https://doi.org/10.5061/dryad.r2280gbmc

Description of the data and file structure

The following data are included. All raster files are on a 1km grid with CRS EPSG: 28992.

• modelInputs.zip
  • predictors
    • climate: All climatic predictor rasters used for the modelling, divided by scenario:
      • H: high emissions, used for SSPs 3 and 5
      • M: medium emissions, used for SSP4
      • L: low emissions, used for SSP1
      • r: reference scenario
      • Predictors have ‘breeding’ and ‘winter’ variants. Breeding means the breeding season and is different for each species (see the accompanying paper), winter is December 1st to February 28th (or 29th in leap years) and is the same for all species. The predictors are:
        • cold_days: number of days with minimum temperature below 0C
        • prec_covar: precipitation coefficient of variation
        • prec_dry: precipitation (mm) in driest 4-week period
        • prec_tot: total precipitation (mm)
        • temp_max: average maximum daily temperature (C)
        • temp_mean: mean daily temperature (C)
        • temp_min: average minimum daily temperature (C)
        • temp_range: average temperature range over the time period (C)
    • ref: All non-climate predictors rasters used for the reference scenario. The numbers in the file names indicate the buffer size
      • The predictors are:
        • altitude (m)
        • arable: proportion of arable farmland
        • CBC: Common Bird Census dataset (kept as zero)
        • clay: clay percentage in soil
        • forest: proportion of forested area
        • France: French dataset (kept as zero)
        • grass: proportion of grass area
        • industry_commerce_transport: proportion of industry, commerce or transport area
        • MAS: Meetnet Agrarische Soorten dataset (kept as zero)
        • mine_dump_construction: proportion of mine, dump or construction area
        • MUS: Meetnet Urbane Soorten (kept as zero)
        • open_unvegetated: proportion of open invegetated area
        • other_rural: proportion of rural area which does not fall under other rural types
        • pasture: proportion of pasture area
        • permanent crops: proportion of permanent cropland
        • sand: sand percentage in soil
        • tree: tree density
        • urban: proportion of urban area
        • urban_green: proportion of urban green area
        • vegetation: proportion of vegetated area
        • water_inland: proportion of inland water area
        • water_wetland_marine: proportion of marine water or wetland area
        • wetland_inland: proportion of inland wetland area
    • SSP1/3/4/5: All non-climate predictors rasters used for each SSP scenario. The numbers in the file names indicate the buffer size
      • The predictors are the same as for the reference scenario
• Outputs
  • predictions: raster files with predicted relative abundance for each species and scenario
  • uncertainty: raster files with the confidence interval for the predictions for each species for the reference period
  • variableImportance: csv files with the variable importance of each predictor for each species
  • anovaResults.csv: results of pairwise comparisons between predictions to test for significant differences. Results shown are afer the application of the Bonferroni multiple comparison test. The numbers after the species indicate the SSP scenario and ‘R’ indicates the reference prediction
• climate_ranges.csv: Climate ranges used in determining areas of France where the current climate is similar to the future Dutch climate. These values were taken from summarised statistics from future climate scenarios produced by the Dutch Meteorological Association: https://cdn.knmi.nl/system/data_center_publications/files/000/071/901/original/KNMI23_klimaatscenarios_gebruikersrapport_23-03.pdf?1696579253 

Code

All code was written in R, using R version 4.0.4 (the SDMaps package does not run in newer versions). The model generation was performed using the SDMaps package v0.15-6.

• makePredictors.R: Generates the current and future predictors (as raster files). This includes the derivation of future land use, as described in the document Deriving future and use.pdf
• runScenarios.R: Model generation and making reference and future predictions
• uncertainty.R: Runs bootstrapped model and calculates confidence interval of predictions
• anova.R: Performs anova on model predictions to check for significant differences

Supplementary material

• Data summary.pdf: A summary of the bird abundance data used for the modelling. The full dataset is available on request from the Dutch Centre for Field Ornithology (https://www.sovon.nl/)) and the Common Bird Monitoring Scheme database for France (https://www.vigienature.fr/fr/suivi-temporel-des-oiseaux-communs-stoc).
• Deriving future land use.pdf: Descibes the land use classifications used and the procedure for deriving future land use, grass and tree cover maps.
• Partial dependence plots.pdf: Shows the predictor maps and partial dependence plots for each species for all model covariates.