Welcome

The Future Weather Generator (FWG) is an open-source Java application for researchers, engineers, architects, and building-performance professionals. It transforms present-day hourly weather files into representative future weather files by applying climate-change signals derived from global and regional climate-model simulations.

FWG uses the established morphing approach to modify EnergyPlus Weather files (.epw) while retaining the hourly sequence, temporal structure, and relationships among the variables in the original weather file. The generated files can be used in dynamic building simulation to investigate future:

• Climate resilience and adaptation strategies.
• Building energy use.
• Heating and cooling demand.
• Indoor and outdoor thermal comfort.
• Overheating risk.
• Renewable-energy performance.

The application is:

• Free and open source.
• Cross-platform.
• Available through graphical and command-line interfaces.
• Suitable for individual weather files and multi-file workflows.
• Usable as a standalone application or as part of reproducible research pipelines.
• Capable of generating results from individual climate models or user-selected multi-model ensembles.

Future Weather Generator is developed and maintained by the CURA Lab – Climate and Urban Resilience in Architectural Engineering Laboratory at the University of Coimbra.

Climate-data distributions

FWG uses a shared software engine with separate climate-data distributions. Each distribution has its own geographic domain, climate-model generation, scenarios, spatial grid, model products, and scientific metadata.

Distribution	Climate framework	Domain and resolution
FWG CMIP6 Global	CMIP6 global climate models and SSPs	Global coverage; model-dependent native grids
FWG CORDEX-CMIP5 Europe (EUR-11)	EURO-CORDEX regional simulations and RCPs	Europe; 0.11° grid, approximately 12.5 km
FWG CORDEX-CMIP5 Middle East and North Africa (MNA-22)	CORDEX-Middle East and North Africa regional simulations and RCPs	Middle East and North Africa; 0.22° grid, approximately 25 km
FWG CORDEX-CMIP5 Africa (AFR-22)	CORDEX-Africa regional simulations and RCPs	Africa; 0.22° grid, approximately 25 km
FWG CORDEX-CMIP5 North America (NAM-44)	CORDEX-North America regional simulations and RCPs	North America; 0.44° grid, approximately 50 km
FWG CORDEX-CMIP5 Central America (CAM-44)	ORDEX-Central America regional simulations and RCPs	Central America; 0.44° grid, approximately 50 km
FWG CORDEX-CMIP5 South America (SAM-44)	CORDEX-South America regional simulations and RCPs	South America; 0.44° grid, approximately 50 km
FWG CORDEX-CMIP5 West & South Asia (WAS-22)	CORDEX-West Asia regional simulations and RCPs	West Asia; 0.22° grid, approximately 25 km
FWG CORDEX-CMIP5 Southeast (SEA-22)	CORDEX-Southeast Asia regional simulations and RCPs	Southeast Asia; 0.22° grid, approximately 25 km
FWG CORDEX-CMIP5 Australasia (AUS-22)	CORDEX-Australasia regional simulations and RCPs	Australasia; 0.22° grid, approximately 25 km
FWG Portugal	High-resolution WRF regional simulation and SSPs	Mainland Portugal; approximately 9 km

CMIP is coordinated by the World Climate Research Programme. Its simulations support climate research and inform assessments such as those produced by the Intergovernmental Panel on Climate Change; the IPCC does not itself produce the CMIP datasets.

---

Shared scientific methodology

Unless otherwise stated for a specialized distribution, FWG climate resources use the following methodology.

Reference and future periods

Climate-change signals are calculated by comparing a reference climatology with future 30-year climatologies. The standard future periods are:

• 2050 timeframe: 2036–2065.
• 2080 timeframe: 2066–2095.

The year used in the timeframe name represents the approximate midpoint of the corresponding 30-year period.

Monthly climate-change signals

For every grid point, climate model, scenario, future period, variable, and calendar month, FWG derives:

• A central climate-change signal.
• A measure of within-model variability associated with that signal.

For additive variables, the central signal is calculated as the future monthly mean minus the reference monthly mean. For multiplicative variables, it is calculated as the ratio between the future and reference monthly means.

Individual-model uncertainty uses the propagated variability of the future-minus-reference change or future-to-reference factor. Multi-model ensemble uncertainty is calculated separately from the dispersion of the central climate-change signals across the selected models.

Climate variables

The climate resources include monthly information for ten principal variables:

• Near-surface air temperature.
• Daily maximum near-surface air temperature.
• Daily minimum near-surface air temperature.
• Specific humidity.
• Surface downwelling shortwave radiation.
• Total cloud cover.
• Surface snow amount.
• Near-surface wind speed.
• Mean sea-level pressure.
• Precipitation.

FWG uses these climate signals directly or together with physical and psychrometric relationships to modify the meteorological variables required by building-simulation weather files.

Spatial interpolation

FWG supports several interpolation methods, depending on the climate grid and selected workflow:

• Nearest valid grid point.
• Unweighted mean of the four nearest valid points.
• Geodesic inverse-distance-squared interpolation.
• Rectilinear bilinear interpolation.
• Adaptive structured-cell interpolation.
• Structured-grid inverse-isoparametric bilinear interpolation.
• Structured-grid triangular barycentric interpolation.

