Combining ecohydrologic and transition probability-based modeling to simulate vegetation dynamics in a semi-arid rangeland

• The model couples a hillslope hydrological CA submodel with vegetation transitions.
• Transitions depend on growing-season soil moisture and grazing pressure.
• Grazing pressure represents timing of herder decision-making to relocate.
• Model behavior showed comparable sensitivity to variation in degree of runoff channelization and grazing intensity.
• Simulations revealed multi-decadal switches in hillslope vegetation cover.

Drylands support pastoralist social–ecological systems around the world. Ecological function in these water-limited environments frequently depends on tightly coupled, nonlinear interactions between water, soil, vegetation, and herbivores. Numerous complexity-based approaches have modeled localized ecohydrological feedbacks to yield insights into dryland landscape organization and emergent dynamics. The relevance of these models to management and sustainability continues to increase as researchers incorporate ecological processes at multiple scales and social–ecological variables like herding practices. However, many processes vary in their importance depending on ecological context, so there is a continuing need to construct models tailored to different contexts. We developed a model for semi-arid rangelands that experience highly variable rainfall, substantial Hortonian runoff during rain events, patchy vegetation structure, and grazing-influenced patch transitions. The model couples an existing, mechanistic cellular automata model of hillslope water balance with a dynamic vegetation model in which probabilistic transitions between bare, annual grass, perennial grass patches depend on soil moisture and grazing intensity. The model was parameterized based on a field site in Kenya, from which we had empirical hydrological measurements and several years of patch-to-hillslope scale measurements of vegetation structure. The model domain is a 100 × 100 grid of 2 × 2 m cells, it simulates seasonal cycles of growing seasons followed by dry seasons, and it computes daily soil moisture based on stochastic rainfall forcings. Patch type transitions can occur twice during each seasonal cycle: at the end of the growing season, with probabilities based on average growing-season soil moisture availability; and at the end of the dry season, with probabilities based on grazing intensity and antecedent growing-season soil moisture. By parameterizing grazing intensity as a per-patch impact, it can be interpreted as the degree of forage depletion at which a herder decides to leave the area. We conducted a series of simulation experiments, principally altering runoff channelization and grazing intensity. The model generated plausible vegetation dynamics across the range of grazing intensities simulated. Vegetation cover fluctuated seasonally, but never collapsed completely, even at the highest grazing intensity. At low to intermediate grazing, we observed multi-decadal switches in fractional perennial cover, triggered by periods of below- or above-average rainfall. At low to intermediate grazing intensities, we noted emergent spatial patterning in the form of a step-like increase in vegetation density in the lower half of the domain. With a vegetation patch transitions governed by mechanistic water balance dynamics as well as grazing intensities that represent herder decision-making, the model holds great potential for further explorations of how land use, climate, and spatial heterogeneity affect the functioning of a dryland pastoralist social–ecological system.

Figure optionsDownload as PowerPoint slide