Methodological Problems in the Modeling of Ecosystems and Ways of Solutions

Abstract

The great variety and complexity of natural ecosystems leads to methodological problems of simulation and lack of understanding of a common framework in the organization of ecosystems prevents the construction of the universal ecosystem models that are suitable for efficient integration in the model of the biosphere and for the successful solution of theoretical and practical problems. The aim of this work was to develop a systematic theoretical basis for universal description, analysis, and modeling of ecosystems as structural units of the biosphere as the highest-level ecosystem. Analysis of the literature and special studies of the authors demonstrated that a reliable basis for universal models is a multidisciplinary approach that includes three types of concepts (substrate, energy and information) and is based on such fundamental properties of living systems as the attractiveness, adaptability, fractal, network organization. It was found that the most promising are the methods of computer mathematics, based on adaptive networks. The network models can be formed using various tools for operation with artificial neural networks that were developed in neuroinformatics. The entire range of the model states, from excessive to minimally possible, can be investigated. In the network models, one can obtain arbitrarily complex static and dynamic regimes correctable both in the process of model tuning (identification) and during operation of the model (adaptation). The models based on the self-organizing adaptive networks can potentially reflect the most general and fundamental properties of the complex natural systems.