Pivoting forestry and agricultural concessions toward conservation in Southeast Asia Authors: Annabel Jia Yi Lim, Yiwen Zeng, Hoong Chen Teo, Jose Don T De Alban, Lian Pin Koh Published: 2026 ABSTRACT AbstractExpansion of forestry and agriculture in Southeast Asia has driven widespread deforestation, biodiversity loss and carbon emissions. Yet large areas of natural forest remain within concessions, representing overlooked opportunities for conservation and climate mitigation. We identify 42 Mha of natural tropical moist forest within 3,754 logging, oil palm, planted wood and rubber concessions across Indonesia, Malaysia, Cambodia and Myanmar. Clearing these forests could threaten 4.6 Mha of Key Biodiversity Areas and release ~1,216 Mt CO2 over 30 years. We assess the economic viability of conserving these forests through carbon market–supported avoided deforestation, estimating break-even carbon prices ranging from US$33 per t CO2 to US$1,677 per t CO2. Our results suggest that substantial areas could be conserved under higher carbon prices, but realising this potential will require complementary finance mechanisms and supportive policy and governance frameworks.Subject terms: Environmental economics, Climate-change mitigation, Environmental impact, Conservation biology, Social sciences Introduction Over the past few decades, Southeast Asia has experienced a dramatic expansion of commercial forestry and agriculture1. While these activities have contributed to economic growth and job creation, they are also linked to widespread deforestation2,3. Collectively, forestry and agriculture were almost entirely responsible for the loss of 39 million hectares of Southeast Asia’s forests between 2001 and 2015, spanning more than 8% of the region’s terrestrial area4. Such pervasive land-use change has imperilled biodiversity, ecosystem services, and local communities5,6, as well as contributed to carbon emissions and climate change7. Forestry and agricultural concessions are drivers of this forest loss8. In many parts of Southeast Asia, governments grant licenses for these concessions to private corporations, giving companies the right to utilise the land for designated resource extraction and/or cultivation over a fixed duration9,10. That concessions are often initially occupied by natural forests means large swathes have been felled for harvesting or to make way for plantations11,12. However, there may still be considerable, albeit yet unknown, amounts of relatively intact natural forests within concessions. If Southeast Asian concessionaires can be sufficiently motivated to protect their remaining forests, and policy environments enable them to do so, they could transform their lands from centres of deforestation into foci for nature conservation and climate action13,14. Green finance—funds from private, public, and non-profit bodies channelled into environmental sustainability initiatives—may provide the incentives and capital to make this forest transition a reality. Green finance encompasses mechanisms such as philanthropy, green loans, and impact investments (e.g., green bonds) to carbon markets, ecosystem service payments, and blended approaches15,16. Returns from green finance could allow concessionaires to meet the operational costs of safeguarding forests, and even generate profits17,18. Nonetheless, for protecting forests through any green finance mechanism to be deemed competitive with forestry and agriculture, returns from the former should be comparable to or exceed those attained from the latter19,20. In this study, we evaluate the potential for forestry and agricultural concessions in Cambodia, Indonesia, Malaysia, and Myanmar to leverage carbon markets in their transition to forest protection. First, we quantify the amount of natural tropical moist forests and their carbon stocks remaining within 3,754 logging, oil palm, planted wood and rubber concessions across Indonesia, Malaysia, Cambodia and Myanmar. Second, we assess the economic viability of protecting these forests through avoided deforestation projects by modelling break-even carbon prices for each concession type in each country, and for each concession. Break-even carbon prices are the minimum prices per tonne of avoided CO2 emissions (i.e., carbon credit) at which transitioning to forest protection becomes competitive with forestry and agriculture. We determine these prices by constructing break-even price curves and applying them to previously reported forestry and agricultural returns in the region. Our research provides one indicator of the potential of green finance in facilitating forest transition in Southeast Asia. Results & discussion Amount of natural forests and carbon stocks within concessions Across Southeast Asia, 42 Mha of natural tropical moist forests remain within 62 Mha of concession land (Table 1). Put in perspective, concessions contain more natural forests at present than the amount of forests lost over the past 15 years across the whole region (39 Mha)4. The bulk of natural tropical moist forests persists inside logging concessions (29 Mha) and planted wood concessions (7 Mha), occupying 9–95% and 13–79% of these concession lands respectively (Table 1). Table 1Amount of natural tropical moist forests and associated aboveground carbon stocks within concessionsCountryConcession typeTotal concession area (ha)Natural tropical moist forestsTotal Aboveground carbon (Mt)UndisturbedDegradedRegrowthCoverage (%)Area (ha)Mean ACD (Mg ha−1)Area (ha)Mean ACD (Mg ha−1)Area (ha)Mean ACD (Mg ha−1)IndonesiaLogging25,805,70021,283,454136.32 ± 0.622,795,866116.88 ± 6.96435,41183.26 ± 0.0395.03,264.35 ± 23.55Oil Palm12,450,3702,075,53699.73 ± 8.311,836,31172.64 ± 8.02825,69147.99 ± 0.0138.1380.01 ± 22.67Planted Wood10,882,3502,976,688104.89 ± 24.772,190,57081.22 ± 16.03843,42455.91 ± 0.0155.2537.28 ± 81.67Overlapping2,734,0441,123,074120.12 ± 35.02522,15091.42 ± 44.95141,08652.52 ± 0.0365.3190.05 ± 45.79Malaysia (Sarawak)Logging4,423,0432,911,760116.22 ± 0.181,046,358107.58 ± 0.42156,46792.52 ± 1.1293.0465.43 ± 0.70Oil Palm1,132,368168,620102.52 ± 1.63132,17688.00 ± 2.3133,03258.02 ± 4.6929.530.84 ± 0.44Planted Wood671,454291,696115.02 ± 0.61210,839105.26 ± 1.1325,54584.15 ± 5.1278.757.89 ± 0.32Overlapping2,247,9291,111,071116.63 ± 0.29622,140108.26 ± 0.4486,76989.37 ± 1.9681.0204.69 ± 0.45CambodiaLogging78,31599485.11 ± 1.815,53881.58 ± 205.2237380.92 ± 2.978.80.57 ± 1.14Oil Palm72,5082,12261.58 ± 4.622,61057.58 ± 3.4747639.11 ± 3.377.20.30 ± 0.01Planted Wood110,4445,90287.26 ± 12.535,85467.33 ± 76.322,58262.39 ± 10.3613.01.07 ± 0.45Rubber937,03241,58894.90 ± 15.7329,36980.22 ± 13.418,25670.60 ± 13.188.56.89 ± 0.77Overlapping195,64036264.77 ± 13.7752557.72 ± 3.7532448.08 ± 4.920.60.07 ± 0.01MyanmarOil Palm308,92492,109103.83 ± 0.7937,41685.62 ± 1.6529,53461.56 ± 3.0651.514.59 ± 