Loss of biodiversity from lower to upper trophic levels reduces overall productivity and stability of coastal ecosystems in our oceans, but rarely are these changes documented across both time and space. The characterisation of environmental DNA (eDNA) from sediment and seawater using metabarcoding offers a powerful molecular lens to observe marine biota and provides a series of ‘snapshots’ across a broad spectrum of eukaryotic organisms. Using these next-generation tools and downstream analytical innovations including machine learning sequence assignment algorithms and co-occurrence network analyses, we examined how anthropogenic pressures may have impacted marine biodiversity on subtropical coral reefs in Okinawa, Japan. Based on 18S ribosomal RNA, but not ITS2 sequence data due to inconsistent amplification across samples, as well as proxies for anthropogenic disturbance, we show that eukaryotic richness at the family level significantly increases with medium and high levels of disturbance. This change in richness coincides with compositional changes, a decrease in connectedness among taxa, an increase in fragmentation of taxon co-occurrence networks, and a shift in indicator taxa. Taken together, these findings demonstrate the ability of eDNA to act as a barometer of disturbance and provide an exemplar of how ecological functions, biotic networks, and coral reef resilience may be impacted by anthropogenic activities.

We here provide raw .fastq files for each sediment and seawater sample collected in this study. The naming convention for .fastq files are as follows: Illumina run, Primer Set, Substrate, Site, Sample ID (e.g. "OKIXX" or "AWFS_XX" or "SEDXX"), Forward Fusion Tag Number (e.g. "F123"), Reverse Fusion Tag Number (e.g. "R123"). Sample IDs "NCXX" indicates a negative extraction control. The year of sampling for each sample is indicated in "Supplementary Materials, Table S1, sequence read information, 18S and ITS2" available with the published paper.