Comparing the different iDNA samplers for biomonitoring
A PERMANOVA test and further visualization of the dissimilarity matrix using NMDS (k = 2; stress = 0.17) show that the gamma diversity did not differ significantly based on iDNA source (p-value = 0.23) (Fig. 5). Although there was no statistically significant separation in the gamma diversity found using the different iDNA methods, a site-by-site comparison of the alpha diversity found with each iDNA method does not support substantial overlap of the vertebrate communities revealed by each iDNA source at the site-level (Fig. 5). Although carrion flies revealed the greatest gamma diversity, when comparing species presence at a site across iDNA datasets, we found that there were many species absences in the carrion fly data at sites where sandflies or mosquito samplers recorded a species presence (Appendix S1: Fig. S4). Carrion fly data missed 181 of the 213 species occurrences from the sandfly data and 42 of the 63 total species occurrences from the mosquito data when examined at the site-level (Appendix S1: Fig. S4). Arrowed vectors show that the significant species in the ordination (p-value < 0.001) were also the most abundant in each iDNA dataset, and thus these species likely drive the separation in the vertebrate communities at sites as described by the different iDNA samplers (Fig. 5). These significant, non-human vertebrate species include cattle (highest RAI in carrion fly data), dog (highest RAI in mosquito data), nine-banded armadillo (highest RAI in sandfly data), capuchin, lesser anteater, and turkey (second highest RAI in mosquito data). This sampler bias was confirmed by our analysis of the most abundant species in each of the iDNA datasets which revealed the differing feeding associations of each insect group (Fig. 6) and we found that these abundant species were responsible for many of the site-level mismatches between the iDNA samplers (Appendix S1: Fig. S4). Additionally, we determined that humans were the primary feeding target of mosquitos (Fig. 6) and human DNA accounted for more than 80% of the total sequence reads in the mosquito iDNA dataset.