Synergies and trade-offs for reaching global biodiversity and climate commitments

Abstract

The Kunming-Montreal Global Biodiversity Framework strives to expand high-quality protected areas globally, yet faces economic constraints, climate change and human pressures. This requires integrating biological and socio-economic complexity while ensuring equity. We explore multi-target ‘30 by 30’ and ‘50 by 50’ (the ‘Half Earth’) protected area expansion strategies that combine species diversity with carbon storage, quantifying their effectiveness (the represented proportion of global species richness and carbon stocks), connectivity and cost-efficiency. Under near-current and future climates, single-criterion strategies – based on five taxonomic groups (plants, amphibians, birds, mammals, and reptiles) or carbon alone – achieve effectiveness of 42.3 ± 4.10% and 68.8 ± 8.39% for both targets, whereas multi-goal strategies reach 44.1 ± 0.42% and 67.0 ± 0.89%. Multi-goal strategies reduce connectivity by 2.35% but slightly improve cost-efficiency (2.06% lower cost) and individual-metric effectiveness (mean rise 0.76%). Moving from single- to multi-target strategies increases conserved-area coverage in high-latitude, higher-income countries (by 3.14% and 5.59%), with a corresponding decline in low-income countries (7.31% and 12.6%); single-goal strategies targeting threatened species allocate more coverage to the Global South. By rebalancing North-South responsibilities and easing burdens on low-income nations needing financial support or compensation mechanisms, these findings improve conservation affordability and provide equity insights for implementing Global Biodiversity Framework and Paris Agreement.

Expanding global protected areas is critical for biodiversity and climate goals. This study finds multi-goal strategies shift burdens from the Global South to higher-income nations, improving conservation, cost-efficiency and distributional equity.

Introduction

Our planet’s biodiversity is increasingly threatened by global climate change and human activities^1,2, leading to the worst state of nature in human history³. These impacts affect the intimate and complex relationship between biodiversity and ecosystem services, directly reducing nature’s contributions to people⁴. While human-driven changes are ubiquitous, affecting all continents and ecosystems, global biodiversity hotspots – originally defined as regions containing high levels of species richness, endemism and threat⁵, but used here more generally as areas containing high levels of biodiversity – are being particularly affected⁶, meaning that coordinating the complex nexus of conservation action is crucial to prevent further biodiversity loss. To halt the ongoing degradation of nature and bend the curve of biodiversity loss, scientists widely agree that protected areas are paramount for safeguarding biodiversity and ensuring the benefits of ecosystem services^7–10. By September 2025, the terrestrial coverage of protected areas and other effective area-based conservation measures (OECMs) had reached 17.53%, while marine coverage stood at 9.85%¹¹. However, a sole focus on area and quantity carries serious potential pitfalls¹², and risks accentuating the decline in quality and functional degradation of protected areas¹³. Consequently, the synergistic quality of multiple ecological functions – provided by a wide diversity of species, organism groups, and functional roles – remains poorly understood.

To tackle the biodiversity crisis, the Kunming-Montreal Global Biodiversity Framework – hereafter ‘the Framework’ – has embraced diverse social and ecological perspectives, setting its nearly 200 state signatories on a pathway towards high-quality, equitable and effective conservation. Target 3 of the Framework commits to protecting at least 30% of terrestrial and marine areas by 2030 (‘30 by 30’)¹³. It focuses on conserving areas crucial for biodiversity and ecosystem functions and services through being ecologically representative and well-connected. Meanwhile, it advocates for equitable public governance regarding the quality of conservation expansion^13,14. However, under environmental and climate change, rising human demand for land has made it increasingly difficult to balance equity with high-quality conservation efforts^15,16. This difficulty further intensifies the challenge of delivering the Framework goals and decision-making strategies. Given global climate change and land use, the synergistic effectiveness and global equity of protected area expansion for national participation remain uncertain.

A successful achievement of the Framework targets relies on integrating multiple conservation goals, which are essential for addressing biodiversity loss¹⁷. In parallel, as we recently marked the 10^th anniversary of the 2015 Paris Agreement on climate change – aiming to keep the rise in global temperature to ‘well below 2 °C above preindustrial levels and to pursue efforts to limit the temperature increase to 1.5 °C’ – we simultaneously must focus on cutting the emission of greenhouse gases, including those derived from deforestation and other forms of habitat destruction, and removing carbon dioxide out of the atmosphere through measures such as increased conservation and restoration. Although global priority mapping has been criticised as risking alienating local contexts and community engagement¹⁸, it is a powerful tool for exploring synergies and trade-offs related to multiple and sometimes competing global commitments for nature and climate, while considering costs and benefits^15,17,19.

Many natural and anthropogenic factors have impacted the efficiency of existing protected areas, while the lack of species survey data has led to notable conservation gaps and a shortfall in systematic conservation planning^1,20. Conservation assessments have often been limited to one or a few taxonomic groups^3,20, and often use proxies for identifying priority areas, such as species diversity^21,22 and phylogenetic diversity^23,24, even though the diversity of one group may not mirror that of another²⁵. The prioritisation of areas for conservation can also vary depending on the methodology applied and data utilised^26,27. With the emergence of global conservation frameworks, traditional strategies are shifting towards multidisciplinary integration and synergy. Conservation effectiveness is now evaluated through synergistic relationships among weighted indicators^15,17, such as biodiversity, ecosystem services, landscape connectivity, and human footprint^9,15,28,29. However, in a dynamic environment, the prioritisation criteria used in these assessments are shifting, and the value of protected areas continues to be enriched through ongoing social and cultural re-evaluations^13,21. For instance, whereas earlier protected areas often targeted specific species or ecosystems, current approaches emphasise multi-objective strategies that integrate biodiversity and ecosystem services, highlighting their systemic conservation. Examples include the added value now being attributed to protected areas in proximity to, or integrated with, urban areas – which can support environmental education and natural ecosystem functions, although those aspects may not have been considered for selecting areas for protection initially^30–32.

The uneven global distribution of economic resources is largely a legacy of human history and the perpetuation of systemic inequities^33,34. Against this background, Target 3 of the Framework advocates recognising indigenous and traditional territories and promoting equitable governance of protected areas³⁵, implying a need to uphold equity in biodiversity decision-making. In the context of conservation science, equity relates primarily to the allocation of costs, responsibilities, and benefits among stakeholders^36,37. However, global biogeographical patterns – ultimately driven by geological and environmental factors such as tectonic uplifts and ocean currents – influence the uneven spatial distribution of biodiversity and ecosystem services^38,39. These heterogeneities in the distribution of biodiversity risk lead to unfair or hard-to-implement policy recommendations, which could further exacerbate global economic inequities.

In previous conservation frameworks, each country often adopted the same protection targets in line with global percentage goals⁴⁰. That approach continuously overlooked the uneven distribution of biodiversity and carbon storage, and the inequities among nations and regions of varying economic wealth⁴¹. These inequities are also reflected in diverse states of environmental degradation and future risks¹⁵. For example, climate change amplifies the vulnerability of low-income countries to environmental threats, and limited resources for monitoring, assessment, and capacity building may exacerbate inequities in climate negotiations over biodiversity conservation^14,42.

