On flat sandy beaches in Washington state, non-native seagrass (Zostera japonica) competes for space with native sand dollars (Dendraster excentricus). Because of the clonal nature of seagrass and the benthic, motile lifestyle of sand dollars, shifts between these two ecological states likely depend on patch dynamics and invasion at territory edges. Is long-term coexistence likely? Our previous experiments and non-spatial models predicted different outcomes depending on the state of the surrounding matrix, suggesting the importance of spatial pattern. How does this spatial component affect state shifts? I examined spatial dynamics at boundaries between sand dollars and seagrass on Orcas Island, Washington in plots with and without experimental disturbance. Fences prevented movement between treatments; but open ends allowed invasion. Presence/absence data were taken in a grid within plots in June and August 2014 & 2015. These data were used to calculate transition rates between states depending on nearest neighbor identity, then probabilistic spatial simulations were run with these rates over a realistic scale. We analyzed results of simulations for spatial autocorrelation using Ripley’s K and measures of mean patch sizes, and compared model distributions to those of original empirical data and aerial images.
Results/Conclusions
Our previous empirical work suggests that seagrass outcompetes sand dollars under low disturbance conditions, but sand dollars are quicker to colonize following disturbance. Analysis of aerial images using Ripley’s K indicates significant clustering above random. Although the current spatial cellular automata models (using information from 0, 4, or 8 neighbors) appear qualitatively spatially structured, the structure is evident at smaller spatial scales than are observed in the field (0.2-0.5 m in models, vs. 1-5 m mean patch diameter in field), and not statistically different from random. Initial spatial pattern for simulations did not alter the final distribution, but varying the number of neighbors used to estimate transition rates did, with greater seagrass dominance when more neighbor information is included. No model predicted extinction of either species, which suggests that coexistence is possible, and likely. However, the mismatch between simulation and observation suggest that more parameters, representative of additional biological aspects of the system, such as aggregating behavior in sand dollars and seasonal dynamics in seagrass, may produce better predictions with more closely matching patch scales.