A Literature Review on the Use of Weather Data for Building Thermal Simulations

Abstract

Thermal simulations of buildings play a critical role in optimizing energy efficiency, thermal comfort, and heating, ventilation and air conditioning (HVAC) systems design. Accurate weather data is essential for reliable simulations, as local weather and climate have a significant impact on energy requirements for space heating and cooling and thermal comfort. This study conducted a literature review regarding the sources, types, and uncertainties of weather data used for thermal simulations of buildings, including typical meteorological years (TMYs) and extreme weather files under current and future climates. Additionally, this paper evaluates methods for weather data processing, including interpolation, downscaling, and synthetic generation, to improve simulation accuracy. Finally, approaches are proposed for constructing weather files for the future and extreme conditions under a changing climate. This review aims to provide a guide for researchers and practitioners to enhance the reliability of thermal modeling through informed construction, selection, and application of weather data.

1. Introduction

The built environment significantly influences global energy demand and greenhouse gas emissions, being responsible for approximately 36% of finial energy use and 39% of energy-related CO₂ emissions worldwide [1]. As climate change intensifies and energy efficiency standards evolve, optimizing building performance through advanced thermal simulation has become indispensable. In building thermal performance simulation (BTPS), accurate weather data is critical for predicting the thermal behavior, energy usage, and indoor comfort of buildings. Weather files serve as a fundamental input, providing detailed climatic information such as temperature, humidity, solar radiation, and wind speed. These files are essential for simulating real-world environmental conditions and assessing how a building’s design will perform in diverse climates. The reliability of these simulations is inherently tied to the quality, representativeness, and temporal resolution of the input weather data.

Traditionally, typical long-term building thermal performance simulation tools utilize standardized weather files such as TMY [2], Test Reference Year (TRY) [3], Representative Meteorological Year (RMY) [4], and International Weather for Energy Calculations (IWEC) data [5]. In a previous article, 32 commonly used online sources were summarized and critically reviewed by Amin and Mourshed [6]. These datasets were often derived from historical weather observations, typically spanning a 15 to 30-year period, to represent “typical” climate conditions of a location.

Hall et al. [7] pioneered the original method for creating a TMY by selecting months from different years, based on Finkelstein–Schafer [8] statistical comparison to the long-term climatic normal. This method laid the foundation for subsequent TMY2 [9] and TMY3 [2] datasets. Alternative methods for generating TMY data were reviewed by Janjai and Deeyai [10]. They compared three methods: the Sadia National Laboratory method [6], the Danish method, and ref. [11] the Festa and Ratto method [12]. They concluded that there was no significant difference between these three methods.

A recent major review by Rady et al. [13] analyzed 530 publications from 1978 to 2023. It outlined limitations in existing TMY databases, including “sparse station data, fixed weighting scheme, and failure to account for microclimate or abrupt climate change”. This review also identified four key research directions for developing more accurate and representative TMY datasets for building thermal performance, listed as follows:

• Expand case study scales;
• Refine TMY meteorological parameters;
• Improve techniques for data processing;
• Enhance validation strategies for improving TMY data.

While TMY data simplifies design processes, its static nature often fails to account for extreme weather events, microclimatic variations, or the accelerating impacts of climate change, which have been summarized in review studies [14,15]. Furthermore, the growing use of probabilistic future climate projections (critical for resilient building design) introduces complexities related to data uncertainty, spatial resolution, and methodological consistency. These challenges are compounded by regional disparities in weather station coverage, data accessibility, and the technical limitations of interpolation or downscaling techniques. These potentially lead to discrepancies between predicted and actual building performance. Recent studies conducted in Sweden [16] and Western Sydney [17] indicate that outdated TMY weather data is unsuitable for BTPS.

The increasing awareness of climate change and its impact on the built environment has driven significant interest in future weather files for building simulations. Future weather files that integrate projected climate scenarios are thus becoming essential data for building simulations to account for the anticipated changes in temperature, humidity, solar radiation, and other meteorological variables.

Recent advancements have seen the development of new methods for generating weather files, including those based on real-time data, stochastic weather generation [18,19], and climate projections [20]. These approaches aim to improve the representativeness and accuracy of weather inputs, especially as the impact of climate change introduces greater variability and extremes in weather patterns.

Buildings are typically designed for long lifespans—often exceeding 50 years—making it essential to consider how future climate conditions may impact their energy performance and thermal comfort. Future weather files, which are developed using climate projections from models such as those produced by the Intergovernmental Panel on Climate Change (IPCC), play a crucial role in this process. These files simulate a range of plausible future climates based on different greenhouse gas emissions and socio-economic pathways [21], enabling designers, engineers, and researchers to assess buildings’ resilience and identify appropriate adaptation strategies.

