NGFS Climate Scenarios — The Definitive Reference

Executive Summary

The NGFS Climate Scenarios are the reference set of forward-looking climate pathways used across the global financial system to assess climate-related risk. Produced by the Network for Greening the Financial System — a coalition of central banks and supervisors — they take a small number of stylised assumptions about how climate policy unfolds and how severe physical climate impacts become, and run them through a chain of climate-economy models to produce internally consistent projections of macro-financial variables: GDP, inflation, carbon prices, energy mix, and physical-risk indicators, over horizons that extend to 2100.

The scenarios come in two families. The long-term scenarios — seven pathways in the current vintage — explore structural transition and physical-risk outcomes to end-of-century and are the ones most people mean by “the NGFS scenarios.” The short-term scenarios, first published in 2025, capture near-term shocks over a three-to-five-year horizon for capital planning and credit-risk work. Both are designed for the financial sector specifically, which distinguishes them from the climate-science scenarios of the IPCC or the energy-system scenarios of the IEA that they draw on.

One development dominates the current picture. In the latest long-term vintage (Phase V, November 2024) the NGFS adopted a new physical-risk damage function that produced markedly larger projected GDP losses than earlier vintages. The academic paper underpinning that damage function was subsequently retracted, and in December 2025 the NGFS issued a statement flagging the affected outputs and committing to a revised methodology for the next release, expected at the end of 2026. Any current use of the Phase V physical-risk numbers must be read against that caveat — a point this page returns to in detail.

What Are the NGFS Climate Scenarios?

The NGFS Climate Scenarios are a set of standardised, publicly available climate pathways designed to give the financial sector a consistent analytical basis for assessing climate-related risks and opportunities. Rather than a single projection, they are a deliberately chosen spread of plausible futures, each defined by a coherent set of assumptions about policy ambition, technology, behaviour, and physical climate outcomes, and each rendered into the economic and financial variables a bank, insurer, asset manager, or supervisor needs.

Three features define them. First, they are conditional, not predictive: each scenario asks “if the world follows this policy-and-physics path, what happens to the economy and the financial system?” — the NGFS is explicit that they are not forecasts. Second, they are internally consistent: the carbon price, energy mix, emissions, temperature, and GDP in a given scenario are all products of the same modelling run, so they cannot be mixed and matched across scenarios without breaking that consistency. Third, they are macro-financial: their output is not degrees of warming for its own sake but the financial transmission of climate change — GDP paths, sectoral output, carbon prices, and physical damages.

They are used to answer supervisory and risk-management questions that a purely scientific scenario cannot: how a disorderly transition would revalue carbon-intensive assets, how chronic physical risk would erode collateral values, how a carbon-price shock would ripple through a credit portfolio. That financial-transmission focus is what makes them a standard rather than an academic exercise.

Why the Scenarios Exist

Before the NGFS scenarios, financial institutions attempting climate stress tests faced a coordination problem. Climate science produced emissions and temperature pathways (the IPCC), energy analysts produced fuel-mix pathways (the IEA), and macroeconomists produced growth models — but nothing joined them into a financial-sector-ready package. Two institutions running the same nominal “2°C scenario” could produce incomparable results because they used different carbon prices, different damage assumptions, and different macro models.

The NGFS was launched at the Paris One Planet Summit on 12 December 2017 by a group of central banks and supervisors to close exactly this gap: to build a common set of scenarios so that climate risk could be assessed consistently across institutions and jurisdictions, and so that supervisory stress tests could be benchmarked. The first vintage arrived in 2020, and successive updates have added scenarios, refined the models, and — increasingly — integrated physical as well as transition risk.

The scenarios matter because they have become infrastructure. Central-bank supervisory stress tests, prudential frameworks, and internal risk models across dozens of jurisdictions are built on them, and disclosure standards point to them as an acceptable source for the resilience analysis they require. A change to the NGFS scenarios — a new damage function, a new scenario, a data caveat — therefore propagates into supervisory exercises and corporate disclosures worldwide.

