CORDEX EUR-11 vs. CMIP6: Climate Data Choice for Risk Analysis

When companies report their physical climate risks under CSRD, the EU Taxonomy or IFRS S2, the methodological choice comes down to a data question: which climate projections carry the analysis? CMIP6 as the global model base with the current SSP framework and a 100 to 200 km grid, or CORDEX EUR-11 as regional downscaling at 12.5 km on an RCP forcing. Each has a clearly bounded strength and a clearly documented limitation. Anyone who does not understand the difference cannot build an audit-proof risk analysis. This authority article maps architecture, scenarios, bias correction, the regulatory frame and offers a decision tree for German and European industrial sites.

Architecture: WCRP, CMIP6 and CORDEX

Both worlds sit under the World Climate Research Programme (WCRP). CMIP6 (Coupled Model Intercomparison Project Phase 6) is the globally coordinated framework for coupled climate models, providing the basis for IPCC AR6 and comprising simulations from more than 100 models in 30 countries. CMIP6 global climate models (GCMs) typically run at 100 to 200 km resolution, the HighResMIP subset at 25 km. That captures large-scale dynamics reliably, while regional differentiation (low mountain ranges, coastal gradients) remains coarse.

CORDEX (Coordinated Regional Climate Downscaling Experiment) was also launched in 2009 under WCRP. The logic: regional climate models (RCMs) are „driven" by GCMs, with the GCM output forming lateral boundary conditions while the RCM simulates regional processes on a finer grid. Dynamical downscaling is the standard method.

EURO-CORDEX is the European arm, producing ensembles on a 0.11° grid (around 12.5 km resolution), short EUR-11. Partners include the German Climate Computing Centre (DKRZ), the Climate Service Center Germany (GERICS) at Helmholtz-Zentrum Hereon, and numerous European research institutions. Forcing is currently primarily CMIP5; CORDEX-CMIP6 is in build-out.

Spatial resolution and core variables

A CMIP6 grid box can span an area that runs from the North Sea coast to Hanover. The difference between an urban heat island, marshland and heath landscape disappears within it. CORDEX EUR-11 resolves mid-mountain effects, coastal gradients and regional circulation considerably better.

Core variables from EUR-11 output for industrial practice:

• tasmax (daily maximum temperature): hot days above 30 °C, tropical nights, cooling demand projection, worker protection
• pr (daily precipitation) and R99p (very heavy precipitation): flood analyses
• CDD (Consecutive Dry Days): drought periods, water supply, fire risk, agricultural supply chains
• sfcWind / wsgsmax (wind speed, max gust): storm risk
• prsn, snw (snow load): roof structures, alpine sites

Important limitation: hail cannot be projected directly at 12.5 km resolution, because convective cells form on a 1 to 20 km scale within minutes. Hail requires a different methodological route, as described in the hail frequency article via ERA5 reanalysis and AR-CHaMo.

RCP vs. SSP: no 1:1 mapping

CORDEX EUR-11 currently runs on a CMIP5 forcing and uses the Representative Concentration Pathways (RCP2.6, RCP4.5, RCP8.5). RCP2.6 corresponds to a strong climate protection pathway (2.6 W/m² additional forcing), RCP8.5 to a „business-as-usual" path. The scenarios are well validated, data-rich and established for years.

CMIP6 uses the Shared Socioeconomic Pathways (SSPs): SSP1-1.9, SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5. SSPs combine the radiative forcing with a socio-economic narrative (population, economy, governance). The strength for corporate assessment: not only physical impacts but also transition risks are addressed.

The rule of thumb RCP8.5 ≈ SSP5-8.5, RCP4.5 ≈ SSP2-4.5, RCP2.6 ≈ SSP1-2.6 is a approximation, not an exact mapping. Both generations agree roughly in forcing, but CMIP6 models show a higher equilibrium climate sensitivity (ECS) on average than CMIP5. For SSP5-8.5 the temperature projections trend higher than the nominally „equivalent" RCP8.5 in EURO-CORDEX. EFRAG and ISSB publish no official mapping table. Instead they recommend covering uncertainty with an ensemble of models and scenarios rather than relying on a single translation. The RCP and SSP complete guide describes the steps for an audit-proof choice.

Limitations of CORDEX EUR-11

No complete SSP set

The existing EURO-CORDEX EUR-11 ensemble (CMIP5 driven) does not cover a complete SSP scenario set. Data ends in the RCP framework. CORDEX-CMIP6 is in build-out; the WCRP published the updated experiment design in April 2025, individual simulations are ongoing. Anyone using CORDEX EUR-11 in 2026 for SSP-based reporting must document the RCP-SSP bridge methodologically.

Bias correction is methodologically required

All regional climate models show systematic errors against observations. In EURO-CORDEX, temperature biases of up to more than 1.5 °C and precipitation biases of up to ±40 per cent have been documented. Bias correction is therefore not optional. Three standard methods:

• Quantile Mapping (QM): Aligns the statistical distribution of model output to the observed distribution. In EURO-CORDEX, six QM methods were applied with three observation datasets.
• ISIMIP3BASD: Parametric QM developed at the Potsdam Institute, trend-preserving, that is, it retains the climate change signal. GERICS has published 8 EUR-11 RCMs bias-corrected to the E-OBS reference openly.
• CORDEX-Adjust: Bias-adjusted EUR-11 and EUR-44 simulations on the ESGF under its own project name.

Boundary problem and the model cascade

RCMs receive their boundary conditions from the driving GCM. Biases in the GCM are partially „transferred" into the RCM output. Convective extreme events such as hail, local thunderstorms and supercells remain parameterised rather than explicitly modelled at 12.5 km resolution. The representation of these hazards stays an essential limitation of the EUR-11 ensemble.

