Info

Figure 12.7. Geographic distribution of the posterior-predictive mean number of avian species present at each of 41,365 locations in Switzerland.

each sample location. In that setting it was natural to consider static features of a community, such as species richness, similarity in species composition among locations, etc. In this section we consider models appropriate for surveys that are undertaken to infer changes in a community which may occur as the duration of sampling increases. For example, sites which are occupied by a species at time

Figure 12.7. Geographic distribution of the posterior-predictive mean number of avian species present at each of 41,365 locations in Switzerland.

Figure 12.8. Estimated relationship between elevation and average probability of occurrence for each of the 134 species observed in the 2001 survey of Switzerland and for each of 3 individual species.

t may become unoccupied (i.e., experience local extinction) at some future time. Similarly, sites which are unoccupied by a species at time t may become occupied (i.e., locally colonized) by that species at some future time. In other words, we consider models for communities that are 'open' to changes in species composition that occur through demographic processes, such as local extinction and colonization.

Of course, the biological mechanisms that underlie these processes are many and varied. Examples that apply over annual time scales include differences in migratory or activity patterns among species. Over longer time scales, differences in responses of species to the introduction of exotic competitors or to major shifts in climate may induce changes in community composition. Whatever the cause(s), a statistical framework for the analysis of changes in community composition is necessary.

Almost all historical work on this subject has ignored the effects of observation error - that is, an observer's inability to distinguish absence of a species from its non-detection. We anticipate that the development of statistical models which account for observer error will become an expanding area of research. In addition, we envision at least 3 kinds of dynamic models - those based on temporal covariates of species occurrence, those which assume some form of conditional dependence in the site-specific occurrence of a species over time, and hybrids of these two. These approaches are briefly described in the following sections.

Was this article helpful?

0 0

Post a comment