Target 2: Enhancing carbon drawdown and storage through additional CSAs

We identified currently unprotected high-carbon areas that must be conserved to meet global climate targets. A by-product of conserving areas high in biodiversity value is that most, but not all such areas, also store the most carbon (Fig. 2). In ecoregions where the median total carbon density is above 215 MT/ha, a total of 29,247,979 km² of terrestrial area storing 1,331,834 megatons of carbon require increased conservation attention for carbon storage. Ninety-two percent of this area is already captured in target 1 (Fig. 2 and table S2), underlining the interdependence of carbon and biodiversity and the importance of these lands to achieve the dual goals of biodiversity conservation and climate stabilization. To bridge the gap for adequate carbon storage beyond areas identified in target 1, the remaining 2,337,246 km² or 1.7% of Earth’s land surface was selected as tier 1 CSAs in target 2 (Fig. 2).

Open in a separate window

Fig. 2

Interdependence of carbon and biodiversity.

Currently unprotected high-carbon areas with median total carbon >215 MT/ha overlap extensively (92.0%) with areas selected under target 1, highlighting the importance of these lands for biodiversity conservation and climate stabilization. Other areas important for biodiversity but of lower carbon value, i.e., <215 MT/ha, are also shown. Additional CSAs, including tier 1 and tier 2 CSAs, are also selected to bridge the gap for adequate carbon storage beyond areas identified in target 1.

In addition, we identified 3,946,581 km² of unprotected land, or 2.9% of Earth’s surface, as tier 2 CSAs. Together, tiers 1 and 2 CSAs add 6,283,826 km² of currently unprotected lands, or 4.7% of global land area, to the Global Safety Net. These land areas store an estimated 131,000 megatons of carbon (Table 1). Indigenous lands (26) contribute extensively to carbon storage. Greater than 74% of all mapped indigenous lands (28,123,013 km²) are tier 1 or tier 2 CSAs, and together, these areas store >931,000 megatons of carbon biomass.

Combined targets

Together, the two targets described above and currently protected areas that form the Global Safety Net cover 50.4% of the terrestrial realm as regions to enhance biodiversity protection and carbon storage (Table 1 and Fig. 1). Approximately 34% of the area in targets 1 and 2 is indigenous land (26). The overlap is particularly pronounced in high β-diversity ecoregions (41%; cluster 2), cluster 3 rare phenomena (49%), and cluster 4 intactness (33%). This includes many ecoregions in the tundra, boreal, tropical forests, and xeric biomes (table S1). We map results on a finer scale across five biogeographic realms—Neotropic, Nearctic, Afrotropic, Palearctic, and Indo-Malayan (Fig. 3, A to D). All mapped layers are available for online viewing at www.globalsafetynet.app.

Open in a separate window

Fig. 3

The Global Safety Net made more visible in a close-up of five biogeographic realms.

Shown here are Neotropic (A), Nearctic (B), Afrotropic (C), and Palearctic and Indo-Malayan (D) (adjacent realms partly included). Existing protected areas are expanded to account for additional lands requiring increased conservation attention (target 1), augmented by additional CSAs (target 2), and connected by potential wildlife and climate corridors (target 3). Numbers in parentheses show the percentage of total land area of Earth contributed by each set of layers. To explore the component terrestrial layers of the Global Safety Net, please visit www.globalsafetynet.app. Indigenous lands are not shown but overlap extensively with proposed areas for increased conservation attention (see table S2 for ecoregions depicted in Fig. 3).

Restoration

One overlooked area of research that should inform future iterations of the Global Safety Net is the restoration opportunities on degraded lands (30). These degraded landscapes could be restored to address both climate and biodiversity concerns. Further, reconnecting forest corridors in degraded lands could offset emissions that will occur before a moratorium on land-based emissions is reached. Focusing restoration efforts on degraded lands that can serve as wildlife corridors could help achieve other objectives, such as the Bonn Challenge (31). Similarly, massive tree-planting programs, if designed using native species and planted to restore corridors, riparian and coastal vegetation, and upper watersheds, could contribute to stabilizing climate and restoring connectivity.

Major opportunities exist for restoration of forests using native plants. Ecoregions such as the Atlantic Forest of Brazil, several forest ecoregions in Madagascar, and the Western Ghats of India are currently underrepresented in this version of the Global Safety Net, which is focused on protection of remaining habitat. Restoration opportunities should drive future iterations and allow for monitoring of recovery efforts. A prime example is the mid-elevation forests of Nepal, previously one of the more deforested and degraded ecoregions, where intensive community forestry programs have led to nearly doubling forest cover in 24 years (32), increasing carbon stocks from 213.42 to 502.03 megatons.

