Integrating Urban Heat Island Impact into Building Energy Assessment in a Hot-Arid City

Abstract

Dense cities usually experience the urban heat island (UHI) effect, resulting in higher ambient temperatures and increased cooling loads. However, the typical lack of combining climatic variables with building passive design parameters in significant evaluations hinders the consideration of the UHI effect during the building design stage. In that regard, a global sensitivity analysis was conducted to assess the significance of climatic variables and building design features in building energy simulations for an office building. Additionally, this study examines the UHI effect on building energy performance in Qatar, a hot-arid climate, using both measurement data and computational modeling. This study collects measurement data across Qatar and conducts computational fluid dynamics (CFD) simulations; the results from both methods serve as inputs in building energy simulation (BES). The results demonstrate that space cooling demand is more sensitive to ambient temperature than other climatic parameters, building thermal properties, etc. The UHI intensity is high during hot and transition seasons and reaches a maximum of 13 °C. BES results show a 10% increase in cooling energy demand for an office building due to the UHI effect on a hot day. The results of this study enable more informed decision-making during the building design process.

1. Introduction

Today, around 55% of the world’s population lives in cities, and this rate is projected to reach as high as 70% by 2050 [1]. The State of Qatar stands out with the world’s highest urbanization rate (99.1%), while its rural population represents only 0.9% [2]. Among different energy-consuming sectors, buildings account for over 40% of the annual energy consumption worldwide and around 55% of the world’s electricity consumption [3,4]. This rate is even higher in Qatar [5]. Thus, accurate estimation of building energy consumption is critical to provide city planners and policymakers with exquisite information on energy use to establish energy-efficient cities and mitigate greenhouse gas emissions and climate change. To this end, building energy simulation (BES) tools are widely used to estimate building energy consumption and investigate the influence of input variables on the energy performance of buildings. However, typical meteorological Yyear (TMY) data have been widely used in BES studies to represent the ambient climate of the building area without considering the local urban heat island (UHI) effect, omitting complex interactions between buildings and the environment.

The UHI refers to the characteristic warmness of an urban area, which is often approximated by comparing the temperature of a city with its surrounding rural areas. Increasing urbanization aggravates the urban climate environment characterized by the UHI effect. The UHI intensity, a crucial indicator of the increased heat in urbanized areas, is defined as the air temperature difference between the urban and rural area [6]. Previous publications have investigated the UHI for various climates; for instance, the daily profile of urban air temperature was studied in comparison to that of rural areas for an observation period of one year in Switzerland [7]. The average diurnal UHI intensity varied from 0 $°\mathrm{C}$ to 2 $°\mathrm{C}$ and peaked at 10:00 pm on a sample clear-sky day on 26 June 2002. The UHI intensity reaches higher temperatures in cities with hot and arid climates. For instance, in the case of Doha, long-term measurement data show that UHI intensity reaches as high as 5 $°\mathrm{C}$ [8]. A lack of knowledge of the local UHI effect will decrease the accuracy of building energy simulation results. Shi et al. [9] presented the huge influence of the local UHI effect on the sensible and latent cooling energy demand of residential buildings during the summer in Hong Kong. The results show sensible cooling demand considering the urban microclimate is approximately twice that of the rural weather, and the latent cooling demand could be up to 96% higher. Heat in cities increases cooling demand, with each 1 °C increase causing an increase of between 8.0 and 15% in building use, placing a considerable burden on decarbonization efforts in cities [10]. In a bibliometric review of urban heat mitigation and adaptation, He et al. [11] summarized the impact assessment and cause identification of UHI. It has been suggested that a holistic and comprehensive understanding of the scope of urban heat and its associated impacts is needed.

A promising solution to investigate the impact of local UHI on building energy use is to couple urban microclimate simulation tools with BES tools since there is no distinct tool that can directly assess the urban microclimate impact on building energy use [12]. Urban microclimate simulation tools predict local ambient conditions regarding different urban configurations. However, the features and the thermal processes of buildings are usually simplified or neglected in these simulation tools. BESs can provide detailed descriptions of the building and its systems using a dynamic model in the building energy performance analysis involving many input parameters. There is a gap in explicitly quantifying the impact of local UHI on building energy consumption compared with other inputs in BES, especially in hot and arid climate zones. The specific objectives of this study are listed as follows:

• To provide insights into the significance of ambient temperature in BES compared to other building passive design parameters by global sensitivity analysis.
• To assess the tempo-spatial UHI effect with the help of year-round observation of six weather stations across the country in hot-arid coastal areas.
• To employ a one-way coupling strategy between urban microclimate simulation and building energy simulation using CityFFD and EnergyPlus, respectively.
• To demonstrate the impact of the UHI effect on building energy performance based on co-simulation results.

