Retaining natural vegetation to safeguard biodiversity and humanity

Global efforts to deliver internationally agreed goals to reduce carbon emissions, halt biodiversity loss, and retain essential ecosystem services have been poorly integrated. These goals rely in part on preserving natural (e.g., native, largely unmodified) and seminatural (e.g., low intensity or sustainable human use) forests, woodlands, and grasslands. To show how to unify these goals, we empirically derived spatially explicit, quantitative, area‐based targets for the retention of natural and seminatural (e.g., native) terrestrial vegetation worldwide. We used a 250‐m‐resolution map of natural and seminatural vegetation cover and, from this, selected areas identified under different international agreements as being important for achieving global biodiversity, carbon, soil, and water targets. At least 67 million km2 of Earth's terrestrial vegetation (∼79% of the area of vegetation remaining) required retention to contribute to biodiversity, climate, soil, and freshwater conservation objectives under 4 United Nations’ resolutions. This equates to retaining natural and seminatural vegetation across at least 50% of the total terrestrial (excluding Antarctica) surface of Earth. Retention efforts could contribute to multiple goals simultaneously, especially where natural and seminatural vegetation can be managed to achieve cobenefits for biodiversity, carbon storage, and ecosystem service provision. Such management can and should co‐occur and be driven by people who live in and rely on places where natural and sustainably managed vegetation remains in situ and must be complemented by restoration and appropriate management of more human‐modified environments if global goals are to be realized.


INTRODUCTION
Despite the dependence of humanity on the natural world, anthropogenic erosion of nature, often irreversible (IPBES, 2019), continues. Although commitments to sustainability abound and are at the core of international agreements, humans continue to degrade and destroy ecosystems at unsustainable rates worldwide (Mackey et al., 2015;Watson et al., 2016). A clear understanding of how much further biodiversity loss can occur before permanent, irrevocable damage is wrought on the biosphere is lacking . However, evidence suggests we are at imminent risk of breaching (or have already exceeded) certain critical planetary thresholds (Steffen et al., 2015). Given that the pressures on the biosphere are accelerating as the human population grows and consumption intensifies (Leclère et al., 2020;Wiedmann et al., 2020), it is important to identify a limit to the loss of nature to prevent further, irreparable damage. Many goals enshrined in multiple global (e.g., United Nations) agreements are implicitly underpinned by, among other actions, retaining and sustainably managing existing natural and seminatural terrestrial vegetation . However, the targets underpinning the achievement of these goals rarely articulate a desired, measurable outcome state (i.e., the amount of natural and seminatural vegetation, where it occurs, and how to manage it) that is required for biodiversity and sustainability objectives to be achieved Elder & Olsen, 2019). Identifying this state provides an opportunity to develop a unified approach to achieving the goals. Dialogue relating to the proposed Post-2020 Global Biodiversity Framework under the Convention on Biological Diversity (CBD) conveys an important shift; the emphasis is on net outcomes for biodiversity: net improvements in ecosystems by 2050 delivered via an interim milestone to increase the area, connectivity, and integrity of natural ecosystems by at least 5% by 2030 (Secretariat of the Convention on Biological Diversity, 2021). In large part, this goal's achievement, and that of the overarching vision of the CBD-to be living in harmony with nature by 2050-will require limiting further losses of terrestrial vegetation. Although restoration will play an important role, and is rightly the focus of significant attention (e.g., the United Nations Decade on Ecosystem Restoration 2021−2030), the retention of the vast majority of remaining vegetation is likely required (Fastre et al., 2021). This is not least because future gains from restoration may take decades to accrue or may have antagonistic outcomes for different environmental imperatives (e.g., biodiversity and carbon goals) (Goldstein et al., 2020).
Several authors argue that approximately 50% of the planet requires dedicated conservation attention, delivered via a range of actions including area-based protection and retention of existing vegetation (Dinerstein et al., 2020;Jung et al., 2021;Locke, 2014;Locke et al., 2021;Noss et al., 2012;Riggio et al., 2020;Wilson, 2016). Where this 50% is, why it warrants conservation attention, and how this should be delivered vary among the studies-a reflection of their different approaches and intents. Together with the opportunities presented by a move toward ambitious net outcomes-based goals (e.g., increased ecosystem extent and condition), as called for under the proposed Post-2020 Global Biodiversity Framework ; Secretariat of the Convention on Biological Diversity, 2021), integrated delivery of various international agreements underscores the key role that the retention of terrestrial vegetation could play in safeguarding natural assets globally. It follows that identification of retention targets that set out how much natural and seminatural vegetation is needed and where is one way to help unite nature conservation and sustainable development goals across the international environmental agenda . However, to date, no global effort to identify areas for retention has been undertaken.
