Critical Habitat criteria and scenarios

IFC PS6 defines Critical Habitat as areas with high biodiversity value, based on a set of five criteria (Table 1). Critical Habitat may also be triggered by other recognized high biodiversity values, described in detail in the IFC Guidance Note 6 (GN6) [10]. These are referred to here as scenarios A and B following the approach in Martin et al. [18]. Scenario A includes additional biodiversity values that are not included within the criteria, for example “Ecosystems of known special significance to Endangered or Critically Endangered species for climate adaptation purpose”. Scenario B refers to internationally or nationally recognized areas of high biodiversity value that are explicitly mentioned within GN6, such as UNESCO natural World Heritage sites and Wetlands of International Importance under the Ramsar Convention (for more detail see S1 Table in [18]). Within GN6, Criteria 1–3 are further divided into Tier 1 and Tier 2 Critical Habitat depending on the relative vulnerability and irreplaceability of the biodiversity present. Mapping these tiers would require more refined datasets than those that are available globally and, consequently, they are not indicated in the screening layer developed here.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Critical Habitat criteria and scenarios.

https://doi.org/10.1371/journal.pone.0193102.t001

Data screening and classification

Relevant spatial datasets were identified through consultation among the authors and other experts based on the following criteria adapted from Martin et al. [18]: (i) direct relevance to one or more Critical Habitat criteria/scenarios; (ii) global in extent; (iii) assembled using a standardised protocol; (iv) the best available data for the biodiversity feature of interest; and (v) sufficiently high resolution to indicate presence of biodiversity on the ground at scales of relevance to business operations. Where data were not published or available for use at the time of the analysis, they were excluded from the study (S2 Table).

Datasets retained for the analysis were classified as ‘likely’ or ‘potential’ Critical Habitat based on two variables: alignment of the dataset with the Critical Habitat definition, and spatial resolution of the dataset indicating presence on the ground (Fig 1). Biodiversity features represented by data with strong alignment with one or more Critical Habitat criteria or scenarios A or B, and high spatial resolution were classified as likely Critical Habitat. Where alignment with Critical Habitat criteria and scenarios was less strong or the spatial resolution of the dataset was coarser, features were mapped as potential Critical Habitat. These categories of Critical Habitat indicate the likelihood of finding Critical Habitat in subsequent on-site assessments. Areas outside of likely or potential Critical Habitat are classified as ‘Unknown’. These areas include both Critical Habitat for which no suitable global-scale biodiversity data currently exist and areas that do not qualify as Critical Habitat based on their biodiversity values.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Screening layer classification scheme.

Classification of data subsets as likely or potential Critical Habitat is based on the strength of alignment with IFC PS6 criteria and scenarios and the spatial resolution of the data (adapted from Martin et al. [18]).

https://doi.org/10.1371/journal.pone.0193102.g001

Data processing and spatial analysis

The identification of terrestrial Critical Habitat follows the methodology developed for screening Critical Habitat in the marine realm outlined by Martin et al. [18]. The analysis uses the same likely and potential Critical Habitat classification scheme, resolution, and shoreline delineation (GSHHG version 2.2.2, [20]). The screening exercise retained datasets in both vector and raster format, all of which were converted into raster layers of 1 km grid cell size in cylindrical equal-area projection. Vector data represent the location and extent of features using points, lines and polygons. Raster data consists of a matrix of cells, where each pixel contains a value representing information. During the vector to raster conversion, likely or potential grid-cell values were assigned upon intersection with the relevant point, polygon or line features, irrespective of the area of overlap or frequency of features. The 1 km grid cell size ensures that point features are visible on the final layer, and that boundaries of point and polygon features are preserved. The final map is a composite of all underlying layers produced by successively combining individual rasters and retaining the highest class for overlapping grid cells. In hierarchical order: likely grid cells (purple in Fig 1) were retained over potential grid cells (pink), in turn being retained over unknown grid cells (grey). A key benefit of using the same methodology for both terrestrial and marine realms is that it can facilitate the combination of terrestrial and marine Critical Habitat screening layers in the future.