Against this background, this study aims to identify priority conservation areas to support biodiversity conservation and carbon storage in response to environmental change, thereby providing spatial insights to inform equitable decisions on priority expansion. We analyse climate change projections of habitat dynamics for threatened species to gauge the effectiveness (defined here as the percentage coverage of indicators resulting from the expansion of protected areas, ecological connectivity and conservation costs) of the 30% terrestrial targets of the Framework and a long-term goal of 50% target (‘50 by 50’) – the ‘Half Earth’ scenario promoted by E. O. Wilson⁴³. We combine global species distribution data and environmental variables to infer high-resolution (1-km) patterns of threatened biodiversity, while accounting for complementarity in our analysis. We develop seven conservation expansion scenarios based on threatened biodiversity and carbon storage, optimise them using functions from the recently developed CAPTAIN software², and analyse conservation gaps and effectiveness of existing protected areas, as well as their land-use threats. While not fully representing all aspects of biodiversity, threatened species are often sensitive indicators of climate change and environmental alterations^44,45. Our quantitative analysis explores differences and challenges in applying scenarios nationally, aiming to identify equitable conservation policymaking.

Results

Conservation scenarios and dynamics under climate change

We employ widely used Maxent models to quantify the potential distribution of threatened species and predict their habitat range shifts under climate change scenarios for 2030 and 2050. We analyse seven conservation scenarios for protected area expansion based on single or combined considerations of threatened plants, threatened animals, and carbon storage (Methods for details). These scenarios identify priority areas for conservation expansion by constraining existing protected areas and allocating additional priority zones in the remaining unprotected land, targeting the top 30% and 50% of global hotspots for biodiversity and carbon storage. Gap analysis is used to assess the extent to which conservation hotspots are covered by existing protected areas, as well as under seven expansion scenarios targeting the top 30% and 50% priority areas. Hereafter, “effectiveness” denotes the proportion of conservation hotspots covered by existing protected areas or designated expansion areas, measured relative to global totals of relevant indicators such as threatened biodiversity or carbon storage (We acknowledge that this term is used differently across the scientific literature, such as referring to how well species and ecosystems fare within a protected area as compared to in reference sites; e.g., Wauchope, et al.⁴⁶ and Rasolofoson, et al.⁴⁷).

Our analyses reveal spatial differences in the conservation patterns of biodiversity and carbon storage scenarios (Fig. 1). Results indicate that across all scenarios, expansions are predicted to be relatively more effective in South America, where coverage reaches 54.4 ± 4.3% under the ‘30 by 30’ target and 78.7 ± 14.9% under the ‘50 by 50’ target, followed by higher protection coverage in Oceania and Africa (Supplementary Data 5). The percentages indicate the total land area under protection (the sum of both newly proposed and existing protected areas) as a proportion of each continent’s total landmass. The ± values indicate the standard deviation calculated across the different conservation scenarios within each continent. Under scenario 4 (biodiversity), where we prioritise the expansion of protected areas based on the integration of threatened plants and animals (Fig. 1d), we find that hotspots are predominantly predicted in low-latitude or tropical regions. Specifically, the coverage reaches a mean of 57.9 ± 2.38% and 94.3 ± 2.00% in South America (notably with less prominent hotspots in the Amazon basin), 30.2 ± 0.44% and 52.1 ± 0.38% in Africa, and 21.0 ± 2.94% and 37.7 ± 4.64% in Asia for ‘30 by 30’ and ‘50 by 50’ targets (Fig. 1 and Supplementary Data 5). Carbon storage priorities, in addition to aligning with biodiversity hotspots, also exhibit a distinct distribution in the higher latitudes of the Northern Hemisphere (in Europe, multi-target coverage increases the protected area expansion by 23.8% and 98.5% compared to considering only biodiversity for ‘30 by 30’ and ‘50 by 50’ targets, while in North America, the coverage was increased by 88.1% and 34.5% (Supplementary Data 5). Common to all scenarios is prioritising conservation hotspots in low-latitude and tropical regions, notably near the Equator and coastal regions within South America, Africa, and Southeast Asia.

Fig. 1. Global conservation priorities for the ‘30 by 30’ target derived from joint biodiversity and carbon storage optimisation, integrating potential threatened species suitability patterns across the 1970–2000 bioclimatic baseline and projected 2030 and 2050 conditions.

‘50 by 50’ priority patterns are presented in Supplementary Fig. 2. Expansion priority areas are identified by combining existing protected areas (approximately 16% of land) with designated conservation hotspots, fulfilling the global terrestrial conservation targets of ‘30 by 30’ and ‘50 by 50’, respectively. Seven expansion scenarios are explored for priority areas, each based on single or combined considerations of threatened plants, threatened animals, and carbon storage. Grey areas, referred to as Non-priorities within PAs, represent areas within existing protected areas that are not designated as priority hotspots; expansion hotspot priorities outside protected areas are shown in purple (Unprotected priorities), whereas priorities within existing protected areas are shown in green (Protected priorities). Under the expansion scenario assumptions, Scenario 1 targets threatened plant diversity exclusively (a); Scenario 2 targets threatened animals (b); Scenario 3 targets carbon storage (c); Scenario 4 integrates threatened plants and animals (d); Scenario 5 integrates threatened plants and carbon storage (e); Scenario 6 integrates threatened animals and carbon storage (f); and Scenario 7 represents a comprehensive combination of plants, animals, and carbon storage (g). Carbon storage is based on current levels and does not account for future changes (Supplementary Fig. 7). The maps were created using ArcGIS Pro software by Esri under license. Copyright ^© Esri. All rights reserved.

Distribution-based projections of threatened species – an essential component of the biodiversity hotspot concept – indicate that by 2030 and 2050, biodiversity hotspot patterns may shift from inland to coastal regions (e.g., in the United States and China) and show declines in equatorial hotspots (Fig. 1, Supplementary Note 4 in Supplementary Information). In addition, the global expansion strategy reveals a core–periphery pattern, in which conservation hotspots in equatorial and low-latitude regions partially overlap with existing protected areas, while adjacent unprotected hotspots delineate priority zones for stepwise outward expansion (Fig. 1 and Supplementary Fig. 13).

Gap analysis of different scenario combinations reveals a significant mismatch between conservation hotspots and the locations of established protected areas (Figs. 1 and 2). Among all scenarios assessed, less than 17.9% of priority areas are classified as protected priorities, representing the proportion of expansion hotspots currently covered by protected areas. In contrast, approximately 43.3% of the expansion area falls within unprotected priorities (which increases to 65.9% in the ‘50 by 50’ target), while more than 39.5% corresponds to non-hotspots within protected areas (decreasing to 16.1% in the ‘50 by 50’ target) (Fig. 2). Multi-goal conservation scenarios demonstrate a higher coverage of priority areas than scenarios focusing on a single goal or solely on a biodiversity target (scenarios 1, 2 and 4) (Figs. 1 and 2). Scenario 3 and those incorporating carbon storage exhibit a higher proportion of protected priorities and a lower proportion of non-hotspot areas within protected areas.

Fig. 2. Conservation gaps across seven conservation expansion scenarios under ‘30 by 30’ and ‘50 by 50’ targets.

