Influence of built environment on building energy consumption: a case study in Nanjing, China

Abstract

Built environment plays a significant role in optimizing building energy consumption. However, few studies have explored the comprehensive effect between built environment metrics on building energy consumption. Thus, this study aims to explore interrelationships between built environment on building energy consumption focused on moderating effect. In this study, we established a built environment measure system from the perspective of land use and land cover, landscape structure and building configuration. This study explored the correlation between built environment and building energy consumption and analyzed the moderating effect of building configuration emphatically. Results show that: for integrated grids group, normalized difference vegetation index (NDVI) and modified normalized difference water index (MNDWI) have a positive influence and impervious area (IA) has a negative influence, with NDVI has the greatest impact. Building floor (BF), building coverage ratio (BCR) and aspect ratio can weaken the positive relationship between NDVI and energy use intensity of grid (${\text{EUI}}_{\text{grid}}$). BCR weakens the positive effect of MNDWI on ${\text{EUI}}_{\text{grid}}$. The moderating effect of building configuration on EUI varies in the same grid group and among different grid groups. For sample 1, BCR inhibits the negative effect of mean perimeter–area ration (PARA-MN) on ${\text{EUI}}_{\text{grid}}$. For sample 2, BF promotes the negative effect of number of patches and land use richness index (R) on ${\text{EUI}}_{\text{grid}}$. And sky view factor inhibits the positive effect of IA on ${\text{EUI}}_{\text{grid}}$. This study reveals the pathways of built environment on building energy consumption. As a result, the keys of optimizing building energy consumption are the reasonable planning and optimization of the urban built environment of different land cover.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s10668-023-02930-w.

Introduction

The rapid urbanization has brought economic vitality to the city, while the energy consumption has also increased dramatically (Abbasi et al., 2021a; Zhuang et al., 2022). Buildings make a great contribution to the total energy consumption of cities, accounting for about 32%, which is the key to urban energy conservation and pollution emissions reduction (Lucon et al., 2014). The main methods of lowering energy consumption of buildings consist of using energy-saving reconstructions of building envelope and changing the built environment (Kim et al., 2019; Wang et al., 2022a). Among these, the elements of built environment can sustainably affect the energy use by modifying the microclimate around buildings, such as land use and land cover (LULC) (Qiao & Liu, 2020; Soydan, 2020), landscape structure (Chen et al., 2011; Haas & Ban, 2014), building configuration (Deng & Wong, 2020; Leng et al., 2020; Wong et al., 2011) and so on. LULC and landscape structure can affect the surface temperature around buildings and alleviate the urban heat island effect (UHI) (Chen et al., 2011; Qiao & Liu, 2020). Building configuration can produce ventilation corridors and shadows, which will affect ventilation efficiency and solar irradiation and further affect the building energy consumption (Deng & Wong, 2020; Wong et al., 2011). Besides, the outbreak of the COVID-19 pandemic has brought global health challenges in recent years, especially mental disorders (Farzadfar et al., 2022; Ge et al., 2022; Liu et al., 2022). Sustainable environment may have a positive social impact, such as by benefiting human health, improving residents’ thermal comfort and quality of life (Aman et al., 2019; Li et al., 2022; Wang et al., 2022a, 2022b; Yuan et al., 2021). Therefore, how to low building energy consumption, improve the urban thermal environment and achieve the goal of sustainable development by optimizing the environment of cities has received attention from policymakers and scholars increasingly (Aman et al., 2022; Wang et al., 2023).

The built environment is defined as human-made physical environment surroundings and conditions (Lin et al., 2021; Travert et al., 2019). Built environment plays an important role in improving energy efficiency and reducing building energy consumption (Anderson et al., 2015). It is evident that land use can improve urban thermal comfort and lower building energy consumption by reducing the temperature around buildings (Bowler et al., 2010; Edan et al., 2021; Qiao & Liu, 2020). Generally, urban impervious surface is conducive to strengthen the urban heat island intensity, while the green surface can provide cooling effect (Edan et al., 2021; Soydan, 2020). Landscape structure is also an important factor of built environment to reduce building energy use. Studies use landscape indicators to quantify urban land use patterns and to explore the effect between fragmentation and complexity of landscape on building energy consumption (Chen et al., 2011).

