Model Catalog / climate

ERA5-Calibrated Climate Baseline

Climate Current active

Observed-state anchored climate profile for calibration-heavy use cases.

Horizon 1940–2025 (historical); 2025–2030 near-term
Geography Global (~31 km reanalysis grid)
Resolution Sub-daily observed state; facility-level via point extraction
Projection years 2020, 2021, 2022, 2023, 2024, 2025, 2026, 2027, 2028, 2029, 2030
0.62
hazard
0.49
transition
0.54
resilience
0.73
confidence
Observed warming trend Precipitation variability Extreme event frequency Wind & solar resource Drought & flood indices Operational disruption correlation
Projected uncertainty vs. observed truth — why the near-term anchor matters

CMIP6 is indispensable for long-run physical risk under different emissions pathways — no other data source can project 2050–2100 climate change with comparable rigour. But for the 1–5 year horizon that drives most portfolio risk decisions today, CMIP6's model drift and internal variability make it less skilful than ERA5. ERA5-calibrated is the observational truth anchor: it contains no structural model uncertainty because it is grounded in 80+ years of actual measurements. Company operational disruption calibration, extreme event attribution, and near-term seasonal risk cannot be done from CMIP6 — they require ERA5.

CMIP6 Core Ensemble ERA5-Calibrated
Time horizon 2025–2100; strongest for 2035+ long-run projection 1940–2025 historical; 2025–2030 near-term extrapolation; not scenario-conditioned
Physical basis 40+ physics-based models; explicit structural uncertainty Observational reanalysis — no structural model uncertainty; reflects what has happened
Near-term skill Low near-term skill — model drift dominates on 1–5yr horizon High near-term skill — observational grounding produces more skilful 1–5yr physical risk estimates
Company validation Cannot be matched to individual company loss records ERA5 event records can be directly matched to corporate disruption data for empirical loss calibration
Attribution capability Attribution requires additional post-processing of ensemble counterfactuals World Weather Attribution methodology uses ERA5 as the factual climate record for event attribution
Use ERA5-calibrated for near-term operational risk calibration, company-level historical loss validation, and extreme event attribution. Use CMIP6 for long-run scenario-conditioned physical risk, regulatory TCFD/NGFS alignment, and uncertainty quantification across warming pathways. The combination — ERA5 for near-term grounding, CMIP6 for long-run trajectory, GFDL for process-credible specific hazards — is CE's physical risk data architecture.

Methodology

ERA5 is the European Centre for Medium-Range Weather Forecasts (ECMWF) global reanalysis product, covering 1940–present at hourly resolution and ~31km spatial resolution. It provides the observed-state anchor for near-term physical risk calibration — grounding CE's climate signals in measured physical reality rather than model projections. CE uses ERA5 trend data to calibrate the current climate state (1.2°C observed warming, precipitation variance trends, extreme heat frequency) and correlates it with actual company operational disruption records (BASF Rhine levels, DHL hub closures, Union Pacific heat advisories) to validate the causal chain from climate variable to economic impact.

Key Mechanisms

  1. Observational anchor: ERA5 reanalysis provides the single best estimate of historical climate state, verified against 80+ years of station data
  2. Trend detection: multi-decadal ERA5 trends in temperature, precipitation, and extreme events are statistically validated before entering the model
  3. Operational disruption correlation: ERA5 weather event data is matched to company operational records to calibrate impact thresholds
  4. Near-term projection: ERA5 initialised short-range climate projections (1-5 year) are more skilful than long-run CMIP6 projections for this horizon
  5. Physical-to-financial bridge: ERA5 observed loss events are matched to insured loss records (Munich Re, Swiss Re) to calibrate the economic cost per unit of physical stress
  6. Extreme event attribution: ERA5 weather events are matched to climate change attribution studies (World Weather Attribution) to partition observed risk between natural variability and climate-change-forced components — supporting regulatory disclosure and litigation risk assessment
  7. Operational impact calibration: ERA5 weather records are matched to corporate operational disruption records to estimate production loss per unit of climate stress — the only physically grounded company-level loss calibration available
  8. Seasonal forecasting bridge: ERA5 initial conditions are used to initialise operational seasonal forecasting models (ECMWF SEAS5, NOAA CFS) providing the 1–3 month physical risk outlook layer

Score & Confidence Methodology

Hazard scores (0–1) are calibrated against IPCC AR6 WG2 Table 16.SM.1 industry-sector exposure bands. Transition pressure scores use NGFS Phase IV scenario families. Confidence intervals are asymmetric where IPCC likelihood language (likely/very likely) maps to the p17–p83 and p5–p95 ranges. Scores are not actuarially certified — see Known Limitations.

