How we work
We build our own models, rebuild published ones and couple models with partner institutions. Each is made so that someone else can rerun it and prove it wrong.
Every programme follows the same six principles, and draws on the same set of capabilities. Neither is a statement of virtue: each one is a step you can see in a project.
Three ways we model
| Way | What we do | Example of current work |
|---|---|---|
| Build | We design and build our own models where no existing model answers the question: from forest growth and soil carbon to agent-based and macroeconomic models. | Forest carbon cost of wood products (model chain built end to end) · Why efficient systems become fragile · The price of slack (planned) |
| Rebuild and verify | We rebuild published models as software that can be checked, reproduce their results, correct them, and only then extend them. | The full climate cost of food |
| Couple | We link models, ours and those of partner institutions, into integrated models that no single team could run alone. | Land, biomass and ocean in Europe · The full climate cost of food (planned links to equilibrium and spatial land-use models) |
Whichever way a model is made, the six principles below apply to it.
Six principles
| Principle | What it means | Where to see it |
|---|---|---|
| Reproduce before improving | Before extending a published model, we rebuild it and check that it gives the same results. | The full climate cost of food |
| Write predictions first | Hypotheses and expected results are recorded before the model is run. | Why efficient systems become fragile |
| Carry uncertainty through | Results come with their range. Structural choices are shown side by side, not hidden in an error bar. | Forest carbon cost of wood products |
| Publish failures | Predictions that failed are reported with those that held. | Why efficient systems become fragile |
| Open code and data | Code and data are released under open licences where contracts and data rights allow. | The price of slack |
| Asymmetric scrutiny | An assumption that makes the answer convenient gets more checks, not fewer. | Forest carbon cost of wood products |
Capabilities
What we bring to a new question, whatever the programme. Each capability links to a project that shows it; the projects are examples, not the limit of what we do.
Modelling systems
- Biophysical modelling of land, forests, soils, food systems and carbonSHOWN IN · The full climate cost of food · Forest carbon cost of wood products
- Complex-systems and agent-based modellingSHOWN IN · Why efficient systems become fragile
- Macroeconomic modelling: stock-flow consistent, input-output, production networksSHOWN IN · The price of slack
- Integrated modelling and model couplingSHOWN IN · Land, biomass and ocean in Europe
- Geospatial modelling and data engineeringSHOWN IN · Forest carbon cost of wood products
Measuring and accounting
- Life-cycle assessment, attributional and consequentialSHOWN IN · Forest carbon cost of wood products · The accounting gap
- Carbon opportunity cost of land and biomassSHOWN IN · The full climate cost of food · Forest carbon cost of wood products
- Material-flow analysis and allocation to productsSHOWN IN · Forest carbon cost of wood products
- Uncertainty analysis: Monte Carlo simulation, and defensible methods compared side by sideSHOWN IN · Forest carbon cost of wood products · The full climate cost of food
- Checking environmental claims and standards for what they omitSHOWN IN · The accounting gap · 24 cement and concrete technologies assessed
Technology and sector foresight
- Technology assessment: chemistry, mineral resources, readinessSHOWN IN · 24 cement and concrete technologies assessed
- Systems archetypes, rebound and supply-limit analysisSHOWN IN · 24 cement and concrete technologies assessed
- Scenarios to 2050SHOWN IN · The full climate cost of food · Land, biomass and ocean in Europe
- Sector decarbonization roadmaps and industrial ecologySHOWN IN · Industrial ecology for construction · 24 cement and concrete technologies assessed
Research engineering
- Rebuilding and verifying published modelsSHOWN IN · The full climate cost of food
- Pre-registration and research governanceSHOWN IN · Why efficient systems become fragile
- Open, reproducible code and dataSHOWN IN · The price of slack
- Systematic literature review and evidence gradingSHOWN IN · Why efficient systems become fragile · The accounting gap
From research to decision
- Policy analysis: land use and forestry, bioeconomy, building carbon, public roadmapsSHOWN IN · Forest carbon cost of wood products · Industrial ecology for construction
- Decision tools for non-specialistsSHOWN IN · Land, biomass and ocean in Europe
- Translating scientific gaps into financial and regulatory riskSHOWN IN · The accounting gap
- Purchasing and investment strategy from technology foresightSHOWN IN · 24 cement and concrete technologies assessed
- Policy and public-affairs work grounded in peer-reviewed evidenceSHOWN IN · Industrial ecology for construction
- Research collaboration within multi-institution consortiaSHOWN IN · Land, biomass and ocean in Europe
Domain knowledge
- Cement and concrete chemistry; mineral resources; natural construction materialsSHOWN IN · Industrial transitions
- Forest science and forest-product economics; agri-food systemsSHOWN IN · Land and ecosystem transitions
- Ecological, post-Keynesian and thermodynamic economics; robustness and viability theorySHOWN IN · Systemic transitions
A worked example
We took a published global agri-food model and rebuilt it as software. Two independent implementations were tested against each other until they agreed, then the corrected version was frozen as a 2020 benchmark. Extension toward 2050 comes only after that step.
How we report status
Every project carries one status. A provisional result is never written as a finding.
- CompletedWork delivered.
- OngoingWork under way. No results are claimed yet.
- Provisional resultsResults exist but have not been reviewed. They are not quoted as findings.
- PlannedDesigned, not started.
- ConceptAt working-paper or proposal stage.
- Peer reviewedReviewed by independent experts before publication.
Open science
Code is released under open licences such as GPL or AGPL; data and methods are released where we can, alongside the outputs they support. Where a contract or a data licence prevents release, the project page says so. Client data is never released.
Independence
Advisory work does not change these standards.
- Every mission draws on a research programme.
- Client data and results stay with the client. Methods are published; a client's name never is.
- We assess claims. We do not certify or endorse them, and we accept no payment for endorsement.
- Any public output states its method and assumptions.
- We do not present a result we cannot explain.
- We argue for a material, sector or policy only on published, reviewable evidence.