You are on page 1of 14

Energy & Buildings 165 (2018) 270–283

Contents lists available at ScienceDirect

Energy & Buildings


journal homepage: www.elsevier.com/locate/enbuild

A method to account for the urban microclimate on the creation of


‘typical weather year’ datasets for building energy simulation, using
stochastically generated data
S. Tsoka a,∗, K. Tolika b, T. Theodosiou a, K. Tsikaloudaki a, D. Bikas a
a
Laboratory of Building Construction and Building Physics, Faculty of Civil Engineering, Aristotle University of Thessaloniki, Greece
b
Department of Meteorology and Climatology, School of Geology, Aristotle University of Thessaloniki, Greece

a r t i c l e i n f o a b s t r a c t

Article history: Predicting buildings’ heating and cooling needs through dynamic simulation methods requires the input
Received 24 July 2017 of hourly weather data, so as to represent the typical meteorological characteristics of a specific location.
Revised 24 October 2017
Hence, the so called ‘typical weather years’ (TWY), mainly deduced from multi-year records of meteo-
Accepted 7 January 2018
rological stations outside the urban centres, cannot account of the complex interactions between solar
Available online 9 February 2018
radiation, wind speed and high urban densities which lead to the formation of the urban heat island
Keywords: effect and to higher ambient air temperatures. As the assumption that climatic parameters at a refer-
Typical weather years ence location of a meteo station are similar for a densely built up area can lead to miscalculations of
Representative days the heating and cooling needs, the aim of this study is to present a computational method for assessing
Microclimate simulation the urban climate’s effect during the generation of typical weather data for dynamic energy calculations.
Envi-met In this vein, a typical ‘urban specific weather dataset’ (USWD), reflecting the microclimatic conditions in
front of a building unit inside an urban district in the city of Thessaloniki, Greece is created based on
microclimate simulations with the Envi-met model; it is then compared with a typical reference weather
dataset (RWD), representing climatic conditions at a reference location of a meteo station. The results
indicate that the proposed method can capture microclimate characteristics; higher dry bulb tempera-
tures were reported during the year inside the urban canyon, compared to the corresponding values at
the reference location, with indicative mean daily deviations up to 1.0 °C and 0.75 °C in February and July
respectively. Wind speed, near the building façade is generally found lower than the corresponding val-
ues at the reference location, due to wind sheltering by neighbouring constructions. Given that climatic
parameters strongly influence the output of energy simulations the proposed computational method pro-
vide a contribution for higher accuracy of building energy simulation in the urban context. Future work
will involve energy performance simulations of a typical building unit with the generated USWD file so
as to evaluate the urban climate’s influence on energy needs.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction under major renovation, both for the residential and the tertiary
sector [1]. Given the growing concerns about energy waste and its
Today, residential and commercial buildings account for the 40% negative impact on the environment due to excessive CO2 emis-
of the final energy use in European Union while heating represents sions, there is an increasing interest towards accurate building en-
the most important part of residential energy consumption. Aiming ergy simulation. The corresponding simulation tools are currently
at the reduction of the energy demand of buildings, the Energy widely used by architects, engineers and designers in order to es-
Performance Buildings Directive2002/91/EC has obliged all mem- timate building energy needs but also to investigate the potential
bers of the EU to adopt a number of measures orientated towards of energy savings through measures that involve equipment, opera-
innovations and practices in order to respond to the growing en- tional schedules or control strategies. The use of advanced Building
ergy demand of the building sector. It has also imposed minimum Energy Simulation (BES) models may generally provide reliable es-
energy performance requirements for new and existing building timates of building energy demand with simulation results fitting
quite accurately with actual measurements [2].
When configuring the simulation, special attention should be

Corresponding author.
given on the accuracy of the climatic dataset used as input for the
E-mail address: stsoka@civil.auth.gr (S. Tsoka). calculation; as the IEA Annex 53 [3] pointed out, the weather file

https://doi.org/10.1016/j.enbuild.2018.01.016
0378-7788/© 2018 Elsevier B.V. All rights reserved.
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 271

introduced in BES models, comprising of 8760 hours of various cli- ergy performance, are intercompared so as to identify the impact
matic values such as dry bulb temperature, dew point temperature, of the urban morphology on the urban climate parameters.
wind speed, solar radiation etc. will strongly influence heating and
cooling loads calculations, systems dimensioning, energy produc- 2. Overview on the methods for ‘typical weather years’
tion from solar panels etc. Given that long term real weather data, generation
measured at a close distance from the respective study area are not
always available, most of the corresponding BES models, adopt the The use of a ‘typical weather file’ for BES has arisen from the
so called ‘typical weather files’; the latter are usually statistically need to compromise accuracy and computationally efficient sim-
processed and mainly derived from multiyear observations, issued ulations. To date, several procedures regarding the acquisition of
by weather stations in the peripheral zones, outside the urban ar- typical weather data have been reported in the literature. One of
eas [4]. the most commonly used hourly data format file for BES is the
As a consequence, complex interactions between solar radia- Typical Meteorological Year format, created by the U.S. National
tion, wind speed and the increased urban densities are ignored Renewable Energy Laboratory in 1978 for 248 locations using long-
[5,6]; urban areas are usually characterized by higher ambient air term observations of solar radiation and weather data from the
temperatures and considerably different microclimate, compared to 1952–1975 [14]. Later updates, in the early 1990 s, introduced the
areas outside the city centres, which is usually referred as the ur- TMY2 format derived from measurements during 1961–1990 [15],
ban heat island effect [7,8]. This differentiation is mainly attributed while the current TMY3 data sets cover 1020 sites across the US
to extensive urban morphological changes comprising of: (i) the using data from 1976–2005 or 1991–2005 [16]. The ‘typical year’
progressive substitution of vegetation and green areas with mineral is composed of 12 ‘Typical Meteorological Months’ [17], selected
surfaces, leading to a decrease of latent heat flux through evapo- from different years and concatenated so as to form a single year
transpiration [9], (ii) the replacement of natural, permeable sur- [18]. A detailed description of the creating process of a TMY is
faces with mineral, rough materials such as concrete and asphalt, presented in [19]. A similar compilation procedure, based on long-
the thermal properties of which, contribute to large quantities of term measurements (usually 20 years), is applied for the Test Ref-
solar radiation stored and then re-emitted as long-wave radiation erence Year (TRY), a technique developed by Chartered Institu-
inside the urban districts [10], (iii) the low albedo of the urban tion of Building Services Engineers (CIBSE) [20]. Finkelstein-Schafer
construction materials, which increases the quantity of solar radi- statistics are estimated for each month and each climatic variable
ation absorbed by building envelopes and urban surfaces, causing in order to define the typical months which are then aggregated
thus, their higher surface temperatures (Tsurf), and severe intensity to form a complete year; yet, there are two major differences with
of long wave radiation emission [11] and (iv) the reduced open- the TMY technique: (a) only the mean value of dry bulb tempera-
ing to the sky, contributing to the entrapment of shortwave and ture and wind speed and the global solar radiation are taken into
long wave radiation inside the canyons inhibiting thus, the urban account, instead of the 9 variables considered in the TMY method,
cooling [12]. As a result, when it comes to the energy performance (b) all three variables are equally weighted whereas on the TMY
simulation of an urban building, the assumption that the climatic technique, the global radiation is considered to be more critical,
conditions at a reference location of a meteo station (i.e. the air- having the highest weight factor. Another technique for generat-
port area) are the same for a densely built up area can lead to ing hourly weather dataset is the Weather Year for Energy Calcu-
inaccurate predictions of heating and cooling loads. This is not at- lation method (WYEC), initially proposed by [21]. It relies again
tributed to the altered solar radiation effects within the urban con- on the definition of 12 representative months in order to form a
text that are usually treated sufficiently by advanced BES models, complete year; still, there are two important differences in com-
but to the fact that the complex urban geometry and surface mate- parison with the previous techniques: individual months are only
rials strongly affect airflow, air temperature and relative humidity. selected if the average monthly dry bulb temperature has a dif-
In this context, the aim of this study is to present a compu- ference of ± 0.2 °C. After the initial selection, in case abnormalities
tational method so as to create typical weather datasets that ac- or extreme events are found, individual days or hours can be ad-
count for the influence of local, site specific surroundings and the justed so as to lead the monthly mean values to come closer to
respective microclimatic conditions. The proposed methodology is the respective long term values. The complete procedure of creat-
depicted in Fig. 1. More precisely, two typical hourly weather files ing hourly weather values with this technique along with the cor-
are generated: responding further improvements were described by [22] and [23].
Finally, a different approach is proposed in the Example Weather
• A reference one, corresponding at a reference location of a me- Year (EWY) method, introduced by [24]. A complete year is now
teo station. defined, the monthly mean weather values of which, contain the
• An urban specific weather dataset, reflecting microclimate con- least abnormalities in comparison with the long-term observations.
ditions inside an urban study area. Thus, a whole representative year instead of representative months
is sought. More precisely, monthly mean values of global and dif-
A stochastic generation procedure is carried out for the creation fuse radiation, daily mean wind speed, mean, maximum and min-
of both datasets; the detailed analysis of the corresponding pro- imum dry bulb temperature and their standard deviation from the
cess is presented in Section 3. It has to be underlined that the long-term mean are estimated; The years that contain monthly
generated ‘reference’ weather file could be theoretically used for means that differentiate more than standard deviation from the
every study area inside the city (hence, no specific microclimatic corresponding long-term mean are rejected and the last remaining
conditions of the district are considered) while the ‘urban spe- year would become the example year chosen [20,25].
cific’ weather dataset is only site specific, representing microcli-
matic conditions of the analysed study area. 3. Methodology for climate data generation
The two datasets are generated so as to be representative of cli-
matic conditions for a long time period, and not just for a single As already mentioned, the current study aims to present a
calendar year (see Section 3). In order to account of local microcli- computational method for the generation of typical weather years
mate conditions in the latter climatic file, the Envi-met non hydro- that account for the influence of local surroundings and the re-
static microclimate model [13] is applied. Finally the major climatic spective, site specific microclimatic conditions. Based on the pre-
variables of the two datasets, that strongly influence buildings’ en- viously mentioned techniques, when aiming at the generation
272 S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283

