Paper: arXiv 2412.14182
Authors: Hendrik Weichel, Aleksandr Zinovev, Heikki Haario, Martin Simon
Abstract
We present a novel Bayesian framework for quantifying uncertainty in portfolio temperature alignment models, leveraging the X-Degree Compatibility (XDC) approach with the scientifically validated Finite Amplitude Impulse Response (FaIR) climate model. This framework significantly advances the widely adopted linear approaches that use the Transient Climate Response to Cumulative CO2 Emissions (TCRE). Developed in collaboration with right°, one of the pioneering companies in portfolio temperature alignment, our methodology addresses key sources of uncertainty, including parameter variability and input emission data across diverse decarbonization pathways. By employing adaptive Markov Chain Monte Carlo (MCMC) methods, we provide robust parametric uncertainty quantification for the FaIR model. To enhance computational efficiency, we integrate a deep learning-based emulator, enabling near real-time simulations. Through practical examples, we demonstrate how this framework improves climate risk management and decision-making in portfolio construction by treating uncertainty as a critical feature rather than a constraint. Moreover, our approach identifies the primary sources of uncertainty, offering valuable insights for future research.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 6.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced statistical methods (Bayesian framework, adaptive MCMC) and complex climate modeling (FaIR model with deep learning emulation), indicating high mathematical density. It includes practical examples, collaboration with an industry partner (right°), and mentions GitHub code for reproducibility, demonstrating significant empirical rigor and backtest-ready implementation.
Research Flowchart
flowchart TD
A["Research Goal: Quantify Uncertainty<br>in Portfolio Temperature Alignment"] --> B["Key Methodology: Bayesian Framework<br>with FaIR Climate Model & XDC"]
B --> C{"Input Data"}
C --> C1["Portfolio Emission Data"]
C --> C2["Decarbonization Pathways"]
C --> C3["Climate Parameter Priors"]
B --> D["Computational Process"]
subgraph D ["Adaptive MCMC & Emulation"]
D1["Parametric Uncertainty<br>via MCMC"]
D2["Deep Learning Emulator<br>for Speed"]
end
C --> D1
D1 --> D2
D --> E["Key Findings & Outcomes"]
E --> E1["Robust Uncertainty<br>Quantification"]
E --> E2["Identification of<br>Uncertainty Sources"]
E --> E3["Improved Climate Risk<br>Management in Portfolios"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a novel Bayesian framework for quantifying uncertainty in portfolio temperature alignment models, leveraging the X‑Degree Compatibility approach with the FaIR climate model and a deep‑learning emulator for fast simulations.
Source sentence
We present a novel Bayesian framework for quantifying uncertainty in portfolio temperature alignment models, leveraging the X-Degree Compatibility (XDC) approach with the scientifically validated Finite Amplitude Impulse Response (FaIR) climate model.
Assumes 6
- The FaIR simple climate model is used instead of a linear TCRE approach, assuming its structure is adequate for portfolio temperature alignment. model
Source sentence
The climate model In this work, we use the FaIR simple climate model rather than a linear TCRE approach which suffers from a number of limitations: Some studies find the TCRE is not fully path-independent as it varies slightly with the rate of CO2 emissions, see, e.g., [28].
- The mapping F from emissions to temperature does not account for measurement noise or modelling bias, which are added later as an additive noise vector ε. statistical
Source sentence
F does neither take into account measure-ment noise nor modelling bias. Therefore, we model these by the additive noise vector ε.
- Parameter calibration is performed on present‑day GHG emission data, assuming these data are representative for the benchmark calculations. data
Source sentence
For now, we limit ourselves to benchmark calculations based on counterparty-specific present-day GHG emission data, leaving the integration of both, future counterparty-specific pathways and the corresponding uncertainty for future research. 2.1.2.
- Model structural uncertainty is acknowledged, implying that different simple climate models may produce different results due to varying assumptions and parameter choices. model
Source sentence
This inevitable structural uncertainty arises from the simplifications and assumptions inherent in the models’ design. Different simple climate models may produce different results due to variations in their underlying assumptions, parameter choices, and structural designs.
