GCRI · Science and methods
Models and scenarios
Models explore what could happen when the real experiment is impossible: a flood spreading through a city, a grid in a heatwave, a price shock moving through food systems. Scenarios set out plausible futures to test against. The Global Centre for Risk and Innovation (GCRI) states each model’s validation, assumptions and uncertainty, and never presents an output as an observation. Part of science and methods.
01Register
Models and scenarios on the record.
The Global Centre for Risk and Innovation
- 01InstitutionThe Global Centre for Risk and Innovation (GCRI)
- 02Four disciplinesWhat every GCRI model must show
- 03Scenario familiesThe scenario frameworks most risk analyses are read against
- 04Calibration and validationHow a model earns trust, and the documentation that shows it
- 05GCRI’s partDocumented models, published as scenario or modelled evidence
- 06CounterfactualsAsking what would have happened otherwise
- 07ScenariosPlausible futures, not predictions
02Four disciplines
What every GCRI model must show.
A model is only as useful as its user’s understanding of where it holds. These four elements are published with every model result.
Validation
Tested against reality
Compared with observations it was not built on, using uncertainty quantification and model validation.
Assumptions
Stated, not hidden
Every key assumption is written down, with how results change if it is wrong.
Uncertainty
Ranges, not single numbers
Results are reported with ranges and the sources of uncertainty that drive them.
03Scenario families
The scenario frameworks most risk analyses are read against.
| Framework | Body and date | What it provides | Caution |
|---|---|---|---|
| Shared Socioeconomic Pathways with CMIP6 forcing (ScenarioMIP) | Research community, O’Neill and others, 2016 | Population, economy and emissions storylines paired with climate model runs | Some high-emission pathways are now considered less plausible; state why one is chosen |
| IPCC AR6 scenario database | IPCC Working Group III, 2022 | Mitigation pathways classed by warming outcome | A pathway is not a plan or a forecast |
| NGFS climate scenarios, Phase V | Network for Greening the Financial System, 2024 | Transition and physical risk scenarios for financial supervision | Designed for finance; physical damage functions are coarse |
| National and sector storylines | Ministries, agencies and operators | Locally coherent stress cases | Must be checked for internal physical and economic consistency |
04Calibration and validation
How a model earns trust, and the documentation that shows it.
| Practice | Reference | What it guards against |
|---|---|---|
| Bayesian calibration with model discrepancy | Kennedy and O’Hagan, 2001 | Tuning parameters to hide a structural error |
| Global sensitivity analysis | Saltelli and others, 2020, among the five rules for models that serve society | Reporting precision the inputs cannot support |
| Out-of-sample and hindcast tests | Standard in hydrology and climate modelling | A model that fits the past but cannot predict |
| ODD protocol for agent-based models | Grimm and others, 2020 | Simulations that no one else can understand or rebuild |
| Model cards | Mitchell and others, 2019 | Using a model outside the conditions it was evaluated for |
05GCRI’s part
Documented models, published as scenario or modelled evidence.
GCRI builds and tests models with national teams and publishes them with their documentation, so another team can rerun or challenge them. Model results never stand in for observation and never become a decision. Reviewed scenario work is designed for the Scenario and Stress-Test series and the Foresight Outlook series; each model would carry a card under NS-12 of the proposed Nexus Standards. Modellers can contribute validation datasets, intercomparison runs and documented failure cases.
06Counterfactuals
Asking what would have happened otherwise.
Many risk and policy questions are counterfactual: what losses a flood defence prevented, or how a programme changed outcomes. GCRI answers them with causal inference and impact evaluation. It states the identifying assumptions. It says so when the data cannot support an answer.
07Scenarios
Plausible futures, not predictions.
- 01
Purpose
Scenarios test decisions against a range of plausible conditions, including severe ones.
- 02
Design
Built with scenario design and stress testing, with drivers and storylines stated.
- 03
Interpretation
Simulated loss or performance is not observed impact, and a scenario carries no probability unless one is estimated and shown.
- 04
Use
Scenario results inform options for competent decision-makers. Analysis is not the decision.
08Where models run
Systems, reports and related pages.
Scenarios, options and decision analysis run in NXS-DSS (decision cases and deliberation). Reviewed scenario work is published in the Scenario and Stress-Test series. Shared modelling methods are organised through the Systems Science and Modelling platform. See also risk science.
Model code, data and parameters are recorded so results can be reproduced, within the rights of each source.
09Sources
Where this page’s references lead.
The bodies and documents this page is read against. Status and editions change; each link goes to the issuing body.
- O’Neill and others, The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6, Geoscientific Model Development 9 (2016)
- IPCC Sixth Assessment Report, Synthesis Report (IPCC, 2023)
- NGFS climate scenarios for central banks and supervisors, Phase V (Network for Greening the Financial System, 2024)
- Kennedy and O’Hagan, Bayesian calibration of computer models, Journal of the Royal Statistical Society Series B 63 (2001)
- Saltelli and others, Five ways to ensure that models serve society: a manifesto, Nature 582 (2020)
- Grimm and others, The ODD protocol for describing agent-based and other simulation models: a second update, Journal of Artificial Societies and Social Simulation 23 (2020)
- Mitchell and others, Model cards for model reporting, Proceedings of FAT* 2019
- NS-12, Software, model and AI/ML artefacts, a proposed standard
10Questions
Questions and answers.
Why should anyone trust a model?
Trust should follow evidence: validation against independent data, clear assumptions, honest uncertainty and openness to challenge. GCRI publishes all four with each model result.
What is the difference between a forecast and a scenario?
A forecast estimates what is likely to happen over a period with a measurable skill. A scenario describes a plausible situation to test decisions against, without claiming it will happen.
Can I see the model code?
Where licences and security allow, code and parameters are released with the publication. Where they cannot be released, the method is described in enough detail to be assessed.
How are model errors corrected?
Through the same correction process as any publication. An error found after release leads to a new recorded version and, where needed, revised results.
Does GCRI use AI models?
Where machine learning is used, it is validated and reported like any other model, with its training data, performance on unseen cases and known failure modes.
ReportsContribute
Contribute evidence from this page.
Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
GCRI
- EmailGateway GW-GCRISend it from this pageEvidence, a correction, a case, a method, a review, a translation or profile input. It arrives tagged with this page, so the editor who owns it receives it.
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- ReportsGCRISee every way to contributeThe contribution types, how each is reviewed, what credit you receive, the licence and what is never accepted.
Never send personal data, sensitive infrastructure detail or anything you have no right to share. A contribution is reviewed before use; sending it confers no authorship or editorial rights, and sponsorship never buys a finding.
The Global Centre for Risk and Innovation
Bring a method, a dataset or a question.
Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.
Legal
Legal documents.
Maps, data and figures on this site are shown for context and general information. They are not advice, forecasts or warnings; the limits of the information are set out in the Disclaimers and Data Notices.