We set out an analytical approach for translating globally agreed goals to spatially explicit guidance on how much and where existing natural and seminatural terrestrial vegetation should be retained, if these goals are to be met. We aimed to develop an approximation of how much natural and seminatural terrestrial vegetation must be retained to underpin the nature-reliant ambitions of global agreements relating to biodiversity conservation, climate stabilization, soil maintenance, and water quality regulation. We examined mechanisms for the management of this vegetation considering that a combination of formal protection and sustainable use will be required to align environmental, social, and economic imperatives, rights, and responsibilities; international cooperation will be crucial for determining country-level contributions to global retention efforts; and considerable additional action will be needed to achieve global goals, including restoration, as well as sustainable management of marine and intensively human-modified biomes, although our focus was on existing natural and seminatural terrestrial vegetation. We sought to add to the literature and associated initiatives on how global environmental agreements can be delivered in a more integrated fashion (e.g., Dinerstein et al., 2020;Jung et al., 2021;Locke et al., 2021;Riggio et al., 2020;Wilson, 2016).

Global goals dependent on natural and seminatural terrestrial vegetation retention
There are 4 global agreements made under UN resolutions for which the retention of terrestrial vegetation is directly relevant: Convention on Biological Diversity (CBD), Framework Convention on Climate Change (UNFCCC), Convention to Combat Desertification (UNCCD), and Sustainable Development Goals (SDGs). These international agreements, each adopted and ratified by the majority of nations, contain clearly articulated statements about what is needed from nature and are the logical starting point to mapping out how much terrestrial vegetation should be retained to help achieve a sustainable future, thus allowing humans to live in harmony with nature by 2050 as per the vision of the CBD. Our focus on terrestrial vegetation is founded on the integral role vegetation in terrestrial ecosystems plays in supporting species, providing services, and regulating local to global climate and hydrological cycles (Ellison et al., 2017).
We examined the goals, objectives, targets, and indicators under each of the 4 agreements to identify statements about desired outcomes that specifically depend on (in whole or part) the retention of terrestrial vegetation. This formed the basis of translating explicit quantifiable milestones and aspirational goals in the respective agreements into terrestrial vegetation retention targets ( Figure 1) (details for each target described below and in Appendix S1). Hereafter, we refer to the retention targets corresponding to the 4 agreements as the biodiversity conservation, carbon storage, soil maintenance, and freshwater quality targets.

Quantifying natural and seminatural terrestrial vegetation retention requirements
In our spatial overlay analysis, we considered areas for retention characterized by mapped terrestrial vegetation. We focused on identifying land-cover types for which retention is most applicable to supporting the global agreements we examined (Figure 1; Appendix S1). We excluded areas that support some vegetation, but for which retention was less applicable vis à vis the global agreements to which the retention targets we examined are linked (e.g., urban areas, agricultural lands dominated by nonnatural land cover [e.g., sown crops, pasture with natural overstory removed]). We focused on natural and seminatural terrestrial vegetation, and in our maps selected land-cover types that included forests, woodlands, shrublands, savannas, and grasslands along a spectrum of conditions and human influences (Appendix S2). Much of the vegetation we captured may be under use or managed (e.g., commercial forests) or occur in cultural landscapes affected by thousands of years of human stewardship (see DISCUSSION).
Our analyses did not extend to marine systems. Nonetheless, our framework for translating internationally agreed goals to spatially explicit quantitative ecosystem retention targets could be transferable to the marine realm. Amenable data sets may include marine ecosystems (Spalding et al., 2007;Spalding et al., 2012), human pressures (Halpern et al., 2015), carbon export in oceans (Henson et al., 2012;Roshan & DeVries, 2017), and species ranges and protected area (data sets produced by IUCN, BirdLife International, UNEP-WCMC; taxa-specific data [Kaschner et al., 2011]).
For each of the 4 retention targets (biodiversity conservation, carbon storage, soil maintenance, and freshwater quality), we produced maps that captured the amount and location of existing natural and seminatural terrestrial vegetation that should be retained to contribute to the achievement of key elements of international environmental agreements. To represent terrestrial vegetation (hereafter, the natural and seminatural vegetation layer) in this analysis, we used the MODIS Land Cover Type product MCD12Q1 at 250 m (native data resolution = 500 m) as was processed and modified by Borrelli et al. (2017). This layer, covering approximately 84% of Earth's surface, is based on the International Geosphere Biosphere Programme system and reports 17 land-cover classes, including 10 terrestrial vegetation classes (Appendix S2). We excluded unvegetated and aquatic land-cover types (permanent wetlands [but see "Freshwater quality target" below], barren, snow or ice [including all of Antarctica], and water).
