An update to the global Critical Habitat screening layer

This Zenodo record contains archived code for:

Dunnett, S., Muge, A., Ross, A. et al. An update to the global Critical Habitat screening layer. Sci Data 12, 1812 (2025). https://doi.org/10.1038/s41597-025-06117-y

Abstract

The International Finance Corporation (IFC) defines Critical Habitat in Performance Standard 6 (PS6) as high biodiversity value areas requiring net biodiversity gain for projects. We present an updated global screening layer of Critical Habitat aligned with IFC’s 2019 guidance. This layer derives from global datasets covering 54 biodiversity features, categorized as ‘Likely’ or ‘Potential’ Critical Habitat based on alignment  with IFC criteria and data suitability. Analysis indicates 53.95 million sqkm (10.58%) and 13.71 million sqkm (2.69%) of the globe can be considered Likely and Potential Critical Habitat respectively, with the remaining 86.73% not overlapping with assessed biodiversity features. This represents a significant increase over previous efforts but likely remains a significant underestimation of actual Critical Habitat. Likely Critical Habitat was dominated by Important Bird and Biodiversity Areas, Intact Forest Landscapes, and protected areas; Potential Critical Habitat by Important Marine Mammal Areas and ranges of IUCN Vulnerable species. Our results can help businesses prioritise impact avoidance and identify opportunities by screening potential development sites for biodiversity features.

The most up-to-date data layers can be found on the UN Environment Programme World Conservation Monitoring Centre data portal. This code and the basic layer are made available under CC BY but note that the drill down layer is made available under a CC BY-NC licence.

The process for compiling the complete Critical Habitat layers comprises three main steps, split across six separate scripts.

While some of the input data can be prepared using the scripts provided here through a custom direct interface with the UNEP-WCMC data portal, many input datasets cannot be provided here and must be sought, with permission, from their respective owners.

Where these are used in the scripts, we have included placeholder file paths starting with "PATH-TO" (or similar). These need to be replaced with their respective actual paths when known.

1. Data processing

1.1 Red_List_Preprocessing.R

Queries the IUCN Red List to produce three input datasets for the analysis: CR, EN, VU species, and a fourth, Great Apes, that did not end up in the final layer. The script includes code, currently commented out, that can add a desired buffer to these data.

1.2 Data_Preprocessing_Raster.R

Pre-processes any input data only available in raster format. This script includes the processing of extremely high resolution tropical moist forest and mangrove data, both of which can take days to complete.

1.3 Data_Preprocessing_Vector.R

The main pre-processing script: pre-processes all vector input data. Some datasets, e.g. warm-water coral reefs, are extremely detailed and take a substantial time to complete.

N.B. At this stage, the lookup table, lookup.csv, needs to be manually edited in Excel or similar to include all features for which data exist. For example, at the moment, many KBA criteria do not have any sites triggering them. So while a criterion may align with the definition of Critical Habitat, the pre-processing scripts will not output any data for them so they need removed from the lookup.

2. Creating Critical Habitat layers

2 Create_Drill_Down_Critical_Habitat_Raster_Layer.R

Combines all standardised input data from 1.1, 1.2, and 1.3 to produce the finished drill down raster layer with accompanying attribute table in WGS 84.

3 Post-processing

3.1 Create_Basic_Critical_Habitat_Raster_Layers.R

"Flattens" the drill down raster layer produced in 2 to only contain one attribute: whether a cell is Likely Critical Habitat (10), Potential Critical Habitat (1), or Unclassified (0).

3.2 Create_Drill_Down_Polygons.R

Polygonises the drill down raster layer produced in 2 and reorders the variables to be more user-friendly in a way not possible with raster data.