TY - BOOK

T1 - Data-based Mechanistic Modelling (DBM) of Nonlinear Environmental Systems.

AU - Fawcett, Christopher P

PY - 1999

Y1 - 1999

N2 - The main focus of the research studies presented in this thesis centre on the application and development of data-based mechanistic (DBM) transfer function models for nonlinear environmental systems. The data-based mechanistic modelling approach exploits the available time series data, in statistical terms, to expose the model structure, generating dynamic stochastic models that are parsimonious in nature and can be interpreted in a physical manner. For nonlinear systems, the DBM approach centres on the use of transfer function models whose parameters are free to vary over time. The presence of such time variable parameters may reflect either nonstationarity or nonlinearity; the latter arising if the variations are also shown to be state dependent. Statistical time series methods for estimating time varying parameters (TVP) and the use of these methods in state dependent parameter modelling (SDPM) are discussed and applied to the modelling of nonlinear ecological population dynamics and hydrological processes. Further, DBM modelling techniques are applied to an oceanic ecosystem simulation model in order to investigate model uncertainty and over-parameterisation, set within the context of data assimilation. In each application, the DBM methodology is shown to successfully identify the system nonlinearities, so providing additional physical insight and ensuring a good explanation of the data with the minimum of model parameters (parsimony). Further, the DBM approach provides an approach to both the evaluation and reduction in the complexity of highly parameterised simulation models.

AB - The main focus of the research studies presented in this thesis centre on the application and development of data-based mechanistic (DBM) transfer function models for nonlinear environmental systems. The data-based mechanistic modelling approach exploits the available time series data, in statistical terms, to expose the model structure, generating dynamic stochastic models that are parsimonious in nature and can be interpreted in a physical manner. For nonlinear systems, the DBM approach centres on the use of transfer function models whose parameters are free to vary over time. The presence of such time variable parameters may reflect either nonstationarity or nonlinearity; the latter arising if the variations are also shown to be state dependent. Statistical time series methods for estimating time varying parameters (TVP) and the use of these methods in state dependent parameter modelling (SDPM) are discussed and applied to the modelling of nonlinear ecological population dynamics and hydrological processes. Further, DBM modelling techniques are applied to an oceanic ecosystem simulation model in order to investigate model uncertainty and over-parameterisation, set within the context of data assimilation. In each application, the DBM methodology is shown to successfully identify the system nonlinearities, so providing additional physical insight and ensuring a good explanation of the data with the minimum of model parameters (parsimony). Further, the DBM approach provides an approach to both the evaluation and reduction in the complexity of highly parameterised simulation models.

KW - MiAaPQ

KW - Environmental science.

M3 - Doctoral Thesis

PB - Lancaster University

CY - Lancaster

ER -