To measure the level of replication needed to detect changes in arboreal arthropods, a data set based on the arthropod faunas of 26 replicate trees of the species Melaleuca linariifolia was analysed. Analysis showed that the number of species, individuals per tree and Chao 1 estimation of species per tree were not normally distributed and that this was best corrected by using log transformed data. It was found that 20 replicates were needed to detect a 23% change in the observed number of species per tree in the study population. The estimation of the ɑ diversity index and the Chao 1 estimation of the number of species per tree required even larger numbers of replicates for the same precision. Detecting changes in the evenness of diversity , measured as Simpson’s D or equitability E, also require d large numbers of replicates.

In practice, more replication than that suggested by this study will be needed as many relevant sources of increase in variance, for example time and location, were avoided and this almost certainly reduced the standard deviation between trees. It is clear that further exploration of the measurement of species level biological diversity is urgently needed. For the present, richness, evenness and the number of individuals which are independent of one another, should all be reported in biodiversity studies.