Selected datasets

Thirty-six data sources were identified as suitable to review for Critical Habitat. Of these, 12 datasets (Table 2) relating to 12 biodiversity features (Table 3) were retained for use in the screening layer. These include datasets that represent biodiversity features explicitly referred to in GN6, those directly aligned with the PS6 criteria, and those for which the existing scientific literature provides an evidence-based justification for alignment with the PS6 criteria (S1 Table). Note that a single dataset may provide spatial information on multiple biodiversity features, which may be classified differently as likely or potential depending on their alignment with IFC PS6 criteria and scenarios.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Datasets incorporated into the Critical Habitat screening layer.

https://doi.org/10.1371/journal.pone.0193102.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Biodiversity features included in the analysis, their alignment with IFC PS6 Critical Habitat criteria and scenarios, and classification as ‘likely’ or ‘potential’ Critical Habitat.

https://doi.org/10.1371/journal.pone.0193102.t003

The datasets can be broadly categorised as: 1) protected areas, 2) Key Biodiversity Areas, 3) threatened ecosystems, 4) critical sites for selected species (tigers and sea turtles), and 5) the distributions of threatened species qualifying under IUCN Red List criterion D. Detailed justifications for the exclusion or retention of individual datasets and features are provided in the supplementary materials (S1 Table and S2 Table respectively). Below we describe these five types of datasets and how, for some datasets, we extracted subsets of the data where there is alignment with Critical Habitat criteria and scenarios.

1. Protected areas. The specific subsets of protected areas included are: natural and mixed World Heritage Sites, Wetlands of International Importance (Ramsar sites), and protected areas under IUCN management categories Ia, Ib and II. These are explicitly referred to in GN6, and always treated as Critical Habitat (paragraph GN115 in [10]). GN6 indicates that areas under IUCN management categories III-VI and areas for which the IUCN management category was not reported or not assigned may qualify as Critical Habitat if underlying biodiversity values align with Critical Habitat criteria. In this study, the top 100 irreplaceable protected areas worldwide as identified by Le Saout et al. [21] were also included, which reflect GN6’s ‘Areas determined to be irreplaceable’ (GN57). Additionally, a subset of all protected areas in the World Database on Protected Areas (WDPA) [22] was included based on an inferred alignment with PS6 criteria 1 and 2. The subset comprises protected areas that overlap with ≥10% of the range of one or more Critically Endangered (CR) or Endangered (EN) species within the IUCN Red List ([23]), or ≥95% of the range of one or more restricted-range species. Restricted range species were defined as those with a range of less than 50,000 km² (IFC GN6), and the percent range overlap was used as a proxy for percent population. It is recognised that percent range is an imperfect measure of percent population due to uneven distributions of individuals, but this was considered to be the best available approach to indicate alignment of protected areas for which the underlying basis for designation is not reported.
2. Key Biodiversity Areas, including Important Bird and Biodiversity Areas (IBAs) and Alliance for Zero Extinction sites (AZEs). KBAs are identified using threshold-based criteria [24], and subsets of the KBA dataset were extracted where these aligned with Critical Habitat criteria. A global standard for the identification of KBAs was recently published [25], but we used the criteria and thresholds for identifying KBAs specified in Langhammer et al. [26] as all non-IBA and non-AZE KBAs identified to date have used these criteria. In addition, they were taken into consideration during the development of the PS6 criteria and consequently display direct alignment. This is expected to change in view of the new KBA Standard and a proposed alignment between the new KBA criteria and PS6 is provided in S3 Table.
3. Threatened ecosystem datasets. GN6 [10] does not provide detailed numeric thresholds to identify highly threatened and unique ecosystems, but points to the IUCN Red List of Ecosystems (RLE) as a suitable standard in development (paragraph GN62). The RLE provides quantitative guidelines to assess the threatened status of ecosystems [27] and RLE thresholds for threatened ecosystems were used here to assess the suitability of globally mapped ecosystems for inclusion in the Critical Habitat data layer under criterion 4. Global scale datasets available for assessment included mangroves [28], saltmarshes [29], ever-wet tropical forests [30], tropical dry forest [31], and tropical montane cloud forest [32, 33]. The RLE is intended to be applied at a finer geographic scale than these global layers. However, as there have been no consistent sub-global assessments of these ecosystem types, each ecosystem dataset was assessed against RLE criteria based on existing scientific literature outlining the global extent and rate of habitat loss of these ecosystem types. We recognise, nevertheless, that there may be substantial regional differences in their vulnerability. For example, past deforestation rates for tropical dry forests range from 2–18% over a 20 year period for different geographical areas [31], and the loss of mangrove forests varies between 3.6% and 1.3% per year in the Americas and Africa, respectively [34]. Integrating regional differences would be valuable but is beyond the scope of this paper. As in all cases, the classification of likely or potential Critical Habitat depends on the alignment with IFC PS6 criteria and scenarios and the spatial resolution of the dataset. Alignment was assessed based on the degree to which the RLE threshold would be met globally. The ecosystem was classed as potential Critical Habitat if regional differences indicated that the RLE threshold would be unlikely to be met consistently across regions.
4. Critical areas for selected species. Only two global datasets at a suitable spatial resolution for selected species were identified: Sea Turtle Nesting Sites and Tiger Conservation Landscapes. The distribution of sea turtle nesting sites [35] was disaggregated based on the species’ IUCN Red List category. All nesting sites were included as potential Critical Habitat under Criteria 3 and 4, and those for Critically Endangered and Endangered species were included as likely Critical Habitat due to alignment with Criteria 1 and 2, respectively. Similarly, we extracted source sites and potential source sites for the Endangered tiger Panthera tigris from the Tiger Conservation Landscapes dataset [36]. Source sites are areas that contain concentrations of tigers that have the potential to repopulate larger landscapes and were therefore aligned with Criteria 1. Potential source sites are those where the case for an area’s inclusion as a Source Site is equivocal and more surveys are required [37]. These sites have therefore been included as ‘potential’ Critical Habitat.
5. Threatened species’ ranges. For most species, range maps from the IUCN Red List are not appropriate to include because they represent limits of distribution which can cover vast areas, particularly for wide-ranging species, and are not suitably refined based on an understanding of suitable habitat. However, species that are categorised as highly threatened based on a restricted distribution or population size were considered to be appropriate for inclusion as potential Critical Habitat. Species categorised as Critically Endangered or Endangered under IUCN Red List criterion D (i.e. global population size estimated to number fewer than 50 or 250 mature individuals respectively) were included based on alignment with PS6 Criterion 1. Species categorised as Vulnerable under IUCN Red List criterion D2 (where the population has a very restricted area of occupancy, typically less than 20 km², such that it is prone to the effects of human activities or stochastic events) were included based on alignment with PS6 Criterion 2 [38].