0.13ACD, aboveground carbon density. Undisturbed, degraded, and regrowth classes are as defined by Vancutsem et al.83. Coverage is the percentage of natural tropical moist forest area out of the total concession area. All uncertainties are expressed as standard errors. Values only pertain to tropical moist forest ecosystems and do not include other forest types, such as Cambodia and Myanmar’s extensive dry forests; numbers for these countries may therefore appear comparatively low. Amount of natural tropical moist forests and associated aboveground carbon stocks within concessions ACD, aboveground carbon density. Undisturbed, degraded, and regrowth classes are as defined by Vancutsem et al.83. Coverage is the percentage of natural tropical moist forest area out of the total concession area. All uncertainties are expressed as standard errors. Values only pertain to tropical moist forest ecosystems and do not include other forest types, such as Cambodia and Myanmar’s extensive dry forests; numbers for these countries may therefore appear comparatively low. One reason for the substantial quantity of natural forests remaining within concessions is that forest conversion typically proceeds when it is expected to generate a return on investment. This is contingent on the costs of commodity production, which can involve non-trivial spending on manpower, machinery, and infrastructure like access roads21,22. It is also contingent on commodity demand; when sales lose momentum and prices weaken, concessionaires may pause expansion until prices recover23. Lastly, with reference to forestry concessions, the common practice of selective logging in Southeast Asia leaves large logging blocks intact, or distributes felling over large areas as opposed to clearcutting24–26. The extensive amounts of natural forests in these concessions signal that there is much at stake. The remaining forests identified in this study are vital to regional conservation. They belong to diverse ecosystems, including lowland rainforests, montane rainforests, mangrove forests, and carbon-dense peat swamp and freshwater swamp forests27. We moreover find that these forests in concessions co-occur with many threatened biodiversity- and endemic-rich habitats, such as the Heart of Borneo28 and 4.6 Mha of Key Biodiversity Areas (KBAs)29. In addition, the fate of natural forests within concessions will have serious ramifications for the climate. Together, natural tropical moist forests in the concessions we studied store 5.2 ± 0.1 Gt C in their aboveground biomass (Table 1). Should deforestation trends in concessions continue, a sizeable portion of this carbon could be released, equivalent to an estimated 1,216 Mt CO2 over the next 30 years. This is more than 20% of all industry sector emissions produced across ASEAN between 2000 and 2023 (5,823 Mt CO2)30. We note that the conservation and climate change mitigation potential of natural forests across all Southeast Asian concessions is probably markedly greater than our estimates here. Better data coverage of concession types and areas would likely reveal that far more natural forests are intact within concessions in the region8. Similarly, incorporating other carbon pools (e.g., belowground biomass carbon), emissions sources (e.g., forest degradation), and natural forest types (e.g., dry forests) would likely increase the overall magnitude of climate benefits31–33. Maintaining these forests for conservation and climate benefits becomes especially critical given the elevated rates of deforestation inside concessions. Around two-thirds (66.4%) of forestry and agricultural concessions in this study exhibit a deforestation rate surpassing the background national rate, by a median of 5.3 times (Supplementary Table S1). Such high deforestation underscores the urgency of implementing protections for natural forests on concession land. Economic viability of protecting forests within concessions The economic viability of protecting natural forests in concessions through avoided deforestation projects can be condensed to a central question: whether the price and volume of carbon credits generated by a project are sufficient to offset its cost. Break-even carbon prices for protecting forests, therefore, vary depending on both the carbon credit volume and the cost of a project. We illustrate this relationship through break-even carbon price curves (e.g., Fig. 1), where the credit volume influences the vertical position of the curve—with higher curves indicating a higher carbon price is generally needed to break even—while the project cost determines the point at which the curve is crossed, i.e., the break-even point. Break-even carbon price curves of (A) logging, (B) oil palm, (C) planted wood, and (D) rubber concessions, using concession-specific deforestation baselines. Curves were produced by computing the pooled mean gradient of the break-even carbon price formula across concessions of the same country and concession type (see Methods). Multiplying this gradient by the annual project cost (US$ ha-1 year-1) yields the break-even carbon price (US$ per t CO2). Graphically, this is analogous to projecting vertically from the x-axis (annual project cost) onto the curve, then horizontally to the y-axis to obtain the break-even price. The solid lines denote the mean curves, while the lighter surrounding bands show the uncertainty range of ±1 standard deviation. These were converted from the standard errors of the pooled mean, and shown instead of standard errors to better illustrate the spread of gradients. The bands are staggered for easier visual distinction. Dotted reference lines demonstrate how mean break-even carbon prices are obtained using the mean equivalent discounted annual project costs calculated for each concession type (see Table 2 for these values). Axes are on a log scale. For curves originating from jurisdictional deforestation baselines, see Supplementary Fig. S1. The volume of carbon credits is intended to represent the estimated emissions avoided by the project’s interventions. Its calculation thus includes a deforestation baseline that approximates the “business-as-usual” rate of deforestation and the aboveground carbon density of the threatened forests. Deforestation baselines are usually derived from historical deforestation rates before the inception of a project, with the implicit assumption that these rates would continue in the project’s absence. Changing the spatial scale used to obtain these historical rates yields different baseline variants, which, in turn, affect the final credit volume and the break-even carbon price34. We focus our discussion on concession-specific baselines, which are anchored to each concession’s own historical rates and thereby capture localised deforestation dynamics. This makes them consistent with the