A novel method for targeting survey effort to identify new bat roosts using habitat suitability modelling

In the UK, four out of 18 bat species are listed on the EU Habitats Directive, including the lesser horseshoe bat ( Rhinolophus hipposideros ), and their population status is closely monitored by visiting known roosts. R . hipposideros predominantly form maternity roosts in buildings, but roosts are impermanent features in the landscape and their distribution changes as bats form new roosts and abandon others. Locating new roosts requires intensive surveys which are challenging and inefficient. In this study, we provide a novel model-based strategy to identify potential R . hipposideros maternity roost sites that can be used to monitor bat populations. First, we model potential maternity roost habitat using record centre data on roost locations across Wales, Great Britain. We then constrain the area identified from modelling using record centre data on locations of bats in areas with no known roosts. We used two variable selection methods and three pseudo-absence data sets (random background points, random points in buildings and target group selection of mammal records) to produce six habitat suitability models. The three pseudo-absence data sets produced different habitat suitability maps, demonstrating the influence of pseudo-absence selection on species distribution models. The six models were combined using weighted mean average to produce an ensemble model that performed better than individual models and that indicated high levels of congruence in areas predicted to have high habitat suitability for maternity roosts. Our model revealed an extensive area (6523 km 2 ; 31% of the area of Wales) containing 18,051 buildings in suitable habitat. Using record centre data on bat activity outside commuting range from known roosts reduced the potential survey area to 133 km 2 (0.6% of the area of Wales) and 207 buildings. Our modelling outputs can be used to direct volunteers and bat surveyors in more targeted and efficient searches.