Dataset DOI: 10.5061/dryad.sbcc2frq3

Description of the data and file structure

This archive holds the data and R code for the article "Unravelling drivers of forest biodiversity: contrasting effects of mean environmental conditions, environmental heterogeneity and landscape context", Journal of Applied Ecology (manuscript JAPPL-2025-01296). It contains everything needed to reproduce:

• (a) data preparation: from the raw survey and inventory data to the community matrices and predictor tables (see "Limitations" for the parts that came from a GIS database and cannot be re-run from this archive);
• (b) the analysis: GAMs of species richness and Shannon diversity, GDMs of β diversity, deviance partitioning, multicollinearity diagnostics;
• (c) the figures and tables of the paper and its Supporting Information.

Study overview

Seven taxonomic groups were surveyed on 135 one-hectare forest plots of the ConFoBi project (Conservation of Forest Biodiversity in Multiple-use Landscapes of Central Europe) in the Black Forest, Baden-Württemberg, Germany:

For each plot we compiled environmental predictors in three groups: local mean conditions, local heterogeneity and landscape context. Landscape context was calculated in buffers of 1, 5, 10, 15 and 20 km². We modelled species richness and Shannon diversity (exp H) with generalised additive models (GAMs) and β diversity (Bray–Curtis) with generalised dissimilarity models (GDMs). For each taxon and response, the landscape scale was chosen by AICc (GAMs) or by deviance explained (GDMs). Full sampling protocols are given in the paper (Methods, Biodiversity data).

Workflow and folder structure

The scripts use paths relative to the archive root. To run them, recreate this folder structure (the raw data are supplied as Raw_data.zip, which unpacks to the Raw_data/ folder below):

.
├── Code/                         R scripts (see "Code/software")
├── Raw_data/                     (a) raw input data, unpacked from Raw_data.zip
│   ├── Biodiversity_data/<subproject folder>/...
│   └── Environmental_data/...
├── Processed_data/
│   ├── Predictors/               local_predictors.csv, landscape_predictors.csv (supplied) + files written by (a)
│   ├── Community_matrices/       written by (a)
│   └── Models/                   written by (b)
├── Analysis_data/                (b) analysis-ready tables: community_*.csv, predictors_*.csv (supplied)
└── Figures/                      written by (c)

Steps b and c can be run directly from the supplied files without repeating step a.

General conventions

• All processed and analysis-ready files are comma-separated text files (.csv, ASCII, decimal point) with one header row.
• Raw files are kept in the original format their data owners delivered, because the preparation scripts read them as they are. Several of them are semicolon-separated: Updated_Plant_Data_ConFoBi_JH_MW_100322.csv, Knuff_ModifiedWindowTrap_CapturedInsectTaxa.csv, 20230109_FinalClassificationActivity_withoutBadWeather_AllNights_toGita.csv, B6_birds_Joao.csv, Allcameratrapdata_Sebastian.csv and lyingdeadwoodvolumes.csv. Two forest inventory files and the bat and camera-trap metadata are .xlsx workbooks. Do not rename the raw files: the scripts address them by name.
• plot_id: integer plot identifier (values between 1 and 188; 135 plots; numbers are not consecutive). It is shared by all processed and analysis files and links community and predictor data. Raw files use the string form CFB001–CFB188, which the scripts convert to integers.
• Missing values: the processed and analysis-ready files contain no missing values. A plot absent from a community file was not sampled for that group (see the row counts below). A value of 0 in a community file means the taxon was not recorded on that plot. In the raw files, missing values are empty cells or NA.
• Coordinates (x, y in local_predictors.csv; plot centres): Gauss–Krüger zone 3 (DHDN, EPSG:31467), metres.

