Metadata Factsheet

3. Goals And Targets Addressed

3a. Goal

N/A

3b. Target

Target 3: Ensure that at least 30% globally of land areas and of sea areas, especially areas of particular importance for biodiversity and its contributions to people, are conserved through effectively and equitably managed, ecologically representative and well-connected systems of protected areas and other effective area- based conservation measures, and integrated into the wider landscapes and seascapes.

5. Definitions Concepts And Classifications

5a. Definition

PARC-representativeness

PARC-representativeness scores are initially calculated separately for each and every grid-cell (both protected and unprotected) within the spatial domain of interest (for the existing global implementation these are 30- arcsecond cells, approximately 900m wide at the equator). For each cell of interest this score is an estimate of the proportional protection of all cells that are ecologically similar to this cell – i.e. the proportion of all ecologically-similar cells that are included in the protected-area system (see ‘Method of computation’ below). In the past the protection status of grid-cells has been based on the World Database on Protected Areas (WDPA), but this can now be readily extended to include data from the World Database on OECMs https://www.protectedplanet.net/.

The aggregate PARC-representativeness score for any larger spatial unit of interest – e.g. a country, an ecoregion, or an ecosystem type – is then derived as a weighted geometric mean of the individual scores of all cells falling within that unit, with the contribution of each cell weighted according to its ecological uniqueness. Aggregate scores for the PARC-representativeness index are therefore expressed in familiar units of the proportional (or percentage) coverage of protected areas and OECMs. However, unlike more basic indicators of proportional protection – e.g. Protected Area Coverage of Ecoregions https://www.bipindicators.net/indicators/protected-area-coverage-of-ecoregions – the PARC-representativeness score obtained for any given spatial reporting unit is a function not only the overall area of protection, but also of the extent to which that protection is spread evenly, and therefore ‘representatively’, across any gradients of environmental and biological variation within the unit, rather than being biased towards particular parts of this variation (e.g. steeper less-productive lands).

PARC-connectedness

PARC-connectedness scores are initially calculated separately for each and every protected grid-cell within the spatial domain of interest. This score reflects the functional connectedness of the cell of interest to other protected cells in the surrounding landscape, using a cost-benefit analysis technique founded on well-established principle of metapopulation ecology (see ‘Method of computation’ below). As for PARC-representativeness, ‘protection’ has, to date, been based on data from the WDPA, but this can now be readily extended to include data from the World Database on OECMs. In the past, derivation of PARC-connectedness scores has also taken into account the connectedness of each protected cell not only to other protected cells, but also to unprotected cells containing primary vegetation in the surrounding landscape, in accordance with the stated ambition of Aichi Target 11 to establish “well-connected systems of protected areas … integrated into the wider landscape”. This extension should be considered as optional for application of PARC-connectedness to post-2020 GBF reporting – i.e. if need be, the index can be readily generated to reflect connectedness between protected cells alone, without any consideration of connectedness to unprotected primary vegetation.

The PARC-connectedness score assigned to each protected cell is expressed as a proportion (0 to 1) of the maximum possible level of connectedness obtainable if that cell were surrounded by a continuous expanse of protection in the surrounding landscape. The aggregate PARC-connectedness score for any larger spatial unit of interest – e.g. a country – is then derived by simply averaging the individual scores of all protected cells falling within that unit, and is therefore also expressed as a proportion.

Composite PARC indicator

The composite PARC indicator is derived by extending the calculation of PARC-representativeness to further weight the contribution of each protected cell by its individual PARC-connectedness score. Therefore, as for PARC-representativeness, aggregate scores for the composite PARC indicator are expressed in familiar units of the proportional (or percentage) coverage of protected areas and OECMs, but with the score of any given reporting unit (e.g. a country) on this scale rigorously adjusted for the effects of both representativeness and connectedness.

5b. Method Of Computation

PARC-representativeness

The existing global implementation of PARC-representativeness generates results using a grid of cells covering the entire terrestrial surface of the planet at 30-arcsecond resolution (these cells are approximately 900m wide at the equator).