Croplands (landcover class consistent with in the UN Food and Agriculture Organization's FAOSTAT data in the map produced by Borrelli et al. [2017]) were excluded because they represent highly modified environments, and our analysis focused on natural and seminatural environments. For the same reason, for grasslands (category as per the Borrelli et al. [2017] map) we masked grazing lands for the year 2000 with a spatial data set that combines agricultural census data with satellite-derived land cover to map pasture extent (Ramankutty et al., 2008). In this data set, the grid values range from 0 to 1 (0, 0% of pixel under pasture; 1, 100% of pixel under pasture). We resampled this layer to 250 m and masked all the grassland pixels from the Borrelli et al. (2017) map that were overlaid by pixels with a pasture grid value >0.5.
The resultant map indicated that approximately 83.8 million km 2 of natural and seminatural terrestrial vegetation remained on Earth (approximately 62% of the non-Antarctic land surface); condition or level of degradation of the vegetation was not accounted for. To determine how much and where this vegetation needed to be retained to allow the global goals to be met, we overlaid existing mapping products linked to each of the 4 targets (described briefly below and in full detail in Appendix S1) with the natural and seminatural vegetation layer. This allowed us to calculate the amount and location of vegetation that occurs in places recognized as of importance for each of the targets. All spatial statistics were calculated using a Mollweide projection, and all mapping analyses were undertaken at a raster pixel resolution of approximately 250 m (i.e., the resolution of natural and seminatural vegetation layer). The workflow for deducing spatially explicit retention targets (e.g., maps, percentage targets), based on the methods and rationales detailed below, is in Figure 2.

Biodiversity conservation target
Language of the first draft of the Post-2020 Global Biodiversity Framework under the CBD (July 2021) focuses on preventing extinctions, recovering threatened species, expanding the protection and management of important sites, and maintaining or enhancing ecosystem extent and resilience (including connectivity and intactness) (Secretariat of the Convention on Biological Diversity, 2021). To translate these various aims to a biodiversity conservation retention target, we used the map produced by FIGURE 2 Workflow for deriving maps for each of the retention targets (biodiversity conservation, carbon storage, soil maintenance, freshwater quality) considered in this analysis (yellow boxes, input or derived data sets; blue boxes, Geographic Information System (GIS) operations; green boxes, outputs). Important areas for biodiversity combine key biodiversity areas, wilderness, protected areas, species distributions, and ecoregions as captured by Allan et al. (2022). All raster inputs were resampled to match the resolution of the natural and seminatural vegetation layer. Allan et al. (2022) to identify natural and seminatural terrestrial vegetation that should be retained for multiple biodiversity conservation objectives ( Figure 1). These maps displayed the spatial distribution and areal requirements needed to conserve 28,594 species and showed existing ecoregional representation targets and protected areas, key biodiversity areas, and all large, contiguous areas with low human pressure (wilderness) (Allan et al., 2022). Detailed methods including the input data used to produce this map and limitations on its interpretation are in Allan et al. (2022). We identified all vegetation from our natural and seminatural vegetation layer that overlapped this map to identify the extent and distribution of terrestrial vegetation that requires retention to contribute to the achievement of the biodiversity conservation objectives captured by Allan et al. (2022).

Carbon storage target
Forests play a key role in storing carbon because they hold 70-90% of terrestrial biomass (Houghton et al., 2009). Goodquality global data of their distribution exist, so they were the focus of this component of the analyses. To determine how much and where terrestrial forest should be retained to contribute to stabilizing levels of atmospheric carbon, we used spatial outputs from a harmonized land-use transition model for shared socioeconomic pathway (SSP) 1 . Spatial outputs for SSP1 are based on the IMAGE 3.0 integrated assessment model (Stehfest et al., 2014) and describe annual transitions between different land-use categories (e.g., primary forest, secondary forest, pasture, and cropland) from 1500 to 2100 at a spatial resolution of 0.5 × 0.5 degrees (approximately 25 km 2 at the equator) (details of data sets and rationale for their inclusion and analysis in Appendix S1). We chose SSP1 (of the 5 SSPs available [Riahi et al., 2017]) because the land-use transitions captured in this scenario have the greatest probability of limiting global warming to <2 • C above preindustrial levels (Hurtt et al., 2020). Using the above, we identified the percentage of primary forest in each grid cell that was retained in 2050 under SSP1. This provided an indication of forest retention required if forests are to effectively help stabilize levels of atmospheric carbon.