Files and variables

1. Analysis-ready files (folder Analysis_data/; inputs to steps b and c)

Community data (7 files)

community_plants_FIXED.csv, community_fungi.csv, community_beetles.csv, community_hymenoptera.csv, community_birds.csv, community_bats.csv, community_mammals.csv

Each file is a plot × taxon matrix with one row per plot. Apart from the identifier and effort columns below, every column is one taxon (species, family or acoustic group, depending on the group). The column header is the taxon name or code, and the cell value is the summed abundance or activity of that taxon on that plot, in the unit given in the table.

• sampling_effort (bats, beetles, birds and mammals only; first column): sampling effort on that plot, used as a smooth covariate in the GAMs and as a predictor in the GDMs to account for unequal effort among plots (units in the table).
• plot_id: plot identifier (first column in fungi, hymenoptera and plants; second column in bats, beetles, birds and mammals).

Bat acoustic groups (columns of community_bats.csv; BatScope classifications grouped as in 20230901_Species_BClass_Guild_Overview_ALH.xlsx). Calls that could not be assigned to one of these groups (e.g. Eptesicus/Nyctalus/Vespertilio mixed calls, unassigned Pipistrellus calls, "Chiroptera") are not included.

Mammal species (columns of community_mammals.csv): Badger Meles meles; Fox Vulpes vulpes; Chamois Rupicapra rupicapra; Hare Lepus europaeus; Polecat Mustela putorius; Lynx Lynx lynx; Marten Martes martes and M. foina (not distinguishable on camera images); Roe_deer Capreolus capreolus; Red_deer Cervus elaphus; Sika_deer Cervus nippon; Raccoon Procyon lotor; Weasel Mustela sp; Wild_cat Felis silvestris; Wild_boar Sus scrofa. Where two cameras operated on the same plot in the same season, one was randomly retained (set.seed(123) in B8_mammals_preparation.R).

File: predictors_local_mean.csv

Local mean environmental conditions per plot. 135 rows, 9 columns.

• plot_id: plot identifier
• ndsm_mean: mean canopy height from the UAV-derived normalised digital surface model (m)
• gap_share: proportion of the plot area in canopy gaps (0–1)
• log_dw_vol_all: natural logarithm of total deadwood volume (lying plus standing), in log(m³ ha ^-1). Plots are 1 ha, so the untransformed value equals the volume per plot (range 4.9–486 m³).
• avg_altitude: mean elevation (m a.s.l.)
• cf_share: proportion of coniferous trees in plot basal area (0–1)
• northness: cosine of aspect (−1 = south-facing, 1 = north-facing)
• eastness: sine of aspect (−1 = west-facing, 1 = east-facing)
• avg_slope: mean slope (degrees)

cf_share, gap_share and avg_slope can be interpreted either as mean or as heterogeneity variables. They are in this file by default and are reassigned to the heterogeneity group in the partitioning sensitivity analysis (03_run_all_models.R).

File: predictors_local_het.csv

Local environmental heterogeneity per plot. 135 rows, 5 columns.

• plot_id: plot identifier
• log_ssci: natural logarithm of the stand structural complexity index (SSCI, dimensionless; Ehbrecht et al. 2017), from terrestrial laser scans, averaged over the scans per plot
• ndsm_sd: standard deviation of canopy height within the plot (m)
• tree_shannon: Shannon diversity index (H, natural log) of tree species in the forest inventory, based on stem counts (dimensionless)
• dw_div: deadwood diversity, the Shannon index of deadwood volume across deadwood types (combinations of category, origin, wood type, decay stage and diameter class; dimensionless), calculated with 01c_calculate_deadwood_diversity.R

Files: predictors_landscape_1.csv, predictors_landscape_5.csv, predictors_landscape_10.csv, predictors_landscape_15.csv, predictors_landscape_20.csv

Landscape context around each plot, calculated within a circular buffer centred on the plot. The number k in the file name is the buffer area in km² (1, 5, 10, 15 or 20 km²; radii 564, 1262, 1784, 2185 and 2523 m). Each file has 135 rows and 3 columns.

