The Global Centre for Risk and Innovation (GCRI)
No single model captures a complex system fully. Model intercomparison runs several models on the same experiment with the same inputs, so that their agreement and disagreement can be seen and explained. It is how climate, crop, hydrology and epidemic modelling communities learned what their results can and cannot support. This page sets out how intercomparisons are designed, what they reveal, and what evidence serves the institutions that rely on model ensembles.
GCRI · In development · service level R0
Model intercomparison is the coordinated running of several models on the same experiments with shared inputs, so that agreement and disagreement can be seen and explained. Common protocols isolate structural differences, spread measures structural uncertainty, and evaluation against observations shows which models capture key processes.
For model intercomparison, in GCRI's spatial-intelligence reading, the map opens on true-colour imagery, flood water, rain and population density with relief in 3D and true colour from NASA GIBS, framed on North America; every other NASA layer, live feed and forecast waits in the layer list.
The interactive map needs JavaScript; the layer list and sources are below.
Layers 4
Hazards and exposure, for this page
Keyboard, with the map focused
Hold Ctrl or ⌘ and scroll to zoom
Latest available
NASA GIBS / EOSDIS · ASTER GDEM (METI, NASA) · Blue Marble (NASA) · © OpenStreetMap contributors (ODbL) · Natural Earth
Loading the latest NASA imagery.
Drag to move. Ctrl or ⌘ and scroll, pinch or the buttons to zoom; hover for the values under the pointer; reset returns to the region.
For model intercomparison, in GCRI's spatial-intelligence reading, the map opens on true-colour imagery, flood water, rain and population density with relief in 3D and true colour from NASA GIBS, framed on North America; every other NASA layer, live feed and forecast waits in the layer list.Satellite and model products as NASA publishes them, each with its own date; indicators, not assessments of any place; claim ceiling INTEGRATED.
Layers and dates
The working screen behind this subject: its structure, terms and states. Names are placeholders and values illustrative.
Model intercomparison · Coupled model A · Scenario set 3
nodes: models · edges: exchanged variables
| Hydrology | 1 day | runoff | Yes |
| Power system | 1 h | demand, outages | Yes |
| Economy | 1 quarter | output, prices | Partial |
| Population | 1 year | exposure | Yes |
Comparing models of the same system, and what disagreement tells you. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
When models share inputs, scenarios and output formats, remaining differences point to model structure and parameters rather than set-up.
The range of results across credible models is a measure of how much the answer depends on modelling choices.
Models share code, data and ideas, so an ensemble can agree for shared reasons and overstate confidence.
Comparing models with observations shows which reproduce key processes, which can inform how ensemble results are interpreted.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| Ensemble | A set of model runs from different models or settings used to estimate a range of outcomes. |
| Protocol | The agreed experiment design, inputs and output formats in an intercomparison. |
| Model genealogy | The shared ancestry of models through common code and components. |
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Models participating in an intercomparison | count | per round | Intercomparison records | REPORTED |
| Ensemble spread | range or standard deviation across models | per experiment | Intercomparison outputs | MODELLED |
| Agreement on sign of change | percent of models | per experiment | Intercomparison outputs | MODELLED |
| Skill against observations | error metric by model | per evaluation | Model evaluation | ESTIMATED |
| Shared code lineage between models | count of related model families | per assessment | Model genealogy studies | ESTIMATED |
| Outputs published in standard formats | percent of outputs | per round | Archive metadata | OBSERVED |
The topic as a computational graph: named sources feed the methods, an evidence review gates the outputs, and the decisions stay with the authorities that hold them.
Fig. 02Nodes carry their type, two metric rows and a state; connectors show what feeds what; dashed frames name the owners; the timeline is the cadence. Design of the pipeline, not a record of a run.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| CMIP6 archive via the Earth System Grid Federation | Climate model outputs from common experiments | Open, CC BY 4.0 for most models (to check per model) |
| ISIMIP repository | Impact model outputs for water, agriculture, health and other sectors | Open, licence per dataset (to check) |
| AgMIP crop model intercomparisons | Crop model ensemble protocols and results | Public, terms per project |
| Scenario Modeling Hub publications for epidemics | Multi-model epidemic scenario projections | Public, terms per hub |
| Literature on model genealogy and ensemble independence | Analysis of shared model components | Licence per publication |
Systems Science and Modelling prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Modelling groups | Which experiments and protocols to adopt | Protocol designs; evaluation reports |
| Institutions using model results | How much confidence to place in ensemble findings | Ensemble summaries with spread |
| Research funders | Whether to fund intercomparison infrastructure | Evidence on value of intercomparison |
Every model embodies choices: which processes to include, how to represent them and how to calibrate them. Two well-built models can therefore produce different answers. When models agree, confidence increases, though not to certainty, since they may share the same flaw. When they disagree, the disagreement is information: it reveals structural uncertainty that no single model can show.
Interpreting an ensemble is not straightforward. Models are rarely independent; many share components, data or lineage. An ensemble's spread may underestimate true uncertainty if the models share errors, or overstate it if poor models are included. Methods for weighting models by performance and independence are an active research area.
Intercomparison also improves models. Diagnosing why one model differs often reveals errors or missing processes, and shared protocols raise standards across a community.
Shared inputs, scenarios, output variables and formats. Designing them takes time and agreement.
Means, spreads and distributions across models, using probability and statistics. Spread is not a probability unless justified.
Weighting models by skill and distinctness, using uncertainty quantification. Weights depend on chosen metrics.
Comparing internal variables to find sources of difference. Requires access to model internals.
Agreement among models is not proof of correctness. Intercomparison results state which models took part and how they relate.
The current service level is R0: information and enquiries. 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.