Restore deforested and degraded terrestrial ecosystems

Restoring deforested and degraded ecosystems — including forests, peatlands, grasslands and coastal wetlands — can help limit global warming and reverse biodiversity loss. Done right, reforestation and other forms of ecosystem restoration can also deliver far-reaching, sustainable development benefits for communities around the world.

Restoring the world’s forests, peatlands, grasslands and coastal wetlands is one of the most promising approaches for combatting the climate crisis and reversing biodiversity loss. If designed well and implemented appropriately, these efforts can also deliver multiple sustainable development benefits.

Restoration and protection must go hand in hand to conserve ecosystems

However, restoration cannot cancel out the impacts of ecosystem loss. For example, it can take decades or longer to fully recover species diversity, ecosystem structure and ecological functions, all of which impact carbon cycling and greenhouse gas (GHG) fluxes. Restoration efforts should therefore complement — but not replace — measures aimed at protecting forests, peatlands, coastal wetlands and grasslands.

Reforestation and other forms of ecosystem restoration generate substantial climate benefits

Reforestation — particularly natural regeneration and assisted natural regeneration — can deliver substantial near-term carbon removals at relatively low costs when compared to more nascent technological approaches like direct air capture. At the same time, swiftly restoring ecosystems that store carbon primarily belowground (including peatlands, grasslands and coastal wetlands) can not only reduce emissions from degrading soils but also enhance carbon sequestration and storage. Together, these measures can contribute nearly one-third of the mitigation needed from terrestrial ecosystems globally in 1.5-degree-C pathways.

Restoring degraded agricultural lands — for instance, by integrating trees in croplands and pasturelands or improving grazing practices across intensively managed grasslands — can generate additional mitigation benefits while enhancing productivity. When implemented alongside ecosystem restoration, these interventions across food and agriculture systems can help spur more durable change.

The importance of adhering to ecological restoration principles to reverse biodiversity loss

However, narrowly focusing on restoration on maximizing carbon removals can backfire. By one estimate, commercial-scale, monoculture plantations of nonnative tree species account for nearly half of the Bonn Challenge commitments to restore forest landscapes. While such initiatives may achieve impressive carbon sequestration rates in the near term, these monocultures lack the functional diversity underpinning ecological resilience and are often more vulnerable to climate impacts and other disturbances that ultimately limit long-term carbon sequestration and storage.

In some instances, these projects may also harm local biodiversity by displacing naturally non-forested ecosystems altogether. For example, scientists have raised concerns about tree-planting efforts occurring across grasslands. But by adhering to broader ecological restoration principles — including focusing on reestablishing a diversity of native species, recovering ecosystem structures and functions, maximizing ecological connectivity, prioritizing natural or assisted natural regeneration wherever possible, and delivering multiple benefits across entire landscapes — restoration efforts can not only avoid unintended negative consequences but also help recover terrestrial biodiversity.

Global efforts to accelerate ecosystem restoration are falling short

Yet globally, a persistent gap remains between restoration’s substantial potential to help combat these crises and current efforts to realize these benefits in practice. The most recent data on gross tree cover gains indicates that progress made in reforesting 100 million hectares by 2030 is well off track, such that pace of collective efforts needs to double to get on track. Current rates of mangrove restoration are also well off track and need to accelerate more than 10-fold by the end of this decade. Even where data is insufficient to assess global action, available evidence suggests that most restoration initiatives are likely small-scale or occurring too slowly, underscoring the urgent need for a step change in action.

State of Forests and Land

Join us on October 27 for the launch of new data and insights on the state of the world's forest and land.

Tracking progress on global outcomes

11877: Reforestation, as measured by total gross gains in tree cover

Status: Well Off Track

Historical Data Methodology and Challenges:

Data limitations pose significant challenges to monitoring reforestation globally, with the total gross area of tree cover gains offering an available global proxy.

However, there are several key limitations in relying on this data from Potapov et al. 2022 to approximate reforestation. First, tree cover gains, alone, do not indicate the full recovery of forests’ pre-disturbance species composition and diversity, or their ecological structures and functions. Although tree cover gains in tree plantations were excluded by using Richter et al. 2024 this data may still include some tree cover gains occurring outside of forests and/or historically forested land, including afforestation (i.e., the conversion to forest of land that has not historically contained forests). Consequently, not all tree cover gains meet the standard definition of reforestation and would not constitute progress made toward these 2030, 2035 and 2050 targets. Second, the data does not include all forms of forest restoration. Because Potapov et al. 2022 measures gross tree cover gains in areas that previously did not have tree cover, the data tracks changes in land use — for example, from abandoned croplands to forests. But the addition of tree cover within degraded, existing forests is not included in this data. Similarly, detecting sparse trees in agricultural landscapes likely requires higher-resolution imagery than that used in Potapov et al. 2022, and so agroforestry is also excluded from this data. Third, because Potapov et al. 2022 used a conservative definition of height change to eliminate noise in the data, tree cover gain may be underestimated in some cases. And finally, because tree cover gain occurs gradually, it is generally more difficult to detect from satellite data within short time frames, limiting the temporal resolution of the data for this indicator. Thus, current global data on tree cover gain represents cumulative total areas from 2000 to 2005, 2005 to 2010, 2010 to 2015, and 2015 to 2020, and annual data is not available. A detailed assessment of the accuracy of this data can be found in Potapov et al. 2022.

