Effectiveness of CCWorldWeatherGen weather file generation tool

Eficácia da ferramenta de progressão de arquivos climáticos CCWorldWeatherGen

Abstract

This paper aimed to analyze the effectiveness of the CCWorldWeatherGen tool, focusing on climate change in São Paulo, São Paulo State, Brazil. For this, dry-bulb temperature, relative humidity, global solar radiation, and wind speed data from the test reference year weather file (1954) and the CCWorldWeatherGen file for the 2020 period (representing the 2011-2040 period) were compared with observational data collected between 2011 and 2023 by the Meteorological Station of the Institute of Astronomy, Geophysics, and Atmospheric Sciences of the University of São Paulo. The accuracy of variables predicted using weather files was evaluated using five statistical measures of error. Annual relative root mean square error (RRMSE) values for dry-bulb temperature, relative humidity, global solar radiation, and wind speed in the morphed weather file were 17.04% (good), 17.95% (good), 31.57% (poor), and 224.44% (poor), respectively. It is concluded that CCWorldWeatherGen is suitable for generating future weather files with complete information, mainly for its practicality. However, this approach requires caution, as sequences depend on the consistency of the weather file used as a basis.

Keywords: Building simulation; Climate change; Future weather file; Meteorological data analysis; Resilient building design.

Resumo

O objetivo deste artigo foi analisar a eficácia da ferramenta CCWorldWeatherGen, considerando as mudanças climáticas na cidade de São Paulo. Para tanto, as variáveis temperatura do ar (bulbo seco), umidade relativa do ar, radiação solar global, e velocidade do vento do arquivo climático de referência, de 1954, e do arquivo climático gerado pelo CCWorldWeatherGen para 2020 (representa 2011 a 2040), foram comparadas com observações da Estação Meteorológica do Instituto de Astronomia, Geofísica e Ciências Atmosféricas da Universidade de São Paulo, obtidas entre 2011 e 2023. A acurácia das variáveis dos arquivos climáticos foi avaliada por meio de cinco indicadores de erro. Os resultados da raiz do erro quadrático médio relativo (RRMSE) anual para a temperatura de bulbo seco do ar, umidade relativa do ar, radiação solar horizontal global, e velocidade do vento foram 17,04% (bom), 17,95% (bom), 31,57% (ruim) e 224,44% (ruim), respectivamente. A ferramenta CCWorldWeatherGen é adequada, especialmente por sua praticidade, para gerar futuros arquivos climáticos com informações completas. Mas seu uso demanda cautela porque as progressões dependem da consistência dos dados do arquivo climático utilizado como base.

Palavras-chave:  Simulação de edificações; Mudanças climáticas; Arquivo climático futuro; Análise de dados meteorológicos; Projeto de edificações resilientes.

Introduction

Climate change is one of the greatest challenges faced by society in the twenty-first century, exerting direct effects on the environment, the economy, and human health (Farah et al., 2019; IPCC, 2023). In view of this, building performance becomes increasingly important, necessitating the design of sustainable and climate-resilient buildings. A valuable approach in this regard is to conduct performance simulations taking into account future climate change scenarios. Thus, it is possible to ensure the thermal comfort of users and reduce energy consumption and pollutant emissions, making buildings important elements in the fight against climate change (Farah et al., 2019; De Wilde; Coley, 2012).

Berardi and Jafarpur (2020) and Guan (2009) stated that there are different methods to obtain future climate data for building performance evaluation. According to Farah et al. (2019) and Guan (2009), the imposed offset method - also known as morphing, according to Belcher, Hacker, and Powell (2005) - is the most practical and reliable approach. A valuable strategy involves the use of the Climate Change World Weather File Generator (CCWorldWeatherGen) tool, which performs morphing (Belcher; Hacker; Powell, 2005), to process EnergyPlus Weather (EPW) files, generating complete weather sequences. Other tools that can be used to generate future weather files include Future Weather Generator, Meteonorm, and Weather Shift (Rodrigues; Fernandes; Carvalho, 2023; Troup; Fannon, 2016; Yassaghi; Mostafavi; Hoque, 2019). However, these tools are less commonly used in the literature, possibly because of limitations associated with the accessibility of local data, the requirement for a private license, or their more recent development.

CCWorldWeatherGen is well accepted by the scientific community, having been applied in simulation studies predicting future climatic conditions in several countries worldwide (Ashrafian, 2023; De Wilde; Coley, 2012; Gonçalves et al., 2024; Kutty et al., 2024; Triana; Lamberts; Sassi, 2018; Wang; Liu; Brown, 2017). The tool, which is free and works as a Microsoft Excel extension, was designed to enable the generation of future weather files for three representative time periods: 2020 (representing the 2011-2040 period), 2050 (representing the 2041-2070 period), and 2080 (representing the 2071-2100 period) (Jentsch; Bahaj; James, 2012). The technique was validated by Jentsch et al. (2013) as a practical strategy to assess climate change in the built environment.

However, it should be noted that generating an adequate representation of the long-term patterns of climate change has proven to be a challenging task (Machado, 2019), mostly because of the unpredictability of weather conditions (Torres, 2014; Santos et al., 2015). Validation studies for tools such as CCWorldWeatherGen are crucial for gaining a better understanding of the reliability of weather sequences and, consequently, better estimating the thermal and energy performance of future buildings. Consolidating our understanding of errors associated with CCWorldWeatherGen predictions is also important to improve the accuracy of weather sequences and enable designers and researchers to more efficiently assess the risks and opportunities associated with climate change.

According to Fernandes and Godinho Filho (2010), forecast errors can stem from two sources. The first type of error is inevitable and should be ignored because it refers to the randomness of the model itself. The second type of error is related to the prediction method and model parameters. When performing error analysis, statistical metrics can be used to assess the discrepancy between observed and predicted values.

This paper aimed to analyze the effectiveness of the CCWorldWeatherGen tool in generating future weather files based on meteorological observations in São Paulo, São Paulo State, Brazil, addressing part of theNunes (2021) research. It investigates four variables of great influence on thermoenergetic simulations of building parameters, namely dry-bulb temperature, relative humidity, global solar radiation, and wind speed. The first two variables are considered the most significant in the context of climate change (Alves, 2014). In view of the relevance of this global problem to building performance, the article attempts to contribute to the assessment of the effectiveness and applicability of CCWorldWeatherGen by comparing projections for the city of São Paulo in the 2020 period with meteorological data collected from 2011 to 2023 by the Meteorological Station at the Institute of Astronomy, Geophysics, and Atmospheric Sciences, University of São Paulo (EM-IAG-USP). Although a more succinct analysis can be performed with a reduced number of indicators, this article evaluated five statistical measures of error, with the aim of providing a more comprehensive and substantial understanding of the accuracy of predicted data as compared with observed values.

Theoretical background

The literature related to the simulation of building performance under future climate scenarios is mainly rooted in the analysis of thermal comfort and energy efficiency. Many concepts, models, and technologies introduced to this field can be used for the development of design criteria and strategies aimed at adapting to future climates and constructing resilient buildings (Gonçalves et al., 2024). However, the methods and tools for generating progressions of weather files, such as CCWorldWeatherGen, necessitate additional research efforts to efficiently generatesimulation parameters for building assessment (thermal comfort and energy efficiency) under future climate conditions. Understanding the current state of the art and describing and applying the CCWorldWeatherGen tool form the foundation for conducting this research.