Furthermore, building characteristics are another way to reduce building energy consumption from the point of view of physical attributes. Building layout (orientation, sky view factor (SVF), etc.), construction intensity (building density, aspect ratio (AR), etc.) and building physical form (orientation, building height, etc.) are keys factors for studies. These factors can change cooling and heating energy use by affecting the shelter between buildings, the availability of sunlight and ventilation. Specifically, narrow urban canyons (high AR) can increase the shadows between buildings, which can decrease the cooling demand of buildings (Chen et al., 2018). A high SVF of buildings can improve the solar irradiation rate and influence the heating energy use (Ahn & Sohn, 2019; Bowler et al., 2010; Morganti et al., 2017).

As shown above, the influence of built environment on building energy consumption has been widely explored. However, previous studies have concentrated on the impact of variables on building energy consumption and the extent to which they affect energy use (Chen et al., 2018; Zhou et al., 2017). For example, most of existing researches take representative buildings or block prototypes as research objects, regulate the building parameters and simulate the building energy consumption to explore the influence of parameters on building energy consumption (Pan et al., 2017). Considering the interaction between influencing factors, the change of a single factor may lead to the change of other factors (Lee & Cheng, 2016). Focusing on the independent effects of built environment is likely to underestimate their comprehensive effect (You & Kim, 2018). To fill this gap, a few studies considered various factors that affect building energy consumption. Li et al. (2021a) considered five dimensions of influencing factors: meteorology, architecture, resident, equipment and energy use behavior, and clarified the interaction among them. D’Oca et al., (2017) integrated architectural physics and social psychology to explore the factors driving interactions between building occupants and control systems. Li et al. (2021b) explored the impact of environment factors on building energy consumption regarding the connection between the natural and social environment. You and Kim (2018) revealed direct and indirect effects of urban form and land use on houses’ energy efficiency. These studies explored the complex relationship of influencing factors from direct and indirect influence paths. However, the important role of the moderating effect in exploring the complex relationship and the influence degree of explanatory variables was ignored (Wang et al., 2021). And the moderating factors are concentrated in social and economic aspects, such as public credit (Wang et al., 2021). The moderating effect of built environment has not been fully taken into consideration.

Therefore, from the perspective of built environment, this study attempts to explore the influence of related factors on building energy consumption with an emphasis on the moderating effect. First, we obtained building information from geographic information system and use EnergyPlus to calculate the building energy consumption of each grid. Then based on the satellite remote sensing image, we used building outline data and Geographic Information System (GIS) platform to calculate LULC, landscape structure and building configuration metrics. Finally, we explored the effect of built environment on building energy consumption based on backward regression. And the moderating effect of building configuration was revealed by hierarchical regression. This study aims to identify the moderation role of built environment via buildings configuration, complementing existing research.

Study area

Xuanwu District is located in the middle of Nanjing, Jiangsu Province (see Fig. 1). In Nanjing, Xuanwu is the central area, with 7 streets under jurisdiction. By 2018, the area has a total area of 75 square kilometers. It is characterized by high urbanization rate, rich vegetation and developed transportation system. The green coverage rate reaches 60%. In 2013, the disposable income of urban households in the district of Xuanwu ranked first in Nanjing, and reaching 65,864 yuan in 2018, with an urbanization rate of 100%. Furthermore, the development of urbanization is accompanied by a sharp increase in energy consumption, and the contribution of buildings to the total urban energy consumption has reached 32% (Lucon et al., 2014; Pérez-Lombard et al., 2008). There are approximately 16,000 buildings in the study area. OF these, residential buildings are the most numerous, accounting for about 40% of the building stock. The area is divided into 300 m × 300 m grids, and the grid is used as the basic unit of building energy consumption analysis. The year 2018 is selected as the base year of calculation in this study.

Fig. 1.

Map of the study area and distribution of sampling grids

Data and methods

Data

Data used in this study include Landsat remote sensing images, building information, roads and land use. Landsat images of MOD13QI and Landsat8 OLI in 2018 originated from the National Aeronautics and Space Administration (NASA, https://search.earthdata.nasa.gov/) and the United States Geological Survey (USGS, https://earthexplorer.usgs.gov/), with 250 m and 30 m spatial resolution, respectively. Data in MOD13QI were collected from May 2018 to August 2018. These images are mainly used for normalized difference vegetation index (NDVI) and modified normalized difference water index (MNDWI) inversion. Building and road information in 2018 were from BIGEMAP, GIS and Baidu map, where building data included information on building footprint floors and types of building use. Land use image in 2018 of 30 m spatial resolution was sourced from Geospatial Information Monitoring Cloud Platform (http://www.dsac.cn/), which is primarily used for extracting landscape data.

