Data from: Climate change is predicted to impact the global distribution and richness of pines (genus Pinus) by 2070
Data files
Apr 08, 2024 version files 268.70 KB
Abstract
Aim: Climate change is altering habitat suitability for many organisms and modifying species ranges at a global scale. Here we explored the impact of climate change on 112 pine species (Pinus), fundamental elements of Northern terrestrial ecosystems.
Location: Global.
Methods: We applied a novel methodology for species distribution modelling that considers uncertainty in climatic projections and taxon sampling, and incorporates elements of species’ recent evolutionary history. We based our niche calculations on climate and soil data and computed projections across multiple algorithms and IPCC scenarios, which were ensembled into one single suitability map. We then used phylogenetic methods to account for recent evolution in climatic requirements by estimating the evolution of climatic niche. Edaphoclimatic and evolutionary analyses were then combined to calibrate the projections in areas showing high uncertainty. We validated our models using naturalized occurrences of invasive pine species.
Results: Our models predicted that by 2070 most pine species (58%) might face important reductions of habitat suitability, potentially leading to range losses and a decrease in species richness, particularly in some regions such as the Mediterranean Basin and South North America, albeit migration might mitigate these shifts in some cases. In contrast, our projections showed increased habitat suitability for approx. 20% of species, which may undergo range expansions under climate change. Moreover, the consideration of recent evolutionary trends modified projected scenarios, decreasing range loss and increasing range expansion for some species. The independent validation endorsed our models for many species and the influence of recent evolution in some cases.
Conclusions: We predict that climate change will impose drastic changes in pine distribution and diversity across biogeographical regions, but the magnitude and direction of change will vary significantly across regions and taxa. Species-level responses are likely to be influenced by regional conditions and the recent evolutionary history of each taxon.
README: Data from: Climate change is predicted to impact the global distribution and richness of pines (genus Pinus) by 2070
https://doi.org/10.5061/dryad.44j0zpcnh
This dataset contains two files: A ZIP file with code and a CSV file with evaluation metrics for the algorithms used to model the distribution of pine species.
Code
The ZIP file (scripts_pine_sdms_v2.zip) includes the code (i.e., scripts) used to perform the main analyses of the manuscript. The scripts are stored in different folders following the original order of the analyses. All scripts are annotated with explanations about the steps and main rationale, thus detailed information is provided in each script about each specific step.
- 1_presences_absences: Processing presence and absence data of pine species.
- 2_predictor_selection: Selection of environmental variables used to model the distribution of pine species.
- 3_modelling: Predicting the suitable areas (in terms of climate) for pine species under current and future conditions using Species Distribution Models (SDMs).
- 4_phylogenetic_analyses: Addition into SDMs of information about the past evolution of pines in order to improve the performance of these models.
- 5_global_results: Generation of global maps summarizing the predictions of all pine species. For example, the number of pines predicted to encounter suitable conditions in each region.
- 6_uncertainty_thresholds_comparison: Analysis of the impact of some modeling choices in the output of the models.
- 7_models_validation: Independent validation of the models in order to assess their performance when exposed to new, unseen data.
Evaluation metrics
The CSV file (means_evaluations.csv) contains evaluation metrics for the models of all species. For each pine species, we considered different algorithms with increasing complexity: Generalized Linear Model (GLM), Generalized Additive Model (GAM) and Random Forest (RF). Each model was evaluated using three metrics: Area Under the Curve (AUC), Kappa and True Skill Statistic (TSS). We ran the same model several times for each species, thus we calculated the mean and standard deviation (SD) for all values of each algorithm and metric combination. Therefore, the table has one row per species and two columns (mean and SD) for each combination of algorithm and evaluation metric.
The column labels of the table are the following:
- species: Specific name for each species.
- glm_kappa_mean: Mean of Kappa for GLM.
- glm_kappa_sd: Standard deviation of Kappa for GLM.
- glm_tss_mean: Mean of TSS for GLM.
- glm_tss_sd: Standard deviation of TSS for GLM.
- glm_auc_mean: Mean of AUC for GLM.
- glm_auc_sd: Standard deviation of AUC for GLM.
- gam_kappa_mean: Mean of Kappa for GAM.
- gam_kappa_sd: Standard deviation of Kappa for GAM.
- gam_tss_mean: Mean of TSS for GAM.
- gam_tss_sd: Standard deviation of TSS for GAM.
- gam_auc_mean: Mean of AUC for GAM.
- gam_auc_sd: Standard deviation of AUC for GAM.
- rf_kappa_mean: Mean of Kappa for RF.
- rf_kappa_sd: Standard deviation of Kappa for RF.
- rf_tss_mean: Mean of TSS for RF.
- rf_tss_sd: Standard deviation of TSS for RF.