Fig. 1. Diagram of the implemented methodology.

of proper climatic datasets, several years of meteorological val- sonal profile, the selection of representative days should be per-
ues are needed; hence, Envi-met simulations are vastly com- formed through a statistical analysis of long-term measurements
puter time consuming (for an area 20 0mx20 0m and a grid size of climatic variables [28]. The corresponding statistical analysis is
1.5mx1.5mx3m, approximately 45 hours were needed), and as a presented in section 3.2.1. The paper is organised in the following
result, simulations are mainly run for a limited period of 24–72 way: At the first stage, the procedure for creating a typical ref-
hours. In our case, it would have been extremely inefficient to erence weather dataset, based on long-term measurements issued
simulate all diurnal cycles of several previous years so as to ob- from a meteo station is presented. At the second stage, the steps
tain the corresponding long-term microclimatic results. The above for generating a typical urban specific weather dataset that con-
mentioned limitation indicated the need to follow a different ap- sider the microclimate parameters are described while the third
proach: instead of simulating several years, 12 representative days, stage involves the comparison of the major climatic parameters of
one of each month, are simulated and the urban specific weather the two datasets that strongly influence the energy performance of
dataset, is then generated from the obtained microclimate results. buildings.
For this reason, Meteonorm, a synthetic year generation model
that uses stochastic methodologies to create time series of hourly 3.1. Generation of a reference weather dataset
data, is applied. When creating a synthetic weather year in Me-
teonorm, monthly averages of the most important weather vari- As previously mentioned, generating typical climate datasets,
ables (i.e. dry bulb temperature, relative humidity, global radia- necessitates the analysis of long-term observations of the ma-
tion and wind speed) can be introduced by the user; daily and jor climatic variables (i.e. air temperature, relative humidity, wind
hourly values are then stochastically generated with intermedi- speed and solar radiation). In this study, two sets of multiyear
ate data having the same statistical properties as the monthly im- measured data, recorded in the meteorological station of the Uni-
ported data, i.e. average value, variance and characteristic sequence versity campus, were provided by the Department of Meteorol-
[26]. It should be mentioned that in order for the generated hourly ogy and Climatology. The first dataset corresponds to the period
weather dataset to be ‘typical’, the imported monthly values of the of 1993–2003 and contains monthly average values of air temper-
climatic variables should correspond to 12 typical meteorological ature, relative humidity, wind speed and solar radiation, while the
months (TMMs), being representative of the weather conditions of second one represents the period from 1958 to 20 0 0, comprising
the area in question and issued from various calendar months in a of average daily values of air temperature. Given that in both cases,
long-term weather database climate data are only provided in monthly or daily average values,
Synthetic years and the corresponding generators have been they cannot be directly used in building performance dynamic sim-
vastly applied in many scientific studies where long term hourly ulation tools where an hourly resolution of weather parameters is
records of climate data were not available [17,27–29]. In the required. Hourly values will be thus stochastically generated from
present study the monthly averages that will be introduced in the the corresponding measured monthly averages, using Meteonorm
synthetic generator are provided as follows: Software which creates intermediate data having the same statisti-
cal properties as the measured data i.e. average value, variance and
• For the reference weather dataset (RWD), climatic averages are
characteristic sequence [26].
issued from the multiyear measured data at the meteorological
The first step for the creation of the reference weather dataset
station of the Aristotle University of Thessaloniki.
is the definition of individual typical meteorological months
• For the urban specific weather dataset (USWD), monthly aver-
(TMMs).Given that the dataset of 1993–2003 contains more
ages are provided by the Envi-met simulated microclimatic re-
weather parameters crucial to building energy analysis (i.e. dry
sults, for the selected 12 representative days.
bulb and dew point temperature, wind speed and total global ra-
It has to be emphasized that in order to create an urban specific diation), the selection of TMMs will be based on the monthly
dataset, the weather parameters of which follow a normal sea- averages of the corresponding climatic variables of the 10 years’
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 273

Table 1
Typical meteorological months for the city of Thessaloniki and the corresponding average values, introduced in Meteonorm.

