A species distribution modelling framework for combining citizen science data from different monitoring schemes

Understanding the environmental drivers of species distributions is essential for developing effective conservation strategies for wildlife populations. Unfortunately, collecting species occurrences data on large-scale studies can be costly. Thus, citizen science (CS) data have become an important source of information for monitoring wildlife populations across broad spatiotemporal scales. CS offers a cost-effective solution for collecting large volumes of species distribution data for multiple taxonomic groups. However, the analysis of such data is challenging as of the most CS data recording schemes do not have a standardized sampling protocol. The lack of well-designed survey protocols can result in the sampling efforts to be biased towards locations that are easy to access, or where species are more likely to be found. The large quantity of data available from different CS recording schemes represents an important source of information that can be combined to improve the estimates and quality of the prediction of species distributions. Yet, reliably combining data sources can be challenging since they can vary considerably in their design, gradients covered, and sampling biases. Thus, in this work we present a general modelling framework using INLA/inlabru for integrating data from multiple sources while correcting for the sampling bias induced by preferential sampling. Our approach is based on defining a marked point process approach to model jointly the different recording schemes by using (i) the locations visited by the people who volunteer in these schemes and (ii) the presence records of a species of interest.