An overview of the state of the art

Over the past few years, research on the impacts of climate change on buildings and the urban environment has evolved from a debate about resilience, primarily through computer simulations (Siu et al., 2023; Tootkaboni, 2021; Nielsen, Kolarick, 2021). There are many definitions of resilience in the built environment. It can be summarized as referring to the ability of buildings to resist future climate events while maintaining adequate performance (Su; Chang; Pai, 2022).

Building simulation tools typically require weather files as input data. Test reference year (TRY) or typical meteorological year (TMY) files representing typical past meteorological observations are commonly used for this purpose. However, there is a growing effort in the scientific community to obtain future climate data better suited for building performance simulations.

Multiple methods have been used to integrate the impacts of climate change into files for computational simulations (Berardi; Jafarpur, 2020; Guan, 2009). Two main approaches are typically adopted to create future climate files for building performance simulation. The first approach relies on historical meteorological data and includes statistical extrapolation, imposed compensation, and stochastic climate modeling. The second approach depends on using climate models as an alternative to historical meteorological data. These approaches are summarized in Table 1.

Bewley and Berzolla (2023) described three commonly used future weather file generator tools, namely CCWorldWeatherGen, Meteonorm, and WeatherShift. Each of these tools uses a slightly different process to project local climate change from GCM outputs. CCWorldWeatherGen uses a high-emission scenario from the IPCC to project future climate change based on the Hadley Centre Coupled Model version 3 (HadCM3). The validity of CCWorldWeatherGen's morphed data depends on how well the HadCM3 model represents the climate in that region. Meteonorm uses a stochastic weather generator that interpolates between GCM outputs using a historical climate database and includes different IPCC scenarios (ranging from high to low emission due to reductions in material intensity and introduction of clean and resource-efficient technologies). WeatherShift employs 14 GCM and performs bilinear interpolation to generate morphed output variables, considering IPCC pathways from high to low emissions.

The review conducted by Nielsen and Kolarik (2021), which evaluates 47 studies that applied future weather data in building performance simulation in 164 locations, stated that the morphing downscaling method associated with the CCWorldWeatherGen tool stood out in applications described in the literature. Another study also highlighted the widespread adoption of the morphing approach and CCWorldWeatherGen in the scientific community, given that building performance researchers tend to select the simplest-to-use weather generator (Campagna; Fiorito, 2022; Nielsen; Kolarik, 2021; Rodrigues; Fernandes; Carvalho, 2023; Tootkaboni et al., 2021). Despite their simplicity, common techniques manage to provide adequate future weather data for building simulation, facilitating the application of diverse climate change scenarios (Tootkaboni et al., 2021).

Although various research approaches have been adopted to obtain future climate data, generate weather file progression techniques, and develop best practices for simulating buildings in future scenarios, special attention should be given to the validity of climate variables. These variables play a decisive role in determining model outputs and influencing building performance indicators, being critical for addressing long-term objectives amidst evolving climate conditions (Berardi; Jafarpur, 2020). There is a need to rethink and critically evaluate established practices and concepts in light of the challenges posed by climate change and its inherent uncertainties.

Table 1
Approaches to generating future climate data

In the literature, there are few studies examining the effectiveness of future weather file tools in building performance analysis. Most available studies focused on comparing future weather file generator tools or future and historic data from the same climate model. Tootkaboni et al. (2021) compared 2050 period weather files generated using CCWorldWeatherGen, Meteonorm, and WeatherShift, a future TMY weather file created using a high-quality climate model database, and an actual weather file for Rome, Italy. All generated future data showed similar mean values higher than the actual weather file. However, the tools provided similar predictions of future energy performance and comfort parameters for buildings, whereas the high-quality climate model database produced different patterns with high variations.

Escandón et al. (2023) compared the efficiency of CCWorldWeatherGen and Meteonorm for generating future climate datasets for the 2050 and 2080 periods using two different scale simulation models, namely a test cell and a multi-family building located in southern Spain. The results indicated that the choice of projection method for generating future climatic data has significant effects on the analysis of thermal comfort and energy demand. However, these effects are notably reduced when an annual evaluation is conducted.

Gonçalves et al. (2024) conducted a brief comparative analysis of climate variables in Belém, Pará, Brazil, using TRY data (corresponding to the year 1964) and Brazilian National Institute of Meteorology (INMET) weather files (corresponding to the year 2010), alongside morphed files for the representative years of 2020 and 2050. The variables included dry-bulb temperature, relative humidity, global solar radiation, and wind speed, depicting the gradual growth of the urban heat island phenomenon in Belém. The findings demonstrated the reliability of the CCWorldWeatherGen tool.

The use of different methods and tools to obtain future climate files has been widely discussed. However, what is less clear in the literature is the degree to which morphed weather files resemble real monitored data and the accuracy of predicted climate variables. To fill this gap, we propose an approach that combines CCWorldWeatherGen climate prediction and climate measurements corresponding to the 2020 period addressed by the tool.

CCWorldWeatherGen tool

The morphing method proposed by Belcher, Hacker, and Powell (2005) was applied by Jentsch, Bahaj, and James (2012) in the development of CCWorldWeatherGen. The tool provides a viable and advantageous method for acquiring comprehensive, consolidated, and scientifically accepted weather information through freely available software (Tootkaboni et al., 2021). It should be noted, however, that the tool does not consider the latest IPCC Assessment Report (AR6). Nevertheless, unlike other techniques, CCWorldWeatherGen algorithms provide projections for all variables included in the reference weather file. Additionally, the tool uses equation routines developed by specialized sources for each variable.

The morphing method has the advantage of respecting the original prediction of typical years (average year over 30 years of data). The following algorithms are generated by the program from three generic operations (shift, linear stretch, and the combination of the two), as explained by Belcher, Hacker, and Powell (2005) and Casagrande and Alvarez (2013).

A shift is applied to the reference climate variable for each month m, as shown in Equation 1:

$x=x_0+{∆x}_m$ Eq. 1

Where:

x is the future climate variable;

x₀ is the reference climate variable; and

Δx_m is the absolute monthly anomaly, as estimated by the projection model.

A stretch is applied to the reference climate variable by scaling the projected monthly mean variance, according to Equation 2:

$\mathit{x}={\mathit{x}}_0{\alpha }_m$ Eq. 2

Where:

x is the future climate variable;

x₀ is the reference climate variable; and

α_m is the fractional change in the monthly mean value, as estimated by the projection model.

A combination of shift and stretch is applied to the reference climate variable by using Equation 3. Thus, the reference variable is shifted by the projected monthly mean anomaly and stretched to the monthly variance of the observed variable.