Coverage and relative composition

Of the total terrestrial area analysed, 10.1% (14,986 x 10³ km²) was classified as likely Critical Habitat and 5.1% (7,607 x 10³ km²) as potential Critical Habitat (Fig 2). The remaining 84.8% (125,794 x 10³ km²) is classified as unknown, either because the known biodiversity features are not suitably aligned with the Critical Habitat definition or because of a lack of data to make this assessment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Global screening layer for terrestrial Critical Habitat.

Likely and potential Critical Habitat are depicted in purple and pink, respectively. Unknown areas are depicted in dark grey. Marine areas are depicted in blue, and were not assessed. The screening layer is developed as a raster of 1 km grid cell size.

https://doi.org/10.1371/journal.pone.0193102.g002

An assessment of the amount of coverage by datasets from individual criteria or scenarios found that scenario B (internationally or nationally recognized areas of high biodiversity value) was the primary driver of area classified as likely Critical Habitat (56% or 8.36 million km²; Table 4). Criterion 4 (highly threatened and/or unique ecosystems) was the primary driver of area classified as potential Critical Habitat (78% or 11.92 million km²). Together, criteria 1 and 4 and scenario B accounted for approximately one-third of all the area designated as likely and potential Critical Habitat. There were no datasets identified that aligned with criterion 5 or scenario A.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Surface areas of likely, potential and combined (likely/potential) Critical Habitat (CH) identified under individual criteria and scenarios.

https://doi.org/10.1371/journal.pone.0193102.t004

The different biodiversity features assessed also had a varied effect on the classification of Critical Habitat (Table 5). Overall, protected areas and Key Biodiversity Areas triggered the largest areas of Critical Habitat, accounting for 45% (10.16 million km²) and 39% (8.73 million km²) of combined likely or potential Critical Habitat, respectively. Within these, the largest coverage is recorded for Important Bird and Biodiversity Areas and protected areas under IUCN management categories I-II, which result in 41% and 31% of likely Critical Habitat, respectively. The main contributors to potential Critical Habitat include ever wet tropical forests (28%), tropical dry forests (26%) and Important Bird and Biodiversity Areas (18%). In many cases, more than one biodiversity feature triggered the classification of a single area as Critical Habitat, for example 20% (3.07 million km²) of likely Critical Habitat is triggered by three or more data subsets.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Surface areas of likely, potential and combined (likely/potential) Critical Habitat (CH) triggered by individual biodiversity features.