2. Literature Review

This section provides an overview of recent research on building energy modeling approaches and the integration of building energy modeling with urban heat island simulations to assess the impacts of UHI on building energy performance. It highlights the use of various coupled models and methodologies to assess the influence of input parameters and ambient air temperature on energy demand.

2.1. Building Energy Modeling

Building energy simulations used a dynamic model to provide comprehensive descriptions of buildings and their systems [12]. Most BES models used today were developed for stand-alone buildings. Thousands of input parameters are used to describe detailed envelope material thermal properties, systems, and equipment usage schedules in BES models. Usually, meteorological data are unavailable for a specific urban area or available only for a few locations. In such a case, BES software models use TMY data to define the external meteorological boundary conditions. TMY data are smoothed and time-averaged based on long-term observations of local weather stations, which are usually located in rural or suburban areas, such as airports [13]. Due to that, the data fail to accurately represent the local and site-specific urban microclimate, resulting in an inaccurate evaluation of the actual building energy performance [14,15,16,17]. Several BES tools have been developed over the years, such as DOE-2 [18], Transient Systems Simulation (TRNSYS) [19,20], e-Quest [21], and EnergyPlus [22]. Among them, EnergyPlus is the most widely used tool [23,24,25]. Historical hourly weather data are acquired through the EnergyPlus Weather (EPW) file, available on the official website of EnergyPlus. The data derived from the measurements in 1980–1999 and based on DATSAV3 hourly weather data archived by the US National Climatic Data Center or the historical average are used as a baseline [26].

In BES modeling, it is necessary to evaluate the influence of input parameters on building energy performance. In this direction, several techniques have been used to perform sensitivity analysis (SA) and uncertainty analysis (UA). SA methods are widely used in building energy analysis, which can be divided into two groups: local SA and global SA [27]. Global SA methods include regression-based methods, screening-based methods, variance-based methods, as well as meta-modeling methods [28]. Standardized Regression Coefficient (SRC) and T-value are examples of the regression-based method. Though global SA focuses more on the influences of uncertain cases over the whole input space, local SA analyzes the effects of uncertain inputs around a base case. Studies indicate that the Monte Carlo Simulation (MCS) and Latin Hypercube Sampling (LHS) methods are very powerful, robust, and flexible [29]. Hou et al. [30,31] summarized the sensitivity analysis applied to BES modeling calibration. Rasouli et al. [32] investigated the thermal performance of a two-story office building in Chicago based on the local SA method. The ventilation rate was identified as the most important factor in the energy demand of HVAC systems. It was also found that all building system sizes and the overall initial investment cost in net-zero energy buildings are most sensitive to the indoor temperature set point [33].

2.2. Urban Heat Island and Its Impact Evaluation

A building’s ambient air temperature can be predicted by urban microclimate using different urban configurations. However, the features and the thermal processes of buildings are usually simplified or neglected in these simulation tools. Therefore, coupling the UHI simulation tools and BES models could be a promising solution to achieve the quantitative evaluation of the urban microclimate impact on building energy performance and indoor thermal conditions [12]. Different integration approaches are currently available to assess the effect of urban microclimate on building energy performance, covering building cooling/heating load, thermal comfort, etc. For instance, Urban Weather Generator (UWG) and building thermal model TRNSYS were coupled to carry out building performance simulation at the district scale [34]. This new method aimed to include the microclimatic modification induced by urban environments in buildings’ cooling demand calculations in hot and arid climates. Another study incorporated ENVI-met, a Computational Fluid Dynamics (CFD) simulation tool, and TRNSYS to estimate thermal outdoor comfort and mean radiant temperature, as well as to study the effects of urban form and vegetation on the microclimate of cities [35]. Another integrated platform was developed by coupling City Fast Fluid Dynamics (CityFFD) and City Building Energy Model (CityBEM) [36,37,38]. The platform produced high-resolution results of building thermal load and microclimate conditions during weather extremes in Montreal, Canada. The Weather Research and Forecasting model (WRF) was integrated with CityFFD to examine the effects of street aspect ratios and anthropogenic regimes on street temperatures [39]. A new flexible and tool-agnostic data schema was proposed to facilitate the exchange of data between urban building energy and urban microclimate models [40].