Built environment metrics measurement

Built environment refers to the whole system of building and surrounding environment (Chen et al., 2011). Based on the literature review, it is evident that built environment can influence building energy consumption and carbon emissions by changing the spatial pattern of surrounding environment and building physics characteristics. In this process, carbon emission has decreased, which can ensure to reduce environmental pressure and promote sustainable development (Abbasi, Adedoyin, et al., 2021b). Specifically, surrounding environment can change the external thermal environment of buildings by affecting the land surface temperature and urban heat island intensity (Woo & Cho, 2018). The building configuration determines the ventilation efficiency and the ability to receive solar radiation of the canyon (Chen et al., 2018). To this end, this study uses land use and land cover (LULC, quantified by NDVI, MNDWI, impervious area (IA), richness (R)) and landscape structure (quantified by number of patches (NP), mean perimeter–area ratio (PARA-MN), mean shape index (SHAPE-MN), mean patch area (AREA-MN)) to express surrounding environment and building floor (BF), building coverage ratio (BCR), SVF, aspect ratio (AR) to quantify building configuration (Table 1). The built environment metrics are quantitatively calculated according to ArcGIS 10.5 and FRAGSTATS 4.0 software.

Table 1.

Definition of built environment indices

Type	Indicator	Definition
Land use and land cover (LULC)	Normalized difference vegetation index (NDVI)	The amount of vegetation and plant growth status. Negative values indicate no vegetation; positive values indicate vegetation coverage and an increase with coverage
	Modified normalized difference water index (MNDWI)	The amount and distribution of water bodies
	Impervious area (IA)	The sum of the areas of roads and buildings in each sample
	land use richness index (R)	The number of land types and the change in the proportion of land types
Landscape structure	number of patches (NP)	The total number of patches in each grid. NP represent the degree of patch fragmentation
	mean perimeter–area ration (PARA-MN)	Mean of the perimeter (m)–area (m2) ratio of patches in each grid
	mean shape index (SHAPE-MN)	Mean of the shape index of patches in each grid. PARA-MN and SHAPE-MN represent the complexity of patch shape
	Mean patch area (AREA-MN)	Mean of area of patches in each grid. AREA-MN represent the degree of patch fragmentation
Building configuration	Building floor (BF)	The average floor of buildings per grid
	Building coverage ratio (BCR)	The ratio of built ground to that of the sample district area
	Sky view factor (SVF)	The mean value of the ratio of the solid angle of visible sky from each point of the façades to the sky vault
	Aspect ratio (AR)	Ratio of building height to the width of the distance between buildings

Energy consumption calculation

Building energy consumption refers to the energy input from outside during the use of the building, including the energy used to maintain the built environment and various activities in the building. In this study, building energy consumption mainly depends on electrical appliances usage, being concentrated on heating, cooling, lighting and equipment energy consumption (Abbasi et al., 2020). Therefore, energy type is mainly electricity in this study. To express the level of building energy consumption, we use the concept of energy use intensity (EUI), which refers to the annual energy consumption per unit area of a building, usually expressed in kWh m^− 2y^− 1. EUI of grid (EUI_grid) was calculated on the basis of building spatial information and energy consumption per unit area of typical buildings (EUI_k). This study calculated the EUI_grid of the study area in 2018. Figure 2 shows a summary of the inversion building energy consumption methodology flow. To get the building energy consumption of each grid, there are two kinds of sources of input data: (1) This study uses remote sensing image to identify the types of building use. (2) With the help of BIGEMAP and geographic information system (GIS), this study obtains building footprint area and number of floors.

Fig. 2.

Energy consumption measurement methodology flow

Identification of building information

Building spatial information comprises building types of use and building total floor area. GIS technology is used to collect information about the building stock. There are approximately 16,000 buildings in Xuanwu District. Visual interpretation method is used to identify the building use types. Table 2 shows 12 building types of use. Additionally, office building consists of large office building, small and medium office building. The total floor area of small and medium-sized office buildings is less than 20,000 m². Subsequently, building total floor area is calculated on the basis of building footprint area and number of floors.

Table 2.

Building use types

Building use types	Specify types
Office building	Small and medium office building, large office building
Public building	Hospital, School, tourism building, gym
Commercial building	Restaurant, hotel, mall, shop
Residential building	Villa, dwelling