project-level approaches and preferences for high spatial differentiation that still dominate baselining practices in the carbon market35,36. However, considering the growing interest in alternative approaches36,37, we also provide details and results for other baseline variants in the Methods and Supplementary Information (Note S2; Tables S2–S3, S5; Figs. S1–S2, S5–S6). We observe large variability in the vertical positions of the break-even price curves between different countries and concession types (Fig. 1). This variability is primarily driven by differences in baseline deforestation rates. For instance, annual deforestation rates in oil palm and rubber concessions are around 1.5 to 12 times greater than the rates in logging and planted wood concessions in the same country (Supplementary Table S1 for specific values). The former two concession types, hence, tend to have positionally lower curves (compare panels A and C to B and D in Fig. 1), reflecting how higher estimates of deforestation and, ergo, potential avoided emissions can make it easier for a project to break even. Differences in forest aboveground carbon densities drive curve variability as well, though their effect is frequently eclipsed by the effect of deforestation rate differences that are often bigger in relative magnitude. For example, although Cambodia’s planted wood concessions have a mean forest carbon density 36–54% lower than their Indonesian and Malaysian counterparts, their 2.5 to 14-fold higher deforestation rate compensates for the fewer potential avoided emissions per hectare, resulting in the lowest break-even curve among the three countries (Fig. 1C). In addition, there is large variability within each concession type in each country, as evidenced by the wide uncertainty ranges of the break-even curves. These ranges incorporate the uncertainty in carbon densities, but also the variation in economic viability among individual concessions. Logging concessions (Fig. 1A) along with Indonesia’s and Malaysia’s oil palm (Fig. 1B) and planted wood concessions (Fig. 1C) have especially wide ranges, indicating that some concessions in these groups require much lower carbon prices to break even than most others in the same group. Our calculation of project costs includes opportunity costs, the returns from forestry and agriculture forgone because the project is pursued instead. Previous reports suggest that the rates of return for oil palm concessions are the highest at US$254.96 ± 56.86 (SE) ha−1 year−1, followed by rubber at US$183.85 ± 86.98 ha−1 year−1, logging at US$80.91 ± 25.12 ha−1 year−1, and planted wood at US$21.95 ± 6.31 ha−1 year−1. As opportunity costs are the major component of project costs, overall project costs closely mirror this ordering (see mean annuities in Table 2), which directly translates into the horizontal positioning of the break-even points (see intersection points of dotted lines in Fig. 1). However, since opportunity costs are assigned assuming static concession types, they may underestimate true project costs and break-even prices in the case of profit-motivated conversions from lower-return to higher-return concession types. In particular, the conversion of logging, planted wood, and rubber to oil palm is a well-recognised phenomenon in the region, fuelled by government policies that facilitate rezoning and by limited legal consequences for speculative forest clearing38–41. Separately, returns may vary appreciably across space within the same concession type; our opportunity costs may thus overgeneralise for some areas (see Methods for elaboration). With greater availability of concession-level data on rezoning and returns, future work could create a spatial model to predict individual concessions’ specific opportunity costs using the probability of inter-type conversion and biophysical and socio-economic characteristics, allowing for more refined cost assignments. Table 2Mean equivalent discounted annual project costs (annuities) and predicted break-even carbon prices, using concession-specific deforestation baselinesConcession TypeMean Annuities (US$ ha-1 year-1)Mean Break-Even Carbon Price (US$ per t CO2) for:IndonesiaMalaysiaCambodiaMyanmarLogging85.47 ± 5.221677.17 (0.00–4673.29)463.35 (0.00–1263.82)53.34 (0.00–108.98)-Oil Palm242.12 ± 11.83319.11 (11.14–626.85)361.48 (0.00–807.47)147.21(27.12–267.06)192.73 (31.48–353.98)Planted Wood32.40 ± 1.3184.08 (0.00–184.40)214.12 (0.00–657.67)32.63 (17.43–47.79)-Rubber178.11 ± 18.09--106.51 (40.08–172.77)-Uncertainties for annuities are expressed as standard errors. Ranges of ±1 standard deviation are presented for break-even carbon prices to better illustrate the spread. Though not practically meaningful as a break-even price, certain ranges extend to zero owing to large propagated uncertainties. Mean equivalent discounted annual project costs (annuities) and predicted break-even carbon prices, using concession-specific deforestation baselines Uncertainties for annuities are expressed as standard errors. Ranges of ±1 standard deviation are presented for break-even carbon prices to better illustrate the spread. Though not practically meaningful as a break-even price, certain ranges extend to zero owing to large propagated uncertainties. The interplay between the aforementioned variables is such that the price advantage from lower project costs can outweigh the disadvantage of a moderately high break-even curve. This is the case for Cambodia’s logging and planted wood concessions, as well as Indonesia’s planted wood concessions. Inversely, price disadvantages from higher project costs can be overcome by low break-even curves, as with Cambodia’s rubber and oil palm concessions and Myanmar’s oil palm concessions. In the end, while the mean break-even carbon prices of these concessions (Table 2) are well above current voluntary market rates (i.e., averagely US$5.30 per t CO2 in 2025 for nature-based avoided emissions)42, they either fall inside the target range forecasted as necessary to fulfil the Paris Agreement climate goals (i.e., US$65–131 per t CO2 by 2030, inflation-adjusted to 2025 USD)42, else are below or close to recent estimates of the social cost of carbon (i.e., the meta-analytically synthesised value of US$283 per t CO243, the U.S. Environmental Protection Agency’s value of US$255 per t CO2 for 202544, and the oft-cited value of US$185 per t CO245). The social cost of carbon represents the monetary damages to society from each additional ton of CO2 emitted. Accordingly, if carbon pricing policies evolve to match the ambitions of the Paris Agreement, and further evolve to more comprehensively internalise social costs of carbon, transitioning to avoided deforestation projects in many of these concessions could become economically viable. Indeed, at a carbon price of US$100 per t CO2, concessions