Data and code for "Joint effects of climate change and urbanization accelerate plant biodiversity loss across urban green spaces"

Data and code for "Joint effects of climate change and urbanization accelerate plant biodiversity loss across urban green spaces" Authors: Kilian Perrelet, Andreas Dietzel, Lauren M. Cook, Fengge Liu, Joan Casanelles-Abella, Marco Moretti Corresponding author: Kilian Perrelet, Zuercherstrasse, Birmensdorf, Zürich, Switzerland, [e-mail] This repository contains the data and R code needed to reproduce the analyses and figures of the study on plant diversity in urban green spaces (UGS) of three Swiss cities (Geneva, Lugano and Zurich). The study combines (i) a field survey of 180 sites, (ii) species thermal niches derived from GBIF occurrences and CHELSA climate data, (iii) downscaled microclimate projections for 2041–2070 and (iv) projections of UGS cover under different urban growth scenarios, to estimate how many species are at risk of losing their thermal niche and how much richness could be lost. Data and code is also available on GitHub (https://github.com/KPerrelet/urban_climate_joint_impact). Workflow 1. Create an empty "figures/" folder and set the R working directory to the repository root.2. Run "01_extract_gbif.R" to regenerate "thermal_niche_per_species.csv" and "02_lulc_change.R" to regenerate "lulcc.RData". Both outputs are already in "02_processed/".3. Run the figure scripts ("fig2.R" – "fig5.R", "figS1.R" – "figS13.R") in any order. "fig5.R", "figS12.R" and "figS13.R" run a Monte Carlo simulation with accumulation curves and take longer than the others.4. To recreate the microclimate projection files in "01_raw/microclimate/", follow the instructions in "microclimate_downscaling.zip". 01_raw/ Raw data. Files in this folder are inputs and are never modified by the scripts (except for the GBIF download, which is saved here). Survey data - "sites.csv": One row per study site (180 sites = 3 cities × 5 UGS types × 12 sites). Gives the ID, city and UGS type of each site.- "native_mat.csv": Site × species matrix of native vascular plants (180 sites × 635 species). The first column ("ugs") is the site ID; all other columns are species. Values are the relative cover (%) of each species at the site. Native and non-native cover together sum to 100 % at each site. In all scripts a species is treated as present at a site when its value is > 0.- "nonnative_mat.csv": Same as "native_mat.csv", for non-native species (180 sites × 447 species). Sites without any non-native species contain only zeros. No species occurs in both matrices.- "growth_form.csv": Growth form of every recorded species (Herb, Shrub, Tree or Liana). GBIF occurrences - "gbif_occurrences_raw.rds": Cleaned GBIF occurrence records ("species", "decimalLongitude", "decimalLatitude", "countryCode", "year") of all recorded species, as downloaded by "01_extract_gbif.R". The file is provided so that the GBIF download does not need to be repeated. gis/ Spatial data. - "sites.shp" (+ ".dbf", ".prj", ".shx"): Polygons of the 180 study sites. Attributes include "ugs" (site ID, matches "sites.csv") and "city". Used to extract climate at site centroids.- "city_boundaries.*": Boundaries of the three cities (built-up area + 500 m buffer), used for the maps in fig2.- "lulc.gpkg": Land-use / land-cover data of the cities (Swiss Arealstatistik), reclassified into the five UGS types plus impervious surfaces, used for the land-cover maps in fig2.- "ch_languages.*": Language regions of Switzerland, used for the overview map in fig2.- "eu_boundaries.gpkg": Country boundaries of Europe, used for the overview map in fig2. microclimate/ Climate rasters (CHELSA V2.1 bioclimatic variables). - "CHELSA_bio01_1981-2010_V.2.1.tif", "CHELSA_bio05_1981-2010_V.2.1.tif": Global CHELSA rasters for the reference period 1981–2010 (bio01 = mean annual temperature, bio05 = warm-season temperature). Used only in "01_extract_gbif.R" to describe the climate at GBIF occurrence points.