Target Methodology:

Reforestation targets are additional to any reforestation that occurred prior to 2020, and these targets are cumulative from 2020 to 2050.

Targets to Limit Warming to 1.5°C: When establishing 1.5°C-aligned reforestation targets, we were unable to determine the ramp-up assumptions from the underlying sources for Roe et al. 2021, because the mitigation potentials and associated area estimates were averaged across the two papers by country and over the 30-year period. Instead, we assumed a linear ramp-up in total reforested area from 2020 to 2050 — that is, that the reforested area would increase each year by the average annual “cost-effective area” provided by Roe et al. 2021 (9.8 Mha/yr) to reach roughly 100 Mha by 2030, 150 Mha by 2035 and 300 Mha by 2050.

To validate that this assumption would provide the bottom-up, cost-effective mitigation potential estimated by Roe et al. 2021 (1.2 gigatons of carbon dioxide equivalent per year [GtCO₂e/yr] from 2020 to 2050), we used the average aboveground and belowground carbon removal rate for “reforestable” land (as defined in Griscom et al. 2017 from Cook-Patton et al. 2020 — 11.57 tonnes CO₂ per hectare per year — to estimate the potential mitigation under the assumption of linear ramp-up in reforested area. The resulting estimate for the annual mitigation potential averaged across the 30-year period is 1.8 GtCO₂e/yr — roughly 0.6 GtCO₂e higher than in Roe et al. 2021. We therefore believe that a linear ramp-up in reforested area is a reasonable assumption, because our estimate meets the mitigation potential identified by Roe et al. 2021.
Targets to Minimize Biodiversity Loss: We did not set an additional target that focuses on minimizing biodiversity losses, because the total area associated with the cost-effective mitigation potential from Roe et al. 2021 — 300 Mha reforested between 2020 and 2050 — is relatively consistent with the total area associated with the maximum sustainable potential for reforestation globally as identified by recent literature, most notably Deprez et al. 2024. This 300 Mha target also falls below the total area available for reforestation that Wang et al. 2025 identifies after excluding places where reestablishing forest cover may spur biodiversity loss, warming from changes in surface albedo, and water stress.

However, while reforesting 300 Mha by 2050 is well aligned with the maximum reforestation potential (305 Mha) estimated by Fesenmyer et al. 2025, it does exceed the study’s estimate of constrained reforestation potential (195 Mha), which accounts for additional food security and biodiversity safeguards. But the methods in Fesenmyer et al. 2025 are sensitive to the tree cover criterion used to define the initial area of land suitable for reforestation (i.e., >60%). Relying on a >50% tree cover criterion, for example, would increase this estimate of constrained reforestation potential to 265 Mha, while employing a >30% tree cover criterion, as well as a lower biomass threshold, would raise this figure further still to 845 Mha.

Progress Assessment Methodology:

Outcome indicators in this system experience high interannual variability in historical data due to both anthropogenic and natural causes. To account for this variability, 10 years instead of 5 years are used to calculate the linear trendline where possible. Following Boehm et al. 2021 and due to data limitations, the average annual rate of change across the most recently available ten-year period (2010-2020) was used to estimate the historical rate of change, rather than a linear trendline.

Keeping the Paris Agreement’s temperature goal within reach will require large-scale carbon removal. Reforestation offers a readily available and relatively cost-effective approach that, when implemented in line with ecological restoration principles, can also help reverse biodiversity loss across the world’s forests.

To help sustain forests’ capacity to sequester and store carbon, as well as safeguard these biodiversity gains, reforestation efforts should also focus on building resilience to intensifying climate impacts — including by reestablishing connectivity, selecting species that can tolerate drought and fire where possible, and adopting management strategies that increase adaptive capacity.

Efforts to scale up reforestation remain well off track. Between 2010 and 2020, a total of 49 million hectares (Mha) — an area roughly the size of Spain — experienced tree cover gain, which offers a best available, though still limited proxy for reforestation and likely represents the upper bound of forest area reestablished during this 10-year period. Getting on track to reforest 100 Mha by 2030, 150 Mha by 2035 and 300 Mha by 2050 will require at least a 2.0-fold acceleration of these recent efforts, such that annual tree cover gains increase from an average of 4.9 Mha between 2010 and 2020 to 10 Mha through the end of this decade.

11882: Peatland restoration, as measured by the total gross area of degraded peatlands that have been rewetted

Status: Insufficient Data

Historical Data Methodology and Challenges:

Peatland restoration targets were adapted from Humpenöder et al. 2020 and Roe et al. 2021, which assume that 0 Mha of degraded peatlands globally were rewetted as of 2015. This assumption, however, does not suggest that peatland restoration is not occurring, as there is evidence of rewetting, for example, in Canada, Indonesia and Russia, but rather speaks to the lack of global data available on peatland restoration.