$\mathit{x}={\mathit{x}}_0+{∆x}_m+{\alpha }_m\left[x_0-{\left(x_0\right)}_m\right]$ Eq. 3

Where:

x is the future climate variable;

x₀ is the reference climate variable;

Δx_m is the absolute monthly anomaly, as estimated by the projection model;

α_m is the fractional change in the monthly mean value, as estimated by the projection model; and

(x₀)_m is the monthly mean value of x ₀.

This is the base algorithm of the morphing method used by CCWorldWeatherGen. Thus, the tool morphs variables from weather files, such as dry-bulb temperature, relative humidity, solar radiation, and ground temperature, among others. Of the 27 climate variables contained in EPW files and covered by CCWorldWeatherGen, only 8 - of little relevance to building performance simulations - are not morphed by the progression routines. More details on calculation routines, as well as equations and sequence definitions, can be found in the technical reference manual of the tool (Jentsch, 2012).

Methods

This research compared data from weather files with measured data and evaluated five error indices, namely mean bias error (MBE), mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and relative root mean square error (RRMSE). Weather data from the TRY file for the city of São Paulo, Brazil, were morphed to the current period (2020, 2011-2040) using the CCWorldWeatherGen tool and compared with data collected from 2011 to 2023 by EM-IAG-USP (2023).

Data analysis, including the statistical evaluation of accuracy (errors) and the preparation of graphs and tables, was performed using Microsoft Excel, according to the procedures described throughout this section. The research method is divided into two subsections. The first subsection presents data collection procedures, including the generation of the weather file in CCWorldWeatherGen and the acquisition of weather station data. The second subsection describes the procedures performed to calculate the accuracy of the predicted weather data.

Weather data

The following variables were used to assess the effectiveness and accuracy of the CCWorldWeatherGen tool: air temperature (dry-bulb), relative humidity, global solar radiation, and wind speed. These variables were chosen because they hold the greatest potential to contribute to building performance simulation among variables contained in the weather files and measured at EM-IAG-USP.

The dataset included variables extracted from the São Paulo weather files from TRY (corresponding to the year 1954) and the 2020 period (representing from 2011 to 2040) under the IPCC scenario A2 and meteorological measurements taken in São Paulo between January 1, 2011, and June 30, 2023, by EM-IAG-USP. The weather file containing projections for the 2020 period was generated by CCWorldWeatherGen based on data from the TRY file for São Paulo. The TRY weather file, widely used in computer simulations, was developed by Goulart, Lamberts, and Firmino (1998) and reviewed by Carlo and Lamberts (2005). It is available from the United States Department of Energy on the EnergyPlus website (U.S. DOE, 2020). It is important to note that, although the reference year of the São Paulo file is outside the recommended range (1961 to 1990), the year is close to and, more importantly, not later than the specified range, which could result in overestimation of sequences, according to the developers of the CCWorldWeatherGen tool (Jentsch; Bahaj; James, 2012; Jentsch et al., 2013).

Weather data recorded in São Paulo city were obtained from EM-IAG-USP (2023), which works independently of this research. This meteorological station carries out important weather observations and provides various data (atmospheric pressure; dry-bulb, wet-bulb, dew point, and ground temperatures; relative and specific humidity; precipitation; solar radiation and brightness; cloud information; wind speed and direction; gust wind speed; water evaporation; and the occurrence of drizzle rain, dew, fog, hail, and frost) upon request through an electronic form.Figure 1 shows the location of EM-IAG-USP within the metropolis, and Figure 2provides a zoomed view of the station.

EM-IAG-USP, registered with the World Meteorological Organization (WMO) under number 83004, is located in the Science and Technology Park (CIENTEC) of the University of São Paulo at Fontes do Ipiranga State Park (PEFI), São Paulo, São Paulo State, Brazil (23°39′S 6°37′W, 799.20 m a.s.l.). As the station is situated in a green area, urbanization effects are slightly reduced (masked) in the said locality. Cities act as amplifiers of climate change effects. Therefore, it is assumed that warming effects are more intense in urban areas, where vegetation is typically reduced, as explained by Alves (2014).

Figure 1
Map depicting the location of the Meteorological Station at the Institute of Astronomy, Geophysics, and Atmospheric Sciences in the city of São Paulo

Figure 2
Zoomed view of the Meteorological Station at the Institute of Astronomy, Geophysics, and Atmospheric Sciences

According to the EM-IAG-USP Technical Report (USP, 2010), the procedures and instruments used in the station follow WMO recommendations (WMO, 1994, 1996). The document also describes that, in addition to observations made by meteorological observers, automatic mechanical instruments (anemographs, barographs, thermographs, hygrographs, pluviographs, and actinographs) are used to record weather data 24 h a day. Observers analyze this information in the form of diagrams on the day following the measurements (USP, 2010).

As for data collection on air temperature (dry-bulb), relative humidity, global solar radiation, and wind speed, it is important to emphasize the following points regarding instruments, procedures, and uncertainty requirements (USP, 2010; WMO, 1994, 1996):

1. air temperature (dry-bulb) with 0.1 K (for > −40 °C and ≤ +40 °C) and 0.3 K (for > +40 °C) uncertainty is measured hourly from 7:00 to 24:00 by visual reading of mercury capillaries on an Assmann-type aspiration psychrometer (electric motor R. Fuess). Hourly estimates are generated between 1:00 and 6:00 (during which no direct readings are taken) based on measurements recorded by a bimetallic ring thermograph with daily rotation (R. Fuess, model 79, no. F-2243);

2. relative humidity with 1% uncertainty is calculated hourly from 7:00 to 24:00 from dry- and wet-bulb temperature readings obtained using the above-mentioned psychrometer and atmospheric pressure readings taken on a fixed-cistern mercury barometer (Kew R. Fuess no. 1010). Hourly estimates are generated between 1:00 and 6:00 (during which no direct readings are taken) based on measurements recorded by a hair harp hygrograph with daily rotation (R. Fuess);

3. daily mean global solar radiation with 0.4 MJ/m² (for ≤8 MJ/m²) and 5% (for >8 MJ/m²) uncertainty is determined by integration of instantaneous global solar irradiance values for the daytime period, measured by a bimetallic Robitzsch 58d actinograph with daily rotation installed at the top of the station tower. Values are calculated from the area obtained between sunrise and sunset. The area is calculated by mechanical planimetry, performed several times by different technicians. In subsequent analyses, verifications are performed by comparing the measurements to those that would be made by an instrument sensitive to a wider region of the solar spectrum (with greater precision), that is, a pyranometer equipped with Schott wg295 domes. The quotient between the daily global solar radiation observed near the ground and that which would be observed if the instrument were installed above the Earth's atmosphere is estimated; and

4. wind speed with 0.5 m/s (for ≤5 m/s) and 10% (for >5 m/s) uncertainty is measured using a Universal 82a anemograph with daily rotation (R. Fuess). Instantaneous wind speed is observed, and the average speed of each hourly interval is estimated from the line associated with the displacement of a portion of air passing through the weather vane.

Given that EM-IAG-USP provides daily global solar radiation data in MJ/m²/day, the values needed to be converted to W/m² for comparison with data extracted from weather files. Therefore, as 1 W equals 1 J/s, that is, 0.0864 MJ/day, global solar radiation in MJ/m²/day was divided by 0.0864 to obtain values expressed in W/m². Similarly, wind speed data from EM-IAG-USP are expressed in km/h, which, to be converted to m/s (measurement unit of EPW weather files), were divided by a conversion factor of 3.6.

