Exploring habitat‐density relationships and model transferability for an alpine bird using abundance models

Abstract Because resources for monitoring and conservation are often limited, a primary objective in applied ecological research is to predict key state variables in one context (the prediction context ) using models fitted to data collected in another context (the estimation context ). Model transferability is concerned with how well models predict across ecological contexts in time and space. While several previous studies have evaluated the transferability of species distribution models, much less is known about how well models that predict spatially explicit population density transfer across contexts. Although abundance and distribution are theoretically interconnected, environmental stochasticity, species interactions, functional responses in habitat selection and human management intervention might reduce the transferability of abundance models more than they affect transferability of distribution models. We examined how well models for predicting willow ptarmigan ( Lagopus lagopus ) densities based on habitat covariates transferred across time and space in central Norway. Utilizing line transect survey data from 11 study areas, we fitted Hierarchical Distance Sampling Models to estimate spatially explicit willow ptarmigan density, based on the underlying habitat in the focal area, and evaluated how well this model transferred to other years and areas. Considerable variations in the estimated habitat‐density relationships were observed across different temporal and spatial contexts. In general, model transferability was relatively low (mean r = 0.10) with some contexts showing negative correlations. Models with coarse scale covariates transferred somewhat better between different areas, and transferability was higher within areas between years than across different areas. Generally, model transferability was higher among geographically proximate areas. Practical implication . This study highlights the advantages of using local data and the challenges of predicting population density in novel areas for species with substantial inter‐annual population fluctuations based solely on habitat associations. In a management context, it is advisable to use current and local data to gain accurate insight into species abundance patterns. This is crucial for obtain reliable assessments of conservation efforts, and the sustainability of harvest regimes. If practical constraints make local data unavailable, it is recommended to use data from nearby areas and consider the scale of key variables.