Applied Energy 303 (2021) 117584

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Impact of weather data and climate change projections in the

refurbishment design of residential buildings in cooling dominated climate
Rosa Francesca De Masi a, *, Antonio Gigante a, Silvia Ruggiero a, Giuseppe Peter Vanoli b
a
University of Sannio, DING - Department of Engineering, Piazza Roma, 21, 82100 Benevento, Italy
b
University of Molise, Department of Medicine and Health Sciences -Vincenzo Tiberio, via G. Paolo II contrada “Tappino”, 86100 Campobasso, Italy

H I G H L I G H T S

• Weather files defined with different methodology and dataset are compared.
• Projections of climate changes are considered in the construction of weather files.
• The case study is a residential building in a cooling dominated climate.
• The composition of weather files greatly influences the energy performance prevision.
• Traditional efficiency measures are not resilient for the expected climate changes.

A R T I C L E I N F O A B S T R A C T

Keywords: The methodology, the site and the dataset as well as the emissions scenario considered in the weather file
Weather file construction definition influence the numerical evaluation of efficiency measures resilience. With a complete statistical and
Climate changes critical approach, the paper analyzes the importance of these aspects by means of a residential case study
Weather file projection
simulated in Benevento, a city of south Italy.
Energy refurbishment
Mediterranean climate
Using data monitored from 2015 to 2020, a current weather file is built with different methodologies. The
comparison indicates that there is not repeatability of the year chosen as a reference for the various months and
thus the resolution of the building energy balance could bring different results. Some future climate projections
are also generated on medium (2050 s) and long (2080 s) term considering different emission scenarios. With
long term projection, the heating degree days are reduced also of − 21% meanwhile the cooling degree days are
more than double compared with the current condition. This suggests a remarked transition towards a dominant
cooling climate for Benevento.
Moreover, when the climate change is considered, the insulation intervention and the installation of double
glazed low emissive window is not resilient because the heating energy need decreases also of − 56%, but the
cooling energy need increases of + 62% (2080 s). If the efficiency measures include also the cool roof and the
external shadings, the cooling demand could be reduced until –33% in some scenarios (e.g. RCP 4.5-50th
percentile) and increased (+31%) in some others (e.g. 2080 s).

disasters in 2020, with catastrophic results for millions of people. This is

the premise of the document “Counting the cost 2020 A year of climate
1. Introduction
breakdown ”published by the Christian Aid organization in December
2020 [1]. For the assessment of climate change, the Intergovernmental
Covid-19 may have dominated the news agenda in 2020, but for
Panel on Climate Change (IPCC) [2] provides a continuous upgrade of
many people, the ongoing climate crisis has caused even bigger danger
data with deep analysis of potential environmental and socio-economic
to their lives and livelihoods. Be they fires in Australia and the United
impacts through the periodic production of Scientific Assessment Re­
States, floods in China, India and Japan or storms in Europe and the
ports. In Europe, the community goals aimed at achieving climate
Americas, almost every part of the globe was touched by climate-related

* Corresponding author.
E-mail addresses: [email protected] (R.F. De Masi), [email protected] (A. Gigante), [email protected] (S. Ruggiero), [email protected]
(G.P. Vanoli).

https://doi.org/10.1016/j.apenergy.2021.117584
Received 22 February 2021; Received in revised form 11 June 2021; Accepted 7 August 2021
Available online 27 August 2021
0306-2619/© 2021 Elsevier Ltd. All rights reserved.
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Nomenclature m number of days during the month (Sandia method)

MAD Mean absolute difference
AP Atmosferic pressure MAPD Mean absolute percentage difference
CC_MT Future projection obtained by CCWorldWeatherGen – µ average value
middle term (2050 s) n number of days during the single month (EN ISO 15927–4)
CC_LT Future projection obtained by CCWorldWeatherGen – Long N total number of days for the month m over all years (EN ISO
term (2080 s) 15927–4)
CDD Cooling degree days (◦ C day) P Precipitation
CDF Cumulative distribution functions p primary parameter (EN ISO 15927–4)
δk absolute difference between long-term CDF and short-term PVGIS Photovoltaic Geographic Information System
CDF RCP Representative concentration paths
DBL_L Double glazing with low emissivity coating RH Relative humidity
DBL_S Double glazing with solar control RP Reflective paint
DBT Dry bulb temperature RS Retrofit scenario
DBTd,ave,w Average daily DBT in daytime hours only – winter months σ2 Variance
(◦ C) SF Solar factor
DBTd,ave,s Average daily DBT in daytime hours only – summer S/V Surface to Volume ratio, [m− 1 ]
months (◦ C) thtot Total thickness, [m]
DHI Diffuse horizontal radiation TI Thermal insulation
DNI Direct normal radiation TMY Typical meteorological year
DPT Dew point temperature TMY2 Typical meteorological year 2
ΔEC Reduction of the energy need for cooling [%] TRY Test reference year
ΔEC,cl,ch Reduction of the energy need for cooling with climate TRY_BN Test reference year built for Benevento
changes [%] TRY_BN Test reference year built by the CTI
ΔEH Reduction of the energy need for heating [%] Ug Glass thermal transmittance [W/m2 K]
ΔEH,cl,ch Reduction of the energy need for heating with climate Ur Roof thermal transmittance [W/m2 K]
changes [%] Us Basefloor thermal transmittance [W/m2 K]
ΔEP Primary energy saving [%] Uw Wall thermal transmittance [W/m2 K]
ΔEPcl,ch Primary energy saving considering the climate changes[%] wD Wind direction (degrees)
ΔTh,ave Average annual hourly temperature difference [◦ C] Wj weight assigned to the j-th index.
ΔTave,monthly Difference in average monthly temperature [◦ C] wS Wind speed (m/s)
ES External shading WS Weighted sum
FS Finkelstein-Schafer W_45-10 Future projection obtained by Weathershift TM – RCP 4.5
GCM General circulation model (10 th percentile)
GHI Global horizontal radiation W_45-10 Future projection obtained by Weathershift TM – RCP 4.5
GHId,tot,s Total daily GHI – summer months (W/m2) (50 th percentile)
HadCM3 Hadley Centre Coupled Model 3 W_45-10 Future projection obtained by Weathershift TM – RCP 4.5
HDD Heating degree days (◦ C day) (90 th percentile)
IS Internal shading W_85-10 Future projection obtained by Weathershift TM – RCP 4.5
IWEC International weather for energy calculations (10 th percentile)
LT Light transmission W_85-10 Future projection obtained by Weathershift TM – RCP 4.5
J ranked order number – SHORT TERM (EN ISO 15927–4) (50 th percentile)
K ranked order number – LONG TERM (EN ISO 15927–4) W_85-10 Future projection obtained by Weathershift TM – RCP 4.5
k Ranked order number (90 th percentile)
Kd Diffuse fraction of global horizontal radiation XMY extreme year
(dimensionless) XMYmin extreme cold year
Kt Clearness index (dimensionless) XMYmax extreme hot year
λ Thermal conductivity [W/m K] y year (EN ISO 15927–4)

neutrality by 2050 [3–4] push towards an ever greater energy efficiency for the other countries, the weather data are not available and some­
of buildings, considering the prevalent contribution of the civil sector to times it occurs to build the files using interpolated data of the nearby
final energy consumptions and therefore to global greenhouse gas locations that could be not representative of the location to be studied.
emissions. It is necessary to take into account the strong climate change The performance gap due to the adoption of conventional or new
underway to design wisely both new buildings and the energy retrofit of weather files defined with data recently monitored in areas with
existing ones. This means that the adoption of updated climate files into different levels of urbanization has been discussed by De Masi et al. [7].
the dynamic energy simulation software used by designers and re­ The adequacy of adopted weather files regards not only the available
searchers is an important topic of the future engineering research. reference period but also the methodology used for the construction of
EnergyPlus, which is one of the leading dynamic energy simulation the meteorological year. The most commonly used methodology is
programs, provides numerous databases containing climate files in EPW called Typical Meteorological Year (TMY) [8]. TMYs are assembled by
format [5]. However, in many cases these files derive from historical compiling the individual months which best represent the long-term
monitoring; for instance, the “Gianni de Giorgio” database (IGDG) [6] is monthly means conditions of different variables. TMY method is based
available for Italy with 68 locations. These files are created using data on Finkelstein-Schafer statistic and a weighting factor is applied to each
collected in the period 1951–1970. For many Italian locations, as also climatic parameter in the typical month The TMY is composed of 12

2
R.F. De Masi et al. Applied Energy 303 (2021) 117584

typical meteorological months (January through December) that are and maximum and minimum hourly average values in each month.
concatenated without modification for creating a single year with Using prototype building models, they have concluded that XMY based
complete data record for primary measurements. These monthly data on hourly maximum and minimum dry-bulb temperature best captured
sets contain actual time-series meteorological measurements and the range of energy use for the XMY. They have suggested that designers
modeled solar values, although some hourly records may contain filled should use three weather files, one TMY and two XMYs based on hourly
or interpolated data for periods when original observations are missing maximum and minimum dry-bulb temperature for evaluating the vari­
from the data archive. ability of building energy performance.
Over the years, the methodology was mainly focused on the con­ Piotr Narowski et al. [20] proposed the creation of extreme years
struction of climate archives for locations in the United States. The first named Untypical Meteorological Years (UMY), based on Weather Year
TMY data set for the United States was produced by Sandia National for Energy Calculations 2 (WYEC2) procedure [21]. From the energy
Laboratories in 1978 for 248 locations using long-term weather and simulations carried out with both UMYs and TMYs files, they have
solar data from the 1952–1975 SOLMET/ERSATZ database [8]. National observed maximum differences of 24% and 46% respectively in the
Renewable Energy Laboratory (NREL) updated the TMY data in 1994 energy demand for heating and cooling.
using data from the 30-year 1961–1990 [9]. In this case the TMY Kočí et al. [22] proposed the comparison between the test reference
methodology has undergone changes and the new collection has been year (TRY) and the critical and positive weather years. In this case, the
identified as TMY2. The main change concerned the introduction of the extreme years were selected according to the methodologies presented
tenth variable, namely the direct normal radiation to which the highest in [23–24]. Erba et al. [25] reported the comparison of the simulation
weighting factor was attributed. In 1997, an initiative called Interna­ results for public housing in Milan with two different extreme weather
tional Weather Year for Energy Calculations (IWEC) has created new datasets, respectively representative of the coldest winter and the
files using the TMY methodology with updated weighting factors [10]. warmest summer. In [26], starting from the approach described by the
The last update of the climate files of the US locations occurred in 2008 standard EN ISO 15927–4, a new method to develop hot and cold
with the creation of the TMY3 data set, containing the file in EPW format extreme reference years was proposed from multi-year hourly weather
for 1200 locations [11]. The same method, as expected in TMY2, was series. Nik et al. [27] have suggested an approach for assessing the
implemented but referring to the input data from the 1961–1990 impact of climate change on buildings and their energy performance,
NSRDB, Version 1.1 and the 1991–2005 NSRDB update. For some lo­ based on the creation of three sets of meteorological data starting from
cations, the TMY3 dataset draws representative months from 1976 to the outputs of one or more regional climate models (RCMs): typical year
2005, which means that the most recent dataset potentially incorporates reduced (TDY), extreme cold year (ECY) and extreme hot year (EWY).
meteorological data measured over 40 years ago, although it serves as The method was applied to assess different future climate scenarios for
the basis for the current simulation. Recently, a new TMY series was an office building in Geneva and for the residential building stock in
created for Europe, with data collected from 2004 to 2018 [12]. Stockholm. He has concluded that a combination of typical and extreme
In Europe the method proposed in the EN ISO 15927–4 standard [13] data sets allows the better probability distributions of future conditions.
is often used. In this case the typical meteorological year is named test Instead, the uncertainties about climate change and how it can affect
reference year (TRY). Although also this standard is based on the the building energy consumption can be studied with future climate
Finkelstein-Schafer statistics, compared with TMY methodology there projections. This type of analysis concerns the resilience concept of
are different criteria for the selection of the typical month and the building. Indeed, according to the literature, the main current challenge
weight is not applied to the evaluated climatic parameters [14]. Basing is to design buildings that are efficient throughout their useful life,
on the standard EN ISO 15927–4, during 2015, the Italian thermo- taking into account present and future climatic conditions. The assess­
technical committee (CTI) has made the test reference year (TRY) for ment of future climate change is an issue that currently presents strong
110 locations, based on measurements made between 1989 and 2014 uncertainties. As explained by Hallegatte et al. [28], there are not
[15]. U.S. Department of Energy’s (DOE) has not integrated the CTI methods for eliminating this uncertainty since the level of GHG emis­
climatic years into EnergyPlus databases because these require additional sions cannot be predicted. The climate projections are generated starting
variables to be compatible with the EPW format. from the outputs of general / global circulation models (GCMs) or
Lupato et al. [16] by means of the comparison of the historical IGDG Regional Climate Models (RCMs) and take into account the emission
database with those defined by the CTI, have concluded that the his­ scenarios of the third and fourth IPCC (SRES) [29] or the Representative
torical weather files lead to underestimate the energy needs for cooling Concentration Pathways (RCP) reports) of the fifth IPCC report (AR5)
and to overestimate those for heating. Furthermore, Pernigotto et al. [30]. According to [31], the RCP scenarios are based on a different
[17] have proposed a possible modification to EN ISO 15927–4 by approach to SRES scenarios and ‘‘they are not necessarily more capable
introducing weighting coefficients for the various meteorological pa­ of representing future developments than the SRES scenarios”. Thus, for
rameters. The study was conducted for five locations in Northern Italy research purposes, using either SRES or RCP scenarios is appropriate.
with 10 years or less in the dataset and for a sample of 48 simplified The GCM models are numerical computer-driven models that simu­
buildings. Both proposed changes have improved the representativeness late the physical processes in earth’s atmosphere and oceans. The GCMs
of the baseline year results. typically have coarse resolutions and provide average daily or monthly
The method in accordance to ISO 15927–4 is also the basis of the values of the meteorological variables, with a high spatial resolution (i.e.
TMY-tool made available by the Photovoltaic Geographical Information several hundred kilometers). Statistical or dynamic downscaling is
System (PVGIS) [18] for Europe, Africa and most of Asia and America. necessary to adapt the data of these models to the resolutions required
Although the typical meteorological years are good enough to by the energy simulation programs of buildings. The dynamic down­
represent the typical behavior, it should be also considered a reasonable scaling method requires switching from GCM to RCM and provides data
range of extreme events that a building could experience over its useful with higher spatial resolutions. However, more computation time is
life. Indeed, extreme weather events are increasingly frequent and required in this case and the availability of regional climate models is
intense in current climates. Furthermore, the strong climate change very limited. Statistical downscaling is mainly done by relying on the
underway forces designers to consider the energy efficiency of buildings “morphing method”, proposed by Blecher in 2008 [32]. This method
also from a future rather than a current perspective. uses mathematical equations and GCM outputs to “transform” current
Regarding this, Crawley et al [19] have proposed a methodology for weather files into future climate projections. To enable the generation of
the extreme meteorological years (XMY). The proposed method foresees morphed-epw weather files, the University of Southampton’s Sustain­
the construction of several extreme years (at least four) for six meteo­ able Energy Research Group (SERG) in 2017 made available the
rological parameters, on the basis of the maximum and minimum daily “CCWorldWeatherGen” tool [33], based on work discussed in [34–35].