• plot_id: plot identifier
• fc<k> (e.g. fc5): proportion of forest cover within the buffer (0–1), from the Copernicus High Resolution Layer Forest Type 2018 (10 m)
• landscape_cf_share<k> (e.g. landscape_cf_share5): proportion of coniferous forest in the forested area of the buffer (fcc<k> / fc<k> in landscape_predictors.csv; 0–1)

File: Processed_data/Predictors/local_predictors.csv

The full table of local plot variables written by 01a_data_curation_predictors.R. It is supplied because (i) step a1 cannot be re-run from this archive, (ii) it is the input to steps a3 and a5, and (iii) 03_run_all_models.R and 04_Model_checking_plots.R take the plot coordinates (x, y) from it for the GDMs and the spatial-autocorrelation tests. 135 rows, 27 columns, sorted by plot_id.

• plot_id: plot identifier
• avg_altitude: mean elevation (m a.s.l.)
• avg_slope: mean slope (degrees)
• aspect: mean aspect (degrees clockwise from north)
• x, y: plot-centre coordinates (m; see General conventions)
• gap_share: proportion of the plot area in canopy gaps (0–1; UAV canopy model)
• ndsm_mean, ndsm_sd: mean and standard deviation of canopy height (m; UAV nDSM)
• tri: mean terrain ruggedness index of the UAV canopy surface model
• gap_share_sd: standard deviation of the gap share among 200 × 200-pixel tiles of the plot (0–1)
• ssci: stand structural complexity index (dimensionless; TLS)
• enl: effective number of canopy layers (dimensionless; TLS)
• dbh, dbh_sd: mean and standard deviation of tree diameter at breast height (cm; trees > 7 cm)
• ba: basal area (m² per plot = m² ha ^-1)
• tree_sr: number of tree species
• tree_shannon: Shannon diversity of tree species (dimensionless)
• cf_share: coniferous share of plot basal area (0–1)
• dw_vol_lying, dw_vol_standing: volume of lying and standing deadwood (m³ ha ^-1)
• dw_cf_share: coniferous share of standing deadwood, by summed DBH (0–1; not used in the analyses)
• temp: mean air temperature from plot data loggers, mean of monthly means (°C)
• precipitation: mean annual precipitation at the plot centre (mm) 
• trems_richness, trems_abundance, trems_shannon: richness, total number and Shannon diversity of tree-related microhabitats (not used in the final models)

File: Processed_data/Predictors/landscape_predictors.csv

The full table of landscape variables written by 01a_data_curation_predictors.R (input to step a5). It is supplied because it was calculated from raster data in the ConFoBi GIS database. 135 rows, 101 columns: plot_id plus 20 variables for each of the five buffer areas. The suffix k = 1, 5, 10, 15, 20 gives the buffer area in km².

• fc<k>, fcb<k>, fcc<k>: share of the buffer classified as forest, broadleaved forest and coniferous forest (0–1; Copernicus Forest Type 2018, 10 m)
• ndvi<k>, ndvi_sd<k>: mean and SD of Sentinel-2 NDVI (dimensionless)
• precipitation_l<k>, precipitation_l_sd<k>: mean and SD of annual precipitation (mm)
• terrain_height<k>, terrain_height_sd<k>, terrain_min<k>, terrain_max<k>: mean, SD, minimum and maximum elevation from a 10 m digital terrain model (m)
• tri<k>: mean terrain ruggedness index (m)
• residential_share<k>, farmland_share<k>, orchard_share<k>, greenland_share<k>, windthrow_share<k>, waterbodies_share<k>: share of LUBW land-cover classes (settlement and industry; farmland; vineyards and orchards; grassland; windthrow; water bodies; 0–1)
• lcc_n_types<k>: number of these land-cover classes present
• lcc_shannon<k>: Shannon diversity of all LUBW land-cover classes (dimensionless)

2. Raw data (Raw_data.zip; inputs to step a)

Raw_data.zip unpacks to the folder Raw_data/ with the sub-folders and file names expected by the scripts. Each sub-folder name gives the ConFoBi sub-project and the data owner. Variables that the scripts do not use are kept as delivered.