Known Failure Modes

  • Backward-looking by design: does not capture the structural shift in risk from future emissions trajectories
  • Understates fat-tail physical risks for the 2040-2100 horizon relative to CMIP6 or GFDL projections
  • Reanalysis uncertainty is low for temperature but higher for precipitation extremes and tropical cyclone intensity — key insurance sector drivers
  • Stationary variance assumption: historical ERA5 statistics assume climate variability is approximately stationary — this understates the changing frequency of extremes as global mean temperature rises; a 1-in-20-year event in ERA5 history may already be a 1-in-10-year event in 2026
  • Uneven global coverage: high quality for North America, Europe, East Asia, and Australia; reduced confidence for data-sparse regions including parts of Sub-Saharan Africa, Central Asia, and small Pacific island states

Best For

historical calibration and near-term climate-state anchoring

Strengths

  • Grounded in observation — no model uncertainty from climate model structural assumptions; reflects what has already happened
  • Near-term accuracy (1-5 year horizon) significantly exceeds CMIP6 ensemble for weather-regime and extreme event frequency
  • Company operational disruption validation: ERA5 allows direct verification of climate-to-impact pathways using real corporate loss records
  • Company-specific operational loss calibration: ERA5 allows direct backtesting of impact models against actual historical disruption records — empirical loss-per-unit-of-stress curves unavailable from any other source
  • Extreme event attribution: enables attribution of recent events (2021 US heat dome, 2022 Pakistan floods, 2023 Mediterranean wildfires) to climate change — supports regulatory disclosure, investor liability analysis, and compensation frameworks
  • ECMWF continuous update cycle: ERA5 is updated in near-real-time with the latest observational assimilation — the most current and accurate historical climate record available to any analytical platform

Maturity & Validation

Model era: Current • Status: active
Core models are internally cross-validated against institutional benchmarks. Advanced modules (DSGE, Monte Carlo, Catastrophe, Commodity) are prototype-grade — not yet independently peer-reviewed. View the full validation record at Validation Registry and current capability status at Capability Registry (JSON).

Scenario Coverage

ERA5 Historical Baseline — 1940–2025 observational record Near-Term Trend Extrapolation — 2025–2030 statistical projection from observed trends Observed State Stress — recent anomaly period (2020–2025) projected forward as physical stress scenario Long-range physical risk projection (2040–2100) — use CMIP6 or GFDL Physical Policy transition scenarios — use IMF WEO / NGFS Compound future hazard scenarios — use CE Physical Hazard Cascade Model

ERA5 is an observational reanalysis product — it cannot project forward under different emissions pathways. For scenario-conditioned long-run risk, CMIP6 or GFDL is required.

Calibration Benchmarks

ECMWF ERA5 Validation Reports (Technical Memoranda) Core model validation: temperature, precipitation, wind speed, and extreme event frequency against station observations
IPCC AR6 WG1 Chapter 1 (Observed Warming Trends) Long-run trend calibration; ERA5 is the primary observed-state input to IPCC AR6 WG1 attribution chapters
Munich Re NatCatSERVICE Loss event calibration: ERA5 weather events matched to insured loss records for physical-to-financial bridge
Swiss Re Sigma Natural Catastrophe Database Catastrophe loss calibration: extreme event ERA5 proxies validated against Swiss Re insured loss statistics
NOAA NCEI Billion-Dollar Disasters Database US-specific extreme event frequency and loss validation; ERA5 hazard intensity calibrated against NCEI economic loss records
Industry Signal Dashboard — projected signals from this model across all tracked industries
Physical Hazard Pressure by Industry
Physical hazard index (0–1) indicating asset and operational exposure to climate-related physical risks.
Transition Pressure by Industry
Regulatory and market pressure from the low-carbon transition — 0 (low) to 1 (high).
Adaptive Resilience by Industry
Resilience index (0–1) — the industry's estimated capacity to adapt to physical and transition risk.
Industry Context
Energy
ERA5 anchors the energy sector's near-term physical risk in observed data. It documents current wind resource availability (NextEra's capacity factors), solar irradiance trends, and hydropower catchment water balance — all driving real-world clean energy production. ERA5 trend data confirms the 1.2°C of observed warming already affecting cooling demand and grid stress events at scale.
Agriculture
ERA5 is the observed-state anchor for agricultural risk — it documents already-occurring shifts in growing season length, precipitation patterns, and extreme heat frequency affecting yields now. The model grounds CE's agriculture calibration in measured change rather than projections, making it particularly valuable for near-term (1–3 year) agricultural outlook. JBS's and Cargill's South American growing region ERA5 data directly calibrates the near-term yield risk signal.
Manufacturing
ERA5 provides the historical record that calibrates current manufacturing physical risk. BASF's Rhine water level data, Rio Tinto's Pilbara heat records, and ArcelorMittal's facility disruption logs are all correlated against ERA5 observed data to establish baseline loss rates. This model reports only what has been observed — making it the most conservative and credible calibration for near-term risk.
Transport
ERA5 documents observed changes in transport route availability: Arctic shipping route navigability (sea ice extent, relevant to Maersk's Arctic corridor), road surface temperature exceedance events (Union Pacific), and port infrastructure exposure to historical storm surge (Delta hub airports). ERA5 provides the empirical calibration anchor against actual disruption records for all four transport sub-modes.
Insurance
ERA5 is the primary data source for historical nat-cat loss calibration. ERA5 weather event data underlies the catastrophe models used by Munich Re, Swiss Re, and AXA to calibrate expected annual losses. CE uses ERA5 trend analysis to identify whether the frequency of loss-threshold events is already shifting — the answer is unambiguously yes across hurricane, flood, wildfire, and extreme heat peril lines globally.
Real Estate
ERA5 provides the observed record of flooding events, heat extremes, and storm damage already affecting real estate valuations. Property price data in repeatedly flooded areas shows a measurable discount emerging in the observed record — Vonovia's assets in German flood zones and British Land's London flood risk mapping are both calibrated against ERA5 precipitation and storm surge datasets.
 Back to Model Catalog  Open in Workbench