Once we identified the proportion of forested areas to be retained in each 25-km 2 cell, we overlapped this with treed natural and seminatural areas (forests, categories 1-5 from the map produced by Borrelli et al. [2017] that formed the basis of the natural and seminatural vegetation layer in our analyses [Appendix S1]). To allocate the spatial distribution of forest within each grid cell to be retained from this layer, we preferentially selected pixels starting with those with the most tree coverage based on the MOD44B continuous fields layer (an input into the MODIS-derived ecosystem layer we used [Appendix S1]). We continued this process until the area of forested pixels selected corresponded with the proportional retention value identified by the index of forest retention (to 2050) produced from SSP1 output at a 25-km 2 grid cell resolution.
Unlike the biodiversity, freshwater, and soil targets that spanned vegetation classes, our target for carbon retention focused only on primary forest. This meant that our representation of terrestrial vegetation required for carbon storage and sequestration is an underestimate, particularly in parts of the world where other (nonforest ecosystems) play a key role in carbon storage. Management (and retention) of other ecosystem types (e.g., wetlands, mangroves, grasslands) can greatly benefit global efforts to mitigate climate change (Goldstein et al., 2020). However, in addition to data set availability limitations, we focused on forested systems because they contain the majority of terrestrial aboveground carbon stocks on Earth (Goldstein et al., 2020) and the spatial resolution of the input data set (approximately 25 km 2 ) (van Vuuren et al., 2017) precluded assessment of some carbon-dense ecosystems that have narrow or patchy geographic distributions (e.g., some mangrove ecosystems).

Soil maintenance target
Embedded in the UN Convention to Combat Desertification and 2030 Agenda for Sustainable Development goals (Figure 1) is the need to avoid degradation of land-based natural capital (e.g., soil and site-specific biotic, hydrological, and geomorphological attributes). The land degradation neutrality (LDN) goal of the UNCCD provides the most explicit and quantitative statement of an outcome-focused target for soil maintenance. The LDN is also captured in SDG 15, target 15.3, and requires that there be no net loss of land-based natural capital relative to 2015 levels (2015 being the year in which the UNCCD adopted LDN and the SDGs were agreed) by 2030 (Cowie et al., 2018). The UNCCD acknowledges area-based conservation as an important tool in achieving LDN. Soil underpins maintenance of land-based natural capital; as it is degraded or lost, land is degraded and natural capital is inevitably lost too.
We focused on soil retention via maintenance of existing forest, woodland, shrubland, savanna, and grassland as a proxy for maintenance of land-based natural capital. Although several threats to land-based natural capital and soil quality are globally significant (salinization, acidification, nutrient depletion, contamination, waterlogging, etc.), for this analysis we focused on soil erosion because it constitutes an irreplaceable loss of soil resource that cannot reasonably be restored (Montanarella et al., 2016) and is the most immediate risk to land-based natural capital if vegetation is cleared. We set out to identify where irreversible soil loss may occur in response to vegetation removal. The UNCCD and UN SDG targets cannot be met if soils are eroded at a faster rate than they form, regardless of how well other threats to soil resources are managed.
Current global rates of soil loss by water erosion have been estimated by Borrelli et al. (2017) with the RUSLE (Revised Universal Soil Loss Equation)-based (Renard et al., 1997) modeling platform Global Soil Erosion Modelling (GloSEM). The RUSLE incorporates rainfall erosivity climate (R factor), erodibility of the soil (K factor), topography (L and S factor), and local farming systems and practices to predict the amount of soil lost (due to inter-rill and rill erosion processes) per unit area and time (t⋅ha −1 ⋅year −1 ) (Renard et al., 1997). Using data from Bor-relli et al. (2017), we modeled the likely rate of soil loss if mapped natural and seminatural terrestrial vegetation was cleared. Conversion to agriculture is the primary driver of clearing of natural and seminatural vegetation worldwide (Campbell et al., 2017;Gibbs et al., 2010). Hence, we focused on quantifying the difference in soil erosion rates in areas of natural and seminatural vegetation before and after conversion to agriculture to identify where this would result in unsustainable soil loss that threatened the UNCCD and UN SDG targets.
The possible new land use for 3252 subnational administrative units of 202 countries from the Global Administrative Unit Layers (United Nations Food and Agriculture Organization, 2015), after clearing for agriculture, was either grazing or cropping depending on the main land use in the region according to Ramankutty et al. (2008). We assumed that administrative units with <50% coverage of grazing would be converted to the dominant current crop for that unit, whereas areas with over 50% of coverage of grazing would be converted to grazed areas. This is based on the assumptions that the prevailing land use in an administrative unit reflects land suitability of the whole unit and that administrative units tend to expand coverage of existing land uses, rather than diversify land uses. This approach overlooks land suitability at a finer scale than administrative unit. However, consistent global data for land use and suitability were not available at a finer scale. The dominant crop per country was identified using data from the United Nations Food and Agriculture Organization (2015). Based on this land-cover change, we assigned C-factor values that measure the effect of cropping on the soil erosion process, which are premised on the mean values reported for each land-cover type identified by Borrelli et al. (2017). The lower the C factor, the lower the erosion rate. All grazing areas had the highest C factor values (0.5).