Simulation of building energy consumption

In this study, typical buildings are taken as simulation objects, and dynamic energy consumption simulation software EnergyPlus is used to calculate energy consumption per unit area for various building models. First, this study defines the architectural attribute of 12 types of buildings according to the design standards of each type of building. Standards include lighting energy-saving standards, air-conditioning energy-saving standards, average household electrical appliances ownership and typical building cases in the study area. Next, we use SketchUp to set up a building geometric model including building shape coefficient, building plane shape, building plane size, building space function type and building floors. Then, through OpenStudio we set the thermal performance parameters of building envelope and building energy system, including lighting system, HVAC system, general energy-consuming equipment and other energy systems. The relevant design paraments and heat parameters of typical buildings are shown in the supplementary data (Table S1). OpenStudio, a visual user interface of EnergyPlus, is a building energy simulation software (https://www.openstudio.net/). The summer cooling period is set from May 15, 2018, to September 30, 2018, and the winter heating period is set from January 1, 2018, to February 28, 2018, and November 15, 2018, to December 30, 2018. Outdoor weather file is obtained from EnergyPlus (https://www.energyplus.net/weather) of typical meteorological year (TMY) format, including dry bulb temperature, dew-point temperature, wind direction, wind speed, solar horizontal radiation intensity, total solar radiation intensity, etc. Finally, we use the energy consumption simulation engine to obtain EUI of representative buildings. Figure 3 shows a summary of energy consumption simulation methodology flow.

Fig. 3.

Energy consumption simulation methodology flow

Calculation of building energy consumption

On the basis of EUI data and total floor area (FA) in each building type to calculate EU and EUI for each grid, as expressed by Eq. (1) and Eq. (2).

$${\text{EU}}_{\text{grid}}=\sum _{k=1}^n{\text{EUI}}_k\times {\text{FA}}_k$$ 1

$${\text{EUI}}_{\text{grid}}={\text{EU}}_{\text{grid}}/{\text{TFA}}_{\text{grid}}$$ 2

where ${\text{EU}}_{\text{grid}}$ is annual building energy consumption per grid; ${\text{EUI}}_{\text{k}}$ is annual energy consumption per unit area of building type k. ${\text{FA}}_k$ is floor area of building type k.${\text{EUI}}_{\text{grid}}$ is annual energy consumption per unit area of a grid;${\text{TFA}}_{\text{grid}}$ is total floor area per grid.

Regression analysis

To explore the correlation between the built environment and the building energy consumption, this study adopts backward regression and hierarchical regression. Backward regression is used to determine influence of LULC and landscape structure on building energy consumption. Hierarchical regression further discusses the moderating effect of building configuration. Firstly, backward regression is conducted between dependent variables (LULC and landscape structure) and independent variable (building energy consumption). Next, this study selects hierarchical regression to examine the moderating effect of building configuration based on the result of backward regression.

Considering variables of LULC and landscape structure are up to 8, the problem of multicollinearity may exist. Therefore, backward regression is used to eliminate the variables that don’t meet the criteria (Sheather, 2009). Variance inflation factor (VIF) can determine whether multicollinearity is a problem. In this study, if the VIF of independent variables are less than 10, we can conclude that multicollinearity is not a problem.

Hierarchical regression can test the moderating effect of variables in the relationship between independent and dependent variables (Sawada & Toyosato, 2021). This study adopts the following steps to test the moderating effect of building configuration. The mean-centered approach is used to dispose of dependent variables and independent variable, which can prevent a high degree of multicollinearity (Yu et al., 2015). Then, to generate the interaction terms, hierarchical regression analysis is conducted.

Additionally, on the basis of first two steps, this study explored the heterogeneity of impact of variables on building energy consumption in different locations. SPSS Statistics 26.0 is used to conduct backward regression and hierarchical regression.

Results

Built environment

Figure 4 shows the spatial distribution of LULC. The areas with high NDVI are concentrated on the central and eastern part of Xuanwu District, where the Bauhinia Mountain is located. High MNDWI is in Xuanwu Lake. High IA is mainly located in the densely built areas. R reflects the abundance of LULC. At the boundary of land use, the index is high.

Fig. 4.

Spatial distribution of LULC of Xuanwu district

Figure 5 shows the spatial distribution of landscape structure. NP and AREA-MN indicate the fragmentation of landscape. PARA-MN and SHAPE-MN indicate shape complexity of landscape. Corresponding to the land use in Fig. 4, it can be seen that NP and PARA-MN of the boundary between vegetation land, water area and impervious area are larger. That means the land here is more fragmented and complex in shape. It also shows that the same land type in the Xuanwu has good connectivity and integrity.

Fig. 5.

Spatial distribution of landscape structure of Xuanwu district

Figure 6 shows the spatial distribution of building configuration. Building configuration is measured with BF, BCR, SVF, AR. The overall building height distribution is more average. Areas with high building coverage are concentrated in impermeable areas. The SVF value is the highest in forest land and water area, and slightly lower in densely built area. The areas with high AR values are mainly distributed in the southwest of Xuanwu District.

