Persistence of spatial structures in tropical tree populations is lower than expected based on aggregation, indicating time lags in underlying dynamics
Spatial patterns play a central role in determining ecological dynamics in most communities on a broad range of spatial and temporal scales. The availability of mapped plant data for large forest dynamic plots has stimulated a large body of spatial analyses. Multiple censuses of the same plot offer important additional insights into spatial processes; indeed, robust inferences about ecological processes require knowledge of how patterns change over time. One important property that can be investigated with multi-temporal data is how long spatial patterns remain coherent or persist before stochastic drift eventually dissipates them. In this study we investigate how aggregation and persistence are related to each other in theory (using spatially explicit models), and empirically in tropical plant populations (using dynamic plot data). Because patterns depend strongly on the spatial scale at which they are defined and observed, theoretical and empirical relationships are presented on a continuous scale axis, allowing for a complete generalization of the results. Understanding of species level dynamics often requires observations over large areas and long temporal intervals. The emergence of a general relationship between aggregation and persistence may enable inference of ecological properties even where spatial or temporal scales become limiting factors.
Using exact solutions of spatially explicit Lotka-Volterra models that incorporate dispersal limitation (DL) and negative density dependence (NDD), we first established theoretical relationships of aggregation and persistence to dispersal and NDD parameters. We found that, theoretically, persistence was proportional to aggregation at all scales for all combinations of dispersal and NDD. For a population dominated by DL, large structures become increasingly denser and more persistent. In contrast, for a population controlled by conspecific NDD, large structures become increasingly less variable and prone to quick dissipation. We then quantified aggregation and persistence for 221 species sampled in the Barro Colorado Island 50 ha forest dynamics plot in seven censuses spanning almost 30 yr. We found, in accordance with theory, that the most aggregated species are more spatially persistent. However the persistence level is significantly lower than expected for the aggregation level, and this is more pronounced for the highly clustered species, which are also the most abundant. This indicates that more complex dynamics involving multivariate interactions induce time lags in responses to aggregation. Such time lags could arise, for example, if it takes time for natural enemies to locate aggregations of their hosts.