Integrating new land cover classes into ecological models to predict their biodiversity impacts

•As the need for sustainable agroecosystems gains recognition, new land cover classes are increasingly emerging in temperate landscapes. Process-based ecological models are often the most suitable initial option for predicting the biodiversity outcomes of such novel systems, particularly when implementation and large-scale baseline data remain scarce. However, there are no accepted guidelines for integrating new land covers into these models. •Using UK silvoarable alley-cropping as a case study, we explore how to introduce this emerging land cover into the established process-based pollinator model, poll4pop. We demonstrate several parameterisation approaches, including proxy land covers, field data, expert opinion and Bayesian calibration. We also provide the first field-scale and seasonally-resolved evaluation of poll4pop, using pollinator abundance data collected at three UK silvoarable sites. •Our results show that models using proxy land cover parameters can capture spatial trends in observed bee abundance where suitable proxies exist, but that predictions are improved by integrating field-derived floral cover. Neither bespoke, expert-derived, land cover attractiveness scores nor Bayesian-calibrated scores improved our model fit, although they did reveal valuable insights into model parameter sensitivity. Overall, poll4pop effectively reproduced observed fine-scale spatial variation in bumblebee and spring-flying solitary bee foraging activity in silvoarable systems. However, seasonal differences between communities resulted in reduced model-predictive performance for summer-flying solitary bees. •We demonstrate that poll4pop is suitable for modelling fine-scale pollinator abundance in complex mixed-cropping systems. We also present a practical framework for integrating new land cover classes into process-based models which can guide future modelling of emerging land use systems.