3
R.F. De Masi et al. Applied Energy 303 (2021) 117584

This meteorological generator is based on the outputs of GCM HadCM3 Yarima et al. [50] have scaled the outputs of a global circulation
[36] and on the emission scenario A2 [29]. Based on two of the RCPs model using the dynamic downscaling method. From simulations car­
(4.5 and 8.5), Arup and Argos Analytics have developed a tool called ried out on a single-family building in Tokyo, they have observed that in
“WeatherShift ™” [37] which applies the morphing procedure on the the month of August of the near future (2026–2035), the total sensible
results of 14 GCM (on about 40 models) available under AR5[30]. Other thermal load and the latent load could increase by 26% and 10%
meteorological generators available for statistical downscaling are respectively. Radhi [51] has studied the influence of climate change on
based on stochastic models [38]. Recently, Moazami et al. [39] have the energy performance of residential buildings in the UAE. The results
provided an overview of the main approaches to create future meteo­ suggest that the cooling demand would increase by 23.5% and that the
rological datasets based on statistical and dynamic downscaling of thermal mass and thermal insulation can significantly mitigate the
climate models and tested them on 16 ASHRAE reference buildings for impact of global warming. For the building stock in Switzerland,
the location of Geneva, Switzerland. The results show that compared to Mutschler et al. [52] have suggested that the cooling demand could
typical conditions, the relative increase in peak cooling load can reach increase from 3 TWh with RCP 2.6 climate scenario, low population
up to 28.5% under extreme conditions. The authors have concluded that growth, low uptake, to over 17.5 TWh with RCP 8.5 climate scenario,
only those meteorological files generated by dynamic downscaling that high population growth, full cooling device uptake.
take into account typical and extreme conditions are reliable to provide As one key finding, Streicher et al. [53] have pointed out the need to
representative boundary conditions for testing the energy robustness of increase the level of retrofit activity and Bamdad et al. [54], very
buildings under conditions of future climatic uncertainty. In [35] future recently, have observed that optimising under future climate conditions
meteorological data produced using the RCM and the transformed data can lead to different optimal building designs. However, there are very
by the GCM for a naturally ventilated building in the UK were compared. few papers bout this topic. For instance, considering an existing sec­
The authors have stated that until the RCM is fully available for all re­ ondary school building in Ankara, Akkose et al. [55] have found that
gions, the modified procedure can be reliable for assessing the energy total energy consumption can be reduced up to 50% with the optimal
performance of buildings. Triana et al. [40] have evaluated the thermal combination of passive measures for maximum impact towards the
and energy performance of two residential buildings in Brazil using the adaptation to the changes in climatic conditions. Similarly, Fereidani
future weather file generated by the CCWorld-WeatherGen tool. A et al. [56] have suggest that in the Middle East region, the most effective
comprehensive statistical analysis of the results of the two future cooling passive measures for climate changes in residential buildings are
meteorological generator tools based on morphing method was pre­ highly reflective roofs, natural ventilation, and thermal insulation.
sented in [41]. Using Weathershift tool, Troup et al [42] have evaluated As underlined by the literature review, the interest on the effects of
the energy performance of an office building prototype for three future climate change on the building energy efficiency is increasing. However,
time windows (2030, 2060 and 2090) from the TMY3 climate files of the research is usually focused on the analysis of the energy performance
Boston, Miami and San Francisco. Due to the increased energy demand that the building with its present configuration will have in featuring
for cooling, an increase in the building’s annual primary energy con­ scenarios. This is an important topic that needs of other research due to
sumption by 2090 was observed up to 4% in Boston, up to 5% in Miami the numerous uncertainties, both related to the input weather files and
and up to 10% in San Francisco. Using the CCWorldWeatherGen tool, to the outputs of the climate models used and the IPCC emission sce­
Ciancio et al [43] have created meteorological input files for 19 Euro­ narios. Indeed, the adoption of reliable current weather files remains of
pean cities and for the future years 2050 and 2080. The obtained results considerable importance both for comparison with short-term analyzes
have showed an increase in cooling demand up to 272% in Rome in and because these files are the starting point for the generation of future
2080. With the help of the same tool, Pagliano et al. [44] have used projections. The available weather files are based on monitoring in areas
validated building energy model of a child care center in Milan (Italy) to not representative of the building location or these are based on old
investigate the changes in building energy use and uncomfortable measures.
thermal conditions. The results indicate that the space cooling and On the other side, considering the importance of intervening on
duration of discomfort in summer will be the major challenges in the existing buildings to achieve medium and long-term community objec­
future. In the same city, with two future scenarios (RCP 4.5 and RCP tives, there are currently few studies that analyze the concept of resil­
8.5), for existing residential building stock, Tootkaboniet al. [45] have ience for buildings subjected to energy retrofits. Instead, it is extremely
found a decrement in heating energy demand up to 30.9% and an in­ important the evaluation of the optimal refurbishment configuration for
crease in cooling energy demand, up to 255.1%. In case of historic the building stock taking into consideration the building operation
buildings González et al. [46] have found that the annual energy con­ under changed climatic conditions. Indeed, the better refurbishment
sumption should increase by almost 15% for future scenarios for 2050 solution could be different if evaluated with present and future weather
for the preservation of works of art due to higher level relative humidity. data, but there are very few papers that propose this analysis. To find
To generate future TMY files, Yassaghi et al. [47] have used both the solutions capable of mitigating the future worsening of climatic condi­
morphing technique and a stochastic meteorological generator for the tions, the sensitivity of refurbishment intervention should be investi­
periods of the near future (2020), mid-century (2050) and end of cen­ gated with consideration of all climates and building stock. Moreover, it
tury (2080). They have observed reductions in heating requirements is also important to consider the effect of extreme events and not only
between 20.5% and 29.6% while the cooling requirements have showed long-term changing for suggesting to designer the best strategy in the
an increase between 25.5% and 41.6%. Liu et al. [48] have developed design of energy efficiency measures.
the future hourly weather data using the downscaled data from multiple Starting from this analysis, the novelty of proposed paper is the
GCMs under different scenarios and evaluated the impacts of climate introduction of a complete approach for evaluating the variations that
change on the energy demand and indoor thermal comfort of mixed- different methodologies and dataset can cause in the construction of the
mode residential buildings in Hong Kong. Results indicate that by the weather file used for the energy simulations of actual and post-retrofit
end of this century, the indoor discomfort percentage in the cooling building configurations. Therefore, monitoring data are discussed as
seasons are expected to increase from 21.9% for TMY to 36.0% and well as the modification that the emissions scenarios could bring to the
50.4% under RCP4.5 and RCP8.5 scenarios, respectively, while the external forcing and thus to the energy balance.
annual cooling load is expected to increase up to 278.8%. By means of an Then, the impact of different methodologies to define the average
eXplainable artificial intelligence model, Chakraborty et al. [49] have and extreme climatic conditions and the use of different meteorological
found a persistent increment of cooling demand from 2020 to 2100, with generators for the creation of future climate projections is studied with
the maximum incremental 87.2% in hot–humid and 37.1%, 47.5%, reference to the prevision of the building performance before and after
85.3%, and 121% in mixed–humid climate regions. the refurbishment.

4
R.F. De Masi et al. Applied Energy 303 (2021) 117584

The case study is a residential building representative of the Italian Table 1

building stock of 1980 located in Benevento a city of south Italy. Site characterization of Benevento.
Benevento has typical Mediterranean climate and its weather file is not Benevento Winter design Summer
available in the public databases. The peculiarity of this climatic zone (ITA) day design day
lies in the fact that the weather conditions determine comparable Latitude 41◦ 7′ 55′ ’ Outdoor − 2 ◦C Outdoor 32 ◦ C
building energy need for heating and cooling. For this reason, the energy Longitude 14◦ 46′ 40′ ’ temperature temperature
retrofit requires the evaluation of efficiency measures that, in most Elevation 135 m
cases, lead to opposite effects in the two seasons. Climate zone C Outdoor 48.8 Outdoor 50 %
relative % relative
A new approach for evaluating the resilience of the refurbishment humidity humidity
scenarios is introduced. This allows understanding if the effectiveness of Heating 1316 Wind speed 4.4 Daily 11 ◦ C
efficiency measures in the current weather scenario is confirmed. Degrees- m/s temperature
Briefly, the case study will show how much the tradition adopted effi­ Day variation
(baseline,
ciency measures (insulation, replacement of window, installation of
20 ◦ C)
shading system and adoption of cool roof) are able to front the climate
change.
The findings of this research can be considered a first step in the based on meteorological data monitored from May 2015 to December
development of tools and methodology for including the climate 2020 in an urban station installed in the city center and in the test-room
changes in the energy analysis of building. Indeed, it is important to take named MATRIX - Multi Activity Test-Room for InnovatingX (41◦ 07′ 54.1
into consideration the effect of prevised emissions scenarios in the se­ “ N, 14◦ 47′ 03.7 ”E), owned by the Department of Engineering of the
lection of energy efficiency measures to design more resilient building. University of Sannio. The work proposed could be easily updated in the
In future works, other climate zones could be analyzed with the same future with different sources as in [60] for studying the effects of tem­
approach in order to reach a global knowledge of the impact of climate perature difference between the dense urban areas and the surrounding
change on the residential buildings. rural and suburban zones with the aim to investigate the urban heat
island phenomenon.
2. Methodological approach for the study Focusing on the adopted source, MATRIX is an experimental labo­
ratory for testing different solutions of building envelope under real
Fig. 1 summarizes the methodological approach adopted in this weather conditions. It is characterized by a large flexibility in terms of
paper. Benevento has a typical Mediterranean climate, with warm to variability of the thermal envelope, heating and cooling technologies,
hot, dry summers and mild to cool, wet winters. It is inside the “Csa” ventilation systems and artificial lighting. The laboratory was described
class according to Koppen classification [57]. in detail by Ascione et al. [61]. The recording of weather data is carried
Table 1 shows the site characterization according to UNI 10349:2016 out by the central weather station placed on the roof of the laboratory, at
[58] and National Decree n◦ 412: 1993 [59]. a height of about 7.20 m from the ground level. This apparatus records
The starting point is the definition of the weather file (EPW) for the most important meteorological variables such as global solar radi­
Benevento (south Italy) since there is not an official weather file avail­ ation, outdoor air temperature and relative humidity, wind speed and
able in commercial or public database. For this city, three datasets are wind direction, in addition to their pressure. For the measurement of
available on the website of Campania region [60]; two sites refer to low- amount of rain, it was installed a rain gauge with capacity of 400 cm3.
density population and activity and another one of urban center. These Finally, in order to measure the global solar radiation and infrared solar
data can be downloaded but are not complete with all needed variables radiation, five pyranometers and six pyrgeometers have been installed
due to failure of the acquisition system and however, these data are on all exposures, vertical and horizontal. The main characteristics of
available starting from 2018. For these reasons, for the proposed studies sensors are shown in Table 2.
these data were not used but The construction of the weather file is As shown in Fig. 1, different methodologies have been compared for

Fig. 1. Methodological approach adopted for the study.

5
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Table 2 set. The created files are also compared with two other files deriving
External sensors technical data. from external weather data sources. The first one was made available by
Sensor Unit Type Accuracy the CTI in 2015 for Benevento [15]; it was built with the UNI EN ISO
15927–4 standard and subsequently converted into EPW. In particular,
Rain Gauge mm/ Cylindrical body of 400 ±2%
h cm3 the TRY meteorological year was built with weather data collected from
Global solar W/ II class thermopile <10 % day-1 2006 to 2009 in a suburban station paced in Airola, a city near Benev­
radiation m2 pyranometer ento. Fig. 2 shows the position of the test-room and the distance, about
Infrared solar W/ Pyranometer − 300 ÷ 300 W/m2 (5.5 20 km, from the weather station of MATRIX.
radiation m2 ÷ 45 μm)
Wind speed m/s Ultrasonic anemometer ±2%
The second one has been obtained for Benevento from the TMY-
Wind direction ◦
N Ultrasonic anemometer ±2◦ RMSE 1.0 m/s TOOL made available by the Photovoltaic Geographic Information
Air temperature ◦
C Pt100 ±0.1% System (PVGIS) of the Joint Research Center [18] based on nine climatic
Relative humidity % Capacitive transducer ± 1.5% RH at 15 ÷ variables. The typical year is generated following the procedure
35 ◦ C
described in EN ISO 15927–4. The definition is based on the setting of
Atmospheric hPa Piezoresistive ±0.5 hPa at 20 ◦ C
pressure transducer the latitude and longitude of the place and on the selection of the time
period to be used. The period 2007–2016 was chosen, since it is the most
recent available. Among the output options, the TMY-tool allows
evaluating what differences in term of the main weather parameter can obtaining the data set directly in the EPW format. The solar radiation
influence the building energy simulation. On the basis of recent de­ data derive from the PVGIS-ERA5 database [62] and the other meteo­
velopments on this matter, first of all the methodologies of the typical rological variables are obtained from the ERA-Inteirm reanalysis [63].
year have been implemented, e.g. a mild year over the entire monitoring ERA-Interim and ERA5 are two datasets of the European Center for
period. The simplest is the methodology IGDG [6]. In this case, the Medium-Range Weather Forecasts (ECMWF) [64]. ECMWF uses its
construction of the typical year is based only on the dry bulb tempera­ forecast models and data assimilation systems to ’reanalyze’ archived
ture values by means of the calculation of its average value and variance observations, creating global data sets describing the recent history of
for each month of the year on the entire population. The same operation the atmosphere, land surface, and oceans. ERA-5 was produced by
is repeated for each month of every single year. The most representative ECMWF and obtained from the Copernicus Climate Change Service In­
typical month is then selected with the average value and variance of the formation website. ERA5 includes a detailed record of the global at­
air temperature closest to the values calculated for that month for the mosphere, the earth’s surface and ocean waves from 1950 onwards. This
entire population. The typical year is finally constructed as a composi­ new reanalysis replaces the ERA-Interim reanalysis (1979 to 2019).
tion of the typical months. Among the most widespread methodologies Among the key features of ERA5 is a horizontal resolution of 31 km,
for the construction of the typical year, the Sandia method [8] was significantly improved than the 80 km of ERA-Interim. For simplicity, in
implemented for the construction of the typical meteorological year the following discussion the two external climatic files will be indicated
(TMY) for Benevento. The modified versions of the aforementioned respectively as TRY_CTI and PVGIS.
method were also examined, e.g. the methodologies relating to the The second step is focused on the necessity to take into account the
construction of the TMY2 and IWEC datasets [9–10]. The Sandia method effects of the strong climatic changes; this has led to consider the con­
and the derived ones envisage the use of several indices obtained from struction of two extreme meteorological years: an extreme cold year
the hourly values of some meteorological variables. The procedure for (XMYmin) based on the minimum average daily dry bulb temperature in
selecting the typical months is based on the Finkelstein-Schafer statistic, each month and an extreme hot year (XMYmax) based on the maximum
with the application of a weighting set for the considered indices. A average daily dry bulb temperature in each month.
further method implemented for the typical year led to the construction Moreover, starting from the TMY (Sandia method) built for Benev­
of the test reference year (TRY) by adopting the procedure described in ento, some future climate projections were generated on medium (2050
technical standard EN ISO 15927–4 [12]. Also in this case the choice of s) and long (2080 s) term. The future climate files of the two meteoro­
the typical months is based on the Finkelstein-Schafer statistic but the logical generators CCWorldWeatherGen and WeatherShift ™ [33,37]
meteorological parameters evaluated are not subjected to a weighting

Fig. 2. Location of Benevento and Airola weather stations.