The PARC-representativeness score for cell i is calculated as:

where sij is the ecological similarity between cell i and each other cell j in the grid, and pj is the protection status of that cell (0 if none of the cell is protected, 1 if all of the cell is protected, and a proportion between 0 and 1 if only part of the cell is protected).

The ‘ecological similarity’ between any pair of cells is the proportional overlap in species composition expected between those locations prior to any habitat transformation, ranging from 0 (for a pair of cells expected to have had no species in common) through to 1 (for a pair of cells expected to have supported exactly the same species). For the global implementation of PARC-representativeness these ecological similarities are predicted as a function of the abiotic environmental attributes and geographical locations of the cells concerned, using a set of generalised dissimilarity models (GDMs; Ferrier et al 2007) previously fitted to occurrence records for more than 400,000 species of vascular plants, vertebrates and invertebrates worldwide. This set of models includes 183 separately fitted GDMs, one for each possible combination of three broad biological groups – plants, vertebrates and invertebrates – and 61 bio-realms (unique combinations of biomes and biogeographic realms, as per WWF’s ecoregionalisation; http://www.worldwildlife.org/biomes). Full details of these models are provided in Hoskins et al (2020).

The aggregate PARC-representativeness score for any larger spatial reporting unit (e.g. a country) is derived as a weighted geometric mean of the individual scores of all cells (both protected and unprotected) falling within that unit:

where the contribution of each cell is weighted by the predicted overlap in species composition between this cell and all other cells, and therefore its ecological uniqueness:

The underpinning logic, and details, of this approach are described further by Ferrier et al (2004), Allnutt et al (2008) and Williams et al (2016).

PARC-connectedness

The existing global implementation of PARC-connectedness generates results using two spatial grids covering the entire terrestrial surface of the planet at the same resolution as for PARC-representativeness. The first grid records the proportion of each cell included within protected areas – i.e. 0 if none of the cell is protected, 1 if all of the cell is protected, and a proportion between 0 and 1 if only part of the cell is protected. The second grid records the proportion of each cell estimated to be covered by primary vegetation (see ‘Data sources’ below for details of this layer). As noted above, under ‘Definition’, the involvement of primary-vegetation data should be considered as optional for application of PARC-connectedness to post-2020 GBF reporting – i.e. if need be, the index can be readily generated to reflect connectedness between protected cells alone, without any consideration of connectedness to unprotected primary vegetation.

The connectedness of each protected cell (value > 0 in the protected-area grid) to other protected cells, and (if considered relevant) to cells containing primary vegetation in the surrounding non-protected landscape, is scored using the cost-benefit analysis (CBA) technique described by Drielsma et al (2007). This raster-based technique, developed originally to analyse the connectedness of habitat for individual species, is founded on well- established principles of meta-population ecology.

To derive PARC-connectedness using the CBA technique each cell within a 500km radius (consistent with Santini et al 2015) of the protected cell of interest is first assigned a ‘benefit’ value, representing the proportion of that cell included in protected areas or covered by primary vegetation (outside protected areas). Each cell in the surrounding landscape is also assigned a ‘cost’ value, indicating permeability to dispersal through that cell, scaled from 0.1 for cells with no protection or primary vegetation through to 1.0 for cells fully protected or covered by primary vegetation.

Using the CBA technique, the connectedness of each protected cell is calculated as a weighted sum of the benefit values of all cells in the surrounding landscape, with the contribution of each cell weighted by the probability of dispersal associated with the least-cost path between that cell and the protected cell of interest. This probability of dispersal is a function of the permeability values of cells along the least-cost path and an estimated median-dispersal parameter, assuming a negative-exponential relationship between distance and dispersal probability. Four different median-dispersal values are used – 200m, 2km, 20km, and 200km – based loosely on the distribution of estimated dispersal capabilities of mammals reported by Santini et al (2015).

Connectedness is therefore calculated four times for each protected cell and these results are then averaged to yield a single weighted sum for the cell. Least-cost path calculations are approximated by adapting the ‘petals’ approach advocated by Drielsma et al (2007) to use a radial geometric stratification (Harwood et al 2022), thereby greatly reducing computation time.