Sustainable or tolerable soil erosion rates (Verheijen et al., 2009) can be defined as rates of soil erosion equivalent to rates of soil formation, consisting of mineral weathering and dust deposition. Soil formation rates vary widely worldwide, depending on climate, geology, topography, and other factors (García-Ruiz et al., 2015). No global map of soil formation rates exists, so we applied a conservative approach that likely overestimates soil formation rates in many locations and therefore underestimates the risk vegetation clearing poses to soil erosion and LDN. A global average soil formation rate of 2 t⋅ha −1 ⋅year −1 was selected. This is an upper estimate of soil formation rates across the globe, based on calculation of rock weathering rates from exported water chemistry globally across a range of lithologies, climates, and biota (Alexander, 1988;Panagos et al., 2015;Renard et al., 1997;Watusuki & Rasyidin, 1992). Using our map of areas where erosion rates were predicted to exceed 2 t⋅ha −1 ⋅year −1 if vegetation was replaced by agricultural land use, we identified areas from our natural and seminatural vegetation layer that should be retained to avoid unsustainable erosion and thus contribute to LDN goals. This upper estimate of soil formation rates means that our approach presents a best-case scenario of the minimum amount of vegetation retention required to avoid net soil erosion after vegetation clearing.

Freshwater quality target
We produced a map of natural and seminatural terrestrial vegetation that should be retained to contribute to freshwater quality maintenance (e.g., via in situ and catchment-level processes associated with filtration of impurities, reduction of sedimentation and pollutant run-off, etc.). The SDGs do not contain an outcomes-based (quantifiable) goal relating to freshwater quality maintenance. Therefore, we assumed that keeping all vegetation in situ, where this is likely to influence the quality of water in or entering water courses and wetlands, was a sound starting point for translation to a quantifiable goal. To translate this to a map, we used 3 different spatial data sets depicting semiaquatic (vegetated) ecosystems, areas of natural freshwater importance globally, and river basins that contribute disproportionately to global freshwater discharge. First, we used the Global Lakes and Wetlands Database 3 (GLWD) (30 Arc second resolution) (Lehner & Döll, 2004). This data set identifies areas that are predominantly water (e.g., lakes, rivers, marshes, flooded forests) and those that are partially water (e.g., intermittent lakes and wetland complexes). We included classes 1-10 from the GLWD 3, but excluded classes 11 and 12 (these cover <50% of the 30 arcsec raster grid on which the GLWD 3 is based (Lehner & Döll, 2004). To refine our analysis to natural and seminatural vegetation associated with (overlapping) large semiaquatic systems, we did not capture these small (<0.5 km 2 ) ephemeral wetland complexes. Some such vegetation may have been captured in the complementary river analysis described below. We overlapped the layer produced using the GLWD 3 (classes 1-10) with our natural and seminatural vegetation map to identify terrestrial vegetated areas congruent with mapped wetlands. We undertook the same process for mapped (nonestuarine or marine) Ramsar sites, whereby terrestrial vegetation overlapped by Ramsar sites was selected.
To complement this wetlands component of our freshwater quality target, we also used the HydroATLAS database developed by Linke et al. (2019) to identify river basins responsible for the majority of the planet's freshwater discharge. HydroAT-LAS captures 12 nested levels of subbasins at the global scale, each depicting consistently sized subbasin polygons at scales ranging from millions (level 1) to tens of square kilometers (level 12). We used level-10 basins to identify basins collectively responsible for 95% of global freshwater discharge (approximately 75,000 individual basins). This allowed us to capture the vast majority of freshwater discharge, mapped at a scale (level-10 basins) commensurate with the resolution of this global, multiple-data-set analysis. This is an arbitrary distinction, but sufficient to identify riparian vegetation mapped at a regional scale. We then selected natural and seminatural terrestrial vegetation in these basins under the rationale that this vegetation plays a role in contributing to the regulation of the quality of water being discharged. To represent the amount and distribution of natural and seminatural vegetation to be retained to contribute to global freshwater quality, we combined the areas of vegetation selected for the GLWD overlay, Ramsar sites, and river basin analyses (there is some overlap among these 3).

