Assessment of species abundance from metabarcoding data
After selecting the spider families that were present in at least 10 communities, we built beta regression models with data for the remaining 14 families. Seven of the families provided appropriate models for the proportion of weight, and six families did so for the models relating proportion of reads and proportion of individuals (Fig. 5). In the families for which an adequate model could not be fitted, there was still a positive correlation between the two variables.
Although the relation between RRA and weight or abundance was positive in all models, its slope varied across spider families. Even in the three families where both models had a p > 0.05, the two curves were almost overlapping in two of the families (Lycosidae and Salticidae) but rather different in the third (Clubionidae). Also, in some families the observations were consistently above or below the 1:1 line. In the plots linking RRA and abundance of individuals, funnel-web (Agelenidae) and cob-weaving spiders (Theridiidae) were above the 1:1 line, while ghost spiders (Anyphaenidae) and ground-dwellers (Clubionidae) were mostly below. In the plots linking RRA and mass, Clubionidae ground-dwellers, Philodromidae crab spiders and Theridiidae cob-weavers were above the 1:1 line, while Agelenidae funnel-weavers, Araneidae orb-weavers, Gnaphosidae ground-dwellers and Lycosidae wolf spiders were mostly below.
DISCUSSION