Application of Sensitivity and Uncertainty Analyses to Linear Time-Invariant Compartmental Models

Gazioglu, Suzan (2002) Application of Sensitivity and Uncertainty Analyses to Linear Time-Invariant Compartmental Models. PhD thesis, University of Glasgow.

Full text available as:
[thumbnail of 13834184.pdf] PDF
Download (37MB)


Chapter 1 reviews the fundamental aspects of modelling and introduces sensitivity and uncertainty issues. Chapter 2 first introduces and reviews the theory of linear, time-invariant compartmental models, then describes a number of methods used to solve model state equations analytically and numerically in order to make predictions. This chapter also describes the methodology of numerous sensitivity analysis methods. In Chapter 3, application of various sensitivity analysis techniques to two 8- compartment global carbon cycle models is presented. For ease of comparison, a measure of similarity between the sensitivity conclusions from different methods is defined based on the top 10 ranked input factors according to each method for each output variable (i.e. for each compartment at chosen time points). Chapter 4 presents the results of the application of various sensitivity analysis methods including non-parametric methods to a more complex 25-compartment global carbon cycle model. An overall informal comparison indicates that the 8-compartment global carbon cycle models used in Chapters 3 and 4 are optimal with respect to efficiency (i.e. both are simple and model codes are not very time-consuming to run), but in return do not have a high degree of stability and reliability since they do not adopt biological and chemical processes. As for the 25-compartment model, it is more complex and more costly to run. These chapters review the applicability of the sensitivity analysis methods to these models which has steady-state constrain. Chapter 5 explores various sources of uncertainty and presents results of uncertainty analysis applied to the three global carbon cycle models that are used in Chapters 3 and 4. Here, we partition the overall prediction uncertainty of an output variable into different components of uncertainty. Finally, Chapter 6 presents conclusions and main findings of the thesis.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Additional Information: Adviser: Marian E Scott
Keywords: Statistics
Date of Award: 2002
Depositing User: Enlighten Team
Unique ID: glathesis:2002-76404
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 19 Nov 2019 14:43
Last Modified: 19 Nov 2019 14:43

Actions (login required)

View Item View Item


Downloads per month over past year