- "Geneva/", "Lugano/", "Zurich/": Downscaled (microclimate) rasters for each city, 38 files per city. Used in fig2, fig3, fig5 and figS6, figS12, figS13 to extract current and future climate at each site. File names follow the pattern "pred_value_corrected_<variable>_<period/scenario>_<model>.tif". | Element | Values ||---|---|| $variable | "bio01" (mean annual temperature, °C), "bio05" (warm-season temperature, °C) || $period/scenario | "hist_1981-2010_CHELSA_V2.1" (historical, no model suffix), or "ssp370" / "ssp585" followed by "2011-2040", "2041-2070" or "2071-2100" || $model | "GFDL-ESM4", "MPI-ESM1-2-HR", "UKESM1" (only for future scenarios) | For example: "pred_value_corrected_bio01_ssp585_2041-2070_UKESM1.tif". The analyses presented in the paper use the 1981–2010 and 2041–2070 periods. Recreating the microclimate projection files. The downscaled microclimate rasters in "01_raw/microclimate/Geneva|Lugano|Zurich/" are not recomputed by any script in this repository. To recreate them, follow the instructions provided in "microclimate_downscaling.zip", which is only available on Zenodo due to file size constraints. 02_processed/ Intermediate data generated by the scripts. They are provided so that the figures can be reproduced without repeating the (long) GBIF download and land-use modelling. - "thermal_niche_per_species.csv": Thermal niche of each species, from GBIF occurrences and global CHELSA climate. Generated by "01_extract_gbif.R". - "lulcc.RData": R objects with the modelled and projected UGS cover. Generated by "02_lulc_change.R". Contains "fitted.df", "observed.city.df", "growth.stats.df" and "baseline_year" (see data dictionary below).- "lulcc.csv": Plain-text export of "fitted.df" (first column is the row index). Scripts Scripts "01" and "02" generate the files in "02_processed/". The figure scripts then read from "01_raw/" and "02_processed/" and write to "figures/". Each figure script is self-contained. Data processing - "01_extract_gbif.R": Derives the thermal niche of every recorded species. 1. Matches the species names of "native_mat.csv" and "nonnative_mat.csv" to the GBIF taxonomic backbone (only EXACT/FUZZY matches with confidence ≥ 80 are kept). 2. Submits and retrieves a GBIF occurrence download (records with coordinates, no geospatial issues, year ≥ 1970, human/machine observations and living specimens, coordinate uncertainty ≤ 500 m). 3. Drops species with fewer than 500 records. 4. Extracts mean annual temperature (MAT, CHELSA bio01) and warm-season temperature (MTWQ, CHELSA bio05) at each occurrence point from the global CHELSA rasters. 5. For each species computes the mean, SD, 2.5 % and 97.5 % quantiles (niche limits) and breadth (upper − lower limit) of MAT and MTWQ. Output: "02_processed/thermal_niche_per_species.csv". - "02_lulc_change.R": Models how the cover of each UGS type changes with population size and projects it into the future. 1. Reads the Swiss land-use statistics (Arealstatistik, 100 m grid points, five survey periods 1979–85 to 2020–25), keeps points inside the three cities and reclassifies the land-use codes into the five UGS types (everything else counts as impervious; forest, crops and water are excluded). Computes the % cover of each UGS type per city and survey. 2. Attaches the cantonal population (historical statistics and Swiss Federal Statistical Office scenarios: AM-00 reference = "Current growth", BM-00 high = "High growth", CM-00 low = "Low growth"; can be downloaded from https://statistik.admin.ch/en/asset/36735051 and https://statistik.admin.ch/en/asset/36197557). 3. Fits one linear mixed model per UGS type, "pct_cover ~ total_pop + (1 | city)" (ML), and extracts slope, p-value and marginal R². 4. Predicts the % cover (± 1 SE of the fixed effects) for each city (including its random intercept) and for a "Consensus" curve (mean population of the three cities, no city effect), for historical and projected populations. Output: "02_processed/lulcc.RData". - "fig2.R": Study design and diversity patterns. - "fig3.R": Proportion of species at risk per site. - "fig4.R": Projected change in UGS cover with urban growth. - "fig5.R": Combined effect of urban growth and climate change on richness. - "figS1.R": Species accumulation curves. - "figS2.R": Species richness per site by UGS type, for native and non-native species. - "figS3.R": Share of herbs, lianas, shrubs and trees in the species richness of each site (% of the species with a known growth form), summarized as mean ± SD per UGS type and origin. - "figS4.R": LCBD per site by UGS type, for native and non-native species. - "figS5.R": Community composition of native and non-native species. - "figS6.R": Thermal status of the species of each city × UGS type under the historical climate and the future scenarios. - "figS7.R": Community-weighted mean MAT breadth and MTWQ breadth (± SD) of the