As depicted in Fig. , AD-MERGE comprises four modules: (i) the Energy Technology Assessment (ETA) module uses a bottom-up approach to model the energy supply sector, separating the production and associated GHG emissions of electric and non-electric energy. This module enables transitions among energy carriers and technologies. (ii) The Macroeconomic (MACRO) module applies a top-down approach to capture the interplay between the energy sector and the broader economy by modeling substitutions between value-added and energy aggregates. (iii) The Climate module represents the causal chain through which GHG emissions increase atmospheric GHG concentrations, alter the radiative forcing balance, and lead to changes in global mean surface temperature. In AD-MERGE, this reduced-form module is internally integrated within the model's optimization framework, allowing climate dynamics to evolve endogenously and enabling explicit feedbacks between climate outcomes, economic activity, energy-system decisions, and adaptation responses. Owing to the thermal inertia of the climate system, driven by a slower heat uptake of the oceans, actual temperature change lags behind potential temperature . Accordingly, the model distinguishes between potential temperature and actual temperature, with the latter adjusting gradually over time. (iv) The Damage module quantifies gross damages, which are economic losses due to temperature changes, assuming no adaptation. Through adaptation policies, these gross damages can be reduced to residual damages. In AD-MERGE, the original market damage function from the MERGE model is replaced with a new function that incorporates these residual damages as well as the costs of adaptation measures, representing net costs of climate change . Reactive adaptation is modeled as immediate spending (i.e. a flow variable), and proactive adaptation is modeled as an investment in capital that provides protection (i.e. a stock variable). These strategies are integrated using a Constant Elasticity of Substitution (CES) function, reflecting the premise that reactive and proactive adaptations are imperfect substitutes. The formulation of the damage module is provided separately in Appendix D, while the formulations of the other modules are presented in the Supplement (Sects. S1–S3).

AD-MERGE is formulated as a non-linear optimization problem, designed to optimize a global Negishi welfare function, which aggregates the discounted utility of consumption across regions . The Negishi welfare function adjusts regional utility weights to equalize the marginal utility of income across regions, maximizing overall social welfare while accounting for regional differences . The implementation of this welfare formulation in AD-MERGE 2.0, including the calculation and updating of regional Negishi weights, is described in Sect. S2. The Negishi objective provides the welfare aggregation structure used to solve the model, while regional distributional outcomes are examined separately. The economic structure of each region is represented using a Ramsey-Solow framework, which captures the dynamics of optimal long-term economic growth. AD-MERGE follows an intertemporal optimization problem under the assumption of perfect foresight, solving for all time periods simultaneously. The model ensures that supply and demand are balanced in each period through the pricing of traded goods, including energy commodities, and a numeraire, which represents all non-energy production.

Table  compares AD-MERGE 2.0 with selected IAMs and provides an overview of how AD-MERGE 2.0's mathematical formulation, climate and damage specifications, carbon dioxide removal (CDR), adaptation treatment, and regional coverage compare with those of other widely used IAMs. The table is not intended to provide an exhaustive taxonomy of IAMs, but rather to situate AD-MERGE 2.0 relative to established frameworks used in climate-policy assessment. An important comparison emphasizes whether existing models provide impact and adaptation representations that are suitable for incorporating adaptation explicitly, including corresponding estimates of gross damages, avoided damages, residual damages, and adaptation costs. Where relevant, the model versions or source references used for the comparison are specified. The information is synthesized from review studies on IAM capabilities, including , , and , and from the Integrated Assessment Modeling Consortium database and model documentation . It is further complemented by model-specific publications and documentation, including , , , , , , , , , , , , and the main methodological papers associated with each model.

To achieve a more granular and region-specific representation of the socioeconomic dynamics, the initial nine regions of AD-MERGE have been divided into 15 regions (Fig. ). The new regional disaggregation enhances AD-MERGE by capturing the diversity of development patterns, challenges, and opportunities across more regions. This improves the model's ability to study regional energy policies, resource management, and the assessment of impacts and adaptation strategies. Several factors, including insights from other IAMs, geographical considerations, political coalitions, and the availability of data have informed this development.