However, despite substantial progress in both climate modeling and building performance simulation, significant gaps remain in the integration and application of weather data. There is a lack of clear guidance on how to select, process, and tailor weather data to meet specific simulation goals, regional contexts, and time horizons. While previous studies have addressed discrete components such as TMY generation or the downscaling of climate model outputs, there is a lack of comprehensive reviews that critically examine the full pipeline from weather file generation to application in building simulations.

Furthermore, as highlighted by Zeng et al. [22], a persistent disconnect exists between the climate modelling and building energy research communities. This gap, rooted in historical separation of disciplines, has limited collaboration and slowed the effective translation of climate science into practical tools for building performance analysis.

This review aims to address these shortcomings by providing a holistic overview of the development and use of weather files in building thermal performance simulations, focusing on generating future weather files for thermal simulations of buildings. It examines key methodologies, recent innovations, and practical limitations in the generation of both typical and extreme weather data for present and future climates, considering climate change. By synthesizing knowledge across disciplines, this study seeks to bridge the gap between climate science and building simulation practice, ultimately contributing to more robust and future-ready building design strategies.

2. Weather Files for Building Thermal Simulations

As proposed by Herrera et al. [14], weather files should ideally include examples of typical conditions, contain data for extreme conditions, and match the temporal resolution needed by simulation tools, which is typically hourly or finer. They should have a geographical resolution that matches changes in local topography in the region of interest, they should explore the effects of the urban micro-climate, they should include potential future climate scenarios, preferably accounting for climate change impacts, and they should demonstrate a validated history of industrial application.

In early assessment models, weather data from typical weather years were used, derived from hourly historical observations spanning up to 30 years at specific locations. However, to account for the impacts of future climate change, it became necessary to incorporate climate projections into these weather files. Additionally, to better respond to severe weather conditions, it is essential to include weather files that represent ‘extreme’ years.

2.1. Files for Typical Weather Conditions

Typical weather files contain hourly data on temperature, humidity, wind speed and direction, global horizontal radiation, and diffuse solar radiation. These files represent average weather conditions over a set of baseline years and are used to estimate buildings’ average energy consumption and carbon emissions. A typical weather file is generated from historical data, usually spanning 15–30 years, depending on data availability.

Table 1. A short list of typical weather files used in various countries (adapted from [14]).

Weather File Name	Region	Sites	Period	References
Representative Meteorological Year (RMY)	Australia	83	1991–2015	[4]
Chinese Standard Weather Data (CSWD)	China	270	1982–1997	[23]
Indian Typical Years from ISHRAE (ISHRAE)	India	62	1991–2005	[24]
Italian ‘Gianni De Giorgio’ (IGDG)	Italy	68	1951–1970	[25]
Typical Meteorological Year (TMY3)	New Zealand	18	1993–2023	[26]
Spanish Weather for Energy Calculations	Spain	52	1961–1990	[27]
Test Reference Year (TRY, CIBSE)	UK	14	1984–2013	[28]
Typical Meteorological Year (TMY)	USA and others	1020	1991–2005	[2,9]
Weather Year for Energy Calculations (WYEC)	USA/Canada	77	1953–2005	[29]
International Weather for Energy Calculations (IWEC)	Worldwide	3012	1991–2005	[5]

provides a representative sample of typical weather files used globally.

Two approaches are used to construct a typical weather year. The first approach identifies a continuous 12-month period as typical. The second approach applies ranking criteria to individual months from the baseline dataset, which are then combined into a composite 12-month year. The Australia RMY, UK TRY, and TMY all use the second approach, computed using the Finkelstein–Schafer (FS) statistic [8]. Each month in the file may come from different years. A comparison study by Kershaw, Eames, and Coley [30] showed that these composite years, derived from the baseline dataset, have advantages over the single-year approach.

Note that TMY, RMY, and TRY datasets (Table 1) are commonly used for regulatory compliance, which may be updated to include more recent data, depending on the region. For instance, the Canadian Weather Year for Energy Calculations (CWEC) data have been updated to 2020 [31], the latest version of CSWD is based on data from 1982–2010, and TMY weather files used for the Title 24 building standards of California were updated based on 20 recent years of data for each station (1998–2017) [24].