More details about data measurements can be found in bulletins and technical reports available on the USP (2024) website.

Statistical evaluation of the accuracy of morphed weather files

To determine the error between weather files and EM-IAG-USP data, we applied commonly used statistical parameters (Chou; Bui, 2014; Despotovic et al., 2016; Leung et al., 2012; Li et al., 2013; Moreno et al., 2013). Five statistical quantitative indicators were used to evaluate the accuracy of measured and predicted variables, as follows:

1. mean bias error (MBE): indicates the average deviation of predicted data (weather file) from the corresponding measured data (meteorological station). MBE reflects the tendency toward overestimation or underestimation of a given parameter, whereby negative values indicate overestimation, and positive values indicate underestimation. Therefore, values close to zero are desirable. Caution is necessary when analyzing MBE values alone, because positive and negative errors can offset each other;

2. mean absolute error (MAE): indicates the mean of absolute differences between predicted and measured values. It was used to measure the degree of proximity between weather file and meteorological station values. MAE represents the arithmetic mean of absolute errors, accounting only for their magnitude, regardless of direction. Lower MAE values indicate higher accuracy;

3. root mean square error (RMSE): quantifies differences between predicted and actual values. Although it is similar to MAE in terms of the output result, the RMSE does not process errors in the same manner; it penalizes the most significant errors. Thus, a single large error is sufficient to produce a poor result. The RMSE is frequently used in scientific studies because it provides a clear understanding of the analyzed data. Lower values indicate better predictive accuracy;

4. mean absolute percentage error (MAPE): quantifies accuracy as a percentage of prediction error. Smaller MAPE values indicate smaller errors. For example, if the calculated MAPE is 10, the prediction is, on average, 10% wrong. Although it is a practical parameter, MAPE divides each error individually, which makes it asymmetrical. High errors have a significant impact on MAPE. Because of this, individual analysis of MAPE can result in a superficial understanding of data accuracy. According to Lewis (1982) and Moreno et al. (2013), accuracy is categorized as high if MAPE < 10%, good if 10% < MAPE < 20%, reasonable if 20% < MAPE < 50%, and inaccurate if MAPE > 50%; and

5. relative root mean square error (RRMSE) is a variant of the RMSE. The variable measures the accuracy of predicted values in relation to the average value of measured data. It normalizes RMSE and is given as a percentage to facilitate comparison between datasets or between variables, with lower values indicating better accuracy. Given its facilitated comparison and calculation rigor, RRMSE was adopted as the conclusive accuracy index in this study. Nevertheless, we underscore that it is important to analyze all calculated indices together. According to Despotovic et al. (2016), Jamieson, Porter, and Wilson (1991), and Li et al. (2013), accuracy is considered excellent if RRMSE < 10%, good if 10% < RRMSE < 20%, fair if 20% < RRMSE < 30%, and poor if RRMSE > 30%.

The indicators MBE, used to understand the amount and direction of the error, MAE and RMSE, used to measure the magnitude of the error in absolute terms, and MAPE and RRMSE, used to define the relative amount (percentage) of the error and classify the accuracy, were calculated according to data on air temperature, relative humidity, and wind speed for all 8,760 h of the year, as well as global solar radiation, obtained for all 365 days of the year (the weather station provides daily global solar radiation values). Calculations were made according to Equations 4 to 8, as follows:

$\mathrm{MBE}=\frac{1}{\mathrm{n}}\times \sum _{\mathrm{i}}^{=}\left({\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right)$ Eq. 4

$\mathrm{MAE}=\frac{1}{\mathrm{n}}\times \sum _{\mathrm{i}}^{=}\left|{\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right|$ Eq. 5

$\mathrm{RMSE}=\sqrt{\frac{1}{\mathrm{n}}\sum _{\mathrm{i}}^{=}{\left({\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right)}^2}$ Eq. 6

$\mathrm{MAPE}=\frac{1}{\mathrm{n}}\times \sum _{\mathrm{i}}^{=}\frac{\left|{\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right|}{{\mathrm{o}}_{\mathrm{i}}}\times 100$ Eq. 7

$\mathrm{RRMSE}=\frac{\sqrt{\frac{1}{\mathrm{n}}\sum _{\mathrm{i}}^{=}{\left({\mathrm{o}}_{\mathrm{i}}-{\mathrm{p}}_{\mathrm{i}}\right)}^2}}{\frac{1}{\mathrm{n}}\sum _{\mathrm{i}}^{=}{\mathrm{o}}_{\mathrm{i}}}\times 100$Eq. 8

Where:

MBE is the mean bias error (°C, %, W/m², or m/s, depending on the climate variable);

MAE is the mean absolute error (°C, %, W/m², or m/s, depending on the climate variable);

RMSE is the root mean square error (°C, %, W/m², or m/s, depending on the climate variable);

MAPE is the mean absolute percentage error (%);

RRMSE is the relative root mean square error (%);

n is the total number of observations;

li is each individual observation;

o is the actual measured value of each observation; and

p is the predicted value of each observation.

Results and discussion

This section presents the results and discussions about the effectiveness of the CCWorldWeatherGen tool. Figures 3 to 6 illustrate the air temperature (dry-bulb), relative humidity, global solar radiation, and wind speed data for São Paulo, in monthly averages, daily averages, hourly averages, and monthly hourly averages for the entire year. The different lines plotted on the graphs represent data from the TRY file (1954), CCWorldWeatherGen sequences for the 2020 period (2011 to 2040), and EM-IAG-USP from 2011 to 2023 (USP, 2023).

Tables 2 to 5 list the errors calculated for each month and the entire year, according to the indicators MBE, MAE, RMSE, MAPE, and RRMSE. It should be noted that, for MBE values, the number indicates the amount and the sign shows the deviation side. A negative sign indicates overestimation of predicted values (data from EPW files) in relation to present-day measured values (data from the weather station); a positive sign indicates underestimation. For MAE, RMSE, MAPE, and RRMSE, larger values and darker tones indicate larger errors.

Dry-bulb temperature

Figure 3 shows the monthly, daily, hourly, and monthly hourly averages of air temperature (dry-bulb) in São Paulo, and Table 2 shows their respective monthly and annual error indices.