Total global SR Tair aver Td aver Ws aver Tadmin aver Tadmax aver
(kWh/m2 ) °C °C m/s °C °C

January 1995 44.47 5.78 4.07 2.18 3.10 9.00


February 1997 63.78 7.63 5.39 2.04 3.50 12.60
March 20 0 0 103.91 9.33 6.79 2.02 4.80 14.50
April 2001 115.86 14.26 11.72 1.67 9.30 19.40
May 2002 145.48 19.60 16.12 1.57 14.90 24.30
June 20 0 0 164.64 24.20 20.61 1.92 18.30 30.10
July 1994 165.10 26.67 23.03 1.9 21.20 32.20
August 2001 155.60 27.60 23.57 1.69 22.50 33.20
September 20 0 0 115.50 21.78 17.66 1.57 16.50 28.40
October 2003 78.96 17.34 15.00 1.63 13.40 22.00
November 1997 43.02 11.80 10.10 1.30 9.20 15.00
December 2003 41.93 7.60 5.87 1.81 5.10 11.30

dataset. More precisely, the proposed method for the selection of climatic variables over the long-term period are needed. As a re-
typical meteorological months is the following: sult, in this study, the database of 1958–20 0 0, comprising of daily
average values of air temperature are used. The procedure that is
• The long-term average monthly values of the decade 1993– carried out for the determination of representative days for micro-
2003, are estimated from the corresponding monthly average climate simulation is the following:
values of air temperature, wind speed, relative humidity and
the monthly total global radiation. • For every day of each month, the mean value of the daily aver-
• The absolute % difference between average monthly values of age of air temperature is calculated, over the multi-year period.
each individual month and the corresponding long-term aver- For example, for the 1st of January and for the air temperature:
age is defined.
1st

• The month M of the year Y in which simultaneous discrepancy Tair mean = Tair aver_1st−1958 +Tair aver_1st−1959 +Tair aver_1st−1960
from the long-term average is the minimum for all 4 parame-

+ · · · +Tair aver_1st−2000 /n
ters is selected as typical.
Where n is the number of years of the long-term period.
In Table 1, the selected typical meteorological months for the
city of Thessaloniki and the corresponding average values of the • For every month, the median of the previously estimated mean
climatic variables that will be introduced in Meteonorm are de- values is calculated. For example, for the air temperature:
picted.  1st 2nd 31st

Median Tair mean, Tair mean . . . Tair mean

3.2. Generation of the urban specific weather dataset • For every day of the period September 2015-August 2016,1 the
deviation of the average day air temperature from the corre-
The non-hydrostatic Envi-met v.4 tool is applied for the eval- sponding median value was estimated. In order to determine
uation of the local microclimate conditions. Simulations at a mi- whether a day is representative or not, the deviation from the
cro scale are generally run for 24–72 hours with a typical resolu- median of the multi-year period should be the lowest possible
tion down to 0.5 m in space and 1- 5 sec in time. Given the above with maximum accepted deviation set to 20%.
time limitation and the significantly large duration of calculation, • In case where more than one day fulfilled this criterion, we re-
simulations for the evaluation of the microclimatic conditions are stricted the selection by simultaneously choosing the day with
mainly conducted for a diurnal cycle and thus, a specific day has to the smaller deviation and eliminating days with precipitation
be defined for the analysis. To date, many studies that have used as the Envi-met model can account of clouds but not of rain.
the Envi-met model do not provide any further description regard-
ing the methodology for selecting the simulation day [4,30–36] or Following the above mentioned methodology, the representa-
are based on random selection [37,38]. In other scientific works, tive days for the city of Thessaloniki were defined and are de-
extreme days during hot summer conditions are selected [39,40] or picted in Table 2. At the next step, microclimate simulations are
days that are characterised as typical [5,41–43]; yet, in the latter performed for all 12 days in order to acquire values of the major
studies, there is no further explanation of how the typical day was climate variables that will be introduced in the weather data gen-
derived and whether its definition was based on annual data or on erator.
longer observations. Within this context, this section aims to de-
scribe a methodology for the selection of representative days for
3.2.2. Microclimate simulation
microclimate simulation using long-term observations of climatic
3.2.2.1. Model description. The Envi-met v.4 three dimensional mi-
variables.
croclimate simulation tool is applied for the estimation of micro-
climatic variables. The model is based on the fundamental laws
3.2.1. Defining representative days for microclimate simulation of fluid dynamics and thermodynamics and can simulate complex
According to Tirabassi [44], a representative day could be de- surface-vegetation-air interactions in the urban environment [46].
fined as a set of 24 hours data, measured at a meteo station, that Two main input files are required for microclimate simulations
has the least differences from all the 24 hours long-term observa- with Envi-met tool: the area input file (Fig. 2b and 2c), where the
tions of the meteo station. In the same vein, Santamouris [45] sug- building layout, vegetation, soil type, receptors (i.e. specific grids
gested that the term “representative day” describes the day the cli-
mate variables of which, are close to the corresponding long-term
averages of the respective month. Taking into account the previous 1
Period in which on-site microclimate observations, necessary for model evaluation,
remarks, in order to define representative days, daily values of the were performed.
274 S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283

Fig. 2. (a) Aerial photo indicating the study area, (b) the Envi-met 3D area model and (c)the Envi-met plan of the study area.

Table 2 mer, through the use of experimental data [49]. The obtained re-
Selected representative days for the city of Thessaloniki and the corresponding air
sults revealed an average temperature difference of around 1.7 °C,
temperature (Tair) deviations from the respective long-term median values.
while peak differences up to 2 °C were observed in the very early
Selected representative day Tair deviation (%) morning and also during nighttime, due to the re-emission of the
4th of January 2016 7% heat that has been trapped in daytime in the urban fabric.
8th of February 2016 11% The densely built-up site, which is part of the municipality of
28th of March 2016 2.7% Thessaloniki, extends to 400m2 and contains 6 blocks of residential
21th of April 2016 1.8%
buildings, mainly constructed between 1970–1980 while the area
16th of May 2016 2%
1st of June 2016 1.1% reached its saturation during 90 s. Open spaces are only limited to
22st of July 2016 0.7% street canyons and courtyards of irregular shape between build-
10th of August 2016 5.5% ing volumes. The aspect ratio (H/W) (i.e. ratio of canyon height
24th of September 2015 0.6%
to width) of the main canyons (Pittakou, Kalliga and Voga) varies
1st of October 2015 1.8%
26th of November 2015 3.3% from 1.0 to 1.6 while there are two narrower canyons with H/W
20th of December 2015 2.7% ratio close to 2. The majority of the buildings are 7–8 storeys high
and building roofs are mostly covered by cement tiles. Ground sur-
face is shielded by asphalt and concrete paving materials and only
in the Envi-met model) and project location parameters are de- a small part is covered by loamy soil and other permeable mate-
fined and the configuration file, containing simulation settings re- rials. Vegetation consists of low trees and bushes placed on both
garding initialization values for meteorological parameters, defini- sides along the main streets of the study area while there is only
tion of output folder names and timings. Envi-met model has been a limited number of a tall mature tree.
applied in a great number of scientific studies dealing with urban
heat island and human heat stress mitigation strategies. The lat- 3.2.2.3. Model configuration. At the first step, numerical simula-
est version, released in 2014, overcomes important limits of the tions using Envi-met v.4 are initially performed for 24 hours di-
previous one, as it takes into account different U-values for the urnal cycles for representative days in March, July, October and
buildings’ envelope components but also thermal mass and heat February so as to assure model accuracy for all seasons of the year.
inertia of the building elements. Another important improvement Intermediate months of each season are selected for model evalua-
concerns the possibility of forcing air temperature and relative hu- tion expect from spring period where onsite measurements for the
midity input parameters by creating a user specified weather pro- selected representative day of April were not available. The area in-
file with hourly Tair and Relative Humidity (RH) data derived from put file is the same for all four simulations. Aiming at the numer-
nearby meteorological stations. The full forcing function is a major ical stability and the minimization of boundary effects which may
improvement, as it provides higher simulation accuracy and more affect the output data [34], 7 nesting grids are set around the main
realistic results [47]. model area. The meteorological input parameters and the ther-
mal properties of the ground surfaces are depicted in Table 3 and
3.2.2.2. Study area description. The study area, (Fig. 2a) is located Table 4 respectively. For the microclimatic simulations, hourly val-
at 3 km east from the center of Thessaloniki (40.65°N 22.9°E). ues of air temperature and relative humidity, recorded at the mete-
The city is situated in the northern part of Greece and placed orological station inside the University Campus during the specific
along the North-East coast of Thermaikos gulf. The municipality of representative days are used as boundary conditions of the model
Thessaloniki, had a population of around 386.0 0 0 in 20 07, while (i.e. for the manual diurnal definition of the variables that enables
the larger metropolitan area had around 1.104.400 inhabitants [48]. the full forcing of Tair and RH); wind speed and direction on the
The climate of the city is characterized by temperate Mediter- model boundaries correspond to the mean daily value, recorded
ranean conditions, with generally hot, dry summers, and mild and at the logger of the meteorological station, at 10 m height. Build-
wet winters; yet, significant seasonal variations can arise. During ing walls and roofs are made of bricks and concrete respectively;
winter, time records indicate daily temperatures that reach 6.5 °C albedo values were set to 0.4 and 0.3 for roofs and walls corre-
on average, falling to 2.1 °C overnight; however, during the coldest spondingly.
winters temperatures can even drop below − 10 °C. During sum- At the next step and after assuring the Envi-met model’s accu-
mer, average high and low temperatures are around 30.5 °C and racy under different seasonal conditions, microclimate simulations
17.5 °C respectively, while the recorded maximum air temperatures are performed for all representative days, depicted in Table 2. Once
of the last decades rarely exceeded 40 °C. Previous scientific stud- again, for the rest of the simulations, the meteorological input vari-
ies have investigated the magnitude of air temperature differences ables at the inflow model boundary (i.e. mean daily wind speed,
between central and suburban areas of Thessaloniki during sum- wind direction, diurnal variation of air temperature and relative
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 275