Fig. 6.

Spatial distribution of building configuration of Xuanwu district

Building energy consumption

The spatial distribution of buildings is shown in Fig. 7. The types of about 4,500 buildings in the study area are unknown. Among the identified building types, residential buildings are the most, accounting for about 40% of the buildings, which are evenly distributed in all streets in study area. Commercial buildings are mainly distributed in Xinjiekou Street, the core area of CBD (central business district) in Nanjing. Office buildings are evenly distributed, with Xuanwuhu Street in the majority. Public buildings are mainly located in Xinjiekou Street, Xiaolingwei Street and Suojincun Street. Table 3 shows the descriptive statistics of building data.

Fig. 7.

Spatial distribution of building types

Table 3.

Descriptive statistics of building data

Types of building	Number of buildings	Total floor area (km2)
Office building	1188	6.37
Public building	2842	7.39
Commercial building	767	2.25
Residential building	6652	23.12

The simulation results of energy consumption per unit area of typical buildings are shown in Fig. 8. As expected, the EUI of office building, public building, commercial building and residential building conform to previous studies (Huang et al., 2013; Wei et al., 2013). There are obvious differences of EUI in different types of buildings (Fig. 8). The EUI of mall and tourism building is the highest. Because compared with other types of buildings, the EUI of mall and tourism building is high due to many energy-consuming equipment, high pedestrian flow density, long service time and high working strength. The EUI of mall and tourism building are closely followed by shops, hospitals and large office buildings. The EUI of school is the lowest because it uses electricity for less time. Due to holidays, schools avoid the summer and winter during the peak period of electricity consumption. The EUI of various types of buildings in the figure will be applied as the basic data of the building energy consumption calculation.

Fig. 8.

Simulation results of building energy consumption

The distribution result of ${\text{EUI}}_{\text{grid}}$ is shown in Fig. 9. Major parts with a high ${\text{EUI}}_{\text{grid}}$ are primarily located in the north-west and middle area of the study area, which are mainly composed of tourism buildings. The lower ${\text{EUI}}_{\text{grid}}$ are quite evenly covered from west to northeast and southeast, which mainly consist of dense residential buildings and schools.

Fig. 9.

Spatial distribution of EUI_grid (kWh/m²/y)

Backward regression

In this study, Model 1 contains only the integrated grid group. Models 2 and 3 represent the grid group under woodland and water area and under cultivated and impermeable area, respectively.

This study uses the backward regression to explore the effect of LULC and landscape structure on ${\text{EUI}}_{\text{grid}}$. The final equation of regression is as follows:

$${\text{EUI}}_{\text{grid}}=47.471+99.115\ast \text{NDVI}+67.150\ast \text{MNDWI}-0.001\ast \text{IA}$$ 3

Based on the model 1’s fitting information, this study shows the multiple correlation coefficient R² is 0.315, which means that the equation can explain 31.5% of ${\text{EUI}}_{\text{grid}}$ (Table 4). NDVI and MNDWI have a positive influence on the ${\text{EUI}}_{\text{grid}},$ while IA has a negative influence. NDVI has the greatest impact to the ${\text{EUI}}_{\text{grid}}$ (0.360) among the three variables, followed by MNDWI (0.283) and IA (− 0.184) having smaller impact on ${\text{EUI}}_{\text{grid}}$.

Table 4.

Results of multiple linear regression analysis

Variable	Model 1		Model 2		Model 3
	Coef	S. E	Coef	S. E	Coef	S. E
AREA-MN			0.445**	0.042
NP			0.488**	15.223
NDVI	0.360***	14.044
PARA-MN			− 0.233**	0.038	0.146***	0.011
MNDWI	0.283***	9.234	0.195***	10.939
IA	− 0.184***	0.000	− 0.210***	0.000	0.124**	0.000

*p < 0.1, **p < 0.05, ***p < 0.01

NDVI and NDWI are believed to reduce surface temperature and building energy consumption, while impervious area will increase urban heat island effect and thus increase building energy consumption (Qiao & Liu, 2020; Soydan, 2020). This is contrary to the results of the findings in this study. This study speculates that this may be due to the matching of buildings and land cover. For example, the middle part of Xuanwu consists of water and woodland and has high vegetation coverage. But there are many tourist buildings and restaurants with high energy consumption intensity. By contrast, the rest of Xuanwu is mainly covered by impervious area and vegetation coverage is relatively small. The buildings are dominated by apartments and schools (Fig. 10), tending to be low energy consumption.

Fig. 10.