Reestablishing peatland hydrology through rewetting is just one form of peatland restoration, and this intervention, alone, does not guarantee the full recovery of these ecosystem’s pre-drainage species composition and diversity, or their ecological structures and functions. Additional indicators — such as those focusing on vegetation species or other land cover characteristics — are needed to assess biodiversity outcomes.

Target Methodology:

Peatland restoration targets are additional to any restoration that occurred prior to 2020, and these targets are cumulative from 2020 to 2050.

Targets to Limit Warming to 1.5°C: Roe et al. 2021 defines the bottom-up, cost-effective mitigation potential for avoided GHG emissions from the restoration of degraded peatlands (0.59 GtCO₂e/yr from 2020 to 2050) as the difference in the global area of rewetted peatlands between two SSP2-RCP2.6 scenarios modeled by Humpenöder et al. 2020, using MAgPIE, in 2035. The first scenario assumes land-based climate policies that include peatland protection and restoration, while the second assumes land-based climate policies that include only peatland protection. The resulting area is roughly 16 Mha of degraded peatlands restored by 2035.

For our 1.5°C-aligned targets, we generally followed the ramp-up assumptions in Humpenöder et al. 2020’s scenario that includes peatland protection and restoration policies, which entail restoring approximately 15 Mha by 2030 and 20 Mha by 2050. Note that our ramp-up assumptions involve restoring 16 Mha by 2035, which ensures alignment with the sector’s total contribution to 1.5°C pathways (13.8 GtCO₂e/yr), as estimated by Roe et al. 2021. We set a second, more ambitious, 1.5°C-aligned target than Roe et al. 2021, because some studies (e.g., Leifeld et al. 2019 argues that restoring nearly all degraded peatlands by around mid-century will be required to limit warming to 1.5°C or below, as emissions from drained peatlands may otherwise consume a large share of the global carbon budget associated with this temperature limit. However, as IPCC 2022 notes, restoring all degraded peatlands may not be possible (e.g., those upon which cities have been constructed, that are subject to saltwater intrusion, that have experienced significant subsidence, or that have already been converted into plantation forests). While it remains to be determined with certainty what percentage can be feasibly rehabilitated, particularly at costs of up to $100/tCO₂, several papers find that restoring roughly 50% of degraded peatlands is needed to help deliver AFOLU’s contribution to limiting global temperature rise to 1.5°C. We followed these studies and set a more ambitious, 1.5°C-aligned target than Roe et al. 2021 for 2050. The lower bound of this range involves restoring 20 Mha as estimated by Humpenöder et al. 2020, while the upper bound of this range entails restoring roughly half of degraded peatlands, recently estimated at 57 Mha globally by UNEP 2022.
Targets to Minimize Biodiversity Loss: Similarly, to minimize biodiversity loss, we set a high-ambition target that calls for rewetting all degraded peatlands, which are currently estimated at 46 Mha by Humpenöder et al. 2020 and 57 Mha by UNEP 2022.

Progress Assessment Methodology:

Even if peatland degradation ended today, previously degraded peatlands could continue emitting roughly 1.9 gigatonnes of carbon dioxide equivalent (GtCO₂e) per year for decades to centuries. Unlike forests, these ecosystems store carbon and nitrogen primarily within their waterlogged soils rather than in aboveground vegetation; consequently, CO₂ and nitrous oxide (N₂O) emissions following disturbances are not immediate and can continue until all peat is fully lost or the soil is rewetted.

The efficacy of restoring peatlands to avoid these greenhouse gas (GHG) emissions depends, in part, on how these ecosystems were degraded (e.g., through drainage, burning or cutting). For example, rewetting peatlands drained by agriculture — a primary driver of degradation — can significantly reduce or even halt carbon losses, as well as enable carbon sequestration. However, restoring these ecosystems’ water tables may also increase methane (CH₄) emissions, particularly in the short term (though these CH₄ emissions vary substantially across different sites). To maximize mitigation benefits, peatland rewetting should occur as quickly as possible, such that the CO₂ and N₂O emissions avoided outweigh any CH₄ emissions released.

For some degraded peatlands, rewetting can also help reduce the potential for fires, recover habitat for wetland species and reestablish ecological functions and services. But for others, such efforts can lead to entirely new peatland ecosystems that bear little resemblance to their near-natural counterparts, even several decades after their water tables were restored. Such disparate results indicate gaps in our collective understanding of the factors that determine successful peatland restoration for biodiversity, as well as the need for additional interventions like removing encroaching trees and shrubs or reintroducing peatland vegetation.

Nonetheless, rewetting 15 million hectares (Mha) — more than a quarter of all degraded peatlands worldwide — by 2030, 16 Mha by 2035 and 20 Mha to 29 Mha by 2050 can help keep the Paris Agreement limit within reach. Although data is insufficient to assess global progress made toward this near-term target, available evidence suggests that peatland rewetting is occurring in countries like Canada, Indonesia and Russia but likely not yet at the pace and scale required. Minimizing biodiversity losses would call for even greater ambition and action: rewetting 46 Mha to 57 Mha of degraded peatlands.