Table 3
Description of the model meteorological boundary configuration for each seasonal simulation.

Meteorological variable Spring Period Summer Period Autumn period Winter period
28th of March 22th of July 1st of October 4th of January

Wind speed at 10m 2.75 m/s 1.15 m/s 2.30 m/s 1.30 m/s
Wind direction WSW SW SW W
Roughness length 0.1 0.1 0.1 0.1
Air temperature forcing Hourly data from Hourly data from Hourly data from Hourly data from
meteo station meteo station meteo station meteo station
Relative humidity forcing Hourly data from Hourly data from Hourly data from Hourly data from
meteo station meteo station meteo station meteo station
Cloud coverage No clouds No clouds No clouds No clouds
Specific humidity at 2500m 7 g/kg 7 g/kg 7 g/kg 7 g/kg
Solar adjustment factor 1 1 1 1
Soil upper layer(0–20 cm) initial temperature (data from meteo station) 284.3K 305 K 292 K 280.3 K
Soil upper layer(20–50 cm) initial temperature (data from meteo station) 285 K 304 K 293 K 282 K
Soil upper layer(50–200 cm) initial temperature (data from meteo station) 290 K 293 K 293 K 290 K
Soil deeper layer initial temperature 293 K 293 K 293 K 293 K
Soil upper layer(0–20 cm) moisture content∗ 50% 50% 50% 50%
Soil upper layer(20–50 cm) moisture content∗ 60% 60% 60% 60%
Soil upper layer(50–200 cm) moisture content ∗ 60% 60% 60% 60%
Soil deeper layer moisture content∗ 60% 60% 60% 60%

Envi-met default values.

Table 4
Thermal properties of ground surface materials [50,51].

Material Emissivity Solar Volumetric heat Thermal


reflectance capacity Cp conductivity λ
(J/m3 K) (W/mK)

Asphalt 0.90 0.10 2.1 × 106 0.7


Concrete tiles 0.90 0.30 2.1 × 106 1.5
Soil 0.95 0.20 3.0 × 106 1.5

humidity) are issued from onsite records of the meteorological sta- Fig. 3. Placement of the logger on a balcony of an apartment in Pittakou Street.
tion for the specific days. The logger was mounted on a tripod and protected from direct and reflected solar
radiation, using a suitable solar radiation shield.

4. Results

4.1. Envi-met model evaluation of the main streets of the study area, at a height of nearly 4.5 m
(Fig. 3). In order to enhance the accuracy of temperature and RH
The accuracy of the microclimate simulation results strongly measurements, the logger was protected from direct and reflected
depends on the initial boundary conditions and the input data; solar radiation with a suitable radiation shield. Measurements of
therefore special attention should be given when defining the ini- air temperature and relative humidity were taken for all 4 seasons
tial parameters [52]. Based on the recommendations of Willmott during the year, at an interval of 30 minutes.
[53], the statistical evaluation of the model was carried out us- The obtained readings are then used to investigate current mi-
ing a set of indices that describe the magnitude of difference be- croclimatic conditions but also to evaluate the microclimate simu-
tween observation and prediction. The recommended measures in- lation tool. Simulations are performed for the representative days
clude the root mean square error (RMSE), the mean absolute er- defined previously and the results are compared with onsite mea-
ror and the index of agreement (d). The RMSE and the MAE pro- surements of air temperature and relative humidity. For this rea-
vide information on the average error and are generally considered son, virtual receptors were placed in the measurement points of
suitable measures for the model’s performance. Regarding the in- the study area, inside the Envi-met model, in order to acquire mi-
dex of agreement, it indicates how possible is for the model to croclimate results at different heights. Simulation results for air
predict a variable without errors with values ranging from 0 to temperature and relative humidity are also compared with obser-
1.A monitoring project of air temperature and relative humidity vations recorded in the meteo station of the Meteorology and Cli-
was carried out during the period September 2015-August 2016, matology Department of the Aristotle University. The comparison
using high accuracy and weatherproof Onset Hobo data loggers between modeled and observed values for both air temperature
with built-in temperature and relative humidity sensors, suitable and relative humidity for all seasons is shown in Fig. 4; the quan-
for outdoor environments. The technical characteristics of the log- titative evaluation using statistical indices is presented in Table 5
gers that were used are the following: while in Table 6, the results of Envi-met evaluation in previous
studies and the corresponding statistical indices are reported. To
• Model name: HOBO U23001.
date, most of the studies that have evaluated the accuracy of the
• Operation range: − 40 °C to 70 °C for the temperature sensor
Envi-met model were mainly based on air, surface and mean ra-
and 0 to 100% for the relative humidity sensor.
diant temperature measurements [30,34,36,38,40,47,52,54–58] and
• Logger resolution: 0.02 °C for the temperature sensor and 0.03%
only a few have evaluated the model’s performance with relative
for the relative humidity sensor.
humidity values [31,59] or specific humidity values [57,34].
The logger was mounted on a tripod and was then placed on a As presented in Fig. 4, during all seasons, air temperatures mea-
first floor balcony of an apartment in Pittakou Street, which is one sured inside the study area are always higher than the corre-
276 S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283

Fig. 4. Evolution of measured and simulated Tair and RH for a representative day in (a) March, (b) July, (c) October, (d) January.
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 277

Fig. 5. (a) 3D representation of the studied apartment in Google earth and (b) in the Envi-met model, (c) position of receptors in front of the studied apartment.