Figure 3
Dry-bulb temperature in São Paulo, São Paulo State, Brazil, according to weather file and meteorological station data

Table 2
Monthly and annual errors of dry-bulb temperature sequences from weather files in relation to weather station data

Air temperature data clearly revealed a recent increase in temperature and a progressive rise in temperature in the future scenario. In relation to TRY data, the increase in average annual air temperature was estimated at 0.91 °C for the 2020 period. Furthermore, it was estimated that the minimum temperature will rise from 16.05 to 17.03 °C by 2040, whereas the maximum will rise from 22.47 to 23.27 °C. Although moderate, the increase in minimum air temperature was larger than the increase in maximum air temperature, which is consistent with some climate change scenarios reported in the literature (Cubasch et al., 2001; Hatfield; Prueger, 2015; Jones et al., 1999; Knowles; Dettinger; Cayan, 2006). Sequences for the 2020 period (current period) showed remarkable fidelity with observed data. The smallest monthly average difference (+0.06 °C) between morphed 2020 data and the observed air temperature occurred in March, and the largest difference (+1.46 °C) occurred in December. In warmer months, temperatures obtained by morphing were more similar to real observations. In March, April, October, November, and December, the temperatures recorded by the weather station were higher than those of the 2020 weather file. On the other hand, during the colder months, the recorded temperatures were lower than those of the 2020 file. There was greater fluctuation in the daily and hourly averages of temperature from weather files than in recorded temperatures. The weather station is located in a wooded area, which may explain the low variation in observed temperatures, as described in some studies demonstrating a positive effect of afforestation on microclimate (Jones, 2021; Martelli; Santos Junior, 2015; Weirich et al., 2015).

In general, compared with observed weather data, the graphs indicate that morphed air temperature data were more representative than data from the TRY weather file. This finding is corroborated by the annual values of all calculated error indices. MBE, calculated for all hours of the year, was 0.83 °C (TRY) and −0.08 °C (2020 period), revealing underestimation of TRY file data and a small overestimation of the morphed 2020 data. MAE and RMSE indices showed that the annual absolute errors were 2.65 °C and 3.42 °C (TRY) and 2.58 °C and 3.35 °C (2020 period), respectively. MAPE and RRMSE results showed annual error percentages of 13.50% and 17.40% (TRY) and 13.32% and 17.04% (2020 period), respectively. Moreover, in individually analyzing the hours of the months with the highest (February) and lowest (July) temperatures, it was found that the RRMSE (adopted as the most conclusive parameter in this study) was 12.11% for TRY data and 11.80% for 2020 data in February and 20.03% for TRY data and 20.89% for 2020 data in July. According to MAPE values, the accuracies were classified as good or high, and, according to RRMSE values, the accuracies were predominantly good.

Relative humidity

Figure 4 shows the monthly, daily, hourly, and monthly hourly averages of relative humidity in São Paulo, and Table 3 shows their respective monthly and annual error indices.

The 2020 weather file revealed a reduction in relative humidity. As shown by the projections, until 2040, the maximum humidity will be 87.22% (May) and the minimum 69.82% (August). Thus, it is estimated that the annual average will decrease from 82.17% (TRY) to 81.33% (2020 period) by 2040. It should be noted, however, that the observed data (2011-2023) had an annual average of 79.76%; in several months, relative humidity was well below the values forecast for the 2020 period, as also indicated by MBE.

Compared with observed weather data, as occurred for air temperature, the relative humidity of the 2020 file had smaller errors than that of the TRY file, according to the five calculated indicators. MBE values showed that data were overestimated in almost all months and in the cumulative year period. The magnitude of MAE and RMSE annual errors for the 2020 weather file was 10.59% and 14.32%, respectively, whereas that of the TRY weather file was 10.79% and 14.35%, respectively. Similarly, MAPE and RRMSE values were 14.47% and 17.95%, respectively, for the 2020 file and 14.73% and 17.99%, respectively, for the TRY file. All parameters were classified as having good accuracy. Of note, milder periods had greater errors (highlighted in dark yellow and dark red in Table 3). The environmental conditions (afforestation) near the weather station may explain the higher observed relative humidity, even in the driest months of the year. The microclimate is influenced by vegetation, which contributes to maintaining humidity, as reported by some studies on the topic (Martelli; Santos Junior, 2015; Kuang, 2020; Weirich et al., 2015). Additionally, it should be noted that no large buildings have been built in the vicinity of EM-IAG-USP (radius of 500-1000 m) in the last 85 years; existing trees have grown, and new trees have been planted throughout the PEFI Park and especially in the CIENTEC Park (USP, 2017).

Solar radiation

Figure 5 shows the global solar radiation values for São Paulo, and Table 4 describes their respective monthly and annual error indices. Given that EM-IAG-USP only provides global solar radiation in daily averages, it was not possible to compile graphs for hourly averages and monthly hourly averages.

Figure 4
Relative humidity in São Paulo, São Paulo State, Brazil, according to weather file and meteorological station data

Table 3
Monthly and annual errors of relative humidity sequences from weather files in relation to weather station data

Figure 5
Solar radiation for São Paulo, São Paulo State, Brazil, according to weather file and meteorological station data

Table 4
Monthly and annual errors of solar radiation sequences from weather files in relation to weather station data

The annual averages of global solar radiation from weather files were 195.94 W/m² (TRY) and 212.04 W/m² (2020 period), whereas the observed annual average (2011 to 2023) was 190.26 W/m². A plausible explanation for the relative differences in this parameter is that hourly solar radiation data from the TRY file have important limitations; they were estimated from hourly cloudiness, which affects accuracy (Carlo; Lamberts, 2005; Goulart; Lamberts; Firmino, 1998).

According to weather sequences, the maximum monthly average radiation will increase from 271.93 (December, TRY) to 276.98 W/m² (December, 2020 period) by 2040, and the minimum will increase from 119.85 (June, TRY) to 157.34 W/m² (July, 2020 period). In comparing observed data (2011 to 2023) and weather file data, we identified differences between the 2020 file (current) and the TRY file (1954). In some months, the observed solar radiation was below the values indicated by weather files. Such differences may be explained by climate variations and the location of the weather station, as well as by the equipment used at the station (actinograph). Guermoui et al. (2022) tackled the challenge of forecasting global solar radiation. The authors observed that the distribution of hourly solar radiation at different hours in different locations varies considerably. Regarding equipment, whereas some pyranometers are classified as first- (2% uncertainty) or second-class (5% uncertainty) instruments, actinographs are considered third-class instruments (15% to 20% uncertainty), having low accuracy, according to WMO (CRESESB, 2008). Brock and Nicolaidis (1984), Dallacort et al. (2004), and Malm and Walther (1980) claimed that an actinograph can generate deviations of up to 19% in the measured information. Although the station's technical report (USP, 2016) states that actinograph data undergo treatment to obtain an estimate of the value measured by a pyranometer, inaccuracies in hourly measurements can occur and cause inconsistencies. Another hypothesis for the calculated errors is an overestimation of the solar radiation predicted by the climate model (Rivington et al., 2008; Ma; Wang; Wild, 2015). Ma, Wang and Wild (2015) evaluated surface incident solar radiation from 48 Earth system models that compose the Coupled Model Intercomparison Project (CMIP5), including HadCM3 used in CCWorldWeatherGen. The authors reported an overestimation of approximately 11.9 W/m² based on information from the Global Energy Balance Archive (GEBA) database.

