Tree cavities offer protected shelters and resources for arboreal vertebrates worldwide. In general, cavities with larger openings are better accessible for predators and are avoided by smaller species for breeding, but can still be attractive for occasional use. The current study explores the diversity of functional use types and species interactions at the largest available tree cavities (entrance width ≥ 10 cm) in a temperate European forest with a low number of large cavity‐breeding species. Year‐round camera observations at 9 cavities (range 0.7–3.5 years) revealed 34 visiting species of birds and mammals, including non‐cavity‐breeding species. The top predator threatening other large‐cavity users was European pine marten (Martes martes), which regularly visited each cavity year‐round, on average every 0.7 months. Tawny owl (Strix aluco) was the only species successfully breeding in cavities, arguably because of its ability to defend the nests. However, other species visited cavities at an average rate of 1.5 visits per day, making predominantly short visits (less than 30 s) interpreted as exploration, searching for food, or inspecting for the presence of owls (mobbing). Making short visits and time segregation with predators was a behavioural strategy to exploit cavities for most species. These results confirm that, similarly to other keystone structures (large arboreal nests, ground burrows, etc.), large tree cavities attract a significant part of the arboreal vertebrate community and enrich their habitats. To sustain these functions in wooded ecosystems, management should provide a surplus of available cavities and diversity of their characteristics even when the apparent number of breeding species is low.

Data collection

Data were collected as images produced by remote cameras (camera traps) installed in front of tree cavity entrances. Cameras were set to still mode, making 3 captures per trigger with the delay time 5 seconds, trigger sensitivity was set to the highest.

The unit of observation was a visit by a vertebrate defined as an image or a series of images of one species obtained within 30 seconds before and after other visits. When animals used cavities for roosting, a ‘visit’ was one roosting event per day, irrespective of its length. If two or more species were observed within one event, they were treated as one visit for each species.

The information recorded for each visit was: animal species, number of specimens, time and date of observation, time spent, activity type. Activity types were classified into five categories: passing with no interest to cavity (excluded from analysis), approaching and looking inside, entering the cavity, roosting and breeding in cavities.

Data pre-processing

Relevant information (animal species, activity types, notes on behaviour etc.) was assigned to images as tags using DigiKam software (www.digikam.org). To get table with event data (observations.csv), image metadata with tag information were extracted from images and grouped into events with 30-sec lag using R package ‘camtrapR’.