Gap analysis is performed by overlaying the required expansion hotspots (under the respective targets) with existing terrestrial protected areas (excluding OECMs, based on the WDPA data), thereby identifying protected priorities, unprotected priorities, and non-hotspots within protected areas.

Conservation costs and effectiveness for different scenarios

Achieving the ‘30 by 30’ target requires an additional expansion of 1,889 million hectares (Mha) of land (approximately twice the size of China), while the ‘50 by 50’ target would necessitate 4,581 Mha (roughly the size of Asia) (Fig. 2 and Supplementary Data 4). Approximately 16% of global terrestrial protected areas achieve an average conservation effectiveness of only 17.7 ± 2.20%, measured as the percentage of functional coverage across five taxonomic groups and carbon storage. These areas aim to safeguard threatened plants and tetrapod animals (reptiles, mammals, birds, and amphibians) and carbon storage (Fig. 3 and Supplementary Data 7). In contrast, for the combined average of species groups and carbon storage, boosting the hotspots to meet the ‘30 by 30’ target could increase conservation effectiveness to a mean of 42.3 ± 4.10% for single-goal and 44.1 ± 0.42% for multi-goal conservation. Achieving the ‘50 by 50’ target could boost comprehensive effectiveness to at least 59.2 ± 11.9%, with single-goal conservation peaking at a mean of 68.8 ± 8.39%, while multi-target strategies maintain 67.0 ± 0.89% effectiveness (Fig. 3a–c and Supplementary Data 7). Across both the ‘30 by 30’ and ‘50 by 50’ targets, multi-target strategies show a modest overall advantage, increasing mean effectiveness by 0.76% compared with single-goal targets.

Fig. 3. Conservation effectiveness of seven different scenarios and existing protected areas in terms of biodiversity, carbon storage, ecological connectivity, and conservation cost.

Radar charts showing conservation effectiveness of threatened species (plants, amphibians, birds, mammals, reptiles) and carbon storage with baseline (a), 2030 (b), and 2050 (c) time periods. Forest habitat connectivity, measured by the equivalent connected patch (ECP) (d), and landscape connectivity, quantified by the patch cohesion index (PCI) (e), are employed to evaluate ecological connectivity under each scenario. Conservation cost is quantified as the proportion of impervious surface (cement-hardened surfaces like paved roads, buildings) (f) and cropland (g) within protected or prioritised areas relative to their total area; higher proportions indicate greater overlap with human land use and therefore higher restoration costs. The quantification of these data (d–g) is the result of plotting with SSP1-2.6, where SSP2-4.5 and SSP5-8.5 are indices that reflect increases or decreases in relation to SSP1-2.6. Changes between SSP models are quantified as index changes.

It does not seem possible to fully maximise the synergies between biodiversity and carbon storage, as they exhibit clear trade-offs (in line with other studies; e.g., Strassburg, et al.¹⁹ and Silvestro, et al.⁴⁸). When biodiversity hotspots are prioritised (scenario 4), average conservation effectiveness reaches 46.5% and 74.8% for ‘30 by 30’ and ‘50 by 50’ targets, respectively. In contrast, prioritising carbon storage (scenario 3) yields effectiveness values of 37.6% and 59.2% (Figs 3a–c, and Supplementary Data 7). Importantly, prioritising carbon storage expansion increases carbon effectiveness by an average of 22.0% (percentage of change) but sacrifices biodiversity by 27.8%. In contrast to scenario 4, changes in the effectiveness of multi-goal strategies fall between the priority of biodiversity and carbon storage alone. For example, the effectiveness of scenario 7, which focuses on the integration of threatened species with carbon storage, shifts to a 21.1% increase in carbon storage and a relatively smaller – 12.9% – decrease in biodiversity.

Multi-goal conservation offers a more balanced level of benefits than single targets, covering threatened species, carbon storage function, connectivity, and cost efficiency. In particular, multi-target scenarios show a total decrease of 2.35% in connectivity metrics – equivalent connectivity probability (ECP, which measures the connected area proportion of forest patches) and patch cohesion index (PCI, reflecting physical patch connectedness at the landscape level) – compared to single-objective effectiveness. The ECP for forest habitats in all scenarios outperforms the connectivity of existing protected areas (14.4% for protected areas, a mean of 34.6 ± 5.50% for all targets). Meanwhile, landscape connectivity (PCI) slightly exceeds that of existing protected areas (91.4%) and shows improvement across the seven scenarios over three time periods, with mean values of 91.8 ± 0.69% under the baseline condition, 93.7 ± 0.47% for 2030, and 93.3 ± 0.59% for 2050. This temporal improvement indicates that spatial configuration optimisation ensures the inclusion of more aggregated and contiguous patches within the expansion selection.

As mainstream land use expands, conservation costs in priority areas are expected to increase rapidly, led by the growing share of impervious surface and cropland within the proposed protected area expansion (Fig. 3). Here, cost is quantified as the proportion of impervious surface and cropland within protected or prioritised areas relative to their total area. The impervious surface (cement-hardened surfaces like paved roads, buildings) within the expansion areas is projected to increase from 0.01% to a mean of 0.03 ± 0.00% and 0.19 ± 0.06%, while cropland increases from 0.85% to a mean of 1.86 ± 0.53% and 5.72 ± 2.21% under the ‘30 by 30’ and ‘50 by 50’ targets (Fig. 3f and g). Different SSPs reveal that green and sustainable pathways (SSP1-2.6) can mitigate the future opportunity cost for conservation. For instance, compared with other SSPs, the combined share of impervious surface and cropland under SSP1-2.6 is on average 27.9% and 14.8% lower for the ‘30 by 30’ and ‘50 by 50’ targets, respectively (Supplementary Data 7). Scenarios 1 and 2, focusing on threatened plants and animals respectively, require the highest conservation costs (average 0.11 ± 0.10% and 4.80 ± 2.88% compared with 0.04 ± 0.02% and 1.73 ± 0.78% for carbon storage on impervious surface and cropland, respectively). In contrast, scenarios 5, 6, and 7, which synergistically integrate carbon storage, show lower costs than a single-target scenario (a total mean decrease of 2.06%). Furthermore, the green pathway (SSP1-2.6) outperforms fossil-fuelled development (SSP5-8.5) in ecological connectivity and cost-efficiency, highlighting the significance of land-use policies and management in shaping protected areas.

Global terrestrial protected area coverage targets

Global conservation scenarios are mapped onto 203 countries, evaluating conservation outcomes and their cost-efficiency (Fig. 4 and Supplementary Data 7). We find that under multi-goal scenarios, protected area expansion coverage increases by 3.14% and 5.59% for high-income countries when shifting from single to multi-target approaches for the ‘30 by 30’ and ‘50 by 50’ targets, respectively, while low-income countries experience decreases of 7.31% and 12.60% (Supplementary Data 6). In contrast, single-species-focused scenarios require a disproportionate share of priority area expansion in tropical and low-latitude countries, e.g., at least 50% land expansion. In some instances, this requirement approaches unrealistic 100% coverage, e.g., Madagascar, Cameroon, Laos, New Zealand (Supplementary Data 4).

