Climatic extremes and wildfire can disrupt soil communities, but the consistency of abundance responses across disturbance types remains unclear. We synthesized 425 abundance contrasts from 30 publications using sample size weights normalized within publications. Multilevel models accounted for clustering by publication. The overall mean reduction was 33.6%, with the clearest statistical support for wildfire. Hydroclimatic and storm estimates were uncertain, and thermal studies were summarized descriptively. Restricting the synthesis to 25 publications with suitable terrestrial exposure and reference comparisons yielded a 41.6% reduction across 322 observations. The average response remained negative under stricter event criteria and across alternative weights, taxonomic subsets and sampling methods. Direct tests detected no differences between body-size classes or between forests and grasslands within wildfire. Temporal analyses did not establish a common recovery pattern. Analyses using reported variances indicated high heterogeneity and a prediction interval spanning zero. Soil fauna abundance therefore declined on average, particularly following wildfire, but this response cannot be assumed across all disturbances or locations. Including soil fauna abundance in disturbance monitoring would improve assessment of belowground impacts that aboveground damage alone may not reveal.