Table 5 (a) the urban heat island effect and the corresponding higher air
Quantitative evaluation of the Envi-met model based on the calculation of RMSE,
temperature inside the urban district is not handled by the model
MAE and d.
since the air temperatures on the model boundaries are issued
RMSE( °C) MAE( °C) d from a meteo station outside the study area, (b) the effect of mul-
Tair ( °C) March 1.14 0.95 0.93 tiple reflections and the entrapment of longwave radiation due to
July 1.02 0.82 0.94 reduced sky view factor are also not taken into consideration dur-
October 1.55 1.50 0.65 ing the calculations, (c) the anthropogenic heat flux resulting from
January 0.85 0.78 0.70
the use of air conditioners, transportation and other human activ-
RMSE(% ) MAE(% ) d
RH(%) March 6.70 2.50 0.90 ities, the effect of which is more prominent in warmer periods, is
July 10.24 7.70 0.97 not taken into account during the microclimate simulation and (d)
October 4.80 3.50 0.98 the sea breeze effect, mainly occurring during the warm period is
January 4.7 4.3 0.97
not taken into account by the model. As previous studies indicated,
in the city of Thessaloniki, the mean frequency of the sea breeze
Table 6
occurrence is around 17 days per month, during the warm period
Overview of deviations in air temperature between simulation and observation in of the year [60]. As the study area is at a rather close distance
other studies. from the sea (approximately 300 m), the observed air temperatures
Study Simulation period Variable RMSE[°C] MAE[°C] d
during summer will be strongly influenced by the sea breeze. The
latter remark could thus explain the overestimation of air temper-
[40] summer Tair 2.79 2.26–3.49 0.86–0.92
atures in July, occurring from 9:00 to 12:30 and with an average
[52] summer Tair 1.51 – 0.97
[55] autumn Tair 1.31–1.53 1.16–1.33 0.40–0.47 discrepancy of 1.10 °C. Concerning relative humidity results, during
[57] summer Tair 0.96–1.68 0.86–1.43 – spring and winter simulations, the model tends to overestimate the
[56] summer Tair 1.06–2.73 – – relative humidity values with better agreement being reported in
[58] autumn Tair 1.11–1.41 0.94–1.21 0.87–0.91
January; However, underestimation and higher deviations were re-
winter 0.66–0.89 0.60–0.61 0.95–0.97
summer 0.52–0.75 0.40–0.62 0.95–0.98
ported during summer simulations, mainly attributed to the fact
[30] autumn Tair 3.84–4.32 3.35–3.67 0.69–0.71 that the study area is strongly influence by the sea breeze and the
winter 2.45–2.77 2.14–2.38 0.75–0.82 corresponding cooler air, introduced inside the urban canyons, a
spring 2.95–3.97 2.56–3.28 0.68–0.84 phenomenon that cannot be reproduced by the model.
summer 1.46–2.92 1.35–2.46 0.75–0.94
Based on the previous remarks and on statistical indices’ val-
ues already accepted in previous evaluation studies (Table 6), Envi-
met can be considered as a reliable tool for microclimate simula-
sponding values recorded in the meteo station. During spring and tion under different meteorological conditions. Microclimate sim-
summer, the average discrepancy is 1.30 °C and 1.19 °C respectively ulations for the rest of the representative days are then carried
while in winter, the average difference is close to 1.20 °C. This can out and the acquired microclimatic averages are introduced in the
be explained by the fact that meteorological station of the Uni- stochastic weather generator so as to create the USWD file.
versity is placed in a park area, and rather far from heavy traf-
fic, with increased vegetation consisting of tall and densely foliated 4.2. Application of the method
trees, lower solar gains, lower building density and thus better mi-
croclimatic conditions. In terms of air temperature results, during The proposed computational method is applied at a1st floor
spring, summer and winter simulations, the Envi-met model tends apartment of Pittakou Street, which is one of the main canyons of
to over-estimate part of night-time temperatures (i.e. from 0 0:0 0 the study area. The hourly weather file is created in order to rep-
to 2:00) and under-estimate day-time temperatures, a feature that resent the immediate climatic conditions in front of the specific
has also been reported in previous studies [40,30,56]. Regarding building unit. Only the main façade is exposed to exterior condi-
autumn simulations, the model always underestimates air temper- tions while the ground surface, ceiling and the rest of the vertical
ature and a similar trend was previously reported in other stud- facades are considered as adiabatic due to the same operational
ies [41]. Some factors that could explain the differences between schedules between the apartments. The ground floor of the anal-
modeled and measured air temperature values are the following: ysed building has a 4.5 m height while typical building floors were
278 S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283

Fig. 6. Inter–annual daily temperatures (a), Relative Humidity (c) and wind speed (e) values for the urban specific weather dataset and the reference weather dataset as
well as the mean monthly values for the same three parameters for the urban specific weather dataset and reference weather dataset) over a typical year.

of 3 m height. Thus, for a 1st floor apartment, microclimate data 4.3. Comparison of the generated annual weather datasets
necessary for BES modelling were extracted for heights between
4.50 m and 7.50 m. Based on the recommendations of Yang et al. The analysis of the air temperature (dry bulb temperatures) be-
[5] and Morakinyo et al. [31], five receptor points providing micro- tween the RWD and the USWD, on a daily basis throughout the
climatic values of air temperature, relative humidity, wind speed, year, showed that both data sets follow a similar inter-annual pro-
shortwave radiation, mean radiant temperature etc. were placed at file with the latter one generally presenting higher Tair values
0.75 m away from the façade (Fig. 5); The obtained microclimate (Fig. 6a). The range of the daily dry bulb temperature vary be-
values from the five receptors were then averaged and the esti- tween −1.4 °C to 31.1 °C and −1.5 °C to 30.4 °C for the urban spe-
mated values were then used as the monthly input data for the cific and the reference weather dataset respectively. Aiming on a
Meteonorm weather generator. more thorough analysis of the results, the mean monthly values
Based on the methodology presented in Section 2, the two for both data sets are calculated (Fig. 6b). For every case, air tem-
typical hourly weather files are generated (i.e. Reference Weather peratures of the RWD file are lower than the corresponding USWD
Dataset, RWD and Urban Specific Weather Dataset, USWD); in values, indicating the effect of different microclimatic conditions
section 4.3, climatic variables of the USWD, involving air temper- inside the densely built-up area, which are sufficiently reproduced
ature, wind speed and relative humidity, are compared with the by the Envi-met model. The smaller differences are found in Jan-
corresponding values of RWD. uary (0.1 °C) where the study area is practically kept in shadow
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 279

Fig. 7. Inter- daily air temperature variation for each month of the year for the Urban Specific Weather Dataset (red curve) and the Reference Weather Dataset (blue line).

