Dataset DOI: 10.5061/dryad.02v6wwqkh

Data description

This dataset contains data and code associated with the study Identifying future climate refugia in central European mountain forests. Using the individual-based forest landscape model iLand, future forest composition in Berchtesgaden National Park (Germany) was simulated over 1,000 years to assess the persistence of mountain forest communities under climate change.

The repository contains summarized iLand simulation outputs for a historical-climate scenario and two future emissions scenarios (RCP 4.5 and RCP 8.5), outputs used to calculate a refugia capacity index, environmental and disturbance variables used to investigate drivers of refugia capacity, Boosted Regression Tree (BRT) inputs and outputs, figure data, and Quarto/R scripts used for the analyses.

Future-climate simulations use the EC-EARTH global climate model with four regional/downscaling realizations:

• CLM: CCLM4-8-17
• KNMI1: KNMI-RACMO22E (r1i1p1)
• KNMI12: KNMI-RACMO22E (r12i1p1)
• SMHI: SMHI-RCA4

Unless otherwise stated, spatial simulation units are 100 × 100 m grid cells. Species codes used in the iLand outputs consist of the first two letters of the genus and the first two letters of the species name (for example, lade = Larix decidua).

All necessary files can be found in RPr_Repository.zip

Files and variables

The data are organized in the following folders and analysis scripts.

0. Base data/

Contains pre-existing or pre-processed input datasets used in the analyses.

1. initial_species_assessment.csv

Summary of vegetation at the start of the simulations (2020) in Berchtesgaden National Park.

disturbance.csv

Disturbance data extracted from iLand outputs and averaged across simulations.

env_variables.csv

Environmental variables used during iLand landscape initialization and subsequently in the environmental-driver analyses.

pretreated_climate.csv

Pre-treated climate data used by the analysis scripts. The file contains 46 columns (V1–V46). Their interpretation is handled in the analysis code; users wishing to reuse this file independently should consult the relevant data-processing section of B. BRT.qmd.

species_param_europe.sqlite

SQLite database containing the European iLand species parameter data used by the analyses.

---

1.1 Summarized simulations/

Contains summarized iLand simulation outputs. The historical and RCP scenario databases contain repeated simulations as separate SQLite tables.

Initial.sqlite

Contains the vegetation state at the start of the iLand simulations.

HC_.sqlite

Historical-climate simulations. Contains 20 tables (rep_1 to rep_20), one for each simulation repetition.

Each table contains:

RCP 4.5 simulation databases

Each file contains 20 repeated simulations (rep_1 to rep_20) and has the same table structure as HC_.sqlite.

• RCP 45_CLM.sqlite — EC-EARTH downscaled with CCLM4-8-17.
• RCP 45_KNMI1.sqlite — EC-EARTH downscaled with KNMI-RACMO22E (r1i1p1).
• RCP 45_KNMI12.sqlite — EC-EARTH downscaled with KNMI-RACMO22E (r12i1p1).
• RCP 45_SMHI.sqlite — EC-EARTH downscaled with SMHI-RCA4.

RCP 8.5 simulation databases

Each file contains 20 repeated simulations (rep_1 to rep_20) and has the same table structure as HC_.sqlite.

• RCP 85_CLM.sqlite — EC-EARTH downscaled with CCLM4-8-17.
• RCP 85_KNMI1.sqlite — EC-EARTH downscaled with KNMI-RACMO22E (r1i1p1).
• RCP 85_KNMI12.sqlite — EC-EARTH downscaled with KNMI-RACMO22E (r12i1p1).
• RCP 85_SMHI.sqlite — EC-EARTH downscaled with SMHI-RCA4.

PNV simulation databases

Each file contains 10 repeated simulations (rep_1 to rep_10) and has the same table structure as HC_.sqlite.

• PNV1.sqlite — every 50 years output.
• PNV2.sqlite — outputs only around year 2000.

---

2. Refugia capacity/

Contains species weights, refugia-capacity values, scenario summaries, and raster outputs.

Main_sp_weight.csv

Species weights used in the refugia-capacity calculation.

Weights are based on the relative decrease in species range over the simulation period.

Refugia-capacity files for individual repetitions

The following files contain refugia-capacity values for each simulation unit and each of 20 repetitions:

• index_HC_.csv
• index_RCP 45_CLM.csv
• index_RCP 45_KNMI1.csv
• index_RCP 45_KNMI12.csv
• index_RCP 45_SMHI.csv
• index_RCP 85_CLM.csv
• index_RCP 85_KNMI1.csv
• index_RCP 85_KNMI12.csv
• index_RCP 85_SMHI.csv

All have the same structure:

Main\_index\_per\_scenario.csv

Mean refugia capacity across the 20 repetitions for each climate/model combination.

index_table.csv

Scenario-level summary of refugia capacity.

