Phylogenetic diversity and North Andean block conservation

Taxa and area

The study area was divided into 0.25° grid cells (Fig. 1). To ensure a representative sample of the high biodiversity in the NAB, we selected taxa from different taxonomic ranks (e.g., family, tribe, genus) across different groups (e.g., Vertebrates, Invertebrates, Plants) that had at least three genes available in GenBank for at least 50% of the known species for the taxon. We selected taxa based on the information available in GenBank. We primarily chose families if at least 70% of the species in the family had a minimum of three genes available in GenBank. We used the family as a “taxonomic group”; if not, we moved to the tribe or genus level. The detailed workflow of the taxon selection rules is shown in Fig. S1. To obtain distributional data for the selected taxa (see Table S1), we retrieved their occurrence data (Telenius, 2011; see Table S2). Invasive and domestic species were excluded from the distribution data to prevent bias introduced by human activities, which promotes the widespread distribution of these species.

To assess the reliability of the data distribution used in this study, we calculated the Half-Ignorance Index using the method proposed by Ruete (2015). The index ranges from 0 (no or minimal ignorance) to 1 (maximum or total ignorance). The O_0.5 parameter (species observation index) was set to a minimum of 10 occurrences to ensure that the presence of a species could be trusted, following the methodology of Chazdon et al. (1998). Using the resulting index values, we generated a raster image that displayed the level of information available for each cell in the study area.

Phylogenetic reconstruction

For each taxonomic group selected (e.g., family, tribe, and genus; see Table S2), we performed a phylogenetic analysis using the genes and species available for each taxonomic group. Once we obtained the genes from GenBank, we selected the best evolutionary nucleotide model under the Akaike Information Criterion (AIC) (Akaike, 1974) for each gene. We performed partitioned reconstruction under the Maximum Likelihood criterion, as implemented in RAxML v8.2.X (Stamatakis, 2014), using the nucleotide model GTR+GAMMA for all genes, followed by branch length optimization of the partitioned data in PhyML v3.0 (Guindon et al., 2010). For branch length optimization, we used the evolutionary nucleotide model calculated for each gene. Not a single tree in the dataset was dated or calibrated.

Index selection

We assessed the relationship between the phylogenetic diversity index (PD) and species richness in various areas. To achieve this, we employed two indices: PDi, which relies on a minimum spanning tree approach (Faith, 1992), and the average taxonomic distinctness index (AvTD), which is based on pairwise distances (Warwick & Clarke, 1998). We calculated PDi and AvTD for each cell in our study area to determine the index that showed the lowest correlation with species richness. Subsequently, we compared the values obtained from each index with the species richness data obtained from these cells. The aim of this analysis was to identify the index with the lowest correlation. For this correlation analysis, we implemented a Bayesian Simple Linear Regression (following Kruschke, 2014), as Bayesian analysis outperforms classical analysis (Wasserstein, Schirm & Lazar, 2019).

Prioritization based on phylogenetic diversity

Once an index was selected, the index value of each cell was used to prioritize the grid cells. The index values of the cells were divided and classified into five quantiles, with the highest index values in the upper quantile (Q5) and the lowest in the lower quantile (Q1).

We used a jackknife approach to evaluate the robustness of the phylogenetic data and the resulting prioritization. We randomly selected subsets of the data (i.e., phylogenies) and recalculated the index values. Finally, with the calculated index values, we classified the cells into five quantiles (Q1–Q5). This procedure was repeated 10 times for each subset. The selected subsets were 23 (24.7%), 46 (49.5%), 60 (74.2%), and 93 (100%) phylogenies, respectively. We used a binomial distribution to estimate the probability of each cell remaining in its initial quantile when all phylogenies were used. This analysis provided a measure of the stability of the prioritization scheme and its sensitivity to variations in the data.

Finally, we measured cell similarity by calculating phylogenetic beta diversity using the method described by Cardoso et al. (2014). Then, using the values of phylogenetic beta diversity, we classified the cells into three different clusters (C1, C2, and C3) based on their similarities using silhouette analysis (Rousseeuw, 1987).

Endemic species

We assessed the endemism of the species by calculating the mean distance between occurrences, and only species with a mean distance of less than 0.25° (~50 km, the distance between the two cells) were considered endemic. Thus, we considered an endemic species as one with a restricted or limited distribution range in the NAB. To understand the impact of endemism on area prioritization, we conducted a jackknife analysis with 100 replicates, randomly removing 25%, 50%, and 75% of endemic species. To assess the influence of endemic species on prioritization, we recalculated the AvTD values for each replicate. Additionally, we evaluated the effect of removing all endemic species from the AvTD.

