PERSISTENCE OF COMMUNITY STRUCTURE - WHAT HAPPENS WHEN YOU CHANGE TAXONOMIC SCALE

It is common in community ecology to use the species as the taxonomic category of interest, yet in rich tropical assemblages containing guilds of very similar species this may not be appropriate. Such assemblages may be organized at the level of guilds rather than at the finer species level. In a ten-year study of assemblages of fish at One Tree Reef, Great Barrier Reef, we found species composition and the number of fish on a given lagoonal patch reef vary greatly across time (Sale and Douglas 1984; Sale et al. in preparation). The mean average proportional similarity of a reef's assemblage to itself at different times (censuses) is usually low at a value of around 0.5. This apparent variability may be ecologically irrelevant noise if organization is at the higher guild level. We have recast our database at the guild level to test this possibility. Thirteen guilds were defined by the diets, foraging habitats and times of the individual species comprising them. Similarity of an assemblage to itself at successive censuses was re-calculated using the number of individuals in each guild instead of the numbers in each species. This analysis yielded significantly higher levels of similarity (P<0.01) among censuses. To test whether this increase in similarity was due solely to the smaller number of categories used to calculate the similarity indices, 5 sets of randomly generated guilds were constructed using a Monte Carlo approach. No significant difference (P>0.05) was found between the average similarity among censuses when assemblages were classified into these "null" guilds and when they were classified according to the real guilds. These results indicate that shifting to the "larger" taxonomic scale of guilds does not reveal a significantly more persistent assemblage structure than that revealed when analysis is at the "smaller" scale of species. There is no evidence of an underlying organization of these assemblages at the guild level.