that can competitively transition will contain 142 kha (logging), 20 kha (oil palm), 864 kha (planted wood), and 21 kha (rubber) of natural tropical moist forests (Fig. 2), which together store an estimated 297 Mt CO2 worth of carbon. Doubling the carbon price to US$200 per t CO2 magnifies the area of natural forests that are within competitive concessions—alternately by a factor of 1.6 (planted wood and rubber), 3.1 (logging), and 8.7 (oil palm) (Fig. 2)—raising carbon storage to roughly the equivalent of 584 Mt CO2. Should these concessions successfully pivot to forest protection and avoid emissions, the positive impact on conservation and climate would be immense. Area and extent of natural tropical moist forests in concessions that can competitively transition to avoided deforestation projects at varying carbon price tiers, using concession-specific deforestation baselines and mean carbon prices. Graphs (left) and maps (right) cover A logging, B oil palm, C planted wood, and D rubber concessions in Indonesia, Malaysia (Sarawak), Cambodia, and Myanmar. Carbon prices here refer to the mean break-even values needed to be commensurate with project costs (which include forgone forestry and agricultural returns), calculated under the assumption that concessions rely only on carbon finance to enable transition. In the graphs, forest areas are summed additively across price tiers, whereas the maps colour-code forests discretely according to the lowest break-even price tier for which transition becomes competitive; colours progress from deep cerise through orange-yellow hues to beige with increasing tiers, and forests requiring a break-even price of greater than US$200 per t CO2 are highlighted in pale blue. Concession boundaries are outlined in white. The map in panel D is zoomed geographically to Cambodia, the sole country for which we had rubber concession data. See Supplementary Fig. S2 for results using jurisdictional baselines, and Supplementary Figs. S3–S4 for results using lower-bound and upper-bound carbon prices. Concession layer sources: Global Forest Watch (GFW, host; multiple providers, see refs. 75–81). Cambodia concession data are licensed under CC BY-SA 4.0; other data were licensed under CC BY 4.0 at the time of download, though Indonesia data are currently subject to provider-specific custom licenses. Additional data from OneMap Myanmar82, CC BY-SA 4.0. Forest layer source: European Commission Joint Research Centre (EC JRC)83, licensed under the Copernicus Regulation of the European Commission. Modified by the authors; no endorsement implied. Notwithstanding the above, numerous concessions still require carbon prices greater than most social cost of carbon estimates43 to competitively transition to forest protection. Although certain Indonesian and Malaysian oil palm and logging concessions can do so at prices beneath US$50 per t CO2, the vast majority of these concessions require prices in excess of US$200 per t CO2 (Table 2 and the pale blue areas in Fig. 2A and B). This is attributable to the high returns for oil palm and the generally low deforestation rates for logging. The results here, however, do not imply that forests with such high break-even price concessions cannot be protected or that their conversion should proceed unchecked. Firstly, while conversion to industrial forestry and agriculture can seem enticingly lucrative, the long-term social welfare benefits from protecting forests should be carefully weighed against such short-term monetary gains. Besides providing important services like climate resilience, water purification, flood and erosion control, and pollination, natural forests support local livelihoods with food, medicine, and revenue from non-timber forest products (NTFPs)46,47. Losing the extensive forests remaining in concessions would entail the loss of many of these benefits. Secondly, exploring a wider spectrum of green finance mechanisms beyond avoided deforestation carbon credits can shift the economic calculus, offering opportunities for high-break-even price concessions to competitively conserve their forests. As an example, for logging concessions—which contain large amounts of degraded forest (Table 1)—substantial additional carbon finance could be secured by avoiding degradation (e.g., suspending logging or switching from conventional to reduced-impact logging practices), or by enhancing carbon stocks through the restoration of degraded forests48. Even then, green finance mechanisms outside of carbon markets will be imperative for bridging funding shortfalls when carbon prices are too low to support forest transition on their own. Reinforcing this need to diversify funding streams are recent developments highlighting the vulnerability of carbon markets to market volatility49,50. A possible option for diversification, then, is using green bonds that finance sustainable projects, which have been on the rise in Southeast Asia and climbed to US$12.8 billion in 202251. Another promising approach is blended finance, where private and public sectors share the burden of capital contribution and project risks. Several blended initiatives concentrating on forest protection are underway or planned for the region, including those overseen by the G20 Bali Global Blended Finance Alliance and the Tropical Forest Forever Facility newly launched at COP30, which respectively hope to mobilise US$100 billion and US$125 billion in investments52,53. The nascent markets for biodiversity and nature credits, which recompense the protection of wildlife and ecosystem services, could supplement conservation returns as well, once they are more developed54. Ideally, the protection of natural forests in concessions should not rely exclusively on carbon markets, but rather on a robust multi-instrument funding environment (consisting of carbon markets and various other mechanisms) that holistically rewards the suite of ecological and social benefits these forests deliver. Even with adequate funding from green finance, structural and institutional challenges can yet impede a transition to conservation in concession landscapes. One challenge is market feedbacks and ensuing leakage34,55. If a large number of concessions pivot to forest protection, reduced supply may raise commodity prices and stimulate production elsewhere. Higher prices could also tempt concessionaires to revert to production if green finance becomes comparatively less attractive. Another challenge is land tenure conflict, which is prevalent in forestry and agricultural concessions across Southeast Asia56. Unclear or disputed land claims can create instability for conservation initiatives by complicating enforcement and discouraging long-term stewardship, increasing risks of impermanence57,58. Moreover, designating industrial concessions for conservation can