Protected areas

To quantify the phylogenetic diversity within PAs (Fig. 1), we identified the grid cells that overlapped with the polygon of each PA to extract and sum their AvTD values. Likewise, we extracted and summed the AvTD values of cells that did not overlap with PAs. To define the polygons of the PAs, we utilized the World Database of Protected Areas (WDPA, UNEP-WCMC, 2022).

Data/script availability

All data and scripts used in this study and more information regarding the methods can be found at https://github.com/oleon12/PhyloDiversity. An interactive map of the results is available at https://rpubs.com/oleon12/PhyloDiv

Index selection

We used Bayesian Simple Linear Regression to determine that PDi values correlated highly with species richness, whereas AvTD presented a low correlation value. The mean slope value of the posterior distribution with PDi was 0.78, while AvTD showed a mean slope value of 0.39 (see Fig. S2). Thus, given the low slope value found with AvTD, we selected this index for the prioritization scheme and the other analyses in this study.

Taxa and data quality

We reconstructed 93 phylogenies at different taxonomic levels, encompassing 1,252 species found in the NAB (see Table S1). The smallest dataset (seven in total) comprised only three genes. In contrast, most taxa were represented by more than four genes, with the Felidae family having the largest dataset of 15 genes (see Table S2). Our results were highly robust because adding new phylogenies to the data had little effect on the outcome. With 46 phylogenies (approximately 50% of the total), the mean probability of correct classification was >50%, and with 64 phylogenies, the mean probability was >80% (Table S3).

The mean ignorance value in our analysis was low (mean = 0.26, SD = 0.27), indicating that most cells were well-sampled and had a good representation of the species present (Fig. 2). Only 38% of the cells had high ignition values (red cells in Fig. 2). The AvTD index implemented here considered the presence or absence of species, thus avoiding potential biases that could have arisen from differences in observation efforts. Approximately 44% of the species had 20 or more observations per cell, suggesting that our results were not significantly affected by differences in sampling efforts across the study area.

Prioritization based on phylogenetic diversity

Q5 cells, which represented the highest priority areas for conservation, accounted for approximately 90% of the phylogenetic diversity of the NAB and covered approximately 20% of the total area. Most of these cells (80%) were in the Andes of Colombia, Ecuador, and Venezuela, whereas the remaining 20% were in the lowlands, particularly the Pacific Ocean (Fig. 3A). In Colombia, Q5 cells were primarily located in the northern and central areas of the western Cordillera (including the Cauca Valley) and eastern Cordillera’s western and eastern flanks. The remaining cells were spread in the “Nudo de los Pastos,” Caribbean Valley, Sierra Nevada de Santa Marta, Pacific Lowlands, Amazonia, and Llanos Piedemonte (Fig. 3A). Similarly, in Ecuador, Q5 cells were mainly distributed in the Andean region, especially over the western flank of the western Cordillera and the eastern side of the eastern Cordillera. Other Q5 cells were spread over the Pacific and Amazonian Lowlands (Fig. 3A). Finally, Venezuela had a few Q5 cells in the Merida Cordillera, while the others were in the “Depresión del Maracaibo” and Lara-Falcón regions (Fig. 3A).

Silhouette analysis revealed three distinct clusters (C1, C2, and C3) that exhibited specific patterns (Fig. 3B). The first cluster (C1) corresponded principally to the Magdalena Valley, Depresión de Maracaibo, Lara-Falcón region, Llanos Piedemonte, southern Pacific Lowlands in Ecuador, and Caribbean Valley (Fig. 3B). The second cluster (C2) corresponded to the Pacific Lowlands, Amazonian Lowlands, and Piedemonte. In addition, some cells in the second group were distributed in the Caribbean Valley, Depresión de Maracaibo, and Lara-Falcón regions (Fig. 3B). The third group (C3) corresponded to the Andean Cordilleras of Colombia, Ecuador, and Venezuela, including Sierra Nevada de Santa Marta and Serrania del Perijá (Fig. 3B).

Endemic species

We identified 221 species endemic to the NAB. When we completely removed the endemic species from the analysis, the classification of the cells showed some changes, but the original category did not change significantly (Fig. 4). As anticipated, removing endemic species decreased the phylogenetic diversity value of specific cells but did not alter their classification, except for a few cells that changed from Q5–Q2, and vice versa (Fig. 4). Removing all 221 endemic species resulted in a marginal decrease of only 1.2% in the overall phylogenetic diversity of the NAB.

Protected areas

Our analysis of the protected areas in the NAB revealed some interesting patterns. The overall species richness and phylogenetic diversity values were higher outside the protected areas (64% and 67%, respectively). Although we found more species and higher phylogenetic diversity values outside protected areas, they still held a significant amount of the region’s total evolutionary information in a relatively small area (9.2% of the NAB’s area).