2.2. Files for Extreme Weather Conditions

The typical weather files discussed above are suitable for calculating typical annual energy consumption to represent long-term energy performance. However, these files, developed based on average weather data, are not suitable for extreme or atypical weather conditions. Extreme weather events, such as heat waves, droughts, or cold snaps, are crucial for analyzing building overheating, thermal comfort, and peak energy demand. Due to climate change, extreme weather events are expected to occur more frequently in the future. The building industry is addressing these challenges by adapting building designs and occupant behaviors to better cope with such events.

As global warming occurs, heatwaves are projected to increase in frequency, intensity, and duration. The IPCC reports that severe heat waves, which used to occur once every 50 years, are now likely to happen once per decade [32].

2.2.1. Files for Extreme Weather Conditions

With climate change, cold extremes in Australia are likely to become less common and less severe, although they can still occur as a result of natural climate variability. This study focuses on heat waves.

The definition of a heatwave can vary based on context, geographic location, and the specific criteria used. Generally, a heatwave is defined as an extended period of abnormally high temperatures that are higher than the expected norm for a particular region and time of year [33]. However, several specific definitions and criteria are commonly used, including the following:

• Meteorological Definitions
  • World Meteorological Organization (WMO): The WMO typically defines a heatwave as a period of at least three consecutive days during which the daily maximum temperature exceeds the average maximum temperature by at least 5 °C [34].
  • Australian Bureau of Meteorology (BoM): The BoM uses a heatwave definition based on the excess heat factor (EHF) index, which considers both the magnitude of high temperatures and their persistence. In Australia, a heatwave is classified when there are three or more unusually hot days in terms of maximum and minimum temperatures compared with the location’s climatological norms [35].
• Threshold-Based Definitions
  • Some definitions use absolute temperature thresholds, such as temperatures above 35 °C or 40 °C, for a set number of days. This type of definition is often used in regions with extreme climates [36].
  • Percentile-based thresholds are also common. For instance, a heatwave can be defined as temperatures above the 90th or 95th percentile of daily maximum temperatures for at least three consecutive days [37].
• Health Impact-Based Definitions
  • Heatwaves can be defined based on their impact on human health, such as when prolonged high temperatures significantly increase the risk of heat-related illnesses like heatstroke or heat exhaustion [38].
  • This type of definition might incorporate factors such as high humidity, heat index (feels-like temperature), and overnight minimum temperatures, which can exacerbate heat stress [38].
• Urban and Environmental Definitions
  • In urban areas, heatwaves might be influenced by the urban heat island (UHI) effect, where built environments experience significantly higher temperatures than rural areas. The relative temperature anomaly compared to surrounding rural areas might be used to characterize heatwaves in cities [39].
• Climate Change Context
  • In the context of climate change, heatwaves are being redefined due to increasing global temperatures. A new classification, “compound heatwaves,” considers consecutive extreme temperature events across multiple regions or recurrent events within a season [40].

2.2.2. Methodology for Development of Weather Files for Atypical Years or Extreme Conditions

Various methods have been employed to generate weather files for extreme conditions or atypical years. For example, the weather parameters used in the Test Reference Year (TRY) computation were weighted differently depending on whether the weather file was intended for heating or cooling analysis, as outlined by Pernigotto et al. [41] in accordance with the international–European technical standard EN ISO 15927-4 (https://www.iso.org/obp/ui/es/#iso:std:iso:15927:-4:ed-1:v1:en, accessed on 6 July 2025). Similarly, Kakamess et al. [42] applied a comparable approach to adjust the climate zone weather parameters, modifying EN ISO 15927-4 to better estimate energy demand for space heating and cooling.

As an alternative to the previously mentioned approaches, specific designed years have been developed to represent atypical years. These are described as follows:

• Design Summer Year (DSY): Developed to estimate the impact of warmer-than-average summers, the DSY was initially used for sizing mechanical cooling systems [43]. It does not account for extreme temperatures of individual months or incident solar radiation. The DSY is determined using average dry bulb temperatures from April to September and represents the year that falls in the middle of the upper quartiles of the dataset, typically the third warmest summer over a 20-year period. However, this method can miss high temperatures in relatively cool summers, such as the summer of 2003, which caused numerous deaths across Europe but was not highly ranked based on average summertime temperature.

