https://doi.org/10.5061/dryad.sj3tx96bs

These files include everything needed to run the DynaCLUE model to replicate our results. This includes the inputs for each scenario and for the model validiation. These files also include the model outputs, i.e. land use maps for the Netherlands for 2050 under each of four scenarios.

Description of the data and file structure

DynaCLUE

The DynaCLUE model is available to download here: https://www.environmentalgeography.nl/site/data-models/data/clue-model/

We used version 2.0.

To run the model with our inputs, save all the model inputs for one scenario in the same folder as the DynaCLUE executable file. Then open DynaCLUE, select the region and demand files, and click 'Run'. The outputs will be saved in the same folder.

Model inputs

The model inputs are contained in 5 zip files: SSP1, SSP3, SSP4, SSP5 and Validation.

Each of these contains the following:

Land use types are referred to by numbers. These are as follows: 0 - Urban, 1 - Pasture, 2 - Crops, 3 - Forest, 5 - Non-forest nature

All raster files are on a 1km grid with CRS EPSG: 28992.

Full details of how DynaCLUE input files are structured is available here: https://www.environmentalgeography.nl/site/data-models/data/clue-model/

Model outputs

Model outputs are in the following files:

Additional files

• rasterTemplate.asc
• land_types.txt
• currentLU.asc
• correlations.xlsx
• predictors.txt
• predictor_names.txt
• predictorFiles.zip
• regressionOutput.csv
• LandPrice_RegressionOutput.csv
• SSP4demand.tif
• SSP5demand.tif

rasterTemplate.asc is a blank raster, providing the grid which is used in all the analysis

land_types.txt lists how the CORINE land use types were classified into our 5 land use types.

currentLU.asc is a raster showing current (2018) land use in the Netherlands, using our 5 land use types.

correlations.xlsx is a spreadsheet containing the correlations between all the different predictors which were considered for use.

predictors.txt is the list of predictors used for each land use type for the logistic regression analysis.

Each environmental predictor used in the model is referred to by a number. Predictor_names.txt shows which predictor has which number.

predictorFiles.zip includes all the predictor data used in the logistic regression analysis.

regressionOutput.csv contains the output of the logistic regression analysis which was used to establish relationships between the land use types and the different environmental predictors.

LandPrice_RegressionOutput.csv contains the output of the regression analysis which was performed to enable the prediction of future agricultural landprices.

SSP4demand.tif was used for calculating potential forest succession under SSP4. It shows the area that was deemed to have the potential for succession. This is the pasture, crop and non-forest nature area from 2018, excluding the wealthy areas and areas near centres for knowledge and business. Comparing this with the total pasture, crop and non-forest nature area from 2018 (see cov_all.0 in above table) suggests that 87% of this area has the potential for succession.

SSP5demand.tif was used for calculating potential forest succession under SSP5. It shows the area that was deemed to have the potential for succession. This is the non-forest nature area from 2018, excluding areas near major population centres. Non-forest nature which has the potential for succession has value '4', while non-forest nature near population centres has value '5'. This suggests that 67% of 2018 non-forest nature area has the potential for succession.

Code/Software

Two R scripts are included:

• historicChanges.R
• regression.R

historicChanges.R determines historic land use changes in the Netherlands from 1990 to 2018 and uses this to calculate the conversion elasticities used in the DynaCLUE model. In addition to the files included here, this also requires the CORINE land use maps for the Netherlands.

regression.R performs logistic regression analysis to determine land use suitability and creates the alloc1.reg file to use as an input to DynaCLUE.

All analysis was performed using R v4.0.4. These scripts require the following packages:

• dplyr (v1.0.7)
• MASS (v7.3-56)
• pROC (v1.18.0)
• raster (v3.5-2)
• rgdal (1.5-23)

Additional files

• Supplementary_tables.docx

This includes several tables complementing the journal article.