Long-term, large-scale studies of community succession provide critical baselines for understanding how ecosystems respond to perturbations such as disturbance, keystone species decline, and changing environmental conditions. Many experiments demonstrate that the outcome of succession is contingent upon the history of species arrival and loss. Although we recognize the importance of historical contingency, the extent to which contingent assembly contributes to predictive power requires further investigation. Our study of rocky intertidal succession asks (Q1) which functional groups distinguish community states over space and time, and (Q2) does knowledge of past community states lead to accurate predictions of future community structure? In a 15-year study at 13 field sites, we assessed community structure in 260 plots following initial clearing in 2006. Using a comparative experimental approach, plots were distributed across four capes in central Oregon, southern Oregon, and northern California. To address Q1, we performed cluster analysis and indicator species analysis to categorize plots into community states and to identify important functional groups. To answer Q2, we used Markov transition matrices and a dynamic Bayesian network (DBN) to model future community structure in individual plots. We assessed the accuracy of model predictions by comparing them to field observations at various forecast times.
Results/Conclusions
(Q1) Several community states, dominated by macrophyte or sessile invertebrate functional groups, emerged over space and time. Most experimental plots reached a “climax” community after five years, marked by a transition from early successional species (e.g., crustose algae and acorn barnacles) to late successional species (e.g., surfgrasses and mussels). In general, three field sites in central Oregon were dominated by sessile invertebrates throughout the study period, while the remaining ten sites were covered primarily by macrophytes, but with a recent increase in sessile invertebrate abundance. The growing populations of invertebrates at these sites may be due to the dramatic decline in Pisaster ochraceus predation as a consequence of sea star wasting disease. (Q2) The accuracy of community structure predictions improved with greater knowledge of past community states (i.e., more years of evidence) in our DBN. Prediction accuracy varied spatially and was related to the dominant functional groups at each site. For instance, plots characterized by mussels or surfgrasses were among the easiest to predict, in part because these community states are persistent with high transition matrix probabilities (>0.8). Our findings suggest that knowledge of historical contingency can be used to predict future community structure.