https://doi.org/10.1371/journal.pone.0193102.t005

Current and future datasets for Critical Habitat

This study presents a single terrestrial layer derived from 12 biodiversity datasets that align with one or more of the criteria defined by IFC PS6 for Critical Habitat or one of two scenarios based on the PS6 standard and guidance notes. The final Critical Habitat screening layer covers 15% of the terrestrial surface. While the distribution of potential or likely Critical Habitat varies across countries and regions, this pattern is in part driven by the availability of data and therefore cannot accurately reflect the actual distribution of Critical habitat.

The screening layer has gaps, as reliable and up-to-date spatial information is not available across all taxonomic groups and ecosystem types. For example, of an estimated 8.7 million species on earth [39], only ca. 85,000 species have been assessed for the IUCN Red List of Threatened Species. Furthermore, where they exist, range maps for most species show distributional boundaries rather than known occupancy owing to a lack of sufficiently high-resolution data on presence or confirmed absence. Likewise, global spatial data on ecosystems often focus on more economically valuable ecosystems, or those that can be mapped using remote sensing, such as forests [40] and wetlands [41]. For nearly all countries worldwide, KBAs have been identified for birds (IBAs) and for highly threatened and range-restricted vertebrates, corals and conifers (AZEs). While KBAs have been identified for other taxonomic groups in many countries, coverage is far from complete [42]. The recent adoption of a new global standard for the identification of KBAs [25], and the launch of the World Database of Key Biodiversity Areas [24] should help to expand rapidly the comprehensiveness of the global KBA inventory. While there are a large number of national and local survey efforts underway to map species and ecosystems, compiling these data at the global level remains a significant task.

Data availability varied across the IFC criteria. Specifically, Criterion 1 (Habitats of significant importance to Critically Endangered and/or Endangered species), Criterion 2 (Habitats of significant importance to endemic and/or range restricted species) and Criterion 3 (Habitats supporting globally significant concentrations of migratory and/or congregatory species) were particularly influential in the final screening layer due to the availability of large polygon-based spatial datasets (i.e. KBAs and protected areas) that aligned well with the IFC criteria.

Criterion 4 (Highly threatened or unique ecosystems) was represented by specific ecosystems (e.g. saltmarshes, mangroves, or tropical dry, montane or ever-wet forests). Expert review of the forest datasets concluded regional variations in their threatened status and/or relatively low levels of certainty in relation to on-ground presence of these mapped ecosystems. Consequently, this resulted in the attribution of potential (rather than likely) Critical Habitat status to these layers (Table 3). Data certainty is higher for datasets on the global distribution of saltmarshes and mangroves, resulting in a classification as likely Critical Habitat. These were compiled based on occurrence datasets (saltmarsh) and Landsat imagery validated through distribution data and published literature (mangroves). Criterion 5 (Areas associated with key evolutionary processes) had no available data as a result of the difficulty associated with mapping these areas.

There are a large number of broad-scale priority-setting approaches that identify regions of the world based on varying degrees of irreplaceability and/or vulnerability. These include Biodiversity Hotspots [43], Endemic Bird Areas [44], Intact Forest Landscapes [45], Centres of Plant Diversity [46], Crisis Ecoregions [47], Megadiversity Countries [48], Last of the Wild [49], High-Biodiversity Wilderness Areas [50], and Global 200 ecoregions [51]. While these areas are aligned with the definition of Critical Habitat (S4 Table) they are spatially extensive and not sufficiently refined to indicate presence on the ground and are therefore excluded from this screening layer. However, they do provide important background context that could support Critical Habitat assessments and inform their interpretation.

This screening layer is envisaged to evolve as new global datasets become available. There are a number of different types of data which could, with some further development, be included in subsequent iterations of this analysis. Of the 36 datasets that were considered for inclusion in this study, 25 were excluded as they were too coarse to indicate presence of the features at a resolution suitable for project planning (e.g. IUCN range data and a number of datasets that are based on these, Crisis Ecoregions, Biodiversity Hotspots); were considered insufficiently aligned with Critical Habitat criteria (e.g. a number of subsets of the Global Lakes and Wetlands database, and datasets of physical features such as mountains and islands); or were not finalised or publicly available (e.g. High Conservation Value areas, and Important Plant Areas [52] (S2 Table).

