AbstractReliable estimates of population trends are important for prioritizing and implementing wildlife management actions. Abundance estimates are the most informative metric of population status but are difficult and costly to achieve for rare or elusive species. For bat species that are highly mobile and nocturnal, abundance estimates are usually only possible at their roosts. Although they are relatively easy to census in individual roosts, it is often unclear whether observed trends are representative of broader populations trends. Dual‐frame sampling is useful for monitoring wildlife populations that exhibit site fidelity and are highly visible but where the proportion of the population in known sites is unknown. We adapted a dual‐frame sampling approach to estimate abundance of Townsend's big‐eared bats (Corynorhinus townsendii) hibernating in a system of >800 caves at Lava Beds National Monument (LBNM) in northern California, USA, in 2013–2022. Beginning with a list of opportunistically discovered hibernacula, we expanded inference to the entire system by randomly selecting sites and stratifying survey effort according to number of bats observed during previous counts. Following 4 years of surveys, we began substituting survey‐informed estimates for some site types to improve efficiency of the monitoring effort. We estimated that a mean of 2,216 ± 112 (SE) bats hibernated at the study site annually and observed fluctuations of up to 22% around the mean population estimate and as much as 21% from the previous year. Despite annual fluctuations, we did not detect a trend in the number of hibernating bats at LBNM over the study period, underlining the challenges of detecting population trends in long‐lived species not exposed to pronounced stressors. Because our analytical approach generates reliable estimates of trend while accommodating annual variability in survey effort among years, it should be attractive to wildlife managers faced with shifting priorities and resources.