Mehoued and Lartigue [41] reported that the building cooling load, considering the urban microclimate, is 15% higher than the one calculated with the city weather station. Toparlar et al. [42] compared an office building constructed in an urban area with the same construction characteristics in a rural area. Buildings in the urban area were found to have 17.3–30.6% more monthly cooling demand in the summertime. A recent study by Hong et al. [43] shows potential deviations of up to 65% of building annual cooling energy use due to the urban microclimate in San Francisco. Zhang et al. [15] proposed a Long Short-Term Memory (LSTM) approach based on the long-term on-site measured weather data to study the variation of typical building energy performance with different weather data inputs. Results revealed a 5.1% decrease for buildings in suburban weather compared to urban microclimate. Shi et al. [9] studied the UHI and its impact on the sensible and latent cooling energy demand of residential buildings during summer in Hong Kong. The results show sensible cooling demand considering the urban microclimate is approximately twice that of the rural weather, and the latent cooling demand could be up to 96% higher. Brozovsky et al. [44] coupled CFD and BES to quantify the different compositions of urban surfaces on building energy demands. The results show that vegetation on the urban surface decreased the cooling energy demand by 28.5%. Liu et al. [45] coupled CFD and BES to quantify the impacts of the urban microclimate on the energy consumption of an academic building in the US. The annual cooling energy demand increased by 2.67% with the microclimate effect. Vallati et al. [46] studied the surrounding microclimate impact on urban building energy consumption using TRNSYS. A significant increase of up to 158% was computed for the annual cooling demand of the building in a street canyon configuration compared with the individual building. Errebai et al. [47] utilized meteorological simulations and weather station observations to study the UHI effects on cooling energy demand. The results indicate that cooling energy demand is significantly underestimated based on reference climate data provided in BES by 25% to 34%. Using downscaled urban weather data at urban morphology building levels to estimate cooling demand, Palme et al. [48] found that when UHI is incorporated, energy demand increases between 15% and 200%.

While there are numerous studies on the UHI effect, few have offered insights into the importance of considering the local air temperature in BES modeling in comparison to the other influential parameters, such as system, usage schedule, and building envelope properties. Fewer UHI studies for hot-arid coastal cities with the availability of year-round meteorological observation across the country, which is probably some of the most challenging climate areas for thermal comfort and sustainable city development. Thus, the present paper investigates the sensitivity of energy demand to various input parameters in BES modeling based on the global SA method, considering a case study of high-rise office buildings in the Marina district of Lusail City, Qatar. The evaluated input parameters simultaneously consider the building envelope, system, internal heat gain parameters, and external meteorological boundary conditions. The range of each parameter in SA is selected according to local standards based on Lusail City GSAS 2 Star Rating Guidelines [49] and the Handbook of Fundamentals (ASHRAE) [50]. In addition, with the help of the year-round observation of six weather stations across the state of Qatar, the tempo-spatial analysis of the UHI effect was conducted in a hot-arid coastal city. Finally, to study the UHI impacts on building energy performance using both experimental measurement data and CFD modeling, the UHI effect was introduced using a CFD and BES co-simulation for buildings in the Marina district of Lusail City, Qatar.

3. Description of Study Area

Over the last two decades, Qatar has experienced rapid population and urban growth [51,52]. To accommodate part of Qatar’s growing population, a new city development project, namely Lusail City, was developed as a flagship project to build a sustainable city on the northeast coast of Qatar. The project aimed to apply sustainability principles such as reducing greenhouse gas emissions, conserving water, and reversing desertification. Lusail City covers a 38 km² area and provides homes to more than 200,000 people. Marina district is a part of Lusail City. Its master plan comprises 36 high-rise commercial buildings, 37 mixed-use high-rise buildings, and 30 residential buildings to be constructed in 10 years. Until now, 20% of the buildings have been constructed, 27% are currently under construction, and 53% are still to be constructed. Pavement widths in the Marina district are determined by traffic volume, surface movement requirements, and under-grade infrastructure tunnel requirements. In order to enhance the sustainability of the district plan, greenery will be integrated into the rooftops, parking areas, and pedestrian areas of the district. To achieve the sustainability goals of the project, it is essential to evaluate the energy performance of buildings within the local urban context. BESs are commonly used to predict the space cooling and heating demands of buildings based on energy conservation for a control volume. Space cooling accounts for approximately 80% of the generated electricity in Qatar [53]. District cooling has the potential to save approximately 40% of the electricity due to the concentration effect of cooling load at one location [31,54]. District cooling load calculations play a vital role in the design and operation phases of a district cooling plant. However, the cooling load of individual buildings in a district is usually overestimated in the BES modeling [55], which results in designing oversized district cooling systems, high initial investment, low operational efficiency, and waste of energy and water. Thus, accurate prediction of building cooling loads is necessary, yet remains a challenge in the optimal design and operation of district cooling systems.

The local environmental setting in Qatar underwent considerable changes as a result of the dynamic evolution of numerous urbanization aspects, including a general change in land use, the number of buildings, the total amount of living space, and the number of cars. Typically, these shifts take the form of potentially significant micro-level variances in the urban climate across the nation. Because of these factors, Qatar is an intriguing case study to examine the sensitivity of climate conditions to a range of urban-scale factors.