Governance: Who Owns the Scenarios

The Network for Greening the Financial System is a voluntary coalition of central banks and financial supervisors that share best practice on environmental and climate risk in the financial sector. As of its most recent membership statements it brings together 148 central banks and supervisors and 23 observers. It is chaired by Sabine Mauderer, First Deputy Governor of the Deutsche Bundesbank, with Fundi Tshazibana of the South African Reserve Bank as Vice Chair; the Secretariat is headed by Yann Marin and provided by the Banque de France.

The scenarios themselves are produced by a dedicated NGFS workstream in partnership with an external academic consortium of climate-economy modellers. The integrated assessment models, the macroeconomic model, and the physical-risk damage functions are run by research institutions — among them the Potsdam Institute, IIASA, the University of Maryland’s Center for Global Sustainability, and the National Institute of Economic and Social Research — under the NGFS’s direction. The NGFS sets the scenario narratives and assumptions; the consortium runs the models; the outputs are published through the NGFS Scenarios Portal and the IIASA Scenario Explorer.

Membership is a matter of ongoing note. The scenarios remain a global reference, but the network’s coverage is not universal: the United States’ principal financial regulators — the Federal Reserve, the Federal Deposit Insurance Corporation, and the Treasury — withdrew from the NGFS in 2025, a departure that is part of the broader retreat of US federal regulators from climate-specific supervisory activity. The scenarios continue to be maintained and used across the large majority of other major jurisdictions.

The Two-Axis Framework and the Four Quadrants

The organising logic of the NGFS long-term scenarios is a two-axis uncertainty matrix. The vertical axis is transition risk — the disruption caused by policy, technology, and market shifts as the economy decarbonises. The horizontal axis is physical risk — the damage caused by the changing climate itself, from chronic effects like heat and drought to acute events like floods and cyclones. Every scenario is a point in that space, and the scenarios are grouped into four quadrant families.

The value of the quadrant structure is that it forces a risk manager to think in two dimensions at once. A common analytical error is to treat “worse climate outcome” as a single dial; the matrix makes explicit that an orderly net-zero path carries high transition risk but low physical risk, while a hot-house path carries the reverse. A robust stress test deliberately samples across quadrants rather than picking points along a single line.

The Seven Long-Term Scenarios

The current long-term framework comprises seven scenarios distributed across the four quadrants. The narratives below are the standing definitions carried through the recent vintages; the specific numbers attached to each (temperature, GDP loss, carbon price) are vintage-specific and are discussed, with their caveats, in the Phase V section.

Two of these are relatively recent additions. Low Demand (in the Orderly quadrant) and Fragmented World (in the Too Little Too Late quadrant) were introduced in the Phase IV vintage, at which point the earlier “Divergent Net Zero” scenario was removed — bringing the count to the current seven. Reading older material that references Divergent Net Zero, or that lists six scenarios, is a reliable sign that it predates Phase IV.

The Model Chain: IAMs, NiGEM, and Damage Functions

The NGFS scenarios are not the product of a single model but of a chain, and understanding the chain is essential to interpreting the outputs correctly. Each link introduces its own assumptions and its own uncertainty.

Integrated assessment models (transition pathways)

The transition side — emissions trajectories, energy mix, carbon prices, land use — is produced by three integrated assessment models (IAMs): GCAM, MESSAGEix-GLOBIOM, and REMIND-MAgPIE. Running three models rather than one is deliberate: it exposes the model-dependence of the results, and the NGFS presents outputs across all three so users can see where the models agree and where they diverge. A carbon price in a given scenario is therefore a range across models, not a single number.

The macroeconomic model (financial translation)

The IAM outputs are passed into a macroeconomic model, NiGEM (the National Institute Global Econometric Model), which converts the physical and transition pathways into macro-financial variables — GDP, inflation, unemployment, interest rates, and sectoral output. This is the step that turns a climate-energy scenario into something a financial institution can use.

Damage functions (physical risk)

The physical-risk side — the economic damage from a changing climate — is estimated through damage functions calibrated to climate and economic data, plus more granular acute-risk modelling for events like floods, droughts, and heatwaves. The choice of damage function is one of the most consequential and most contested modelling decisions in the whole chain, because it largely determines how severe the projected GDP losses are. This is precisely where the Phase V change, and the subsequent retraction, occurred.