Of datasets currently under development, the IUCN Red List of Ecosystems (RLE) is particularly noteworthy in the context of this assessment. As yet the RLE is only represented by 20 case studies spread across the world [53]. Although it is expected to be several years before global maps are available, national maps will be released gradually over the coming years. The RLE is specifically referred to within the IFC criteria and in later years will be the principal global dataset that aligns with Criterion 4 on highly threatened and/or unique ecosystems. Other datasets that will likely become available and inform future iterations of this screening layer include refined range data based on extent of suitable habitat for threatened and endemic species, and globally collated assessment data for High Conservation Value areas. It is therefore expected that over time the screening layer will become more comprehensive both geographically as well as in its coverage of taxonomic groups and ecosystem types.

Interpretation and use of the Critical Habitat screening layer

The screening layer of Critical Habitat can provide information for companies in the early stages of project development by highlighting areas of potential or likely Critical Habitat presence. By providing information on the underlying biodiversity features that trigger classification, this screening layer can inform subsequent more detailed Critical Habitat assessments by companies and support the evaluation of biodiversity risk of proposed operations by finance institutions.

However it is important to recognise the limitations of a data layer based on currently available, global spatial data. In particular, the classification of an area as ‘unknown’ in this assessment can include areas of Critical Habitat for which there are no available global data or areas without the necessary biodiversity values to trigger a Critical Habitat designation. There are many elements of the Critical Habitat definition for which no appropriate global data are available. In addition, the resolution, age, and completeness of the datasets create the potential for omissions (where datasets omit the presence of the biodiversity feature) and commission (where a biodiversity feature is reported as present but is actually absent). Definitive identification of Critical Habitat that is needed prior to planning projects and mitigation measures will require finer scale data (e.g. records of threatened species at particular locations, which are increasingly available through citizen science repositories) combined with in-situ biodiversity surveys by qualified assessors to confirm or dismiss the presence of Critical Habitat and more accurately define boundaries. Once confirmed and defined and operators decide to continue, IFC PS6 requires those values for which it is designated to be enhanced to deliver a net gain of biodiversity and therefore robust mitigation measures are required to achieve this.

As presented in this paper, a static compound layer means that some areas of Critical Habitat may not be visible when viewed at a global scale and it precludes the ability to interrogate the layer to identify the underlying features triggering classification. The utility and impact of this work would be increased by making the layer available through an interactive platform, such as the Integrated Biodiversity Assessment Tool (IBAT, [12]) where users could zoom in to specific areas of interest and add contextual layers by providing access to relevant information on biodiversity features not included in the screening layer. For example, protected areas which are not all incorporated into the Critical Habitat screening layer are still subject to a set of requirements on the basis of their legal designation and / or international importance [6]. Equally some of the broad-scale priority-setting approaches such as Endemic Bird Areas and Biodiversity Hotspots referred to in the earlier section that are aligned with the Critical Habitat definition (S4 Table) but excluded on the basis of their spatial resolution, are included in IBAT and would add important context to this screening layer. Tools such as IBAT are regularly updated as new data become available and data users are therefore accustomed to the evolving nature of these datasets.

Using this layer in synergy with the marine version [18] will provide insight regarding global Critical Habitat across all realms. Results presented in Martin et al. [18] revealed a lower proportional coverage of Critical Habitat in the marine realm, with 1.6% and 2.1% of the area classified as likely and potential Critical Habitat. This is due to the lower number of datasets available for the marine realm, particularly in areas beyond national jurisdiction and the deep seas [54; 55] as well as the spatial resolutions of those datasets. This can be seen in differences between the spatial properties (geometries) of the features used in the two layers; the terrestrial features are generally better resolved, showing greater detail around feature boundaries.

This study provides a methodology to integrate a number of global biodiversity datasets and make them accessible and relevant to business at a crucial stage in the project life cycle. By tailoring the layer to the biodiversity specificities of a particular standard to which business are complying (IFC PS6) there is a direct and much needed application for this work. This study has highlighted a range of opportunities for expanding and improving the data used within this screening layer and the next step is to operationalise this as an accessible, evolving and updateable data product aimed at supporting improvements to biodiversity screening and management.