Future Climate Data for Building Simulation

Why Future Weather Matters for Buildings

Buildings designed and built today will operate for 50–100 years. A school completed in 2025 will still be in service in 2075. The climate that building will experience over its lifetime is measurably different from the climate it was designed against — and the gap grows with each decade of continued greenhouse gas emissions.

Historical weather files like TMYx represent past typical conditions. Designing against them alone risks:

• Undersized cooling systems that cannot meet peak loads in a warmer climate
• Overheated spaces in buildings designed for natural ventilation
• Underperforming envelopes whose insulation and glazing ratios were optimised for conditions that no longer exist
• Stranded assets — buildings that require expensive retrofits decades before the end of their design life

"The climates of the future will be different from today's and from those of the recent past used to derive current design data. Engineers, architects, and developers need to understand these changes to ensure that buildings are designed for the climates they will actually experience."
— Belcher, Hacker & Powell (2005), Constructing design weather data for future climates

"Buildings built today will experience significantly different climates in 2050 and 2080 from those that currently exist. Early consideration of projected climate change during the design process is therefore essential."
— Jentsch, Bahaj & James (2008), Climate change future proofing of buildings

This service provides future climate EPW files for the continental United States, enabling architects and engineers to simulate building performance under mid-century and end-of-century projected conditions alongside their historical baseline.

The Data: Argonne National Laboratory WRF Downscaling

The future climate EPW files served here are derived from the dataset:

Argonne National Laboratory. (2023). Dynamically Downscaled Hourly Future Weather Data with 12-km Resolution Covering Most of North America [data set]. https://dx.doi.org/10.25984/2202668

Published on OpenEI submission #5974.

How the data was produced

The dataset uses dynamic downscaling — a physically-based approach in which a high-resolution regional climate model is driven by the output of a global climate model (GCM).

CCSM4 produces boundary conditions (temperature, pressure, wind, humidity) at the coarse scale of a GCM (~100–200 km). WRF then resolves local topography, land use, and atmospheric dynamics at 12 km, capturing effects that GCMs cannot — mountain ranges, coastlines, orographic precipitation.

The WRF output is bias-corrected before distribution, meaning systematic offsets between the model and observed climate are reduced using statistical techniques. This makes the files more suitable for direct use in building simulation than raw model output.

Dynamic downscaling vs. morphing

Two broad approaches exist for generating future EPW files:

Morphing (e.g. Belcher et al. 2005, Epwshiftr) statistically shifts historical weather observations by climate-model-derived delta values. A historical TMYx file is "stretched" or "shifted" variable-by-variable using monthly change factors from a GCM. This preserves the diurnal and seasonal patterns of the original station data but imposes GCM-derived changes on top of them.

Dynamic downscaling (used here) runs an independent physical simulation of the regional atmosphere for the future period, constrained by the GCM at its boundaries. It does not start from historical observations — it is a forward simulation of what the atmosphere would do under the projected large-scale conditions. This is computationally more expensive but physically more consistent, particularly for variables like wind and precipitation that depend strongly on local terrain.

Neither approach is universally superior. Morphed files preserve local station characteristics; dynamically downscaled files are more physically self-consistent but carry GCM biases (hence the bias correction step).

About the EPW Files Served by This Service

The Argonne WRF dataset contains continuous hourly projections for multi-decadal future periods. The individual EPW files labeled "2050" and "2090" in this dataset are individual WRF-simulated years, not statistical composites analogous to how a TMY file is constructed.

This is an important distinction for practitioners:

• A TMYx file is assembled from 30+ years of observations by selecting the most "typical" month from each calendar month across the record — it represents the long-run central tendency.
• A WRF-simulated future year is a physically consistent time series for a specific projected year. It includes year-specific weather sequences that may include atypical runs of weather that a TMY construction would have filtered out.

For the purposes of annual building energy simulation, both are used in similar ways — as an 8,760-hour annual weather input. But users should be aware that the future files carry more year-to-year variability than a TMY-constructed file would.

Climate Scenarios: RCP 4.5 and RCP 8.5

The Representative Concentration Pathways (RCPs) are standardised greenhouse gas concentration trajectories developed for the IPCC Fifth Assessment Report (AR5, 2013). They describe different futures based on total radiative forcing — the net change in energy flux in the atmosphere — by the year 2100, expressed in watts per square metre (W/m²).

