An Ecohydrological Cellular Automata Model Investigation of Juniper Tree Encroachment in a Western North American Landscape

Woody plant encroachment over the past 140 years has substantially changed grasslands in western North American. We studied encroachment of western juniper (Juniperus occidentalis var. occidentalis) into a previously mixed shrub–grassland site in central Oregon (USA) using a modified version of Cellular Automata Tree–Grass–Shrub Simulator (CATGraSS) ecohydrological model. We developed simple algorithms to simulate three encroachment factors (grazing, fire frequency reduction, and seed dispersal by herbivores) in CATGraSS. Local ecohydrological dynamics represented by the model were first evaluated using satellite-derived leaf area index and measured evapotranspiration data. Reconstructed pre-encroachment vegetation cover percentages and the National Land Cover Database (NLCD 2006) vegetation map were used to estimate parameters for encroachment factors to represent juniper encroachment in CATGraSS. Model sensitivity experiments examined the influence of each encroachment factor and their combinations on trajectories of modeled percent cover of each plant functional type and emergent spatial vegetation patterns in the modeled domain. Simulation results identified grazing as the key factor leading to juniper encroachment, by reducing shrub and grass cover and promoting the formation of juniper tree clusters. Reduced fire frequency and increased seed dispersal by grazers further amplified juniper encroachment into grassland patches between clusters of juniper trees. Each encroachment factor showed different consequences on modeled vegetation patterns. Time series of modeled plant cover and spatial patterns of plant functional types were found to be consistent with an existing conceptual model described in the literature. The proposed model provides a tool that can be used to improve our understanding of the drivers and processes of woody plant encroachment and vegetation response to global change.