exacerbate existing tensions if wielded to reinforce concessionaire claims at the expense of local and Indigenous communities’ customary land rights59. Finally, multifarious incentives and pressures may prompt concessionaires to continue extractive activities over conserving forests. These include generous policies promoting industry expansion (e.g., tax breaks and increased subsidies)60–62, massive private and corporate shareholder capital as well as bank lending and underwriting directed towards forest-risk commodity sectors63, and the threat of license revocation for underdeveloped concessions64. Addressing these barriers will be crucial for the wide adoption and effective implementation of forest protection in concessions. Domestic market leakage can be mitigated through national or jurisdictional-level accounting34, but international leakage is more difficult to control and will require multilateral cooperation on trade rules (e.g., coordinated product-emissions standards)65. To maintain the competitiveness of conservation even when commodity prices spike, green finance disbursements could be adjusted in response to commodity price fluctuations, perhaps drawing on a buffer reserve accumulated during years of excess funding66. Building on these guardrails, concessions that receive green finance should be contractually obliged to prevent reversion to extraction-induced deforestation (though where such commitments cannot ultimately be guaranteed, temporary delays in emissions can still have climate value and could be correspondingly rewarded67). In parallel, strengthening land tenure security will be paramount. Governments can assist in this goal by passing clearer legislation, boosting transparency in land-use planning, and setting up accessible dispute resolution systems56,68. Concessionaires, meanwhile, must aim to collaborate on conservation with local and Indigenous communities living in or around concessions, co-creating alternative livelihood opportunities (e.g., forest stewardship, restoration work, NTFP enterprises) and arranging for inclusive benefits-sharing (e.g., equitable distribution of green finance returns)69. Such social efforts, which enhance welfare, reduce conflict likelihood, and bolster project longevity, are integral to improving conservation outcomes70. Additionally, policymakers and banks should enact reforms that temper pro-expansion fiscal incentives, such as phasing out harmful subsidies and steering credit and investment towards sustainable intensification63,71. Coupled with redirections of private and corporate capital into green finance mechanisms (e.g., to commit to ESG principles), these actions can counterbalance the appeal of continued extraction. In theory, governments could withdraw underdeveloped concessions and contract directly for carbon outcomes. However, management by existing concessionaires and communities may in many contexts be the pragmatic route, reflecting considerations of property rights, political reception, and the speed and efficiency of deploying conservation at scale72,73. Governments can thus support transitions to conservation by formally permitting concessionaires, local and Indigenous communities and/or other stakeholders to helm forest protection projects in concessions, as well as by streamlining procedures for relicensing concessions as dedicated conservation concessions74. Collectively, these social and regulatory efforts will help shape a favourable concession landscape for implementing forest protections and for ensuring said protections endure. Summary Abundant natural tropical moist forests remain within Southeast Asian forestry and agricultural concessions. These forests play important roles in preserving regional biodiversity and mitigating climate change, yet the majority are facing elevated deforestation rates and are in imminent danger of being lost. We show that carbon finance for avoided deforestation can safeguard large tracts of these forests by empowering concessionaires to pivot from commodity-driven activities to forest protection. In order to fully realise this potential, however, a variety of green finance mechanisms should be leveraged alongside carbon markets, and projects should be buttressed by social and regulatory efforts to achieve broader, longer-term success. Our study enhances understanding of the scale and value of natural forests within concessions, and clarifies pathways for concessionaires, communities, investors, and policymakers to effectively transition these threatened forests towards nature conservation and climate action. Methods Unless otherwise specified, calculations in this study were performed in R version 4.4.3, primarily using the data manipulation package ‘dplyr’ v1.1.4, and spatial analysis packages ‘terra’ v1.8-42 and ‘exactextractr’ v0.10.0. Concession data We investigated four forestry and agricultural concession types: logging, oil palm, planted wood, and rubber. Logging concessions are here defined as lands designated for the felling of natural forests to supply natural wood for the timber industry. This is distinct from planted wood concessions, which are lands designated for establishing fast-growing, managed tree plantations—frequently using species like acacia, teak, and sengon—that are then harvested for a variety of wood products, e.g., timber, wood fibre, and paper pulp. The spatial allocations of concessions in Cambodia, Indonesia, and Malaysia (Sarawak) were retrieved from the Global Forest Watch (GFW) Open Data Portal, and were updated as of 2020, 2018, and 2019 respectively at the time of download75–81. Concession information was provided to GFW by a wide array of sources, including governmental bodies (e.g., Indonesia’s Ministry of Forestry), NGOs (e.g., AidEnvironment, LICADHO, Earthsights Investigations), community associations (e.g., Sarawak Dayak Iban Association), and companies (e.g., APRIL). These concession datasets had often been harmonised or merged across sources. Myanmar’s concession data, updated as of 2018, was retrieved from the OneMap Myanmar Geoportal, a repository for verified government maps hosted by the country’s Ministry of Natural Resources and Environmental Conservation82. This was reprojected to the WGS84 coordinate reference frame to be consistent with other data layers. All in all, our study looked at 2,949 concessions in Indonesia (logging: 557; oil palm: 1,854; planted wood: 538), 572 concessions in Malaysia (logging: 256; oil palm: 248; planted wood: 68), 178 concessions in Cambodia (logging: 3; oil palm: 8; planted wood: 10; rubber: 157), and 55 concessions in Myanmar, which were all oil palm. Compiling accurate and up-to-date concession data is challenging due to highly mutable concession statuses and boundaries, asynchronous spatial mapping, and limited transparency arising from the political