To improve sensitivity to extreme weather conditions, several variations of the DSY have been developed, including the following:

• Probabilistic Design Summer Year (pDSY): Introduced by Hacker, Belcher, and White [44] and further refined by Eames [45], this variation incorporates probabilistic methods to better capture extreme events;
• Summer Reference Year (SRY): Developed by Jentsch, Eames, and Levermore [46], the SRY aims to provide a more accurate representation of typical summer conditions;
• Near Extreme Design Reference Year (DRY): Proposed by Watkins, Levermore, and Parkinson [47], and Du, Underwood, and Edge [48], the DRY focuses on near-extreme conditions to better estimate energy demand for space heating and cooling;
• The Extreme Meteorological Year (XMY) builds upon the concept of the TMY, using the same weather variables and similar weightings [49]. The XMY is designed to represent a year with the most extreme weather conditions, featuring the hottest summer and the coldest winter. This is achieved by selecting months with the highest and lowest hourly average values from the basis years (1999–2013).

It should be noted that there is no internationally standardized threshold or statistical definition for what constitutes “extreme” in this context. For example, some studies use percentile thresholds (e.g., top 5% of temperature), while others rely on absolute values or ranking based on cumulative metrics. This lack of standardization makes comparisons across studies or climates challenging and highlights a need for clearer guidance on threshold criteria.

• The Untypical Meteorological Year (UMY) was developed based on the Weather Year for Energy Calculation 2 (WYEC2) methodology, with modified weight parameters [50]. In an UMY, the most critical factors are the maximum and minimum dry bulb temperature, solar radiation, and wind speed. The results obtained using UMY weather files are comparable to those of TMY2 under normal conditions. However, the UMY provides better predictions of maximum energy consumption during extreme weather events.
• The Hot Summer Year (HSY) has two versions: HSY1 and HSY2 [51]. Both versions use the same base period from 1975–2006.

HSY1 selects the summer year (June, July, August) with the highest weighted cooling degree hours (WCDHs). This metric incorporates not only the magnitude but also the duration of elevated temperatures above a reference base temperature (typically 24 °C), effectively weighting both intensity and persistence of heat. The exact base temperature and weighting function may vary across studies, leading to inconsistencies in the interpretation of what constitutes the “hottest” year.

HSY2 identifies the summer year with the most hours of physiologically equivalent temperature (PET) exceeding 23 °C, a thermal comfort index that combines air temperature, humidity, wind speed, and radiation [52]. The 23 °C threshold represents the lower bound for slight thermal stress in moderate climates, but its applicability may vary depending on the regional context. Again, the selection of this PET threshold is not universally standardized and may not reflect the same level of thermal stress in all climates or populations.

While both HSY1 and HSY2 provide a structured method to identify hot years based on thermal load or comfort indices, there is currently no consensus on threshold definitions, base temperatures, or selection criteria, which limits comparability between regions or applications. Standardization of metrics and thresholds, particularly those accounting for regional thermal comfort expectations, remains a key area for future development.

• The Three Most Extreme Heatwaves method was proposed by Machard et al. [20] for studying the resilience of buildings, based on the three criteria of the most intense, the most severe, and the longest heatwaves. The method developed by Ouzeau et al. [37] was adopted in that study [20] to select heatwaves in terms of three temperature relative thresholds from a multiyear period. The threshold of 99.5% is applied to detect a temperature peak and a potential heatwave; 97.5% is used to determine the heatwave duration (days between the heatwave starts and ends) and severity (degree days above the threshold). The heatwave stops when the temperature is lower than this threshold for more than three consecutive days. The 95% threshold is applied to end the heatwave when the temperature drops below this level. In this proposed method, each heatwave features three criteria: intensity (maximum daily mean temperature in °C during the heatwave), duration (in days), and severity (aggregated temperature above 97.5% threshold in °C.day).

2.3. Future Weather

Frequency and intensity of extreme weather events are expected to increase due to climate change. The built environment will face climate change-related challenges, including risks of overheating.

Possible future carbon emission scenarios were developed by the Intergovernmental Panel for Climate Change (IPCC) using different social-economic scenarios [53,54]. These emission scenarios have evolved into Representative Concentration Pathways (RCPs), which are now provided as inputs to both Global Circulation Models (GCMs) and finer resolution Regional Climate Models (RCMs) [55].

The GCMs and RCMs produce average weather data over regions or numerical grids. The grid sizes are dependent upon the model resolution. The numerical grids of GCMs are too coarse to represent the impacts of climate change on the built environment. RCMs can produce finer-scale geographic and land surface information and provide weather and climate data at horizontal resolutions up to 25 km. Higher resolutions (10–20 km horizontal resolution) can be obtained by combining several RCMs through multi-model ensembles and dynamical downscaling techniques [56,57]. An alternative to RCMs is to ‘downscale’ GCM models to a finer grid using statistical techniques.