Indigenous lands

The overlay of mapped indigenous territories with spatial targets 1 to 3 reveals an extensive overlap of 37% and underscores the central role that indigenous peoples and their lands play to preserve biodiversity and regulate Earth’s atmosphere (26, 33, 34). Another observation is echoed by other conservation biologists who have examined maps of indigenous lands and global biodiversity priorities: A 30% area–based target for protection by 2030, as advocated by many groups to the Convention on Biological Diversity, effectively already exists when accounting for indigenous lands, should effectively conserved lands be formally acknowledged by governments as other area-based effective conservation measures (OECMs) (35). In short, the “30 × 30” target is far less ambitious when viewed through this perspective. Many conservation organizations, indigenous peoples, and local communities have called for an area-based target of “at least 50%” under the Convention on Biological Diversity. Explicit in these calls is to allow for the protection of the land rights and traditional management practices of communities most at risk to food insecurity, the negative impacts of land degradation, and climate change.

Mapping the elements

The current version of the Global Safety Net is formulated from 11 biodiversity layers (fig. S1, A to K, and table S1). We partitioned two of the above datasets to calculate a median pixel values: IUCN range-size rarity raster (median = 0.006) (17) and small-range vertebrates raster (median = 24) (15). For both datasets, only pixels greater than or equal to the median values were used. In the case of rare plant species, to be conservative, we excluded pixels containing only one to two rare plant species. The rationale here is that some of these are known from one to a few specimens. All raster data were converted to vector data (polygon) for further analysis.

We overlaid each of these biodiversity data layers with all terrestrial protected areas (14) to remove areas already set aside for conservation. To remove double counting, we subtracted any overlapping areas with previous datasets. For example, all AZEs are included as KBAs. We ingested resulting layers into the Google Earth Engine to derive remaining habitat in each layer using percent tree-cover maps (42) in forested ecoregions (except boreal forests) and excluded globally significant patterns of human land use and populations (“anthromes”) in nonforested ecoregions (43) [see (1) for detailed methods]. We selected all nonoverlapping unprotected areas within each of layers 1 to 4 and only the remaining habitat for layers 5 and 6 as contributions toward target 1. For layers 1 to 5 within “species rarity,” we added a 1-km buffer around all unprotected sites except layer 6, rare plant species, as the size of a “rare plant pixel” was ~10,000 km².

To estimate carbon storage potential by biodiversity layer to construct (Fig. 2, Table 1, and table S1), we first overlaid a map of total carbon biomass (24)—which includes above ground, below ground, and soil carbon—with terrestrial ecoregion boundaries (1) to derive the median carbon density for each ecoregion. To determine CSAs, we selected ecoregions with a median total carbon density >215 MT/ha as candidates for tier 1 CSAs. Ecoregions with a median total carbon density between 50 and 215 MT/ha were designated as tier 2 CSA candidates. Ecoregions with low levels of carbon density (<50 MT of total carbon per hectare) were not selected as potential sites for additional CSAs. We then selected all remaining habitat outside protected areas after removing any overlap with the 11 biodiversity layers to derive the polygons for tier 1 and tier 2 CSAs.

On the basis of the best available literature, we designed wildlife corridors to meet the ecological requirements of the most wide-ranging species that must disperse as part of their life histories and climate corridors that would allow species movement up and down mountainsides, along riparian corridors, or across human-dominated landscapes (44). The connectivity analysis was a computationally intensive analysis that included producing a cost-distance matrix, weighing land cover classifications, buffering, and processing. The cost-distance matrix surface was developed as a surface intended to represent varying levels of resistance for wildlife to move along a landscape with regard to vegetation cover, slope, roads, and other land uses. While future iterations should be more specific to ecoregions and local fauna, for this first global scoping phase, we used continents as the unit of analysis and corridors were modeled considering variables that are potentially important for the gene flow of terrestrial species generally. We weighted both variables and classes, depending on the type of data, so that higher weights were given for factors that have higher costs. Land cover data were obtained from the European Spatial Agency with a spatial resolution of 300 m and was reclassified considering the degree of anthropized areas. Urban areas and water bodies were excluded from the modeling. Roads, railways, and mining areas were buffered. The design of the corridor network and the links between core areas was done with the Linkage Mapper Toolkit of the Circuitscape project (www.circuitscape.org). A full description of the methods is available from the authors.

