Temporal changes in taxon abundances are positively correlated but poorly predicted at the global scale

Linking changes in taxon abundance to biotic and abiotic drivers over
space and time is critical for understanding biodiversity responses to
global change. Furthermore, deciphering temporal trends in relationships
among taxa, including correlated abundance changes (e.g. synchrony), can
facilitate predictions of future shifts. However, what drive these
correlated changes over large scales are complex and understudied,
impeding our ability to predict shifts in ecological communities. We use
two global datasets containing abundance time-series (BioTIME) and biotic
interactions (GloBI) to quantify correlations among yearly changes in the
abundance of pairs of geographically proximal taxa (genus pairs). We use a
hierarchical linear model and cross-validation to test the overall
magnitude, direction, and predictive accuracy of correlated abundance
changes among genera at the global scale. We then test how correlated
abundance changes are influenced by latitude, biotic interactions,
disturbance, and time-series length while accounting for differences among
studies and taxonomic categories. We find that abundance changes between
genus pairs are, on average, positively correlated over time, suggesting
synchrony at the global scale. Furthermore, we find that abundance changes
are more positively correlated with longer time-series, with known biotic
interactions, and in disturbed habitats. However, the magnitude of these
ecological drivers alone are relatively weak, with model predictive
accuracy increasing approximately two-fold with the inclusion of study
identity and taxonomic category. This suggests that while patterns in
abundance correlations are shaped by ecological drivers at the global
scale, these drivers have limited utility in forecasting changes in
abundances among unknown taxa or in the context of future global change.
Our study indicates that including taxonomy and known ecological drivers
can improve predictions of biodiversity loss over large spatial and
temporal scales, but also that idiosyncrasies of different studies
continue to weaken our ability to make global predictions.