during all day and the solar radiation effect is significantly low, the corresponding RWD value by 0.5 m/sec. This can be explained
while the highest ones are observed in September (1.5 °C). by the fact that the parameter of wind speed was not considered
In terms of the relative humidity values, it should be mentioned during the definition of the representative days for microclimate
that the stochastic model primarily simulates the dew point tem- simulation. As a result, potential wind gusts or extreme wind con-
peratures which were then transformed/converted to relative hu- ditions during that day, provoking an important rise of the average
midity values (%). Contrarily to the temperature results, the inter- daily value of wind speed, were not taken into account.
annual daily values of this parameter differs quite substantially On a final step, the performance of the method was investi-
(the correlation coefficient does not exceed 0.4), with the USWD gated on an hourly basis. More precisely, for both USWD and RWD
values being higher, in general, during the cold period of the year, datasets and for each month, the mean hourly values of the three
and lower to the RWD ones during the warmer season (Fig. 6c). examined parameters are calculated and compared. Fig. 7 illus-
This finding is confirmed by the mean monthly values with posi- trates the results concerning the hourly air temperature values. It
tive differences (USWD – RWD) for January, February, March and is observed that both datasets follow the expected inter-daily tem-
November and negative differences for the period March – August perature variability with the lowest values early in the morning
and October. (Fig. 6d). and the maximum ones two or three hours after mid-day. The air
Regarding wind speed, considerable differences are generally temperature values of USWD are generally higher than the RWD
observed in the day to day USWD and RWD datasets (Fig. 6e) with ones, with the largest differences detected during the month of
wind speed values of the latter one, being generally higher than September. The reported discrepancies between the study area and
the urban specific values. In the same vein, the mean monthly the park area of the meteo station are the result of complex mi-
values analysis showed that for the majority of the months, the croclimatic conditions, reproduced by Envi-met model and mainly
RWD wind speeds are higher, with the largest differences found in attributed to the increased building density of the study area, con-
January (1.3 m/sec) followed by June and July with (1.0 m/sec) and tributing to lower longwave radiation losses, to the thermal prop-
the lowest ones in September (0.3 m/sec) and October (0.02 m/sec) erties of the construction materials, enhancing the storage of emit-
(Fig. 6f). The observed deviations can be attributed to the atten- tance of sensible heat to the air and also to the reduced green ar-
uation of the wind speed inside the urban district due to the in- eas leading to lower amounts of latent heat. The only exception is
creased building heights and high building density, compared to for the month of January where the USWD and RWD mean hourly
the corresponding values observed at the reference location, in an values are very similar and the computed curves almost match
open field and less sheltered by constructions. However, it is worth each other. As already mentioned, the whole study area is kept in
mentioning that during May, the USWD wind speed is higher than
280 S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283

Fig. 8. Inter- daily relative humidity variation for each month of the year for the Urban Specific Weather Dataset (red curve) and the Reference Weather Dataset (blue line).

shadow for most of the daytime and thus, solar radiation does in- days, are neglected (Fig. 9). The latter remark could also explain
crease the ambient air temperature. the fact that, during October, the USWD and the RWD time series
The behavior of the method in the reproduction of the dew are almost identical.
point temperature-relative humidity is rather more complicated
(Fig. 8). In general, the maximum is observed during the early 5. Discussion
morning hours (5 to 7 pm) and the minimum two or three hours
after mid-day. This pattern is almost identical during all months. To date, most of the BES models adopt weather files, based on
However, during the cold period of the year, the mean hourly rel- long-term measurements recorded at meteorological stations out-
ative humidity model values of the USWD dataset are higher than side the city centres and consider buildings as stand-alone [6].
the RWD ones, which could be partly attributed to the entrap- As mentioned by Guattari et al. [61], Typical Meteorological Years
ment of humidity in front of the building façade due to low so- (issued from multiyear records of meteo stations away from city
lar radiation values and reduced wind speed. On the other hand, centres) could be considered more consistent for energy perfor-
for the warm period (i.e. April-October) during which solar radi- mance calculations of buildings located in new residential areas in
ation intensity is more important, USWD relative humidity values the peripheral zones, characterized by lower building densities and
are lower than the corresponding of the RWD file with the greatest higher vegetation coverage. However, when it comes to simulations
deviations generally reported from 13:00 to 18:00 in the afternoon. of urban buildings and given that climatic parameters strongly in-
Regarding mean hourly wind speed, in most months, both fluence the output of energy simulations [62], accounting of com-
datasets follow a similar inter-day profile with maximum values plex interactions between climatic variables and increased urban
being observed between 14:00 and 17:00 and a secondary maxi- densities, is of high importance. In this study, a typical urban spe-
mum (much smaller one) from 2:00 to 3:00 in the morning. Once cific weather dataset, reflecting the microclimatic conditions in
again, USWD wind speed values are significantly lower, compared front of a building unit inside an urban district in Thessaloniki,
to the RWD ones, as wind speed inside the urban district is con- was generated based on microclimate simulations with the Envi-
siderably reduced due to obstacles created by building volumes. met model; it was then compared with a typical reference weather
Hence, the opposite phenomenon is observed during May, where dataset, representing climatic conditions at a reference location of
the urban specific mean hourly values are higher than the cor- a meteo station.
responding values at the reference location. As previously men- The analysis revealed that dry bulb temperature is generally
tioned, this exception is attributed to the fact that the parame- higher inside the urban canyon, compared to the corresponding
ter of wind speed was not taken into account during the selec- values at the reference location, with indicative mean daily devi-
tion of representative days; as a result, extreme wind speed gusts ations reaching 1.0 °C, 0.63 °C and 0.75 °C in February, March and
or even very calm wind conditions occurring in the representative July respectively. The tendency of the urban area to present higher
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 281

Fig. 9. Inter- daily wind speed variation for each month of the year for the Urban Specific Weather Dataset (red curve) and the Reference Weather Dataset (blue line).

dry bulb temperatures was confirmed with onsite measurements, wall and the corresponding cooling of the building unit envelope
carried out in the study area and the meteo station during 20 will be considerably reduced. The importance of accounting of the
consecutive days in February, March and July 2016; the average wind speed reduction inside the urban canyons and its influence
daily difference of air temperatures was estimated close to 1.10 °C, on the buildings energy simulation results has already been em-
1.50 °C and 1.20 °C respectively. It has to be mentioned that the phasized in previous studies [67]. However, since wind speed pa-
magnitudes of air temperature deviations, obtained by the onsite rameter was not taken into consideration during the definition of
measurements and the computational method are different as the the representative day, due to lack of long term daily wind speed
calendar year 2016 and the corresponding climatic parameters can- data, potential wind gusts or very calm conditions were not cap-
not be considered as typical; yet, the datasets generated by the tured resulting thus in the opposite profile for the month of May
proposed methodology are created so as to be representative of where USWD wind speeds are higher than the corresponding val-
climatic conditions for a long time period, and not just for a sin- ues of the reference location. In order to solve this problem, long-
gle calendar year. The higher ambient air temperatures reported term daily values of wind speed would be also needed; the statis-
inside the urban areas are expected to have a great impact on tical procedure of the Section 3.2.1 could be then followed and the
the estimated building energy needs for heating and cooling pur- selection of the representative day would be based at the simul-
poses [63]. On the one hand, higher winter temperatures will have taneous lowest deviation from the long-term median, for both cli-
a positive impact on lowering the heating energy needs but on the matic parameters; yet, it has to be emphasized that the parameter
other hand, in cities characterized by cooling dominated climatic of wind speed is highly influenced by local features and temporal
conditions, such as Thessaloniki [64], high urban air temperatures variations can be very large; defining representative wind speed
will considerably increase cooling energy needs and peak electric- conditions can be thus a rather difficult task to achieve. Similarly
ity demand. Findings of a recent literature review suggest that the for the parameter of relative humidity, simulated values inside the
rise of the peak electricity load per degree of air temperature in- urban district were found generally higher during the cold period
crease, ranges between 0.45% and 4.6% [65], while the comparative of the year, and lower than the RWD ones during the warmer sea-
analysis of existing scientific results revealed that the average in- son. Based on indications of existing studies, this differentiation
crease of the cooling load due to the urban warming is close to can have a significant impact on the dynamic energy simulations
13%, depending on the intensity of the phenomenon and the build- as the latent loads of the building unit will considerably differ, af-
ing characteristics [66]. fecting the sizing of the HVAC system [62,68]. To sum up, find-
In terms of wind profile, the acquired simulation results indi- ings of the present paper indicate the importance of accounting
cated that wind speed inside the urban canyon, near the build- of local microclimatic conditions, when aiming at high accuracy of
ing unit façade is generally lower than the corresponding values building energy calculations. Future work comprises of two runs of
at the reference location, due to wind sheltering by neighbour- energy performance simulations of a typical building unit, by re-
ing buildings. As a result, the convective heat flux normal to the placing the existing weather file in the BES model with the two
282 S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283