Future TMY (fTMY) and extreme weather files are widely used for building simulation, which can be created using the following three methodologies:

• Morphing Techniques: This is a widely used method where historical weather data is morphed to represent future conditions. The morphing method typically adjusts temperature, solar radiation, and humidity based on GCM outputs [58];
• Stochastic Weather Generators: These generators use probabilistic models to create synthetic weather data that incorporates the variability and extremes predicted in future climate scenarios (https://www.ipcc-data.org/guidelines/pages/weather_generators.html, accessed on 24 June 2025). This method accounts for uncertainty and the range of possible future climates;
• Downscaling Climate Models: Global climate models (GCMs) simulate climate data on a coarse scale. To create weather files relevant for specific locations, GCM outputs are statistically or dynamically downscaled. Statistical downscaling uses historical climate data to adjust GCM outputs, while dynamic downscaling employs regional climate models (RCMs) [59].

These three methods are detailed in the following sections.

2.3.1. Morphing Techniques

The morphing technique is one of the most widely used methods for generating future weather files. It modifies historical weather data (such as TMY files) to reflect projected changes in climate variables like temperature, solar radiation, and humidity, based on outputs from climate models. The basic idea is to “morph” historical data by applying climate change signals, typically derived from GCMs/RCMs.

∘

The process is as follows:

▪

Step 1: Climate change projections (e.g., temperature rise, changes in humidity) are extracted from GCMs/RCMs for future time periods;

▪

Step 2: These changes are applied to the historical weather data in a linear or proportional manner;

▪

Step 3: The morphed data reflects future conditions, maintaining the temporal patterns of the historical data but with modified magnitudes.

∘

Key examples include the following:

▪

Belcher, Hacker, and Powell [59] proposed a morphing method that adjusts TMY data based on future climate predictions from the UK Met Office’s Hadley Centre model. They applied changes to temperature, humidity, and solar radiation using climate projections for different future time horizons (e.g., 2020, 2050, 2080);

▪

With requirements from Department of Climate Change, Energy, the Environment and Water (DCCEEW) for future weather files for building energy simulations, CSIRO has developed and published projected fTMY climate files [60] by adopting one of the widely accepted morphing techniques introduced by Belcher et al. [59] with three RCPs (RCP 2.6, 4.5 and 8.5) for four averaged time periods: the 2030s, 2050s, 2070s, and 2090s.

▪

WeatherShift is a tool that allows users to adjust historical weather files to reflect future climate scenarios. It uses climate models to predict temperature, humidity, and solar radiation shifts [61];

▪

The CCWorldWeatherGen tool morphs TMY weather data to produce future weather files based on climate projections from GCMs [62].

∘

Advantages include the following:

▪

The resulting model maintains the same temporal structure and patterns of the original weather data, making it easy to integrate into existing simulation tools;

▪

The process is simple and computationally efficient.

∘

Limitations include the following:

▪

This method assumes that future temporal weather patterns (e.g., seasonal variations) will remain similar to historical patterns, which might not always be the case;

▪

This resulting model may not adequately account for extreme weather events or changes in variability.

2.3.2. Stochastic Weather Generators

Stochastic weather generators use statistical models to simulate weather sequences that are consistent with observed weather patterns but include variability and extremes that may occur in future climates. These generators can model both average conditions and the variability and extremes that might be underrepresented in simple morphing techniques.

∘

A typical process is as follows:

▪

Step 1: A statistical model is built using historical weather data to replicate daily or hourly weather patterns;

▪

Step 2: Projections from climate models are used to adjust the parameters of the statistical model (e.g., mean temperature, frequency of extreme events);

▪

Step 3: The generator produces synthetic weather sequences based on the adjusted parameters.

∘

Key examples include the following:

▪

WGEN65 [18] models the sequential nature of precipitation using a Markov chain, where the likelihood of precipitation on a given day depends only on the previous day’s conditions. The maximum and minimum temperatures, along with solar radiation, are conditioned on whether the day is classified as wet or dry. The relationships among these three variables are preserved using cross-correlation coefficients across different time steps;

▪

LARS-WG (Long Ashton Research Station Weather Generator [63]) is a tool that generates daily weather data using a stochastic process, and it has been adapted for generating future weather scenarios. It uses GCM outputs to adjust the parameters of the weather generator, simulating future weather conditions for specific locations.

∘

Advantages of these methods include the following:

▪

They provide a more accurate representation of extreme weather events and variability compared with morphing methods;

▪

They can generate multiple realizations of future weather, capturing a range of possible outcomes.

