Coupling Agent-Based Models with Bayesian Belief Networks in Social-Ecological Systems Modeling: The Role of Uncertainty in the Resilience of Semi-arid Riparian Corridors in the Sonoran Desert

摘要

Resiliency, or the ability of ecosystems to absorb change, is a particularly challenging goal for systems with a high interconnectivity of their social and ecological sub-systems, dubbed social-ecological systems. Exploring potential effects of policy decisions and/or changes in climate are difficult to study under traditional research methods; however, social-ecological system modeling approaches have overcome many of these difficulties. Uncertainty, particularly in regards to human decision-making, remains a challenge. I developed a series of models to test the use of Bayesian modeling techniques in addressing human-based sources of uncertainty. The models were developed in two semi-arid river basins in the Sonoran Desert to represent model social-ecological systems with three sub-systems: social---urban and agricultural water demand, hydrological---pumping groundwater from the aquifer, and ecological---changes in riparian vegetation communities. The rancher decision-model, developed in the Rio San Miguel basin, used cognitive mapping and Bayesian modeling to express key decisions as a series of probabilities under a variety of environmental conditions. The social-ecological system model of the Rio San Miguel basin, which utilized the rancher decision-model to incorporate real-world uncertainty of the human decision-making process, exposed trade-offs between scenarios that benefited riparian vegetation versus rancher well-being. The Upper San Pedro River Watershed social-ecological system model tested the effects of policy decision-making, specifically on population growth and water conservation, on the spatial distribution of riparian vegetation. The model predicted degradation of the riparian corridor under all population growth/water conservation scenarios, particularly for marsh vegetation and in a specific stretch of the river. Additionally, the model demonstrated how changing the likelihood of adopting new water conservation programs could alter the probability distribution of varying levels of deterioration in the riparian corridor. Since many of the social-ecological systems models are created to aid in decision-making for natural resource managers, we believe that the outputs from models that incorporate Bayesian techniques will have a more accurate representation of the range of possible outcomes of management decisions.