Input data: Background
To correct for geographical and temporal sampling biases of the
occurrence data, I created a weighted target-group background (Fig. 3b;
sensu Anderson 2003) using data from the Global Biodiversity Information
Facility (GBIF.org 2022). Target-group method uses species observed with
similar methods (i.e., Sherman and pitfall traps) as background,
assuming that points have similar sampling bias as the focal species
(Phillips and Dudík 2008, Phillips et al. 2009), and therefore
cancelling out any spatial and temporal bias in the occurrence data
(Reside et al. 2010, VanDerWal et al. 2013). To ensure similar sampling
as the occurrences, I used occurrence records of preserved specimens
from the same collections where the training occurrence data were
obtained. The Mexican collections used were: Colección Nacional de
Mamíferos, Instituto de Biología (CNMA); Museo de Zoología
’Alfonso L. Herrera’, Facultad de Ciencias (MZFC); and Colección
de Mamíferos, Universidad Autónoma Metropolitana Unidad Iztapalapa(UAMI). From these collections, I used records of Mexican small
non-volant mammals captured between January 1971 and December 2000 (same
date range as the occurrence data) from the orders Soricomorpha,
Rodentia (only families Cricetidae and Heteromyidae), and
Didelphimorphia (only genus Marmosa ) as background points.
Background points were removed if they were not located within a
one-degree point buffer around the training occurrences (between 102 to
106 km at the latitude of occurrences). Finally, I used the
CoordinateCleaner R package to identify and remove background points
mistakenly georeferenced in the ocean (Zizka et al. 2019).