Fig. 1: Conceptual figures illustrating the time series analysis of estuarine macroinvertebrates and multiple global change drivers.

a Convergent cross-mapping, an equation-free, nonlinear time series analysis^{6,29,30,31,32} was used to assess the causal strength of multiple global change drivers (climate, freshwater, and sediment variables) on macroinvertebrates. b Time-varying sensitivity of macroinvertebrates to a given driver was quantified by a locally weighted state-space regression method^30,34,35,36. c Linear mixed-effects models were used to examine how biological traits determined the temporal mean and variability of species sensitivity to a given driver. The models were weighted by the causal strength of each driver on species abundance. In this figure, body size is assumed to be smallest in the small bivalve, intermediate in the worm, and largest in the large bivalve, whereas lifespan is assumed to be shorter in the worm and longer in both bivalves. We predicted that larger-bodied, longer-lived species would exhibit either more positive sensitivity or a reduced degree of negative sensitivity to intensified drivers, because they tend to be better competitors and more able to tolerate or avoid environmental stress. Accordingly, the large bivalve is expected to show positive temporal mean sensitivity, whereas the small bivalve may show negative sensitivity. In contrast, the worm, intermediate in size but short-lived, may exhibit both positive and negative sensitivity at different times, yielding a temporal mean close to zero. We further predicted that species with shorter lifespans would show greater temporal variability in sensitivity due to faster generation turnover. Thus, the worm is expected to show the largest variability (standard deviation) in sensitivity across time. Illustrations created in the Mind the Graph platform.