∘

Limitations include the following:

▪

The primary limitation of stochastic weather generators is their reliance on statistics derived from historical weather observations. Consequently, extreme events are unlikely to be accurately represented, as such events are rare or may not have occurred within the limited historical record used to create the generator. Additionally, there is an inherent assumption that future weather patterns will mirror those observed historically, which may not account for changes due to factors like climate change;

▪

These processes require detailed statistical knowledge to set up and calibrate;

▪

The quality of the generated weather files depends on the robustness of the underlying statistical model.

2.3.3. Downscaling Climate Models

GCMs provide global climate projections at a coarse spatial resolution (typically 100–300 km grid cells), which is insufficient for building-scale simulations. Downscaling refers to the process of converting coarse-resolution outputs from GCMs into higher-resolution data suitable for local or regional analysis.

∘

Types of downscaling include the following:

▪

Statistical Downscaling uses relationships between historical large-scale climate variables (e.g., pressure systems, jet streams) and local weather observations to predict local weather patterns based on GCM outputs;

▪

Dynamical Downscaling employs RCMs, which use physical models to simulate climate processes at a finer scale over a specific region.

∘

The process is as follows:

Step 1: Selection of climate models

• Global Climate Models (GCMs) provide large-scale climate projections but at a coarse resolution;
• Regional Climate Models (RCMs) are used for higher resolution data, and these are often driven by GCM outputs.

Step 2: Downscaling

• Dynamical downscaling involves running high-resolution RCMs that use GCM outputs as boundary conditions. This method is computationally intensive but provides detailed local climate projections [64];
• Statistical downscaling establishes statistical relationships between large-scale climate variables and local climate conditions. This method is less computationally demanding and can be applied to specific locations or grids [65].

Step 3: Bias Correction

• The downscaled model outputs can be tested against (gridded) observations, e.g., AGCD (Australian Bureau of Meteorology Australian Gridded Climate Data) and Queensland Department of Science, Information Technology and Innovation SILO (Scientific Information for Land Owners), or model data, e.g., BARRA reanalysis (Bureau of Meteorology’s Atmospheric high-resolution Regional Reanalysis for Australia) [66] to identify and correct biases;
• Adjustment involves applying methods (varying from simple scaling approaches to more complex univariate techniques) to adjust the model outputs, ensuring they align more closely with observed data [67].

Step 4: Generation of Weather Files

• Data Integration requires integrating the downscaled and bias-corrected data into weather files, which include variables such as temperature, radiation, precipitation, and wind speed;
• The projected historical weather data are validated against independent observed data to ensure accuracy and reliability [67,68].

∘

Advantages include the following:

▪

This approach provides high-resolution data that captures local topography, urban heat islands, and microclimates;

▪

With the advantage of using a multi-model ensemble, it is more physically realistic than stochastic methods, particularly for extreme events.

∘

Limitations are as follows:

▪

Dynamical downscaling is computationally expensive and requires significant computing power;

▪

The accuracy of the downscaled data depends on the quality of the GCM output and the downscaling methodology.

To provide key examples, a sample of widely used international climate models (CMIP—Coupled Model Inter-comparison Project) is summarized as follows:

• CMIP3/5-Coupled Model Inter-comparison Project Phase 3 and Phase 5 [69,70]

The Coupled Model Intercomparison Project (CMIP) was initiated by the World Climate Research Program’s Working Group on Coupled Modelling to provide a standardized experimental framework. It aims to study the output of coupling comprehensive three-dimensional atmospheric GCMs with oceanic general circulation models, land-surface processes, and sea-ice models. CMIP3 is based on the Special Report on Emissions Scenarios (SRES) defined in the IPCC Fourth Assessment Report (AR4) [71], while the newer CMIP5 uses the Representative Concentration Pathways (RCP) for greenhouse gases [70]. Both CMIP3 and CMIP5 have been utilized to generate projections of future global climate conditions using numerous GCMs. Various downscaling approaches are employed for CMIP3 and CMIP5 models, typically achieving a spatial resolution of approximately 50 km. European initiatives like ENSEMBLES [72] and EURO-CORDEX [56] provide alternatives through extensive use of RCMs. Based on EURO-CORDEX [56], Ouzeau et al. [37] developed weather files for heat waves analysis for France, and Machard et al. [20] applied the method developed by Ouzeau et al. [37] to generate current and future projected weather files for building simulations across 15 major cities worldwide, including typical and extreme weather years.