What the Scenarios Output

The deliverable of an NGFS scenario is a large set of time-series variables, published to 2100 (long-term) at global, regional, and country level, accessible through the NGFS Scenarios Portal and the IIASA Scenario Explorer. The variables that matter most to financial users fall into three groups.

The single most-used transition variable is the carbon price (or shadow carbon price), because it is the mechanism through which the models drive emissions reductions and the most direct input to asset-repricing analysis. In the orderly scenarios the carbon price rises early and steadily; in Delayed Transition it stays low then jumps sharply after 2030; in the hot-house scenarios it stays low throughout. That shape difference is often the heart of a transition stress test.

Short-Term Scenarios

Historically the NGFS scenarios were exclusively long-term, extending to 2100 — well suited to strategic and structural analysis but a poor fit for the three-to-five-year horizons that dominate capital planning, credit provisioning, and near-term supervisory stress tests. In May 2025 the NGFS published the first vintage of short-term scenarios to close that gap.

The short-term scenarios model near-term shocks to the economy and financial sector — abrupt climate-policy shifts and extreme weather events — and trace how those shocks transmit to and are amplified by the financial system over a few years rather than decades. They are narrative-driven, pairing a policy or physical shock with its macro-financial propagation, and are designed to sit alongside the long-term scenarios rather than replace them: long-term for structural and strategic questions, short-term for cyclical and capital-planning ones.

Version History and Where We Are Today

The scenarios have been revised on a roughly annual-to-biennial cadence since 2020. Each vintage updates the underlying economic and climate data, the policy commitments reflected, and the model versions — and, periodically, the scenario set itself. Reading a scenario as if it were unchanged across vintages is a common error, because both the narratives and the numbers move.

Alongside the scenarios, the NGFS maintains a companion Guide on Climate Scenario Analysis, first published in 2020 and updated in November 2025, which sets out best practice in scenario design, data, and modelling for institutions running their own analyses. The Guide is the practical bridge between the raw scenario data and a usable supervisory or internal exercise.

Phase V and the New Damage Function

Phase V, published in November 2024, was dominated by a single methodological change: a new physical-risk damage function. Earlier vintages had used a damage function based on Kalkuhl and Wenz (2020); Phase V switched to the approach of Kotz et al. (2024), which incorporated additional climate drivers beyond mean temperature — temperature variability, annual precipitation, the number of wet days, and extreme daily rainfall — and modelled more persistent effects of climate shocks, with impacts accumulating for up to a decade after an event.

The consequence was a step change in projected damages. The NGFS reported that estimated global losses from chronic physical risk were significantly higher than in the previous vintage — by a factor of roughly two to four across scenarios in 2050 — and its own communications noted that GDP losses by 2050 could be two to four times greater than previously estimated. In the high-warming scenarios this translated into headline figures of GDP losses on the order of 15% by 2050 at around 2°C and around 30% by 2100 at higher warming, figures that were widely cited and that fed into some official economic projections.

The Phase V package also incorporated the latest GDP and population projections from the IPCC’s Shared Socioeconomic Pathways and country-level climate commitments as of March 2024, and came with supporting documents: an overview of the update, an explanatory note on the new damage function, updated technical documentation, and companion notes on climate adaptation and the state of climate finance. The transition side of Phase V was an incremental update; the physical-risk side was the headline — and, as it turned out, the vulnerability.

The Kotz et al. Retraction and the Data Caveat

The academic paper underpinning the Phase V physical-risk estimates, Kotz et al. (2024), was subjected to post-publication scrutiny at Nature — two formal “Matters Arising” critiques — and was ultimately retracted from the journal. The retraction concerned methodological and data issues in the paper; the authors indicated they intended to resubmit a revised version for peer review. Because Phase V had adopted that paper’s damage function for its physical-loss estimates, the retraction reached directly into the scenario outputs.