The resulting weighted sum for each protected cell is expressed as a proportion of the maximum possible sum if that cell were surrounding by a continuous expanse of protected cells within the 500km radius, thereby yielding a connectedness score for that cell between 0 and 1. PARC-connectedness for any given spatial reporting unit (e.g. IPBES region, country) is then derived by averaging these scores across all protected cells within the unit, thereby expressing overall connectedness as a proportion (also ranging between 0 and 1).

Composite PARC indicator

The composite PARC indicator is derived by using PARC-connectedness scores for protected cells to further weight the contribution that each of these cells makes to the assessment of protected-area representativeness. Building on the calculation of PARCrepri from above, the composite PARC score for cell i is calculated as:

Where PARCconnj is the PARC-connectedness score of protected cell j (between 0 and 1).

Further building on the calculation of PARCreprreporting_unit from above, the aggregate score for the composite PARC indicator across any larger spatial reporting unit (e.g. a country) is then:

5c. Data Collection Method

All data used to derive the PARC indices are described below under ’Data sources’.

5d. Accessibility Of Methodology

All methods and underlying data used to derive the global implementation of the PARC indices are published in accessible peer-reviewed locations (see ‘Method of computation’ and ‘Data sources’ for details).

Options for deriving and reporting the PARC indices at national scale are described below under ‘National/regional indicator production’.

5e. Data Sources

The existing GDM models of ecological similarity used to derive the global implementation of PARC-indicator were fitted using the following input data (see Hoskins et al 2020 for details):

• Global 30-arcsecond gridded environmental surfaces for: Min Monthly Min Temperature, Max Monthly Max Temperature, Max Diurnal Temperature Range, Annual Precipitation, Actual Evaporation, Potential Evaporation, Min Monthly Water Deficit, Max Monthly Water Deficit, Soil pH, Soil Clay Proportion, Soil Silt Proportion, Soil Bulk Density, Soil Depth, Ruggedness Index, Topographic Wetness Index (Sources: http://www.worldclim.org/ https://www.soilgrids.org/ http://www.earthenv.org/).
• Global occurrence records for all terrestrial species within the following taxa, extracted from data accessible through the Global Biodiversity Information Facility (GBIF http://www.gbif.org/): vascular plants, amphibians, reptiles, birds, mammals, ants, bees, beetles, bugs, butterflies, centipedes, dragonflies, flies, grasshoppers, millipedes, snails, moths, spiders, termites, wasps.

For each year of interest, a 30-arcsecond grid of protected-area coverage is derived from all protected-area boundaries included in the World Database on Protected Areas (https://www.protectedplanet.net/). This grid records the proportion of each grid-cell included within protected areas – i.e. 0 if none of the cell is protected, 1 if all of the cell is protected, and a proportion between 0 and 1 if only part of the cell is protected. Any protected area for which only a centroid and areal extent are provided (rather than an explicit boundary) is assumed to be circular in shape. As noted above, under ‘Definition’, future derivation of the PARC indices can be readily extended to include data from the World Database on OECMs.

The grid of the proportion of each 30-arcsecond cell estimated to be covered by primary vegetation in any given year, is derived from a global time-series of land-use change generated by CSIRO using statistical downscaling of coarse-resolution land-use data using finer-resolution covariates, including remotely-sensed land cover and abiotic environmental attributes. This was achieved using the technique described by Hoskins et al (2016), extended by employing Version 2, in place of Version 1, of the Land Use Harmonisation product https://luh.umd.edu/, MODIS Vegetation Continuous Fields https://modis.gsfc.nasa.gov/data/dataprod/mod44.php as remote-sensing covariates in place of discrete land-cover classes, and constraints derived from ESA CCI

land-cover mapping https://www.esa-landcover-cci.org/ (see also Di Marco et al 2019).

5f. Availability And Release Calendar

Global implementation available now.

Raw gridded results for PARC-representativeness and PARC-connectedness, for the entire planet, are accessible via the CSIRO Data Access Portal https://data.csiro.au/ and the UN Biodiversity Lab https://unbiodiversitylab.org/.

Aggregated country-level results for PARC-representativeness and PARC-connectedness are accessible via the Biodiversity Indicators Partnership Dashboard https://bipdashboard.natureserve.org/.

