Extended Data Fig. 2: Summaries of GAMMs predicting Lepidoptera diversity changes through time in the tropical Americas and Asia.

a. Parametric coefficient estimates and smooth fits. Estimates are on the log-link scale. As in the models of order-level diversity change, the time scale adjustment controls for differences in the temporal scale of reported taxon occurrences between time series. Adjustment was unnecessary for non-Papilionoidea superfamilies (moths) in the Americas as all utilized datasets were of monthly temporal intervals. Also as before, the ONI smooth described short-term fluctuations in diversity attributable to climate oscillations. The Richness’s of Papilionoidea (butterflies) in the Americas and Asia and other Lepidoptera superfamilies in the Americas were predicted to be in significant long-term decline. Inverse Simpson’s diversity indices for both Papilionoidea and non-Papilionoidea Lepidoptera in the Americas were similarly predicted to be in significant long-term decline, but not the Papilionoidea in Asia. b. Visualizations of ONI smooth functions. Translucent bands represent 95% confidence intervals, derived from standard errors. The Richness’s of Papilionoidea (butterflies) and other Lepidoptera superfamilies (moths) in the Americas declined significantly during El Niño compared with La Niña, as did the Inverse Simpson’s index of non-Papilionoidea. There was no significant equivalent effect in Lepidoptera time series from Asia, although data from that region were lesser in temporal coverage and did not coincide with strong El Niño events.