In an era of increasing anthropogenic disturbances, there is a pressing need to forecast how ecological variation affects restoration outcomes. Plant-plant interactions, including negative density-dependence (NDD), are an important source of variation that could alter plant performance depending on the spatial characteristics of plant neighborhoods. A local neighborhood can be defined in terms of the number, identity, and spatial proximity of neighbors to the focal plant.
We explored the effect of NDD on plant performance in a series of spatially-explicit simulations parametrized on a long-term sagebrush common garden experiment. The simulation workflow included three steps. First, we fit spatial Bayesian models to estimate demographic rates (i.e., growth and survival) and the magnitude of spatial interactions using field-data from a common garden experiment located in Idaho, USA. This common garden contained a set of big sagebrush plants (Artemisia tridentata) representative of high intraspecific variation, including three subspecies and two ploidy variations. Second, we used agent-based models (ABM) to simulate the performance of plants under incrementally increasing interspaces between neighbors. Finally, we quantified the magnitude of among subspecies:ploidy group differences in simulations with and without consideration of spatially-explicit plant locations. We applied linear analysis of variance to examine group differences in the ABM outputs to parallel frequently-used analyses of common garden data.
Results/Conclusions
We found that smaller interspaces lead to greater differences among groups due to density-dependence processes. Starting from a minimum distance of 0.5 m, the F-statistic, which measures how the spatially-explicit (full) model differed from the base model, diminished and eventually converged asymptotically to zero when the interspaces between plants reached 3 m. The correlation between simulation outcomes from the full and base models ranged from 0.92 to 1 when the distance became greater. From interspaces greater than 2 m, the standard error of the mean correlation between the base and full models overlapped 1, i.e., the effect of density-dependence processes was not discernible.
Overall, we demonstrate the utility of spatially-explicit agent-based models to guide planting designs for big sagebrush that account for plant-plant interactions. Model simulations showed that NDD can alter ecological inference when spatial interactions are not accounted for. Our study emphasizes the need to account for intrinsic sources of variation when quantifying demographic outcomes of both plant growth and survival in experimental designs and restoration field trials.