sensitivity of land tenure. As such, the concession data used in this study, while assembled from the most recent publicly available maps of individual concessions, is incomplete and may contain areas since redrawn or revoked. The inclusion of some revoked and possibly inactive concessions may have lowered our calculated deforestation rates (see the section “Deforestation rates and baselines”), and somewhat biased our estimated mean break-even carbon prices upwards relative to what would be expected with a fully up-to-date concession dataset (see the section “Modelling break-even carbon prices”). Natural tropical moist forest cover and deforestation data Our analyses relied on the Tropical Moist Forest (TMF) maps83, a 30 m-resolution product that tracks yearly changes to rainforest and moist deciduous forest cover on a pantropical scale. These changes encompass deforestation, degradation, and recovery, permitting the separation of land classes into undisturbed forest (closed evergreen or semi-evergreen forest free from disturbances since 1990), degraded forest (closed evergreen or semi-evergreen forest whose canopy is temporarily disturbed for less than 2.5 years and remains forested after), regrowth forest (previously deforested land that has regrown forest cover for at least three years), and deforested land (previously forested land that has become non-forested after a long-term canopy disturbance lasting more than 2.5 years). The former three classes are regarded as forest by the TMF product, a categorisation we followed as well. The TMF product applies a commodities mask to its forest classes to ensure that they reflect natural forests, rather than agrarian tree cover, as far as possible. This commodities mask combines multiple published data products on tree and agricultural plantations. Further, it is updated annually with new plantation areas visually interpreted from high-resolution images. In deciding to use this map, we thus diminished (but did not eliminate) the chances of developed forests being included in our assessment of natural tropical moist forests within concessions. Although two of the countries we examine (Cambodia and Myanmar) are known to have extensive dry forests, these are not captured in our results owing to the moist forest specificity of the TMF maps. Carbon density data Our analyses employed the ESA Biomass Climate Change Initiative (CCI Biomass) map of forest aboveground biomass densities for 202284, which has a spatial resolution of 100 m. We multiplied aboveground biomass densities by the IPCC default carbon conversion factor of 0.4785 to derive aboveground carbon densities. The CCI Biomass product, which synergises multiple Earth observation datasets through a standardised retrieval methodology, is more temporally updated than other forest biomass or carbon density maps while offering finer spatial granularity. Moreover, it is accompanied by per-pixel standard deviations, making it one of the few global products covering all the studied countries that has a dedicated uncertainty layer86. Nevertheless, like many satellite-based products, the CCI Biomass map is known to systematically underestimate densities in high-biomass regions, e.g., dense tropical forests, making explicit treatment of uncertainty essential for our study87. It is also affected by strong spatial autocorrelation, necessitating non-conventional uncertainty propagation (see Supplementary Note S1). Quantifying natural forests, carbon stocks, and biodiversity in concessions Some concession boundaries from different concession types overlapped in space. This could stem from several factors: conflicting land claims, maps that were not revised after concessions were revoked or redrawn, mistakes during the digitisation process, or genuine mixed use, e.g., oil palm interspersed with planted wood cultivation. To prevent the duplication of concession expanses—and in turn forest expanses—in our calculations involving area summation, we created a fifth concession type representing this overlap. The overlapping concession layer consisted of all intersecting areas between the original four concession types, along with a class of concessions labelled “mixed rubber” in Cambodia’s rubber concession data. The overlapping layer was subsequently masked from the original concession types, generating four new non-overlapping layers containing concession lands purely designated for logging, oil palm, planted wood, or rubber. Concession layer processing at this stage was carried out in QGIS version 3.26. By conducting zonal statistics with the TMF map for 2024—the latest available year of TMF data at the time of this study—we obtained the cumulative area of undisturbed, degraded, and regrowth natural tropical moist forests for each concession type in each country we examined. To obtain the percentage of natural forest coverage in the five concession types, we divided the cumulative natural forest areas by the total areas of their respective concession types. Next, we clipped CCI Biomass-derived aboveground carbon densities to the natural forest extents and computed a mean carbon density for undisturbed, degraded, and regrowth forests per concession type in each country. Uncertainty propagation for these mean carbon densities is described in Supplementary Note S1. Carbon density values (Mg C ha−1) were multiplied by corresponding cumulative natural forest areas (ha) to arrive at carbon stocks (Mt C). We determined the amount of these carbon stocks that would be released over 30 years by summing, across all concessions, the product of the total proportion of the original forest lost (estimated from concession-specific deforestation rates), the corresponding natural forest area, and the corresponding mean carbon density—see the succeeding section for deforestation rate calculation details. Amounts were stoichiometrically converted to get equivalent carbon dioxide emissions (Mt CO2). We evaluated biodiversity by identifying which ecoregions in the Ecoregions2017 map27 coincided with the natural forest extents, noting distinct forest biomes. We also retrieved KBAs from The World Database of Key Biodiversity Areas29, and computed the area of intersection between natural forest extents and the polygonal KBAs. Deforestation rates and baselines We used the areas of undisturbed, degraded, and regrowth natural forests from years 2014 to 2024 of the TMF product to calculate annual deforestation rates (% year-1). The formula we adopted was the Puyravaud annual rate of deforestation, which is derived from the exponential decay model for the Compound Interest Law and thereby accounts for shrinking forests as deforestation progresses88. It is expressed as DR=[1/(t2−t1)]×ln(A2/A1), which takes the start (t1) and end year (t2) of the period for which deforestation rate is calculated over, as