5g. Time Series

Time series available: 1970, 1980, 1990, 2000, 2010, 2012, 2014, 2016, 2018, 2020.

Next planned update: 2022.

5h. Data Providers

Organisations producing data used to derive the PARC indices are listed under ‘Data sources’.

5i. Data Compilers

CSIRO (Australia’s national science agency), in partnership with GEO BON and GBIF.

5j. Gaps In Data Coverage

The global implementation of the PARC indices assesses only terrestrial protected areas. However, good potential exists to extend this coverage to the marine and freshwater realms, especially given that the GDM modelling paradigm, which underpins the PARC-representativeness approach, is increasingly being applied effectively at regional scale in both of these realms.

5k. Treatment Of Missing Values

All spatial data layers used to derive the PARC indices cover the entire terrestrial surface of the planet at 30- arcsecond grid resolution (see ‘Data sources’).

6. Scale

6a. Scale Of Use

Scale of application (please check all relevant boxes): Global, Regional, National

Scale of data disaggregation/aggregation: The existing global implementation is derived from data and models covering the entire terrestrial surface of the planet at 30-arcsecond grid resolution (these cells are approximately 900m wide at the equator).

Global/ regional scale indicator can be disaggregated to national level: Yes

6b. National Regional Indicator Production

Three main options exist, or are anticipated, for countries to report national-level changes in the PARC indices. These options differ both in terms of ease of implementation, and in terms of the extent to which they mobilise national versus global data.

Option 1: In this option countries would simply extract the raw gridded PARC results (at 30-arcsecond grid resolution) for their country from the relevant globally-generated layers (for details of access to these results see ‘Availability and release calendar’ above). They could then use standard GIS processing to aggregate and report results by any desired set of spatial units – e.g. provinces, ecosystem types. This is the easiest option to implement, but results would be totally dependent on the quality of the global inputs employed, and would not benefit from incorporation of any better-quality national data.

Option 2: Countries would here make use of the existing global GDM modelling of spatial variation in species composition (the most challenging, and computationally demanding, component of the indicator’s workflow) but would replace the globally-generated data on protected areas and OECMs with best-available national data. This option would also open up potential for countries to evaluate the contribution that alternative additions to the system of protected areas and OECMs might make to improving the present PARC score for their country, thereby providing a foundation for prioritising such additions. The rigour of results generated through this option would, however, still be constrained to some extent by the spatial resolution, and quality, of the global biodiversity modelling employed. CSIRO is currently exploring, with partner organisations, potential avenues for giving countries direct access to the analytical computing capability needed to implement this option.

Option 3: In this most demanding option countries would derive the PARC indices for their country from scratch, not only employing best-available national data on protected areas and OECMs, but also making use of best- available biological and environmental data to refine the modelling of spatial variation in species composition – potentially at a finer spatial resolution than that employed in the global implementation (see Williams et al 2016 for an example of this option implemented for Australia). As for Option 2, CSIRO is exploring potential avenues for giving countries direct access to the analytical computing capability needed to implement this option.

6c. Sources Of Differences Between Global And National Figures

Some modest differences between country-level results generated using the three options outlined above are to be expected, as a function of differences in the resolution and quality global versus national sources of data.

6d. Regional And Global Estimates And Data Collection For Global Monitoring

6d.1 Description Of The Methodology

Regional and global aggregates for the PARC indices should ideally be derived directly from raw gridded results across the region concerned, using the rigorous aggregation procedure described under ‘Method of computation’, rather than from pre-aggregated country values.

6d.2 Additional Methodological Details

6d.3 Description Of The Mechanism For Collecting Data From Countries

7. Other MEA And Processes And Organisations

7a. Other MEA And Processes

CBD Aichi Target 11 reporting.

IPBES Core Indicator, used in the IPBES Global Assessment (Chapter 3) and Regional Assessments.

7b. Biodiversity Indicator Partnership

Yes

10. Data Reporter

10a. Organisation

CSIRO (Australia’s national science agency)

10b. Contact Person

Simon Ferrier: simon.ferrier@csiro.au