Fig. 4. Differences in percentages of country territories required to meet globally set biodiversity targets ‘30 by 30’ under seven expansion scenarios, together with the current coverage of existing protected areas.

The percentage (%) measures the ratio of prioritised area within the country to the national area under the global targets. In the legend’s segmentation thresholds: 16% corresponds to existing terrestrial protected areas alone (excluding OECMs, based on the WDPA data); 25% is the approximate area of the latest expanded protected areas and other effective area-based conservation measures (OECMs) announced by the WDPA; 30% and 50% are the expanded areas of ‘30 by 30’ and ‘50 by 50’ targets, respectively; 40% is a threshold quantifying the transitional phase between 30% and 50%. The maps were created using ArcGIS Pro software by Esri under license. Copyright ^© Esri. All rights reserved.

Existing protected areas show that coverage in most countries is below 16%, with only a few achieving the 30% and 50% conservation targets. Among the continents, South America exhibits the most pronounced expansion pressure (Supplementary Data 5). For instance, single-criterion scenarios require 57.9% coverage to satisfy the 30% global target and 94.3% for the 50% global target. In contrast, multi-target scenarios reduce the required coverage by 7.44% and 27.0%, respectively, relative to these single-criterion scenarios. Larger expansion of multi-target areas results in a lower shared conservation burden for South America, Africa, Asia, and Oceania. Given these expansion pressures, we also estimate coverage considering complementarity and find that less than 30% of the global area is sufficient to cover all simulated threatened-species conservation units (Supplementary Figs. 3, 4 and 6), which helps identify priority areas under constraints of limited land and funding.

Discussion

Multi-objective synergies and trade-offs in conservation prioritisation

A prerequisite for effectively expanding protected areas is recognising the uneven distribution of biodiversity, ultimately resulting from millions of years of evolutionary processes, geological transformations and climate change^38,39,49. This natural unevenness, combined with the unequal development of national economies, results in differential conservation needs and responsibilities to achieve global biodiversity goals. Our findings confirm the uneven spatial distribution of diversity, showing that it also applies to threatened plants and animals, which are predominantly concentrated in tropical and low-latitude areas (Fig. 1 and Supplementary Fig. 3). These areas also coincide with regions where low-income countries are situated, primarily in Latin America, Africa, and Southeast Asia (Supplementary Data 4)⁴¹. However, under the 30% expansion target, priority hotspots for threatened plants exclude several important regions, such as the Amazon rainforest and the Sanjiangyuan region, the latter being characterised by sparse vegetation and grasslands (Fig. 1). When integrating carbon storage, areas with high forest cover become expansion hotspots, although some open-canopy (sparsely vegetated and grassland) areas are not included, potentially highlighting the discrepancy between the distribution patterns of threatened species and overall biodiversity (Supplementary Fig. 5). By incorporating carbon storage, these effective scenarios can evolve into multi-goal conservation strategies, thereby reaching broader biodiversity and climate goals.

Our findings suggest that unequal allocation of priorities is due to the spatial diversity and temporal dynamics of biodiversity and carbon storage, which are in turn influenced by global climate change and land-use policies^20,50. Implementing a multi-goal approach involves acknowledging climate-driven dynamics of biodiversity and carbon storage from a biogeographic perspective (i.e., modelling the likely distribution shifts in response to climate change), as biodiversity and ecosystem functions are inherently dynamic⁴. While ambitious long-term goals may not fully avert the peril of species extinctions⁵¹, our models can help inform expansion directions through understanding species’ dynamic distributions and changes, reducing the risk of loss.

Inadequate species data and single-effective evaluations have affected the quality of priority expansion, causing uncertainty in national participation and responsibility allocation^20,41,52. With the increasing availability of biodiversity and carbon storage data, integrative multiple-target assessments are increasingly employed to identify additional highly effective conservation priority areas (protection hotspots outside of existing protected areas). Our findings indicate that the multi-goal expansions exhibit strengths in biodiversity representativeness and climate goal alignment. For example, under the multi-objective optimisation (scenario 7), the effectiveness of biodiversity conservation is reduced by an average of 12.9% conversion compared to scenarios prioritising pure biodiversity or carbon storage, with a 21.1% increase for carbon storage (Supplementary Data 7). In other words, multi-target conservation balances comprehensive species groups and carbon storage while slightly decreasing overall ecological connectivity (2.35%) and lowering costs (2.06%) (Fig. 3 and Supplementary Data 7). These characteristics are bolstered by the synergy and trade-off aspects of multi-goal conservation, enhancing global targets at the country level^9,17,52. Countries participating in global conservation targets can leverage these synergies and trade-offs inherent in multi-objective approaches to enhance equitable target allocation^14,36,41. In addition, we find that, under a maximum area constraint, complementarity-based unit selection achieves conservation targets for all modelled species more rapidly than richness-based approaches (Supplementary Fig. 3). However, this assumes that all species within areas designated as protected areas are de facto safeguarded, and achieving broader ecological representativeness still requires coordination with other objectives. Nevertheless, this approach can help decision-makers identify relatively high-priority land resources and reduce the conservation burden on biodiversity-rich but low-income countries, thereby supporting more systematic protection and fairer allocation of responsibilities.

Prioritising biodiversity conservation areas may necessitate trade-offs with food security and essential cropland requirements^53,54, which are linked to national land-use policies and are not explicitly modelled in our analyses. Unlike plant taxa, for which occurrence data are pre-filtered by habitat preference and elevation to ensure accuracy, animal taxa are subject to a posteriori environmental filtering. This approach may introduce biases in realised niche estimation and dilute the model’s responsiveness to climate change^55,56. Specifically, including unsuitable habitats during the calibration phase risks overestimating occurrence probabilities in marginal areas. Consequently, the model’s sensitivity to bioclimatic drivers may be compromised, leading to generalised projections (Supplementary Table 3). Beyond these modelling constraints, standard spatial optimisations relying on aggregate richness often neglect complementarity, inevitably leading to redundant protection. To address this, we incorporate a dynamic rarity-weighted richness (RWR) index that explicitly accounts for species-specific minimum area targets. Our results reveal that covering all simulated threatened species requires significantly less than 30% of terrestrial land (Supplementary Figs. 3 and 4), demonstrating that ambitious expansion targets must be underpinned by rigorous complementarity principles to ensure effective, representative biodiversity protection rather than surplus coverage.

The priority expansion strategy developed here seeks to reconcile synergies and trade-offs across multiple goals, enhancing our understanding of equity in global conservation targets. The economic burden of prioritised expansion in tropical or low-latitude countries (associated with the costs of enacting conservation as well as opportunity costs) can be substantial when pursuing single goals or focusing solely on high biodiversity hotspots. One financial upside is the potential for these countries to receive more funding support to enhance conservation efforts for threatened species, habitat protection and restoration. Additionally, the trade-off effect of multi-targeting combined with carbon storage potentially dilutes the protection shares of biodiversity hotspot countries, suggesting a partial shift of burdens to the global north or high-income countries. Integrating complementary species protection minimises land expansion that is not strictly required to meet international commitments. When coupled with other conservation targets, this approach requires less expansion in biodiversity hotspots, further easing the disproportionate financial strain that exists on lower-income countries. In reality, complementary species protection assumes 100% effectiveness within protected areas. Beyond delineating protection boundary, improving the effectiveness of existing protected areas remains essential³¹ — particularly for areas identified as non-priorities within PAs in our results, which should be strengthened and must not be downgraded or removed from their current protected status (Figs. 1 and 2).