In December 2025 the NGFS issued a formal statement addressing the situation. It advised users to be aware of the retraction when interpreting and applying Phase V results, identified precisely which outputs were affected, and confirmed that a revised physical-risk methodology would be used for the next long-term vintage, expected end of 2026. The statement is careful about scope, and reproducing that scope accurately matters:

The NGFS framed the episode as part of normal scientific debate and reiterated that its scenarios are illustrative tools rather than forecasts, adding the standing warning that real economic impacts could be more severe than modelled — for instance if tipping points are reached — and that nature-related and tail risks are not fully captured. External commentators split sharply on interpretation: some argued the affected Phase V physical-risk results should be set aside entirely until re-based, while others held that the direction of travel (larger damages than older vintages assumed) was likely robust even if the specific magnitudes were not. Institutions that had already built the Phase V figures into official projections faced pressure to revisit them.

How Financial Institutions Use the Scenarios

The scenarios are applied across a spectrum of financial-sector activities, from binding supervisory exercises to voluntary internal risk management.

The financed-emissions and portfolio-alignment work that sits behind much of this — measuring the emissions attributable to loans and investments — is governed by separate accounting standards such as the PCAF financed-emissions standard. The NGFS scenarios supply the forward-looking risk pathways; PCAF supplies the emissions baseline the pathways are applied to. The two are complementary layers of a climate-risk programme, not substitutes.

Interaction with TCFD, IFRS S2, and CSRD

The NGFS scenarios are voluntary, but the disclosure regimes that reference them are increasingly not. Climate-disclosure standards require forward-looking scenario analysis, and the NGFS scenarios are among the sources those standards point to.

TCFD and IFRS S2

The recommendations of the Task Force on Climate-related Financial Disclosures established scenario analysis as a core expectation of climate disclosure. Those recommendations have been carried into the ISSB’s IFRS Sustainability Disclosure Standards: IFRS S2 requires an entity to assess the resilience of its strategy to climate-related changes using climate-related scenario analysis, and the NGFS scenarios are among the authoritative sources — alongside the IPCC and IEA — recognised for that purpose. The TCFD recommendations reference page covers the disclosure architecture the ISSB inherited.

CSRD and ESRS E1

Under the EU’s Corporate Sustainability Reporting Directive, ESRS E1 requires disclosure of climate-related risks and, for many undertakings, resilience analysis that scenario analysis supports. The NGFS pathways are a common input to the transition-plan and physical-risk elements of ESRS E1 reporting, though ESRS requires substantially more than a scenario run alone.

Prudential and jurisdictional requirements

Beyond corporate disclosure, prudential frameworks increasingly mandate scenario analysis. European supervisors have moved to integrate climate and ESG risk into stress testing under the banking and insurance directives, the Bank of England embeds climate scenarios in its supervisory framework, and jurisdictions from Australia to Singapore have adopted ISSB-aligned requirements. In the United States, even as federal regulators stepped back, state measures such as California’s SB 261 require TCFD-aligned climate-risk disclosure that effectively calls for scenario analysis.

Relationship with IPCC and IEA Scenarios

The NGFS scenarios are frequently confused with the IPCC’s and the IEA’s, and the differences are what make the NGFS set fit for financial-risk work.

The practical distinction is that the IPCC tells you how much the world might warm, the IEA tells you how the energy system might change, and the NGFS tells you what either might do to a balance sheet. A firm assessing climate risk needs the NGFS layer precisely because the scientific and energy scenarios stop short of the financial variables. The IPCC AR6 and IEA reference pages cover the upstream sources the NGFS draws on.

Selecting and Applying Scenarios

Applying the scenarios well is a discipline in its own right, and the NGFS Guide on Climate Scenario Analysis exists to support it. A defensible application runs roughly as follows.