6
R.F. De Masi et al. Applied Energy 303 (2021) 117584

were compared, both based on the statistical downscaling performed the choices done considering the current climatic conditions will be also
through the “morphing method” [32] but with important differences in appropriate to respond to climate changes. In other words, the aim is to
terms of global circulation models (GCMs) and of the IPCC emission find a method for evaluating the resilience of a refurbishment design
scenarios. that can guide the choices of designer and owner thinking about the
Considering all these weather files, the first quantitative analysis is building of the future, without necessarily requiring further
the comparison of the different generated current and future climatic interventions.
conditions. This comparison regards the main meteorological variables One limitation of the proposed approach could be the extension of
in term of statistical distribution of average and extreme values. More­ adopted data. Indeed, many past studies used the past 10–20 years’ data.
over, a rigorous method for calculating heating and cooling degree days However, the paper wants propose a methodology and it is aimed to
is implemented for a preliminary evaluation of the incidence on the start a discussion on the reliability of adopted data and the main dif­
building energy consumption. Furthermore, a method for the agreement ferences that could occur with different methodology and datasets.
assessment between some climate files generated by the data of the However, when other data for Benevento will be available the proposed
Benevento weather station, “Urban weather data files”, and those weather file will be updated. Another critical point is the fact that the
coming from external weather data sources is proposed. case study is developed for one city; however, it wants to be remarked
The third step is the study of the impact of different data sources and that the application is proposed for discussing the implications of the
climatic projections on the selection of energy efficiency measures for developed methodology that can be implemented for further studying
the refurbishment of a residential building representative of the Italian with more representative zones and building kind of use. Moreover, the
building stock of the 1980 s. Therefore, the numerical model of a single- findings can be generalized to the Mediterranean area and to energy
family house was built with reference to the information contained in retrofit design for climate in which the cooling energy consumptions are
the TABULA project [65–66]. The model obtained was further custom­ comparable or higher than heating energy need.
ized to make it suitable for a typical Mediterranean climate such as that
of Benevento. The model is simulated by means of EnergyPlus 8.9 [67], 3. Data and methods for weather file definition
through the graphical interface DesignBuilder [68]. More in detail, first
of all, the effect of the different weather files is evaluated on the state of An EPW file (EnergyPlus / ESP-r Weather format) needs a lot of in­
fact (base case) of the building. This allows to highlight if some meth­ formation on the climate of the location, based on the weather station or
odology determines particularly unfavorable scenarios during the climate model from which the weather data is derived. Not all the in­
summer and winter period and what may be the mistake committed by formation provided by the EPW format are mandatory for the use in the
the designers in using more or less updated data. building energy simulation (BES). It is certainly essential to have the
Then, the energy retrofit is evaluated by combining six energy effi­ hourly values, for an entire year (8760 values), of the main meteoro­
ciency measures on the building envelope. The energy needs for heating logical variables that influence the energy needs of a building. In this
and cooling and the global primary energy are compared, also consid­ section the adopted data are illustrated as well as the procedures for the
ering the reference base case. The aim is the evaluation of possible dif­ weather file construction.
ference in the prevision of energy saving when different data sources are
used by designers but also the evaluation of building resilience. Indeed, 3.1. Climatic data
the availability of current and future climate files allows a sensitivity
analysis on the behavior of different retrofit solutions when the external Thanks to the external sensors of the test-room MATRIX, for the
solicitation changes compared with the moment during which the purposes of this study it was possible to have the following meteoro­
design has been realized. This analysis can suggest to designers and logical variables on an hourly basis:
researcher the aspect to be focused for improving the building resilience
in a typical Mediterranean climate. Indeed, it has been demonstrated by - Dry bulb temperature – DBT [◦ C];
Shen et al. [69] that global climate change will alter the optimal solution - Dew point temperature – DPT [◦ C];
of energy retrofit and its influence varies from building to building, - Global horizontal radiation - GHI [W/m2];
location to location. - Relative humidity – RH [%];
More in detail, two comparisons are proposed. The first one, ac­ - Atmospheric pressure – AP [Pa];
cording to a traditional approach, consists in the calculation of the en­ - Wind speed – wS [m/s];
ergy saving considering in both simulations the same weather file; for - Wind direction – wD [degrees];
instance, the base case simulated with TMY is compared, for the previ­ - Precipitation – P [mm].
sion of the energy improvement, with the refurbished configuration
simulated with TMY. More in detail, in the discussion, ΔEP will be the In addition to the global horizontal radiation (GHI), its diffuse
primary energy saving considering only the heating and cooling ser­ component (DHI) and the direct normal solar radiation (DNI) are
vices; ΔEH and ΔEC will indicate the reduction of the energy need for needed. A simplified method for calculating DHI and DNI from GHI is
heating and cooling respectively. In this case, the designer can know the used in this study. This method starts from the consideration that when
effect of considered efficiency measures on the near term and that is, DHI is not directly measured, it can be calculated by the estimation of
until the environmental conditions can be described with the weather diffuse fraction of GHI (Kd) [70], where:
file, without considering extreme events or climate changes.
Kd = DHI/GHI (1)
The second approach, more appropriate for the evaluation of the
resilience of designed interventions, is based on the estimation of the This ratio depends by the clearness index, Kt, that represenst the
energy saving on the long term; this means that the base case simulated, percentage of extra-atmospheric radiant energy that reaches the ground.
for instance, with TMY weather file, is compared with the refurbished Various models are available in literature for the calculation of Kd
scenarios simulated with the weather file created by projections of starting from the clearness index. Those obtained for the Mediterranean
climate change. For this discussion, ΔEPcl,ch will be the primary energy area treat monthly average data rather than single daily or hourly
saving considering only the heating and cooling services; ΔEH,cl,ch and values. For the present study, the Erbs regression model [71] is used; it
ΔEC,cl,ch will indicate the reduction of the energy need for heating and allows the hourly evaluation of the diffuse fraction of GHI.
cooling respectively. The Erbs correlations used are as follows:
This type of comparison which is not common in the scientific
Kd = 1.0 − 0.09∙Kt , 0 ≤ Kt ≤ 0.22 (2)
literature and in the design approach, can be helpful for understanding if

7
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Table 4
Kd = 0.9511 − 0.1604∙Kt + 4.388∙Kt2 − 16.638∙Kt3 + 12.336∙Kt4 , 0.22 ≤ Kt
Weighting factors in different TMY methods.
≤ 0.80
INDEX TMY TMY2 IWEC
(3)
Maximum dry bulb temperature 1/24 1/20 5/100
Minimum dry bulb temperature 1/24 1/20 5/100
Kd = 0.165, Kt > 0.80 (4) Mean dry bulb temperature 2/24 2/20 30/100
Maximum dew point temperature 1/24 1/20 2.5/100
Furthermore, the technical relations (astronomical and geometric)
Minimum dew point temperature 1/24 1/20 2.5/100
that allowed the calculation of the hourly clearness index, starting from Mean dew point temperature 2/24 2/20 5/100
the hourly values of global horizontal radiation, are reported in Table 3. Maximum wind speed 2/24 1/20 5/100
Finally, downstream of the calculation of the diffuse fraction of GHI Mean wind speed 2/24 1/20 5/100
and of the calculation of DHI, the direct normal radiation (DNI) is ob­ Horizontal solar radiation 12/24 5/20 40/100
Direct normal solar radiation 5/20
tained from the following equation [72]:
– –

GHI = DNI∙cosḮz + DHI (5)

1 ∑m
FS = δk (7)
3.2. Procedures for developing TMY, TMY2 and IWEC m k=1

where m is the number of days during the month and δk is the ab­
TMY, TMY2 and IWEC methodologies are considered for the con­
solute difference between long-term CDF and short-term (month under
struction of an EPW file for Benevento. In Table 4 the indices and
study) CDF at xk.
weighting values for the TMY, TMY2 and IWEC datasets are compared.
The FS obtained for each single year of monitoring are multiplied by
In all cases, the horizontal solar radiation has the maximum impact
the corresponding weights, obtaining the weighted statistic (WS):
meanwhile the direct normal solar radiation appears only in the TMY2.
The mean dry bulb temperature has higher weight compared with the N
∑ index

maximum and minimum value. WS = Wj (FS)j (8)

In all three cases, the reference procedure is the Sandia method,
j=1

proposed below: where Wj is the weight assigned to the j-th index.

Step 1 - For each index and for each month, the long-term cumulative Step 2 - From the lowest WS values, five months are identified for the
distribution function (CDF) and the short-term CDFs are calculated. The choice of the typical month and these are classified on the basis of their
long-term CDF uses one month of the full set of years (2015–2020 in this closeness to the long-term mean and median values of the WS.
study), while the short-term CDF uses data from one month of just one Step 3 - The persistence of mean dry bulb temperature and daily
year. The CDF for each variable is calculated by: global horizontal radiation are evaluated by determining the frequency
⎧ and length of runs of consecutive days with values above and below
⎨k − 0.5 0, x < x(1)
fixed long-term percentiles. For the mean daily dry bulb temperature,
Sn (x) = , xk ≤ x ≤ x(k+1) (6)
⎩ n
1, x ≥ x runs above the 67th percentile (consecutive warm days) and below the
(n)
33rd percentile (consecutive cool days) are determined. For the global
where n is the total number of elements and k is the ranked order horizontal radiation, the runs below the 33rd percentile (consecutive
number (k = 1, 2, 3, …, n-1). Next, for the choice of each typical month, low radiation days) are determined. The persistence criterion excludes
the difference between the short-term CDFs and the long-term CDF is the month with the longest run, the month with the most runs, and the
calculated for each index using Finkelstein-Schafer statistics (FS): month with zero runs. The data are used to select from the five candi­
dates the month to be used in the TMY. The highest-ranked candidate
Table 3 month from Step 2 that meets the persistence criteria is used in the TMY.
Correlations used for decomposition of hourly global horizontal radiation. For TMY2, as a consequence of adding the index for direct normal ra­
Ref. diation, the persistence check in Step 3 is modified to determine the
frequency and run length below the 33rd percentile (consecutive low
[70] GHI Iex: Extraterrestrial horizontal
Kt =
Iex
; (5)
radiation (W/m2). radiation days) for daily values of direct normal radiation.
[72] Iex = cosḮz ∙Ien ; (6) Ien: extraterrestrial direct Step 4 - The 12 selected months were concatenated to make a com­
normal radiation (W/m2) plete year and discontinuities at the month interfaces were smoothed for
ϴZ : zenith angle (degrees) 6 h using curve fitting techniques.
[78] cosḮz = sinφ∙sinφ + cosδ∙cosφ∙cosω; (7) δ: declination (degrees); φ:
latitude of the weather site
(degrees); ω: hour angle
(degrees). 3.3. Method according to EN ISO 15927–4
[ ]
[72] 360 n: day of the year
δ = 23.45∙sin ∙(n + 284) ; (8)
365 The typical year built with this methodology of the standard EN ISO
[79] 360 SOT = Solar time (h).
ω =
24
∙(SOT − 12); (9) 15927–4 [13] is named test reference year (TRY). The algorithm starts
[79]
SOT = LST +
(λ − λR )
+ EOT; (10)
LST: Local standard time (h); with the identification of a set of primary climatic parameters p (dry
15 λ: Longitude of the weather site bulb temperature, horizontal global solar radiation and relative hu­
(degrees);
midity, as suggested by the technical standard). The wind speed is
λR: Longitude of the time zone
in which the weather site is considered a secondary parameter.
situated (degrees); For the three primary climatic parameters, the following operations
EOT: Equation of time (h) are applied:
[79] EOT = d’: day angle (degrees)
− 0.128∙sin(d’ − 2.8)0.165∙sin(2∙d’ +
19.7); (11)
• calculation of the daily average value for each month m of each year
[79] 360 n: day of the year
y belonging to the historical records under examination;
d’ = n∙ ; (12) • calculation of the normalized cumulative distribution of daily
365 ( )
[72] 360 Isc : Solar constant = 1367 W/ average values for each month across all years of the historical record
Ien = Isc ∙(1 + 0.033∙cos n∙ ; (13)
365 m2
(long term CDF):

8
R.F. De Masi et al. Applied Energy 303 (2021) 117584

K(i) contained in the current file given as input. Regarding the reference
ϕ(p, m, i) = (9)
N+1 interval of HadCM3 data, the authors specify that the use of EPW files
with data subsequent to the period from 1961 to 1990 leads to over­
where K(i) is the rank order of the ith day and N is the total number of
estimate climate change impacts. Starting from the typical meteoro­
days for the month m over all the available years;
logical year realized for Benevento (TMY), in this study two future
projections were generated using this tool, one in the medium term
• for each year, calculation of the cumulative distribution of the daily
(2050 s), indicated as CC_MT, and another long term (2080 s) indicated
mean values within each month (short term CDF), by sorting all
as CC_LT.
obtained values with increasing order for that month and that year
by using equation (10):
3.4.2. Weathershift TM
J(i) Weathershift is a generator of future meteorological files and pro­
F(p, y, m, i) = (10)
n+1 vides for the application of the morphing method on the offsets of 14
GCM (out of a total of 40 models). In detail, the models used are relative
where J(i) is the rank order of the i-th value of the daily mean values
within that month and that year and n is the number of days for the to the phase 5 of the Coupled Model Intercomparison (CMIP5) project.
Offsets are available for eight time periods, from the short term
specific month;
(2011–2030) to the long term (2081–2100), relative to the 1976–2005
baseline period and for the representative concentration paths (RCPs)
• for each month, calculation of the Finkelstein-Schafer statistical
4.5 and 8.5. As reported in [37], the four closest grid points of the model
parameter:
are considered for each location and the offsets are calculated by per­
∑ forming a bilinear interpolation. The use of multiple models allows
n
FS(p, y, m) = |F(p, y, m, i) − ϕ(p, m, i)| (11)
i=1
generating a cumulative distribution function (CDF) for each variable
using linear interpolation between the model values. Because the offset
values for each model is linked with each other, a physically consistent
• for each month and for each primary climatic parameter, composi­ relationship between the values of the variables is maintained. In this
tion of the ranking list of years based on the increasing value of the study, future climate projections are generated for the time period
Finkelstein-Schafer statistic; 2041–2060 (referred as 2050 s), for RCP 4.5 and RCP 8.5, and using the
• for the three months with the lowest total ranking, calculation of the offsets on mean daily temperature change of the 10th, 50th and 90th
deviation of the monthly mean wind speed from the corresponding percentiles. These six future projections within the study are indicated as
multiyear calendar-month mean. The month with the lowest devia­ W_45-10, W_45-50, W_90-10 for the scenario RCP 4.5 and as W_85-10,
tion in wind speed is selected as the best month for the TRY. W_85-50 and W_85-90 for RCP 8.5. It is important to clarify that, unlike
CCWorldWeatherGen, in this case a free tool is not made available to
Finally, the standard provides for the adjustment of weather users, but it is possible to use the Weathershift online interface [73] to
parameter hourly values in the last 8 h of each month and in the first 8 h generate future climate projections for the selected location or starting
of the following month by interpolation, to ensure a smooth transition from the EPW file that the user provides as input.
when the individual months are merged to form the reference year.
The test reference year made for Benevento in this study was indi­ 3.4.3. Important considerations on future weather data
cated as TRY_BN. With both used generators, there is an overestimation of climate
change since, for the TMY variables subjected to the morphing method,
3.4. Future weather data the offsets refer to periods earlier than the monitoring in the urban
station (2015–2020). This is mainly true in the case of future climate
Two tools are used starting from the typical meteorological year projections obtained with CCWorldWeatherGen, since it refers to the
(TMY) built for Benevento. Both tools perform a statistical downscaling baseline period 2061–2090 of the HadCM3 model (Table 5).
based on the so called ’morphing’ methodology for climate change Furthermore, it should be remembered that WeathershiftTM refers to
transformation of weather data, developed by Belcher, Hacker and two representative concentration paths, RCP 4.5, an intermediate sce­
Powell [32]. It can be preliminarily concluded that although the current nario (substantial reduction of emissions), and RCP 8.5 (business as
climate files are not adequate to assess the effects of climate change on usual), high emissions scenario. In general, the RCPs – originally
the energy needs of buildings, their use currently remains the starting RCP2.6, RCP4.5, RCP6, and RCP8.5 – are labelled after a possible range
point for the generation of future climate projections to be used in BES of radiative forcing values in the year 2100 (2.6, 4.5, 6, and 8.5 W/m2,
tools. respectively). The radiative forcing, as defined in AR4, is the measure of
the influence that a factor has in altering the balance of energy entering
3.4.1. CCWorldWeatherGen tool and leaving the earth and atmospheric system. This index indicates the
The climate change world weather file generator (CCWorldWea­ importance of the factor as potential climate change mechanism. The
therGen) changes the current EPW files into EPW or TMY2 climate values of the radiative forcing refer to 1750, pre-industrial conditions.
change weather files compatible with most building performance The RCP 8.5 scenario is certainly more comparable to the A2 (SRES)
simulation programs. The CCWorldWeatherGen uses the IPCC TAR scenario to which the CCWorldWeatherGen tool refers. In particular, the
model summary data of the HadCM3 A2 experiment which is available A2 storyline and scenario describes a very heterogeneous world. The
from the IPCC DDC [36]. As reported in [35], the HadCM3 A2 summary
data is provided as monthly values for each grid point of the HadCM3
data grid for a simulated 1961–1990 baseline climate and for three Table 5
future time slices, the 2020 s, 2050 s and 2080 s [20]. As three HadCM3 Generated future projections and the characteristics of the meteorological
generators.
A2 model experiments are available for each time slice, average values
are generated for each climate parameter prior to conducting the CCWorldWeatherGen WeatherShiftTM
‘morphing’. This also helps to overcome the problem of data gaps in the Projected time period 2050, 2080 2080
individual experiments. In detail, the CCWorldWeatherGen tool loads the IPCC scenario(s) A2(SRES) RCP 4.5, RCP 8.5
results of the HadCM3 A2a, A2b and A2c experiments for the four grid GCM(s) HadCM3 10th, 50th, 90th percentile
Baseline period 1961–1991 1976–2005
points closest to the point identified by the longitude and latitude