Sources of variation

Here, we identify five potential sources of variation in our results that could be improved in future iterations of the Global Safety Net. We also point to how variants in methods or data sources differ from other, recent efforts to map global biodiversity (3, 15).

1) Total areal extent of the terrestrial realm

The total land surface we used to produce the Global Safety Net is based on Earth’s entire terrestrial area excluding Antarctica, which amounts to 134,126,000 km². Much of Antarctica includes rock and ice, and the 18 tundra ecoregions on the continent do not contribute to the key targets of the Global Safety Net. The total land area calculated by this method is closely comparable to that adapted by the World Database on Protected Areas from which they derive the 15.1% terrestrial coverage that is the standard used in other biodiversity analyses in preparation for the Convention on Biological Diversity. As a result, the Global Safety Net does differ from other studies—Allan et al. (3)—that use a larger total terrestrial area estimate of about 146,000,000 km². The 44% of Earth’s terrestrial area that Allan et al. (3) call for increased conservation attention amounts to 64 million km². In contrast, the 45.7% of Earth’s terrestrial area included under Global Safety Net for currently protected areas and target 1 totals 61.3 million km².

2) Potential but limited error of inclusion of nonhabitat from cluster 1 datasets

When applying layers 1 to 4 of cluster 1, we used the original polygons provided by the authors of each dataset. As a result, these four data layers include varying amounts of nonhabitat within each selected polygon. We did not apply habitat suitability modeling to refine these datasets as it would further fragment these critical areas for narrow-range and rare species, which could have detrimental effects for biodiversity conservation, especially where some of these adjacent nonhabitat areas are prime candidate for restoration or reconnecting via wildlife corridors. For those reasons, we used the original polygons, including nonhabitat areas, in our analysis for cluster 1. The inclusion of nonhabitat areas is essentially moot because of its limited spatial extent: Summing the entire area of nonhabitat from layers 1 to 4 adds only 388,089 km² (or 0.3% of the total land surface of Earth to the 2.3% selected for cluster 1). We suggest removal of nonhabitat is best performed at the ecoregion scale by local experts and done on a case-by-case basis.

3) Remaining habitats in layers 5 to 11

A goal for the Global Safety Net is to identify near-term opportunities to achieve the global nature conservation target, i.e., areas where additional protection can have the most effective conservation outcome. Thus, we focused on suitable natural habitat remaining without the need for major restoration. We therefore selected only intact or semi-intact habitat remaining outside currently protected areas to derive potential contributions to the Global Safety Net from layers 5 to 11 (clusters 1 to 4) and from additional areas for carbon storage (CSAs).

4) Indigenous lands

We overlaid the Global Safety Net with the most recent global map of lands managed or controlled by indigenous peoples (26) to determine the extent to which such lands overlap with the existing network of protected areas. The intent was to illustrate the role such lands could have in enhancing biodiversity protection and carbon storage if this were the intention of peoples managing such areas. Two sources of variation are noted: (i) Many indigenous peoples’ lands remain unmapped. Blank areas merely indicate that no publicly available datasets currently indicate the presence of indigenous peoples from those areas; the map, however, should not imply absence of indigenous peoples. (ii) The scale at which indigenous lands are mapped in is based on multiple public datasets varying greatly in spatial resolution (26). For example, polygons in the Sahara and in the tundra ecoregions are much larger and more coarse grained than those mapped within the United States and the Brazilian Amazon.

We would like to thank D. Basiero, N. Burgess, S. Butchart, S. Garnett, C. Jenkins, R. Noss, S. Pimm, A. Plumptre, and two anonymous reviewers for comments on the manuscript. We thank the KBA Secretariat and UNEP-WCMC for sharing several data layers used in the analysis. Funding: This work was supported by a grant from One Earth. Author contributions: E.D., A.R.J., M.F., K.B., D.O., and C.V. designed the study. E.D. and C.V. drafted the manuscript. C.V. helped obtain data layers and permissions from authors. A.J. and M.F. performed the spatial analysis. A.T.L.L. and S.F. processed analysis output, generated the tables, and contributed to the Materials and Methods section. F.P.-D. generated the figures. G.P.A. and T.B. contributed to β-diversity and carbon sections and methods and editing. T.B., C.V., K.B., and A.R.J. helped create the website to serve as repository for data and also to rerun analyses on Google Earth Engine. All authors participated in manuscript and figure editing. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper are available at www.globalsafetynet.app concurrent with publication and may be requested from the authors.