Jointly estimating spatial sampling effort and habitat suitability for multiple species from opportunistic presence?only data

Building reliable species distribution models (SDMs) from presence-only information requires a good understanding of the spatial variation in sampling effort. However, most cases, effort is unknown, leading to biases SDMs. This study proposes method jointly estimate parameters and densities avoid such biases. The particularly suited analysis massive but highly heterogeneous data. proposed based on estimating over units mesh parallel with environmental density multiple using marked Poisson process model. Based simulations realistic settings, we examined performance robustness parameter estimations. We also analysed large-scale citizen science dataset (Pl@ntNet), including around 300,000 occurrences 150 plant species. found that was correctly estimated when true constant within cells mesh. Estimation bias arose drivers strongly covaried cells. Otherwise, inference correct robust Running model real provided an map relative for 15% French territory. exotic invasive consistent prior first depending environment, as explicit function, occurrence data An asset few frequently observed greatly contribute effort, thereby improving estimation all other approach can thus provide SDM large opportunistic datasets, broad many species, datasets programmes.