Production of vegetation retention map and percentage target calculations
We produced 4 maps that showed the extent of natural and seminatural vegetation that should be retained to contribute to biodiversity conservation, carbon storage, soil maintenance, and freshwater quality imperatives under international agreements (workflow in Figure 2). On each map, we also showed natural and seminatural vegetation that was not captured by the respective retention targets. The maps allowed us to calculate the percentage of vegetation requiring retention compared with the total extent of natural and seminatural terrestrial vegetation remaining in 2012 and the terrestrial surface of the planet (excluding Antarctica). We combined the 4 maps into a single global map to calculate an overarching retention target, that is, the total amount (and percent remaining) of natural and seminatural terrestrial vegetation that needs to be retained to contribute to key elements of the CBD, UNFCCC, UNCCD, and SDGs. To translate these findings to discrete units, we broke down natural and seminatural vegetation retention values by country.

RESULTS
We found that approximately 67 million km 2 (79%) of Earth's remaining natural and seminatural terrestrial vegetation (per our input mapping layer), covering approximately half the terrestrial (excluding Antarctica) surface of the planet, should be retained to help achieve key components of internationally agreed goals associated with biodiversity conservation, carbon storage, soil maintenance, and freshwater quality (Table 1; Figure 3). This value represents the overarching retention target for natural and seminatural terrestrial vegetation-the amount needed for coachievement of various environmentally based goals that are fundamental elements of 4 international agreements. This target-which should be considered based on the limitations of the input data we examined-establishes (broadly) where retention of natural and seminatural terrestrial vegetation is necessary to reduce the risk of compromising 1 or more of the environmental goals enshrined in the agreements that nations have committed to. Full achievement of these goals relies on many other factors in addition to this retention, including appropriate management of marine and more human-dominated biomes. The spatial (raster) data of vegetation retention that were produced for each of the 4 targets and that were used to produce Figures 3 and 4 and the summary data presented in Table 1 are available from the Dryad Digital Repository (https://datadryad.org/stash/share/ rqPq2g_WsbrAx6nVB-FUxYe7-rL0L3cJOWb0QY5DGtU).
Biodiversity conservation required the largest extent of natural and seminatural vegetation retention (Table 1; Figure 4), with over 43 million km 2 (51% of remaining vegetation) identified as required to contribute to this goal with its various objectives (species persistence; ecosystem representation; securing important sites, such as existing KBAs; retention of contiguous areas of low human industrialized pressure [ecologically intact areas]). Ecologically intact areas were responsible for the largest areal component of the total. Achieving soil maintenance goals required the next greatest area of natural and seminatural vegetation to be retained, followed by carbon storage, and freshwater quality (Table 1). Soil maintenance relied on vegetation retention throughout much of the tropics and subtropics, whereas freshwater quality maintenance relied on retention of riparian vegetation throughout all major river basins, so its contribution to the overarching retention target was widely dispersed globally. Carbon storage depended on retention of extensive areas of tropical and boreal forests. There was relatively low overlap among the spatial distribution of retained natural and seminatural vegetation satisfying multiple goals-approximately 27% of retained vegetation contributed to 2 targets, whereas only 2% contributed to all 4 targets (Table 1). Areas important for at least 3 targets were concentrated extensively throughout the Amazon and Congo basins and the islands of New Guinea and Borneo (Figure 3). There were also notable concentrations of areas important for at least 3 targets scattered throughout central America, West Africa, and Southeast and eastern Asia (Figure 3). Both ecologically intact areas with low human pressure (north Asia and far north North America) and areas under intense pressure (eastern Madagascar, Brazil's Atlantic coast, and Sri Lanka) contributed natural and seminatural vegetation essential for multiple targets (Figure 3).
Individual countries differed markedly in their retention requirements (Figures 4a-d & 5; Appendix S3). The area for retention required to meet the overarching target of 79% was disproportionately shared among several large countries, with Russia, Canada, Brazil, the United States, and Australia accounting for over half (52%) the extent. Of the 66 countries with retention targets of at least 90% of existing natural and seminatural terrestrial vegetation extent, the majority (n = 63) were in Africa, Asia, and the Americas.

DISCUSSION
We found that approximately 79% of the remaining extent of natural and seminatural terrestrial vegetation we considered should be retained because any loss from these areas could compromise the achievement of key elements of globally agreed environmental goals. This equates to keeping at least half of the planet's non-Antarctic land surface under natural and seminatural terrestrial vegetation coverage-a striking finding that aligns with other recent analyses that call for greatly increased ambition to conserve at least half of the natural world (Dinerstein et al., 2020;Locke, 2014;Noss et al., 2012;Wilson, 2016). Importantly, achieving such retention is not predicated on the exclusion of people from natural and seminatural landscapes and it should not compromise development imperatives. Rather, changes in the way the biosphere is managed and exploited can make this ambition a reality (Tallis et al., 2018). This approach to linking global environmental and sustainability goals via their dependence on natural and seminatural terrestrial vegetation could be extended to other biomes, where appropriate data are available.

