On the importance of standardization by species in the exploratory analysis of community data

Background/Question/Methods In the exploratory analysis of community data, standardization should be an important consideration. Many researchers simply accept the defaults provided in software, which may not be optimal to achieve efficient summarization of large, complex multivariate data sets and elucidate community trends that are related to ecological factors. A survey of published articles over the past ten years reveals that most authors use either no prior standardization or standardize by sampling unit (SU). Previous research using simulated community data with known structure suggests that standardization by species increases the rank correlation between dissimilarity measures and ecological distance and improves the performance of ordination methods in recovering community patterns. Practitioners who have explicitly considered standardizing by species, usually decide not to do so, arguing that there is a risk of giving too much weight to low-abundance, infrequent, uninformative (LIU) species.

Results/Conclusions Using vegetation data from a variety of ecosystem types, I demonstrate that failure to standardize leads to results that are overwhelmingly dominated by those few species that attain high abundances, with the less abundant species being effectively ignored. I then use simulated community data with increasing proportions of LIU species to test the effects of standardization by species on ordination performance. The results clearly support the routine use of standardization by species maximum in community analysis. Even with data containing a majority of LIU species, this standardization results in a more powerful consensus among species, improving the recovery of community patterns.