well as the forest areas (A1 and A2) in those years. Based on this formula, DR is a continuous rate of change that is always ≤ 0, and the total proportion of the original forest lost after i years can be calculated as 1−eDRi. The volume of carbon credits issuable to an avoided deforestation project is determined in part by the estimated difference between the project's deforestation rate and a deforestation baseline. There is no universally accepted way of building a baseline, with carbon market participants trialling a diverse array of baselining techniques36,89. This diversity can be ascribed, among other factors, to divergent conceptions of appropriate deforestation reference areas and time periods. We tested four deforestation baselines from deforestation rates calculated at different scales of reference areas: national baselines, jurisdictional baselines, type-specific baselines, and concession-specific baselines. Refer to Supplementary Note S2 for in-depth descriptions, justifications, and processing steps for each deforestation baseline tested. The baselines in this study are predictive (i.e., ex-ante), congruent with baseline methodologies prescribed by most carbon standards. They are therefore well-suited for assessing economic potential, but might not reliably characterise true environmental impact, which is better assessed ex-post using contemporaneous deforestation rates and project performances90. Project opportunity costs Opportunity costs in this study were the financial returns from forestry and agriculture that would be forgone as a result of concessions transitioning to avoided deforestation projects. We collated returns (e.g., Net Present Values) from logging, oil palm, planted wood, and rubber industrial concessions by reviewing published literature, restricting our search to papers that based their numbers on real Southeast Asian concessions, rather than best guesses. We strove to gather more recent numbers, and as such, most of our source papers were dated to the 2010s–2020s (year range: 2007–2023). We converted returns that were in the currency of their native country to US$ using 2023 exchange rates. We annualised returns for papers that provided aggregated returns over the course of a growth rotation, but for which the time to harvest for the crop in question was known. For papers that reported returns in the form of a range, we took the mean of the upper and lower limits. If a paper reported a series of returns from several alternate scenarios, we took the value from the scenario that either most closely matched our research’s assumptions (e.g., 10% discount rate) or assumed the maximum crop productivity (e.g., 76%–100% capacity). All in all, 5, 9, 9, and 10 return values were used to inform our mean estimates of rubber, logging, oil palm, and planted wood annual opportunity costs, respectively. The source papers are documented in Supplementary Table S6. Because area-based returns for industrial oil palm were available only for Indonesia and Malaysia, the mean oil palm opportunity cost we applied may overstate true costs in Myanmar and Cambodia, where less extensive oil palm development could imply lower returns. Our break-even carbon prices for oil palm concessions in these two countries (computed in the succeeding section) should therefore be interpreted as being conservatively high, in the absence of more representative return values. In addition, we did not consider the opportunity cost of the overlapping concession type due to the ambiguity of what these concession areas represent. We consequently refrained from modelling break-even carbon prices of the overlapping type, and used the original unmasked concession layers for the downstream analyses. Modelling break-even carbon prices Philipson et al.91 developed equations for the purpose of calculating carbon prices needed to offset forest restoration costs. We modified these equations to attain a new formula catered to avoid deforestation projects. The break-even carbon price for a project protecting natural forests in a single concession, BE (US$ per t CO2), is thus: The denominator of Eq. (1) sums the carbon credit revenue accrued over the project lifespan of n years (t CO2 ha-1). The numerator of Eq. (1) sums the total project cost over the project lifespan (US$ ha-1). Both allow for a risk-adjusted discount rate of r. Annual project cost, xi, encapsulates establishment, maintenance, and opportunity costs. Carbon credit revenue in a single year, CCi, is expanded in Eq. (2). EI is the Effectiveness Index of the project which weights its success at avoiding deforestation, FC is the forest carbon density in the concession (Mg C ha-1), and BC is the proportion of buffer credits set aside to address the risk of non-permanence. DR is the annual deforestation rate calculated using the Puyravaud formula (see the section “Deforestation rates and baselines”); the eDR(i−1)−eDRi component of Eq. (2), therefore, represents the proportion of the original forest lost exclusively in the ith year of the project. To get FC for individual concessions, we excluded non-forested expanses from the carbon density layer using the TMF 2024 map, then conducted zonal statistics at the level of concessions. Uncertainty propagation for FC proceeded as outlined in Supplementary Note S1. DR was adjusted according to the deforestation baseline tested and reflected a 10-year average from 2014 to 2024. We fixed a project lifespan of 30 years, a discount rate of 10%, and a 20% buffer for credits, following previous literature’s appraisals of default values from the carbon market praxis50. We also fixed the Effectiveness Index at a constant 80% for simplicity, based on a survey of real values used by avoided deforestation projects in Southeast Asia. Nonetheless, as part of our sensitivity analysis, we trialed different Effectiveness Indices (70%, 90%, 100%), buffer credit proportions (10%, 30%, 40%), discount rates (5%, 8%, 15%), and number of years for estimating the baseline deforestation rate (5 years, i.e., 2019–2024). Relevant results for these sensitivity scenarios are presented for concession-specific and jurisdictional baselines in Supplementary Tables S4 and S5, respectively. For each concession, we calculated the denominators of Eq. (1) for the four deforestation baselines we tested. We next variously grouped concessions by country, concession type, and/or the baseline tested. We modelled the pooled mean denominator and associated standard error of each of these groups by employing the R package ‘metafor’ v4.8-0 to perform a random-effects meta-analysis. Models used restricted maximum likelihood to account for both within- and between-concession variance. We then computed a pooled mean gradient of n / pooled mean denominator, where the gradient is the first component in this rearrangement of