Unifying multiple goals
Our method, founded on retention targets, offers a complementary approach to other proposals for setting global environmental targets in that it establishes what is needed (desired outcome), which can then lay a foundation for how to get there (mechanisms). One such proposal is the Global Safety Net, which provides a spatially explicit representation of where vastly increased conservation efforts need to be concentrated across 50% of the terrestrial realm to achieve complementary biodiversity and carbon objectives (Dinerstein et al., 2020). Based on a detailed suite of biodiversity and carbon data, the outputs of this analysis provide key input to future land-use planning and decision-making. Dinerstein et al. (2020) suggest that 50% of the terrestrial realm requires conservation attention, very close to our finding (although locations differed). However, the approach we used to arrive at 50% was very different from that of Dinerstein et al. (2020). Our focus was on the reliance of multiple biodiversity and sustainability objectives that are explicitly captured in several international agreements to retain existing vegetation across a spectrum of land-use types. Although protected areas will always be a cornerstone of conservation efforts, even ambitions to increase their coverage could, in the absence of a more holistic framing, fall short of securing all the ecosystems needed to keep for the full suite of services that nature provides, as well as providing sufficient space for wild species to thrive and for evolutionary processes to continue (Maron et al., 2021).
Our results revealed limited overlap among targets in some parts of the world-a likely function of our input data sets being constrained to certain elements of the biota (e.g., ecoregional representation targets for biodiversity, primary forest only for carbon, focus on vegetation retention in subset of river basins). Because our approach is a coarse-scale approach FIGURE 4 Natural and seminatural terrestrial vegetation requiring retention to meet each of the 4 retention targets considered in this analysis: (a) biodiversity conservation, (b) carbon storage (primary forest only), (c) soil maintenance, and (d) freshwater quality (light green, natural and seminatural terrestrial vegetation [i.e., other natural vegetation] not selected for retention). In terms of strategic priorities, some of the smaller areas identified in (a) are of greater importance in terms of protecting endangered or range-limited endemic species than some of the larger intact areas; however, both are important to an overall global biodiversity conservation strategy. for identifying sites for retention, opportunities for synergies among targets are likely to be greater than we found. Our method is not an optimization and differs from spatial prioritizations. Jung et al. (2021) used an optimization approach to identify high-priority sites for species conservation (but not ecosystem representation nor wilderness areas), carbon retention, and water provisioning. This approach, which identified priority sites, not the amount and spatial distribution of land needed to meet specific outcomes-based targets, is valuable in identifying what and where particular actions should be most urgently undertaken. Many highly biodiverse regions, such as the Amazon and Congo basins and the islands of Borneo and New Guinea, stand out prominently on maps produced in Jung et al. (2021) and through our analyses (Figure 3b).

Implementation opportunities and considerations
Despite our focus on natural and seminatural vegetation, all vegetation is shaped to some degree by human activity and, in many cases, long-term and ongoing stewardship (Ellis et al., 2021;Fletcher et al., 2021). Thus, the way in which this vegetation is best managed to maximize its contributions to each of the global goals is a key consideration. For example, the biodiversity values supported by some areas may require management regimes that are best supported by strict protection. Other sites may be better suited to sustainably managed mixed use (e.g., intensification of existing agricultural practice while retaining remnant natural and seminatural vegetation) landscapes. Our maps do not rule out agricultural expansion globally, but rather highlight where clearing for agriculture is likely to result in challenges to achieving goals such as the SDGs and LDN and should therefore be reconsidered or managed under strict practices (Fastre et al., 2021).
Achieving retention targets will require a carefully developed mixture of protection and sustainable management of natural and seminatural terrestrial vegetation (in parallel with restoration of degraded sites), alongside transformative changes in consumption patterns (supply and demand of food and resources) (Leclère et al., 2020). Imperatives to meet national socioeconomic goals are likely to be particularly acute in many countries, presenting a conflict between the achievement of global environmental goals and development. Many pragmatic decisions and trade-offs will be needed. Well-managed protected areas are long-standing mechanisms for securing ecosystems (Maxwell et al., 2020) that can be complemented by newer approaches, such as other effective area-based conservation measures (Dudley et al., 2018), including privately held land (Clements et al., 2018). Stewardship by Indigenous people and other local communities with connection to the land and knowledge of its management will be necessary for the preservation of large swathes of natural and seminatural landscapes, where such retention is aligned with the aspirations of Indigenous custodians and rights holders (Garnett et al., 2018). Security of tenure and access is an essential enabling condition for this stewardship to be equitable and undertaken in a manner consistent with local peoples' aspirations. Although land under intensive production may contribute to certain ecosystem services, strategies favoring commercial land uses that retain important natural and seminatural vegetation offer a further pathway to retention. For example, sustainable forest management (e.g., via certification schemes [Kleinschroth et al., 2019]) or wildlife-friendly livestock grazing (Fynn et al., 2016) is likely to contribute more to retaining such services than conversion to intensive single-crop agriculture or removal of tree cover for pasture.