Show 2 more
- Adaptive MCMC (DRAM) can adequately explore the 20‑dimensional posterior despite the high dimensionality. statistical
Source sentence
In this work, we use Delayed Adaptive Metropolis (DRAM) [36]. Since this is practically infeasible, we rely on adaptive MCMC sampling methods to explore the posterior distri-bution.
- The deep‑learning emulator approximates the FaIR model outputs with sufficient accuracy for near real‑time use. model
Source sentence
To enhance computational efficiency, we integrate a deep learning-based emulator, enabling near real-time simulations. As a result, the run time is reduced from 12 minutes to 0.06 seconds. 5.
Data
- Counterparty‑specific present‑day GHG emission data macro unknown
Source sentence
For now, we limit ourselves to benchmark calculations based on counterparty-specific present-day GHG emission data, leaving the integration of both, future counterparty-specific pathways and the corresponding uncertainty for future research. 2.1.2.
- SSAB 2022 Scope 1‑3 emissions macro public
Source sentence
Scope 1, 2 and 3 emissions for the year 2022 were obtained from SSAB’s 2023 annual report [42].
- Gross value added (GVA) for SSAB fundamentals public
Source sentence
Gross value added was estimated using the income approach, which involves summing the following data points extracted from the annual report: operating profit, employee compensation, depreciation and amortization.
- Synthetic emission uncertainty distributions (lognormal and normal) synthetic unknown
Source sentence
Blue lognormal distribution, Green normal distribution. Specifically, we sample from two distributions: the emission uncertainty distribution described in Test Case 2 and the posterior distributions of climate model parameters obtained from Test Case 1.
- Deep‑learning emulator training data (FaIR model runs) synthetic unknown
Source sentence
The emulator takes emissions in the base year as input and calculates the projected mean earth temperature and 90% credible interval under different scenarios (SSP1-RCP1.9, 11 SSP1-RCP2.6, SSP2-RCP4.5, SSP3-RCP7.0, and SSP5-RCP8.5). For this task, we use a standard feed-forward neural network with three dense layers of 20 neurons.
Costs
Not applicable
Finds 5
- The Bayesian methodology significantly outperforms traditional sensitivity‑based Monte Carlo simulations in terms of estimation uncertainty. performance
Source sentence
The results highlight that this methodology significantly outperforms traditional sensitivity-based Monte Carlo simulations in terms of estimation uncertainty.
- Integrating emission uncertainty quantification improves the overall analysis by capturing a wider range of uncertainty sources. statistical
Source sentence
In the second test case, we integrate emission uncertainty quantification as outlined in Subsection 4.2, showcasing the importance of accounting for uncertainties in reported emission data. This demonstrates how their integration improves the overall anal-ysis by capturing a wider range of uncertainty sources.
- The deep‑learning emulator reduces run time from 12 minutes to 0.06 seconds, enabling near real‑time simulations. performance
Source sentence
As a result, the run time is reduced from 12 minutes to 0.06 seconds. 5.
- MCMC simulations applied to FaIR’s parameter uncertainty markedly narrow the confidence intervals of temperature forecasts. statistical
Source sentence
Carlo (MCMC) simulations applied to FaIR’s parameter uncertainty significantly narrow the confidence intervals of temperature forecasts. This improvement underscores the value of Bayesian methods in enhancing forecast reliability and reducing uncertainty in long-term climate modeling.