9
R.F. De Masi et al. Applied Energy 303 (2021) 117584

underlying theme is self-reliance and preservation of local identities. n ⃒ ⃒

Fertility patterns across regions converge very slowly, which results in 1 ∑ ⃒̂ ⃒
MAD = × ⃒ Y i − Yi ⃒ (15)
continuously increasing global population. Economic development is n i=1
primarily regionally oriented and per capita economic growth and
∑n ⃒⃒ ⃒
⃒
technological change are more fragmented and slower than in other ̂
i=1 ⃒ Y i − Yi ⃒
storylines. MAPD = ∑n × 100 (16)
i=1 Yi

where n is the number of months considered in each season, while Yi

3.5. HDD and CDD and Ŷi are, respectively, the reference average monthly value (calculated
for TMY or TRY_BN) and the average monthly value obtained the other
In order to compare the hourly temperature values of the examined two files.
weather files, the degree days for heating and cooling are calculated.
The method reported in the standard UNI 10349–3 [74] is used. It is 4. Case study
based on the calculation of the heating degree day (HDD) and the
cooling degree day (CDD) by summing the positive differences between The section describes the main characteristics of the adopted case
the base temperature and the hourly external temperature (degree- study and the designed energy efficiency measures.
hours) and then dividing by 24. The daily degree days are then summed
over the desired period. The following equations are used:
( ) 4.1. Base case
∑D
1 ∑ 24
( )+
HDD = Tref − Text,h (12)
d=1
24 h=1 The case study is a single-family house representative of the Italian
building stock of the 1980 s. The case study was carried out using the
d

( ) TABULA web-tool [66]. The building is representative of the middle

∑
D
1 ∑ 24
( )+
CDD = Text,h − Tref (13) climatic zone (from 2100 to 3000 HDDs) and of the sixth building age
d=1
24 h=1
d class (from 1976 to 1990).
The structure is composed by three levels: a semi-basement, used as
where Tref is the reference temperature assumed 20 ◦ C during the
garage, and two floors above the ground. The building envelope is
heating season and 26 ◦ C during the summer period; Text,h is the external
characterized by a pitched roof with a brick-concrete slabs, brick wall
mean hourly temperature; D is the total number of days (d) that are
and double glazed windows with air gap and wood frame. Table 6
calculated considering the heating /cooling season for the considered
summarizes the thickness (thtot) and thermal transmittance of the wall
climatic zone.
(Uw), the roof (Ur) and the slab on the ground (Us); it also indicates the
solar factor (SF) and the glass thermal transmittance (Ug) of the win­
3.6. Proposed method for agreement assessment dows. The insulation level is very low if it is compared with the current
Italian requirements. For instance, for the climatic zone Benevento (C
The definition of current climate files with reliable hourly climate zone), if the building was undergone to retrofit intervention, the
data is a critical issue. In addition to the importance of using recent data, threshold value of thermal transmittance should be 0.34 W/(m2 K) for
it is necessary that these would be collected in a meteorological station external walls and 0.33 W/(m2 K) for the roof [75].
close to the building for which the energy needs must be analyzed. Thus, Regarding the installed systems, a gas boiler (poor efficiency) with
when a multi-year meteorological monitoring is not available, the atmospheric burner and chimney lower than 10 m, placed in an un­
designer uses comparable weather data sources. For this reason, the two heated room, is used for heating service. It produces hot water at 80 ◦ C
climate files of Benevento deriving respectively from the TRYs made for radiators placed in every room. The domestic hot water is produced
available in 2016 by the CTI and by the TMY-tool of the PVGIS was by the same boiler coupled with a storage tank with medium level of
considered. Before comparing the effects on building energy simulation, insulation. The total efficiency of the gas boiler was set equal to 0.9. For
it is evaluated the consistency of the most important meteorological the summer air conditioning, it has been assumed that in the main room
variables in comparison with the weather file defined by means of a split-system has been installed with a nominal cooling capacity of 3.25
monitored data at the urban station of Benevento. kW and a nominal performance coefficient equals to 3.0. The lighting
In detail, the proposed method is based on the analysis of the hourly system consists of fluorescent lamps with a specific rated power of 9.9
records of dry bulb temperature (DBT) and global horizontal irradiation W/m2.
(GHI), with the following indices: These data have been used for defining a numerical model simulated
with Energy Plus v 8.9. In the heating season, a set point temperature of
1) DBTd,ave,w: average daily dry bulb temperature in daytime hours 20 ◦ C has been set with an operating period (10 h per day) from
(from 9:00 to 19:00) during the winter months (from November to November 15th to March 30th. In the cooling season the set-point
March), [◦ C]; temperature is set to 26 ◦ C and a cooling period is assumed from 1st
2) DBTd,ave,s: average daily dry bulb temperature in daytime hours June from 30th September (10 h per day).
(from 9:00 to 19:00) during the summer months (from June to
September), [◦ C]; Table 6
3) GHId,tot,s: total daily global horizontal irradiation during the summer Building envelope data (thickness and thermal transmittance).
months (from June to September); SINGLE- Building Building envelope
FAMILY age
The agreement assessment was based on the calculation of the bias, HOUSE class
(SFH)
the mean absolute difference (MAD) and the mean absolute percentage Walls Roof Basefloor Windows
difference (MAPD):
TABULA II (1976 – thtot = thtot = thtot = 0.28 SF = 0.75
1 ∑n ( ) web-tool 1990) 0.41 m 0.27 m m
bias = × ̂ i − Yi
Y (14) Uw = Ur = Us = 1.04 Ug = 2.08
n i=1
0.74 W/ 1.20 W/ W/m2 K1 W/m2 K1
m 2 K1 m 2 K1

10
R.F. De Masi et al. Applied Energy 303 (2021) 117584

The building’ model rendering is shown in Fig. 3 where the main Table 7
geometrical data are also reported as the surface to volume ratio (S/V). Description and properties of energy retrofit measures.
Moreover, it is reported the gross opaque and glazed surface for each N ID Description Properties
exposure. It is considered the state of fact of building (base case). measure

1 TI Increase in the level of thermal • Rookwool: λ = 0.04 W/

4.2. Energy retrofit measures insulation of the building envelope m K; ρ
using: • EPS: λ = 0.04 W/m K; ρ
- Rockwool for external walls Benefit on Building
Several passive measures have been studied for the energy refur­ - Expanded Polystyrene for roof envelope:
bishment of the opaque and glazed envelope. First of all, the thermal - Uw = 0.34 W/m2 K
insulation of the external walls and the roof is considered. In both cases, - Ur = 0.33 W/m2 K
2 RP Application of high-reflective paint Reflective paint:solar
the thermal insulation layer was applied on the external side; it consists
on roof and increase the level of reflectance = 0.83;
of rockwool panels for the walls and expanded polystyrene (EPS) for the roof thermal insulation using EPS thermal emissivity = 0.9
roof. The second intervention consists in the application of a high- 3 DBL _L Double glazing with low emissivity Ug = 1.00 W/m2 K
reflective paint on the external roof’ surface. In this case, with the aim coating LT = 74 %
to contain the heat losses during the wintertime [75], also the applica­ SF = 54%
4 DBL_S Double glazing with solar control Ug = 1.00 W/m2 K
tion of the insulation has been simultaneously considered. For the coating LT = 70 %
windows, two alternatives are evaluated: low emissive or solar control SF = 0.33%
double glazing system. Moreover, the integration of external fixed and 5 IS Internal shielding systems Blind with high reflectivity
internal highly reflective mobile shielding systems is evaluated. All in­ slats with seasonal
schedule
terventions have been chosen in accordance to Italian Ministerial Decree
6 ES External shielding systems Horizontal protrusions,
26/06/2015 [76], for fulfill the threshold values for the thermos- 1.0 m Overhang
physical parameters. These measures are described in detail in
Table 7; herein LT indicate the transmission in the visual range.
The proposed efficiency measures have been combined and thus 9
Table 8
energy retrofit scenarios (RS) have been simulated. These are summa­ Retrofit scenarios: composition and acronyms.
rized in Table 8.
N. ID Composition

5. Results and discussion 1 RS1 RP + DBL_S + IS

2 RS2 RP + DBL_S + ES
3 RS3 TI + DBL_L
5.1. Application of different methodologies for the weather file definition 4 RS4 TI + DBL_L + RP
5 RS5 TI + DBL_S + RP
Some preliminary evaluations are needed on the data collected be­ 6 RS6 TI + DBL_L + RP + IS
tween 2015 and 2020 at the Benevento weather station. Fig. 4 shows the 7 RS7 TI + DBL_S + RP + IS
8 RS8 TI + DBL_L + RP + ES
monthly average values of the dry bulb temperature for each year of 9 RS9 TI + DBL_S + RP + ES
monitoring and over the entire population (long-term).
This trend helps to understand the implications of the methodologies
(as IGDG and XMY) based exclusively on the comparison of the dry bulb values is due to temporal phases not included in the multi-year moni­
temperature values or when the extreme conditions are excluded. For toring or to anomalies in the data detection of the weather station. The
instance, in January the extreme monthly average temperatures are variance (σ2) identifies the dispersion of the hourly temperature values
those relating to the years 2018 and 2015, respectively equal to 7.9 ◦ C around the calculated monthly average value. The smaller variance in­
and 3.7 ◦ C; instead, the month closest to the long-term monthly average dicates that the values of the variable are concentrated around the mean
values is January 2020, with an average DBT of 5.9 ◦ C. This month is value. In some months the variance is very high, for instance in
chosen for the definition of the IGDG file and it is clear that there is an December and January mainly during the 2016 but also for the spring
important difference with the extreme conditions recorded during the months during the 2017. This means that the external conditions are
years. For the summer period, July is the month for which the mean extremely variable and the adoption of the mean value could be not
value is most comparable in the analyzed years except 2015 during representative of the real condition.
which there is the higher value (27.1 ◦ C). The value of σ2, for the same month, is usually different along the
Tables 9 shows the complete set of data for the choice of the typical year. The square root of the variance is the standard deviation; it is
months for the IGDG weather file. The months with the average value usually comparable along the year since, for instance during 2018 it
and variance of the dry bulb temperature closest to the values calculated varies from ± 4.3 in January to ± 5.5 in April. April, June and August are
for that month on the entire population, are chosen. The absence of some the months with the higher standard deviation.

Fig. 3. Rendering of the building model and general data of the case study.

11
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 4. Monthly average values of DBT for all monitoring years (2015–2020).

Table 9
Mean and variance values in each month of dry bulb temperature for each month.
Jan Feb Mar Apr Mag Jun Jul Aug Sep Oct Nov Dec
2015 µ [◦ C] 18.7 22.0 27.1 25.7 21.1 15.8 11.0 6.0
σ2 [◦ C2] – – – – 28.6 27.1 31.6 30.0 34.1 17.1 25.5 22.7
2016 µ [◦ C] 7.8 10.0 15.1 16.9 21.6 24.8 23.7 5.4
2
σ [◦ C2] 29.8 – 18.3 28.6 26.3 26.1 33.3 25.8 – – – 21.6
2017 µ [◦ C] 3.7 9.3 11.3 13.1 18.2 24.2 25.6 19.7 12.2 10.3 6.6
2
σ [◦ C2] 19.6 18.6 30.1 30.6 28.9 27.6 35.6 – 24.7 11.2 18.8 25.0
2018 µ [◦ C] 7.9 6.4 10.1 16.0 18.9 22.3 25.7 24.2 21.6 17.6 11.3 7.6
σ2 [◦ C2] 18.8 19.5 18.9 31.0 26.1 21.4 25.2 28.7 31.6 19.0 18.6 22.7
2019 µ [◦ C] 4.5 7.8 10.7 13.2 14.5 24.7 26.0 26.6 22.0 17.1 13.4 8.2
σ2 [◦ C2] 13.7 22.4 24.7 27.1 16.4 38.8 30.5 35.6 26.3 31.8 13.7 19.5
2020 µ [◦ C] 5.9 9.4 10.1 13.2 18.3 21.3 25.5 25.9 22.0 14.8 11.8 8.5
σ2 [◦ C2] 26.3 24.5 25.3 36.4 28.0 26.9 30.9 27.8 34.6 24.5 18.3 14.6
Long term µ [◦ C] 6.0 8.2 10.4 14.1 17.6 22.7 25.8 25.2 21.3 16.2 11.6 7.1
σ2 [◦ C2] 24.5 22.9 23.7 32.2 28.1 29.7 31.7 30.7 31.0 25.8 20.1 22.3

Fig. 5. Minimum daily mean values of DBT in each month − 2015–2020.

12
R.F. De Masi et al. Applied Energy 303 (2021) 117584

The choice of the most suitable months for the construction of Finally, Table 11 shows the composition of all meteorological years
extreme meteorological years is based on different criteria. (typical years and extreme years) that will be used in the further
Fig. 5 and Fig. 6 show respectively the minimum and maximum daily analysis.
mean values in each month, for each year of monitoring. In the case of It can be noted that it is not possible to establish a clear connection
the extreme cold year (XMYmin) the reference months are selected by between the different methodologies. There is not repeatability of the
different years in each month; for instance: January 2017 has the min­ year chosen as a reference for the various months. This already un­
imum value (-3.3 ◦ C), February 2018 (-1.86 ◦ C) and then March 2020. derlines that the energy simulations can give basically different results
The minimum value is very different along the years, for instance in depending on the chosen methodology since the external forcing could
January it has been 4.8 ◦ C during 2018 and thus, compared with the be not comparable.
value of 2017, there is high difference. Lower differences are recorded
during the summer period, considering July, the lowest value has been
5.2. Preliminary analyzes on current climate files
recorded during 2016 (18.2 ◦ C) and the higher (21.4) during 2017.
In the case of XMYmax, the difference in the monitored years are less
Before to analyze the impact in term of energy simulation, it is
marked. Considering July, the extreme values of DBT are 30.2 ◦ C in the
interesting the evaluation of the hourly distributions of the main
2015 and 27.9 ◦ C in the 2016; in December the maximum value
meteorological variables of the current climate files (typical and
(15.8 ◦ C) has been recorded during the 2017 and the minimum (9.5 ◦ C)
extremes).
during the 2016.
The first comparison concerns the methodologies used for the con­
Another interesting comparison is between the methodologies based
struction of the “urban” climate files of Benevento, and thus the files
on the original and modified Sandia method (TMY, IWEC and TMY2
obtained from the monitored data by MATRIX weather station. The box
files) and the methodology of the EN ISO 15927–4. In both cases, the
plots in Fig. 8 and Fig. 9 show, respectively, the distributions of the
Finkelstein–Schafer statistic is applied, but, as extensively described in
hourly values of DBT and GHI. The first unexpected conclusion is that
the previous sections, the approach is different, both for the number and
the statistical distributions seem comparable in both cases, also if the
type of climatic parameters and for the steps that lead to the choice of
base months are different. However, it is important to remark that these
each typical month. For example, a comparison is shown on the choice of
are global distributions and thus the extreme values can occur in
the month of June. Fig. 7 reports the CDF curves for the main parameters
different hours, days or period and this influences the resolution of the
considered with EN ISO 15927–4 standard. Compared to the long-term
heat balance on the building.
CDF, only the CDFs of the years with lowest and highest FS are shown.
Considering the temperature, the biggest differences, as expected,
June 2015 and June 2018 are therefore the two candidate months.
concern the XMYs. In the case of XMYmax the average is equal to 16.1 ◦ C
Finally, the final comparison on the average values of wind speed, it is
and it is the highest between the methodologies. This distribution has a
1.08 m/s during June 2015 and 1.20 m/s during June 2018 meanwhile
maximum value of 41.6 ◦ C meanwhile between the other files the
the long term average value is 1.13 m/s. Thus, the choice falls on June
maximum is 39.5 ◦ C (without considering the abberant values). The
2015.
distribution of XMYmin indicates that the average value of DBT is 14.1 ◦ C
Table 10 shows the calculation of the FS in the case of TMY, TMY2
and the minimum value is − 7.6 ◦ C meanwhile in the other cases it varies
and IWEC methodologies, for all the considered parameters during June
from − 5.4 ◦ C (XMYmax) and − 3.5 ◦ C (IGDG).
of each year. More in detail, the calculation of FS for GHI and DNI is
Among the typical climatic years, TMY is the one that has a distri­
based on the total daily value. The calculation of DNI is used only in the
bution with greater variability over the year. The average value is equal
case of TMY2.
to 15.5 ◦ C while the maximum value and the minimum value is
It is noted that, compared to the previous method, the minimum
respectively 39.5 ◦ C and − 4.9 ◦ C. It can be concluded that with the
value of FS occurred in 2018 only for the average DBT while it never
Sandia method, the typical climatic year of Benevento is generated with
occurred in 2015. Subsequently, the calculation of the WS-weighted
more extreme conditions. TRY_BN and IGDG have the same maximum
statistic and the evaluation of the persistence criteria determine the
and minimum values and almost coincident average value. However,
exclusion of June 2018 and June 2015 from the candidate months; thus
IGDG has a distribution characterized by greater variability between the
for these methodologies the representative month is June 2020.
first and third quartiles. As regards the global solar radiation on the

Fig. 6. Maximum daily mean values of DBT in each month − 2015–2020.