Conservation efforts must recognise that protection shares are proportionate to the amount of funding, which may weaken support for impoverished countries, particularly under already limited global financial resources¹². We do not endorse the potential threats posed by such implementation efforts and emphasise the need for multi-target balance mechanisms to accelerate global conservation target implementation and increase global cooperation and financial transfers. It has also been argued that, at least initially, strengthening the existing protected areas may be a better use of increased conservation funds allocated to some countries, like Madagascar, rather than expanding protected areas that are not properly resourced and managed³¹.

Equity implementation of conservation priorities

Our study indicates the dual conservation effectiveness of the ‘30 by 30’ target and ‘Half Earth’ global initiatives, with expansion scenario 7 achieving 63.5% and 82.9% protection of biodiversity and carbon storage, respectively. Previous studies on species protection and ecosystem storage gains align with our results of increased overall benefits^17,19,43,51. Achieving ambitious targets requires the unification and synergy of global and local goal implementation. It also depends on conservation adoption mindsets, bottom-up national interests and priorities, and equitable practices at national and local scales^15,52. We also recognise that protected area expansion and development are inseparably linked to land-use policies and management strategies. We highlight three key lessons from our research findings that are relevant to political leaders and policymakers.

Priority area expansion should account for spatial synergies and trade-offs across multiple biodiversity and climate objectives to expedite global target implementation. The global protected area system encompasses various forms of conservation implementation, such as the designation of national parks and nature reserves^20,32. Other effective area-based conservation measures (OECMs) are being considered for priority expansion, applicable to reducing natural habitat loss or maintaining multiple ecosystem functions and their service benefits⁵². The types of protected areas are evolving as their values are continually enriched, increasingly reflecting actual conservation quality through integrative multi-conservation benefits. Multi-goal conservation utilises its synergies and trade-offs to develop a balanced portfolio for differentiated integrative strategies. This means increasing biodiversity and ecosystem service synergies primarily in tropical regions, while higher latitudes retain fewer synergy types^15,17,52. Differentiated multi-goal conservation demands international cooperation and is grounded in the global conservation agenda. It involves developing multi-goal strategies following the Convention on Biological Diversity and the United Nations Framework Convention on Climate Change^9,41.

Multi-goal conservation must consider the dynamic nature of biodiversity patterns to accurately pinpoint national priority areas. The expansion of protected areas faces challenges in adapting to the dynamic shifts in biodiversity caused by climate change². Our bioclimatic model shows that expanding conservation priorities based on the changing dynamics of threatened species results in varying levels of conservation effectiveness (Figs. 1 and 2). By integrating dynamic biodiversity patterns into conservation strategies, we can forestall and anticipate trends in the expansion of protected areas across various countries amidst climate and land-use changes in order to avoid larger gaps. By predicting decades-long biogeographical dynamics (such as geographic range shifts and local extirpations), multi-goal conservation strategies can identify spatial and temporal patterns that sustain habitat stability over time. In addition, integrating species’ dynamic trajectories into systematic conservation planning and applying the principles of irreplaceability and complementarity enables species-level multi-objective optimisation⁵¹, which enhances both conservation quality and allocative equity. Specifically, in this study, we incorporate the dynamic RWR index to comprehensively characterise biodiversity by combining species-specific distribution ranges with richness. This helps to account for complementary protection and prevent site redundancy, potentially easing expansion pressures and prioritising funding for low-income, high-biodiversity nations. Conservation optimisation methods that explicitly model species-specific spatial and temporal dynamics while targeting multiple objectives offer promising avenues to further improve this selection⁴⁸.

Expanding priority areas based on the proximity principle represents a promising option in practice. In biodiversity hotspots, particularly in low-latitude and tropical regions, increasing conservation targets tend to drive multi-goal hotspots to expand outward from existing protected areas (Fig. 1 and Supplementary Fig. 13). Given the importance of maintaining network integrity and ecological connectivity, proximity-based expansion may provide a more feasible pathway and is often easier to achieve than expansion in areas without surrounding natural ecosystems from which species can migrate. This approach can also help avoid conflicts with land use and development policies that may arise when establishing new protected areas in different locations⁶. Another advantage is that it addresses critical challenges in high-human-impact areas by balancing hotspot protection and ecological connectivity needs, prioritising proximate areas over non-adjacent ones. This is particularly critical given that previous studies have revealed that nearly one-third of protected areas are situated in high-human-activity regions or are exposed to unprotected surroundings^6,57. In addition, conservation costs associated with such expansion efforts necessitate increased financial support and the development of innovative transaction mechanisms^9,15,57. Methods like carbon trading for climate mitigation, payments for ecosystem services, biodiversity credits, ecotourism, and other recognised fundraising modes could help mobilise the necessary resources^17,32,58.

In conclusion, our study emphasises accelerating the expansion of priority areas through multi-objective conservation strategies, providing insights into how the Global Biodiversity Framework and the Paris Agreement could be integrated and delivered jointly. On the one hand, multi-objective strategies achieve a balance between benefitting different species groups and carbon storage through synergies and trade-offs. On the other hand, integrating multi-objective conservation leverages inherent balancing mechanisms to provide insights for national implementation of global biodiversity and climate mitigation goals. Future efforts could focus on dynamic strategies at finer spatial resolutions, enhancing global collaboration, and securing the necessary funding for biodiversity conservation actions, particularly in biodiverse, low-income countries.

Methods

Overview of methods

First, we mapped the richness and dynamic rarity-weighted richness (RWR) index of threatened plants and tetrapod animals for the near-current (as baseline conditions), 2030, and 2050 time periods to delineate biodiversity hotspots. We also mapped carbon storage, representing a key aspect of nature’s contributions to people in relation to climate change mitigation. Second, we established seven biodiversity conservation scenarios, prioritising plants, animals, and carbon storage. Ecological connectivity (ECP and PCI) of existing protected areas or expansion areas was quantified as one measure of conservation quality. We also assessed the likely benefits for protecting species and carbon storage and the cost-effectiveness resulting from land-use changes. Thirdly, we quantified the conservation gap (defined as the land area mismatch between protected areas and assumed expansion hotspots) across various scenarios with assumed expansions of priority areas, analysing the fulfilment of global ‘30 by 30’ and ‘50 by 50’ targets within countries. The foundational methodological framework is illustrated in Supplementary Fig. 1.

