Ecological Modelling 220: 1968-1983. 2009.
Abstract. We present a method of multi-criteria assessment for the analysis of process model uncertainty that
combines analysis of model structure, parameters and data requirements. There are three components
in calculation and definition of uncertainty.
(1) Assessment criteria: Uncertainty in a process model is reduced as the model can simultaneously simulate
an increased number of assessment criteria selected to test specific aspects of the theory being
investigated, and within acceptable limits set for those criteria. This reduces incomplete specification
of the model—the characteristic that a model may explain some, but not all, of the observed features
of a phenomenon. The calculation required is computation of the Pareto set which provides the list of
simultaneously achieved criteria within specified ranges.
(2) Parameter values: Uncertainty in a process model is reduced as the distribution of values for parameters
simulating multiple assessment criteria within their acceptable limits becomes unimodal and
with reduced range. This reduces non-uniqueness in the model—the characteristic that there may be
alternative representations and no a priori way of deciding between them. The calculation required is
analysis of parameter values within each group of the Pareto set, for their distributions and possible
correlations which contributes to a definition of non-uniqueness in terms of whether multiple modes
in parameter values are equally plausible.
(3) Data and information: Uncertainty in a process model is reduced as the acceptable limits for assessment
criteria are defined with increasing precision. The calculations required are to define acceptability
ranges for the assessment criteria either through empirical investigations or through inference from
related theories. This provides a definition of the relationship of the model to empirical and theoretical
construction thought to be important.
Three types of criteria should be used in assessment of uncertainty: (i) Tests of the quantitative domains
of the model outputs. (ii) Examination of features explicitly predicted by the theory that the model
describes. (iii) Tests between alternative structures in model formulation. The ability of a model to satisfy
these criteria simultaneously is calculated using evolutionary computation and results are presented as
a Pareto optimal set. Deficiencies in the model are revealed by manipulating acceptable limits around
the assessment criteria, inclusion and exclusion of different criteria, and examining the distribution of
model parameter values as non-dominated groups achieving different combinations of assessment criteria
form in the Pareto optimal set. We illustrate this procedure through application to a simple model
of competition between plants in single species even-aged populations. We use six assessment criteria
and test an alternative to the model.