IAS-pDT modelling workflow — 2. abiotic data

This article details the processing of abiotic data within the IAS-pDT modelling workflow. These processed data serve as predictor variables in the species distribution models. All data preparation adheres to the FAIR principles (Findable, Accessible, Interoperable, and Reusable) to ensure scientific integrity and reproducibility.

Reference grid

The workflow employs the European Environment Agency (EEA) reference grid, standardized at a 10×10 km resolution across the study area. This grid utilizes the ETRS89-LAEA Europe coordinate reference system (CRS; EPSG:3035).

Corine land cover and habitat data

The CLC_Process() function manages the processing of Corine Land Cover (CLC) data within the IAS-pDT workflow. It computes the percentage coverage and predominant classes per grid cell across all three CLC levels, alongside EUNIS and SynHab habitat classifications. A custom crosswalk was used to transform the CLC Level 3 data into ecologically meaningful EUNIS and SynHab habitat classes. The resulting data serve the following purposes:

• model grid selection: Grid cells with at least 15% land cover (default threshold) are retained for modelling.
• habitat-specific modelling: The percentage coverage of SynHab habitat types informs model fitting by:
  1. excluding grid cells with zero coverage of the relevant habitat type during model fitting.
  2. serving as a potential predictor variable in the models.

Biogeographical regions

The BioReg_Process() function retrieves and processes the biogeographical regions dataset from the EEA. It extracts the names of biogeographical regions corresponding to each reference grid cell, enabling the quantification of species presence across these regions.

CHELSA climate data

The CHELSA_Process() function manages the retrieval and processing of CHELSA (Climatologies at High Resolution for the Earth’s Land Surface Areas) climate data for the study area across multiple climate scenarios. CHELSA delivers high-resolution global datasets encompassing a range of environmental variables for current conditions and future projections. These processed data are integrated into the species distribution models as predictor variables. For each environmental variable, the dataset encompasses 45 future scenarios, derived from the combination of five CMIP6 climate models, three Shared Socioeconomic Pathways (SSPs), and three future time periods (refer to the CHELSA technical specifications for details). Additionally, future climate model outputs are aggregated to generate ensemble predictions for each SSP and time period combination.

Climate model	Institution
mpi-esm1-2-hr	Max Planck Institute for Meteorology, Germany
ipsl-cm6a-lr	Institut Pierre Simon Laplace, France
ukesm1-0-ll	Met Office Hadley Centre, UK
gfdl-esm4	National Oceanic and Atmospheric Administration, USA
mri-esm2-0	Meteorological Research Institute, Japan

Shared Socioeconomic Pathway	Description
ssp126	SSP1-RCP2.6 climate as simulated by the GCMs
ssp370	SSP3-RCP7 climate as simulated by the GCMs
ssp585	SSP5-RCP8.5 climate as simulated by the GCMs

Railways and roads intensity

The Railway_Intensity() and Road_Intensity() functions retrieve and process railway data (sourced from OpenRailwayMap) and road data (sourced from the Global Roads Inventory Project; GRIP), respectively, for the study area. Road intensity reflects site accessibility, habitat disturbance levels, and IAS dispersal potential, while railway density serves as a proxy for IAS dispersal routes. The summed lengths of railways and roads per grid cell, transformed to a logarithmic scale (log₁₀), are incorporated into the models as predictor variables.

River length

The River_Length() function processes data from the EU-Hydro River Network Database to compute river lengths categorized by Strahler order for each grid cell. The Strahler order, a hierarchical classification of river networks, assigns higher numbers to larger, more significant river segments. For each grid cell, the function calculates the cumulative length of rivers at or above a given Strahler order (e.g., for Strahler 5, it includes rivers with Strahler values of 5 or greater). The total river length per grid cell for Strahler order 5 and above, transformed to a logarithmic scale (log₁₀), serves as a potential predictor variable in the species distribution models.

Sampling efforts

To address the opportunistic bias inherent in the presence-only data, the total number of vascular plant observations per grid cell from the Global Biodiversity Information Facility (GBIF) is employed as a proxy for sampling effort. The Efforts_Process() function manages the request, retrieval, and processing of this sampling effort data, encompassing over 260 million occurrences as of March 2025. Beyond total observations, the function also determines the number of vascular plant species per grid cell. The processed data support two primary applications:

• grid cell filtering: The number of species per grid cell enables optional filtering to exclude areas with insufficient sampling effort (e.g., grid cells with fewer than 100 observed vascular plant species are excluded).
• sampling bias correction: The total number of observations per grid cell, on a logarithmic scale (log₁₀), is incorporated as a predictor variable in the models to account for sampling bias. To mitigate this bias during predictions, this predictor is fixed at a constant value representing optimal sampling effort across the study area, as detailed in Warton et al. (2013).

---

Previous articles:
↠1. Overview
Next articles:
↠3. Processing biotic data
↠4. Model fitting
↠5. Model post-processing