Reviewing and benchmarking ecological modelling practices in the context of land use

Despite habitat loss and degradation are the primary drivers of biodiversity loss, different conclusions have been drawn about the importance of land‐use or land‐cover (LULC) change for biodiversity. Differences may be due to the difficulty of framing a coherent model design to assess LULC effects. Recommendations have previously been identified for the design of statistical models and failing to follow them can risk misidentification of drivers, misinterpretation of predictions, overconfidence, high uncertainty, and incorrect management recommendations. We review modelling practices in statistical models assessing biodiversity responses to LULC, and investigated relationships between modelling practices and citations by scientific articles and policy documents. We benchmarked practices across model approaches, political extents, and objectives. From 346 model applications, we found that more than half of the model applications have justified ecologically‐relevant predictors, have used 1 km² or lower LULC spatial resolution, have used fine LULC thematic resolutions, performed validation or communicated uncertainty. However, we found that the model approach and political extent were strong determinants of the misuse of modelling recommendations. Top–down models followed less frequently three recommendations out of six, compared to other model approaches. Global studies used coarser LULC thematic and spatial resolution than studies at other extents, and thus potentially underestimated the relationships between LULC and biodiversity. Global studies were however more frequently cited by both scientific studies and policy documents. Modelling recommendations are not universally applied, especially because of methodological tradeoff, technical difficulties in their applications and data requirements. However, the multiples risks associated with the misuse of modelling recommendations, particularly in large‐scale modelling exercises, raise concerns on model interpretation and policy support from science, regarding the impacts of LULC on biodiversity.