Assumption Registry

Every modelling assumption with rationale, counterarguments, failure conditions, and alternative approaches.

4
Critical
5
High
4
Medium
0
Low
JSON
A-CLM-001 Climate HIGH validated_against_cmip6
TCR is the appropriate sensitivity metric for near-term warming projections (2025–2060).
TCR reflects transient (realised) warming over the policy-relevant horizon. ECS (equilibrium) overestimates near-term warming since ocean heat uptake is slow.
Supporting evidence

IPCC AR6 WG1 Section 4.3; Gillett et al. (2021) Geophysical Research Letters

Counterarguments

For projections beyond 2100, ECS becomes more relevant; TCR underestimates long-run commitment. Some models show TCR/ECS ratio departing from 0.6 standard assumption.

Failure conditions

Would fail if rapid warming commitment (zero-emission commitment) substantially exceeds current IPCC estimate of 0.1–0.5°C

Alternative assumptions (2)
  • Use ECS with explicit ocean heat uptake parameterisation
  • Use TCRE (transient climate response to cumulative emissions) directly
Affected models: ClimateModelService Scenarios: all Owner: Climate Science Team Reviewed: 2026-01-15
A-CLM-002 Climate HIGH annually_updated
The 2025 baseline emissions are 57.4 GtCO₂e/yr under current policies.
UNEP Emissions Gap Report 2024 is the most current and comprehensive global GHG accounting.
Supporting evidence

UNEP EGR 2024 (Nov 2024); Global Carbon Project 2024; Climate Watch GHG data

Counterarguments

Methane emissions may be underestimated by 40–60% per recent atmospheric measurements (Worden et al. 2017). Land use emissions highly uncertain (±2 GtCO₂e).

Failure conditions

If methane inventory corrections materially revise total; if COVID rebound was larger/smaller than estimated

Alternative assumptions (1)
  • Use GCP 2024 estimate of 54–56 GtCO₂ (CO₂-only); note non-CO₂ gases add ~10 GtCO₂e
Affected models: ClimateModelService, GapAccountingEngine Scenarios: all Owner: Climate Science Team Reviewed: 2026-02-01
A-CLM-003 Climate MEDIUM partially_validated
Linear interpolation between 2025, 2030, 2040, 2050 emissions waypoints is sufficient for pathway modelling.
Linear interpolation is standard in scenario planning tools and avoids spurious precision between anchor years.
Supporting evidence

IPCC AR6 WG3 scenario methodology (similar approach in C1 pathway representation)

Counterarguments

Technological deployment follows S-curves, not linear ramps. Linear interpolation underestimates early-decade inertia and late-decade acceleration.

Failure conditions

Causes systematic over-optimism in 2025–2030 window; use S-curve deployment for technology-specific projections

Alternative assumptions (1)
  • Logistic (S-curve) deployment for each technology sub-sector
Affected models: ClimateModelService, GapAccountingEngine Scenarios: all Owner: Climate Science Team Reviewed: 2026-01-15
A-DAM-001 Physical Damage CRITICAL calibrated_to_literature
The quadratic DICE damage function is used as the central damage estimate.
DICE is the most widely cited integrated assessment model damage function, enabling comparison with published literature.
Supporting evidence

Nordhaus (2023) DICE-2023; meta-analysis in Howard & Sterner (2017) with 27 studies

Counterarguments

Quadratic form dramatically underestimates high-temperature damage. Weitzman 'Dismal Theorem' suggests heavy tails dominate expected utility. Burke et al. (2015) non-linear function gives 3–5× higher damage at 4°C.