The Marina district is in the southern part of Lusail City, comprising high-rise towers (number of floors ranging from 15 to 40) for office, residential, mixed-use, hotel, and retail use connected to a continuous boardwalk (Figure 1b). The district totals 3.1 million m² of built-up area, and the population of this district is 40,760, of which 27,000 are residents. It is bounded in the south by Lagoons Canal, to the west by Road B, to the east by the Arabian Gulf shoreline, and to the north by Qatar Entertainment City. The urban geometrical data were extracted from the information provided by the utility representatives and through the literature search. The Marina district is projected to be the future downtown of Lusail City in Qatar, which is under colossal construction. Our case study focuses on a typical high-rise office building in the Marina district of Lusail City, Qatar. The selection was made based on the fact that all office buildings in the Marina district adhere to the same construction requirements outlined in local and national standards. In addition, the availability of relevant information regarding the geometry and features of this building contributed to the decision to choose this building.

4. Methodology

This section outlines the methods employed in the present work. Firstly, high temporal granularity weather data were collected from a locally installed weather station and five rural airport weather stations across the country. Additionally, the microclimate was assessed using CityFFD, a microclimate tool based on CFD simulations. The coupling of building energy simulation and microclimate modeling enabled the analysis of UHI effects on building energy performance. In conducting a global sensitivity analysis, key climatic variables and building design features that have a substantial impact on building energy modeling were identified.

4.1. Weather Data Collection

High temporal resolution meteorological information is acquired from a local weather station set up in the Marina district of Lusail City at coordinates 25.399952 E and 51.519568 N. The sensors of the weather station report hourly data on temperature, solar radiation, wind direction, wind speed, relative humidity, and precipitation. Wind speed and temperature sensors were positioned 5 m and 4 m above the ground, respectively, and the measuring pole was 8 m away from the building.

In addition to the local weather station installed in the Marina district of Lusail City, a total of five airport weather stations were installed in different locations across Qatar: Doha International Airport, AI Ruwais, Dukhan, Umm Said, and Abu Samra. These have been designated as AWS-1 to AWS-5 in this study. The location of each station is shown in Figure 1a. The airport meteorological information was collected from the OGIMET website based on the “climate” package in R 3.6.3 [56], which is specialized in the automation of meteorological data downloading. Hourly meteorological data are available for these stations during summer, while every three-hour data are available during winter. UTC +3 time zone is used in simulations and the time difference between the airport weather station and local measurement is accounted for. The five airport meteorological stations were installed in different locations across Qatar. All aforementioned weather data were collected from August 2020 to August 2021.

4.2. Urban Microclimate Modeling

CityFFD is used as a microclimate simulation tool, which is based on the semi-Lagrangian approach: a fast and stable numerical model suitable for simulating urban-scale airflow conditions. It overcomes the huge computational load of traditional CFD tools in modeling urban scale microclimate and capturing neighborhood building impact on aerodynamics [36]. CityFFD is already validated in previous studies by the numerical and experimental data in the microclimate field [36,57]. In this case, the computational domain size was 4000 × 4900 × 1110 m³. A grid convergence study was performed, the select meshing setting shows a 1 m grid size close to the building. The total grid number is 65.7 million. A 24 h period was simulated with 1 h timesteps and 4000 iterations per timestep. The turbulence closure was achieved by the large eddy simulation (LES) [58].

4.3. Building Energy Modeling

EnergyPlus calculates the local outdoor wind speed and air temperature separately for each zone, as well as the temperature of surfaces exposed to the outdoor environment according to the US Standard Atmosphere [59] and Handbook of Fundamentals (ASHRAE 2005) [50]. However, the variation in barometric pressure is ignored in most situations [60].

There is a strong correlation between altitude and air temperature. Air temperature decreases almost linearly with altitude at a rate of $~$1 °C per 150 m across the troposphere. Wind speed increases with altitude whereas barometric pressure gradually decreases with altitude. Therefore, tall buildings could experience significant differences in local atmospheric conditions between the ground floor and the top floor [60,61].

Variation in outdoor air temperature is calculated using reference values for a given altitude independent of climate and seasonal differences [60,61], based on the US Standard Atmosphere (1976) [59]. The following formulas describe the relationship between air temperature and altitude in any given layer of the atmosphere.

$$T_z=T_b+L\left(H_z-H_b\right)$$

(1)

$$H_Z=\frac{Ez}{\left(E+Z\right)}$$

(2)

$$T_b=T_{z,met}-L\left(\frac{E_{Z_{met}}}{E+Z_{met}}-H_b\right)$$

(3)

where $T_z$ and $T_b$ are the air temperature at altitude $z$ and the base of the layer, i.e., ground level for the troposphere, respectively; $L$ is the air temperature gradient ($L=-0.0065 \mathrm{K}/\mathrm{m}$ in the troposphere); $H_b$ is offset equal to zero for the troposphere; $H_z$ is geopotential altitude; $E$ is the radius of the Earth, which equals 6356 km; $Z$ is altitude; $T_{z,met}$ is weather file air temperature (measured at the meteorological station); and $Z_{met}$ is the height above the ground of the air temperature sensor at the meteorological station. The default value for $Z_{met}$ for air temperature measurement is 1.5 m above the ground level.