13
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 7. CDF in June - EN ISO 15927–4: a) DBT; b) RH; c) GHI.

Table 10
FS for weather indices in TMY, TMY2 and IWEC.
FS DBT DPT WS GHI DNI

AVG MIN MAX AVG MIN MAX AVG MAX TOT TOT

2015 0.099 0.115 0.087 0.150 0.108 0.186 0.135 0.048 0.076 0.114
2016 0.099 0.054 0.095 0.054 0.078 0.046 0.034 0.033 0.067 0.076
2017 0.185 0.125 0.153 0.053 0.030 0.038 0.105 0.160 0.135 0.121
2018 0.065 0.072 0.079 0.081 0.140 0.059 0.082 0.054 0.135 0.164
2019 0.252 0.146 0.255 0.176 0.086 0.136 0.241 0.185 0.051 0.083
2020 0.141 0.115 0.143 0.047 0.100 0.120 0.164 0.139 0.073 0.055
min 0.065 0.054 0.079 0.047 0.030 0.038 0.034 0.033 0.051 0.055

regards CDDs, an anomalous result concerns the value of approximately

Table 11
21 ◦ C day obtained for PVGIS which is really different than the average
Composition of typical and extreme meteorological years of Benevento.
value obtained with the other typical climatic years. This indicates that,
Month TMY TMY2 IWEC IGDG TRY BN XMYmin XMYmax in the summer period, the air temperature values of this dataset rarely
1 2017 2020 2020 2020 2020 2017 2016 exceed 26 ◦ C. Another interesting observation can be made by looking at
2 2020 2019 2017 2019 2017 2018 2019 the CDDs obtained for TRY_CTI and TRY_BN. Despite the different
3 2019 2019 2019 2019 2020 2020 2017
monitoring behind these two typical years, their values are quite com­
4 2017 2019 2020 2017 2016 2020 2019
5 2015 2015 2015 2020 2017 2019 2020 parable but at the same time these are really different from the other
6 2020 2020 2020 2015 2015 2016 2019 methodologies. Thus, for this case study, it seems that the methodology
7 2017 2020 2016 2020 2020 2016 2020 in accordance with EN ISO 15927–4 could lead to the choice of typical
8 2019 2015 2019 2015 2016 2020 2019 months with less critical temperature values, mainly in summer. The
9 2015 2020 2019 2018 2017 2018 2015
highest value, (269 HDDA), is obtained with the DBT of the extreme hot
10 2017 2020 2017 2020 2015 2017 2020
11 2018 2018 2018 2018 2018 2017 2019 year, while with TMY there is the highest value among the typical years
12 2017 2017 2017 2018 2018 2016 2017 (209 ◦ C day).
Fig. 11show the monthly average values of DBTd,ave,w, DBTd,ave,s e
GHId,tot,s , for the 4 climate files for which the agreement assessment
horizontal plane, XMYmin and XMYmax, whose composition was based
between “external” and “urban” weather datasets is studied.
exclusively on air temperature values, have the lowest average values,
The bias, MAD and MAPD values are reported in Table 12. As regards
respectively equal to about 157 W/m2 and 164 W/m2. In all other cases,
the winter season, compared to TMY both TRY_CTI and PVGIS have a
the distribution of GHI is comparable with the average value that varies
positive bias instead using as comparison TRY_BN the bias is equal to
from 162 W/m2 for TMY2 and 168.2 W/m2 for IWEC. The highest value
− 0.3 ◦ C for TRY_CTI and to 0.3 ◦ C for PVGIS. In both cases PVGIS has a
is equal to 706 W/m2 and concerns both the IGDG and IWEC distribu­
positive bias and this could justify the lower value of the heating degree
tion, if the aberrant values are excluded. TMY, TMY2 and IWEC show,
days compared to other typical years. Moreover, compared to the Sandia
due to the same choice of the summer months n, a maximum value equal
method, the distortion is greater. This is clear for both external datasets
to 1081 W/m2.
in terms of MAD and MAPD since the MAPD is very high in both cases,
In order to compare the hourly temperature values, the degree days
equal to 20.6% for TRY_CTI and 29.6% for PVGIS. Considering the
for heating and cooling were calculated. Fig. 10a and Fig. 10b show the
summer season, for DBT the bias is always negative and the MAD is
values obtained for HDDs and CDDs respectively. The HDDs are com­
higher taking TMY as a reference.
parable between the typical climatic years obtained from the urban
The MAD obtained for PVGIS in comparison with TMY is equal to
station of Benevento (TMY, TMY2 and IWEC) as well as the value ob­
3.9 ◦ C while compared to TRY_BN it is equal to 2.7 ◦ C. These values seem
tained for IGDG is coincident with those indicated as TMY (1656 ◦ C
to confirm the result obtained in term of CDDs for PVGIS, since in the
day). The highest value, as expected, is obtained with the extreme
summertime the DBT is on average much lower than the reference
weather year based on the average minimum temperature, (1807 ◦ C
datasets. However, for the summer period it is also important to consider
day). Compared to TMY, the HDDs of the extreme cold year increase by
the GHI since this parameter significantly affects the energy needs of
9.2%, while in the case of PVGIS there is a reduction of 16.7%. As

14
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 8. DBT distribution - current weather files of Benevento.

Fig. 9. GHI distribution - current weather files of Benevento.

Fig. 10. Degree days: a) HDD; b) CDD.

buildings. From the average of the total daily values (Fig. 12c), it seems TRY_BN and respectively equal to 5.4% and 11.5% compared to TMY.
that, with the exception of June of TRY_BN, the two external datasets More in general, this analysis indicates that the external weather data­
have always a higher value than the reference data. This result was sets are characterized by dry bulb temperature distributions and global
unexpected after the observations made on the DBT in the summer horizontal radiation not in accordance with the distributions obtained
season. In terms of bias, the differences reported by TRY_CTI and PVGIS for the same climatic variables with the data collected in the urban
compared to the two reference files are comparable, while passing to the station by applying the Sandia method and the methodology required by
absolute values there are values of MAD and MAPD greater in the EN ISO 15927–4. The PVGIS file turned out to be less consistent despite
comparison with TRY_BN. In detail, the MAPD values of TRY_CTI and it has been generated by means of data the same geographical co­
PVGIS are respectively approximately 8.5% and 12.1% compared to ordinates of MATRIX. Both for TRY_CTI and for PVGIS it must be taken

15
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 11. Average monthly values: a) DBTd,ave,w; b) DBTd,ave,s; c) GHId,tot,s.

Table 12
Values of bias, MAD and MAPD for the assessment agreement + 9.
Bias MAD MAPD(%)

Ref: TRY_BN Ref: TMY_BN Ref: TRY_BN Ref: TMY_BN Ref: TRY_BN Ref: TMY_BN

CTI PVGIS CTI PVGIS CTI PVGIS CTI PVGIS CTI PVGIS CTI PVGIS

DBTd,ave,w − 0.3 0.3 0.3 0.9 1.1 1.9 2.2 3.1 9.5 16.7 20.9 29.6
DBTd,ave,s − 0.5 − 2.7 − 1.7 − 3.9 1.0 2.7 1.8 3.9 3.8 10.3 6.6 14.0
GHId,tot,s 254.4 608.5 310.4 664.5 499.5 707.9 310.4 664.5 8.5 12.1 5.4 11.5

Fig. 12. Comparison between the urban weather files and TRY_CTI - ΔTh,ave.

into account that the files have been generated starting with data file TRY_CTI this phenomenon is not taken into account.
collected in periods prior to the monitoring carried out at the urban Finally, in Fig. 13, a comparison of the wind speed distributions is
station. In the case of PVGIS, the incidence of the spatial resolution of proposed. Also in this case, there is a sensible difference with the data of
the GCM used must be considered, which inevitably leads to not not urban station. Indeed, the average value of ws for TRY_CTI is 2.35 m/
including phenomena related to the local microclimate. s while the average value for the climate files constructed for Benevento
In addition, a different level of urbanization that characterizes the varies between 1.16 m/s (TRY_BN) and 1.33 m/s (XMYmax). This result
meteorological station of CTI leads to a different representation of the is significant in term of external forcing because the wind speed affects
effects of the local microclimate. In this regard, Fig. 12 shows the the convective heat exchange as well as the infiltration and ventilation
average annual temperature difference ΔTh,ave; this index is the mean loads both in term of energy needs that in term of indoor comfort.
value of the positive hourly temperature difference calculated between All these observations highlight the importance of using reliable
the TRY_CTI (referred to Airola site) and all other files derived by MA­ datasets with recent data monitored in a site with characteristics com­
TRIX monitoring. parable to the place for which the building analysis must be developed.
This value exceeds 4 ◦ C for all climatic files of the urban station. The Indeed, this choice inevitably affects the building energy simulation and
maximum and minimum values occur with the extreme years and are contributes to the uncertainty in the future climate forecast.
respectively 5.0 ◦ C and 4.2 ◦ C. This result indirectly provides informa­
tion on the intensity of the urban heat island phenomenon that affects
5.3. Comparison of future weather files
the most urbanized areas and which in the summer season involves an
increase in the energy needs of buildings for cooling. Using the climate
Fig. 14 shows the comparison between the distribution of DBT in

16
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 13. Wind speed distributions - current weather files of Benevento.

Fig. 14. DBT distribution - current weather file (TMY) and future climate projections.

case of TMY and for the climatic projections. It is interesting to observe are equal to about 26.7 ◦ C and 29.4 ◦ C respectively. Instead this dif­
the variation in the average annual values that depends both by the ference is greatly reduced for the winter months; considering February,
effects of climate change but also by the uncertainties in the projections the average monthly DBT is 9.3 ◦ C, 10.6 ◦ C and 11.7 ◦ C respectively for
of the different models considered for the same time horizon. TMY, CC_MT, CC_LT. More in general, if the temperature differences are
In the case of the medium-term (2050 s) and long-term (2080 s) calculated with the average monthly values there is a proportional in­
projections obtained by CCWorldWeatherGen, indicated respectively crease passing from medium term and long term scenarios, with the
with CC_MT and CC_LT, the average temperatures are respectively evident effect of the global warming. Briefly, the temperature difference
17.2 ◦ C and 18.8 ◦ C, while in the case of TMY this value is 15.5 ◦ C. In the between CC_MT and the TMY goes from 1.16 ◦ C (April) and 2.70 ◦ C
case of the medium-term projections generated using WeathershiftTM, (August) and it varies between 2.40 ◦ C (April) and 4.89 ◦ C (August)
there is a progressive increase in the distributions of DBT passing from when CC_LT is considered.
the scenarios of 10th to the 90th percentile; in the scenario RCP 8.5 the For the other projections, the most adequate medium-term evolu­
temperatures are higher than RCP 4.5 for each percentile. Comparing tions seem to be described by the 50th percentile. In the case of W_45-50
the outputs of the two generators based on the 2050 s time period, for the temperature difference is limited during the winter months, between
both RCP 4.5 and RCP 8.5, in the 50th percentile the DBT distributions 1.33 ◦ C (February) and 1.52 ◦ C (November), and it is comparable to
(W_45-50 and W_85-50) are more comparable to the CC_MT trend. CC_MT during the summer. In this season the great variation compared
Indeed, the average values of W_45-50 and W_85-50 are respectively with TMY is obtained in July since the average monthly DBT is 25.8 ◦ C
17.2 ◦ C and 17.8 ◦ C. For the models relating to the 90th percentile, the for TMY and 28.3 ◦ C for W_45-50. The values of the 90th percentile are
temperature distributions are in both cases (W_45-90 and W_85-90) in all months much higher than the differences obtained with CC_MT
more in line with the box plot obtained for CC_LT (2080 s). The and more in line with what was seen for CC_LT. Indeed, considering July
average values of W_45-90 and W_85-90 are respectively about 18.1 ◦ C the average monthly value is 29.3 ◦ C for W_45-90 and respectively
and 18.7 ◦ C. 28.2 ◦ C and 30.2 ◦ C for CC_MT and CC_LT.
More in detail, the comparison of the monthly mean temperatures for In term of temperature difference with TMY, W_85-90 is character­
TMY and for CC_MT and CC_LT indicates that the temperature differ­ ized by variation between 2.55 ◦ C (March) and 4.49 ◦ C (August); this
ences are greater in the case of CC_LT. The maximum monthly average scenario is comparable with CC_LT, indeed in the same months, these
temperature variation occurs in August. In this month, the average value differences are 2.61 ◦ C (March) and 4.89 ◦ C (August). For these two
of CC_LT is equal to 31.6 ◦ C, while the average values of CC_MT and TMY scenarios, there is also a sensible variation of the average value of

