Biases in data availability have serious consequences on scientific inferences that can be derived. The potential consequences of these biases could be more detrimental in the less-studied megadiverse regions, often characterized by high biodiversity and serious risks of human threats, as conservation and management actions could be misdirected. Here, focusing on 134 bat species in Mexico, we analyze spatial and taxonomic biases and their drivers in occurrence data; and identify priority areas for further data collection which are currently under-sampled or at future environmental risk. We collated a comprehensive database of 26,192 presence-only bat records in Mexico to characterize taxonomic and spatial biases and relate them to species’ characteristics (range size and foraging behavior). Next, we examined variables related to accessibility, species richness and security to explain the spatial patterns in occurrence records. Finally, we compared the spatial distributions of existing data and future threats to these species to highlight those regions that are likely to experience an increased level of threats but are currently under-surveyed. We found taxonomic biases, where species with wider geographical ranges and narrow-space foragers (species easily captured with traditional methods) had more occurrence data. There was a significant oversampling towards tropical regions, and the presence and number of records was positively associated with areas of high topographic heterogeneity, road density, urban and protected areas, and negatively associated with areas which were predicted to have future increases in temperature and precipitation. Sampling efforts for Mexican bats appear to have focused disproportionately on easily-captured species, tropical regions, areas of high species richness and security; leading to under-sampling in areas of high future threats. These biases could substantially influence the assessment of current status of, and future anthropogenic impacts on, this diverse species group in a tropical megadiverse country.

We generated a presence-only database for all bat species that occur in Mexico using information compiled from the National Commission for Knowledge and Use of Biodiversity (CONABIO), Global Biodiversity Information Facility (GBIF), and Mammal Networked Information System (MaNIS). We searched for published literature in English and Spanish using Web of Science, Google Scholar, and Scielo to obtain additional information on Mexican bat occurrence data using the search words “chiroptera”, “M*xico”, “bat”, “bats”, “record*”, “occurrence”, “registro*”, “quiropter*”, “murci*lago*”. Additionally, we requested unpublished material from Mexican researchers.

To determine range size, we counted the total number of grid cells in which the species is expected to occur using species range maps (IUCN, 2015). We used foraging space as a surrogate for sampling method. Three explanatory variables represent species richness: (i) bat species richness, that is, the number of all species that are expected to occur in each grid cell, estimated by overlaying the IUCN range maps of all the bat species distributed in Mexico (IUCN, 2015) and counting how many coincide in each cell, (ii) elevational heterogeneity was estimated as the difference in elevation within each grid cell  based on a digital elevation model at 60 m resolution (INEGI, 2017) and (iii) the coverage of federal protected areas estimated as the percentage of the protected areas polygons within each grid cell (CONANP, 2016). Other three variables represent the accessibility of each cell: (i) road density estimated as the sum within each grid cell of roads length in km (INEGI, 2015), (ii) percentage of urban areas estimated as the percentage of the city polygons within each grid cell (INEGI, 2014) and (iii) human population density by municipality and estimated by assigning the value or average of values given by the map polygons that coincide in each cell (CONABIO, 2010). The remaining variable measures the security level of each grid cell, using reports of the Centro Nacional de Información, which aggregates information on average rate of homicides between 1996-2000 per state (SEGOB, 2016). All variables had a correlation less than 0.6 or -0.6.

Explanation on the metadata is contained within the excel file