The local wind speed calculations were performed as described in Chapter 16 of the Handbook of Fundamentals (ASHRAE 2005) [50]. The wind speed measured at a meteorological station is extrapolated to another altitude with the following equation:

$$V_z=V_{met}{\left(\frac{{\delta }_{met}}{z_{met}}\right)}^{{\alpha }_{met}}{\left(\frac{z}{\delta }\right)}^{\alpha }$$

(4)

where $z$ is altitude (the height above ground); $V_z$ is the wind speed at altitude $z$; $\alpha$ is the wind speed profile exponent at the site; $\delta$ is the wind speed profile boundary layer thickness at the site; $z_{met}$ is the height above ground of the wind speed sensor at the meteorological station; $V_{met}$ is wind speed measured at the meteorological station; ${\alpha }_{met}$ is the wind speed profile exponent at the meteorological station; and ${\delta }_{met}$ is wind speed profile boundary layer thickness at the meteorological station. The wind speed profile coefficients, $\alpha$, $\delta$, ${\alpha }_{met}$, and ${\delta }_{met}$, are variables that depend on the roughness characteristics of the surrounding terrain.

To investigate the impact of the UHI effect on building energy consumption, a detailed commercial BES model was developed using OpenStudio, SketchUp, and EnergyPlus, as shown in Figure 2. The target building was Sendian Tower, which is a typical high-rise commercial building in the Marina District of Lusail City, Qatar. It comprises 2 basement levels, a ground-floor lobby lounge, 27 additional floors, and a penthouse. For the difficulty of gathering some parameters, realistic assumptions were made during the modeling of the target building. Each floor is considered a thermal zone. There are in total 31 thermal zones. The heating, ventilation, and air conditioning (HVAC) system for each zone is a four-pipe fan coil system for space cooling. People density is 0.057 person/m². Lighting power density is 10.65 W/m², and electricity equipment power density is 7.64 W/m². The timestep is 10 min, which serves as the driving timestep for heat transfer and load calculations in the zone heat balance model.

Table 1. Target building characteristics.

Archetype/Feature	Description
Location	The Marina district of Lusail City, Qatar.
Building type	Commercial building
Year of construction	2017
Number of floors	Two basements, 1 ground floor, 27 typical floors, and a penthouse.
Total building area	41,619.09 m2
Number of thermal zones	31
HVAC system	The four-pipe fan coil system
Thermostat setting	24 $°\mathrm{C}$ from 6:00 to 22:00 and 26.7 $°\mathrm{C}$ for the rest of the day.
People density	0.057 person/m2
Lighting power density	10.65 W/m2
Electricity power density	7.64 W/m2

summarizes the characteristics of Sendian Tower.

In the present version of EnergyPlus, air temperature, wind direction, and wind speed are obtained from the TMY file, a meteorological file, and assumed identical for all the surfaces in the scene. For the link to the urban microclimate model, the EnergyPlus model is adapted, allowing for the allocation of individual outdoor air temperature and wind speed to each surface element of the building envelope.

4.4. Coupling Strategy

A one-way coupling strategy was used to perform a co-simulation of CFD and BES. A simplified simulation framework is illustrated in Figure 3. The CFD model calculates local microclimate based on airport meteorological data and urban geometry. The airport meteorological data includes hourly air temperature, wind speed, wind direction, and humidity, from August 2020 to August 2021. The site-specific air temperature around the target building was extracted from the CFD results and then integrated into the EPW file as EPW_local for EnergyPlus meteorological boundary condition, taking into account the urban microclimate impact. With the help of SketchUp and OpenStudio models, the building cooling load and surface temperature were calculated in EnergyPlus.

4.5. Sensitivity Analysis

A global sensitivity analysis was conducted to consider the influences of uncertain cases over the whole input space in this study. A typical SA study consists of six steps as follows: (1) determining input variations, (2) creating building energy models, (3) running energy models, (4) collecting simulation results, (5) running sensitivity analysis, and finally, (6) presenting sensitivity analysis results [27]. Besides the ambient air temperature, building energy consumption was influenced by many input parameters, including weather data, thermal properties of the envelope, and internal gains. The building envelope parameters and their variations are defined by ASHRAE or Qatar local standards (Lusail City GSAS 2 Star Rating Guidelines) [49]. In total, 14 input parameters were evaluated in this study; the range and data source are summarized in

Table 2. Input parameters and ranges of values.