Proportion of built-up area under woodland and water area

In order to further analyze the influence of LULC and landscape structure on ${\text{EUI}}_{\text{grid}}$. In this study, the grids of tourism buildings and malls with high EUI under woodland and water area are analyzed separately (model 2). The remaining grids focus on impervious areas and water areas around the Xuanwu (model 3). Analyzing the two grid groups separately contributes to avoid the statistical error caused by the spatial distribution of buildings and land use.

Based on the model 2’s fitting information, this study shows the multiple correlation coefficient R² is 0.143, which means that the equation can explain 14.3% of ${\text{EUI}}_{\text{grid}}$. IA and PARA-MN have a negative influence on the ${\text{EUI}}_{\text{grid}}$. However, MNDWI, NP and AREA-MN have a positive influence. NP has greater impact on ${\text{EUI}}_{\text{grid}}$ (0.488), followed by AREA-MN (0.445), PARA-MN (0.233), IA (0.210) and MNDWI (0.195).

This study shows the multiple correlation coefficient R² is 0.024 of model 3. It means that the equation can explain 2.4% of ${\text{EUI}}_{\text{grid}}$. IA and PARA-MN have a positive influence on the ${\text{EUI}}_{\text{grid}}$. PARA-MN has greater impact on the ${\text{EUI}}_{\text{grid}}$ (0.146) than IA (0.124).

Hierarchical regression

Integrated grid group

The premise of hierarchical regression is that the independent variable has a correlation effect on the dependent variable. Therefore, the correlation analysis between LULC and landscape structure and ${\text{EUI}}_{\text{grid}}$ was conducted in this study. The results of the correlation analysis are shown in the supplementary data (Table S2). The results show that NDVI, MNDWI, IA, R, NP, PARA-MN, SHAPE-MN, AREA-MN are correlated with building energy consumption. Then, the regulatory effect of building configuration between LULC and landscape structure and ${\text{EUI}}_{\text{grid}}$ is analyzed. The coefficient of the interaction item was used as an indicator to judge the moderating effect.

For the integrated grid group, there is a significant positive correlation with NDVI, R, MNDWI, NP, PARA-MN and SHAPE-MN on ${\text{EUI}}_{\text{grid}}$, whereas there is a significant negative correlation with IA and AREA-MN. Looking at the interaction item coefficients in Table 5, we see that BF has negative moderating effects on the relationships between NDVI, R, IA, NP, AREA-MN, NP, SHAPE-MN and ${\text{EUI}}_{\text{grid}}$. BF inhibits the positive effects of NDVI, R, NP and SHAPE-MN on ${\text{EUI}}_{\text{grid}}$ and inhibits the negative effects of IA and AREA-MN.

Table 5.

Regression analysis results for BF

Independent variable	Coef	S. E	Independent variable	Coef	S. E
NDVI	0.406***	10.447	R	0.125***	5.801
BF	− 0.173***	0.640	BF	− 0.229***	0.73
NDVI*BF	− 0.109***	3.735	R*BF	− 0.105**	2.8
IA	− 0.457***	0.000	AREA-MN	− 0.111***	0.009
BF	− 0.187***	0.652	BF	− 0.231***	0.731
IA*BF	0.201***	0.000	AREA*BF	0.098**	0.004
NP	0.103**	3.198	SHAPE-MN	0.116***	20.24
BF	− 0.232***	0.731	BF	− 0.228***	0.732
NP*BF	− 0.091**	1.357	SHAPE*BF	− 0.099**	12.191

*p < 0.1, **p < 0.05, ***p < 0.01

We see that BCR and AR have negative moderating effects on ${\text{EUI}}_{\text{grid}}$ in Table 7. BCR inhibits the positive effect of NDVI and MNDWI on ${\text{EUI}}_{\text{grid}}$ and inhibits the negative effect of IA. Besides, AR inhibits the positive effect of NDVI on ${\text{EUI}}_{\text{grid}}$ and inhibits the negative effect of IA (Table 6).

Table 7.

Regression analysis results for SVF

Independent variable	Coef	S. E	Independent variable	Coef	S. E
NDVI	0.31***	10.305	R	0.096**	5.289
SVF	0.366***	21.616	SVF	0.462***	21.545
NDVI*BF	0.144***	119.47	R*SVF	0.077**	68.657
SHAPE-MN	0.09**	18.415	MNDWI	0.167***	8.890
SVF	0.462***	21.556	SVF	0.436***	21.672
SHAPE*SVF	0.076**	262.223	MNDWI*SVF	0.12**	133.264
IA	− 0.338***	0.000	AREA-MN	-0.09**	0.008
SVF	0.326***	22.43	SVF	0.464***	21.516
IA*SVF	− 0.296***	0.001	AREA *SVF	-0.074**	0.101
NP	0.088**	2.904
SVF	0.466***	21.492
NP*SVF	0.077**	37.073

*p < 0.1, **p < 0.05, ***p < 0.01

Table 6.

