Methodology and Reproducibility Roadmap

Introduction & Scope

CE (Climate Economics) is an integrated assessment and risk modeling platform in beta. It combines physical climate variables, macroeconomic dynamics, sectoral emissions pathways, transition policy analysis, and financial risk quantification into a unified analytical framework targeting investor, government, and corporate strategy use cases. Several advanced modules are prototype-grade — see the maturity notice above and the capability matrix for full status.

CE is designed to be transparent, reproducible, and auditable. All model parameters are externally sourced (linked below), calibration choices are documented, and code is version-controlled. CE differs from IAMs (MESSAGEix, REMIND, GCAM) in scope: rather than modeling full global energy systems endogenously, CE provides risk synthesis calibrated against IAM outputs. Where CE is less mature than established IAMs, this is explicitly documented in the framework positioning and audit layer.

Modeling philosophy

• Calibrated, not invented: All parameters drawn from published peer-reviewed studies or authoritative institutional sources (IPCC, IEA, IMF, World Bank).
• Transparent uncertainty: Core outputs carry calibrated confidence intervals. Monte Carlo engine (prototype) runs N=10,000 draws; not yet independently validated. Sensitivity cases (low/base/high) are available for all scenarios.
• Institutional comparability: Results benchmarked against NGFS Phase IV, IEA WEO, UNEP EGR, and IPCC AR6 at each release.
• Separation of signal and noise: Structural baselines separated from shock contributions and policy contributions in all economic outputs.

Versioning & Audit Trail

Reproducibility commitments

• All scenario JSON files are version-tagged and archived at /data/scenarios/
• All adapter/calibration files (economic inputs, physical climate variables, shocks registry) are committed to version control with change log
• API endpoints return structured JSON with methodology_endpoint links for machine-readable methodology
• Monte Carlo seeds are parameterised (default: seed=42) ensuring exact reproducibility of probability outputs

Physical Climate Variables

Six physical hazard dimensions computed for each scenario: heat stress, flood risk, drought risk, wildfire risk, coastal risk, storm intensity. Each dimension is scored 0–1.

Computation methodology

• Base exposure: Industry-level vulnerability table calibrated to IPCC AR6 WG2 Chapter 11 (extreme weather) and sector-specific literature
• Pathway multiplier: Scales 0.6 (Aggressive Decarbonisation) to 1.4 (No Action) based on NGFS climate scenario family
• Shock overlays: Event-specific physical hazard deltas from shocks registry (calibrated to historical event magnitudes)
• Confidence: Decreases with number of active shocks; baseline 0.68, floor 0.30

Economic Model

Signal/noise decomposition separates structural baseline from shock and policy contributions. Four macro signals: GDP growth, inflation, investment intensity, labor market indicator.

Structural baseline

Calibrated to IMF World Economic Outlook (growth baselines), World Bank Global Economic Prospects, and IEA energy investment data.

Emissions Accounting

Bottom-up gap accounting matching UNEP EGR and IEA NZE methodology. Technology categories: renewable power, efficiency, electrification, fuel switching, methane, land use, CDR, industrial CCS.

Carbon budget accounting follows IPCC AR6 WG3 Table SPM.2 with CE adjustment to 2025 reference year using observed 2024–2025 emissions.

Damage Functions

Physical damage to economic output modeled using quadratic temperature-GDP relationship calibrated to Burke, Hsiang & Miguel (2015) pooled global regression:

Transition damage (stranded assets): Dietz & Stern (2015) high-damage specification; fossil asset write-down fractions from NGFS orderly/disorderly scenario families.

Energy-System Optimization

Wright's Law learning curves for 8 clean-energy technologies. LCOE(t) = LCOE(0) × (Q_t/Q_0)^(-b) where b = log₂(1–LR). Learning rates from Lafond et al. (2018).

Capacity mix uses simplified least-cost merit-order dispatch with capacity credit adjustments for variable renewables and a 25% firm-power floor. Not a full mixed-integer LP — designed for indicative trajectory analysis, not utility-grade planning.

IEA WEO 2023 STEPS/APS/NZE growth rates used for scenario comparison.

Monte Carlo Engine

N=2,000 stratified random samples over 8 parameter distributions. Output: annual emissions fan chart (p10/p25/p50/p75/p90), peak warming histogram, cumulative budget gap distribution, tornado sensitivity chart.

Distributions: Normal for baseline emissions and abatement potential; Triangular for policy effectiveness and methane abatement; Uniform for net-zero residual; Log-normal for CDR deployment (right-skewed). Carbon budget drawn from Normal with mean=250 GtCO2, std=60 GtCO2 per IPCC AR6.

DSGE / CGE Macroeconomic Model

New Keynesian three-equation system with climate extensions. IS curve, Phillips curve, Taylor rule. Climate enters as incremental physical damage shock (δDamage × 8) and carbon cost supply shock (P_CO2 × 0.00004). Stranded assets enter IS curve as annual write-down flow. Labor productivity loss modeled separately from output gap.

