Self-organising map methods in integrated modelling of environmental and economic systems

Abstract

The need for better techniques, tools and practices to analyse ecological and economic systems within an integrated framework has never been so great. Many institutions have made tremendous efforts in the implementation of sustainable environment management based on ‘integrated’ approaches, as opposed to that of late 20th century's in-depth knowledge or ‘reductionism’ concepts. However, achieving sustainable environment management seems remote, as our understanding of ecosystem response to human influence is insufficient to predict the environmental outcome of proposed development activities. This has left environmentalists and land developers wrangling over the reliability of current environmental modelling techniques, assessment methodologies and their results. As a result, ecosystems continue to deteriorate with commensurate biodiversity loss. The paper elaborates on how self-organising map (SOM) methodologies within the connectionist paradigms (connectionist paradigms refer to the late 20th century neural network architectures) of artificial neural networks (ANNs) could be applied to disparate data analysis at two different scales: regional (using river water quality monitoring data to evaluate ecosystem response to human influence) and global (for modelling of environmental and economic system data and trade-off analysis) within an integrated framework to inform sustainable environment management.

Introduction

The need for better techniques, tools and practices to analyse ecological and economic systems within an integrated framework at wider scales has never been so great. Many environmentally concerned communities, scientists and international institutions, agree that better modelling techniques and tools are needed for an integrated analysis of human interaction with naturally evolving, highly complex and diverse ecosystems. This will allow humans and their activities to be sustained by natural systems (Graedel et al., 2001). The emphasis is on developing integrated interdisciplinary modelling techniques; an absolute contrast to the late 20th century's in-depth knowledge-based approaches. Many studies detailed in Section 2 reveal this fact. These studies elaborate on the environmental issues critical to human well-being with recommendations and measures for integrated analysis of ecological and economic indicators conducive to advancing co-ordinated efforts from a range of professionals to preserve our global ecosystem from further degradation. However, despite such efforts, policymakers and land developers continue to pay no attention to scientific predictions of the long-term detrimental effects to the environment, and argue about the reliability of current environmental impact assessment methods. This is due to a belief that environment sustainability invariably leads to socio-economic loss (Buckeridge and Tapp, 1999). Meanwhile, ecosystems continue to deteriorate with commensurate biodiversity loss (Reid, 2000). Thus, in practice achieving sustainable environmental management seems remote. The urgent need for new approaches to achieve sustainable environment management, the challenges faced in introducing interdisciplinary research efforts and the drawbacks with the current ecological modelling methods are explained. Thereafter, Kohonen's (1995) self-organising map (SOM) based artificial neural network (ANN) applications to integrated analysis of ecological and economic system data (at regional and global scales) are illustrated with examples.