11853: Mangrove restoration, as measured by gross direct gains in mangrove extent

Status: Well Off Track

Historical Data Methodology and Challenges:

Murray et al. 2022 estimates gross area of mangrove gain from 1999 to 2019, defining gain as mangrove establishment in areas where mangroves were not present in 1999. The vast majority of these gains, however, are due to natural, broad-scale coastal processes like the colonization of new sediments or inland migration, with only 8% likely attributable to direct human interventions like mangrove planting and other restoration activities, including those that reestablish tidal regimes. We therefore used this proportion of mangrove gains (i.e., “direct” gains attributable to human interventions) as a proxy for mangrove restoration. More specifically, we calculated the most recent data point for mangrove restoration by taking 8% of the total gross mangrove area gained from 1999 to 2019 (180,000 ha), which equals 15,000 ha.

However, there are a number of limitations in using these “direct” mangrove gains as a proxy for mangrove restoration. As with forests, mangrove trees grow gradually, and therefore mangrove gain is more challenging to monitor on shorter timescales, as gains may not be detected until mangrove trees reach a certain level of maturity. Therefore, recently established mangroves may not be included in these estimates. Moreover, the establishment of mangrove trees does not always indicate the recovery of these ecosystems’ species composition and diversity, or their ecological structures and functions. And, in some cases, the addition of mangrove trees can lead to negative consequences (e.g., the loss of other coastal wetland ecosystems, such as salt marshes or seagrass meadows) or short-lived gains if tree planting is not implemented appropriately. Therefore, this proxy may include mangrove gains that would not be considered mangrove restoration. A detailed assessment of the accuracy of this data can be found in Murray et al. 2022.

A final limitation of relying on gross gains in ecosystem extent to approximate restoration is that this data only measures changes in land use — for example, from abandoned oil palm plantations or aquaculture ponds to mangrove forests. The addition of mangrove trees within existing, degraded mangroves represents another form of restoration, but it is not included within the Murray et al. 2022 data.

Target Methodology:

Mangrove restoration targets are additional to any restoration that occurred prior to 2020.

Targets to Limit Warming to 1.5°C: Roe et al. 2021 defines the bottom-up, cost-effective mitigation for enhanced carbon sequestration from mangrove restoration (0.01 GtCO2e/yr from 2020 to 2050) as 30% adoption of the technical potential from Griscom et al. 2020, expanded to include nontropical countries. Technical potential is defined as the restoration of mangroves lost since 1996, excluding those lost to erosion or urbanization Griscom et al. 2020. We therefore calculated 30% of the area associated with the technical potential (810,000 ha) to derive our 1.5°C-compatible target. Following ramp-up assumptions from Griscom et al. 2020, we set our target to achieve this restoration by 2030 and, therefore, present a 2030 target only.

Griscom et al. 2020 notes that this target is conservative as it excludes mangrove forests lost before 1996, and previous studies suggest that mangrove losses in the 1980s and 1990s were significant: the world may have lost as much as 35% of its mangrove forests during these two decades. This 1.5°C-compatible target, therefore, likely represents the area of mangroves that, at a minimum, needs to be restored to achieve climate mitigation goals.
Targets to Minimize Biodiversity Loss: We relied on the technical potential for mangrove restoration from Roe et al. 2021 to set this target. As above, because it is defined as the restoration of all mangroves lost since 1996, excluding those lost to erosion or urbanization (Griscom et al. 2020), it is conservative.

Progress Assessment Methodology:

Outcome indicators in this system experience high interannual variability in historical data due to both anthropogenic and natural causes. To account for this variability, 10 years instead of 5 years are used to calculate the linear trendline where possible. Following Boehm et al. 2021 and due to data limitations, however, the average annual rate of change across the most recently available time period (1999-2019) was used to estimate the historical rate of change, rather than a linear trendline.

Restoring mangrove forests not only enhances their ability to sequester and store carbon but may also reduce greenhouse gas (GHG) emissions that would have otherwise continued for decades after certain disturbances, such as the loss of soil carbon following drainage for palm oil production. If implemented following ecological restoration principles, these efforts can also help reverse biodiversity losses — for instance, by recovering critical habitats for species, including nurseries for many of the world’s fisheries. Mangrove restoration can also reestablish their ecological functions and services, including their ability to reduce acidification of surrounding seawaters.

Ensuring the long-term survival of these restored ecosystems will also require projects to prioritize building resilience to climate impacts. This entails, for example, selecting species that can tolerate extreme storms, as well as both increasing river connectivity to deliver the sediment mangroves need for vertical accretion and identifying restoration sites along coastlines that can accommodate inland migration to help these ecosystems avoid drowning from rising sea levels.

Progress made in restoring these coastal wetlands remains lackluster. Best available estimates of global gains in mangrove extent indicate that these forests expanded across just over 180,000 hectares (ha) between 1999 and 2019. But direct human interventions like planting mangroves or reestablishing tidal regimes accounted for just 8% of these total gains (15,000 ha).

