Land use changes within watersheds have spatially explicit dynamics and involve decision making by individuals. The role of the spatial dimension of human behavior and its impact on land use change has been analyzed using agent-based modelling approaches. Agent-based land use change has received a significant theoretical attention; however, these models lack empirical implementation and testing due to the lack of spatial modelling tools and data that can capture human land use dynamics.This research presents a methodology for projecting land use conversions through the implementation of a spatially explicit agent-based simulation model in the Opequon Creek watershed of Berkeley County, West Virginia. Empirical estimates for factors that influence the land use conversion probability are captured using a spatial logistic regression model. Then, agentbased probabilistic land use conversion (APLUC) model is programmed on Python language within a geographic information system (GIS) to explore the impacts of policies on land use conversion decisions using estimates from actual land use change from 2001-2011. A series of model runs are executed under buffer policy scenarios. Three policy scenarios are developed: (1) a scenario where there is no policy implemented, (2) a scenario where 50 ft buffer zones are applied to all streams, and (3) a scenario where 50 ft buffers are applied only on critical source areas (CSAs) watersheds. The land use patterns project in APLUC model are driven by individual land conversion decisions over 50 model runs of 10 iterations each under each policy scenario. The APLUC model is validated at sub-basin level and outcomes are analyzed to identify the influence of various land use policies on land use patterns. The results show that a 50 ft buffer policy everywhere in watershed, greatly reduced the residential land use conversions. Spatial patterns generated under a 50 ft buffer policy in CSAs only showed that future projected land use changes occurred close to major highways. In the baseline policy, most conversions occurred near existing residential land use and urban centers. Results from the APLUC model also suggests that forest is serving as "distant amenity" for residential land conversion.Finally, the impacts of these three policies on water quality are estimated using an ArcSWAT model, a graphical user interface for SWAT (Soil and Water Assessment Tool). This model indicates that the 50 ft buffer policy in CSAs is most effective among the three policies in reducing the pollutant loads. This study suggests that carefully designed policies, which discourage residential land use conversion in CSAs, result in less pollutant loads by shifting the location of residential conversion to less critical areas where agricultural land is dominant in the watershed. iii ACKNOWLEDGEMENTS I wish to express my deepest gratitude to my advisor, Dr. Alan Collins, for his constant encouragement, direction and patience throughout the entire process of my dissertation work. Without his outstanding guidance a...