Raw_data/Biodiversity_data/B2_plants_Jan_Helbach/

• Updated_Plant_Data_ConFoBi_JH_MW_100322.csv (semicolon-separated; long format): plotID (CFB plot code), variable (species, Genus.species), value (mean cover across subplots, %), blanquet (cover class). The first, unnamed column is a row index.

Raw_data/Biodiversity_data/B3_beetle_families_Anna_Nolan/

• Coleoptera_families_sample_period.csv: one row per trap sample. sample_id, plotid, trapid, collection_unit (top/bottom), collect_start, collect_end (dd.mm.yyyy), sampling_period (1–5), unknown_sum (unidentified beetles), then one column per beetle family (<family>_sum, number of individuals). Some specimens sent for DNA analysis are not included.

Raw_data/Biodiversity_data/B3_insect_orders_Anna_Knuff/

• Knuff_ModifiedWindowTrap_CapturedInsectTaxa.csv (semicolon-separated): arthropod orders per window-trap sample (Knuff et al. 2019). Used only to add trap samples that contained no beetles (zero rows) to the beetle data, so that sampling effort is complete. Sample_ID; one column per order with the number of individuals (_Larven = larvae); PlotID, TrapID, Sampling_Unit, Collect_start, Collect_end, Sorted_by, Period.

Raw_data/Biodiversity_data/B3_hymenoptera_Nolan_Rappa/

• SP_builder_species_matrix.csv: trap-nest corner × species matrix of nest builders. Plot, Corner (NW/SE), plot_corner, total_abundance_nests, then one column per species (number of provisioned brood cells, pooled over the two trap nests at a corner).
• SP_parasitoid_species_matrix.csv: as above for parasitoids and kleptoparasites. Plot, Corner, Plot_corner, then one column per taxon.

Raw_data/Biodiversity_data/B5_bats_Marlotte_Jonker/

• 20230109_FinalClassificationActivity_withoutBadWeather_AllNights_toGita.csv (semicolon-separated; one row per logger-night): plot_id, subplot_id (logger), date, X, Y (logger coordinates), the 13 acoustic-group columns (one-minute intervals with calls), and the weather covariates min_temp, mean_temp (°C), mean_rain_h(hours with rain), mean_rain__mm, mean_wind_h (hours with strong wind), mean_windspeed (m s⁻¹), mean_relh(%). Nights with minimum temperature < 5 °C, > 3 h of rain (≥ 0.2 mm h⁻¹) or > 3 h of strong wind (≥ 8 m s⁻¹), and nights with battery failure, are already excluded.
• 20230901_Species_BClass_Guild_Overview_ALH.xlsx: documentation (not read by the code). Lookup from BatScope classifications to the acoustic groups. Columns: bat_orginal_batscope_output, count (total one-minute intervals), short_name, b_classification, guild, regroup, final-classification (group used here), Richness_Diversity (groups used in the analysis).

Raw_data/Biodiversity_data/B6_birds_Joao_Pereira/

• B6_birds_Joao.csv (semicolon-separated; one row per point-count visit): plot_id, date_surv (dates written with Portuguese and German month abbreviations; parsed in B6_birds_preparation.R), start_time, visit_id, then one column per species (number of individuals seen or heard). Collected by M. Basile (2017–2018), J. Pereira (2019–2021) and helpers (Basile et al. 2020, 2021).
• species_to_remove.csv: species excluded from the analysis. species (scientific name), abbreviation (column code; empty if the species was not in the data), reason.