EFRAG, TCFD, ISSB and EU Taxonomy as application frame

ESRS E1-9 is the central regulatory hook: „anticipated financial effects from material physical and transition risks". Companies must disclose monetary effects of material physical climate risks on assets and revenues, before and after adaptation measures. The Bundesbank analysis is clear: ESRS E1 requires disclosure of whether and how scenario analyses were used and lists publicly available scenarios as options. EFRAG published revised exposure drafts in July 2025 under the Omnibus Simplification Package with up to 57 per cent fewer data points. The scenario analysis obligation remains: at least two contrasting scenarios, a low-emission path (RCP2.6 / SSP1-2.6) and a high-emission path (RCP8.5 / SSP5-8.5).

TCFD and IFRS S2 are method-agnostic in the choice of specific climate models. IFRS S2, applicable since January 2024, explicitly requires a stress test of business strategy against various climate pathways including a 1.5 °C scenario. The documentation depth is decisive: auditors expect traceable data provenance, described methodology, justified scenario choice. CORDEX EUR-11 is well suited because the underlying publications are peer-reviewed and broadly accepted in the scientific community.

The EU Taxonomy requires a Climate Risk and Vulnerability Assessment (CRVA) under the DNSH criterion (Do No Significant Harm). The Austrian guidance KlimTAX gives methodological recommendations. CORDEX EUR-11 as a regional product fits particularly well, because DNSH requires site-specific statements. For an ISO 14091-compliant climate risk analysis in practice, the data question is a separate methodological module that must be documented audit-proof.

Decision tree: CORDEX, CMIP6 or ISIMIP

CORDEX EUR-11 is clearly preferable for physical climate risk assessment of specific sites in Germany or Europe, for example heat stress at production plants, flood risk at logistics centres or winter sports infrastructure in Alpine regions. GERICS has published climate outlooks for all 401 German districts based on 85 EURO-CORDEX simulations, enabling site-specific orientation.

CMIP6 or statistical downscaling are the better option for global supply chain analysis, the full SSP scenario coverage and transition phases where CORDEX-CMIP6 is not yet complete. ISIMIP3b delivers bias-adjusted CMIP6 data for five global models (GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2-HR, MRI-ESM2-0, UKESM1-0-LL) on a 0.5° grid with eleven variables and SSP1-2.6, SSP3-7.0 and SSP5-8.5. The ISIMIP methodology (ISIMIP3BASD) counts as the methodological gold standard for trend-preserving bias correction.

Data portals: DKRZ, DWD, Copernicus, GERICS, ISIMIP

Five access points matter for German and European site assessment:

• DKRZ ESGF Node (esgf-data.dkrz.de): the German node of the global Earth System Grid Federation, the primary distribution channel for raw EURO-CORDEX EUR-11 simulations (NetCDF) and CORDEX-Adjust data. Connected to the WDCC for FAIR-compliant long-term archival with DOIs.
• DWD ESGF Node and climate data portal (esgf-data.dwd.de): DWD's view on regionalised climate data for Germany, with HYRAS observation data as the reference for bias adjustment.
• Copernicus Climate Data Store (CDS) (climate.copernicus.eu): user-friendly web formats for CORDEX data. CORDEX4CDS cooperates with ECMWF on ensemble preparation. Also contains the IPCC Atlas dataset (CMIP5, CMIP6, CORDEX harmonised) for screening analyses, 22 climate indices.
• GERICS / Helmholtz Climate Service Center: prepared climate outlooks for all 401 German districts based on 85 EURO-CORDEX simulations. Bias-adjusted EUR-11 dataset using ISIMIP3BASD, openly licensed via the WDCC under Creative Commons Zero.
• ISIMIP repository (data.isimip.org): bias-adjusted CMIP6 data at 0.5° resolution, harmonised across sectors (hydrology, energy, agriculture). Methodological gold standard for impact studies.

Bias correction as a methodological requirement

Bias correction is not an academic luxury but a prerequisite for valid statements. Three practical points:

Quantile Mapping (QM) is the industry standard. The model variable distribution is matched quantile by quantile to the observation reference (for example E-OBS for Europe). Weakness: plain QM can „wash out" the climate change trend unless implemented in a trend-preserving way.

ISIMIP3BASD solves this with parametric QM that explicitly preserves the trend for each quantile. Well documented, peer-reviewed, used at several European institutions. For EUR-11 data over Germany, the GERICS dataset with ISIMIP3BASD bias correction is currently the most solid freely available starting point.

Methodological uncertainty from bias correction: method choice, reference dataset and calibration period affect results. In EURO-CORDEX, six bias adjustment methods were combined with three observation datasets; bias-adjusted data is available on the EUR-11 grid for 24 RCM-GCM combinations, mainly for RCP4.5. Daily precipitation and daily temperature extremes are primarily covered.

CORDEX-CMIP6 is in build-out

The CORDEX-CMIP6 project closes the gap between SSPs and regional resolution. WCRP published the updated experiment design in April 2025; ERA5 evaluation runs for the European domain (EUR-12) are ongoing. The Copernicus Climate Change Service (C3S) has tendered a 2.73 million euro contract for machine-learning-based downscaling of CMIP6 projections to fill gaps in CORDEX-CMIP6. For practice in 2026/27 this means: CORDEX EUR-11 (CMIP5 forcing) remains the practice-ready reference database for high-resolution European climate projections. CORDEX-CMIP6 will improve the methodological base medium-term.

Frequently asked questions

Further resources

• Climate risk analysis for companies: Methodology to ISO 14091, CORDEX-EUR-11 base
• ISO 14091 in practice: What banks and insurers expect
• RCP and SSP climate scenarios: A complete guide
• Hail frequency Germany: Why hail takes a different methodological route
• Physical vs. transition climate risks
• Integrating climate risks into financial planning