Influences of neighborhood diversity on the temporal dependence of patch dynamics in an urban wetland

Background/Question/Methods

Discrete Markov chain models are commonly used to simulate the patch dynamics of plant communities through time. In this modeling framework, the current state of a patch is assumed to be dependent only on the state of that patch in the immediately preceding time step. Using a time-series of aerial photographs, we tested the assumption of temporal dependence of the states of plant patches under a Markov framework in an urban wetland co-dominated by cattail (Typha spp.) and nonnative, invasive common reed (Phragmites australis) in Buffalo, New York. Further, to better understand the temporal dynamics of the states of the plant patches in relationship to the spatial configuration of their neighboring patches, we examined the influence of Moore neighborhood patch type diversity on the temporal dependence of patch states.

Results/Conclusions

At the landscape scale, wetland plant community patches were temporally autocorrelated which suggests that a Markov chain approach may be appropriate for modeling plant community dynamics at a broad spatial scale where the effects of spatial variation may be obscured. However, for individual patches, neighborhood diversity strongly influenced the temporal dependence of patch states such that as neighborhood diversity increased, temporal dependence declined exponentially. In the case of patches in the most diverse neighborhoods our analysis suggests that patches are temporally independent of one another. By influencing the likelihood and trajectory of patch transitions, neighborhood patch type diversity imposes limits of the utility of spatially explicit, deterministic models of plant community patch dynamics. For vegetation management and restoration ecology, diverse landscapes may be desirable for wildlife habitat, ecosystem services, or biodiversity conservation, but these findings suggest that spatially explicit predictions of patch dynamics in spatially heterogeneous landscapes will require embracing a probabilistic rather than deterministic perspective.