As part of this refined disaggregation, large countries such as Canada (CAN), Mexico (MEX), Brazil (BRA), and Russia (RUS) are now represented as standalone regions. This update complements the original AD-MERGE model, where major economies such as the United States (USA), China (CHN), and India (IND) were already treated as separate regions. Australia and New Zealand, previously aggregated with Canada in the “Canada, Australia, and New Zealand” region, are now represented as a separate region (ANZ), distinctively. Furthermore, previously grouped “Eastern Europe and the former Soviet Union” have been divided into two distinct regions, Russia and Eurasia (OEA). Finally, the “rest of the world group” is disaggregated into three distinct continental groups: Central and Latin America (CLA), Africa (AFR), and other Asian countries (OAS). Moreover, new regions such as Japan and South Korea (JSK) and the Middle East (MEA) have been introduced, while Western Europe (WEU) continues to be represented in the model. Table  presents the list of newly defined regions, their abbreviations, and descriptions of the countries included within each region.

The ETA module is critical in modeling the conversion of primary energy sources into secondary energy carriers through a variety of energy conversion technologies. In AD-MERGE, 20 distinct energy conversion technologies were incorporated within the model, each competing to provide various types of secondary energy. In AD-MERGE 2.0, these technologies have been extended to 36 distinct energy conversion technologies, including nine new power generation technologies and five new hydrogen generation technologies (as shown in Fig. ).

In earlier model versions, renewable technologies were grouped under broad categories of “advanced high-” and “low-cost” technologies, which included an aggregated representation of carbon-free options such as modern nuclear, biomass, wind, and solar. AD-MERGE 2.0 replaces these generic carbon-free technologies with specific technologies, including solar photovoltaics (PV), concentrated solar power (CSP), bioenergy with CCS (BECCS), onshore and offshore wind, and distinct nuclear technologies. Furthermore, AD-MERGE 2.0's power sector now features explicit fossil fuel technology options. Natural gas technologies are differentiated into combined-cycle and open-cycle gas turbines, with the option for combined-cycle to be paired with carbon capture and sequestration (CCS). Additionally, coal technologies now distinguish between subcritical, supercritical, and ultra-supercritical technologies, which can also be equipped with CCS. The model also includes battery storage and pumped hydro storage (PHS), considering short-term and seasonal storage derived from curtailment.

The new AD-MERGE 2.0 framework employs a granular approach to resource characterization, categorizing oil and gas into 13 extraction cost tiers, where categories 1 to 10 denote conventional sources, and 11 to 13 represent unconventional reserves. Coal is divided into five cost categories that include bituminous, sub-bituminous, and lignite extraction costs with a similar approach to . To ensure consistency between resource inputs and nuclear technologies electricity generation, the model tracks aggregated macro-level fuel resource consumption, thermal efficiency parameters, and associated long-term waste management requirements with a similar approach to . The energy reserves and resources estimates incorporate the latest data from and .

DACCS offers a promising approach to remove CO₂ directly from ambient air and store it, which facilitates offsetting emissions from hard-to-abate sectors. In current IAMs, DACCS has been shown to contribute significantly to achieving mitigation pathways and building negative emissions portfolios . In our model, we represent two DACCS technologies: a liquid-scrubbing process (e.g., hydroxide–carbonate) and a solid‐sorbent process (e.g., temperature vacuum swing adsorption, TVSA), based on the latest literature and their technological readiness . Each technology is characterized by the corresponding energy use, efficiency, and levelized costs over time, reflecting the assumed cost reductions. Appendix  provides the assumptions and parameter values. Our modeling framework endogenously selects between liquid and solid DAC technologies, representing their competition with other mitigation and removal options. Furthermore, the model integrates additional negative emissions strategies, such as BECCS and afforestation, to comprehensively evaluate pathways towards more stringent carbon targets. BECCS is represented within the energy system module using broad regional resource potential and techno-economic assumptions. However, AD-MERGE 2.0 does not include an explicit endogenous land-use framework that allocates land across competing uses such as food production, bioenergy crops, and forests. Accordingly, land-use change, land competition, and land availability are not modeled directly, and afforestation and bioenergy expansion are not linked through a shared land-allocation structure. CDR deployment should therefore be interpreted as conditional on aggregated regional potential assumptions rather than as a full assessment of land-use feasibility under explicit competing land demands.