Climate models are continually being updated. The next generation, CMIP6 is now available, emphasizing enhanced spatial resolution, the inclusion of new physical processes, and the integration of biogeochemical cycles [73]. CMIP6 involves contributions from over 100 climate modelling groups worldwide, using a diverse range of GCMs and Earth system models (ESMs). It uses Shared Socioeconomic Pathways (SSPs) combined with RCPs to explore future climate outcomes. These scenarios include the following:

▪

SSP1-2.6: Low emissions pathway, reflecting a future focused on sustainability and green development;

▪

SSP2-4.5: Moderate emissions scenario, representing a “middle-of-the-road” trajectory with no major deviations from historical patterns;

▪

SSP3-7.0: High emissions pathways, characterized by regional rivalry, limited international cooperation, and high challenges to mitigation and adaptation.

▪

SSP5-8.5 (very high emissions, fossil-fuel development).

CMIP6 represents a significant step forward in climate modelling, proving critical insights into the past, present, and future of Earth’s climate system. Its scientific contributions are as follows:

▪

Improved projections of global and regional climate change;

▪

Enhanced understanding of climate feedback, such as cloud and ice–albedo feedback;

▪

Insights into extreme events, sea-level rise, and carbon cycle dynamics;

▪

Support for policy-relevant climate assessments.

CMIP6 data is stored in the Earth System Grid Federation, a distributed database that provides open access to model outputs for researchers worldwide. Its outputs are widely used by scientists, policymakers, and stakeholders to inform climate mitigation and adaptation strategies.

• UKCIP02/UKCP09-UK Climate Impacts Program 2002 and UK Climate Projections 2009

UKCIP02 predicts climates at a spatial resolution of 50 km, while UKCP09 improves this to 25 km. UKCIP02 climate projections were based on four of the emission scenarios published by the IPCC in 1990 for three future time-slices, whereas UKCP09 uses three of the IPCC emission scenarios and employs a probabilistic projection methodology [74]. This methodology accounts for climate modelling uncertainties by combining the outputs from HadCM3—a global climate model developed by UK Met Office—with projections from a diverse ensemble of international climate models. UKCP09 represents an evolution of the projections achieved by UKCIP02.

Building on the success of UKCP09, the UK Climate Projections 2018 (UKCP18) further updates the probabilistic projections. UKCP18 provides high-resolution future climate projections at a global scale of 60 km scale and for the UK at 12 km. Additionally, the 12 km climate model has been further downscaled to 2.2 km, a resolution previously used for short-term weather forecasts. This enables more realistic simulation of high-impact events, such as localized heavy summer rainfall.

• Australian Climate Futures [75]—a tool that allows users to explore regional climate projections with a range of datasets

The tool was developed based on CSIRO’s Representative Climate Futures Framework [76,77], which incorporates projections from GCMs and RCMs as well as statistically downscaled results. The GCM data can be obtained from CMIP 5 (representative carbon scenarios of the Fifth Assessment Report) or CMIP3 (representative carbon scenarios of the Fourth Assessment Report).

Projected changes from the latest (CMIP5) models are available from up to 40 GCM simulations, 16 CCAMs (CSIRO’s Conformal-Cubic Atmospheric Models), and 22 BOM-SDM (the Bureau of Meteorology Statistical Downscaling Model) simulations. These data are available for analysis across 14 future time periods, ranging from 2025 to 2090 in five- year increments and four carbon emission scenarios (RCP2.6, RCP4.5, RCP6.0, and RCP8.5).

Projected data from earlier (CMIP3) models are available from up to 18 GCM simulations. The projections are available for three future time periods—2030, 2055, and 2090— and cover three emissions scenarios: B1, A1B, and A2.

Users can explore and obtain data for projected monthly, three-monthly, six-monthly, and annual changes in up to 16 variables, including temperature, relative humidity, wind speed, solar radiation, rainfall, etc.

• WRF (Weather Research and Forecasting Model) [78]

WRF is an RCM that is commonly used for downscaling GCM outputs to generate high-resolution future weather data for specific locations. WRF can simulate local weather dynamics, including wind, temperature, and humidity at a fine resolution.

2.4. Selecting the Right Weather Data for Building Thermal Performance Simulations

For building thermal performance simulation, it is necessary to resolve a series of highly non-linear equations that govern energy and mass transfer relationships, as buildings are rather complex dynamic systems. Depending on the simulation objectives, different weather data are required. Related weather data issues for building simulation were reviewed by Hensen [15] and Crawley [79]. To select the right weather data for thermal simulations of buildings, we should understand the characteristics of weather data.

Table 2. Some common weather data used for thermal simulations.

