Print PageEcologists often want to use spatial data that shows contrasting patterns from successive life history stages in order to make inferences about the spatial scale and pattern of the ecological processes driving the life history transition. When the data from the focal life history stages are not commensurate (e.g., measured in slightly different areas or via different sampling methods), this can be a challenging problem. We discuss how one can in general use (1) simple theoretical models of spatial processes and (2) standard spatial statistical techniques (spatial autocorrelation and spectral analysis) to reconstruct the spatial patterns of demographic transitions. We show an example using data on Pinus elliotti (slash pine) seed and seedling distributions, estimating the pattern of spatial variation in seed-to-seedling survival, as well as a range of simulated examples that provide a sanity check on our results and suggest the sample sizes required for reasonable power to detect spatial patterns in underlying processes.
Results/Conclusions
We carefully fitted a wide range of possible spatial correlation models to the seed and seedling data (a surprisingly difficult numerical computation) and find that (penalizing models for additional complexity) simple exponential correlation models fit the data best, possibly because the data set is noisy. The scales of seed and seedling patterns are similar (in the range of 5-10 m in each case), which suggests a fine-scale pattern of seed-to-seedling survival. However, our simulation analyses also suggest that these data are on the lower end of the range at which we can reliably detect patterns.
