The use of agent-based models as virtual laboratories for exploring human-environment interactions in land-use systems.

摘要

Land-cover changes from human land-uses are seen as primary sources of global environmental change. The multitude of forces influencing land change and the complexity of their interactions present substantial challenges for connecting land-use changes to their causes at and between local, regional, and global scales. Due to this complexity, simulation modeling is a primary tool for conceptualizing land-use system structure and testing causal explanations of observed patterns of land changes. Inductive modeling methods, which have been the dominant approach in land change science, have advanced understanding of global patterns and driving forces of land-use and -cover change (LUCC). However, because land-use systems exhibit nonlinear dynamics driven by adaptive responses of human decision-makers, deductive modeling methods that build process-based understanding of land change are also needed. Using two different agent-based models as virtual laboratories, this dissertation explores how individual land users' decision-making processes translate into observed land-use patterns across a spectrum of different land-use systems. The first model represented the decision-making models and interactions of consumer, farmer, and developer agents in land and housing markets in an expanding ex-urban land-use system, which produced development patterns characteristic of 'urban sprawl' from the bottom-up. The same modeling framework was adapted to experimentally test alternative developer decision-making models, and results indicated cognitive biases and credit constraints may lead to inefficient land-use patterns and impediments to adopting ecological subdivision designs. The second model tested mechanistic explanations of empirically-based, system-level patterns of agricultural change due to individual land-user responses to changing environmental, demographic, and economic conditions. Results demonstrated threshold transitions between alternative livelihood strategies defined by interactions between market influence and population pressure. The model was then expanded and applied to a range of real land-use systems. Comparisons across six sites illustrated the varying effects of common processes driving land-use change, and revealed important mechanisms that influenced agents' land-use decisions and resulting patterns of LUCC. Ultimately, agent-based virtual laboratories made explicit and testable connections between individual decision-making processes and system-level patterns of LUCC. I conclude with a synthesis of the insights gained from model experiments, and put forth a framework for a general land systems theory.