NEIGHBORHOOD MODELS OF CLONAL GROWTH IN THE WHITE CLOVER TRIFOLIUM-REPENS

The importance of spatially local conditions in plants and the movement of clones over time are complications that have limited the development of population dynamic models for clonal plant species. We develop a field-calibrated, spatial model of white clover (Trifolium repens) population dynamics. We relate rates of stolen growth and dieback observed in the field to local clover density, and we use field data on the morphology of clover clones to develop a stochastic simulation of how clones spread through space. The size and shape of clover clones varied considerably in the field but still could be depicted accurately by our simulation model. Stolen growth in white clover increased with density at low clover densities and decreased with density at high clover densities. To our knowledge, this is the first demonstration of positive, density-dependent growth in a clonal plant species. Stolon dieback rates were density-independent. An unstable, lower ''equilibrium'' point, together with local (as opposed to global) density dependence, caused the formation and maintenance of patch structure in our model. Model predictions indicated that lawn populations should consist of a mosaic of clover and grass patches, the largest of which persist for substantial periods of time. Clover aggregations persisted in our simulations despite increased dispersal from areas of high population growth. Incorporation into the model of natural variation for clonal growth rates (at each local density of clover) had little effect on model predictions. We tested model predictions with 4 yr of data on the spatial and temporal distribution of white clover in a lawn population. We found positive and significant spatial autocorrelation at low distance classes (up to 192 cm), negative and significant spatial autocorrelation at intermediate distance classes (288-481 cm), and positive and significant spatial autocorrelation at high distance classes (577-673 cm). Temporal autocorrelation analyses indicated the location of clover and grass patches cycled over time. Thus, our field data revealed significant patch structure that corresponded to model predictions but did not support the prediction that patch structure is maintained for extended periods of time.