Qualitative models in ecology and their use in learning environments

This thesis is concerned with the development of qualitative modelling approaches that can be used in educational contexts for simulation and explanation about ecological systems. Students have to learn about complex systems, and computers have great potential for providing tools to support ecology teaching. Most of the simulation models created so far for this purpose are based on mathematical equations. However, quantitative data is often missing. Moreover, mathematical models hardly can support explanations because they lack explicit representation of concepts about the system being modelled and of the causal relations between the modelling components. Qualitative Reasoning has the potential for handling these problems, as it provides ontologies and techniques for building models with qualitative and incomplete knowledge. Accordingly, different modelling formalisms are explored and compared in this thesis. The specific domain chosen is the ecology of the vegetation of the Brazilian cerrado. Recurrent issues in scientific research and teaching are the effects of fire on flowering, germination, establishment, mortality of the cerrado plants, and on the succession of cerrado communities. The qualitative knowledge involved in these issues is represented in a domain theory of plant population dynamics. To implement this domain theory, a framework is proposed in which the structure of the system being modelled is represented as a combination of the conceptual, the causal and the mathematical components. The conceptual structure includes knowledge about the objects, their quantities, quantity relations, typical scenarios, and the mechanisms causing changes in the system. The causal structure is a representation of how changes start and propagate within the system. The mathematical structure is a description of the constraints between the quantities and the procedures for calculating their values. General guidelines for building qualitative models to be used in education are also discussed. A number of models about different ecological problems are developed using the ontology provided by the Qualitative Process Theory (QPT) of Forbus, the qualitative algebra developed for Guerrin' s System of Interpretation of Measurements, Analysis and Observations (SIMAO), and the data structures adopted in the qualitative reasoning shell GARP developed by Bredeweg. Some models are implemented in GARP. In models about the life cycle of cerrado plants, qualitative equations are used to assess the magnitude of the quantities, which in turn are used to determine state transitions. In models about the dynamics of populations and communities, state transitions are determined by the assessment of the causal influences between the quantities. These models can be used for the automatic generation of different kinds of explanations in learning environments. It is shown that if only the mathematical and the causal structures are explicitly represented, the models support explanations about how the values are being calculated. However, if the conceptual structure is also explicit, it is possible to explain why the calculations are being done on the basis of domain knowledge. A set of topics for the explanatory discourse about concepts expressed in qualitative simulations is presented, and the potential of qualitative models for supporting explanations in learning environments is discussed.