This service provides two scenarios: RCP 4.5 and RCP 8.5.

RCP 4.5 — Moderate emissions (stabilisation scenario)

Radiative forcing stabilises at 4.5 W/m² by 2100. Under the CMIP5 multi-model mean, this corresponds to a likely global mean surface warming of approximately +1.1–2.6°C above the 1986–2005 baseline by 2100, with a central estimate around +1.8°C (IPCC AR5, Table SPM.2). Mid-century warming (2046–2065) is projected at approximately +1.4°C above the same baseline under the multi-model mean.

RCP 4.5 requires substantial policy action to reduce emissions — significant deployment of carbon capture and storage, expansion of renewables, and improved energy efficiency. It does not require immediate net-zero but assumes the world makes meaningful progress on decarbonisation.

For building design, RCP 4.5 represents a moderate, policy-action future:
- A reasonable lower bound for US CONUS buildings with 30–50 year design lives
- Appropriate for projects with strong sustainability targets or in jurisdictions with aggressive climate policy
- Useful as the lower bound in a scenario pair with RCP 8.5

RCP 8.5 — High emissions scenario

Radiative forcing reaches 8.5 W/m² by 2100. Under the CMIP5 multi-model mean, this corresponds to a likely global mean surface warming of approximately +2.6–4.8°C above the 1986–2005 baseline by 2100, with a central estimate around +3.7°C (IPCC AR5, Table SPM.2). Mid-century warming is projected at approximately +2.0°C above the same baseline under the multi-model mean.

RCP 8.5 is sometimes labelled "business-as-usual," but this framing is contested. Recent analysis suggests that very high coal consumption projections embedded in RCP 8.5 are unlikely given current energy trends, though the scenario remains physically plausible and has not been ruled out. Its primary value for building design is as a high-end stress-test scenario: if a building performs acceptably under RCP 8.5, it is robust across a wide range of possible futures.

For building design, RCP 8.5 is appropriate for:
- Infrastructure and buildings with long design lives (50+ years)
- Critical facilities (hospitals, data centres, emergency services)
- Worst-case risk assessment and resilience planning
- Projects in regions already experiencing significant warming

Which scenario to use?

There is no single correct answer. Simulating both and understanding the spread is more useful than selecting one scenario as "the answer":

"No individual scenario should be thought of as a prediction of future conditions. The goal is to understand the range of plausible outcomes and to design systems that are robust across that range."
— van Vuuren et al. (2011), The representative concentration pathways: an overview

A practical simulation workflow:
1. Run baseline simulation against TMYx 2011–2025 — understand current performance
2. Run against RCP 4.5 – 2050 — mid-century, moderate emissions
3. Run against RCP 8.5 – 2050 — mid-century, high emissions
4. Optionally run 2090 variants for long-lived infrastructure or stress testing

The difference between your RCP 4.5 and RCP 8.5 results reveals your building's climate sensitivity — how much performance degrades as conditions worsen. A small spread suggests a resilient design; a large spread identifies specific systems or passive strategies that need attention.

Spatial Coverage and Resolution

Future climate files are available for the contiguous United States (CONUS) only, organised by PUMA (Public Use Microdata Area) centroids at 12 km resolution.

The dataset covers 2,368 PUMA grid points across CONUS. For a given location, this service finds the nearest PUMA centroid and returns files for that grid cell. The distance to the nearest centroid is included in every API response.

Hawaii, Alaska, Puerto Rico, and international locations are not covered by this dataset. For alternative global datasets, see Alternative Datasets below.

Known Limitations

These limitations are not defects — they are inherent to the nature of climate projections and the approaches used to generate future EPW files. Understanding them is part of using the data responsibly.

1. These files represent projected conditions for a single simulated year, not a statistical composite

Unlike TMY files (which are statistically constructed from decades of observations), the 2050 and 2090 files are individual WRF-simulated years. A single simulated year may include runs of weather — prolonged heat, cool spells — that a multi-year statistical average would smooth out. Year-to-year variability is embedded in the file.

