Predicting the abundance of European stream macroinvertebrates using biological attributes.

Is there a relationship between the abundance of organisms and particular biological attributes? To assess this old, yet still acutely debated key question of ecology, we have used large databases on 312 stream macroinvertebrate genera (from 27 orders) that describe (1) invertebrate abundance at 527 least human-impacted European stream sites, (2) 11 biological traits (size, life-history, food, among others) described in 61 biological trait categories (BTCs; e.g. small, intermediate or large size) and (3) 14 attributes indicating specialization (AISs; e.g. species richness, size and food diversity). We applied interactive procedures to obtain models (for BTCs, AISs and a mixture of both descriptions) explaining as much as possible of the abundance variability of the genera with the lowest number of significant and ecologically meaningful attributes and assessed the predictive power of these models (in crosswise validations) by comparing predicted and observed abundances. Mean European invertebrate abundance increased with BTC affinities favouring viability in stream systems (e.g. attachment to the stream bottom to resist the flow, aquatic passive dispersal with the flow, exploitation of abundant food sources) and decreased with BTC affinities disfavouring this viability (e.g. drag force increase associated with larger body size, flow exposure associated with aerial respiration). Abundance consistently decreased with specialization of the genera (e.g. low species richness, oddity of their overall BTC profile from an "average" European genus). The model including a mixture of a few BTCs and AISs had the greatest predictive power: it predicted 35% of the observed abundance (ln-transformed) variability of the genera; these predictions were marginally affected by taxonomy (using orders as categorical variables). We conclude that a better appreciation of the influence of the examined taxonomic diversity, number and type of biological attributes, environmental system and spatial scale could enable abundance predictions using different sets of biological attributes for different taxonomic groups and systems.