Global efforts to restore another 240,000 ha of mangrove forests by 2030 are well off track and will require these direct gains, which increased by an average of roughly 750 ha per year during this 20-year period, to accelerate more than 10-fold. Similarly, recent estimates of mangrove restoration from the Food and Agriculture Organization of the United Nations indicate that progress made in reestablishing these coastal wetlands over the last two decades falls well short of the direct gains needed through 2030 to help combat the climate crisis, and minimizing biodiversity losses would call for even greater ambition: restoring 810,000 ha of deforested and degraded mangroves.

11883: Salt marsh restoration, as measured by gross direct gains in salt marsh extent

Status: Right Direction

Historical Data Methodology and Challenges:

Murray et al. 2022 estimates gross area of tidal marsh gain from 1999 to 2019, defining gain as tidal marsh establishment in areas where tidal marshes were not present in 1999. The vast majority of these gains, however, are due to natural, broad-scale coastal processes like the colonization of new sediments or inland migration, with only 30% likely attributable to direct human interventions like marsh vegetation planting and other restoration activities, including those that reestablish tidal regimes. We therefore used this proportion of tidal marsh gains (i.e., “direct” gains attributable to human interventions) as a proxy for tidal marsh restoration. More specifically, we calculated the most recent data point for tidal marsh restoration by taking 30% of the total gross tidal marsh area gained from 1999 to 2019 (116,000 hectares [ha]), which equals 35,000 ha.

The establishment of tidal marsh vegetation does not always indicate the recovery of these ecosystems’ species composition and diversity, or their ecological structures and functions. And in some cases, the addition of tidal marsh vegetation can lead to negative consequences, like the spread of invasive, alien species and the loss of tidal mudflats. Therefore, this proxy may include tidal marsh gains that would not be considered tidal marsh restoration. A detailed assessment of the accuracy of this data can be found in Murray et al. 2022.

A final limitation of relying on gross gains in ecosystem extent to approximate restoration is that they only measure changes in land use — for example, from abandoned rice paddies to tidal marshes. The addition of tidal marsh vegetation within existing, degraded tidal marshes represents another form of restoration, but it is not included within the Murray et al. 2022 data.

Target Methodology:

1.5°C-compatible targets for restoring salt marshes are not available, while those focused on minimizing biodiversity loss are either outdated (e.g., Griscom et al. 2017 relies on data published more than two decades ago to estimate a technical potential for salt marsh restoration), not global in scope (e.g., Jankowska et al. 2022 scales U.S. data on the area available for salt marsh restoration in the United States to the global level), or lack sufficient data to translate into a quantitative target (e.g., Díaz et al. 2020 calls for a 20% net gain in area by 2050 but does not specify a baseline year). Accordingly, we can only indicate the direction of change required, as informed by IPBES 2019 and IPCC 2022.

Progress Assessment Methodology:

Outcome indicators in this system experience high interannual variability in historical data due to both anthropogenic and natural causes. To account for this variability, 10 years instead of 5 years are used to calculate the linear trendline where possible. Following Boehm et al. 2021 and due to data limitations, the average annual rate of change across the most recently available time period (1999-2019) was used to estimate the historical rate of change, rather than a linear trendline.

Salt marsh restoration can enhance carbon sequestration and storage and avoid greenhouse gas (GHG) emissions that would have otherwise continued for decades after certain disturbances, like the drainage of these coastal wetlands for urban development projects. If implemented following ecological restoration principles, these efforts can also help reverse biodiversity losses (e.g., by recovering critical habitats for species, including refuges for migratory birds and nurseries for fisheries) and re-establish marshes’ ecological functions and services.

Ensuring the long-term survival of these restored ecosystems will also require measures to adapt to intensifying climate impacts, particularly sea level rise. For instance, increasing river connectivity can help deliver the sediment that marshes need to accrete vertically and keep pace with rising tides, while reestablishing these coastal wetlands across shorelines with sufficient space for inland migration can also help prevent drowning.

Best available estimates of global gains in salt marsh extent indicate that these ecosystems expanded across nearly 120,000 hectares (ha) between 1999 and 2019, though direct human interventions like planting marsh vegetation or reestablishing tidal regimes accounted for roughly 30% of these total gains (35,000 ha). This data indicates that efforts to restore these ecosystems are heading in the right direction.

11860: Grassland, savanna and shrubland restoration, as measured by total gross gains in these ecosystems’ extent

Status: Right Direction

Historical Data Methodology and Challenges:

Gain, as a proxy for restoration, is defined as a transition from cropland or pastureland to grassland, shrubland and savanna that persists for at least five years. To monitor annual gains in the global extent of grasslands, savannas and shrublands, we used three complementary datasets:

The grassland extent collection 1.1 by Global Pasture Watch’s (GPW), which provides annual, 30 meter (m) maps of grassland extent ecosystems from 2000-2024. A cross-validation assessment for this GPW product is available in the source publication, Parente et al. 2024.
The Global Land Analysis and Discovery (GLAD) land cover product, which consists of annual (2015-2024) maps of land cover at 30 m spatial resolution. An accuracy assessment of the GLAD land cover product is available in the source publication, Potapov et al. 2022.
The Global Forest Change (GFC) dataset, which is available on Global Nature Watch, provides tree cover loss from 2001 to 2024 at 30 m spatial resolution. An accuracy assessment of the GFC land cover product is available in the source publication, Hansen et al. 2013.