Failure conditions

Fails at temperatures above ~3.5°C where tipping points, migration, and conflict become dominant damage channels not in DICE

Alternative assumptions (3)
  • Burke et al. (2015) non-linear damage function — gives ~23% global GDP loss at 3°C
  • Weitzman (2010) tipping-point adjusted damages
  • META model (Dietz et al. 2021) — explicit tipping cascade
Affected models: DamageModelService Scenarios: all Owner: Economics & Damage Team Reviewed: 2026-01-20
A-DAM-002 Physical Damage HIGH partially_validated
Catastrophe model hazard distributions are treated as stationary with a climate non-stationarity adjustment factor applied.
Current cat model vendor hazard datasets are calibrated to historical periods. A multiplicative adjustment factor scales AAL for projected climate change.
Supporting evidence

RMS (2022) climate change adjustment methodology; Swiss Re Institute climate risk reports

Counterarguments

Hazard non-stationarity may be discontinuous (tipping points, jet stream shifts). Multiplicative factors may be insufficient for extreme tail events.

Failure conditions

Fails for perils with strong non-linear hazard amplification under warming (tropical cyclone intensification at high warming levels)

Alternative assumptions (1)
  • Fully dynamic hazard modelling using CMIP6 GCM outputs at peril-level
Affected models: CatastropheModel, DamageModelService Scenarios: SSP3-7.0, SSP5-8.5 Owner: Catastrophe Risk Team Reviewed: 2026-01-20
A-ECO-001 Economics MEDIUM convention_based
Climate damages enter the production function multiplicatively (Ω scaling), not additively.
Multiplicative entry avoids GDP going negative at high damage levels; consistent with most IAM literature.
Supporting evidence

Dietz & Stern (2015); DICE model family convention

Counterarguments

Additive formulation (Stern-style) can result in GDP below zero at extreme warming — Weitzman argues this is realistic for civilisational collapse scenarios.

Failure conditions

Results in underestimation of damage at extreme warming; cannot represent true tipping-point economic collapse

Alternative assumptions (1)
  • Additive damage formulation with floor constraint at 0
Affected models: EconomicModelService Scenarios: high_warming Owner: Economics Team Reviewed: 2026-02-01
A-ECO-002 Economics MEDIUM stylised_fact
The MAC curve is smooth and convex — no technology step discontinuities.
Smooth MAC curves are standard in economic optimisation models and allow analytical treatment.
Supporting evidence

McKinsey MAC curve (2020); IPCC AR6 WG3 Chapter 3

Counterarguments

Real technology deployment follows lumpy investment decisions and has lock-in effects. The MAC curve is empirically non-convex around major technology transitions.

Failure conditions

Fails around major technology thresholds (e.g. when grid-scale storage crosses parity — cost drops discontinuously)

Alternative assumptions (1)
  • Step-function MAC with technology-specific discontinuities
Affected models: EconomicModelService, GapAccountingEngine Scenarios: all Owner: Economics Team Reviewed: 2026-02-01
A-ECO-003 Economics HIGH debated
Carbon markets function efficiently with full price signal transmission to emitters.
Efficient carbon markets are the standard assumption in policy optimisation literature.
Supporting evidence

Theoretically grounded; supported by EU-ETS empirical studies

Counterarguments

Real carbon markets have significant transaction costs, banking/borrowing rules, exemptions, price floors/ceilings, and political interference. Coverage rates remain low globally (~23% of emissions, World Bank 2024).

Failure conditions

Overestimates carbon revenue and price signal effectiveness in jurisdictions with weak carbon market institutions

Alternative assumptions (1)
  • Partial efficiency with explicit coverage rate and leakage parameters
Affected models: FiscalModelService, EconomicModelService Scenarios: policy_optimistic Owner: Policy & Finance Team Reviewed: 2026-02-01
A-TECH-001 Technology CRITICAL historically_not_validated
Breakthrough technologies scale from current TRL to commercial deployment by 2040 under optimistic scenario.
IEA NZE and IPCC AR6 WG3 scenarios require significant pre-commercial technology deployment.
Supporting evidence

IEA NZE 2050 technology deployment milestones; IRENA breakthrough technologies assessment

Counterarguments

Historical technology scale-up timelines rarely match policy projections. Nuclear fusion, advanced geothermal, and DAC have repeatedly missed deployment milestones. Technology-valley-of-death between demonstration and commercial scale is empirically persistent.