Ultimately, achieving targets, such as the retention concept we presented, requires changes in behaviors and decision-making at multiple scales by many different people, and the burden must be managed to avoid exacerbating inequity (Obura et al., 2021). In the context of retention targets, countries (and communities therein, especially Indigenous communities) that support a high proportional extent of natural and seminatural vegetation requiring retention could carry a disproportionate burden when it comes to achieving a key priority target, such as the one we presented. Despite the local (and broader regional and global) benefits of keeping nature in situ, an expectation that communities and nations retain the vast majority of their natural and seminatural vegetation presents complex challenges, due to issues around power imbalances, governance, capacity, and competing (development) aspirations . Recognizing that broad-scale, intensive development is but 1 of several key pathways to improved socioeconomic conditions, the support (finance, capacity and empowerment; distribution of the cost of retention) of other (external) actors may be needed (but only if requested) to ensure that development can occur sustainably in these places while not compromising local ecosystem service provision or global environmental goals . Meaningful engagement with rights holders and other stakeholders at multiple scales, from the outset of any decision-making on development and retention objectives, will be essential to these endeavors (Dawson et al., 2021;Obura et al., 2021).

Limiting vegetation loss
Our analysis acts as a starting point for establishing where and how much terrestrial vegetation should be retained, thereby providing a proof of concept for deriving retention targets to help unify global environmental agreements. The overarching 79% target we identified is not an apex target because it does not encompass all land-cover types or the marine realm. Simply maintaining vegetation in situ will not be sufficient to deliver on goals, such as stopping extinctions, especially where pervasive threats (e.g., illegal harvest, invasive species) imperil biodiversity. Retention must also be complemented by restoration, which will be especially critical for depleted habitats for many threatened species (Allan et al., 2022). The targets also do not cover services that are more challenging to relate to vegetation cover at this broad scale, such as pollination and regulation of air quality.
Data limitations required us to make many simplifying assumptions. Improving the target-specific spatial inputs and the resolution and accuracy of the vegetation data (including information on condition) may allow for a more refined retention map and target values to be produced. This is especially applicable to our exclusion of sparsely vegetated barren areas (per the input land-cover map we used), which cover ecosystems that support unique biota, are vital for combatting desertification, and contribute to carbon storage. Data resolution issues precluded our inclusion of these areas; thus, our results and maps should be considered in the context that these areas also have a key role to play in delivering global environmental agreements. Because we examined terrestrial vegetation, our focus was primarily on the aboveground components. Belowground contributions of these areas are also important. For example, the carbon benefits of retaining forested peatlands are greatly amplified by belowground carbon storage (Hastie et al., 2022). Noting these points, our conceptual and methodological framework provides a foundation on which retention target analyses can be refined-such as by prioritizing vegetation retention based on above-and belowground contributions.
Our analyses set a minimum threshold for retention. We did not comprehensively account for all elements of the biota linked to each goal (e.g., our carbon target only considered primary forest and not other vegetation types or soil carbon; we did not consider biodiversity in unvegetated environments), as well as the relatively narrow scope of goals that we considered (i.e., sole focus on elements of international agreements most directly linked to terrestrial vegetation retention, etc.). Additional goals can only increase this value.
With ongoing losses of natural and seminatural vegetation a certainty and numerous questions around the capacity of many ecosystems to recover (at least in time frames congruent with human lifetimes and time-bound international environmental agreements), balancing losses with gains to achieve net outcomes cannot be relied on to maintain ecosystems at the required retention amount. Retaining vegetation well above the overarching target will act as a buffer for losses that may yet be incurred in areas identified as critical to retain. A small amount of loss could be absorbed in areas that our maps indicate are less critical for retention for any of the 4 targets we considered. However, we urge great caution in this interpretation because we did not assess other values these sites hold (see above barren areas). This vegetation must not be considered worthless regarding its contribution to nature and people.
Increasingly, nations and other influential actors are coalescing around the idea that transformative change to the way humans live and manage the biosphere is needed. Time is running out to halt and reverse the degradation of nature. We propose that a good starting point for initiating this change is asking the simple question: What does humanity need and want from nature? An answer to this question manifests here in the form of a retention target for terrestrial vegetation. For the benefit of biodiversity, and for people too, society needs to keep a great deal of the world's remaining natural and seminatural vegetation in place. Our results provide a starting point to the questions of where and how much this should be.