Calibration: σ=1.5 (Christiano et al. 2005), κ=0.13 (Calvo stickiness), φ_π=1.5, φ_x=0.5 (Taylor 1993). Illustrative model — full production use requires rational-expectations matrix solution and sector disaggregation.

Climate Pathway Calibration

CMIP6 multi-model mean anchor temperatures from IPCC AR6 WG1 Table SPM.1 / Figure SPM.8 for SSP1-1.9, SSP2-4.5, SSP3-7.0, SSP5-8.5. Linearly interpolated between anchor years. p10/p90 likely ranges based on CMIP6 ensemble spread (Hausfather et al. 2022 emergent constraint applied).

Carbon budget exhaustion uses TCRE = 0.45°C/1000 GtCO2 (IPCC AR6 central). Budget-exhaustion timeline solved analytically from quadratic integral of declining emissions trajectory.

Commodity & Land-Use Markets

Cobweb price dynamics with climate supply shocks. Yield loss calibrated to Schlenker & Roberts (2009) nonlinear temperature-yield relationship. Energy-food pass-through: 12% food price sensitivity to oil price (World Bank 2022 Commodity Markets Outlook). Biofuel demand pull modeled for corn, soybeans, and palm oil.

Land cover simulation: annual change rates from FAO Global Forest Resources Assessment 2020. Carbon emissions from deforestation calculated using IPCC Tier 1 carbon stock factors. Deforestation policy scenarios calibrated to REDD+ effectiveness literature.

Actuarial Catastrophe Model

Generalised Extreme Value (GEV) distribution fitted to 7 perils. Parameters (μ, σ, ξ) calibrated to Swiss Re sigma database (1970–2023) and Munich Re NatCatSERVICE industry loss data. Occurrence Exceedance Probability (OEP) curves at return periods 2–1,000 years.

Climate loading: multiplicative hazard frequency increase per °C additional warming above 1.2°C (2025 baseline). Rates: tropical cyclone +7%/°C, riverine flood +10%/°C, wildfire +14%/°C (most climate-sensitive). Loading rates from IPCC AR6 WG1 Chapter 11 and Swiss Re CatNet research.

TVaR-99 approximated as 1.4× loss-at-100yr-return-period. Portfolio scaling: industry aggregate / $3 trillion global insured market proxy.

Primary Data Sources

Calibration Standards

Parameter selection criteria

• Parameters drawn from meta-analyses or authoritative reviews where available (not single studies)
• Uncertainty ranges set to published p10–p90 or "likely range" bounds where available; CE expert judgment otherwise (documented)
• Industry-level adjustments validated against sector-specific literature (e.g., Caldecott et al. for stranded assets, Rentschler for energy poverty)
• Scenario families aligned to NGFS taxonomy to ensure institutional comparability

Cross-validation benchmarks

At each major release CE cross-validates against NGFS macro outputs, UNEP EGR emissions gap, IEA NZE technology deployment curves, and IPCC AR6 carbon budget milestones. Deviations beyond ±15% trigger parameter review.

Validation Records

References

1. IPCC (2021). Climate Change 2021: The Physical Science Basis. Contribution of WG1 to AR6. Cambridge University Press.
2. IPCC (2022). Climate Change 2022: Mitigation of Climate Change. Contribution of WG3 to AR6. Cambridge University Press.
3. Burke, M., Hsiang, S.M., & Miguel, E. (2015). Global non-linear effect of temperature on economic production. Nature, 527, 235–239.
4. Christiano, L.J., Eichenbaum, M., & Evans, C.L. (2005). Nominal rigidities and the dynamic effects of a shock to monetary policy. JPE, 113(1), 1–45.
5. Dietz, S., & Stern, N. (2015). Endogenous growth, convexity of damages and climate risk. Economic Journal, 125(583), 574–620.
6. Hausfather, Z., et al. (2022). Evaluating the performance of past climate model projections. GRL, 49(1).
7. IEA (2023). World Energy Outlook 2023. International Energy Agency, Paris.
8. IRENA (2023). Renewable Power Generation Costs in 2022. International Renewable Energy Agency, Abu Dhabi.
9. Lafond, F., et al. (2018). How well do experience curves predict technological progress? Energy Policy, 112, 55–65.
10. NGFS (2023). NGFS Climate Scenarios for central banks and supervisors, Phase IV. Network for Greening the Financial System.
11. Schlenker, W., & Roberts, M.J. (2009). Nonlinear temperature effects indicate severe damages to U.S. crop yields. PNAS, 106(37), 15594–15598.
12. Swiss Re Institute (2024). sigma 1/2024: Natural Catastrophes in 2023. Swiss Re Group.
13. Tokarska, K.B., et al. (2020). Past warming trend constrains future warming in CMIP6 models. Science Advances, 6(12), eaaz9549.
14. UNEP (2024). Emissions Gap Report 2024. UN Environment Programme, Nairobi.