AD-MERGE 2.0 builds on the stock and flow framework of and . The “optimal adaptation” serves as a theoretical benchmark rather than a policy forecast. As noted by the AR6 WGII report , real-world adaptation is constrained by local institutional barriers and varying adaptive capacities. By modeling adaptation as an explicit capital stock, rather than a reduced-form percentage of GDP, our approach captures the opportunity costs and investment inertia inherent in climate resilience. This structural detail is essential for analyzing the energy-climate-adaptation nexus, as it reveals how adaptation investments may compete with the capital flows required for mitigation .

In AD-MERGE 2.0, the impacts of climate change and the costs and benefits of adaptation were calibrated across the 15 regions included in our analysis (detailed in ). It distinguishes between reactive and proactive adaptation. Reactive adaptation covers reaction to climate change stimuli, whose costs and benefits are confined to the current period and do not carry over; in our model, these measures are treated as flow variables reflecting decentralized, low-investment actions. Proactive adaptation, in contrast, involves large-scale anticipatory investments that incur upfront costs and generate a stock of adaptation whose benefits unfold over multiple future periods. The adaptation stock is modeled as a form of capital that depreciates over time and is treated as a regional public good, as its costs and benefits extend beyond individual decision-makers. Together, these two modeling approaches capture both immediate individual-level coping strategies and long-term collective resilience measures consistently across all 15 regions.

Existing damage assessments provide useful benchmarks for different parts of the calibration. Empirical macroeconomic studies such as and meta-analyses such as and inform the broader comparison of aggregate damage-function severity, while sectorally disaggregated assessments such as COACCH damage function provide direct evidence for impact-specific damage calibration. However, these studies generally do not provide explicit adaptation-cost and effectiveness estimates; they are primarily used to benchmark climate damages, while adaptation calibration relies on more limited impact-specific evidence and model-specific assumptions. The objective here is to derive regionally aggregated damage and adaptation relationships that are consistent with the AD-MERGE 2.0 structure and suitable for calibrating both reactive and proactive adaptation, thereby allowing these adaptation responses to be represented explicitly as decision variables and studied jointly with mitigation and climate damages.

The gross damage and adaptation functions in AD-MERGE 2.0 are calibrated using the same functional form as in . This functional form and the calibration procedure are described in Appendix . The assessment focuses on six climate impacts for which there are reliable regional data. These impacts are river flooding, total energy demand, coastal impacts, labor productivity, health, and tourism. The data sources used to quantify each impact are documented in and summarized in Table . For each impact in a specific region, a relationship was estimated over temperature and economic growth to allow more data points. Impacts in a specific region were then aggregated for different levels of temperature change and economic growth. Based on these observations, a regional damage curve was parameterized to reflect relationships between temperature change and economic losses as a percentage of GDP. Specifying losses as a percentage of GDP allows us to incorporate the effect of economic growth on total economic costs.

Figure  presents a comparative assessment of global damage functions, illustrating climate-induced economic losses as a percentage of GDP under a no-adaptation assumption. The revised damage curve in AD-MERGE 2.0 reflects the integration of recent estimates of sectoral and regional vulnerabilities, leading to a more pronounced increase in economic losses with rising temperatures. The updated curve exhibits mid- to high-end estimates from the literature, such as , and projects a higher damage trajectory than the earlier AD-MERGE 1.0 version, which is closer to the more conservative projections of . This upward shift in damage estimates suggests greater sensitivity of economic outcomes to warming, underscoring the importance of accounting for heterogeneous sectoral and regional impacts in integrated assessment modeling and of regularly updating damage estimates in IAMs. The corresponding regional damage functions (Appendix , Fig. a).