To calculate grassland, savanna and shrubland restoration, we needed to separate temporary gains from those that are permanent or semipermanent, which approximate restoration, by harmonizing the GPW and GLAD datasets into a single combined annual map. To that end, we reclassified the GLAD land cover from classes from Potapov et al. 2022 into seven categories: bare ground, short vegetation, trees, water, snow/ice, cropland and built-up areas (Table 1). We then replaced the GLAD short vegetation class with the GPW grassland categories (cultivated and natural/seminatural), where available. Because GLAD land-cover maps are not available between 2011 and 2014, we took a two-step approach to assign a class for these years. First, we assigned the 2015 annual GLAD classes to all years between 2011 and 2014. Second, we replaced the GLAD short vegetation class with the corresponding GPW grassland category (cultivated or natural/seminatural) for each year from 2011 to 2014, following the same approach described above. This reclassification produced a combined GLAD-GPW product from 2011 to 2024 that enabled spatially explicit, annual tracking of grassland, savanna and shrubland gain, which can be regularly updated contingent on the continued availability of input data layers.

Additionally, since our definition of grasslands is very broad and overlaps with areas of tree cover, we applied a mask based on tree cover loss data (GFC) available on Global Nature Watch. We identified all pixels with >30% tree canopy cover in 2000 that experienced tree cover loss at any time between 2001 and 2024 and excluded them from the grassland, shrubland and savanna restoration analysis. This ensures that we are not conflating post-deforestation regrowth with grassland, shrubland and savanna restoration. Qualitative assessment showed that at the global scale, the masking is conservative and does not reduce grassland gain notably.

Table 1: Definitions for Global Pasture Watch class categories

Class	Short definition	Called here	Original source
Cultivated grassland	Areas where grasses and other forage plants have been intentionally planted and managed, as well as areas of native grassland-type vegetation where they clearly exhibit active and “heavy” management for specific human–directed uses, such as directed grazing of livestock.	Pastureland	Parente et al. 2024
Natural/seminatural grassland	Relatively undisturbed or sparsely managed native grasslands/short-height vegetation, such as open shrublands and savanna, as well as areas that have experienced varying degrees of human activity in the past, which may contain a mix of native and introduced species due to historical land use and natural processes.	Grasslands, shrublands and savannas	Parente et al. 2024
Bare ground	Land with 0-19% vegetation fraction.		Potapov et al. 2022
Short vegetation	Non–treed vegetation cover with 20% or more vegetation fraction.
Tree cover	An area with tree height ≥5 m at the Landsat pixel scale; includes wildland, managed, and planted forest, agroforestry, orchards and natural tree regrowth.
Water	Open surface water (inland) that covers ≥50% of a pixel for 20% or more of the year and is not obscured by tree canopy, floating vegetation, bridges or ice.
Snow/ice	Land covered by glaciers and snow that remains throughout the entire year.
Cropland	Land used to produce annual and perennial herbaceous crops for human consumption, forage, and biofuel; excludes tree crops, permanent pastures and shifting cultivation.
Built-up	Man-made land services associated with infrastructure, commercial and residential uses; defined as pixels containing any man-made surfaces even if they are not dominant.

We completed this analysis at 30 m spatial resolution globally from 2016 to 2024, with 2011 as a baseline year, using the harmonized and masked dataset described above. Gains of grassland, savanna and shrubland were counted as a proxy for restoration if they met two criteria: a pixel transitioned from cropland or pastureland to grassland, savanna or shrubland; and it remained as grassland, savanna or shrubland for at least five consecutive years.

To estimate annual grassland, shrubland and savanna gains, we summed up the area of all pixels in the pastureland and cropland classes that transitioned to the grassland, shrubland and savanna class. Built-up classes were not included, because built-up loss was not present in the GLAD land cover products. We excluded all gains of grassland, shrubland and savanna from the remaining GLAD classes, such as snow, ice, bare ground, short vegetation or water, as they were likely temporary. Such transitions are often driven by natural seasonal dynamics or interannual variability, such as melting snowpack, retreating glaciers or temporary drying of water bodies, and do not represent intentional or lasting ecosystem restoration. This analysis also does not include transitions from areas of tree cover to grassland, shrubland and savanna. Such areas are not treated as genuine gains, as they can represent post-deforestation landscapes rather than the restoration or reestablishment of historically open ecosystems. Including them would risk conflating ecosystem loss with restoration.

