Pattern-Oriented Validation and Sensitivity Analysis Methods for Spatially Explicit Agent-Based Models
MetadataShow full item record
Spatially explicit agent-based models (ABMs) have been used to capture dynamic spatio-temporal phenomena at an individual-level. In the ABMs, the phenomena under study are driven by interactions among heterogeneous individual agents and the environment. The behaviors and interactions of agents are often specified from incomplete knowledge, which may be responsible for uncertainty in the model outcomes. To measure this uncertainty and examine the accuracy of the model outcomes in relation to referenced observations, model validation and sensitivity analysis (SA) should be performed. As a theoretical framework, pattern-oriented modeling (POM) helps in evaluating model fit by comparing simulated results with multiple patterns of observations. In this dissertation, I proposed and tested pattern-oriented validation and SA methods for spatially explicit ABMs. Spatially explicit ABMs of vector-borne disease transmission in a community were used as the case study. The specific objectives of this dissertation are threefold: (1) to test hypotheses about the impact of the characteristics of dengue virus transmission on model outcomes, (2) to propose a conceptual framework of pattern-oriented validation method for spatially explicit ABMs, and (3) to perform pattern-oriented SA for spatially explicit ABMs. I tested the influence of model specifications and parameterizations of the dengue virus (DENV) transmission process on dengue outbreak patterns at macro and micro space-time scales. The results indicate that model specifications about spatial configurations of residential area, mosquito population density, and immunity status of individual humans are of importance to reproduce observed patterns of DENV outbreaks at multiple space-time scales. Thus, the assumptions about behaviors and interactions of agents need to be carefully considered and embedded in the models. I also found that the impacts of input parameters on variability in the space-time patterns differ by space-time scales. This dissertation highlights the potential important of the use of multiple space-time scale patterns in validation and SA for spatially explicit ABMs.