Identification of biodiversity hotspots under climate change

We selected threatened species (plants, mammals, birds, reptiles, and amphibians) to model areas of high threatened biodiversity, which contain high levels of species richness, endemism and threat⁵. As defined by the International Union for Conservation of Nature (IUCN), we collected the threatened species (defined as ‘threatened, extinct and near threatened’), including categories: Extinct (EX), Extinct in the Wild (EW), Critically Endangered (CR), Endangered (EN), Vulnerable (VU) and Near Threatened (NT). We chose to focus on threatened species, since they often occupy narrow ecological niches, or serve as indicators of biodiversity vulnerability to climate change and human disturbances, necessitating in-situ conservation^44,45. To reduce model assumptions, we did not attempt to predict changes in IUCN Red List category for individual species over time⁵⁹. The potential spatial distribution of threatened species and environmental change variables were analysed through Maxent model. The Maxent v3.4.4 is a machine learning tool for species distribution modelling, proven effective in habitat suitability prediction using presence-only data^60,61. It is widely used and performs more robust and accurate simulation outputs than other species distribution models, even with smaller species datasets^62,63. Finally, the modelled potential habitat suitability of species was used to derive species richness and the dynamic RWR index. The dynamic RWR index serves as a strategy in systematic conservation planning to enhance the complementarity principle by incorporating species-specific minimum conservation areas and an iterative algorithm, while ensuring that the broad distribution ranges of certain simulated species or families are not exaggerated. Detailed results were provided in Supplementary Note 3 & 4, and these indices were used to identify conservation priority areas.

Species distribution selection

We compiled data for threatened species covering terrestrial habitats, capturing plants (6957 species with 369,917 non-repetitive records, 80.9% of threatened species in the downloadable dataset), mammals (1686 with 325,993, 96.3%), birds (2285 with 341,798, 87.5%), reptiles (2297 with 422,875, 99.4%) and amphibians (2916 with 81,522, 95.6%) (Supplementary Table 1). The datasets analysed in this study included the IUCN Red List’s point and polygon data, the Global Biodiversity Information Facility (GBIF), and BirdLife International (Data Availability). The plant data sourced from IUCN mainly consisted of polygon layers, with a few point records included. The polygon layers drawn by experts at IUCN represent the approximate range of distribution⁶¹. Although IUCN range maps can sometimes differ substantially from species ranges estimated from point occurrences, they both have been shown to reveal similar macroecological patterns⁶⁴. Even so, relying solely on IUCN point data might miss the nuances of potential habitats recently updated due to political factors, expert judgement, and land cover changes²⁰. Therefore, our study employed a hybrid approach, converting these IUCN polygons into a set of random point records (Supplementary Fig. 8 and Supplementary Data 1). We then integrated GBIF and IUCN data into modelling to identify more potentially suitable habitats. Given the size of expert-drawn polygons and their global scale, we generated a random species point within an approximately 404-km² polygon. This process accounted for seed dispersal distances, vertebrate migration, and the accessibility of connectivity patches^65–67. We used ArcGIS Pro to randomly generate points within the IUCN polygons’ attribute table, subsequently employing the ‘split by attributes’ tool to segregate the data into single-species files. These files were then batch processed via Model Builder, utilising the ‘create random points’ and ‘add XY coordinates’ tools to generate species locations with geographic coordinates based on the predefined random point count. In cases of family-level Maxent modelling, we scrutinised the amalgamated species locations to preclude duplicate points within a 1-km resolution of fishnet spatial distribution, avoiding spatial sampling bias and mitigating collinearity concerns in model predictions^20,68. Moreover, data were downloaded from GBIF and selected for human observations and preserved specimens, with an elevation greater than zero, georeferenced coordinates, and Red List entries, as detailed in Supplementary Note 5. The GBIF data were downloaded and cleaned, ensuring the data were reprojected and possessed accurate geodetic datum WGS84 coordinates. Terrestrial species occurrences in marine areas and duplicated uncertain locations were eliminated, following a similar rationale as in packages such as CoordinateCleaner (R package), to minimise significant impacts on simulated species patterns arising from georeferencing errors, imprecisions and biases^63,69,70. All records obtained from GBIF and IUCN data for the hybrid approach are detailed in Supplementary Data 2. All data were clipped according to the world boundary for integration into subsequent modelling.

Environmental variables

We selected a set of 27 natural environmental variables for modelling species habitats, encompassing bioclimatic factors, geographical features, soil conditions, and vegetation attributes (Supplementary Table 2). These variables were selected from abiotic factors considering the natural environments that currently better represent the species’ habitat exposure²⁰. Our species distribution models also considered indirect human impacts, estimated from land cover and vegetation layers, reflecting varying degrees of human disturbance. To mitigate collinearity arising from high correlations among these environmental variables, we conducted correlation analysis using the ‘pair’ function in R to filter the selection⁷¹ (Supplementary Fig. 9). All variables were identified at a 1-km spatial resolution raster for use in the modelling process (Supplementary Note 5).

Modelling and setting of threatened species patterns

Following the method of Huang, et al.²⁰, the model was structured to run three iterations to ensure computational stability and minimise stochastic errors, producing raster layers at 1-km spatial resolution. Modelling was conducted for individual species. Due to the total environmental variables occupying very large data volumes, which can impact data reading/writing efficiency and computation time, we adopted a two-pronged modelling approach. Individual species with more than 300 occurrence records (combined GBIF and IUCN data) were modelled individually, while those with fewer than 300 points were combined together and modelled at family level to reduce computing time from years to months (Supplementary Fig. 8). This approach assumed that threatened species within the same genus exhibit niche similarity⁷², and species within the same family share similar potential ranges of ecological niches in response to abiotic conditions^73,74, such as bird distribution ranges⁷⁵. Additionally, GBIF notes that the same species may contain duplicate records, which may originate from various institutions or genetic sequencing data⁷⁰. To address this potential pitfall and improve simulation accuracy, we filtered out duplicate GBIF records before analysing individual species, which may have resulted in some individual species data with fewer than 300 points (Supplementary Fig. 8 and Supplementary Data 1). After Maxent modelling, the resulting family- and species-level distributions were overlaid to illustrate the richness and dynamic RWR index of plant and animal species. We validated model accuracy through global model fit of machine learning (AUC and TSS)⁶³. The threshold-independent AUC and threshold-dependent TSS are valid methods commonly used for species distribution models^60,76,77. They provided a comprehensive examination of global, category, and group aspects through species modelling of individual species or families. The generated biodiversity hotspot patterns were compared with those from other studies using linear regression as cross-validation (Supplementary Figs. 10–12)⁷⁸. This approach highlighted global synergies and regional hotspot differences in hotspot identification (low values ‘0-0.6’ of the coefficient of determination, R²), revealing variations in biodiversity patterns⁷⁹. The generated outcomes were validated through AUC (total average=0.96, range 0.62-0.99) and TSS (total average=0.88, range 0.47-1.00) (Supplementary Data 1). Cross-referencing with existing literatures confirmed our findings’ reliability and unveiled regional variations in potential threatened species richness (Supplementary Note 5). The model fitting R² for threatened birds was 0.536, followed by mammals (maximum R² = 0.51), plants (maximum R² = 0.43), reptiles (R² = 0.34), and amphibians (maximum R² = 0.18).