1. Set the horizon. Short-term scenarios (three to five years) for credit risk, provisioning, and capital planning; long-term scenarios (to 2050 or 2100) for strategy, business-model resilience, and structural asset-repricing.
2. Span the quadrants. Select at least three scenarios that occupy different risk profiles — typically an orderly baseline, a high-transition-risk disorderly path, and a high-physical-risk hot-house path — rather than points along a single line.
3. Choose the model view. Decide whether to use a single IAM, a median across the three, or the full range; document the choice, because carbon prices and pathways differ materially across GCAM, MESSAGEix-GLOBIOM, and REMIND-MAgPIE.
4. Map variables to exposures. Connect carbon prices to carbon-intensive assets, physical-risk indicators to real-estate and agricultural collateral, and macro variables to portfolio-wide overlays.
5. Apply the physical-risk caveat. Where Phase V chronic-physical-loss variables are used, flag the Kotz et al. retraction and consider sensitivity around the magnitude until the revised methodology arrives.
6. Document and disclose. Record the vintage, scenarios, models, and horizon used, so the analysis is reproducible and auditable — and so a change of vintage can be tracked.

For the disclosure-facing side of this workflow — turning a scenario run into a TCFD- or IFRS-S2-aligned resilience narrative — the GreenCalculus TCFD scenario-analysis calculator and its methodology document how scenario pathways feed a structured resilience assessment.

Common Misinterpretations

Common Application Errors

1. Using a stale vintage. Running Phase III or IV numbers, or citing six scenarios or Divergent Net Zero, when the current long-term vintage is Phase V with seven scenarios.
2. Reporting Phase V physical-loss figures without the caveat. Quoting the headline GDP-loss numbers as settled, post-December-2025, without disclosing the Kotz et al. retraction and the affected-variable scope.
3. Averaging across IAMs silently. Presenting a single carbon-price or GDP figure without stating whether it is one model, a median, or a range — obscuring the model-dependence the NGFS deliberately exposes.
4. Mismatching horizon to question. Using long-term (2100) scenarios for three-year capital planning, or short-term scenarios for structural asset-repricing.
5. Cherry-picking a single scenario. Running only one pathway — often the most benign — rather than spanning the quadrants, which defeats the purpose of a stress test.
6. Treating chronic and acute physical risk as additive without care. Aggregating a chronic damage function with separately-modelled acute damages can double-count where the damage function already embeds some acute drivers — an issue the NGFS has flagged for the Phase V damage function specifically.
7. Failing to document the run. Not recording vintage, scenarios, models, and horizon, so the analysis cannot be reproduced or compared when the next vintage lands.

Criticisms and Limitations

The NGFS scenarios are the most widely used climate-risk pathways in finance, and they attract substantive criticism that a complete reference must set out plainly.

Physical-risk damages are hard to model — and were mis-modelled in Phase V

The damage-function problem is the deepest limitation. The magnitude of physical-risk GDP losses is highly sensitive to the choice of damage function, and the Phase V adoption of a since-retracted function is the sharpest illustration: a methodological choice moved headline losses by a factor of two to four, and then had to be flagged. Physical-loss numbers should be treated as directional and uncertain, not precise.

Tipping points, nature, and compound risk are under-captured

Critics, including within the actuarial and central-banking communities, argue the scenarios may exclude the most severe systemic outcomes — tipping points, cascading failures, nature loss, climate migration — and so understate true risk. The NGFS acknowledges this and urges users to adjust for missing drivers themselves.

Continued-growth assumptions may flatter the results

The scenarios generally assume economies keep growing, and even growing faster, despite high warming. Commentators have argued this makes even the larger Phase V damage estimates conservative, because it embeds an optimistic macro backdrop into a high-damage climate.

Model-dependence and opacity

Results differ across the three IAMs and the macro model, and the mapping from scenario to financial variable involves many assumptions that are not always transparent to end users. Two firms using the same scenario but different model views can reach different conclusions.

Governance and coverage

The withdrawal of major US federal regulators from the NGFS raises questions about the universality of the standard, and the retraction episode drew scrutiny of how external academic inputs are quality-assured before they enter a framework used for supervision and disclosure.

Future Evolution

Three trajectories will shape the scenarios over the next several years.

The end-of-2026 long-term vintage. The next release will carry a revised physical-risk methodology, replacing the retracted Phase V damage function, alongside the usual updates to data, policy commitments, and model versions. It is the release that will re-base the physical-loss numbers and resolve the current caveat, and long-lead physical-risk work should plan around it.

