Allocation of CLC classes to groups.

The habitat suitability for pollinators is an important factor for biodiversity and crop-based ecosystem services. Most flowering plants, including wild plants, rely on pollination ecosystem services for fructification and reproduction. Suitable nesting sites and accessible floral food resources are crucial to the abundance of pollinator insects. Therefore, the suitability for pollinators and the pollination service itself are influenced by the type of land use, the composition of the land cover and structures in the landscape. One way to estimate pollinator habitat suitability is to use computer models such as ESTIMAP and InVEST. Both models calculate the habitat suitability based on spatial land use data and their suitability as nesting and feeding habitats. Besides the decision for a model, the selected spatial dataset also has important effects on the modelling results. In Germany, a large number of Land Use and Land Cover (LULC) datasets is available, such as the European CORINE Land Cover, CORINE Urban Atlas, Biotope types, ATKIS or Open Street Map. These datasets differ in terms of spatial and thematic resolutions, LULC types, and abundance of structural elements, which are crucial input factors for modelling with ESTIMAP and InVEST. We applied and tested both models on the basis of two different datasets in the study area Hannover region in central Germany. A literature-based estimation and expert-based questionnaire determined the biophysical properties required for modelling with ESTIMAP and InVEST. The differences between the results based on two different spatial datasets and the differences between the models were estimated and the results that can be obtained by using freely available data were investigated and compared with the results based on non-publicly available data. The comparison of the results shows that the proportion of near-natural structures in the landscape are a decisive factor for modelling results. The comparison of the models shows that ESTIMAP estimates a higher influence of small structures in the landscape than InVEST, resulting in a higher pollinator habitat suitability. The median similarity index of the two models is between 0.68 to 0.93 for the highly detailed biotope type dataset and 0.40 to 0.79 for the less detailed Corine dataset. The results provide a guidance on how to choose the appropriate model and data to assess pollination ecosystem services.