17
R.F. De Masi et al. Applied Energy 303 (2021) 117584

temperatures in the winter months; for instance, in December the DBT is energy need is comparable with those obtained with TMY2, while its
6.58 ◦ C in the reference TMY scenarios, 9.48 ◦ C and 9.38 ◦ C respectively energy need for cooling (22 kWh/ m2 year) is equal to the value obtained
for CC_LT and W_85-90. However, these refer to different time horizon. with TMY. This seems indicate that among the weather data, the dry
Similar observations can be done for the box plots obtained for the bulb temperature has the greatest impact on the resolution of the
hourly values of DNI, shown in Fig. 15. building energy balance.
However, climate change in this case is more evident when analyzing The results obtained for the extreme years are also consistent with
the future long-term projection (CC_LT) than the medium-term pro­ the construction criteria. When XMYmax is used, the cooling energy need
jections of both used meteorological generators. The average value of increases of around + 25% compared with TMY meanwhile the heating
CC_LT is approximately 154 W/m2, while in the cases of W_85-90 and request is reduced of around − 14%. Conversely, XMYmin is the extreme
W_45_90 the average values shown by the box plots are respectively cold year, with the maximum heating energy need, equal to 35 kWh/ m2
about 146 W/m2 and 144 W/m2. In the transition from CC_MT to CC_LT year (+16.4% compared to TMY). This result allows understanding the
the average value of the box plot increases of around 11%. The value of importance of considering extreme events during the design of the
the 3rd quartile goes from 173 W/m2 to about 198 W/m2 while the refurbishment interventions, because it is clear that the actual config­
maximum value, excluding the aberrant points, is equal to 493 W/m2 for uration is much more sensible to the heat weaves phenomenon than to
CC_LT and 432 W/m2 for CC_MT. colder winter also if it is not well insulated. The actual building envelope
In general, despite the uncertainties connected to the considered is not adequate to prevent overheating phenomenon and the selection of
projections, it can be certainly considered for Benevento a future tran­ the efficiency measures should be oriented to the increase of building
sition towards a dominant cooling climate. This is clearer with the thermal inertia.
estimated values of heating and cooling degree days (Fig. 16). The results obtained with TRY_BN and TRY_CTI are in accordance
More in detail, the HDDs decrease and thus it is expected a reduction, with the calculation of degree days. More in detail, compared with TMY,
in all scenario, of the heating request. This HDD reduction is high in case the heating request is lower of around − 11% with TRY_BN and − 13%
of long term scenarios (-21%) as well as the increment of the CDDs that with TRY_CTI. This result also suggests that the adoption of data
are more than double compared with the current condition. This in­ monitored not in the city center has lower impact on the prevision of the
dicates the transition to a dominant cooling climate. heating request when the same methodology is adopted. Indeed, this
difference becomes important in the summer season. Indeed, for the case
6. Energy simulation results study, the difference between TRY_BN and TRY_CTI is around 11%.
When TMY is used as reference, the cooling energy need is lower of
All proposed weather files of the current conditions are used for around − 25% for TRY_CTI and it is lower than all other simulations.
evaluating the effect on the prevision of the energy request. Then, the Thus, the data monitored out of the city center, are not representative of
projections are considered for studying the resilience of the refurbish­ the external conditions because the cooling request is lower than the
ment design. expected one. When the base case is unbalanced, it could bring wrong
choices during the refurbishment.
The most different results are obtained with the file named PVGIS. As
6.1. State of fact already indicated by the other proposed analysis, these data are not
consistent for the prevision of the building energy performance. Indeed,
Fig. 17 shows the heating and cooling thermal energy need by the heating energy need is lower of around − 37% compared with TMY
varying the weather file used in the simulation program. For these data and this leads wrong estimation in the evaluation of the most convenient
the net conditioned area is 198 m2. Considering the methodology of the efficiency measure.
Typical Meteorological Year, the results obtained with TMY2 and IWEC It is also interesting the comparison of TRY_BN that is defined with
are comparable since there is also a difference of around 3.3% in the the same methodology but different data. Also in this case, the heating
heating season and 0.4% for the cooling energy need. With the original demand is really low (-29%) and thus also if the same procedure is
Sandia method (TMY) that attributes the greatest weight to the average adopted the database seems not representative for the evaluation of the
daily dry bulb temperature, the heating energy need is the highest heating request. Instead, the prevision of cooling request has a high
among the typical years with an increase of around 9% compared with variation compared with the TMY (-24%) but lower difference when the
IWEC. Instead this difference is lower for the prevision of the cooling same method is applied (-9.3%). Furthermore, from this observation it
energy need, since in both cases the increment is lower than 2.9%. can be deduced that an evaluation of the PVGIS file based on the DBT
Also the IGDG method seems a reliable method despite it is based variable could not justify the result obtained with the energy simulation
exclusively on the dry bulb temperature values. Indeed, the heating

Fig. 15. DNI distribution - current weather file (TMY) and future climate projections.

18
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 16. Degree days for future projections: a) HDD; b) CDD.

Fig. 17. Energy need for heating and cooling with current EPW files – Base case.

of the base case. Indeed, considering the calculation of the degree days, cooling energy need. However, when the refurbishment is designed,
PVGIS showed an extremely lower CDD value compared to the other both aspects must be considered because the right approach consists in
files; this extreme condition is not confirmed by the dynamic simulation. the reduction of the energy needs with a deep design of the passive so­
The importance of solar radiation evaluation for the summer energy lutions and then the consideration of high performance plants with
balance is remarkable. Instead the results in term of HDDs are enough in renewable sources and optimal management strategies. The passive
agree with the prevision of the winter energy request. solutions, that influence the thermal request, have also an important
The same analysis can be done considering the primary energy effect on the indoor comfort. Briefly, the analysis on the primary energy
consumption of the building due to winter and summer needs for the with different weather file underlines the importance of considering the
base case. In the calculation of the primary energy a conversion factor of extreme phenomena more and more frequent in Mediterranean climate,
0.554 is considered for the electricity [77]. when the aim is the optimization of the performance of the building-
In the case of TMY, the primary energy, only for heating and cooling HVAC system. Finally, once again, it can be underlined that the adop­
services, is 74 kWh/ m2 year; when the files composed with the same tion of PVGIS database is not adequate, because the primary energy is
methodology but different weights are considered, the differences on the 49.6 kWh/ m2 year, a value very low compared with all other cases:
global consumption is quite limited: − 4.5% for TMY2 and − 6.5% for –33% respect to TMY and –24% compared with TRY_BN that is defined
IWEC. Similarly, for IGDG method, the difference is only of − 3.6%. This with the same methodology but monitored data.
leads to the conclusion that all these files could be appropriate for the
estimation of the building energy performance also if the weather var­
iables have different weights in the determination of the typical year. 6.2. Retrofitted building
Another interesting conclusion can be highlighted when the extreme
files are considered. Indeed, in the case of XMYmin for which the total The analysis of refurbishment scenario refers to the interventions
primary energy is higher of around 8.3% compared with TMY mean­ described in Table 7 and Table 8, where also the nomenclature is pro­
while the it is lower of around − 3.5% when XMYmax is considered. This posed. All efficiency measures are passive solutions aimed at the
result seems divergent compared with the analysis in term of thermal reduction of the energy need; future research will regard the impact of
energy need. Instead, it must be considered the role of the plant system. interventions on the heating and cooling systems.
Indeed, the hot water is provided by a boiler with good efficiency but the Since it has been demonstrated that when the same methodology is
cooled water is provided by an electric chiller with higher performance used, also with different weights, the results are comparable, only the
coefficient. For this reason, when the primary energy is considered, the TMY file will be used for the current scenario; thus the simulation results
increment in the heating request has a greater weight compared with the for TMY2, IWEC and IGDG are not discussed.
First of all, the application of the single efficiency measure is

19
R.F. De Masi et al. Applied Energy 303 (2021) 117584

reported. The replacement of windows causes, above all, the reduction results for the two most significant redevelopment scenarios are dis­
of the cooling energy need. For each weather file, the calculated energy cussed. These are RS3 (TI + DBL_L) and RS9 (TI + DBL_S + RP + ES), for
saving has the same magnitude. For instance, the adoption of double which it is obtained respectively the maximum reduction of heating
selective windows improves the summer behavior with ΔEC equal to energy need and of energy need for cooling. The following figure shows
− 21.4% in the TMY scenario, − 24.5% in the TRY_BN scenario, –22.5 % also ΔEH and ΔEC for each weather data scenario.
and − 19.3% with XMYmin and XMYmasx. The most different previsions Fig. 18a shows the heating and cooling request when the insulation
are obtained with the file created with external database since ΔEC is of the opaque envelope is associated to the installation of the low-
− 26.8% with TRY_CTI scenario and − 30.8% with PVGIS. The heating emissive windows; these interventions have bene considered for
request is characterized by a slightly increase for all cases except the reducing the heat losses due to conduction and for increasing in the
simulation with PVGIS; in this case ΔEH is + 6.1%. winter season the heat gains due to glassed surface that emits low levels
Similar conclusions are obtained when the installation of internal or of radiant thermal energy. However, the insulation, mainly of the roof,
external shading system is considered. For instance, ΔEC is − 11.9 % with contributes also to increase its inertia and for this reason also the sum­
TMY and –18.4% with PVGIS meanwhile ΔEH is + 2.7% for TMY sce­ mer energy need is reduced. Furthermore, the considered window has a
nario and + 5.0% with PVGIS. lower solar factor than the clear glasses of the base case; the direct solar
The discrepancy in the external forcing could modify the evaluation gains are reduced in the cooling season. It can be observed that the
of the benefit of a proposed measure mainly when it regards the glazed obtainable energy savings are comparable for all scenarios; in term of
component for which the radiation is a fundamental contribute in the primary energy it varies from − 26.8% when the XMYmin file is adopted,
solution of the energy balance. Indeed, it influences the radiative ex­ to − 28.8% for TRY_CTI.
change but also the amount of solar inner gains that can reduce the Also with the package of measures named RS9, in Fig. 18b, the re­
heating request and increase the cooling one. In this view, it seems really ductions obtained with the current climate files are comparable, mainly
important the adoption of data monitored in a site representative of the as regard the winter season. In this case, the adoption of selective win­
building location. dows and external shading combined with thermal insulation and cool
The efficiency measure with the great influence on the building paints greatly reduces the energy need for cooling.
behavior is the insulation of opaque envelope (TI, table 7). Indeed, in This analysis suggests to designer that the package RS9 according to
term of primary energy, the results are comparable in all cases. In detail, the current scenario is feasible for improve both the winter and summer
ΔEP varies between –22.3% with XMYmin and –23.7% with TRY_CTI. performance in Mediterranean climate as Benevento.
This result has to be commented considering the building energy bal­ An important finding can be remarked and extended to refurbish­
ance. The insulation contributes, mainly, to reduce the heat losses due to ment design in climate with high incidence of solar radiation. The
conduction heat flux in the building material. With a simplified effectiveness of efficiency measures to control or reduce the incoming
approach, the main external forcing is the dry bulb temperature of the solar radiation depend by the adopted methodology for selecting the
winter period. The calculated HDDs for the used weather file show typical months in the weather file. Indeed, as reported in Fig. 18b, the
limited differences and this conclusion is found also in the prevision of prevision of energy saving varies from − 45.8% with TMY to − 55.3%
the energy saving. with PVGIS that starts from different data as well as ΔEC is − 52% when
Finally, the application of an insulated cool roof (RP, table 7) im­ data of a rural station is used (TRY_CTI). This conclusion is confirmed
proves both the winter and summer behavior. However, it has to be from the results of the application of double selective windows and the
taken into account that the case study building is characterized by two external solar shading. Indeed ΔEC is − 38.1% with TMY and it becomes
heated and cooled floors and this type of intervention affects mainly the − 51.9% with PVGIS or − 35.2% with XMYmax. Thus, it is extremely
energy balance of the upper floor. important to pay attention to the input data because the economic
For a climate as Benevento, considering the adoption of TMY file, analysis length to the energy results could bring profitable or not results.
ΔEC is − 12.2% and ΔEH is − 4.3%; almost the same results are obtained Fig. 19 shows the primary energy for heating and cooling for the base
with the extreme XMYmax. This indicates that the difference in the case and all considered energy efficiency packages. The represented
energy balance of the base case are equally transposed in the refurbished findings allow remarking two main aspects that can be useful to designer
case and thus, the evaluation of the short term performance is not and researchers more in general for the Mediterranean climate condi­
greatly affected by the consideration of most extreme conditions. tions. First of all, the percentage variation of the primary energy and
As in the other cases, the greatest variations are obtained with the thus the calculated energy saving has the same magnitude for all
external database, mainly for the summers season; indeed, ΔEC is − 14% considered weather files; this is a deceptive starting point for designers
and − 14.8% respectively with TRY_CTI and PVGIS. because it can be assumed that the error in the estimation of the actual
Considering the packages of measures in Table 8, first of all, the performance is translated in the prevision of refurbished performance.

Fig. 18. Energy need for heating and cooling with current EPW files: a) RS3; b) RS9.

20
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 19. Primary energy for heating and cooling – Base case and retrofit scenarios.

Thus, if the results are compared only in term of percentage, it can be therefore, the base case would be lower heating energy need than the
assumed that all weather files bring to comparable conclusions in term refurbished one. Considering the summer behavior, the finding is that
of refurbishment design. On the other side, the value of the primary the proposed intervention is not resilient because the increase of the
energy demand and the composition in term of heating and cooling cooling demand for the extreme events of climate changes is not
request is really different both in the base case and in the refurbished balanced by the installation of solar control windows. Indeed, ΔEC,cl,ch is
ones. Therefore, in Mediterranean area, it is important the evaluation of − 4.6% for the scenarios W_45-10 meanwhile is always positive for the
solar radiation in the construction of weather file for improve the other weather files. More in particular, ΔEC,cl,ch is + 88.9% in the pro­
readability of the summer behavior. Indeed, for instance, in case RS9, jection CC_LT and + 48.3% for CC_MT. Therefore, for the current sce­
the refurbished building passes from 74.1 kWh/ m2 year to 53.7 kWh/ nario, ΔEC is − 21.4% and this suggests that the efficiency measure
m2 year when the weather file is TMY; instead, the primary energy should be applied; instead the projection at 30 or 50 years underlines the
changes from 63.8 kWh/ m2 year to 45 kWh/ m2 year when the available poor impact that the investment causes. However, it must be also
file TRY_CTI is used, or from 49.6 kWh/ m2 year to 35.9 kWh/ m2 year underlined that if the building would not be undergone to refurbish­
when the PVGIS source is used. ment, its performance in long term scenario would be worse. Indeed,
The resilience of the refurbishment design is studied, as said, considering CC_LT the heating energy need would be 49.2 kWh/ m2 year
considering the energy saving calculated comparing to the weather file in the not refurbished building and 42 kWh/ m2 year in the refurbished
obtained with the climate projections. The reference scenario, for the one.
base case, is the one simulated with TMY. The same conclusions have been found for the efficiency measures of
First of all, Fig. 20 reports the results for the intervention with the the shading system application. Considering for instance the internal
higher impact in term of cooling request (20a) and heating request application for the current scenario (TMY) the heating energy need is
(20b). Considering Fig. 20a, in term of energy need, the refurbished comparable to the base case (ΔEH =+0.4%) meanwhile ΔEC is − 9.5%.
building simulated with the future climatic data is characterized by With the future weather data, the heating demand decreases, ΔEH,cl,ch
lower heating demand compared with TMY. This is attributable only to varies from − 49.7% with CC_LT and − 7.5% with W_45-10 but it is due to
the increase of outdoor overheating. Indeed ΔEH is + 1.6% when TMY is the climate change because considering for instance the CC_LT scenarios
considered and thus the reduction of solar gains increases the heating the energy need would be 3075 kWh for the building not refurbished and
demand. Instead, ΔEH,cl,ch varies from − 48.2% with CC_LT to – 6.1% 3085 kWh in case of intervention. Instead, the cooling demand increases
with W_45-10. On the contrary, looking to the heating demand, the and the efficiency measure seems to be not resilient (e.g. ΔEC,cl,ch =
refurbished building could get worse than the base case because +66.1% with CC_MT and + 50.2% with W_85-50).
considering the scenario CC_LT, the heating energy need would be 15.5 These result invite designers and researchers to reflect on the eval­
kWh/ m2 year if the building were not undergone to the efficiency uation of the impact of an intervention. More in detail, they should
measure, meanwhile it is 16 kWh/ m2 year in the refurbished scenario. introduce in the designing approach a sensitivity analysis for estimating
Indeed, the traditional clear windows would allow for better exploita­ the resilience of proposed efficiency measures with the aim to guide the
tion of heat gains due to increase of outdoor temperature and radiation; tenants and users to select the best scenario, according to their

Fig. 20. Energy need for heating and cooling: a) Double solar control windows; b) Roof and wall insulation.

