Data and code from: Historical temperature stability and environmental drivers shape patterns of phylogenetic diversity in cactaceae

Reliable biodiversity data are fundamental for assessing species distributions, detecting knowledge gaps, and informing conservation strategies. We present a comprehensive dataset containing 435,694 georeferenced occurrence records for 1,892 species of Cactaceae across the Americas. This large-scale compilation provides an overview of the spatial distribution of cactus species, revealing geographic and taxonomic biases in sampling efforts. The dataset offers a valuable resource for advancing ecological and biogeographic research and supports future analyses aimed at understanding the diversity, distribution, and conservation needs of this emblematic plant family. This database is an expansion of the database compiled by Amaral et al. (2022).

Methods
Our database expands on the foundational dataset compiled by Amaral et al. (2022), incorporating additional occurrence records of Cactaceae species sourced from publicly available digital biodiversity platforms. We retrieved data from three main repositories: (i) GBIF (Global Biodiversity Information Facility), (ii) iNaturalist (iNat), and (iii) iDigBio (Integrated Digitized Biocollections), using the “spocc” package in R version 4.4.1 (Chamberlain, Ram & Hart, 2021; R Core Team, 2024), with data indexed up to October 9, 2023. Taxonomic validation was performed using the Caryophyllales.org checklist (Korotkova et al., 2021), allowing the removal of misidentified taxa and entries with spelling inconsistencies. Records with spatial inaccuracies or vague locality information were excluded using the CoordinateCleaner package (Zizka et al., 2019). The final dataset comprises 435,694 georeferenced occurrences representing 1,892 Cactaceae species.

Phylogenetic metrics Mean Pairwise Distance (MPD) and Mean Nearest Taxon Distance (MNTD) were calculated from a phylogenetic matrix including all Cactaceae species with valid occurrence data. MPD and MNTD values were standardized using a null model approach, resulting in the metrics ses.MPD and ses.MNTD (standardized effect sizes), to control for species richness effects.

To assess the direct effects of abiotic variables on ses.MPD and ses.MNTD, we used structural equation models (SEM) implemented through the piecewiseSEM package (Lefcheck, 2016) in R version 4.4.1. Causal paths were defined based on an initial hypothetical model. Multicollinearity among predictors was evaluated using the Variance Inflation Factor (VIF), calculated with the car package (Fox & Weisberg, 2007), and variables with VIF ≥ 3 were excluded. Path significance was estimated using maximum likelihood, and model fit was assessed with Shipley’s d-separation test using Fisher’s C statistic (p > 0.05 indicating a good fit).

We also employed GWpath models (Geographically Weighted Path Models) (Barreto et al., 2019) using the spgwr package (Bivand & Yu, 2017) in R. This approach applies Geographically Weighted Regression (GWR) to estimate local models for each grid cell, based on a Gaussian kernel weighting function defined by distance (Fotheringham et al., 2009). The GWR model was built using the causal paths from the adjusted piecewiseSEM model. A bandwidth of 800 km was used. We then mapped the strongest environmental predictors of phylogenetic diversity for each grid cell, using categorical maps following the method of Barreto et al. (2019).