2. Single climate model — CCSM4 does not represent the full range of uncertainty

This dataset uses one global climate model (CCSM4). Different models produce different regional projections due to differences in parameterisation, ocean-atmosphere coupling, and cloud feedbacks. CCSM4's equilibrium climate sensitivity (ECS) of 3.2°C lies within the CMIP5 model range of 2.1–4.7°C, making it a reasonable central choice, but it does not represent the tails of model uncertainty.

"Uncertainty in climate projections arises from three main sources: internal variability, model uncertainty, and scenario uncertainty. For near-term projections (2020–2050), internal variability and model uncertainty dominate; for long-term projections (2050–2100), scenario uncertainty becomes increasingly important."
— IPCC AR6 WGI, Chapter 4 (2021)

3. 12 km resolution does not fully resolve urban microclimates

A 12 km grid cell can span an urban core, its suburbs, and surrounding land. Urban heat island effects — which can add 2–5°C to nighttime temperatures in dense cities — are not fully resolved at this resolution. For urban sites, future temperatures may be warmer than the grid cell projection suggests, because the UHI effect itself may intensify with climate change.

4. RCP scenarios are not probability forecasts

RCPs are conditional projections: what would happen given different emissions trajectories. The probability of any particular RCP being realised depends on policy, technology, and economic choices that are outside the scope of climate modelling.

5. Bias correction reduces but does not eliminate model errors

The WRF output has been bias-corrected against observations. This reduces systematic errors but does not eliminate all discrepancies between modelled and real-world climate. Bias correction can also introduce artefacts if the correction relationships change under future conditions (a phenomenon called non-stationarity).

6. These data are not appropriate for every use case

This data is provided for design exploration and informational purposes. Before using these files for code-compliance, insurance, financial risk assessment, or regulatory submissions, compare them against alternative datasets and consult a qualified climate scientist or building performance specialist.

Alternative Datasets

The ORNL fTMY dataset is CMIP6-based (SSP scenarios) and covers more scenario pathways. It is a strong complement for users who want to compare RCP-era projections (this service) against newer SSP-based projections (ORNL).

Data Attribution

This dataset is a product of Argonne National Laboratory, a U.S. Department of Energy national laboratory. As a U.S. Government work, it is in the public domain within the United States.

Required citation:

Argonne National Laboratory. (2023). Dynamically Downscaled Hourly Future Weather Data with 12-km Resolution Covering Most of North America [data set]. https://dx.doi.org/10.25984/2202668

References

• Argonne National Laboratory. (2023). Dynamically Downscaled Hourly Future Weather Data with 12-km Resolution Covering Most of North America. https://dx.doi.org/10.25984/2202668

• Bass, B., New, J.R., Rastogi, D., & Kao, S-C. (2022). Future Typical Meteorological Year (fTMY) US Weather Files for Building Simulation. Zenodo. https://doi.org/10.5281/zenodo.6939750

• Belcher, S., Hacker, J., & Powell, D. (2005). Constructing design weather data for future climates. Building Services Engineering Research and Technology, 26(1), 49–61. https://doi.org/10.1191/0143624405bt112oa

• Crawley, D.B. (2008). Estimating the impacts of climate change and urbanization on building performance. Journal of Building Performance Simulation, 1(2), 91–115. https://doi.org/10.1080/19401490802182079

• IPCC. (2013). Climate Change 2013: The Physical Science Basis. Working Group I Contribution to the Fifth Assessment Report. Cambridge University Press. https://doi.org/10.1017/CBO9781107415324

• IPCC. (2021). Climate Change 2021: The Physical Science Basis. Working Group I Contribution to the Sixth Assessment Report. Cambridge University Press. https://doi.org/10.1017/9781009157896

• Jentsch, M.F., Bahaj, A.S., & James, P.A.B. (2008). Climate change future proofing of buildings — Generation and assessment of building simulation weather files. Energy and Buildings, 40(12), 2148–2168. https://doi.org/10.1016/j.enbuild.2008.06.005

• Jia, H., Chong, A., & Ning, B. (2023). Epwshiftr: Incorporating open data of climate change prediction into building performance simulation. Proceedings of Building Simulation 2023. https://doi.org/10.26868/25222708.2023.1612

• van Vuuren, D.P., et al. (2011). The representative concentration pathways: an overview. Climatic Change, 109, 5–31. https://doi.org/10.1007/s10584-011-0148-z