According to the MBE, in some months, values were underestimated in the TRY weather file, whereas, in other months, data were overestimated. In the 2020 file, solar radiation was overestimated in all months, except for September, and in the cumulative annual value. The magnitude of the error of morphed data was 48.18 W/m² (MAE) and 60.00 W/m² (RMSE) for the whole year, and errors were higher than those of the TRY file. The MAPE and RRMSE of global solar radiation, considering the annual daily average, were 20.84% (reasonable) and 27.49% (fair), respectively, for the TRY file and 25.82% (reasonable) and 31.57% (poor), respectively, for the 2020 file. For the daily average in February, MAPE and RRMSE were respectively 17.90% (good) and 22.00% (fair) for TRY data and 19.10% (good) and 24.26% (fair) for 2020 data. The MAPE and RRMSE of daily averages in July were respectively 21.33% (reasonable) and 26.27% (fair) for TRY data and 28.75% (reasonable) and 32.51% (poor) for 2020 data.

Wind speed

Figure 6 shows the monthly, daily, hourly, and monthly hourly averages of wind speed in São Paulo, and Table 5 shows their respective monthly and annual error indices.

There were no noticeable differences between the wind speeds of the TRY file and the morphed 2020 file (Figure 6). The annual wind speed averages of weather files were 3.98 m/s (TRY) and 4.03 m/s (2020). The projections show that the maximum monthly average wind speed will decrease from 4.65 m/s (October, TRY) to 4.60 m/s (October, 2020 period) by 2040, whereas the minimum will increase from 3.16 m/s (June, TRY) to 3.22 m/s (June, 2020 period).

Wind speed was the climatic variable with the greatest discrepancy between observed and projected values. Hourly data from weather files indicated wind speeds of up to 16 m/s, whereas observed values did not exceed 4 m/s. The MAE and RMSE of the cumulative annual variable were 2.65 m/s and 3.30 m/s, respectively, for TRY data and 2.70 m/s and 3.36 m/s, respectively, for morphed 2020 data. The wind speed errors of weather files (annual MAPE and RRMSE), considering all hourly data, were 199.23% (inaccurate) and 220.45% (poor) for the TRY file and 203.19% (inaccurate) and 224.44% (poor) for the 2020 file.

Figure 6
Wind speed in São Paulo, São Paulo State, Brazil, according to weather file and meteorological station data

Table 5
Monthly and annual errors of wind speed sequences from weather files in relation to weather station data

These high discrepancies may be explained by the local characteristics of the data source. TRY parameters were collected at an overly open area (airport), and EM-IAG-USP is located in the center of the metropolis, surrounded by vegetation. Fujibe (2011) and Takaneet al. (2017) explained that changes in temperature and wind speed are related to changes at the microscale, such as surrounding buildings, tree growth, and other factors influencing the localized temperature increase around the observation sites. A decrease in wind exposure is accompanied by a reduction in wind speed and upward heat diffusion, resulting in an increase in temperature (as discussed in section Dry-bulb temperature); thus, temperature and wind speed trends have a negative correlation. USP (2017) reported that wind speed observations were hindered by tree growth, as trees act as a barrier. Such a barrier might have been responsible for the increase in roughness and consequent reduction in wind intensity. To mitigate this condition and comply with WMO requirements, EM-IAG-USP conducts tree pruning periodically.

Given the similarity between weather files and the differences in relation to the actual observed data, it is believed that the high error of the 2020 file is due to its data being based on the TRY file. The difficulty of obtaining similar wind speed values between measured and modeled data should be considered, as discussed by Laudari, Sapkota, and Banskota (2018). The authors observed that modeled wind speed is always greater than measured data, having a moderate correlation. This factor is evidence of the difficulty in predicting climate change and validating the tool.

Conclusions

In response to the growing need for studies on the thermoenergetic analysis of buildings amidst climate change, this paper contributed by analyzing the effectiveness of the CCWorldWeatherGen weather file generation tool in comparison with meteorological data from TRY and more recent observations in São Paulo, São Paulo State, Brazil. This tool has some interesting applications. It is possible to generate weather file progressions for analysis of the performance of buildings in the face of climate change; however, it is limited to the consistency of the reference weather file.

Weather is difficult to predict. This fact, added to the limitations of the TRY file for São Paulo and EM-IAG-USP (e.g., location and equipment accuracy), explains the large errors in global solar radiation and wind speed sequences generated by CCWorldWeatherGen. Nevertheless, the tool generated temperature and relative humidity sequences that showed good fidelity with real observations for the 2011-2023 period, as compared with TRY data.

The practicality of CCWorldWeatherGen to generate complete, free-form future weather files is underscored. However, the fact that the tool was conditioned to a local weather file based on meteorological data from 1961-1990 can be a limiting factor influencing the high discrepancies between observed and predicted wind speed and global solar radiation. Given these differences, the tool was partially validated. Dry-bulb temperature and relative humidity had good accuracy with measured data, but solar radiation and wind speed did not.One option would be to combine dry-bulb temperature and relative humidity obtained from morphed data and solar radiation, wind speed, and other variables obtained from the TRY file to form a weather file with reduced errors. However, this might not be feasible, as the work would be exhaustive and could compromise the practicality of obtaining future weather files.

Validation of weather sequence techniques is complex and relative, depending on current and past data limitations and inconsistencies. Furthermore, the method used for accuracy analysis can influence the results, as was observed here by differences in error indices. Thus, this study contributed modestly, within the available possibilities, by demonstrating that CCWorldWeatherGen sequences are suitable for building simulations. No inconsistencies or overly weighted data were observed in the generated weather file in relation to the reference weather file. However, it is important to consider the origin of the reference weather file and the effects of microclimate and other particularities of the specific location (environment) of the simulated building in future conditions.

It is worth noting that this article focused on analyzing the effectiveness of the CCWorldWeatherGen tool, as observed through four climate variables - typically, weather files encompass 27 climate variables - under the described conditions of observations from EM-IAG-USP, in São Paulo, São Paulo State, Brazil. Additionally, it is important to reiterate that the climate projections considered the A2 emissions scenario, which is associated with numerous uncertainties in pollutant emission estimates and climate predictions. The focus of the study was not to assess whether the considered scenario is the most appropriate for analyses or to determine the outcomes of climate change.

Future research is recommended to expand scientific knowledge on the topic and address the limitations of this study. Supplementary studies should consider additional variables from weather files and analyze data from different locations and measurement conditions, as well as other climate scenarios and periods, seeking to understand if, and how, these methodological variations influence the effectiveness of the progression tool.

Acknowledgments

We thank the Weather Station of the Institute of Astronomy, Geophysics and Atmospheric Sciences (IAG) of the University of São Paulo (USP) for making available the meteorological observations, and the Brazilian Federal Agency for Support and Evaluation of Graduate Education (CAPES) for funding this research.