Deeper integration of missing risk drivers. The NGFS has signalled intent to better capture acute physical risk, and separately has begun work on nature-related economic and financial risk, publishing technical recommendations for nature-scenario design. Bringing tipping points, compound events, and nature loss into the frame is the direction of methodological travel.

Consolidation as disclosure infrastructure. As IFRS S2 adoption spreads and prudential scenario-analysis requirements mature across jurisdictions, the NGFS scenarios are likely to become still more embedded as the default financial-risk pathways — even as their governance and coverage evolve. The market infrastructure built around them tends to persist across methodological revisions.

Frequently Asked Questions

Sources and References

Every factual and numerical claim on this page reconciles to the primary NGFS publications below. Where the NGFS has published a definitive document, it is cited directly; secondary commentary is used only for interpretation, and all values are hardcoded to their dated source.

Primary NGFS publications

• NGFS, NGFS Climate Scenarios for central banks and supervisors — Phase V, November 2024 (long-term scenarios; new physical-risk damage function).
• NGFS, Phase V package supporting documents — overview of updates, explanatory note on the new damage function, and updated technical documentation of the modelling framework and assumptions.
• NGFS, NGFS Short-Term Climate Scenarios, first vintage, May 2025 (3–5 year near-term shocks).
• NGFS, Guide on Climate Scenario Analysis, updated edition, November 2025 (first edition 2020).
• NGFS, Statement regarding physical risk estimates in Phase V of NGFS long-term scenarios, December 2025 (Kotz et al. (2024) retraction; affected-variable scope; revised methodology for the next vintage end of 2026).
• NGFS Scenarios Portal and IIASA NGFS Scenario Explorer — the data-access surfaces for scenario variables and the Climate Impact Explorer.
• NGFS, Technical recommendations for the development of scenarios for nature-related economic and financial risks.
• NGFS press releases on the Phase V publication and the updated Guide on Climate Scenario Analysis (membership figures, governance).

Underlying models and scientific basis

• Integrated assessment models: GCAM, MESSAGEix-GLOBIOM, REMIND-MAgPIE.
• Macroeconomic model: NiGEM (National Institute Global Econometric Model).
• Physical-risk damage functions: Kalkuhl and Wenz (2020) (pre-Phase V); Kotz et al. (2024) (Phase V, subsequently retracted from Nature).
• IPCC Shared Socioeconomic Pathways (SSPs) — socioeconomic and climate inputs.
• IEA World Energy Outlook and Net Zero Roadmap — energy-system inputs.

Disclosure and supervisory context

• IFRS Sustainability Disclosure Standards, IFRS S2 (Climate-related Disclosures) — resilience analysis via climate-related scenario analysis. ISSB, 2023.
• Task Force on Climate-related Financial Disclosures (TCFD) — scenario-analysis recommendations.
• European Sustainability Reporting Standards, ESRS E1 (Climate change). EFRAG, 2023.
• European Supervisory Authorities and Bank of England climate stress-testing frameworks; California SB 261 climate-risk disclosure.

Related GreenCalculus reference pages

• IFRS S2 — Climate-related Disclosures
• TCFD Recommendations
• CSRD / ESRS E1
• PCAF Financed Emissions Standard
• IPCC AR6
• IEA Global Energy Review
• California SB 253 / SB 261
• TCFD scenario-analysis calculator
• TCFD scenario-analysis methodology

Editorial scope and methodology

Editorial scope. This page documents the NGFS Climate Scenarios as they stand on the date in the header above. The scenarios are revised periodically and the physical-risk methodology is under active revision; primary NGFS publications should be consulted for any time-sensitive application. GreenCalculus is independent of the NGFS and provides this reference for informational and educational purposes.

Methodology. Every value is hardcoded against its dated NGFS publication, including the flagged status of the retracted Phase V physical-risk inputs. No live-reference shortcodes are used on this page — the scenarios at any given vintage are a historical record. When the NGFS publishes a new vintage or updates the physical-risk caveat, this page is reviewed and revised through the GreenCalculus changelog process at /changelog/.