Including linear landscape elements improves fine-scale species distribution models of bees and hoverflies

Abstract Context Species distribution models combine environmental and occurrence data to map and predict habitat suitability for a species. Predicting species distributions in agricultural regions can be challenging due to historical under-sampling in those areas and the presence of small landscape elements or non-productive land adjacent to the agricultural fields. Objectives Here we model bees and hoverflies in Dutch agricultural regions at a 10 m resolution to measure the improvement in habitat suitability predictions when landscape elements are included as model parameters. Methods We compared models excluding landscape elements with those incorporating them in two ways, as distance to the nearest element and as percentage local cover, and further assessed the latter against null models with randomized landscape variables. We evaluated the models using both withheld training data and independent field-collected data from three distinct agricultural regions. Results Our results demonstrate that integrating linear landscape elements into species distribution models significantly improves model performance by correcting the overprediction in agricultural fields and under prediction in field edges. Herbaceous landscape elements were especially important for modelling the distribution of pollinators, showing statistically significant model improvements for the withheld test data and independently collected data. Conclusions Integrating landscape elements into distribution models is becoming both increasingly relevant and more feasible. It is gaining relevance as European policies promote the expansion of semi-natural habitats within agricultural landscapes, while advances in the availability of high-resolution spatial data are making such integration more achievable. We recommend that modelers integrate these elements to more accurately predict biodiversity in agricultural habitats. The resulting maps provide a more robust estimate of habitat suitability, underpinning conservation and management strategies.