generated datasets. The quantitative evaluation of the obtained re- [16] S. Wilcox, W. Marion, Users Manual For TMY3 Data Sets, National Renewable
sults will then indicate the effect of microclimatic conditions of the Energy Laboratory, Golden, CO, 2008.
[17] R.S. McLeod, C.J. Hopfe, Y. Rezgui, A proposed method for generating high res-
study area, on heating and cooling needs. olution current and future climate data for Passivhaus design, Energy Build. 55
(2012) 481–493.
6. Main conclusions [18] J.M. Finkelstein, R.E. Schafer, Improved goodness-of-fit tests, Biometrika 58 (3)
(1971) 641–645.
[19] M. Gazela, E. Mathioulakis, A new method for typical weather data selection
This paper presents a computational method for assessing the to evaluate long-term performance of solar energy systems, Sol. Energy 70 (4)
urban climate’s effect during the creation of typical weather years (2001) 339–348.
[20] M. Ozdenefe, J. Dewsbury, Simulation and real weather data: a comparison for
for dynamic building energy calculations. In this vein, the results of Cyprus case, Build. Serv. Eng. Res. Technol. 37 (3) (2016) 288–297.
Envi-met microclimate simulations for 12 representative days are [21] L.W. Crow, Development of hourly data for weather year for energy calcula-
used as input in Meteonorm stochastic weather generator so as to tions (WYEC), including solar data, at 21 stations throughout the US, ASHRAE
Trans. 87 (1) (1981) 896–900.
create an urban specific weather dataset (USWD), which is finally [22] L.W. Crow, Weather year for energy calculations, ASHRAE J. (United States) 26
compared with a reference stochastically created weather dataset (1984).
(RWD) reflecting climatic conditions at a meteo station. The analy- [23] J.R. Augustyn, WYEC2 user’s manual and software toolkit, ASHRAE Trans. 104
(1998) 32.
sis initially revealed that the Envi-met model can sufficiently re- [24] M. Holmes, E. Hitchin, An example year for the calculation of energy demand
produce special microclimatic conditions of an urban area. Fur- in buildings, Build. Serv. Eng. 45 (1) (1978) 186–189.
thermore, for all 12 months of the year, air temperatures of the [25] W. Wong, K. Ngan, Selection of an “example weather year” for Hong Kong,
Energy Build. 19 (4) (1993) 313–316.
RWD file are inferior to the corresponding USWD values, indicating
[26] Meteotest, Meteonorm, Global Meteorological Database, Version 7.1, Handbook
the effect of dissimilar climatic conditions inside the dense urban Part I, (2015).
district compared to the meteo station. In addition, USWD wind [27] F.J.S. de la Flor, S.Á. Domínguez, J.L.M. Félix, R.G. Falcón, Climatic zoning and its
speed values are generally found lower, compared to the RWD ones application to Spanish building energy performance regulations, Energy Build.
40 (10) (2008) 1984–1990.
due to the attenuation of wind flow, provoked by building vol- [28] S. Oxizidis, A. Dudek, A. Papadopoulos, A computational method to assess the
umes. Further research would be necessary so as to quantify the impact of urban climate on buildings using modeled climatic data, Energy
impact of the differentiated microclimatic conditions of the urban Build. 40 (3) (2008) 215–223.
[29] H. Radhi, A comparison of the accuracy of building energy analysis in Bahrain
area on heating and cooling needs of an urban building unit. using data from different weather periods, Renewable Energy 34 (3) (2009)
869–875.
Acknowledgements [30] B.C. Hedquist, A.J. Brazel, Seasonal variability of temperatures and outdoor hu-
man comfort in Phoenix, Arizona, USA, Build. Environ. 72 (2014) 377–388.
[31] T.E. Morakinyo, K.K.C. Dahanayake, O.B. Adegun, A.A. Balogun, Modelling the
The authors would like to thank the Department of Meteorol- effect of tree-shading on summer indoor and outdoor thermal condition
ogy and Climatology of the Aristotle University in Thessaloniki, for of two similar buildings in a Nigerian university, Energy Build. 130 (2016)
721–732.
the provision of climate data. [32] E. Ng, L. Chen, Y. Wang, C. Yuan, A study on the cooling effects of greening in
a high-density city: an experience from Hong Kong, Build. Environ. 47 (2012)
References 256–271.
[33] F. Salata, I. Golasi, R. de Lieto Vollaro, A. de Lieto Vollaro, Urban microclimate
[1] E.E.-. Commission, Directive 2002/91/EC of the European Parliament and of the and outdoor thermal comfort. A proper procedure to fit ENVI-met simulation
Council of 16th December 2002 on the energy performance of buildings, Off. outputs to experimental data, Sustainable Cities Soc. 26 (2016) 318–343.
J. Eur. Communities (2003) L1/65–L1/71, 04/01/2003. [34] X. Yang, L. Zhao, M. Bruse, Q. Meng, Evaluation of a microclimate model for
[2] A. Yezioro, B. Dong, F. Leite, An applied artificial intelligence approach towards predicting the thermal behavior of different ground surfaces, Build. Environ.
assessing building performance simulation tools, Energy Build. 40 (4) (2008) 60 (2013) 93–104.
612–620. [35] A. Qaid, H.B. Lamit, D.R. Ossen, R.N.R. Shahminan, Urban heat island and ther-
[3] I.E. Annex, 53, Annex 53 total energy use in buildings: analysis & evaluation mal comfort conditions at micro-climate scale in a tropical planned city, En-
methods. ergy Build. 133 (2016) 577–595.
[4] D. Kolokotsa, K. Gobakis, S. Papantoniou, C. Georgatou, N. Kampelis, [36] N.L. Alchapar, C.C. Pezzuto, E.N. Correa, L.C. Labaki, The impact of different
K. Kalaitzakis, K. Vasilakopoulou, M. Santamouris, Development of a web based cooling strategies on urban air temperatures: the cases of Campinas, Brazil and
energy management system for University Campuses: The CAMP-IT platform, Mendoza, Argentina, Theor. Appl. Climatol. (2016) 1–16.
Energy Build. 123 (2016) 119–135. [37] E. Carnielo, M. Zinzi, Optical and thermal characterisation of cool asphalts to
[5] X. Yang, L. Zhao, M. Bruse, Q. Meng, An integrated simulation method for mitigate urban temperatures and building cooling demand, Build. Environ. 60
building energy performance assessment in urban environments, Energy Build. (2013) 56–65.
54 (2012) 243–251. [38] M. Taleghani, L. Kleerekoper, M. Tenpierik, A. van den Dobbelsteen, Outdoor
[6] J. Allegrini, V. Dorer, J. Carmeliet, Influence of the urban microclimate in street thermal comfort within five different urban forms in the Netherlands, Build.
canyons on the energy demand for space cooling and heating of buildings, En- Environ. 83 (2015) 65–78.
ergy Build. 55 (2012) 823–832. [39] C. Skelhorn, S. Lindley, G. Levermore, The impact of vegetation types on air and
[7] H. Akbari, A. Rosenfeld, H. Taha, Summer Heat Islands, Urban Trees, and White surface temperatures in a temperate city: a fine scale assessment in Manch-
Surfaces, Energy Analysis Program, Center for Building Science, Applied Science ester, UK, Landscape Urban Plann. 121 (2014) 129–140.
Division, Lawrence Berkeley Laboratory, University of California, 1990. [40] W.T. Chow, A.J. Brazel, Assessing xeriscaping as a sustainable heat island miti-
[8] H. Akbari, C. Cartalis, D. Kolokotsa, A. Muscio, A.L. Pisello, F. Rossi, M. Santa- gation approach for a desert city, Build. Environ. 47 (2012) 170–181.
mouris, A. Synnefa, N.H. Wong, M. Zinzi, Local climate change and urban heat [41] J.A. Acero, J. Arrizabalaga, Evaluating the performance of ENVI-met model in
island mitigation techniques–the state of the art, J. Civil Eng. Manage. 22 (1) diurnal cycles for different meteorological conditions, Theor. Appl. Climatol.
(2016) 1–16. (2016) 1–15.
[9] E. Gago, J. Roldan, R. Pacheco-Torres, J. Ordoñez, The city and urban heat is- [42] J. Spangenberg, P. Shinzato, E. Johansson, D. Duarte, Simulation of the influence
lands: a review of strategies to mitigate adverse effects, Renewable Sustainable of vegetation on microclimate and thermal comfort in the city of São Paulo,
Energy Rev. 25 (2013) 749–758. Revista da Sociedade Brasileira de Arborização Urbana 3 (2008) 1–19.
[10] M. Santamouris, A. Synnefa, T. Karlessi, Using advanced cool materials in the [43] A. Middel, K. Häb, A.J. Brazel, C.A. Martin, S. Guhathakurta, Impact of ur-
urban built environment to mitigate heat islands and improve thermal comfort ban form and design on mid-afternoon microclimate in Phoenix Local Climate
conditions, Sol. Energy 85 (12) (2011) 3085–3102. Zones, Landscape Urban Plann. 122 (2014) 16–28.
[11] H. Taha, Urban climates and heat islands: albedo, evapotranspiration, and an- [44] T. Tirabassi, S. Nassetti, The representative day, Atmos. Environ. 33 (15) (1999)
thropogenic heat, Energy Build. 25 (2) (1997) 99–103. 2427–2434.
[12] T.R. Oke, The energetic basis of the urban heat island, Q. J. R. Meteorolog. Soc. [45] M. Santamouris, Advances in building energy research, Earthscan (2012).
108 (455) (1982) 1–24. [46] Envi-met, ENVI-met model description, in, http://www.envi-met.com/
[13] S. Huttner, Further Development and Application of the 3D Microclimate Sim- #section/intro.
ulation ENVI-met, Mainz University, Germany, 2012. [47] H. Lee, H. Mayer, L. Chen, Contribution of trees and grasslands to the mit-
[14] I.J. Hall, R. Prairie, H. Anderson, E. Boes, Generation of a Typical Meteorological igation of human heat stress in a residential district of Freiburg, Southwest
Year, in, Sandia Labs., Albuquerque, NM (USA), 1978. Germany, Landscape Urban Plann. 148 (2016) 37–50.
[15] M. William, K. Urban, User’s Manual for TMY2S, National Renewable Energy [48] ESYE, Concise Statistical Yearbook 2007, http://dlib.statistics.gr/portal/page/
Laboratory, Golden, CO, USA, 1996 Also can be accessed at http://rredc.nrel. portal/ESYE, (2008).
gov.solar/old_data/nsrdb/tmy2.
S. Tsoka et al. / Energy & Buildings 165 (2018) 270–283 283

