Spatial patterns of biodiversity in the Atlantic Forest of Brazil are well characterized. However, there is no consensus on the biological processes underlying these patterns, and multiple competing hypotheses have been proposed, several of which center on climatic stability. Here, we ask if Late Quaternary climatic stability predicts contemporary population structure and genomic-level diversity in two palm species: Syagrus botryophora and S. pseudococos (Arecaceae)

We first use species occurrence data to model the distribution of suitable environments in 62 time-slice climate projections over the last 120 thousand years, and summarize stability over that period. We then use >25,000 RADseq-generated SNPs to i) describe the spatial patterns of genomic variation in both species, ii) test how well genomic variation is explained by isolation by distance and by the environmental resistance imposed by historical instability (isolation by resistance) and iii) test for a correlation between genetic diversity and historical stability.

The contemporary range of S. botryophora has been relatively stable over the last 30 thousand years and there are two isolated regions of high stability for S. pseudococos. The genomic data recovers a clear pattern of isolation by distance in S. botryophora and two structured populations in S. pseudococos. Consequently, the contribution of isolation by resistance to overall genetic structure is much higher in S. pseudococos. Genetic diversity is not significantly correlated with historical stability in either species.

Based on the concordance between historical stability and genetic structure, Late Quaternary climate stability may have maintained population connectivity within S. botryophora and promoted intraspecific divergence in S. pseudococos. Conversely, historical stability does not seem to be driving spatial patterns of genetic diversity. This study supports the primary role of climatic stability in determining spatial population structure, but not genetic diversity, in the Atlantic Forest.

We sampled 11 localities of S. botryophora and 13 of S. pseudococos, including at least three individuals per locality across the entire range of each species. These data were supplimented with occurrences from the Global Biodiversity Information Facility (GBIF.org). To ensure taxonomic and locality accuracy, we included only those GBIF occurrences based on museum specimens.

To characterize climate stability for each species over time, we built SDMs by combining occurrence data for the species of interest with spatial environmental data to model how the species occupies geographic space (Guisan et al., 2017). We used the R package ‘ENMeval’ v2.0.0 to build a suite of models, varying the shape of the response curve (linear, or linear and quadratic) and model complexity (regularization multipliers ranging from 0.1 to 4.5 in increments of 0.1; (Muscarella et al., 2014). For the modeling procedure, we used the machine learning algorithm MaxEnt as implemented in the R package ‘dismo’ v1.1-4 (Hijmans et al., 2013), fitting a total of 90 models. To choose the best model from this suite of SDMs, we prioritized those with less model overfitting. First, we selected those models with the lowest 10% omission rates, within 0.05 of the minimum 10% omission rate. From this subset, we chose the most accurate model as measured by the average Area Under the Curve (AUC) of the testing sets (Guisan et al., 2017).

Please refer to ReadMe file.