This Geography and Spatial Sciences (GSS) project will develop novel methods for understanding and categorizing uncertainty in complex spatial models of human and environmental systems interactions. Increasingly scientists are utilizing such models to understand the dynamics of various processes including urban expansion, climate change, deforestation, and nonpoint pollution of water bodies. An inherent complexity in these processes, leading to model uncertainty, is the variability over both space and time. For example, the dynamics of urban expansion in Detroit are different than those in Cleveland, and their dynamics today are different than those ten years ago. Models of such systems typically require a large number of explanatory variables that describe the economic, social, political, and environmental components of the overall system. To address model uncertainty, scientists use sensitivity analysis: a technique employed to understand how different values of an explanatory variable affect the outcome of interest. For example, how do increasing payments to farmers affect their willingness to convert agricultural lands for conservation purposes? This project will develop a new theoretical framework for applying sensitivity analysis to complex human-environment systems models exhibiting great variability over space and time. Specifically, the framework will facilitate 1) the identification of explanatory variables that account for the greatest variability in model outcomes leading to model simplification; 2) the exploration of the model outcome variability over space and time; and 3) the simulation of system dynamics to understand the implications of various policy scenarios. The developed framework will be tested on a model designed to understand agricultural land conservation decisions and their effects on lakes, local economies, and people.
This project will contribute greatly to scientists', policy-makers', and citizens' abilities to understand the extent to which various drivers in human-environment systems models affect model outcomes. Moreover, the project will facilitate the development of simpler systems models with greater explanatory power. As such, the methods developed will improve the transparency of such models potentially allowing greater community use, understanding, and participation in such modeling exercises. The framework will also provide methods for prioritizing data acquisition, improving the robustness of complex human-environment models, and enhancing the usefulness of such modeling endeavors for policy and decision making to address major societal problems. The project will train students, through graduate student mentoring and a modeling certification graduate program, to employ the new methods developed in order to describe and solve complex human-environment systems problems. The results of our research, including a spatiotemporal sensitivity analysis toolbox (ST-SA), will be disseminated electronically online, in peer reviewed manuscripts, and at professional conferences.
