Results
An extreme wind event in 2018/19 caused the homogenization of a key environmental variable while exposing Eurosta populations to other environmental differences, ultimately creating biotic heterogeneity and correspondingly variable selection. The wind event of 2018/19 knocked down 96% of all goldenrod stems (compared to 51% in the next most extreme year; P<0.001 pairwise differences between 2019 and any other years), reducing spatial variation in this key environmental factor (Fig. 1A; P<0.002 for all pairwise differences between 2019 and other years).
Consistent with a higher proportion of stems being knocked over, bird attack was less frequent for the 2018/19 generation (21% attack rate; P<0.001 for all 2019-other year contrasts, except P=0.16 for 2019 versus 2017). However, spatial variation in bird attack rate was actually greater in 2019 than other years (Fig. 2A; P<0.03 in all pairwise comparisons). This counterintuitive result occurred because when most stems were knocked down, the distance of a population from trees (another source of environmental variation) became an important determinant of bird attack (Fig. 2B). When most stems remain standing, bird attack rate does not vary among populations. However, when most stems are knocked down, bird attack is greater near the trees (Stems knocked down x distance to trees interaction: P=0.001).
No other enemy differed in mean or spatial variance among years (all P>0.2), and as expected attack rates of other enemies were unaffected by stems being knocked over or by the distance of the population from the tree line (all P>0.25). Patterns of mortality were consistent with the effects of year on the attack rates of particular enemies. Mortality was lowest in 2019 (Fig. 1A; 55% compared to an average of 72% in other years; P<0.003 in all pairwise combinations except 2019 versus 2017 where P=0.10). Conversely, spatial variation in mortality was most variable in 2019 (Fig. 1A; P<0.03 in all pairwise combinations except 2019-2018 where P=0.19).
These patterns of survival and biotic interactions shaped differences in selection. In terms of mean directional selection, 2019 was not an exceptional year. The mean β-value of +0.09 fell within the range of directional selection observed in other years (e.g. β=0.16 in 2017 and β=-0.22 in 2015), and only differed significant from 2015 (Fig. 3A; P=0.001, all other pairwise comparisons P>0.5). However, 2019 did exhibit exceptionally variable patterns of selection. Spatial variation in directional selection was greater in 2019 than in all other years (Fig. 3B; variance in 2019=0.31, average variance in other years=0.19, all P<0.03). The increased spatial variation in selection was partly driven by more variable bird attack (changing the shape of the fitness surface) and partly by more variable survival (Fig 3B; stretching and compressing the fitness surface; interaction term: P=0.002). By exposing gall-makers to alternative environmental variation, the windstorm of 2018/19 created biotic heterogeneity in attack and survival, and corresponding spatial variation in patterns of natural selection.
Heterogeneity in survival and selection created corresponding ecological and evolutionary consequences in the subsequent generation. Specifically, densities in the 2019/20 generation increased as a result of reduced mortality in the 2018/19 generation (all pairwise comparisons P<0.04), but densities tended to be more spatially variable (P<0.03 for all pairwise comparisons except 2019-2017 where P=0.15). Finally, gall sizes exhibited a likely evolutionary response to selection—galls tended to increase or decrease in size for positive versus negative directional selection coefficients respectively (Fig. 4; P=0.04). In short, an extreme event created biotic heterogeneity with ecological and evolutionary consequences.