When calibrating adaptation, we estimate the costs and benefits of each form of adaptation, namely stock (reactive) and flow (proactive) (see ). The costs are given as a percentage of GDP, whereas the benefits are given as a fraction of gross damages avoided. This is referred to as the level of adaptation and represents the difference between gross and residual damages. Then, optimal adaptation levels and concomitant adaptation costs are estimated for each impact for a specific region and adaptation type based on data or other literature. A maximum level of adaptation is also estimated, reflecting the point at which the adaptation cost function would become extremely high. An adaptation cost and benefits function over temperature is estimated per impact in a specific region. Adaptation costs and benefits are then aggregated over impacts for each region, and adaptation coefficients are calibrated to capture these estimates. Accordingly, the resulting optimal adaptation levels should be interpreted as benchmark outcomes within the model structure under idealized implementation assumptions, rather than as directly achievable policy targets.

The IPCC uses the SSP, RCP and C1–C8 warming categories to describe potential socioeconomic futures and climate scenarios collectively. The SSPs outline distinct socio-economic futures that influence the drivers of GHG emissions and thereby the resulting radiative forcing levels (RCPs) used in climate scenarios. The RCPs describe the potential atmospheric concentrations of GHGs. In the AR6 framework, the temperature categories C1–C8 distinguish these scenarios based on their projected temperature outcomes by the end of the 21st century relative to the pre-industrial era (1850–1900). The AR6 scenario ensemble includes a diverse set of quantitative scenarios developed by different modeling groups, using various frameworks that incorporate these SSP, RCP, and C-category structures.

AD-MERGE 2.0 is calibrated using 2015 as the base year and aligns with the SSP version 3.0 demographic and economic projections from 2015 through 2100, ensuring consistency with the latest available SSP datasets . Specifically, the model integrates the most recent population projections from the SSP2 scenario provided by and incorporates GDP trajectories based on estimates by . The new calibration also involved revising key model parameters and adjusting variables, including base year values for primary and secondary energy production, existing production capacities, and total trade volumes, based on . Emission-related variables, such as global warming potentials, cumulative emissions in the base year, energy and non-energy emissions, and radiative forcing, were updated using data from AR6 WGI report and Synthesis Report . In addition to hydrogen (see Sect. ) and VRE (see Sect. ), the technical parameters for other emerging technologies were revised based on . To assess consistency, the AD-MERGE 2.0 Baseline scenario results were compared with the SSP2 scenario of the AR6 database , with a particular focus on warming levels categories C7–C8, as these categories correspond to our baseline emission and temperature trajectories.

Figure a presents total GHG emission projections. In AD-MERGE 1.0, which uses 2000 as a base year, the projected emissions are well below the observed emissions for 2015 and 2020 (see “Baseline 1.0” in the figure). Afterwards, emissions increase steadily to a level approaching an RCP8.5 scenario. The emissions estimates from the updated AD-MERGE 2.0 (see “Baseline 2.0” in the figure) are consistent with the observed emissions in 2015 and 2020 and follow the RCP7.0 scenario closely over time (see “SSP3-7.0” in the figure). In this figure, the warming categories C6-C7 (limit warming between 3 and 4 degrees) and C8 (exceed warming of 4°) are also provided for comparison. Non-energy-related emissions are treated exogenously in the model, with their trajectories calibrated to align with the warming level categories corresponding to each of our main scenarios, based on the representative scenarios in the AR6 WGIII report .

In Fig. b–d, AD‐MERGE 2.0 outputs, represented by data points, consistently fall within the interquartile ranges of the AR6 scenario ensemble for SSP2 categories C7 and C8. These categories were selected as a comparison benchmark since the AD-MERGE Baseline scenario aligns with this range of projected warming. The projections of primary energy supply (Fig. b) and electricity generation (Fig. c) closely track the median trends, reflecting the model’s ability to capture regional patterns of electrification, technological diffusion, and decarbonization efforts. In addition, GDP trajectories (Fig. d), shown in alignment with the interquartile ranges across regions, further support the consistency of the economic and demographic assumptions embedded in the model calibration. Additional base year and near-term validation of the main drivers of the model are provided in Appendix .