Based on the scale dependence of environmental filtering⁸⁰, and given the broader mobility and migration of animal taxa, all selected environmental variables were included in the modelling. The Maxent contribution results showed that bioclimatic variables were the dominant drivers, highlighting the macroecological importance of abiotic factors (Supplementary Note 5). To minimise sensitivity associated with random points extracted from IUCN range maps, threatened plant distribution modelling was restricted to ten key bioclimatic variables. To improve model accuracy and avoid overestimating suitable distribution ranges for all animal taxa, we filtered the raw Maxent outputs based on species-specific elevation ranges and habitat preferences, a method adapted from Jung, et al.⁵¹ — whereas for plant taxa, occurrence points were filtered prior to modelling. The species-specific data for these filters (Supplementary Data 3) were sourced from the IUCN website for each species, and the habitat layers were derived from Jung, et al.⁵⁵. Thus, sites that were considered suitable by the Maxent model but fell outside the species’ preferred habitat and elevation ranges were excluded from further analysis.

We analysed the effect of future bioclimatic changes on threatened species locations by applying the SSP2-4.5 scenario of the global climate model, GISS-E2-1-G. We adopted the GISS-E2-1-G climate model under CMIP6 because its equilibrium climate sensitivity and transient climate response fall within the likely range based on historical constraints, avoiding the high-sensitivity bias observed in some CMIP6 models^81,82. The SSP2-4.5 scenario is considered as the ‘Middle of the road’, which is closer to the current and future scenarios of human activities and social development, where the temperature increase is not expected to exceed 2 °C by 2050 (based on bioclimate data for the years 1970-2000 as baseline).

Spatial estimation of carbon storage potential

Potential carbon storage was quantified based on aboveground and belowground biomass carbon and the density of carbon stored in vulnerable soil organic matter, adopting the methodology of Zhu, et al.⁵². The carbon density data of above and belowground biomass, with a 250 m resolution (t/ha), were sourced from Spawn, et al.⁸³. The above- and below-ground biomass are consistent with woody carbon stocks reported by IPCC Tier-2 and Tier-3 at the national level, with R² of 0.96 and 0.95, respectively. It delineates the global spatial distribution of carbon density provided annually by various vegetation types. Vulnerable soil organic carbon was derived from the latest 250 m resolution datasets by Hengl, et al.⁸⁴. These data used a 10-fold repeated cross-validation with a R² of 0.69 and root mean squared error of 0.72, illustrating their improvement and reliability. It estimated carbon density in mineral soil layers up to 30-cm and organic soil layers up to 200-cm, classified according to the United States Department of Agriculture criteria. The organic layer was extracted using a mask of the Histosol layer with a probability exceeding 5%⁸⁵. It was defined as carbon stock in soil layers at risk of loss due to land-use changes over the next 30 years by the IPCC guidelines for national greenhouse gas inventories⁵¹. Vulnerable soil organic carbon was adjusted to t/ha by multiplying by 10 in ArcGIS Pro. All carbon storage data were resampled to 1-km resolution and combined to produce a total carbon storage density layer (Supplementary Fig. 7).

Inference of ecological connectivity

We evaluated the connectivity of priority and protected areas at both the forest-type and landscape scales. These quantifications indicated the concern that human-designated protected areas should not become isolated from major forest habitats or other landscapes, avoiding isolated conservation practices⁸⁶. The Framework’s Target 3 underscores the necessity for areas of particular importance to be effectively protected and remain ‘well-connected’³⁵. The recognition of ecological connectivity as a marker of quality in protected area design is increasingly emphasised¹³.

Regarding forest types, we measured connectivity within protected or priority forests relative to the total assessed forest area. We employed modifications to Saura, et al.²⁹’s indicators for protected connected lands (ProtConn) and equivalent connected areas (ECA) (code in Code Availability). This allows for flexible adaptation to different scale conditions of the assessment area. The equivalent connected patch (ECP) is calculated as follows (1):

$$ECP=100\times \frac{\sqrt{{\sum }_{i=1}^{n+t}{\sum }_{j=1}^{n+t}{a}_i{a}_j{p}_{ij}^{*}}}{Ap}$$ 1

in which, $a_i$ and $a_j$ are the attributes of assessment area i and j, $A_p$ is the total patch area of the assessment area, $p_{ij}^{*}$ is the maximum product probability of all paths connecting nodes i and j. The ECP indicator was calculated using the command-line version of the Conefor 2.6 package for batch processing⁸⁷. The parameters for calculating ECP were derived from land use data on forest patches. We used a dispersal distance of 1000 m as an evaluation criterion based on Wang, et al.⁸⁸ approach. The forest type of lands use was resampled to 50 km pixels to minimise disturbances from small-scale forest fragmentation, followed by generating node and distance files of forest patches for batch analysis.

In terms of the landscape scale, we considered the inherent complexity and heterogeneity of landscape patches. Utilising the latest Fragstats v4.2 (fragstats.org), we calculated the patch cohesion index (PCI). The PCI quantifies the aggregation of landscape patches of the same type⁸⁹. The logarithmic decline observed with increasing land-use types more accurately captures ephemeral spatial connectivity. The formula for PCI is as follows (2):

$$PCI=\left[1-\frac{{\sum }_{i=1}^m{\sum }_{j=1}^n{c}_{ij}^{*}}{{\sum }_{i=1}^m{\sum }_{j=1}^n{c}_{ij}^{*}\sqrt{b_{ij}^{*}}}\right]\times {\left[1-\frac{1}{\sqrt{Z}}\right]}^{-1}\times 100$$ 2

in which, $c_{ij}^{*}$ is the perimeter of patch ij in terms of number of cell surfaces. $b_{ij}^{*}$ is the area of patch ij in terms of the number of cells. Z is the total number of cells in the landscape. The PCI calculation was based on a 1-km resolution land use dataset⁹⁰, resampled to 2.5-km with a moving window of 5-km, employing default settings for other parameters (Data Availability).

Conservation scenario and expanded conservation priority setting

Following the Framework and the ‘Half Earth’ scenario initiatives, we established terrestrial conservation targets of ‘30 by 30’ and ‘50 by 50’, expanding additional priority areas on top of existing protected areas. These scenarios aim to achieve 30% terrestrial land coverage by 2030 and 50% by 2050^8,35. The scenario settings were constructed using the spatial synergies and trade-offs between biodiversity and carbon storage to compare the effectiveness of multi-goal quantitative and functional protection under the Framework initiative, aiming to identify the best global implementation strategies. We emphasised that using their synergies and trade-offs enhances understanding of how to accelerate the expansion and application of priority areas. We explored multi-objective spatial optimisation combining biodiversity (species richness and dynamic RWR index) and carbon storage, and incorporated empirical minimum conservation area targets at the level of individual threatened species to explicitly enhance the principle of complementarity. As their synergies garner greater attention than the trade-offs for global implementation, we did not explore theoretical discussions regarding how their trade-off relationship might challenge the concept of strong sustainability^17,52.

To enhance complementarity, we introduced a dynamic RWR index, adapted from standard rarity indices⁹¹. While standard rarity-based algorithms generally prioritise species-level rarity to maximise coverage^88,89, dynamic RWR employs a stepwise greedy algorithm. This approach iteratively ranks conservation units to maximise species coverage based on minimum area targets⁵¹, thereby avoiding the redundancy often found in static calculations (Supplementary Fig. 6). Detailed formulas and code are available in Supplementary Note 3 and the Data Availability, respectively. By integrating this index into the optimisation process, we examined whether the ‘30 by 30’ and ‘50 by 50’ area-based expansion scenarios sufficiently cover all simulated threatened species.