Use	Description	Weather Data Required	Note
Compliance analysis of energy performance of fully conditioned buildings	Energy savings (generally annual) and energy code compliance (i.e., NatHERS star rating) are derived by comparing results for design variants with considering thermal comfort.	Representative single-year hourly data are sufficient. See Table 1 for some examples.	For this application, the need for absolute accuracy is minimized and short-time inaccuracies are not significant.
Performance of un- or semi- conditioned buildings	To investigate performance under non-average conditions, such as overheating analysis and floating space temperature problems for solar passive design.	Extreme weather data and/or multi-year data are needed	Typical data are often not adequate.
Model calibration, actual energy savings estimation and building troubleshooting	Applications involve the performance of existing buildings during a specific period at a particular site.	Weather data observed during the study period at or near the building site are required.	Historical data, such as TMY and RMY, are generally of little use.
Equipment sizing	Peak heating and cooling loads of HVAC system need be determined by designers	Weather data of design-day or near-extreme conditions are needed.	See Section 2.2.2 for some examples of the weather data.
Natural ventilation design	Natural ventilation design hinges on climate-responsive strategies, airflow modeling, and integration with passive systems.	Local wind data are needed. UHI impact should also be considered.	Data from airport stations are unreliable for nearby sites
Renewable energy systems	To determine output of solar PV and wind turbine under short-term weather data (solar radiation, wind speed and direction).	Short-term (less than hour) weather data (incident radiation, wind velocity) are required.	Standard hourly data may produce unreliable results for solar PV and wind turbine systems as they have non-linear characteristics.

describes some common weather data used for building simulations.

As boundary conditions, appropriate weather data must be selected carefully for the processes being simulated. Using inappropriate data can lead to significant modelling errors. However, in some cases, simplified or representative data may be sufficient, and pursuing more detailed or ‘accurate’ data offers little added value.

3. Conclusions

This review has presented a concise review of major methods for developing typical and extreme weather files for building performance simulations under current and future changing climate conditions. While the field has made significant progress, particularly in the standardization of generation of typical years (e.g., TMY, TRY, RMY), several challenges and knowledge gaps remain. The key insights and critical reflections are summarized below.

Typical Weather Files are well-established for use in assessing average building performance. Approaches such as the Finkelstein–Schafer (FS) method have been widely used to construct representative composite years. However, these files inherently mask atypical and extreme weather conditions, limiting their utility for stress testing or resilience studies. There is a lack of guidance on how these files should evolve under a changing climate, especially as historical stationarity becomes less reliable.

Extreme Weather Files are increasingly being used to evaluate thermal comfort, overheating risk, and peak energy demand. While methods like DSY, HSY1/HSY2, and XMY offer structured approaches to capturing hot summers or cold winters, there is no universal standard for defining or generating such files. This fragmentation limits comparability between studies and jurisdictions. Additionally, most methods focus on temperature extremes, often neglecting combined extremes (e.g., temperature and humidity) or compound events (e.g., heatwaves plus power outages). Heatwave Resilience Files represent a promising direction, incorporating duration, severity, and intensity dimensions. However, these remain largely experimental and have not been systematically validated against observed health or building performance outcomes. Integration with urban microclimate and vulnerability data would greatly enhance their applicability. Future Weather Files based on morphing, stochastic generators, and downscaled climate models each bring unique strengths. Morphing is efficient but oversimplifies future variability. Stochastic generators offer better temporal realism but rely on assumptions about future variability that may not hold. Downscaling GCM/RCM models provides physically grounded projections, especially for extremes, but at high computational and technical costs. Moreover, validation of these future files remains limited. Most studies evaluate only mean monthly temperatures, with fewer addressing variability, extremes, or inter-annual trends. Systematic validation against observed extremes and independent climate datasets (e.g., reanalyses) is urgently needed. A critical gap is the lack of unified frameworks to select or generate appropriate weather files for a given simulation objective. Inappropriate weather input can lead to serious modelling errors; however, there is little consensus or guidance for practitioners on when to use a typical year, extreme year, or future scenario. Similarly, the interplay between urban heat island effects and future climate extremes is underexplored in current file generation approaches.

• Recommendations for Future Work

To strengthen the robustness and applicability of weather files in building simulation, future research should focus on the following areas:

• Standardizing methods for extreme weather file construction and expanding their scope to include compound events and urban microclimate effects;
• Improving validation practices for both present and future weather files, including comparison with observed extremes and independent datasets;
• Integrating weather file generation with evolving climate science, especially regarding representation of uncertainty and climate model ensembles.

In conclusion, while the field has matured in generating typical and extreme weather data for current conditions, future development of weather files still requires more rigorous validation, standardization, and integration with resilience and adaptation frameworks. Addressing these gaps will be critical to ensuring reliable, future-proof building simulations in a changing climate.