The effect of disturbance on community recovery is environment-dependent. For each time series, recovery was calculated as the pairwise dissimilarity between post-disturbance samples and pre-disturbance controls. Each point is a pairwise comparison, colored by study. Microbiomes which recover their pre-disturbance state will exhibit negative slopes; microbiomes which continue to drift away from their pre-disturbance composition over time will exhibit positive slopes. Thin regression lines for each time series are colored by study, and a solid black line indicates the modeled mean response across time series per environment. The 95% CI is displayed as a gray shaded area, and environments for which overall trends deviate from zero are indicated with an asterisk (*) on the bottom right corner