Gains were registered only when a pixel that had transitioned from cropland or pastureland remained classified as grassland, shrubland and savanna for five consecutive years. This persistence criteria minimizes false positives and allows time for initial vegetation recovery. This also avoids classifying cropland fallow periods as gain (the Food and Agriculture Organization of the United Nations classifies fallow areas as arable land as long as the fallow periods are less than five years).

Together, this approach produces a conservative estimate of grassland, savanna and shrubland gains by requiring strong evidence of a lasting transition and filtering out short-term, ambiguous or post-deforestation changes.

Relying on grassland, shrubland and savanna gains as a proxy for restoration has several limitations. It is subject to the errors of commission and omission present in the input GLAD and GPW datasets. Area estimates for this analysis were derived using pixel counting, which involves summing the number of 30 m pixels classified as grassland, shrubland and savanna and converting this count to an area estimate. However, this method does not account for classification uncertainties or biases. Gross gains capture the extent of land cover change from annual cropland and pastureland into grassland, savanna and shrubland. As an area-based indicator, this analysis tracks changes in ecosystem extent only and does not assess changes in ecosystem condition, function or biodiversity within existing grasslands, savannas and shrublands. These transitions may result from abandoned agricultural land reverting through natural regeneration, intentional restoration efforts aimed at reestablishing native vegetation, or land management practices, such as controlled burning or managed grazing, that reduce woody encroachment and maintain open vegetation structures.

While restoration is an important pathway for increasing the extent of these ecosystems, not all gross gains detected in this analysis reflect deliberate or ecologically desirable restoration. Some observed transitions may result from land abandonment rather than active restoration, and in some cases, regrowth may be dominated by invasive species or woody plant encroachment. Further, long fallow periods within rotational or shifting agricultural systems may be misclassified as gains when in fact they are part of a temporary cycle.
This analysis does not capture the extent or severity of degradation within grassland, savanna and shrubland. While it is true that pixels with very heavy signs of livestock impact on the landscape tend to be classified as pastureland (and therefore are outside of the grassland, shrubland and savanna class), subtle or functionally degraded areas may remain within it. For instance, a pasture experiencing woody encroachment or secondary succession might be reclassified as grassland, shrubland and savanna, thereby appearing as a gain, even if the land’s ecological function for biodiversity or grazing has declined. This type of shift may represent either a degraded pasture or a naturally regenerating landscape, depending on the ecological context.
This area-based definition of restoration does not capture restoration activities within existing grasslands that enhance ecosystem function without resulting in detectable land cover change. Practices such as silvo-pastoral management, soil and water conservation measures (e.g., terracing or half-moons), or other management interventions can meaningfully contribute to restoration but would not be detected by this analysis.
Because the establishment of ecological structure and function takes time, the gross gain indicator may capture vegetation changes before full restoration outcomes (e.g., biodiversity recovery, soil carbon build-up) are achieved. Therefore, some gains may reflect early-stage transitions rather than mature, functioning and recovered ecosystems. This is particularly relevant for grassland systems, where soil carbon recovery occurs slowly, meaning that observed gains in extent may not correspond to substantial near-term carbon sequestration benefits.
Classifying transitions from cropland and pastureland to grassland, shrubland or savanna as restoration is difficult without knowledge of the original land cover, as such changes may not represent true recovery toward a historical reference state and may not align with restoration objectives for a given landscape.

Target Methodology:

1.5°C-compatible targets for restoring grasslands, savannas and shrublands are not available. Similarly, we could not identify recent estimates of technical potential for restoring these ecosystems, which can serve as a basis for establishing targets focused on minimizing biodiversity loss, while high-ambition goals like those published in Díaz et al. 2020 do not specify a baseline year and, consequently, cannot be translated into a quantitative target. Accordingly, we present directional, qualitative targets informed by Syktus and McAlpine 2016, Yang et al. 2019, Díaz et al. 2020, Bai and Cotrufo 2022 and Walker et al. 2022.

Progress Assessment Methodology:

Outcome indicators in this system experience high interannual variability in historical data due to both anthropogenic and natural causes. To account for this variability, 10 years instead of 5 years are used to calculate the linear trendline where possible. However, due to data limitations, we calculated a linear trendline for this indicator with 9 years of data from 2016 to 2024.

Restoring grasslands, savannas and shrublands, particularly those cleared for agricultural production, represents an important yet often overlooked opportunity for climate mitigation and biodiversity conservation. Because these ecosystems store a substantial share of their carbon belowground, expanding their extent across former croplands and pasturelands can help initiate the long-term recovery of soil organic carbon stocks and reduce emissions from degraded soils.

If implemented following ecological restoration principles, such transitions can help reestablish ecological functions and services like filtering out pollutants from water, protecting soils from erosion and providing food to livestock. Restoring these ecosystems can also help reverse biodiversity loss by providing habitat for wildlife, including critical pollinators. Such gains encompass seminatural grazing lands as well, where sustainable livestock grazing can support, rather than undermine, ecological function and biodiversity.

Between 2016 and 2024, a total of 32 million hectares (Mha) of former croplands and pasturelands — an area roughly the size of Vietnam — experienced gross gains in grassy vegetation, which offers a globally consistent proxy indicator of grassland, savanna and shrubland restoration. Recent rates of change are heading in the right direction and have increased by an average 190,000 ha per year over the past nine years.