References

• ALVES, C. A. Resiliência das edificações às mudanças climáticas na região metropolitana de São Paulo. São Paulo, 2014. Master Thesis (MSc in Architecture and Urban Planning) - Universidade de São Paulo, São Paulo, 2014.
• ASHRAFIAN, T. Enhancing school buildings energy efficiency under climate change: a comprehensive analysis of energy, cost, and comfort factors. Journal of Building Engineering, v. 80, 107969, 2023.
• BELCHER, S. E.; HACKER, J. N.; POWELL, D. S. Constructing design weather data for future climates. Building Services Engineering Research and Technology, v. 26, p. 49-61, 2005.
• BERARDI, U.; JAFARPUR, P. Assessing the impact of climate change on building heating and cooling energy demand in Canada. Renewable and Sustainable Energy Reviews, v. 121, 109681, 2020.
• BEWLEY, J. L.; BERZOLLA, Z. M. Building energy modeling and resilient department of defense installations: accounting for climate change. Alexandria: Institute for Defense Analyses, 2023.
• BROCK, F. V.; NICOLAIDIS, C. E. Instructor’s Handbook on Meteorological Instrumentation. Boulder: Atmospheric Technology Division, National Center for Atmospheric Research, 1984.
• CAMPAGNA, L. M.; FIORITO, F. On the impact of climate change on building energy consumptions: a meta-analysis. Energies, v. 15, n. 1, 354, 2022.
• CARLO, J. C.; LAMBERTS, R. Processamento de arquivos climáticos para simulação do desempenho energético de edificações. Florianópolis: Laboratório de Eficiência Energética em Edificações, Universidade Federal de Santa Catarina, 2005. Eletrobrás/PROCEL, AET N 02/04.
• CASAGRANDE, B. G.; ALVAREZ, C. E. Preparação de arquivos climáticos futuros para avaliação dos impactos das mudanças climáticas no desempenho termoenergético de edificações. Ambiente Construído, v. 13, n. 4, p. 173-187, out./dez. 2013.
• CHOU, J. S.; BUI, D. K. Modeling heating and cooling loads by artificial intelligence for energy-efficient building design. Energy and Buildings, v. 82, p. 437-446, 2014.
• CENTRO DE REFERÊNCIA PARA ENERGIA SOLAR E EÓLICA SÉRGIO BRITO. Centro de Pesquisas de Energia Elétrica. Solarimetria e Instrumentos de Medição. Rio de Janeiro: CRESESB, 2008. Available: Available: http://www.cresesb.cepel.br/index.php?section=com_content⟨=pt&cid=311 Access: 07 Mar. 2024.
  » http://www.cresesb.cepel.br/index.php?section=com_content⟨=pt&cid=311
• CUBASCH, U. et al Projections of future climate change. In: HOUGHTON, J. T. et al Climate change 2001: the scientific basis. Cambridge: Cambridge University Press, 2001.
• DALLACORT, R. et al Análises do comportamento de um actinógrafo bimetálico (R. Fuess-Berlin-Steglitz) em diferentes tipos de cobertura do céu. Acta Scientiarum, v. 26, n. 4, p. 413-419, 2004.
• DE WILDE, P.; COLEY, D. The implications of a changing climate for buildings. Building and Environment, v. 55, p. 1-7, 2012.
• DESPOTOVIC, M. et al Evaluation of empirical models for predicting monthly mean horizontal diffuse solar radiation. Renewable and Sustainable Energy Reviews , v. 56, p. 246-260, 2016.
• ESCANDÓN, R. et al How do different methods for generating future weather data affect building performance simulations? A comparative analysis of Southern Europe. Buildings, v. 13, n. 9, 2385, 2023.
• FARAH, S. et al Integrating climate change into meteorological weather data for building energy simulation. Energy and Buildings, v. 183, p. 749-760, 2019.
• FERNANDES, F. C. F.; GODINHO FILHO, M. Planejamento e controle da produção: dos fundamentos ao essencial. São Paulo: Atlas, 2010.
• FUJIBE, F. Urban warming in Japanese cities and its relation to climate change monitoring. International Journal of Climatology, v. 31, n. 2, p. 162-173, 2011.
• GONÇALVES, E. L. S. et al Multiscale modeling to optimize thermal performance design for urban social housing: a case study. Applied Thermal Engineering, v. 236, 121379, 2024.
• GOULART, S.; LAMBERTS, R.; FIRMINO, S. Dados climáticos para projeto e avaliação energética de edificações para 14 cidades brasileiras. Florianópolis: Núcleo de Pesquisa em Construção/UFSC, 1998.
• GUAN, L. Preparation of future weather data to study the impact of climate change on buildings. Building and Environment , v. 44, n. 4, p. 793-800, 2009.
• GUERMOUI, M. et al New soft computing model for multi-hours forecasting of global solar radiation. The European Physical Journal Plus, v. 137, 162, 2022.
• HATFIELD, J. L.; PRUEGER, J. H. Temperature extremes: effect on plant growth and development. Weather and climate extremes, v. 10, p. 4-10, 2015.
• INTERGOVERNMENTAL PANEL ON CLIMATE CHANGE. AR6 synthesis report: climate change 2023.Genebra, 2023.
• JAMIESON, P. D.; PORTER, J. R.; WILSON, D. R. A test of the computer simulation model ARC-WHEAT1 on wheat crops grown in New Zealand. Field Crops Research, v. 27, p. 337-350, 1991.
• JENTSCH, M. F. et al Transforming existing weather data for worldwide locations to enable energy and building performance simulation under future climates. Renewable Energy, v. 55, p. 514-524, 2013.
• JENTSCH, M. F. Climate Change Weather File Generators: Technical reference manual for the CCWeatherGen and CCWorldWeatherGentools - Version 1.2. Southampton: University of Southampton, 2012.
• JENTSCH, M. F.; BAHAJ, A. S.; JAMES, P. A. B. Manual CCWorldWeatherGen. Climate change world weather file generator. Version 1.7. Southampton: University of Southampton , 2012.
• JONES, B. A. Planting urban trees to improve quality of life? The life satisfaction impacts of urban afforestation. Forest Policy and Economics, v. 125, p. 102408, 2021.
• JONES, P. D. et al Surface air temperature and its changes over the past 150 years. Reviews of Geophysics, v. 37, n. 2, p. 173-199, 1999.
• KNOWLES, N.; DETTINGER, M. D.; CAYAN, D. R. Trends in snowfall versus rainfall in the western United States. Journal of Climate, v. 19, n. 18, p. 4545-4559, 2006.
• KUANG, W. Seasonal variation in air temperature and relative humidity on building areas and in green spaces in Beijing, China. Chinese Geographical Science, v. 30, p. 75-88, 2020.
• KUTTY, N. A. et al A systematic review of climate change implications on building energy consumption: impacts and adaptation measures in hot urban desert climates. Buildings , v. 14, n. 1, 13, 2024.
• LAUDARI, R.; SAPKOTA, B.; BANSKOTA, K. Validation of wind resource in 14 locations of Nepal. Renewable Energy , v. 119, p. 777-786, 2018.
• LEUNG, M. C. et alThe use of occupancy space electrical power demand in building cooling load prediction. Energy and Buildings , v. 55, p. 151-163, 2012.
• LEWIS, C. D. Industrial and business forecasting methods. London: Butterworths, 1982.
