to analyse complex forest and freshwater biological community structures at limited scales. On the other hand, growing concerns over conventional methods, their soundness and ability to model large volumes of data are seen as of little use, leading to arguments on the results derived from them. Case study chapters illustrate how SOM methods could be best applied to analyse often 'cryptic' ecosystems in a manner similar to that applied in modelling highly complex and diverse industrial system dynamics. Furthermore, SOM based data clustering methods, used for financial data analysis are investigated for integrated analysis of ecological and economic system data to study the effects of urbanisation on natural habitats.SOM approaches prove to be an excellent tool for analysing the changes within physical system variables and their effects on the biological systems analysed. The Long Bay-Okura Marine Reserve case study elaborates on how SOM based approaches could be best applied to model the reserve's intertidal zone with available numeric data. SOM maps depicted the characteristic microclimate within this zone from ecological monitoring data of physical attributes, without any geographical data being added. This kind of feature extraction from raw data is found to be useful and is applied to two more case studies to study the slow variables of ecosystems, such as population dynamics, and to establish their correlation with environmental variations. SOM maps are found to be capable of distinguishing the human induced variations from that of natural/ global variations, at different scales (site, regional and global) and levels using regional and global data. Hence, SOM approaches prove to be capable of modelling complex natural systems incorporating their spatial and temporal variations using the available monitoring data, this is a major advantage observed with SOM analyses.In the third case study, potential use of SOM techniques to analyse global trends on the effects of urbanisation in environmental and biological systems are explored using the World Bank's statistical data for different countries. Many state and international institutions, concerned over global environmental issues, have made attempts to develop indicators to assess the conditions of different ecosystems. The enhancements with SOM approaches against the currently recommended indicator system based on information pyramid and pressure-state-response (PSR) models are elaborated upon.The research results of SOM methods for ecosystem modelling, similar to that applied to industrial process modelling and financial system analysis show potential. SOM approaches (i.e. cluster, dependent component, decision system and trajectories/ time series analyses) provide a means for feature extraction from the available numeric data at different levels and scales, fulfilling the urgent need for modelling tools to conserve our global ecosystem. They can be used to bridge the gap in converting raw data into knowledge to inform sustainable ecosystem management. Increasingly, traditional methods based on Before-After-Control-Impact (BACI) designs and Analysis of Variance (ANOVA) are seen to be unsuitable for ecological data analysis, as they are unable to detect human induced environmental impacts from that of a natural cause. This thesis proves that SOM techniques could be applied to modelling not only a natural systems complexity but also its functioning and dynamics, incorporating spatial as well as temporal variations, to overcome the constraints with conventional methods as applied in other stated disciplines.
dc.format.mediumapplication/pdfen_US
dc.publisherAuckland University of Technology
dc.subjectBiotic communities
dc.subjectEcosystem management
dc.subjectClassification
dc.subjectEnviornmental studies
dc.titleSoft systems analysis of ecosystems
dc.typeThesis
thesis.degree.grantorAuckland University of Technology
thesis.degree.levelDoctoral Theses
thesis.degree.nameDoctor of Philosophy
thesis.degree.disciplineFaculty of Science and Engineeringen_US
dc.rights.accessrightsOpenAccess


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record