Number	Parameters	Unit	Range of Values	Source
1	Air temperature	°C	8.9–46	Doha TMY weather data
2	Wind direction	°	10–360	Doha TMY weather data
3	Wind speed	m/s	0–25.7	Doha TMY weather data
4	Cooling set point	°C	21–26	[49]
5	Lighting power density	W/m2	0–9	[49]
6	Relative humidity	%	5–100	Doha TMY weather data
7	Wall insulation U-value	W/(m2K)	0–0.3	[49,63]
8	Infiltration	ACH	0.1–0.2	[63]
9	Ventilation	m3/s/person	0.00047–0.00247	[63]
10	Occupancy density	m2/person	19–24	[63]
11	Solar reflectance of interior diffusing blinds roll	/	0.4–0.8	[64]
12	Roof insulation U-value	W/(m2K)	0–0.25	[65]
13	Window solar heat gain coefficient	/	0–0.22	[63]
14	Window U-value	W/(m2K)	0–1.8	[63]

. The Latin Hypercube Sampling (LHS) method [29], using the R “lhs” package, was applied due to providing good convergence of parameters space with relatively few simples. A total of 605 different combinations of the parameters were utilized as inputs in this work. A parametric simulation was conducted using the EnergyPlus model to generate hourly and annual cooling loads using RStudio script. The R “eplusr” package [62] was used to perform the parametric simulation and automatically collect input and output datasets.

This study applied a sensitivity analysis (SA) method and sensitivity value index (SVI) method [27,66] to compensate for the difference in sensitivity analysis methods and target output. The SVI method is integrated with the standardized regression coefficient (SRC), random forest variable importance, and T-value method. The importance ranking among input parameters for the target high-rise office building was identified using the SA approach. The SVI calculation is performed based on the following formula [44]:

$$SVI \left(\%\right)={\displaystyle \sum }_{l=1}^m\frac{\frac{{{\displaystyle \sum }}_{j=1}^k\left(\frac{V_{i,j}}{{{\displaystyle \sum }}_{i=1}^n\left|V_{i,j}\right|}\right)}{k}}{m·k}\times 100$$

(5)

where $v$ is the value of a sensitivity analysis method, $i$ is a parameter, $n$ is the total number of the parameters, $j$ is a sensitivity method, $k$ is the total number of sensitivity methods ($k=3$), $l$ is the target output, and $m$ is the total number of target outputs ($m$= 1; building cooling demand).

5. Results and Discussion

We present our findings in the conclusion section. The global sensitivity analysis conducted on a typical office building in the study area revealed significant variables that influence building energy modeling. Further, we demonstrated variations in ambient temperature and UHI intensity around the target building. Through analysis of UHI effects on building cooling energy loads, specifically on a typical hot day, we gained insight into the importance of considering UHI effects in building energy modeling. These findings collectively contribute to a better understanding of the factors influencing building energy performance. They emphasize the need to account for UHI effects in sustainable building design.

5.1. Global Sensitivity Analysis

The global SA of a typical high-rise office building was conducted as described in Section 4.5. The SA results are listed in

Table 3. Results of sensitivity analysis.

Parameters	SRC	Random Forest	T-Value	SVI	Rank
Air temperature	0.53	204.71	68.95	24.09	1
Wind direction	0.20	135.48	30.04	12.00	2
Wind speed	0.16	58.74	23.36	7.53	3
Cooling set point	0.13	60.93	20.97	6.85	4
Lighting power density	0.10	42.88	16.39	5.15	5
Relative humidity	0.01	58.41	0.66	2.56	6
Wall insulation thickness	0.02	7.08	3.67	1.05	7
Infiltration	0.02	6.50	3.74	1.04	8
Ventilation	0.01	3.97	1.33	0.44	9
Occupant density	0.01	6.06	0.90	0.43	10
Solar reflectance of interior diffusing blinds rolls	0.01	4.217	0.81	0.34	11
Roof insulation thickness	0.01	3.52	0.85	0.32	12
Window solar heat gain coefficient	0.003	2.694	0.52	0.22	13
Window insulation	0.0001	2.899	0.02	0.12	14

. The impact of 14 input parameters was ranked from 1 to 14 according to their SVI values. The most critical input parameter is indicated as 1 and the least as 14. The results show that three model weather input parameters—air temperature, wind speed, and wind direction—were the most critical parameters with the highest SVI values, followed by the cooling set point and lighting power density. However, solar irradiation plays the most crucial role in the energy model simulation. It was observed that most of the surrounding buildings are far away from the studied high-rise office building, thus the shading caused by other buildings has a slight impact on the target building. Therefore, we did not consider the solar-related parameters in the co-simulation. The conduction transfer function (CTF) thermal model, used in this study, was the best choice for the analysis of building energy simulation, especially in a hot climate. Notwithstanding, it does not account for the combined transport of heat and moisture within building envelopes [67]. Thus, the CTF model probably underestimates the relative humidity (RH) impact on building energy simulation.