Regression analysis results for BCR and AR

Independent variable	Coef	S. E	Independent variable	Coef	S. E
NDVI	0.22***	12.619	MNDWI	0.165***	8.805
BCR	− 0.352***	14.877	BCR	− 0.452***	12.065
NDVI*BCR	− 0.2***	83.4	MNDWI*BCR	− 0.26***	141.126
IA	− 0.186*	0.000	NDVI	0.422***	11.377
BCR	− 0.312***	35.271	AR	− 0.035	7.89
IA*BCR	0.372***	0.001	NDVI*AR	− 0.205**	72.629
IA	− 0.503***	0.000
AR	0.047	8.132
IA*AR	0.242***	0.001

*p < 0.1, **p < 0.05, ***p < 0.01

Table 7 shows that SVF has positive moderating effects on ${\text{EUI}}_{\text{grid}}$. SVF promotes the positive effect of NDVI, R, MNDWI, NP and SHAPE-MN on ${\text{EUI}}_{\text{grid}}$ and promotes the negative effect of IA and AREA-MN.

Grid group under woodland and water area

For grid group under woodland and water area, NDVI, MNDWI and AREA-MN have significant positive correlations with ${\text{EUI}}_{\text{grid}}$, whereas R, IA, NP and PARA-MN have significant negative correlations. The results are provided in the supplementary data as Table S3. According to the results in Table 8, we see that BF has positive moderating effects, while BCR has a negative moderating effect. BF promotes the negative effect of NP and R on $\text{EUI}$ and promotes the positive effect of AREA-MN. And BCR inhibits the negative effect of PARA-MN on ${\text{EUI}}_{\text{grid}}$.

Table 8.

Regression analysis results for BF and BCR

Independent variable	Coef	S. E	Independent variable	Coef	S. E
NP	− 0.161**	4.836	AREA-MN	0.204***	0.013
BF	− 0.245***	1.658	BF	− 0.242***	1.662
NP*BF	− 0.193***	2.493	AREA*BF	0.18***	0.007
R	− 0.15**	8.134	PARA-MN	− 0.14**	0.022
BF	− 0.243***	1.661	BCR	− 0.339***	50.826
R*BF	− 0.188***	4.497	PARA*BCR	0.166**	0.329

*p < 0.1, ***p < 0.01, **p < 0.05

Grid group under cultivated and impermeable area

According to the results in the supplementary data (Table S4), IA, NP and PARA-MN have significant positive correlations with ${\text{EUI}}_{\text{grid}}$, whereas AREA-MN has a significant negative correlation for grid group under cultivated and impermeable area. SVF has a positive moderating effect on the relationship between PARA-MN and ${\text{EUI}}_{\text{grid}}$, while has a negative moderating effect on the relationship between IA and ${\text{EUI}}_{\text{grid}}$ in Table 9. Besides, BF has a positive moderating effect on the relationship between IA on ${\text{EUI}}_{\text{grid}}$. Specifically, SVF inhibited the positive effect of IA and promoted the positive effect of PARA-MN. And BF promotes the positive effect of IA on ${\text{EUI}}_{\text{grid}}$.

Table 9.

Regression analysis results for SVF and BF

Independent variable	Coef	S. E	Independent variable	Coef	S. E
IA	0.100*	0.000	PARA-MN	0.122**	0.011
SVF	0.101*	17.621	SVF	0.097*	17.542
IA*SVF	− 0.123**	0.001	PARA *SVF	0.147***	0.189
IA	0.089*	0.000
BF	− 0.037	0.476
IA*BF	0.116**	0.000

*p < 0.1, **p < 0.05, ***p < 0.01

Discussion

In this study, we established a built environment metrics system from three types: LULC, landscape structure, and building configuration. Then, we use simulation approaches to calculate building energy consumption on the basis of EnergyPlus. Finally, this study employs stepwise regression and hierarchical regression to analyze the impact of built environment and the moderating effect of the building configuration. The results reveal the effective pathways of built environment to reduce building energy consumption.

Impact of LULC and landscape structure

The changes of land use play an important role on the land surface temperature (LST) and consequently influence building energy consumption. Studies show that vegetation and water can alleviate the UHI by inhibiting the rise of LST, thereby reducing building energy consumption (Shivaram et al., 2021). Impervious area alterations can change heat capacity, albedo and thermal conductivity, resulting in urban heat island effect and increasing building energy use (Adulkongkaew et al., 2020).