Raw_data/Biodiversity_data/B8_mammals_Sebastian_Schwegmann/

• CamEvents_Spring19.csv, CamEvents_Aut19.csv, EventsSp20.csv, EventsAu20.csv, EventsSpring2021.csv: one row per camera-trap event and season, as exported from the image-management software (German column names). The script uses ORDNER (folder), BILDNAME (image name), DATUM (date, dd.mm.yyyy), ZEIT (time), URSACHE(species or other trigger, in German), ANZTIERE (number of animals) and CamID (camera). The other columns are software metadata.
• Allcameratrapdata_Sebastian.csv (semicolon-separated): one row per camera deployment. PosID, CamID, PlotID, Position (C, IN, IS), DateSetup, TimeSetup, DateRet, TimeRet, Trapnights (used as sampling effort), plus camera-site habitat variables, which are described in Metadaten_Kamerafallen.xlsx.
• Metadaten_Kamerafallen.xlsx: documentation (not read by the code). Definitions (in German) of the columns of Allcameratrapdata_Sebastian.csv.

Raw_data/Biodiversity_data/B9_fungi_Max_Wieners/

• CFB_B9_P1_1_occurence_data_2022-11-16_formatted.csv: one row per record (species × deadwood object × visit). Plot_ID, subplot, sample_event_ID, date, taxon_RL (species), then attributes of the deadwood object: dw_tree_gen, dw_tree_fam, dw_tree_ord, dw_tree_class (host taxonomy), dw_pos (LY = lying, ST = standing, dw_l (length, m), dw_rad (radius, cm), dw_moss, dw_bark (moss and bark cover, %), dw_contact(ground contact, %), dw_dec_stage (decay stage I–V).

Raw_data/Environmental_data/

• 2018-08-29_A1_JF_Full_Forest_Inventory_korrekted.xlsx: full tree inventory. Sheet Legend gives the species codes, sheet Metadata gives the survey metadata, and there is one sheet per plot (sheet name = plot number) with Name, Number, Date, Species (code), SpeciesText, Diameter (DBH, mm), Height1. Codes 8, 21 and 22 are standing deadwood.
• 2018-11-05_A1_JF_forest_inv_ba-plot.csv: basal area per species group and plot (m² ha ^-1), with total and cf_share (coniferous share, used as cf_share).
• 2018-11-05_A1_JF_forest_dw-volume_plot.csv: plot_id, dw_volume (standing deadwood volume, m³ ha ^-1).
• lyingdeadwoodvolumes.csv (semicolon-separated): PlotID, deadwoodvolume (lying deadwood volume, m³ ha ^-1).
• all_dead_wood.csv: one row per deadwood object on line-intersect transects. PlotID; Deadwoodcategory (1 = stump, 2 = lying); Origin (1 = natural, 2 = artificial); Type (1 = hardwood, 2 = softwood); Decaystage (1 = fresh … 5 = almost completely decayed); DBH (cm); transectlength, Lengthskiddingtrail (m); fruiting.body, hollow_trunk (1 = present); comments. Two DBH values are corrected in 01c_calculate_deadwood_diversity.R.
• 2020-06-16_A1_JF_TLS_indicators.csv: structural indicators per terrestrial laser scan (several scans per plot; PlotID). SSCI and ENL are used; the other columns are point-cloud statistics (MFRAC, height quantiles zq*, etc.).
• Final_TreM_data_complete.xlsx: tree-related microhabitat survey (one row per tree; sheet Meta data explains the codes). Read by 01a but not used in the final models.
• 2022-03-20_CanopyLoss_Mean.csv: name (plot), mean_loss, mean_loss_with_NA (mean canopy cover loss, 0–1). Read by 01b_preparation_responses.R, which writes cover_loss.csv (plots with loss > 0.25). Not used in the final analyses.

3. Data not included in this archive (limitations)

• ConFoBi GIS database. 01a_data_curation_predictors.R reads plot location and terrain (elevation, slope, aspect, coordinates), UAV canopy metrics (gap_share, ndsm_mean, ndsm_sd, tri, gap_share_sd), plot precipitation and all landscape variables from a PostgreSQL/PostGIS database of the ConFoBi project. This database is not publicly accessible. The results of these queries are supplied as local_predictors.csv and landscape_predictors.csv, and all later steps run from these files. The landscape layers were derived from the sources listed under "Access information".