5.2. Urban Heat Island Effect

Weather in Qatar is tropical maritime with two main seasons: a cold season from December to February and a hot season from May to October, with March, April, and November being the transition months [68]. In the study area, the Marina district consists of 36 high-rise commercial buildings, 37 mixed-use high-rise buildings, and 30 residential buildings. Figure 1 illustrates the layout of the buildings in the study area. The UHI intensity was estimated by calculating the difference between the air temperature of the local and rural stations. Based on the observed data from the local weather station and each airport weather station across Qatar, the maximum station difference between the air temperature of the local weather station and each airport weather during each season was calculated and plotted in

Table 4. Maximum differences between the air temperature of the local weather station and each airport weather station during each season.

	AWS_1	AWS_2	AWS_3	AWS_4	AWS_5
Cold season	5.67	7.87	8.17	7.93	9.65
Hot season	5.20	11.58	10.30	11.99	9.50
Transition	6.39	11.41	11.81	7.71	13.07
All seasons	6.39	11.58	11.81	11.99	13.07

during the period between August 2020 and August 2021. It can be seen in Table 4 that the maximum air temperature difference in each season varies from 5.2 $°\mathrm{C}$ to 13.07 $°\mathrm{C}$. The significant UHI effect occurs in hot and transition seasons, and a noticeable difference in wind speed was observed in the transition season. Different spatial distribution characteristics of rural air temperature were observed. The air temperature data collected from AWS_5, situated southwest of the local weather station, were higher during cold and transition seasons. The minimum air temperature difference was observed to be between the local weather station and AWS_1 due to a short distance. The results indicate that the selection of rural station data influences the UHI intensity.

In order to evaluate the UHI intensity in Marina District, 22 evaluating points are selected to present the temperature variations. These locations include the high-density residential building area, high-density commercial building area, as well as near-wall of the target building. The locations of evaluating points are illustrated in Figure 1c.

The hourly temperature variations of the evaluating points against the temperature data of AWS1 on a hot day are presented in Figure 4. As mentioned earlier, the UTC +3 time zone was used in the simulations. The peak air temperature occurred at different locations and times, such as 47.9 $°\mathrm{C}$ at 10:00, 44.3 $°\mathrm{C}$ at 13:00, and 46.4 $°\mathrm{C}$ at 13:00, in the residential area, commercial area, and target building, respectively. Here, the peak temperature is recorded as the highest value in the temperature data obtained from all evaluation points in each observation area. In addition, the average peak temperature of the evaluation points is calculated as 46.0 $°\mathrm{C}$, 43.8 $°\mathrm{C}$, and 45.6 $°\mathrm{C}$ in the residential area, commercial area, and target building, respectively. On the other hand, the peak temperature of AWS1 was measured as 41.9 $°\mathrm{C}$ at 9:00. All peak temperatures of simulated evaluating points are greater than the AWS1. These results indicate the spatiotemporal variation of the thermal intensity across the Marina district due to a number of factors such as building density, building envelope material and its thermal properties, and location of building stocks.

The spatiotemporal variation of the UHI intensity within the Marina district on a hot day is analyzed by plotting UHI intensity graphs based on the variation of the temperature difference between the simulated evaluation point and measured data from AWS1 by time (Figure 5). The UHI effect is observed only in the daytime between 6:00 and 18:00. The UHI intensity reaches a maximum of 6.2 $°\mathrm{C}$ at 10:00, 5.4 $°\mathrm{C}$ at 14:00, and 8.2 $°\mathrm{C}$ at 14:00 in the residential area, commercial area, and target building, respectively. The average maximum UHI peaks are 4.6 $°\mathrm{C}$, 4.5 $°\mathrm{C}$, and 6.5 $°\mathrm{C}$ in the residential area, commercial area, and target building, respectively. Variation of UHI intensity among the simulation areas is due to the type of building stocks (building archetype) and location. The residential area, for example, is closer to the inlet airflow due to its location where the wind provides more cooling effect, resulting in lesser UHI intensity in this area. The peak temperature and UHI intensity of each area are summarized in

Table 5. Summary of the peak temperature and UHI intensity data on selected areas and around the target building in the Marina district.

	Temperature (°C)			UHI Intensity (°C)
	Time	max	avr max	Time	max	avr max
Residential	10:00	47.9	46.0	10:00	6.2	4.6
Commercial	13:00	44.3	43.8	14:00	5.4	4.5
Target	13:00	46.4	45.6	14:00	8.2	6.5

Time: UTC +3, max: maximum; avr max: average maximum; Residential: residential area; Commercial: commercial Area, Target: target building.

