Discovery of Ecological System Dynamics Models
Supported by Data Descriptions
In the environmental sciences, ecological system dynamics models
are widely used as a tool for representation of knowledge about
ecosystems, followed, in many cases, by simulation, prediction and
decision making. The bases for model building are ecological
data and human expertise in ecological understanding, data analysis
and modeling. The initial hypothesis of this research project is that
there are connections between ecological data specifications and
ecological system dynamics models which, given a precisely defined
model purpose, can be understood and formalised, bringing about the
possibility of semi-automating model design supported by ecological
data descriptions.
Some current project subgoals are:
Development of an ontological conceptual framework for ecological
data description. A first version of this has been prepared using the
Ontolingua server.
Development of a library of transformations reflecting the possible
connections between ecological data descriptions and model
substructure.
Devising and application of heuristics for design decisions and
refinement. The heuristics will be applied to data descriptions and
model substructure designs (generated through the library of
transformations), giving an optimised model substructure together with
its data support rationale.
We envisage a highly interactive system working on
connections from data to model and vice versa: the user (modeler) may
input a model structure and the system searches for data properties
that support it, or, based on properties of a certain data set, the
system may output a proposed model that can be analysed and further
improved and refined by the modeler. Next, the modeler may perform
simulation runs of the model using existing system dynamics modelling
tools, such as AME or Stella.