21
R.F. De Masi et al. Applied Energy 303 (2021) 117584

expectations and to the lifetime of the intervention. extreme projections, the objective of energy saving could be reached
Fig. 20b refers to the insulation measure. In this case, the combined both in winter and in summer. Focusing on the summer period, ΔEC,cl,ch
effect of less severe external conditions and lower thermal transmittance is –32.8% and − 5.8% respectively with the 50th percentile of RCP4.5
guarantees important reduction of heating demand; ΔEH,cl,ch is − 34.4% and RCP 8.5, thus there is a significant energy saving compared to the
in the case of W_45-10 and − 63.9% with CC_LT. If the building was not base case simulated with the current weather file. Moreover, the cooling
refurbished, the heating demand in the future projection would be energy demand, if the building was not undergone to the refurbishment,
higher. Considering the W_85-90 scenario, the request would be 4080 will be 36 kWh/ m2 year, and thus, higher than the value reported in
kWh without the insulation intervention and 2795 if the insulation was Fig. 21b for the same scenarios. Instead, when the reference is CC_MT or
applied. Instead, in all cases an increment of cooling demand is ob­ CC_LT, ΔEC,cl,ch is respectively + 0.4% and + 30.7% and thus the same
tained; ΔEC,cl,ch is + 82.7% in the scenario CC_LT or + 41.4% in the case consideration of the previous case can be applied to these results.
of W_85-50. Also in this case, this result is attributable to the climate However, it is important underline that, in Mediterranean climate, a
change and the designed intervention cannot be decisive on the energy refurbishment action mainly finalized to the reduction of the summer
performance. However, the not refurbished building would have worse energy request, could be improve the resilience of the building more
performance; for instance, in the case W_85-50 the heating request than a simple intervention of thermal insulation. For some emissions’
would be 36 kWh/ m2 year without the intervention and 31.5 kWh/ m2 scenarios, the prevision of the energy behavior indicates a reduction of
year with the insulated envelope. Thus, it can be concluded that in any the cooling request also if the external forcing is more severe. On the
case the intervention is advisable, but the effects are lower compared to other side, the expected energy saving is lower compared to the one
the evaluation done for the current weather file. calculated with a weather file representative of the current conditions.
Fig. 21 shows the prevision of heating and cooling energy need with Indeed, according to the current design practice, or the considered RS9,
the weather files that take into account the climate changes, for the most the expected reduction of the cooling energy need would be − 45.8%.
interesting packages of efficiency measures. The package RS3 (Fig. 21a) Thus, the simulation of the retrofit configuration with the current
combined with less severe condition, causes important reduction of the weather file (TMY in this case study) gives most encouraging indications
heating need, with a minimum (2155 kWh) in the long term (20800) to the designers with positive environmental and economic indicators,
scenario. The energy saving, considering as base case the one TMY, for starting from the value of energy indices.
the RCP 4.5 scenario varies from − 36.6% (10th percentile) and − 50.8% It is also interesting what happens in term of primary energy. Fig. 22
(90th percentile) and for RCP 8.5 from − 41.0% (10th percentile) to shows the requests for the refurbished building with the current file
− 55.6% (90th percentile). The prevision for the 50th percentile is not in (TMY), the future projections and also for the building in the base case.
agree with the CC_MT scenario, as it was observed in the preliminary This last case indicates the primary energy request if the building was
analysis of the weather data. Once again, an accentuated effect of global not refurbished with the extreme forcing due to climate changes. Firstly,
warming is generally observed, passing from W_45-10 to CC_LT. it can be observed that the refurbished configuration is characterized by
Therefore, according to these future projections Benevento would tend lower primary demand compared with the base case simulated with the
to be a dominant cooling climate with critical effects on the increase in projections. For instance, considering the worse scenario (CC_LT), the
energy need for cooling. In the case of RS3 with CC_MT ΔEC,cl,ch is 28.8% RS1 assures a reduction of primary energy of around − 10% meanwhile
and it becomes + 12.1% and + 23.3% respectively for W_45-50 and with S6, RS7, RS8 and RS9 of about − 28%. However, the interpretation
W_85-50. The highest increase is obtained with CC_LT (+61.2%), but if of the results starting from the primary energy could be misleading.
the behavior of the refurbished and not refurbished building is Indeed, it is not distinguishable the particularity of a typical Mediter­
compared for the same climate forcing, it can be concluded that the ranean climate. More in detail, the sharp decrease in heating needs tends
proposed intervention, also if not resilient, helps in the reduction of the to offset the significant increase in cooling needs and this effect is
energy demand. Indeed, the not refurbished building on long term sce­ accentuated by moving from the medium-term to the long-term pro­
nario, will be characterized by a cooling demand of 49.2 kWh/ m2 year, jection. For this reason, all refurbishment scenarios seem to be appro­
instead the refurbished one by 35.9 kWh/ m2 year. The projections with priate to front the climate change. Instead, if the balance is analyzed by
the 10th percentile give different results; indeed when W_45-10 is considering the energy need the criticalities are clear. For instance,
considered, ΔEC,cl,ch is − 8.1%, instead, the cooling demand remains considering the insulation of the building envelope, the application of a
unchanged with W_85-10. This finding indicates that with not extreme cool paint on the roof, the installation of double low-emissive windows
emission projections and thus with the use of decisive actions to control and internal shading (RS6) is characterized by ΔEPcl,ch equals to –33%
climate change, some packages of building efficiency measures could be with CC_MT and − 31% with CC_LT higher that the results obtained with
effectively resilient. the current weather file (-29%, TMY). If the energy needs are analyzed,
The scenario RS9 is the most encouraging because, except for the it is clear that this result is mainly attributable to the decrement of

Fig. 21. Energy need for heating and cooling with climate projections: a) RS3; b) RS9.

22
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 22. Primary energy for refurbished building – current weather file (TMY) and future climate projections.

heating demand as consequence of climate changes. Indeed, ΔEH,cl,ch is and therefore to different results in the energy needs for heating and
− 52.2% and − 62.9% respectively for CC_MT and CC_LT, but the cooling cooling of the building. There is no way to ascertain that one climate
demand increases since ΔEC,cl,ch is respectively + 12.9% and + 44.5%. projection has higher readibility than another one, but the use of mul­
More in general, about the proposed method, the indication for de­ tiple scenarios allows testing < the robustness of a retrofit project
signers is that for evaluating the resilience of the efficiency measures, ensuring the resilience of the building in the long term. In the previous
the whole primary energy saving is not a good index. Indeed, it is not paragraph, the differences between the results obtained for all the
possible to understand the effectiveness of proposed retrofit measures medium-term (2050 s) and long-term (2080 s) climate projections have
because the great decrement of the heating request, due to climate already been discussed. At this point, it is important to carry out a
change, balances the increment of the cooling need. This is true for cities sensitivity analysis of the energy needs for heating and cooling of the
in Mediterranean area. building to varying factors of uncertainty in the climatic projections.
Furthermore, it has to be considered that the normative indications This assessment cannot involve the medium and long term projections of
could change in future to take into account the new European objectives the CCWorldWeatherGen tool since these are based on unique climate
and thus more performant indicators could be added. About it, the model and the same emission scenario. Instead for the projections
proposed methodology and the analysis of proposed results could be generated with Weathershift the analysis is based on two factors of un­
useful both from the point of view of the legislator and the designer. certainty, i.e. the general climate model (10th, 50th and 90th percentile)
Indeed, the legislator could boost this type of research with the aim to and the emission scenario (RCP 4.5 and RCP 8.5). First of all, consid­
receive information on which type of intervention would be more useful ering the refurbishment scenarios RS3 and RS9, Fig. 23 shows that the
to support with national economic incentives for making the building sensitivity of the building energy performances in the two retrofit sce­
stock resilient to climate change on short or long term. Moreover, re­ narios is very similar to the variation of the global climate model and
searchers can support the legislators in identifying new performance with the same emission scenario (RCP 4.5). Basically, even if the retrofit
standards; for instance, in areas with a climate comparable to Benev­ scenarios lead to different values of energy needs for heating and cool­
ento, given the expected increase in consumption in cooling, more ing, comparable percentage variations are observed with respect to the
stringent parameters in the summer passive control. On the other hand, global climate model relating to the 10th percentile. On average, with
designers can define an order of importance on the types of measures to the GCM of the 50th percentile, the heating energy need is reduced by
be recommended according to the climatic zone, also offering customers 9.4 % and the cooling energy need increases of 23.6%. Considering the
the benefits of evolving conditions or moving ahead of the regulatory 90th percentile the energy needs for heating and cooling undergo an
requirements. average variation of − 21.3 % and + 44.1 % respectively. The plotted
Finally, all proposed results suggest that the traditional measures are trend lines indicate the same sensitivity of the building energy perfor­
not properly indicated to front the climate change. Thus, this study mances to varying GCM, regardless of the considered retrofit scenario.
underlines that researchers must study and evaluate the effectiveness of Indeed, the trends relating to the energy needs for heating and cooling
innovative measures aimed at the reduction of the cooling energy need are, with a good approximation, linear in both cases and the slopes are
by using weather data capable of reproducing the effects of climate comparable.
change. Indeed, considering the prevision of the emissions’ scenarios the Therefore, with the 90th percentile model, the building with the
current approach aimed to satisfy a specific energy efficiency target is climate projections provided by the Weathershift tool will always have
not for designing the buildings of the future. the lowest energy need for heating and the highest energy need for
cooling. On the contrary with the 10th percentile model, the higher
6.3. Sensitivity analysis for the uncertainty of the weather projection energy need for heating will be obtained. Similar deductions are made
models by evaluating, with the same global climate model, the transition from
the RCP 4.5 to the RCP 8.5 (Fig. 24). Also in this case, for the two retrofit
In this study, two tools for the generation of the future climate scenarios RS3 and RS9, the percentage variations obtained are compa­
projections from the TMY file were examined. The need to consider rable both in terms of energy need for heating (Fig. 24a) and energy
more meteorological generators and therefore more future climate sce­ need for cooling (Fig. 24b). On average, the energy need for heating is
narios is due to the need to assess the resilience of the retrofitted reduced by 6.8%, 10% and 9.5% with the 10th, 50th and 90th percen­
building taking into account the inherent uncertainties of future climate tiles respectively. At the same time, the energy need for heating in­
projections. As can also be seen from Table 5, the sources of uncertainty creases on average by 9.4%, 10.9% and 10% with the 10th, 50th and
are due to various factors and in particular to the Global Climate Models 90th percentiles respectively. Furthermore, Fig. 2X shows how different
(GCM) or Regional Climate Models (RCM), emission scenarios, initial GCMs respond to the same emission scenarios. More in particular, the
conditions and spatial resolution of GCMs / RCMs. The variation of one GCM relative to the 50th percentile always shows a greater percentage
or more uncertainty factors leads to a different future climate scenario variation passing from RCP 4.5 to RCP 8.5. Instead, the GCM relative to

23
R.F. De Masi et al. Applied Energy 303 (2021) 117584

Fig. 23. Sensitivity analysis on GCM - RS3 & RS9: a) Heating energy need; b) Cooling energy need.

Fig. 24. Sensitivity analysis on RCP - RS3 & RS9: a) ΔEh; b) ΔEC.

the 10th percentile always shows a smaller difference. percentile and − 42.2% to − 47.5% with 90th percentile. In terms of ΔEC,
Finally in Fig. 25 the average percentage variation of heating and cl,ch,avg, it goes from –22.8% to − 15.3% with the 10th percentile, from
cooling energy need (ΔEH,cl,ch,avg and ΔEC,cl,ch,avg) are compared to the − 3.9% to + 7.0% with the 50th percentile and from + 12.7% to + 24.3%
base case of the building simulated with TMY for evaluating the influ­ with the 90th percentile.
ence of the factors of uncertainty considered on the assessment of the By varying the general climate model (GCM) from the 10th to the
resilience of the building. 90th percentile, ΔEH,cl,ch,avg varies from − 27.1% to − 42.2% with RCP
From Fig. 25 it can be seen that passing from the RCP 4.5 scenario to 4.5 and from − 31.9% to − 47.5% with RCP 8.5. At the same time, ΔEC,cl,
the RCP 8.5 scenario, ΔEH,cl,ch,avg varies from − 27.1% to − 31.9% with ch,avg varies from –22.8% to 12.7% with RCP 4.5 and from − 15.3% to
the 10th percentile model, from –33.8% to − 40.2% with the 50th − 24.3% with RCP 8.5. Therefore, it can be concluded that in the context

Fig. 25. Sensitivity analysis of the building resilience on uncertainty factors.

24
R.F. De Masi et al. Applied Energy 303 (2021) 117584

of this study the resilience of the building examined is affected to a change. Only for some emissions’ scenarios, the adoption of solar con­
greater extent by the uncertainty due to the climate model rather than trol windows, external shading and cool roof could determine a reduc­
the uncertainty due to the emission scenario. tion of the cooling request also if the external forcing is more severe (for
the case study, –32.8% and − 5.8% respectively with the 50th percentile
7. Conclusion of RCP4.5 and RCP 8.5). On the other side, the expected energy saving is
lower compared to the one calculated with a weather file representative
The paper proposes a critical analysis on the effect of different of the current conditions. Therefore, the future research should be
dataset and methodology, including the projection of climate change, focused on innovative materials and technologies for the reduction of
when it is defined a weather file for the building energy performance the cooling energy need.
simulation. Finally, this papers wants suggest a methodology useful for designers
First of all, by means of statistical approach the weather data are and researchers in the evaluation of the most effectiveness energy
compared for evaluating the main discrepancies and then the impact on refurbishment to front climate change but also for legislators that need
the building behavior is analyzed. The main findings of the proposed to establish new threshold values for the building sector and new
case study (residential building) can be generalized to comparable incentive strategies.
climates. Further analyzes are under evaluations for obtaining results for other
More in detail, the statistical comparison of the weather files in­ climates and typology of the building stock.
dicates that the selection of efficiency measures requires the adoption of
representative climatic data for studying the composition of the energy CRediT authorship contribution statement
demand and the impact of different choice on the annual energy bal­
ance. The results clearly indicate that the solar radiation is a variable Rosa Francesca De Masi: Conceptualization, Data curation, Formal
with a great impact on the prevision of energy saving and thus in the analysis, Investigation, Methodology, Supervision, Writing – original
Mediterranean zone this must be considered in the selection of typical draft, Writing – review & editing. Antonio Gigante: Conceptualization,
years for the construction of the weather file. However, considering the Data curation, Formal analysis, Investigation, Methodology, Supervi­
simulations of the building energy performance, the weather files built sion, Writing – original draft, Writing – review & editing. Silvia Rug­
starting from the same data and the same methodology but with giero: Conceptualization, Data curation, Formal analysis, Investigation,
different weights for the climatic variables, are comparable with Methodology, Supervision, Writing – original draft, Writing – review &
maximum difference in the primary energy request of 9%. Instead, when editing. Giuseppe Peter Vanoli: Conceptualization, Data curation,
the extreme hot file is used, the cooling energy need increases of around Formal analysis, Investigation, Methodology, Supervision, Writing –
+ 25% meanwhile with the extreme cold file the heating demand in­ original draft, Writing – review & editing.
creases of + 16.4%. This results underline the importance of considering
extreme events for understanding if the actual configuration is much
Declaration of Competing Interest
more sensible to the heat weaves than to colder winter. The adoption of
data monitored in rural or urban site leads to different representation of
The authors declare that they have no known competing financial
the local microclimate and thus of the urban heat island phenomenon.
interests or personal relationships that could have appeared to influence
The average annual temperature difference can also exceed 4 ◦ C and the
the work reported in this paper.
CDDs can differ of more than 44%; this discrepancy brings a difference
in the prevision of the cooling energy need of around 25%.
References
Considering the medium term (2050 s) projection of the CCWorld­
WeatherGen, the temperature difference with the current weather data [1] Kramer, D., J. Ware, Counting the cost 2020: A year of climate breakdown,
varies from 1.16 ◦ C (April) and 2.70 ◦ C (August) and between 2.40 ◦ C December 2020. https://reliefweb.int/report/world/counting-cost-2020-year-cli
(April) and 4.89 ◦ C (August) when the long-term (2080 s) projections are mate-breakdown-december-2020.
[2] IPCC. Climate change 2007: the physical science basis. Contribution of working
considered. In the case of the medium-term projections generated using group I to the fourth assessment report of the intergovernmental panel on climate
WeathershiftTM, there is a progressive increase of temperature passing change. Cambridge, UK and New York, NY, USA: Cambridge University Press;
from the scenarios of 10th to the 90th percentile. Considering the 90th 2007.
[3] Zhai ZJ, Helman JM. Implications of climate changes to building energy and
percentile and RCP 8.5, the temperature variation is between 2.55 ◦ C
design. Sustainable Cities and Societies 2019;44:511–9.
(March) and 4.49 ◦ C (August). A more general conclusion can be find [4] E. Comission, A Policy Framework for Climate and Energy in the Period from 2020
with reference to all proposed projections. The comparison of weather to 2030 – Communication from the Commission to the European Parliament, the
data, in the preliminary analysis, has indicate that the 50th percentile of Council, the European Economic and Social Committee and the Committee of the
Regions, (n.d.).
the projections obtained by WeathershiftTM is comparable with the mid- [5] U.S. Department of Energy. Weather Data Sources. https://energyplus.net/weathe
term projection and thus to CC_MT. Indeed, the analysis of the simula­ r/sources.
tions underlines that each scenario has a different impact on the forecast [6] Italian Climatic data collection “Gianni De Giorgio” (IGDG), EnergyPlus source
(05/02/2021) https://energyplus.net/sites/all/modules/custom/weather/weathe
of energy savings. Moreover, despite the uncertainties connected to the r_files/italia_dati_climatici_g_de_giorgio.pdf.
considered projections, it can be certainly considered for Benevento a [7] De Masi RF, Gigante A, Ruggiero S, Vanoli GP. The impact of weather data sources
future transition towards a dominant cooling climate, with an expected on building energy retrofit design: case study in heating-dominated climate of
Italian backcountry. J Build Perform Simul 2020;13(3):264–84.
increase of CDD that varies from + 15% (10th percentile RCP 4.5) to + [8] Hall IJ, Prairie RR, Anderson HE, Boes EC. Generation of a typical meteorological
104% (90th percentile RCP 8.5). This finding can be generalized for year, Proceedings of the 1978 annual meeting of the American Section of the
other cities with comparable degree days. International, Solar Energy Society 1978; 669–671.
[9] Marion W, Urban K. Users manual for TMY2s: Derived from the 1961–1990
The evaluation of energy savings indicates that the most profitable National Solar Radiation Data Base (No. NREL/SP-463-7668). National Renewable
solution in the current scenario is the insulation of the opaque envelope Energy Lab., Golden, CO (United States), 1995.
and the application of a cool paint for the roof, associated to the [10] ASHRAE, International weather for energy calculations (IWEC weather files) user’s
manual, Version 1.1, 2012.
installation of the selective windows and external shadings. In this case,
[11] Wilcox S, Marion W. Users manual for TMY3 data sets. (Revised) 2008.
the reduction of the heating energy need varies from − 19 % to –23% and [12] Climate One Building Climate one building: repository of free climate data for
the cooling energy saving from − 44% to − 55%. building performance simulation, 2020. http://climate.onebuilding.org/.
Instead, the traditional passive measures (insulation, solar control or [13] European Committee for Standardization (CEN). EN ISO 15927-4:2005,
Hygrothermal performance of buildings - Calculation and presentation of climatic
low emissive windows, cool roof, shading system) are not resilient when data - Part 4: Hourly data for assessing the annual energy use for heating and
the energy savings are calculated taking into account the climate cooling.