Regional model grids retain their native topology and geographic coordinate information. FWG does not create artificial regional grids by independently sorting latitude and longitude.

Individual models and ensembles

Users may generate weather files from:

• One selected climate model or GCM–RCM product.
• Several models processed independently.
• A user-selected multi-model ensemble.

Ensemble statistics are calculated independently for every location, scenario, period, month, and climate variable.

Solar radiation decomposition and photometric modeling

Solar radiation and daylight variables in an EPW file are physically interdependent. After the global horizontal radiation climate-change signal has been applied, FWG reconstructs the corresponding direct and diffuse components and then recalculates the dependent photometric fields.

This shared processing stage is applied consistently to:

• Corrected present-day baseline files.
• Present-day control files.
• Individual-model future weather files.
• Multi-model ensemble weather files.
• Climate-uncertainty variants.

Solar radiation decomposition

FWG uses the morphed global horizontal irradiation to estimate:

• Direct normal irradiation.
• Diffuse horizontal irradiation.
• Direct horizontal irradiation, as an intermediate quantity.

For each weather-file interval, the reconstructed components satisfy the solar-radiation closure relationship:

Global horizontal irradiation = diffuse horizontal irradiation + direct normal irradiation × solar projection factor

The projection factor is calculated from interval-aware solar geometry rather than from a single unqualified hourly solar-position value.

All candidate estimates are processed using common physical constraints, including:

• Nonnegative direct and diffuse radiation.
• Exact closure among global, direct, and diffuse components.
• Astronomical night-time behavior.
• Limits derived from extraterrestrial radiation.
• Low-sun attenuation.
• Restrictions on direct normal radiation near sunrise and sunset.
• Model-specific eligibility conditions.
• Consistent treatment of partially sunlit intervals.

These controls prevent mathematically possible but physically implausible combinations of global, direct, and diffuse radiation.

Available decomposition methods

Method	Year	Description
Constrained multimodel ensemble	—	Calculates the median of eligible Erbs, Orgill–Hollands, reduced Reindl, and Maxwell DISC candidates after applying the common physical constraints. This is the default method.
Orgill–Hollands	1977	Hourly diffuse-fraction formulation based on the clearness index.
Erbs, Klein & Duffie	1982	Piecewise hourly diffuse-fraction model adapted to interval-integrated EPW radiation.
Maxwell DISC	1987	Estimates direct normal irradiation from global radiation, solar geometry, air mass, atmospheric pressure, and extraterrestrial radiation.
Reduced Reindl, Beckman & Duffie	1990	Estimates the diffuse fraction from the hourly clearness index and solar elevation.
Boland logistic	2008	Logistic diffuse-fraction formulation based on the hourly clearness index.
Ridley–Boland–Lauret	2010	Multiple-predictor model using hourly and daily clearness indices, apparent solar time, solar altitude, and short-term persistence.
Engerer, Method 2	2015	Multiple-predictor diffuse-fraction model incorporating clearness, solar geometry, solar time, clear-sky behavior, and cloud-enhancement effects.
Paulescu–Blaga	2019	Two-predictor diffuse-fraction model based on hourly and daily clearness indices.

The constrained multimodel ensemble intentionally contains only four constituent methods:

• Erbs, Klein & Duffie.
• Orgill–Hollands.
• Reduced Reindl, Beckman & Duffie.
• Maxwell DISC, when the interval is within its eligible solar-geometry domain.

The Boland, Ridley–Boland–Lauret, Engerer, and Paulescu–Blaga formulations remain available as individual sensitivity-analysis methods but are not included in the default constrained ensemble.

Default and advanced use

The Standard FWG workflow uses the constrained multimodel ensemble. This reduces dependence on a single empirical formulation while preserving explicit physical closure.

The Advanced Climate Change workflow allows users to select an individual decomposition model.

Photometric models

After the solar-radiation components have been finalized, FWG regenerates the four photometric fields defined in the EPW format:

• Global horizontal illuminance, in lux.
• Direct normal illuminance, in lux.
• Diffuse horizontal illuminance, in lux.
• Zenith luminance, in candelas per square meter.

FWG provides two photometric methodologies.

The Perez (1990) method is the default and fully generative photometric model. It calculates the four photometric outputs from:

• Global horizontal irradiance.
• Direct normal irradiance.
• Diffuse horizontal irradiance.
• Solar geometry.
• Sky clearness and brightness conditions.
• Dew-point temperature and associated atmospheric information.

EPW SolarQC integration

Before climate morphing, EPW SolarQC evaluates the baseline solar and photometric package. It checks:

• The temporal basis of the EPW solar fields.
• Global, direct, and diffuse component closure.
• Extraterrestrial-radiation limits.
• Radiation assigned to astronomical night.
• Missing or invalid radiation values.
• Consistency between radiation and photometric fields.
• Eligibility for source-calibrated luminous efficacy.

Depending on the evidence and selected correction policy, the baseline may be:

• Accepted unchanged.
• Corrected and revalidated.
• Rejected when a defensible automatic correction cannot be established.

The original baseline EPW is never overwritten. When correction is necessary, FWG creates a separate corrected baseline and records the actions, model selections, quality-control results, and residual limitations in the output provenance.

---