.

Mean absolute error (MAE) represents the average absolute difference between simulation results and corresponding measurement values over the dataset. It provides a linear score by assigning equal weights to all the individual differences in the average. On the other hand, root mean squared error (RMSE) involves squaring each difference between simulation and measurement data, averaging them, and taking the square root of the average. As a result, the RMSE emphasizes large errors by assigning relatively greater weight to them. The MAE and the RMSE were calculated together to diagnose the variation in the UHI intensity in the prediction values of the simulation on a hot day (Figure 6). The higher difference was obtained in the daytime between 8:00 and 15:00, with both errors peaking at 14:00, which indicates the UHI intensity reaches the maximum at 14:00 in the Marina district.

5.3. UHI Impact on Cooling Load

To evaluate the microclimate impact on building energy load, the date 9 August 2021, was selected as the evaluating day to represent a typical hot day. The temperature profile of the Marina district at 14:00 on a hot day was evaluated through CFD simulation (Figure 7). The wind direction from the north can be observed on temperature gradients adjacent to the buildings in the evaluation area. Building surface temperature varies among different building stocks, depending on many factors, including building characteristics, envelope material, and shading equipment use.

Urban microclimate impact on building cooling load on a hot day was investigated by inputting CFD results into EnergyPlus. The hourly cooling energy load intensity—the cooling energy used per unit floor area—was simulated in EnergyPlus. The average values of all evaluating points were considered as the local microclimate air temperature, which was integrated into the EPW file as EPW_local for following BES modeling taking into account the urban microclimate impact. The results are presented against the other simulation results with the data of each airport weather station in Figure 8. The minimum and maximum cooling loads were observed at 3:00 and 14:00, respectively. The daily cooling load obtained with the input of CityFFD was higher than that with data from each was weather station. With the input of CityFFD results, the daily Energy Use Intensity (EUI) was computed as 0.53 kWh/m², whereas the average daily EUI based on the five weather station data found as 0.48 kWh/m² (10% lower). The difference in cooling load was higher in the daytime than at nighttime due to the UHI effect in the daytime. With the input of CityFFD results, the daytime cooling load between 8:00 and 18:00 was computed as 11.9 MWh, whereas the average daytime cooling load based on the five weather station data was found as 10.4 MWh. The difference between results, based on CityFFD and each weather station data, indicates the microclimate impact on building energy consumption. The available literature related to the urban microclimate impact on building cooling energy consumption shows a wide variation among different studies, due to several factors such as different study areas, building characteristics, and the method of considering the urban microclimate impact.

6. Conclusions and Outlook

This paper investigates the spatiotemporal characterization of an urban heat island (UHI) and its impact on building cooling load for a high-rise office building in the Marina district of Lusail City, Qatar. Global sensitivity analysis was conducted contributing to a better understanding of the urban ambient temperature impact on the building energy performance compared to other BES parameters. With the help of high temporal-resolution observed data collected from six weather stations across the country, the UHI effect in Qatar was analyzed representing the hot-arid coastal climate. CFD analysis was conducted to evaluate the UHI effect in the case study area. The CFD results obtained from CityFFD were inputted into EnergyPlus to simulate the building energy consumption. As such, one-way coupling enabled the evaluation of the impact of the UHI effect on building energy performance. The following conclusions can be drawn from this study.

• According to the impact ranking of input parameters in global sensitivity analysis, the most critical input parameter is the air temperature, followed by wind direction and wind speed.
• The air temperature difference between the local weather station data and airport weather station data indicated the UHI effect of the urban area.
• The spatiotemporal variation of UHI intensity observed in residential and commercial areas in the Marina district stems from a number of factors, such as building density, thermal properties of building envelope material, shading equipment use, and the location of building stocks.
• The difference between MAE and RMSE results is minimal in the nighttime and maximum in the daytime, indicating the high UHI intensity during the daytime.
• The building cooling load obtained with the input of CityFFD results was higher than with weather station data. The difference clearly indicates the significance of considering the UHI impact in building energy simulation.

This study shows the significance of considering urban microclimate impact in BES studies. Coupling CFD and BES enables defining the meteorological boundary conditions accurately and obtaining realistic energy predictions of buildings within an urban context. Despite its clear advantages, research on implementing coupling strategies between urban microclimate and building energy modeling is very limited. Thus, future studies are recommended to focus on the implementation of different coupling strategies using available simulation tools and coupling platforms to consider the urban microclimate impact in building energy modeling. In addition, a cross-analysis between the UHI effect and building thermal performance may provide more comprehensive insights. It is recommended that future studies be conducted to examine the effects of building thermal performance on the UHI phenomenon, and a detailed analysis of regional cooling disparities influenced by UHI effects is highly valuable.