This data approximates just one facet of restoration: the re-expansion of grassland, savanna and shrubland ecosystems across agricultural lands, including both croplands and pasturelands. It does not track ongoing efforts to improve habitat connectivity, or those dedicated to restoring the ecological integrity of grasslands, shrublands and savannas that have experienced degradation but not full conversion.

Across many of these degraded ecosystems, particularly where disturbances were less intensive, ecological structures and functions can often be restored through targeted management. Improving fire and grazing practices — including sustainable grazing that mimics the role of native herbivores in maintaining open vegetation — can reduce woody encroachment, maintain herbaceous cover and support belowground root systems. Natural regeneration, native species seeding and transplanting, and removal of invasive plants can also be effective in restoring these degraded ecosystems.

Even with active management, old-growth grasslands, savannas and shrublands — ecosystems that have been shaped by millennia of fire, herbivory and gradual soil development — harbor high levels of plant diversity and complex belowground root and bud systems that are often quite slow to recover. Secondary grasslands, for example, can take a century or more to reassemble the plant species richness of their old-growth counterparts.

Key enablers and barriers to change

Other shift Other shifts needed to transform the system

Not only is land-use change across forests, peatlands, grasslands and coastal wetlands the leading direct driver of terrestrial biodiversity loss, but it also releases substantial emissions and weakens these ecosystems’ ability to sequester carbon.

Unsustainable management of terrestrial ecosystems, particularly forests, grasslands, savannas and shrublands, is exacerbating the climate and biodiversity crises. When tailored to local contexts and implemented appropriately, improved management practices can reduce emissions, increase carbon sequestration and bolster climate resilience, all while supporting productivity and local livelihoods.

Sharply reducing pollution in forests, grasslands, peatlands and coastal wetlands is essential to halt biodiversity loss, maintain ecosystem services and protect human health. Key contaminants include heavy metals, persistent organic pollutants such as PFAS, pesticides, nutrient pollution, plastics and artificial light.

Human-driven overexploitation of wild species — including legal and illegal logging, hunting and trade — is a major driver of biodiversity loss and species extinction. Reversing this trend requires better monitoring of wild populations, stronger protection of threatened species, sustainable management of species use and effective regulation of wildlife trade.

Restore deforested and degraded terrestrial ecosystems

Restoration and protection must go hand in hand to conserve ecosystems

Reforestation and other forms of ecosystem restoration generate substantial climate benefits

The importance of adhering to ecological restoration principles to reverse biodiversity loss

Global efforts to accelerate ecosystem restoration are falling short

Tracking progress on global outcomes

Key enablers and barriers to change

Other shift Other shifts needed to transform the system

Effectively halt deforestation, other forms of conversion, and degradation across terrestrial ecosystems

Sustainably manage terrestrial ecosystems to reduce GHG emissions and enhance carbon removals

Sharply reduce pollution in terrestrial ecosystems

Halt the overexploitation of terrestrial wild species and ensure sustainable use

Restoration and protection must go hand in hand to conserve ecosystems

Reforestation and other forms of ecosystem restoration generate substantial climate benefits

The importance of adhering to ecological restoration principles to reverse biodiversity loss

Global efforts to accelerate ecosystem restoration are falling short

Tracking progress on global outcomes

11877: Reforestation, as measured by total gross gains in tree cover

11882: Peatland restoration, as measured by the total gross area of degraded peatlands that have been rewetted

11853: Mangrove restoration, as measured by gross direct gains in mangrove extent

11883: Salt marsh restoration, as measured by gross direct gains in salt marsh extent

11860: Grassland, savanna and shrubland restoration, as measured by total gross gains in these ecosystems’ extent

Key enablers and barriers to change

12743: Countries with national restoration targets for terrestrial ecosystems

12741: Countries with at least one policy, law or regulation mandating restoration following ecosystem conversion or degradation

12747: Integrated spatial planning, as measured by the share of the world’s land covered by a participatory, integrated and biodiversity-inclusive spatial plan

12750: Individuals’ tenure security, as measured by the share of adults who consider it likely or very likely that they will lose rights to their home or land against their will in the next five years

12746: Indigenous and community land rights, as measured by the share of land legally designated for or owned by Indigenous Peoples, Afro-descendant Peoples and local communities

12751: Locally led restoration projects with participatory and inclusive decision-making processes

12748: Countries upholding the right to free, prior and informed consent for Indigenous Peoples, Afro-descendent Peoples and local communities

12749: Countries with at least one restoration seed bank

12752: Stronger regulatory enforcement, as measured by the share of countries scoring relatively highly on regulatory enforcement

12742: Finance for restoration, as measured by total finance for nature-based solutions

12744: Terrestrial restoration monitoring, as measured by the share of restored terrestrial ecosystems monitored over the long term

12753: Access to forest extension services, as measured by the ratio of extension agents to landholders

Other shift Other shifts needed to transform the system

Subscribe for updates