Failure conditions

Most scenarios relying on >5 Gt from currently sub-commercial technologies have no historical precedent of comparable scale-up

Alternative assumptions (1)
  • Conservative scenario: only TRL≥7 technologies achieve significant deployment by 2040
Affected models: GapAccountingEngine Scenarios: optimistic Owner: Technology Team Reviewed: 2026-02-05
A-TECH-002 Technology CRITICAL literature_range
BECCS ceiling is 4.5 EJ/yr biomass feedstock (approximately 2.5 Gt CDR/yr).
Land use and water constraints from IPCC AR6 WG3 Chapter 12 and dedicated land use analysis.
Supporting evidence

IPCC AR6 WG3 Chapter 12.3; Fajardy et al. (2018) Nature Climate Change

Counterarguments

Some integrated assessment models use up to 200 EJ/yr BECCS. Realistic sustainable biomass potential is highly contested.

Failure conditions

If BECCS ceiling is set too high, net-zero scenario becomes unachievable in practice due to competing land uses (food, biodiversity, carbon sinks)

Alternative assumptions (2)
  • No-BECCS scenario (Dooley et al. 2018)
  • High-BECCS scenario (IEA NZE low-land-use variant)
Affected models: GapAccountingEngine, LandValuationService Scenarios: all Owner: Land & Natural Capital Team Reviewed: 2026-02-05
A-UNC-001 Uncertainty HIGH known_limitation
Parameter distributions are independent (zero covariance) in the Monte Carlo simulation.
Independence simplifies implementation and is the default in most MC scenario tools.
Supporting evidence

CE internal implementation; standard practice in ensemble climate-economic modelling

Counterarguments

Technology abatement potential, policy effectiveness, and carbon budget are positively correlated — all are high under optimistic scenarios. Ignoring correlation underestimates tail risk in pessimistic scenarios.

Failure conditions

Underestimates P5 (pessimistic tail) by approximately 15–25% based on CE sensitivity test

Alternative assumptions (1)
  • Correlated sampling using Cholesky decomposition; pairwise correlations from IPCC scenario ensemble
Affected models: MonteCarloEngine Scenarios: uncertainty_analysis Owner: Quantitative Methods Team Reviewed: 2026-01-10
A-UNC-002 Uncertainty MEDIUM known_limitation
Normal and triangular distributions adequately represent parameter uncertainty.
These distributions are analytically tractable and well-understood. They match the symmetric-unimodal character of most climate parameter posteriors.
Supporting evidence

IPCC AR6 uncertainty characterisation guidance; CE parameter calibration

Counterarguments

Many climate-economic parameters have heavy right tails (damage function, carbon budget) that are poorly captured by normal distributions. Damage costs may follow power-law tails.

Failure conditions

Underestimates catastrophic tail risk; power-law tails for damage would materially shift P95

Alternative assumptions (2)
  • Log-normal for damage parameters
  • Pareto tails for catastrophic risk parameters
Affected models: MonteCarloEngine Scenarios: uncertainty_analysis Owner: Quantitative Methods Team Reviewed: 2026-01-10
A-GOV-001 Governance CRITICAL contested
NDC commitments stated by governments will be implemented with the effectiveness rates assumed in NGFS Phase IV scenarios.
NGFS scenarios are the institutional standard for financial sector climate risk assessment.
Supporting evidence

NGFS Phase IV Technical Documentation (2023); Climate Action Tracker NDC assessment

Counterarguments

Historical NDC implementation rates are significantly below stated commitments. Climate Action Tracker currently rates 75% of NDCs as 'insufficient'. Policy reversal risk is not quantified.

Failure conditions

If policy implementation remains at current rates, 2°C pathway becomes infeasible with existing policies by 2028 (Climate Action Tracker)

Alternative assumptions (1)
  • Realistic implementation scenario with 60–70% NDC delivery rate
Affected models: ClimateModelService, EconomicModelService, FiscalModelService Scenarios: policy_optimistic, orderly_transition Owner: Policy Team Reviewed: 2026-02-10