The conservation scenarios were established as follows: (1) Scenario 1 prioritises threatened plants, focusing only on plants for priority delineation; (2) Scenario 2 considers threatened animals, including reptiles, mammals, amphibians and birds; (3) Scenario 3 is carbon storage priority, reflecting carbon sequestration functions by natural vegetation; (4) Scenario 4 is a combined priority of plants and animals reflecting mainstream biodiversity conservation goals; (5) Scenario 5 is the integration of threatened plants with carbon storage; (6) Scenario 6 represents the integration of threatened animals and carbon storage; (7) Scenario 7 is the integration of threatened species with carbon storage.

After defining the scenarios, we normalised raw species richness, dynamic RWR index, and carbon storage data to produce standardised index layers ranging from 0 to 100 (Supplementary Note 4 and Data Availability). These layers were then processed using utility functions from the CAPTAIN software² (Conservation area prioritisation through artificial intelligence; optimisation code and resources available in Data Availability).

This study employed the CAPTAIN software to rank cells based on weighted products of biodiversity and carbon storage data layers, divided by conservation cost, to optimise conservation priorities across seven different scenarios. Distinct objective weights were assigned within the simplified scoring formula (Supplementary Table 4). Equal temporal weighting assigned 0.25 to species richness and dynamic RWR in each period (baseline, 2030, and 2050), giving each metric a total weight of 0.75 across time. These weights represent relative contributions rather than normalised proportions (total = 1.5) and help identify stable species habitat suitability while addressing climate velocity and projection uncertainty, ensuring conservation continuity across the planning horizon. A higher carbon weight (2.0, relative to biodiversity [1.5]) amplifies synergistic opportunities where biodiversity and climate benefits converge⁴¹. Multiplicative functions were selected because they penalise areas lacking key conservation attributes and identify locations where multiple conservation values genuinely co-occur rather than compensate for each other. Incorporating species richness and dynamic RWR as biodiversity proxies into the optimisation algorithm ensures the selection of minimal planning units to maximise the co-occurrence coverage of species hotspots and rarity, thereby generating a prioritisation scheme explicitly characterised by complementarity. Furthermore, to validate the role of species complementarity, the spatial characteristics of the optimisation outcome derived from the equal-weighting approach described above are compared against those produced using dynamic RWR alone as the biodiversity proxy (Supplementary Note 3). The human footprint index was incorporated to penalise the selection of areas with intensive anthropogenic pressure, effectively excluding degraded or non-habitat areas to ensure implementation feasibility⁹², while the minimum denominator threshold (min_den=0.1) maintains numerical stability.

The seven optimised conservation scenarios were subsequently refined in ArcGIS Pro using the ‘raster calculator’ tool, which masked existing protected areas and identified priority areas in the remaining unprotected land for the ‘30 by 30’ and ‘50 by 50’ targets. We conducted a detailed spatial analysis of protection expansion scenario 7 to examine the spatial configuration of priority hotspots relative to adjacent protected areas (Supplementary Fig. 13).

Conservation gap, cost, and effectiveness analysis

This study used the terrestrial protected areas data from the August 2023 version of the World Database on Protected Areas for a gap analysis with scenario-based priority areas. Protected area shapefile data were converted to 1 km raster format for comparison, which excludes isolated protected patches smaller than 1 km².

We applied gap analysis to identify conservation gaps and prioritise areas for protection expansion by overlaying existing protected areas with predetermined conservation hotspots²⁰. Conservation gaps occur where non-priority areas fall within protected areas, while expansion priorities consist of conservation targets located outside existing protected areas. We examined conservation effectiveness by quantifying the represented proportion of the global total biodiversity index and total carbon stocks prioritised for conservation. Additionally, the quality of protected areas was evaluated through ecological connectivity metrics (ECP and PCI). For explicit determination of conservation cost-efficiency in priority and existing protected areas, we employed current and future land-use scenarios focusing on impervious surface and cropland percentages.

Conservation costs were calculated by quantifying the percentage of impervious surface and cropland area in protected or prioritised areas in relation to their total area. We used the ‘reclassify’ and ‘extract by mask’ tools in ArcGIS Pro to extract and obtain the coverage area of impervious surfaces and cropland within each conservation scenario. We used the percentages of land covered by impervious surfaces and cropland as a proxy for the conservation costs or restoration expenses required for future scenario-based conservation efforts, as these land use types are considered areas of high human modification¹⁵. A higher percentage signifies a potentially higher conservation cost. The land use data were sourced from the Shared Socioeconomic Pathways dataset⁹⁰. The land-use data considers three shared socioeconomic pathways (SSP) — SSP1-2.6 (green and sustainable trajectory), SSP2-4.5 (moderate development), and SSP5-8.5 (fossil-fuelled development). The rationale for choosing these two indicators was based on their capacity to measure direct anthropogenic drivers and threats from land use changes. This approach also facilitated an evaluation of future biodiversity conservation trends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: Y.B., Z.H., Z.F., A.A. Methodology: Z.H., D.S., J.W. Visualisation: Z.H. Funding acquisition: Y.B., Z.F., A.A. Supervision: Y.B. Writing—original draft: Z.H. Writing—review and editing: A.A., D.S., Z.H., Z.F., J.W., Y.T., J.L., S.Z., M.A., X.L., R.W., S.X., Q.Z., C.C., Q.W., H.C., Y.B.

Peer review

Peer review information

Nature Communications thanks Francesco Sabatini and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This research was funded by the 14th Five-Year Plan of the Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences (E3ZKFF7B), Yunnan Science and Technology Department (202501AS070088); Yunnan Revitalisation Talent Support Program “Innovation Team” Project (202405AS350019), National Natural Science Foundation of China (32471735, 72404267), the Swedish Research Council (2019-05191), the Swedish Foundation for Strategic Environmental Research MISTRA (Project BioPath), the Kew Foundation, ETH Zurich.

Data availability

The complete data generated and analysed in the study that support the findings of this study have been deposited in the Figshare repository at https://doi.org/10.6084/m9.figshare.26820685https://doi.org/10.6084/m9.figshare.26820685. Raw species occurrence data were obtained from the Global Biodiversity Information Facility (GBIF) for Birds⁹³, Amphibians⁹⁴, Mammals⁹⁵, Reptiles^96,97, and Plants⁹⁸. Species distribution and conservation status data were accessed via the IUCN Red List of Threatened Species (IUCN⁹⁹ and BirdLife International¹⁰⁰). Spatial boundaries for protected areas were sourced from the World Database on Protected Areas (WDPA)¹⁰¹. Global country administrative boundaries used as basemaps were obtained from Natural Earth (https://www.naturalearthdata.com), which is in the public domain. Furthermore, all maps presented in the main text and Supplementary Information were created using ArcGIS Pro software by Esri. ArcGIS^® is the intellectual property of Esri and is used herein under license. Copyright ^© Esri. All rights reserved. For more information about Esri^® software, please visit www.esri.com.

Code availability

The code used to analyse the data and generate the results reported in this study has been deposited in the Figshare repository at https://doi.org/10.6084/m9.figshare.26820685https://doi.org/10.6084/m9.figshare.26820685.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76324-7.