Raster files

• hc_refugia_potential.tif — refugia potential under historical climate.
• rcp45_refugia_potential.tif — refugia potential under RCP 4.5.
• rcp85_refugia_potential.tif — refugia potential under RCP 8.5.
• diff_rcp85_rcp45.tif — difference in refugia potential between RCP 8.5 and RCP 4.5.
• diff_rcp85_hc.tif — difference in refugia potential between RCP 8.5 and historical climate.
• diff_rcp45_hc.tif — difference in refugia potential between RCP 4.5 and historical climate.

---

3. Figures/

Refugia files used for Figure 3

• Refugia_HC.csv
• Refugia_RCP 45CLM.csv
• Refugia_RCP 45KNMI1.csv
• Refugia_RCP 45KNMI12.csv
• Refugia_RCP 45SMHI.csv
• Refugia_RCP 85CLM.csv
• Refugia_RCP 85KNMI1.csv
• Refugia_RCP 85KNMI12.csv
• Refugia_RCP 85SMHI.csv

All have the same structure:

---

4. BRT/

Contains datasets and outputs used for the Boosted Regression Tree analyses of environmental drivers of refugia capacity.

Main analysis tables

• BRT_env_variables.csv — environmental and forest-structure predictor table used for BRT analyses.
• env_variables.csv — environmental-variable table used during preparation of the BRT predictors.
• env_variables_climate.csv — climate summaries for historical, RCP 4.5, and RCP 8.5 scenarios.
• treated_climate.csv — processed climate table used by the BRT workflow.
• env_disturbance.csv — bark-beetle and wind-disturbance summaries by simulation unit and climate scenario.
• functional_div_mat_abundance.csv — species-abundance matrix used in the functional-diversity calculations.
• model_perf_.csv — BRT model-performance summary.
• variable_imp_HC.csv, variable_imp_RCP45.csv, variable_imp_RCP85.csv — relative variable importance for each climate scenario.

BRT_env_variables.csv contains the following columns:

Partial-dependence data

Files named \[variable]\_\[scenario].csv contain the data used for BRT response/partial-dependence plots. Scenarios are HC, RCP45, and RCP85. Variables represented include:

• bb — bark-beetle disturbance
• cos — cosine-transformed aspect
• dist_PNV — distance/difference relative to potential natural vegetation as used in the BRT workflow
• func_red — functional redundancy
• prec — precipitation
• silt_pct — silt percentage
• sin — sine-transformed aspect
• slope — slope
• soil_depth — soil depth
• stand_str — stand-structure variable
• TRI_max — terrain roughness metric
• wind — wind disturbance

Each partial-dependence CSV contains the predictor value and the corresponding fitted response (y).

BRT plots

The folder also contains exported variable-importance and partial-dependence plots (.jpg and .pdf), including plot_var_imp.pdf and plots for individual predictors.

---

Supplementary/

Contains data and code used to generate supplementary material.

• Supplementary.qmd — Quarto script for supplementary figures/tables.
• env_variables_climate.csv — climate-variable dataset used for supplementary analyses.
• var_in_time.csv and var_in_time_large.csv — intermediate data used by the supplementary workflow.

---

Code and software

Analysis scripts

Run scripts from the repository root so that relative file paths resolve correctly.

1. A. Refugia capacity.qmd
   Run first. This script calculates species weights and refugia capacity under historical climate, RCP 4.5, and RCP 8.5; creates scenario summaries and raster outputs; and prepares data used in the figures.
2. Rev_functional_diversity.qmd
   Calculates/assembles functional-diversity, functional-redundancy, species-richness, and stand-structure variables used in the environmental-driver analysis.
3. B. BRT.qmd
   Uses the refugia-capacity and environmental datasets to prepare predictors, fit Boosted Regression Tree models, assess model performance and variable importance, and generate partial-dependence outputs.
4. Supplementary/Supplementary.qmd
   Generates supplementary outputs from the prepared datasets.

Software

The analyses were performed with:

• R version 4.3
• RStudio (optional, recommended)
• Quarto version 1.4 for rendering .qmd files
• SQLite data accessed from R through DBI/RSQLite

Required R packages

Packages used by the supplied analysis scripts include:

• blockCV
• corrplot
• DBI
• dbplyr
• dismo
• dplyr
• gbm
• ggplot2
• grid
• gstat
• hillR
• patchwork
• raster
• RSQLite
• rmarkdown
• scales
• sf
• tidyr
• viridis
• viridisLite

They can be installed in R with:

install.packages(c(
  "blockCV", "corrplot", "DBI", "dbplyr", "dismo", "dplyr", "gbm",
  "ggplot2", "grid", "gstat", "hillR", "patchwork", "raster", "RSQLite",
  "rmarkdown", "scales", "sf", "tidyr", "viridis", "viridisLite"
))

Access information

The dataset is associated with:

Identifying future climate refugia in central European mountain forests

Dryad dataset DOI: 10.5061/dryad.02v6wwqkh

The repository is designed to be self-contained for interpretation of the archived analysis outputs. Users reproducing analyses should preserve the folder structure because the Quarto scripts use relative paths to access input and output files.