Eq. (1): With these gradients, we plotted break-even carbon price curves over an assumed range (i.e., 0 to 1000 US$ ha-1 year-1) of equivalent discounted annual project costs, otherwise known as annuities (the second component of Eq. (3) above). Standard errors of the mean denominators were first converted to standard deviations, then propagated analytically to obtain the ranges of uncertainties in gradients. These results are visualised in Fig. 1 and Supplementary Fig. S1. We report gradients for each country, concession type, and baseline combination, so as to accommodate BE calculation for custom project costs and baseline choices (Supplementary Table S2). To predict the carbon prices that would make transitioning to avoided deforestation projects competitive, we estimated annuities per concession type, a (US$ ha-1 year-1): where EC is the one-time establishment cost (expenses for project design, enforcement, surveying, etc.), MC is the annual maintenance cost (expenses for monitoring, administration, sustainable livelihoods, etc.), PP is the productive proportion of concession land from which commodities can be cultivated and/or extracted, and OC is the mean concession type annual opportunity cost estimated in the section “Project opportunity costs”. We set the establishment cost to US$25 ha-1 and the maintenance cost to US$10 ha-1 year-1, once again following previous literature50. We set a productive proportion to 90% based on concessions’ license obligations and reported non-production area, assuming a higher end of productivity for conservativeness in our break-even analysis. We plugged gradients and annuities into Eq. (3) to yield BE per concession type in each country, as well as for individual concessions. In the second case, uncertainties were also propagated from standard errors for annuities, computed from the variation in annual opportunity costs for each concession type. These were statistically combined with gradient uncertainties to obtain an overall standard error of BE per individual concession. Overall standard errors were then added to or subtracted from the mean concession BE to create upper-bound and lower-bound estimates, respectively. The mean break-even prices were used to categorise concessions that could competitively pivot to avoided deforestation projects at carbon prices of 25, 50, 75, 100, 125, 150, 175, and 200 US$ per t CO2, or else if a price exceeding US$200 per t CO2 was necessary. We calculated and mapped the cumulative areas of natural forests in the competitive concessions at each price tier (Fig. 2 and Supplementary Fig. S2). Analogous mappings were produced using lower-bound and upper-bound break-even prices (Supplementary Figs. S3–S6). Only concession-specific and jurisdictional baselines were assessed for this last phase of the analyses, as they are most conceptually consistent with the baselines favoured by carbon market participants at present36,37. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Peer Review file Reporting Summary Acknowledgements Our gratitude goes to R.A. Chisholm and D.C.J. Yeo for their helpful suggestions and feedback. L.P.K. is supported by the Singapore National Research Foundation (NRF-RSS2019-007) and the Singapore National Research Foundation and Agency for Science, Technology and Research LCER Phase 2 funding programme (U2303D4101). Z.Y. is supported by the NTU Start Up Grant (03INS002001C210). A.J.Y.L. is supported by the Lee Kong Chian Scholarship and the President’s Graduate Fellowship at the National University of Singapore. H.C.T. and J.D.T.D. declare no relevant funding. Author contributions L.P.K. and Y.Z. conceived and supervised the study. A.J.Y.L., Y.Z., and J.D.T.D. developed the methodological workflow. A.J.Y.L. and H.C.T. carried out the analyses. A.J.Y.L., H.C.T., and J.D.T.D. collected the data. A.J.Y.L. led the writing and produced the figures. All authors (A.J.Y.L., Y.Z., H.C.T., J.D.T.D., L.P.K.) were involved in writing and editing. Peer review Peer review information Nature Communications thanks Robert Heilmayr, Diana Parker and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Data availability The concession data supporting the findings of this study are available from Global Forest Watch Open Data Portal at https://data.globalforestwatch.org/ (country- and concession type-specific weblinks are listed in the References), as well as the OneMap Myanmar Geoportal at https://www.onemap-geoportal.gov.mm/. Other data supporting this study are obtainable from the EC JRC’s Tropical Forest Monitoring site at http://forobs.jrc.ec.europa.eu/TMF, the CEDA Archive repository for CCI Biomass with the identifier 10.5285/95913ffb6467447ca72c4e9d8cf30501, and the World Database of Key Biodiversity Areas at https://www.keybiodiversityareas.org/ (available upon request). Break-even carbon price gradients calculated for individual concessions and for all combinations of sensitivity scenarios are deposited in the Zenodo repository at 10.5281/zenodo.19279762. Code availability The R scripts used in this study’s analyses can be downloaded from the Zenodo repository at 10.5281/zenodo.19279762. Competing interests J.D.T.D. is presently employed at an environmental services company developing nature-based carbon projects; this employment began after his contributions to this study were completed. All other authors (A.J.Y.L., Y.Z., H.C.T., L.P.K.) declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors jointly supervised this work: Yiwen Zeng, Lian Pin Koh. Contributor Information Annabel Jia Yi Lim, Email: lim.annabel@u.nus.edu. Yiwen Zeng, Email: yiwen.zeng@ntu.edu.sg. Lian Pin Koh, Email: lianpinkoh@nus.edu.sg. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-72921-8. References Associated Data Supplementary Materials Supplementary Information Peer Review file Reporting Summary Data Availability Statement The concession data supporting the findings of this study are available from Global Forest Watch Open Data Portal at https://data.globalforestwatch.org/ (country- and concession type-specific weblinks are listed in the References), as well as the OneMap Myanmar Geoportal at https://www.onemap-geoportal.gov.mm/. Other data supporting this study are obtainable from the EC JRC’s Tropical Forest Monitoring site at http://forobs.jrc.ec.europa.eu/TMF, the CEDA Archive repository for CCI Biomass with the identifier 10.5285/95913ffb6467447ca72c4e9d8cf30501, and the World Database of Key Biodiversity Areas at https://www.keybiodiversityareas.org/ (available upon request). Break-even carbon price gradients calculated for individual concessions and for all combinations of sensitivity scenarios are deposited in the Zenodo repository at 10.5281/zenodo.19279762. The R scripts used in this study’s analyses can be downloaded from the Zenodo repository at 10.5281/zenodo.19279762.