• LI, M. F. et al General models for estimating daily global solar radiation for different solar radiation zones in mainland China. Energy Conversion and Management, v. 70, p. 139-48, 2013.
• MA, Q.; WANG, K.; WILD, M. Impact of geolocations of validation data on the evaluation of surface incident shortwave radiation from Earth System Models. Journal of Geophysical Research: Atmospheres, v. 120, n. 14, p. 6825-6844, 2015.
• MACHADO, J. M. Avaliação de desempenho térmico em edificações multifamiliares considerando as mudanças climáticas futuras. Vitória, 2019.Master Thesis (MSc in Architecture and Urban Planning) - Universidade Federal do Espírito Santo, Vitória, 2019.
• MALM, W. C.; WALTHER, E. G. A review of instrument-measuring visibility-related variable. Las Vegas: Environmental Protection Agency, 1980.
• MARTELLI, A.; SANTOS JUNIOR, A. R. Arborização urbana do município de Itapira-SP: perspectivas para educação ambiental e sua influência no conforto térmico. Revista Eletrônica em Gestão, Educação e Tecnologia Ambiental, v. 19, n. 2, p. 1018-1031, 2015.
• MORENO, J. J. M. et al Using the R-MAPE index as a resistant measure of forecast accuracy. Psicothema, v. 25, n. 5, p. 500-506, 2013.
• NIELSEN, C. N.; KOLARIK, J. Utilization of climate files predicting future weather in dynamic building performance simulation-a review. Journal of Physics: Conference Series, v. 2069, 012070, 2021.
• NUNES, G. H. Influência das mudanças climáticas na sensibilidade global de parâmetros termofísicos de habitações brasileiras. Londrina, 2021. Master Thesis (MSc in Civil Engineering) - Universidade Estadual de Londrina, Londrina, 2021.
• RIVINGTON, M. et al Downscaling regional climate model estimates of daily precipitation, temperature and solar radiation data. Climate Research, v. 35, n. 3, p. 181-202, 2008.
• RODRIGUES, E.; FERNANDES, M. S.; CARVALHO, D. Future weather generator for building performance research: an open-source morphing tool and an application. Building and Environment , v. 233, 110104, 2023.
• SANTOS, T. S. et alIncertezas das projeções de mudanças climáticas: análise preliminar. Ciência e Natura, v. 37, n. 1, p. 63-68, 2015.
• SIU, C. Y. et al Evaluating thermal resilience of building designs using building performance simulation - A review of existing practices. Building and Environment , v. 234, 110124, 2023.
• SU, Q.; CHANG, H.-S.; PAI, S.-E. A comparative study of the resilience of urban and rural areas under climate change. International Journal of Environmental Research and Public Health, v. 19, n. 15, 8911, 2022.
• TAKANE, Y. et al Factors causing climatologically high temperatures in a hottest city in Japan: a multiscale analysis of Tajimi. International Journal of Climatology , v. 37, n. 3, p. 1456-1473, 2017.
• TOOTKABONI, M. P. et al A comparative analysis of different future weather data for building energy performance simulation. Climate, v. 9, n. 2, 37, 2021.
• TORRES, R. R. Análise de incertezas em projeções de mudanças climáticas na América do Sul. São José dos Campos, 2014.PhD Thesis (PhD in Meteorology) - Instituto Nacional de Pesquisas Espaciais, São José dos Campos, 2014.
• TRIANA, M. A.; LAMBERTS, R.; SASSI, P. Should we consider climate change for Brazilian social housing? Assessment of energy efficiency adaptation measures. Energy and Buildings , v. 158, p. 1379-1392, 2018.
• TROUP, L.; FANNON, D. Morphing climate data to simulate building energy consumption. In: ASHRAE AND IBPSA-USA SIMBUILD: BUILDING PERFORMANCE MODELING CONFERENCE, Salt Lake City, 2016. Proceedings [...] Salt Lake City, 2016.
• U.S. DEPARTMENT OF ENERGY. Weather data. Available: https://energyplus.net/weather Access: 29 Feb. 2020.
  » https://energyplus.net/weather
• UNIVERSIDADE DE SÃO PAULO. Estação Meteorológica do Instituto de Astronomia, Geofísica e Ciências Atmosféricas. Meteorological data measured from 01/01/2011 to 06/30/2023: dry-bulb temperature; relative humidity; global solar radiation; wind speed. São Paulo, 2023. Available: Available: http://www.estacao.iag.usp.br/sol_dados.php Access: 30 Nov. 2023.
  » http://www.estacao.iag.usp.br/sol_dados.php
• UNIVERSIDADE DE SÃO PAULO. Estação Meteorológica do Instituto de Astronomia, Geofísica e Ciências Atmosféricas. Estação Meteorológica - IAG/USP. São Paulo, 2024. Available: Available: http://www.estacao.iag.usp.br/index.php Access: 11 Mar. 2024.
  » http://www.estacao.iag.usp.br/index.php
• UNIVERSIDADE DE SÃO PAULO. Estação Meteorológica do Instituto de Astronomia, Geofísica e Ciências Atmosféricas.Relatório técnico da estação meteorológica do IAG/USP Nº1: Instrumentos e Procedimentos. São Paulo, 2010. Available: Available: http://www.estacao.iag.usp.br/boletim.php Access: 22 Feb. 2023.
  » http://www.estacao.iag.usp.br/boletim.php
• UNIVERSIDADE DE SÃO PAULO. Estação Meteorológica do Instituto de Astronomia, Geofísica e Ciências Atmosféricas. Relatório técnico da estação meteorológica do IAG/USP Nº3: Comparação entre Actinógrafos. São Paulo, 2016. Available: Available: http://www.estacao.iag.usp.br/boletim.php Access: 26 Dec. 2023.
  » http://www.estacao.iag.usp.br/boletim.php
• UNIVERSIDADE DE SÃO PAULO. Estação Meteorológica do Instituto de Astronomia, Geofísica e Ciências Atmosféricas. Relatório técnico da estação meteorológica do IAG/USPNº5: Crescimento das árvores ao redor da EM-IAG-USP e efeito nas observações meteorológicas. São Paulo, 2017. Available: Available: http://www.estacao.iag.usp.br/boletim.php Access: 07 Mar. 2024.
  » http://www.estacao.iag.usp.br/boletim.php
• WANG, L.; LIU, X.; BROWN, H. Prediction of the impacts of climate change on energy consumption for a medium-size office building with two climate models. Energy and Buildings, v. 157, p. 218-226, 2017.
• WEIRICH, R. A. et alArborização urbana para mitigação das condições microclimáticas em Goiânia, Goiás. Revista Ecologia e Nutrição Florestal-ENFLO, v. 3, n. 2, p. 48-58, 2015.
• WORLD METEOROLOGICAL ORGANIZATION. Guide to hydrological practices. WMO No. 168. Geneva, 1994.
• WORLD METEOROLOGICAL ORGANIZATION. Guide to meteorological instruments and methods of observation. WMO No. 8. Geneva, 1996.
• YASSAGHI, H.; MOSTAFAVI, N.; HOQUE, S. Evaluation of current and future hourly weather data intended for building designs: a Philadelphia case study. Energy and Buildings , v. 199, p. 491-511, 2019.

Edited by

• Editor:
  Enedir Ghisi

• Editora convidada:
  Kelen Dornelles

Publication Dates

• Publication in this collection
  07 Oct 2024
• Date of issue
  2024

History

• Received
  27 Jan 2024
• Accepted
  26 Mar 2024