Code/software

All analyses were run in R 4.2.2 (also tested under R 4.3.3). Run each script with the working directory set to the archive root. Exception: 01a_data_curation_predictors.R sets its own working directory (step a1, which cannot be re-run). Required packages:

• Data preparation (a): vegan, openxlsx, reshape2, tidyverse (dplyr, tidyr), corrplot, data.table; for 01a only: DBI/RPostgreSQL, rpostgis, foreach, doParallel, mgcv, fmsb, ape, gdm, FactoMineR
• Analysis (b): vegan, mgcv, MuMIn, gdm; for gdm_varImp_bugfix.R: pacman, pbapply, parallel, doParallel, foreach
• Figures and tables (c): ggplot2, mgcv, gdm, texreg, visreg, reshape2, viridis; for 04: mgcViz, DHARMa, visreg, ape, gstat, fields, nlme

Scripts

Data preparation (a)

• 01a_data_curation_predictors.R: curates the raw plot, inventory and GIS data into local_predictors.csv and landscape_predictors.csv. Needs the ConFoBi database (see Limitations); the supplied output files replace this step.
• 01b_preparation_responses.R: runs the taxon scripts below in the required order (B5_bats_preparation.R must run before B2_plants_preparation.R and B6_birds_preparation.R, which use the bat matrix as the reference list of the 135 plots). It also writes cover_loss.csv.
• B2_plants_preparation.R: converts the long-format plant cover data into a plot × species matrix.
• B3_beetle_families_preparation.R: adds zero-catch samples, computes trap-unit days and pools beetle families per plot.
• B3_hymenoptera_preparation.R: sums brood cells over trap-nest corners and combines nest builders and parasitoids.
• B5_bats_preparation.R: pools bat activity per plot and counts logger-nights.
• B6_birds_preparation.R: parses survey dates, pools counts per plot, counts visits and removes the species in species_to_remove.csv.
• B8_mammals_preparation.R: merges the five seasons of camera-trap events, assigns plots, keeps one camera per plot and season, counts trap nights and pools mammal species per plot. Also writes mammals_samples.csv (cleaned event table).
• B9_fungi_preparation.R: counts records per species and plot.
• 01c_calculate_deadwood_diversity.R: calculates deadwood diversity (dw_div).
• 02a_biodiversity_data_exploration.R: combines the community matrices into communities.Rds and produces exploratory figures (richness, Shannon diversity, records and effort).
• 02b_predictors_selection.R: derives the final predictors (summed and log-transformed deadwood volume, log SSCI, northness, eastness, landscape coniferous share), checks correlations (|r| < 0.7) and writes select_pred_list.Rds.
• pairs_helpers.R: helper functions for pairwise correlation plots of the predictors (used by 02b).

Analysis (b)

• 03_run_all_models.R: full analysis pipeline. It fits the richness and Shannon-diversity GAMs (k = 5 thin-plate smooths, log link, REML; Gamma family for Shannon diversity; Poisson, quasi-Poisson or negative binomial for richness depending on the group; the landscape scale is chosen per model by AICc). It fits the β-diversity GDMs (Bray–Curtis, with geographic distance) and computes variable importance by matrix permutation. It partitions the explained deviance among the three predictor groups (unique and shared fractions), with a sensitivity analysis that moves cf_share, gap_share and avg_slope to the heterogeneity group. Finally, it computes multicollinearity diagnostics (VIF with stepwise removal at VIF > 10; GAM concurvity). Set the path variables data_dir and out_dir at the top of the script.
• gdm_varImp_bugfix.R: patched GDM variable-importance function (gdm.varImp2) used in the paper. If this file is present, the pipeline uses it so that Fig. 2 is reproduced exactly; otherwise gdm::gdm.varImp is used (same method, slightly different permutation values).

