COS 279-5 - Community descriptors are more sensitive to parasite species resolution than to host sampling completeness

Background/Question/Methods

To study the ecological and evolutionary processes structuring a host-parasite community we need a representative sample of the community. However, acquiring a representative sample of a host-parasite community is difficult as not all host and parasite species are equally abundant or equally easy to identify. Therefore, the two main sampling issues are the failure to capture the full range of host species (host sampling completeness) and the failure in identifying parasite species correctly (parasite species resolution). This study aims to understand how decreasing host sampling completeness and reduced accuracy of parasite species resolution affect commonly-used community descriptors. To answer this question, we built synthetic communities that mimicked empirical host-parasite communities. We used a quantitative niche model to generate weighted bipartite networks. This approach allowed us to create many replicates of the same synthetic community. We evaluated how the Rao index of diversity and four bipartite-network indices (modularity, nestedness, connectance and specialisation) varied along a decreasing gradient of host sampling completeness and parasite species resolution.

Results/Conclusions

The two sampling issues did not covary. The descriptors were more sensitive to parasite species resolution than to host sampling completeness. All descriptors were affected in communities with less than 70% of parasite species resolution. The descriptors differed in sensitivity to host sampling completeness: diversity, connectance and specialisation were robust; nestedness was reasonably robust (increased up to 1.2 times); and modularity was sensitive (underestimated by half of the true values). Only in severely under-sampled communities, nestedness did not significantly differ from nestedness of random assemblages. Therefore, these descriptors could still be useful to categorically describe communities with low sampling effort. The loss parasite species resolution made the confidence intervals wider. The gradient in parasite species resolution led the communities to vary in size. Then, the loss of parasite species resolution in small communities resulted in a greater change among them than among large communities. Finally, we recommend avoiding the study of under-sampled communities or those with low resolution in parasite species identification. We should choose the most robust descriptors if it is known beforehand that the sampling issues are unavoidable. To study community dynamics over time or space one should only compare community data acquired with an equivalent sampling effort.