species at each site, by UGS type and origin. - "figS8.R": Same as figS7 but split by growth form (Herb, Liana, Shrub, Tree), i.e. community mean MAT and MTWQ breadth per site × growth form, by UGS type and origin. - "figS9.R": Thermal niche of native vs. non-native species (species-level). - "figS10.R": Same as figS9 but by growth form: niche position and breadth per origin × growth form, with ANOVA + Tukey post-hoc letters. - "figS11.R": Same as fig4 but separately for each city: observed and fitted % cover of each UGS type per city (rows) with the Low / Current / High growth projections. - "figS12.R": Effect of urban growth alone on richness. Same simulation as fig5 but only the "Historical" species pool (no climate filter) is used, so that the loss of richness (shown as the proportion of persistent species, mean and 50 %/75 % intervals) is caused only by the change in the number of sites of each UGS type under the Low / Current / High growth scenarios. - "figS13.R": Change in the native / non-native composition of the remaining species pool. Same simulation as fig5; for each urban growth scenario, UGS type and emission scenario, shows the % of native vs. non-native species among the remaining species, and the relative change in the non-native share compared with the historical climate. Data dictionary "01_raw/sites.csv" | variable | description ||---|---|| "ugs" | Unique site ID (e.g., "GE_AlHH1"); matches the row names of the community matrices and the "ugs" field in "gis/sites.shp" || "city" | City: Geneva, Lugano or Zurich || "ugs_type" | Urban green space type: Allotment, Garden, Park, Real Estate or Ruderal | "01_raw/native_mat.csv", "01_raw/nonnative_mat.csv" | variable | description ||---|---|| "ugs" | Site ID || *species columns* | One column per species (binomial name, matching "thermal_niche_per_species.csv" and "growth_form.csv"); relative cover (%) of the species at the site, > 0 = present | "01_raw/growth_form.csv" | variable | description ||---|---|| "species" | Species name || "growth_form" | Herb, Shrub, Tree or Liana | "02_processed/thermal_niche_per_species.csv" | variable | description ||---|---|| "species" | Species name || "n_occ" | Number of GBIF occurrences used (after filtering and climate extraction) || "mat_mean" | Mean of the mean annual temperature (CHELSA bio01, °C) across occurrences || "mat_sd" | Standard deviation of MAT across occurrences || "mat_lo" | Lower niche limit: 2.5 % quantile of MAT || "mat_hi" | Upper niche limit: 97.5 % quantile of MAT || "mat_breadth" | MAT niche breadth ("mat_hi − mat_lo") || "mtwq_mean" | Mean of the warm-season temperature (CHELSA bio05, °C) across occurrences || "mtwq_sd" | Standard deviation of MTWQ across occurrences || "mtwq_lo" | Lower niche limit: 2.5 % quantile of MTWQ || "mtwq_hi" | Upper niche limit: 97.5 % quantile of MTWQ || "mtwq_breadth" | MTWQ niche breadth ("mtwq_hi − mtwq_lo") | A species is considered to persist at a site if the site's future MAT and MTWQ are both inside the species' "[lo, hi]" ranges; species without a niche are counted as unknown. "02_processed/lulcc.csv" and "fitted.df" in "lulcc.RData" | variable | description ||---|---|| "city" | Geneva, Lugano, Zurich, or "Consensus" (population-level prediction without city effect) || "year" | Year (1985–2024 historical, 2025–2055 projected) || "total_pop" | Population used for the prediction || "urban_scenario" | Historical, Low growth, Current growth or High growth || "ugs_type" | Urban green space type || "fit" | Predicted % cover of the UGS type (negative values set to 0) || "lwr", "upr" | Lower / upper bound of the prediction (± 1 SE of the fixed effects) | Other objects in "lulcc.RData" | object | description ||---|---|| "observed.city.df" | Observed data used to fit the models. Columns: "city", "ugs_type", "year" (last year of the survey period), "pct_cover" (observed % cover of the UGS type), "total_pop" || "growth.stats.df" | Summary of the models, one row per UGS type: "ugs_type", "slope" (change in % cover per unit of population), "p_slope", "r2_marginal" || "baseline_year" | Reference year for the present (2025) | The current land-use maps can be downloaded from https://statistik.admin.ch/de/asset/32376237 and https://statistik.admin.ch/en/asset/36824946.The current future population projections can be downloaded from https://statistik.admin.ch/en/asset/36735051 and https://statistik.admin.ch/en/asset/36197557.