We explore five main scenarios using AD-MERGE 2.0: Baseline

This scenario represents our calibration scenario. This scenario assumes the continuation of current global trends in energy use, GHG emissions, and economic development in the absence of additional climate policies. This scenario excludes climate damages from decision-making, providing a counterfactual framework to evaluate the effectiveness of climate policy interventions when such damages are considered.

The Baseline scenario of AD-MERGE 2.0 projects a persistent increase in total GHG emissions throughout the 21st century, with a notable increase until 2040, followed by a more moderate trend until 2100. This trajectory is consistent with the high-emission pathway of SSP3-7.0 and yields an estimated temperature increase of 3.6 °C by 2100 (see Fig. a). The Baseline scenario highlights the need for additional decarbonization measures and shows the deviation from effective carbon mitigation strategies. The REF-NA scenario depicts a continuous decline in emissions falling well below Baseline levels, leading to a potential temperature rise to 2.3 °C. Although this trend demonstrates that the inclusion of climate damages and current policies helps shift emissions toward lower trajectories, it underscores that additional efforts are still needed to limit the 2 °C goal set by the Paris Agreement . The implementation of adaptation measures in the REF-OA scenario results in a reduction in climate-related damages. This, in turn, decreases the marginal benefits of mitigation, leading to a reduction in mitigation efforts and ultimately higher temperature increases that reach 2.4 °C by the end of the century. The APS-OA scenario, which replicates the IEA-APS emission trajectory for energy-related CO₂ emissions, represents a more ambitious climate mitigation pathway. In this scenario, alignment with the announced policy targets results in an emission pathway that limits global warming to 2 °C by 2100. Our scenarios fall into the following AR6 temperature categories: Baseline is between C7–C8 categories, REF scenarios between C5–C6, and APS between C3–C4.

Figure b depicts the trajectories of energy-related CO₂ emissions and allows comparison with IEA scenarios. As discussed in Section , AD-MERGE 2.0 Baseline is consistent with the SSP3-7.0, although the emissions are slightly higher in the short run. For the REF-NA scenario, energy-related CO₂ emissions broadly follow the IEA-STEPS trajectory in the short run but fall below it in the long run, suggesting AD-MERGE responds more strongly to similar policy assumptions. Consistent with the pattern for total GHG emissions, the REF-OA scenario reduces energy-related emissions, but leaves them slightly higher than under REF-NA. Under APS-OA scenario, emissions further decline over the century, falling to less than 8 Gt CO₂ per year by 2100. Non-CO₂ GHG emissions under the Baseline 2.0 scenario continue to rise steadily throughout the century, as shown in Fig. c and d. In contrast, the REF-NA and APS-OA scenarios exhibit consistent downward trends in emissions of nitrous oxide (N₂O) and methane (CH₄) (see Fig. c and d) with the APS-OA scenario achieving the highest reductions over the century.

Primary energy is defined as energy inputs to conversion and transformation processes, including fuels used for electricity generation, whereas final energy represents energy delivered to end-use sectors after accounting for conversion losses and energy-sector own use. Consistent with IPCC AR6 WGIII reporting conventions , we use the direct-equivalent method for non-combustible electricity sources, counting one unit of electricity from non-combustible sources as one unit of primary energy, rather than converting output to a fossil-fuel-equivalent input through an assumed efficiency. Considering these definitions, the energy supply projections (Fig. ) show a significant shift in the global energy landscape under the decarbonization scenarios, REF-NA, REF-OA, and APS-OA, compared to the Baseline scenario. In the Baseline scenario, primary energy supply reaches 1221 EJ and final energy use reaches 875 EJ by 2100, whereas all alternative scenarios result in reduced levels of both metrics. The reduction is primarily driven by mitigation policies that increase production costs and the inclusion of climate damages, which both yield GDP losses, resulting in lower energy demands.

