Summary Seasonal droughts are predictable components of arid‐land stream hydrology, and many arid‐land aquatic taxa have adapted to their extreme environment. However, climate change is altering this predictable hydrology, producing longer and more severe droughts and creating novel disturbance regimes for resident organisms. The hydrologic transitions from flowing stream to fragmented pools to dry stream bed are frequently associated with steep decreases in taxonomic and functional diversity, referred to as thresholds of biodiversity loss. Less is known about how taxa respond between these thresholds, as fragmented pools gradually dry and abiotic conditions intensify. While an increasingly extreme environment may be expected to reduce taxonomic and trait richness, species adapted to predictable seasonal fragmentation may be resistant to declining water levels until all surface water is lost. We used aquatic mesocosms to test two competing hypotheses of the relationship between richness and pool drying for arid‐land stream invertebrates: (i) the drought vulnerability hypothesis (richness gradually decreases with drying) and (ii) the drought resistance hypothesis (richness remains constant until complete drying occurs). We inoculated replicate mesocosms with aquatic invertebrates from arid‐land streams in Arizona, U.S.A., and applied three drying treatments representing a continuum of drying stress commonly observed in local streams during the summer dry season (water depths: 10, 7 and 1 cm). Mesocosms were covered to restrict dispersal and colonisation processes and to isolate resistance (in situ survival of species) from resilience (community recovery following disturbance). After 45 days, we destructively sampled all invertebrates in the mesocosms and calculated various taxonomic and functional trait metrics. Taxonomic richness and composition did not differ between drying treatments, providing strong support for the drought resistance hypothesis. Severe drying was associated with lower invertebrate abundances and higher densities than the moderate and control treatments. This finding suggests that density‐dependent processes generated by decreased available habitat may be more important determinants of community composition during droughts than abiotic stress in this system. We observed a near‐complete overlap of trophic traits (body size and functional feeding group) and resistance traits (respiration mode and diapause) among the three treatments. This high functional redundancy may provide a buffer against changes to ecosystem functioning, even in cases of severe‐drying‐induced habitat contraction and fragmentation.

Top predator losses affect a wide array of ecological processes, and there is growing evidence that top predators are disproportionately vulnerable to environmental changes. Despite increasing recognition of the fundamental role that top predators play in structuring communities and ecosystems, it remains challenging to predict the consequences of predator extinctions in highly variable environments. Both biotic and abiotic drivers determine community structure, and manipulative experiments are necessary to disentangle the effects of predator loss from other co‐occurring environmental changes. To explore the consistency of top predator effects in ecological communities that experience high local environmental variability, we experimentally removed top predators from arid‐land stream pool mesocosms in southeastern Arizona, USA, and measured natural background environmental conditions. We inoculated mesocosms with aquatic invertebrates from local streams, removed the top predator Abedus herberti (Hemiptera: Belostomatidae) from half of the mesocosms as a treatment, and measured community divergence at the end of the summer dry season. We repeated the experiment in two consecutive years, which represented two very different biotic and abiotic environments. We found that some of the effects of top predator removal were consistent despite significant differences in environmental conditions, community composition, and colonist sources between years. As in other studies, top predator removal did not affect overall species richness or abundance in either year, and we observed inconsistent effects on community and trophic structure. However, top predator removal consistently affected large‐bodied species (those in the top 1% of the community body size distribution) in both years, increasing the abundance of mesopredators and decreasing the abundance of detritivores, even though the identity of these species varied between years. Our findings highlight the vulnerability of large taxa to top predator extirpations and suggest that the consistency of observed ecological patterns may be as important as their magnitude.