Network analysis
To understand bacterial interactions and community structure (i.e.,
complexity) across sites, land cover types, and the four sampling times
(00:00, 06:00, 12:00, 18:00), we evaluated co-occurrence association
networks of bacterial ASVs. We constructed networks at the phylum
taxonomic level to examine broad complexity and the genus level to
visualise higher resolution complexity, recognising that the accuracy of
species-level associations would be low due to the high similarity
between the 16S rRNA gene from closely related species. In the evaluated
networks, vertices (also known as ‘nodes’) represent ASVs and edges
(also known as ‘links’) connect a pair of ASVs if their frequencies are
significantly associated (absolute abundance
>0.75, p = ≤0.01 for phylum and >0.95,p = ≤0.01 for genus). The type of association, whether positive
(representing a mutualistic interaction) or negative (representing an
antagonistic interaction), was denoted with blue and red edges,
respectively. To account for compositional bias associated with ASV
data, we used SparCC (Friedman & Alm, 2012) to define
associations. Only ASVs with sequence counts >10 were
included for phylum and >100 for genus to improve
visualisation, selection rigour and computational processing. Randomly
permuted (n = 100) data were used to estimate the statistical
significance of associations. We used the R package Matrix (Bates
et al., 2023) to create a matrix from the given set of values andigraph (Csárdi et al., 2023) to visualise and evaluate the plots.