Summary1. A spatially‐extensive data set of stream macroinvertebrate communities from 49 northern New Zealand sites sampled over a 10‐year period was analysed to assess relationships between the environment (catchment land‐cover, landscape position and regional‐scale weather patterns), and (i) community persistence and stability based on the constancy of species occurrence (presence–absence) and abundance (per cent composition), respectively and (ii) the temporal variability of various community condition metrics.2. No significant relationship was evident between community stability or persistence within sites and a gradient of increasing land‐use stress (LUS) based on types of upstream land‐cover, with interannual mean similarities at all sites falling within the standard deviations recorded at long‐term reference sites. In contrast, condition metrics were inversely related to LUS. Land‐use appeared to be a key factor influencing community composition among sites whereas climate influenced stability and persistence within sites.3. Three response trajectories of community variability to increasing LUS were distinguished based on the coefficient of variation of mean interannual similarities, such that (i) persistence and stability appeared to be more variable at developed sites with low LUS and at sites with high stress relative to the variability experienced naturally, but similar to the natural range at intermediate levels of stress (‘sinusoidal response’); (ii) variability in Ephemeroptera, Plecoptera and Trichoptera metrics increased at low‐to‐moderate stress and then accelerated rapidly at highly developed sites (‘stepped‐exponential response’) and (iii) variability in a pollution tolerance metric increased rapidly and then maintained a similar level of variability along the remaining stress gradient (‘plateau response’).4. The results of this study have implications for biomonitoring approaches that assume high temporal persistence and stability of communities to define site impairment. Misclassifications caused by interannual variability could lead to misinterpretation of site condition, if conclusions are based on single annual surveys. Temporal variability in stability and persistence may help distinguish low levels of impairment where a predominantly healthy fauna experiences increased environmental fluctuations.