Figures and tables (c)

• 06_make_figures.R: reads the objects written by 03_run_all_models.R and writes each figure as a vector PDF and a 600-dpi PNG: Fig. 1 (effects of predictors on richness and Shannon diversity), Fig. 2 (variable importance for β diversity), Fig. 3 (deviance partitioning) and the supplementary concurvity figure.
• 05_Manuscript_figures_tables.R: LaTeX tables of the GAM results (texreg; Tables 2–3); per-taxon Supporting Information figures of significant GAM effects and GDM variable importance; mean R² and deviance explained. It also contains an earlier version of Figs 1–2, superseded by 06_make_figures.R.
• extract_gam.R: texreg extract method for mgcv GAMs (used by 05).
• 04_Model_checking_plots.R: model diagnostics (gam.check, DHARMa residuals, partial-effect plots), Moran's I tests of residual spatial autocorrelation, GDM I-spline plots and an exploratory attempt at spatial correlation structures for bat Shannon diversity.
• 05a_Presentation_figures.R: additional figures for a presentation; not needed to reproduce the paper.

How to reproduce

1. Unpack Raw_data.zip and arrange the other files as shown under "Workflow and folder structure".
2. (a) Run 01b_preparation_responses.R, 01c_calculate_deadwood_diversity.R, 02a_biodiversity_data_exploration.R and 02b_predictors_selection.R in this order, then to regenerate the files in Analysis_data/.
3. (b) Run 03_run_all_models.R.
4. (c) Run 06_make_figures.R, 05_Manuscript_figures_tables.R and, optionally, 04_Model_checking_plots.R.

Reproducibility notes

• Plot alignment. 03_run_all_models.R has an ALIGNMENT switch at the top. With "positional", predictor rows are taken in ascending plot-ID order and paired with the community rows in the order those appear in the file. With "byID", rows are matched by plot_id. The two settings give identical results for six of the seven groups, whose community rows are in ascending plot-ID order. The rows of the plant matrix are not in ascending order, so under "positional" 131 of the 133 plant plots are paired with another plot's predictor values; "byID" is the correct pairing. The published Tables 2–3 and Figs 1–2 were produced with "positional", and the published Fig. 3 with "byID". Set the switch accordingly to reproduce a given figure.
• Stochastic step. GDM variable importance uses random permutations. A seed is set in the script, but minor numerical differences between runs are expected.
• Fig. 1 effect categories. The significance of each effect is taken directly from the GAMs, while the shape category (monotonic, hump-shaped, etc.) is classified algorithmically in 06_make_figures.R. It may differ in a few cells from the manually assigned categories in the published figure (the manual assignments are in 05_Manuscript_figures_tables.R).

Access information

Data were derived from the following sources:

• Biodiversity data collected by ConFoBi sub-projects B2 (vascular plants; Helbach et al. 2022), B3 (beetles, trap-nesting Hymenoptera; Knuff et al. 2019), B5 (bats), B6 (birds; Basile et al. 2020, 2021), B8 (mammals) and B9 (fungi). Environmental data from sub-project A1 (forest inventory, deadwood, TLS) and the ConFoBi tree-microhabitat survey.
• Copernicus Land Monitoring Service, High Resolution Layer Forest Type 2018 (10 m): forest cover and forest type.
• LUBW (Landesanstalt für Umwelt Baden-Württemberg) land-cover classification
• Digital terrain model, 10 m.
• Sentinel-2 NDVI

References cited: Ehbrecht, M. et al. (2017) Agricultural and Forest Meteorology 242: 1–9. Helbach, J. et al. (2022) Ecology and Evolution 12: e8534. Knuff, A.K. et al. (2019) Methods in Ecology and Evolution 10: 1820–1825. Basile, M. et al. (2020) Journal of Environmental Management 264: 110401. Basile, M. et al. (2021) Ecological Indicators 133: 108402.

Citation

Please cite the associated published article and this dataset.

Contact

Carlos M. Landivar Albis, Department of Forest Growth, Silviculture and Genetics, Austrian Research Centre for Forests (BFW), Vienna, Austria. Email: carlos.landivar@bfw.gv.at