- Applying the framework to SSAB shows a temperature improvement from 2.56 °C to 2.50 °C in the SSP2‑RCP4.5 scenario when adopting green steel production. performance
Source sentence
In the SSP2-RCP4.5 scenario, this transition results in a temperature improvement from 2.56°C to 2.50°C, demonstrating the tangible benefits of industry-wide emission reduc-tions.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 16 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (58)
- Uncertainties in the Emissions Database for Global Atmospheric Research (EDGAR) emission inventory of greenhouse gases (2021) — link
- A New Scenario Framework for Climate Change Research: The Concept of Shared Socioeconomic Pathways (2014) — link
- Net Zero by 2050 (2021)
- Climate change 2021: The physical science basis (2021)
- Climate Change 2014 Synthesis Report — link
- Good practices for climate-related and environmental risk management (2022) — link
- Walking the talk - Banks gearing up to manage risks from climate change – results of the 2022 thematic review on climate-related and environmental risks (2022) — link
- European Banks’ corporate loan books - a pilot study on banks’ alignment with the temperature target of the P aris agreement (2024) — link
and 50 more
- Measuring Portfolio alignment (2020) — link
- Comparing and weighting imperfect models using D-probabilities (2020)
- Provision of climate services - the XDC model (2020)
- Temperature-driven biodiversity change: disentangling space and time (2018)
- Climate change 2013: the physical science basis: Working Group I contribution to the Fifth assessment report of the Intergovernmental Panel on Climate Change (2014)
- Climate change and the global redistribution of biodiversity: substantial variation in empirical support for expected range shifts (2023)
- Changes in temperature alter the relationship between biodiversity and ecosystem functioning (2018)
- FaIR v1.3: a simple emissions-based impulse response and carbon cycle model (2018)
- Modeling uncertainty in integrated assessment of climate change: A multimodel comparison (2018)
- Estimating the likelihood of GHG concentration scenarios from probabilistic Integrated Assessment Model simulations (2022)
- Corporate debt value under transition scenario uncertainty (2024)
- Green investment and asset stranding under transition scenario uncertainty (2023)
- Energy transition under scenario uncertainty: a mean-field game of stopping with common noise (2024)
- Sustainable investing with ESG rating uncertainty (2022)
- Unlocking ESG premium from options (2023)
- NGFS Climate Scenarios Database: Technical Documentation (2020)
- The Alignment Cookbook 2 - A technical panorama of the alignment methodologies and metrics used by and applied to the financial sector, with a view to inform consolidated alignment assessments (2024)
- The Alignment Cookbook - A Technical Review of Methodologies Assessing a Portfolio’s Alignment with Low-carbon Trajectories or Temperature Goal (2020)
- EU P aris-Aligned and Climate Transition-Benchmarks: A Case Study (2021)
- A new scenario framework for climate change research: the concept of shared socioeconomic pathways (2014)
- The roads ahead: Narratives for shared socioeconomic pathways describing world futures in the 21st century (2017)
- A modified impulse-response representation of the global near-surface air temperature and atmospheric concentration response to carbon dioxide emissions (2017)
- FaIR v2.0.0: a generalized impulse response model for climate uncertainty and future scenario exploration (2021)
- FaIR -calibrate v1.4.1: Calibration, constraining and validation of the FaIR simple climate model for reliable future climate projections (2024)
- DRAM : efficient adaptive MCMC (2006)
- An adaptive M etropolis algorithm (2001)
- On M etropolis- H astings algorithms with delayed rejection (2001)
- Transforming the future of steel Annual Report 2023 (2023)
- Fact Book 2022 Smarter steels for people and planet (2023)
- CORPORATE RESPONSIBILITY REPORT 2022/23 (2023)
- TCFD Reference Table Task Force on Climate-related Financial Disclosures Fiscal Year 2022/2023 (2023)
- Environmental and Energy Management
- Fossil freedom is just around the corner!
- Offizieller DAC-Umrechnungskurs
- USD/SEK - US Dollar Swedish Krona
- Uncertainty quantification of FaIR climate model (2021) — link
- Do investors care about carbon risk? (2021)
- Model uncertainty and liquidity (2009)
- Sustainable investing in equilibrium (2021)
- Credit default swaps and corporate innovation (2019)
- Market expectations of a warming climate (2021)
- Implied temperature rise of equity portfolios: A sensitivity analysis framework (2024)
- Portfolio alignment to a 2° C trajectory: Science or art? (2020)
- Optimal business model adaptation plan for a company under a transition scenario (2024)
- Bridging socioeconomic pathways of $\text CO _2$ emission and credit risk (2024)
- An Efficient SSP -based Methodology for Assessing Climate Risks of a Large Credit Portfolio (2024)
- Trajectory sensitivity of the transient climate response to cumulative carbon emissions (2014)
- Cumulative carbon emissions and economic policy: in search of general principles (2019)
- The shared socio-economic pathway ( SSP ) greenhouse gas concentrations and their extensions to 2500 (2020)
- Portfolio construction with climate risk measures (2022)
0 of 58 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.