25
R.F. De Masi et al. Applied Energy 303 (2021) 117584

[14] Kim S, Zirkelbach D, Künzel HM, Lee JH, Choi J. Development of test reference change impacts. Energy Reports, Available online 18 May 2021, In Press, Corrected
year using ISO 15927–4 and the influence of climatic parameters on building ProofWhat are Corrected Proof articles. https://doi.org/10.1016/j.
energy performance. Build Environ 2017;114:374–86. egyr.2021.04.012.
[15] CTI TRY documentation (21/12/2020) https://try.cti2000.it/. [46] Muñoz González CMª, León Rodríguez AL, Suárez Medina R, Ruiz Jaramillo J.
[16] Lupato G, Manzan M. Italian TRYs: New Weather Data Impact on Building Energy Effects of future climate change on the preservation of artworks, thermal comfort
Simulations. Energy and Building 2019;185:287–303. and energy consumption in historic buildings. Appl Energy 2020;276:115483.
[17] Pernigotto G, Prada A, Gasparella A, Hensen JLM. Analysis and Improvement of the https://doi.org/10.1016/j.apenergy.2020.115483.
Representativeness of EN ISO 15927–4 Reference Years for Building Energy [47] Yassaghi H, Gurian PL, Hoque S. Propagating downscaled future weather file
Simulation. J Build Perform Simul 2014;7(6):391–410. uncertainties into building energy use. Appl Energy 2020;278:115655.
[18] PVGIS, Photovoltaic Geographical Information System, Joint Research Centre. [48] Sheng L, Ting Kwok Y, Ka-Lun Lau K, Wai Tong H, Wai Chan P, Edward NG.
https://re.jrc.ec.europa.eu/pvg_tools/it/#TMY. Development and application of future design weather data for evaluating the
[19] Crawley DB, Lawrie LK. Rethinking the TMY: is the ‘typical’meteorological year building thermal-energy performance in subtropical Hong Kong. Energy Build
best for building performance simulation? 14th conference of international 2020;209:109696.
building performance simulation association, Hyderabad, India. 2015. [49] Chakraborty Debaditya, Alam Arafat, Chaudhuri Saptarshi, Başağaoğlu Hakan,
[20] Narowski P, Janicki M, Heim D. Comparison of Untypical Meteorological Years Sulbaran Tulio, Langar Sandeep. Scenario-based prediction of climate change
(UMY) and their influence on building energy performance simulations. Proc. of impacts on building cooling energy consumption with explainable artificial
Conference “Building Simulation–BS2013”, Le Bourget-du-Lac. 2013:1414–21. intelligence. Appl Energy 2021;291:116807. https://doi.org/10.1016/j.
[21] ASHRAE. 1997a. WYEC2 Weather Year for Energy Calculations 2, Toolkit and apenergy.2021.116807.
Data, Atlanta: ASHRAE. [50] Arima Y, Ooka R, Kikumoto H, Yamanaka T. Effect of climate change on building
[22] Kočí J, Kočí V, Maděra J, Černý R. Effect of applied weather data sets in simulation cooling loads in Tokyo in the summers of the 2030s using dynamically downscaled
of building energy demands: Comparison of design years with recent weather data. GCM data. Energy Build 2016;114:123–9.
Renew Sustain Energy Rev 2019;100:22–32. [51] Radhi H. Evaluating the potential impact of global warming on the UAE residential
[23] Kočí J, Maděra J, Černý R. Generation of a critical weather year for hygrothermal buildings – A contribution to reduce the CO2 emissions. Build Environ 2009;44
simulations using partial weather data sets. Build Environ 2014;76:54–61. (12):2451–62.
[24] Kočí J, Kočí V, Maděra J, Černý R. Determination of the positive weather year for [52] Mutschler Robin, Rüdisüli Martin, Heer Philipp, Eggimann Sven. Benchmarking
application in hygrothermal simulations. WIT Trans Model Sim 2015;59:97–107. cooling and heating energy demands considering climate change, population
[25] Erba S, Causone F, Armani R. The Effect of Weather Datasets on Building Energy growth and cooling device uptake. Appl Energy 2021;288:116636. https://doi.
Simulation Outputs. Energy Procedia 2017;134:545–54. org/10.1016/j.apenergy.2021.116636.
[26] Pernigotto G, Prada A, Gasparella A. Extreme Reference Years for Building Energy [53] Streicher Kai Nino, Berger Matthias, Panos Evangelos, Narula Kapil, Soini Martin
Performance Simulation. J Build Perform Simul 2020;13(2):152–66. Christoph, Patel Martin K. Optimal building retrofit pathways considering stock
[27] Nik VM. Making energy simulation easier for future climate – Synthesizing typical dynamics and climate change impacts. Energy Policy 2021;152:112220. https://
and extreme weather data sets out of regional climate models [RCMs]. Appl Energy doi.org/10.1016/j.enpol.2021.112220.
2016;177:204–26. [54] Bamdad Keivan, Cholette Michael E, Omrani Sara, Bell John. Future energy-
[28] Hallegatte S. Strategies to adapt to an uncertain climate change. Global Environ, optimised buildings — Addressing the impact of climate change on buildings.
Change 2009;19(2):240–7. Energy Build 2021;231:110610. https://doi.org/10.1016/j.enbuild.2020.110610.
[29] IPCC special report on emissions scenarios (SRES): summary for policymakers. [55] Akkose Gizem, Meral Akgul Cagla, Dino Ipek Gursel. Educational building retrofit
Geneva, Switzerland: a special report of IPCC Working Group III Intergovernmental under climate change and urban heat island effect Journal of Building. Engineering
Panel on Climate Change; 2000. 2021;40:102294. https://doi.org/10.1016/j.jobe.2021.102294.
[30] IPCC, Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II [56] Azimi Fereidani Nazanin, Rodrigues Eugénio, Gaspar Adélio Rodrigues. A review
and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate of the energy implications of passive building design and active measures under
Change, IPCC, Switzerland, Geneva, 2014. climate change in the Middle East. J Cleaner Prod 2021;305:127152. https://doi.
[31] IPCC, Climate Change 2013: The Physical Science Basis. Contribution of Working org/10.1016/j.jclepro.2021.127152.
Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate [57] Peel MC, Finlayson BL, McMahon TA. Updated world map of the Koppen-Geiger ¨
Change, Cambridge University Press, Cambridge, United Kingdom and New York, climate classification. Hydrol Earth Syst Sci 2007;4(2):1633–44.
NY, USA, 2013, p. 1535. [58] Italian Organisation for Stardardisation (UNI). UNI 10349-1:2016, Heating and
[32] Belcher SE, Hacker JN, Powell DS. Constructing design weather data for future cooling of buildings -Climatic data -Part 1: Monthly means for evaluation of energy
climates. Build Serv Eng Res Technol 2005;26(1):49–61. need for space heating and cooling and methods for splitting global solar irradiance
[33] Climate Change World Weather File Generator for Worldwide Weather Data – into the direct and diffuse parts and for calculate the solar irradiance on tilted
CCWorldWeatherGen: Sustainable Energy Research Group. http://www.energy. planes.
soton.ac.uk/ccworldweathergen. [59] President of the Republic. Regulation containing prescriptions for the design,
[34] Jentsch Mark F, Bahaj AbuBakr S, James Patrick AB. Climate change future installation, operation and maintenance of heating systems in buildings in order to
proofing of buildings—Generation and assessment of building simulation weather limit energy consumption, implementing art. 4, par. 4 of Low 9 January 1991, n.
files. Energy Build 2008;40(12):2148–68. 10. Decree 26.08.1993 no. 412; 1993 (in Italian).
[35] Jentsch MF, Patrick James AB, Bourikas L, Bahaj AS. Transforming existing [60] Climatic data – Campania region (in Italian): https://cemec.arpacampania.it/mete
weather data for worldwide locations to enable energy and building performance oambientecampania/php/misure_suolo.php.
simulation under future climates. Renewable Energy 2013;55:514–24. [61] Ascione F, De Masi RF, de Rossi F, Ruggiero S, Vanoli GP. MATRIX, a multi activity
[36] IPCC Data Distribution Centre, HadCM3 climate scenario data download. test-room for evaluating the energy performances of ‘building/HVAC’ systems in
https://www.ipcc-data.org/sim/gcm_clim/SRES_TAR/hadcm3_download.html. Mediterranean climate: Experimental set-up and CFD/BPS numerical modeling.
[37] Dickinson R, Brannon B. Generating future weather files for resilience. Plea 2016 Energy and Building 2016;126:424–46.
Los Angeles - 36th International Conference on Passive and Low Energy [62] Hersbach Hans, Bell Bill, Berrisford Paul, Hirahara Shoji, Horányi András, Muñoz-
Architecture. Cities, Buildings, People: Towards Regenerative Environments. Sabater Joaquín, et al. The ERA5 global reanalysis. Quartely Journal of the Royal
[38] Remund J, Kunz S. METEONORM: Global meteorological database for solar energy Meteorological Society 2020;146(730):1999–2049.
and applied climatology, Fabrikstrasse 14, CH-3012 Bern (Switzerland): Meteotest; [63] Berrisford P, Dee DP, Poli P, Brugge R, Fielding M, Fuentes M, Kållberg PW,
1997. Kobayashi S, Uppala S, Simmons A. The ERA-Interim archive Version 2.0. 2011. htt
[39] Moazami Amin, Nik Vahid M, Carlucci Salvatore, Geving Stig. Impacts of future ps://www.ecmwf.int/node/8174.
weather data typology on building energy performance–Investigating long-term [64] European Centre for Medium-Range Weather Forecasts (ECMWF) https://www.
patterns of climate change and extreme weather conditions. Appl Energy 2019; ecmwf.int/en/forecasts/datasets/browse-reanalysis-datasets.
238:696–720. [65] Ballarini Ilaria, Corgnati Stefano Paolo, Corrado Vincenzo. Use of reference
[40] Triana Maria Andrea, Lamberts Roberto, Sassi Paola. Should we consider climate buildings to assess the energy saving potentials of the residential building stock:
change for Brazilian social housing? Assessment of energy efficiency adaptation The experience of TABULA project. Energy Policy 2014;68:273–84.
measures. Energy and Buildings 2018;158:1379–92. [66] TABULA WebTool; (06/02/2021), https://webtool.building-typology.eu/#bm.
[41] Moazami Amin, Carlucci Salvatore, Geving Stig. Critical Analysis of Software Tools [67] U.S. Department of Energy. Energy Plus simulation software, Version 8.9.
Aimed at Generating Future Weather Files with a view to their use in Building [68] DesignBuilder Software; 2018, Version 6.0.
Performance Simulation. Energy Procedia 2017;132:640–5. [69] Shen Pengyuan, Braham William, Yi Yunkyu. The feasibility and importance of
[42] Troup L, Eckelman MJ, Fannon D. Simulating future energy consumption in office considering climate change impacts in building retrofit analysis. Appl Energy 2019;
buildings using an ensemble of morphed climate data. Appl Energy 2019;255: 233-234:254–70.
113821. [70] Huang KT. Identifying a suitable hourly solar diffuse fraction model to generate the
[43] Ciancio Virgilio, Salata Ferdinando, Falasca Serena, Curci Gabriele, Golasi Iacopo, typical meteorological year for building energy simulation application. Renewable
de Wilde Pieter. Energy demands of buildings in the framework of climate change: Energy 157 2020:1102e1115.
An investigation across Europe. Sustainable Cities and Society 2020;60:102213. [71] Erbs DG, Klein SA, Duffie JA. Estimation of the diffuse radiation fraction for hourly
https://doi.org/10.1016/j.scs.2020.102213. daily and monthly-average global radiation. Sol Energy 1982;28(4):293–302.
[44] Pagliano Lorenzo, Carlucci Salvatore, Causone Francesco, Moazami Amin, [72] T. Baracu, A. M. Croitoru, A. Badea, Theoretical investigations of the solar
Cattarin Giulio. Energy retrofit for a climate resilient child care centre. Energy radiation at location of the passive house “politehnica” from Bucharest, U.P.B. Sci.
Build 2016;127:1117–32. Bull., Series C, Vol. 76, Iss. 4, 2014.
[45] Tootkaboni MP, Ballarini I, Corrado V. Analysing the future energy performance of [73] www.weathershift.com.
residential buildings in the most populated Italian climatic zone: A study of climate

26
R.F. De Masi et al. Applied Energy 303 (2021) 117584

[74] UNI 10349-3: 2016, Heating and Cooling of Buildings – Climatic Data – Part 3: [77] Atmospheric emission factors of greenhouse gases in the national electricity sector
Accumulated Temperature Differences (Degree-days) and Other Indices. and in the main European countries, ISPRA 2020. [in italian]: Rapporto317_2020.
[75] Ascione Fabrizio, De Masi Rosa Francesca, Santamouris Mattheos, Ruggiero Silvia, pdf (isprambiente.gov.it).
Vanoli Giuseppe Peter. Experimental and numerical evaluations on the energy [78] Khalil SA, Shaffie AM. A comparative study of total, direct and diffuse solar
penalty of reflective roofs during the heating season for Mediterranean climate. irradiance by using different models on horizontal and inclined surfaces for Cairo.
Energy 2018;144:178–99. Egypt, Renewable and Sustainable Energy 2013;27:853–63.
[76] Decree of the Ministry of Economic Development 26/06/2015 “Applicazione delle [79] CIBSE, CIBSE Guide J - Weather, solar and illuminance data, The Chartered
metodologie di calcolo delle prestazioni energetiche e definizione delle prescrizioni Institution of Building Services Engineers, London, 2002.
e dei requisiti minimi degli edifici”, 15 July 2015 (in Italian).

27