Did we measure enough environmental variables? Insights from multiscale spatial analysis
| Authors | |
|---|---|
| Year of publication | 2013 |
| Type | Conference abstract |
| MU Faculty or unit | |
| Citation | |
| Description | Variation partitioning of species composition into components explained by environmental and spatial variables is often used to identify a signature of niche- and dispersal-based processes in community assembly. Such interpretation, however, strongly depends on the quality of available environmental data. Spatially structured variation not explained by environment (component [c]), which is believed to carry legacy of dispersal-based processes, contains unknown proportion of variation attributable to unmeasured environmental variables. We can never measure everything, but it is useful to know if we measured really too less or we are getting close to perfect. To evaluate this, we used multiscale spatial analysis of component [c], based on PCNM analysis. The prevalence of broad-scaled spatial patterns indicates lack of important environmental variables influencing species composition, while dominance of medium- and fine-scaled patterns is more likely to result from dispersal and other population processes. We demonstrate the method using vegetation data from permanent plot in subtropical broadleaf forest in Taiwan with detailed environmental information including topographical and soil variables. Results show that if only topography is included, [c] is represented by substantial amount of broad-scaled spatial variation, indicating that yet other environmental variables were not considered. Including soil variables into analysis, however, considerably increases importance of medium- and fine-scaled spatial patterns. Recent studies of forest permanent plots often use only topography to quantify environmental control imposed on vegetation; our results indicate that topography itself is poor surrogate for environment and that other variables, such as soil, need to be considered. |
| Related projects: |