In the Baseline scenario, the global energy supply continues to grow steadily, with fossil fuels maintaining their predominance in the energy supply throughout the century. By 2100, fossil fuels will account for 75 % of the total primary energy supply, with coal leading at 28 %, closely followed by oil at 27 %. Renewable energy sources, such as wind, solar, and biomass, see only limited expansion, contributing just 15 % to the overall energy mix. In contrast, the REF-NA and REF-OA scenarios show a marked decline in coal and oil use, paired with a notable increase in renewable energy, 40 % and 32 % of the total energy supply, respectively. In the APS-OA scenario, this share increases further to 55 %, due to the stronger policy commitment to decarbonization.

Across all three decarbonization scenarios, final energy use undergoes a major transformation, marked by a decline in fossil fuel reliance and a growing role for electricity, hydrogen, and renewables. Electrification accelerates as the share of electricity in final energy rises from 29% in the Baseline to 38 % in REF-NA, 36 % in REF-OA, and 40 % in APS-OA, reflecting lower decarbonization costs in the power sector and rapid advances in renewable and storage technologies. Hydrogen also moves from a marginal role in the Baseline scenario, where it contributes just 2 % of final energy (mainly gray hydrogen), to a more central position in the policy scenarios, doubling to 5 % in REF-NA and REF-OA and reaching 9 % in APS-OA by 2100. This shift is driven by the scale-up of green hydrogen from renewables and blue hydrogen with CCS, enabling deeper decarbonization of hard-to-electrify sectors and expanding hydrogen’s role as a flexible energy carrier in a low-carbon future.

Bioenergy and carbon-removal technologies become increasingly important contributors to deep decarbonization in all policy scenarios. Throughout the century, the share of biomass in global final energy use rises from just 5 % in the Baseline to 14 % in REF-NA, 11 % in REF-OA, and 19 % in APS-OA by 2100. Most of the biomass is supplied via Fischer–Tropsch synthesis and bioenergy with carbon capture and storage. BECCS is absent in the Baseline, but by 2100 it accounts for roughly 72 % of bioenergy use in REF-NA, 66 % in REF-OA, and 78 % in APS-OA, positioning it as the dominant negative-emissions option in these pathways. The strong uptake of BECCS not only supplies low-carbon fuels and electricity but also delivers net removals that offset residual emissions from hard-to-abate sectors, enabling the tighter carbon budgets embodied in the policy scenarios. A detailed breakdown of regional BECCS deployment and feedstock sourcing is provided in Appendix  to illustrate the geographic concentration of bioenergy resources and capture potential.

Figure  illustrates the economic implications of global and regional climate change damages and adaptation options. Figure a compares global climate change costs, i.e., the sum of climate change damages and adaptation costs, as a percentage of global GDP from 2030 to 2100 under different scenarios. Costs in the Baseline scenario, given no adaptation, represent gross damages and increase significantly, reaching about 10 % of GDP by 2100, clearly highlighting the economic burden associated with climate inaction. Implementing mitigation targets lowers these impacts to about 6.0 % in REF-NA and 5.2 % in APS-NA, without any adaptation measures. In contrast, the REF-OA and APS-OA scenarios demonstrate the substantial potential for targeted mitigation and adaptation measures, significantly reducing climate change costs to less than 2 % of GDP by 2100. This trend underscores adaptation's pivotal role in minimizing climate-induced economic impacts.

