Differential abundance analysis
Separately within each of the four plots (2 sites x 2 land cover types),
globally differentially abundant ASVs across sampling times were
identified using the ANCOM-BC algorithm (Lin & Peddada, 2020) by
assessing log-fold-changes in ASV abundances at 06:00, 12:00, and 18:00
sampling times, compared to midnight (00:00) as the intercept/baseline
sampling time. We used the ancombc() function in the RANCOMBC package (https://github.com/FrederickHuangLin/ANCOMBC)
with settings including alpha = 0.05 significance level, p-value
adjustment for multiple comparison using the method of Holm (1979), taxa
with fractional prevalence less than 0.1 (= prv_cut) were excluded in
the analysis, structural zeros were detected (struc_zero = TRUE) with
taxa classified as structural zeros using asymptotic lower bounds
(neg_lb = TRUE), and a conservative variance estimator was used for the
test statistic (conserve = TRUE). We used the ANCOM-BC global
test result which identifies differentially abundant taxa between at
least two groups across three or more different groups. To determine
whether differentially abundant ASVs were shared across the plots, we
constructed Venn diagrams to display the number of overlapping or
non-overlapping differentially abundant ASVs across plots and comparison
periods (06:00 cf. 00:00, 12:00 cf. 00:00, and 18:00 cf. 00:00).