[49] T. Giannaros, D. Melas, P. Kontogianni, A. Angelopoulos, An observational study [59] N. Müller, W. Kuttler, A.-B. Barlag, Counteracting urban climate change: adap-
of the urban heat island in the greater Thessaloniki area: preliminary re- tation measures and their effect on thermal comfort, Theor. Appl. Climatol. 115
sults and development of a forecasting service, in: AIP Conference Proceedings, (1-2) (2014) 243–257.
2010, p. 991. [60] C.S. Sahsamanoglou, The sea breeze in Thessaloniki, Bull. Hell, Meteorol. Soc 1
[50] E. ISO, 10456: Building materials and products-Hygrothermal properties- (1976) 19–33.
Tabulated design values and procedures for determining declared and design [61] C. Guattari, L. Evangelisti, C.A. Balaras, On the assessment of urban heat island
thermal values (ISO 10456: 2007), in, CEN, 2007. phenomenon and its effects on building energy performance: a case study of
[51] E. ISO, 13370: Thermal performance of buildings-Heat transfer via the ground- Rome (Italy), Energy Build. (2017). https://doi.org/10.1016/j.enbuild.2017.10.050.
Calculation methods (ISO 13370: 2007), in, CEN, 2007. [62] M. Bhandari, S. Shrestha, J. New, Evaluation of weather datasets for building
[52] A. Maleki, K. Kiesel, M. Vuckovic, A. Mahdavi, Empirical and Computational energy simulation, Energy Build. 49 (2012) 109–118.
Issues of Microclimate Simulation, in: Information and Communication Tech- [63] H. Akbari, M. Pomerantz, H. Taha, Cool surfaces and shade trees to reduce en-
nology-EurAsia Conference, Springer, 2014, pp. 78–85. ergy use and improve air quality in urban areas, Sol. Energy 70 (3) (2001)
[53] C.J. Willmott, Some comments on the evaluation of model performance, Bull. 295–310.
Am. Meteorol. Soc. 63 (11) (1982) 1309–1313. [64] K. Tsikaloudaki, K. Laskos, D. Bikas, On the establishment of climatic zones
[54] G. Kyriakodis, M. Santamouris, Using reflective pavements to mitigate urban in Europe with regard to the energy performance of buildings, Energies 5 (1)
heat island in warm climates-Results from a large scale urban mitigation (2011) 32–44.
project, Urban Climate (2017). [65] M. Santamouris, C. Cartalis, A. Synnefa, D. Kolokotsa, On the impact of urban
[55] W.T. Chow, R.L. Pope, C.A. Martin, A.J. Brazel, Observing and modeling the noc- heat island and global warming on the power demand and electricity con-
turnal park cool island of an arid city: horizontal and vertical impacts, Theor. sumption of buildings—a review, Energy Build. 98 (2015) 119–124.
Appl. Climatol. 103 (1-2) (2011) 197–211. [66] M. Santamouris, Regulating the damaged thermostat of the cities—status, im-
[56] R. Emmanuel, H. Fernando, Urban heat islands in humid and arid climates: pacts and mitigation challenges, Energy Build. 91 (2015) 43–56.
role of urban form and thermal properties in Colombo, Sri Lanka and Phoenix, [67] J. Liu, M. Heidarinejad, S. Gracik, J. Srebric, The impact of exterior surface con-
USA, Climate Res. 34 (3) (2007) 241–251. vective heat transfer coefficients on the building energy consumption in ur-
[57] B. Jänicke, F. Meier, M.-T. Hoelscher, D. Scherer, Evaluating the effects of façade ban neighborhoods with different plan area densities, Energy Build. 86 (2015)
greening on human bioclimate in a complex urban environment, Adv. Meteo- 449–463.
rol. 2015 (2015). [68] Y. Sun, L. Gu, C.J. Wu, G. Augenbroe, Exploring HVAC system sizing under un-
[58] M. Roth, V.H. Lim, Evaluation of canopy-layer air and mean radiant temper- certainty, Energy Build. 81 (2014) 243–252.
ature simulations by a microclimate model over a tropical residential neigh-
bourhood, Build. Environ. 112 (2017) 177–189.

You might also like