The divergence between the NA and OA scenarios by 2030 reflects their roles as bounding cases rather than near-term forecasts. NA represents a counterfactual no-adaptation pathway, while OA illustrates an idealized optimal-adaptation pathway. In the OA scenario, adaptation is deployed from 2020 onward at a model-implied cost-effective level for each impact category, whereas NA assumes zero protective adaptation. Early reductions in climate change impacts are driven by both reactive and proactive adaptation. Reactive measures can be scaled within a single model period. Proactive measures, such as investments in seawalls, irrigation infrastructure, and resilient energy and health systems, accumulate more gradually through depreciable investments, although they still yield benefits within a 10-year model period. Because these scenarios abstract from real-world policy and implementation frictions, their differences around 2030 should be interpreted as illustrative of the model's internal damage-adaptation dynamics. Notably, the costs of REF-OA and APS-OA diverge from 2060 at the global level. This global divergence leads to regional disparities, where the impacts and benefits of adaptation are more pronounced. A more detailed regional analysis would reveal greater variation and highlight areas where adaptation efforts are most effective (see Fig. d).

Figure b and c decomposes the total climate change costs into residual damages, adaptation costs (reactive and proactive), and avoided damages for the REF-OA and APS-OA scenarios, respectively. When comparing avoided damages (adaptation benefits) with adaptation costs, the high cost-effectiveness of adaptation is evident, where adaptation benefits outweigh costs threefold. On a global scale, proactive and reactive adaptation costs are initially dominated by higher investments in proactive measures, but over time they converge to similar magnitudes. Figure b and c also allows interpretation of the corresponding APS-NA and REF-NA scenarios. In the absence of adaptation, total climate-change costs would consist of the sum of residual damages and avoided damages shown here, since the avoided-damage component represents the damages prevented by optimal adaptation. Thus, APS-NA can be inferred from Fig. c as the level of climate-change costs that would prevail under the APS mitigation pathway without adaptation, while APS-OA shows how those gross damages are redistributed into residual damages, adaptation expenditures, and avoided damages once optimal adaptation is introduced. The same interpretation applies to REF-NA relative to REF-OA in Fig.b. Therefore, comparing REF and APS scenarios further shows that stronger mitigation in the APS scenario lowers both the gross damages (in NA case) and the residual damages (in OA case).

The baseline climate impact analysis reveals a profound inequality in exposure, characterized by extreme impacts in the Global South and significant heterogeneity among developed economies. In particular, regional climate change impacts (damages and adaptation costs) as a share of GDP in Fig. d highlight that vulnerable regions such as Africa and Other Asia regions face disproportionately high climate-related damages, where their climate change costs exceed 18 %–21 % of GDP. In comparison, high-latitude regions of Russia and Canada experience low impacts. Other vulnerable regions are Central and Latin America and India, which experience significantly higher than average climate-related damages, particularly under the Baseline scenario. Western Europe and the Middle East experience a comparable rate of impacts. While mitigation alone (REF-NA) reduces the absolute ceiling of damages, it fails to alter this structural disparity.

The analysis underscores adaptation as a critical divergence point, although its effectiveness varies markedly between the 15 regions. Regions with high baseline vulnerability, such as Africa, Other Asia, and Central and Latin America, alongside those with high adaptive capacity like the USA, China, and Western Europe, experience the greatest relative gains from adaptation measures. This contrast highlights a significant disparity: while mitigation is universally beneficial, adaptation delivers disproportionately large benefits to vulnerable developing regions that have contributed the least to historic emissions. By disaggregating these regions, this study reveals inequalities that were previously masked in AD-MERGE 1.0, emphasizing the need for comprehensive policy frameworks that integrate both strategies to achieve a more equitable distribution of climate risk.

Overall, while mitigation policies under the REF-NA and APS-NA scenarios achieve a general reduction in impacts, the incorporation of adaptation measures provides a crucial additional layer of protection, significantly lowering residual damages. The cumulative benefit of combining these strategies is evident in all scenarios, yet distinct regional vulnerabilities persist. Even under the most ambitious combined strategy (APS-OA), Africa and India remain critically exposed, retaining residual damage exceeding 3 % of GDP, while the Middle East and Other Asia face impacts above 1.5 %. These findings demonstrate that, while combined strategies effectively minimize economic consequences globally, they do not fully eliminate high-risk exposure of the most vulnerable regions.