Surprisingly, results demonstrate that NDVI and MNDWI have a positive effect for integrated grid group, and IA has a negative effect for integrated grid group and group under woodland and water area, while IA has a positive influence for grid group under cultivated and impermeable area. The reason for this phenomenon is matching of ${\text{EUI}}_{\text{grid}}$ and land use pattern. The area mainly covered by water and vegetation has a high proportion of buildings with high energy consumption intensity (Sample 1), while the area mainly covered by impervious area has a high proportion of buildings with low energy consumption intensity (Sample 2). Therefore, the impact of land use on building energy consumption in the overall sample is different from the theoretical results. Then, this study divided two types of samples to avoid the influence of uneven distribution of land use and building type.

For sample 2, IA and PARA-MN have a positive influence. The effect of IA is the same as in previous studies (Ezimand et al., 2021; Fu et al., 2022; V et al., 2012). PARA-MN depicts the overall shape complexity of patches, and a high value refers to greater complexity (Zheng et al., 2020). The complex landscape areas in the study area are mainly located at the junction of different land types, with various building types including tourism buildings of high energy consumption. Therefore, the building energy consumption increases with the complexity of landscape shape.

For sample 1, IA has a negative effect on ${\text{EUI}}_{\text{grid}}$, while MNDWI has a positive effect. Considering the influence of vegetation area and impervious area on energy consumption is also affected by other characteristics. The location of trees is the key factor affecting in building energy consumption (Skelhorn et al., 2016). Larger area of building envelope and building volumes in commercial buildings lowers the potential impacts of vegetation. Strategic placement (on the south or west side, for instance) of vegetation is needed in order to achieve noticeable benefits (Bowler et al., 2010).

Impact of building configuration

Deep street canyons are created by a cluster of high-rise buildings, which can reduce the availability of natural ventilation and solar gains (Ahn & Sohn, 2019). Additionally, a taller building height means that more shadows around the buildings and ventilation corridors will be produced (Adulkongkaew et al., 2020). For integrated grids group, BF can weaken the positive relationship between NDVI and ${\text{EUI}}_{\text{grid}}$. Because shadow and ventilation corridors can reduce the heat of built-up surfaces and enhance energy flow between vegetation and their surrounding areas. However, for sample 2, BF promotes the positive effect of IA on ${\text{EUI}}_{\text{grid}}$. High-rise buildings can hinder cold air diffuse to the ambient urban environment. The temperature around buildings will be increased, which promotes the positive effect of IA on ${\text{EUI}}_{\text{grid}}$.

For sample 2, BF promotes the negative effect of NP and R on ${\text{EUI}}_{\text{grid}},$ while it promotes the positive effect of AREA-MN. High-rise buildings with higher surface roughness are conducive to heat exchanges at vertical and horizontal dimensions (Yuan et al., 2021). Moreover, more shadows are cast with high-rise buildings to directly reduce LST in the ambient area. Promotion of the negative effect of NP and R is because of shielding between buildings reducing the demand for cooling in summer. Promotion of the positive effect of AREA-MN may be due to the heat island effect intensified by the microclimate around buildings.

Generally, the BCR is used to describe density (Chen et al., 2011). Compact buildings are not conducive to temperature diffusion, resulting in lower heating energy consumption in winter (Yang et al., 2018). Specially, a more clustered urban development results in lower heat losses, thus lowering heat loads of buildings with cold climates. For integrated grid group, BCR weakens the positive effect of NDVI and MNDWI on ${\text{EUI}}_{\text{grid}}$. High-density buildings with low ventilation rate can reduce the evaporation of water (Yuan et al., 2021). Thus, the vegetation and the water surface are promoted to absorb more heat intensity. However, in hot climates, BCR is generally associated with an increase in cooling building use. High BCR of buildings is not conducive to air circulation, and heat dissipation is not easy. The temperature around buildings is increased, which facilitates the formation of urban heat island effect. For sample 1, BCR inhibits the negative effect of PARA-MN on ${\text{EUI}}_{\text{grid}}$ possibly because cooling energy use in summer is proportional to the increase in BCR (Woo & Cho, 2018).

The decrease in the SVF reduces solar radiation reaching the buildings and influences the thermal environment of the street canyons (Liu et al., 2021; Strømann–Andersen & Sattrup, 2011). For integrated grid group, SVF strengthens the positive relationship between NDVI, MNDWI and ${\text{EUI}}_{\text{grid}}$. A high SVF of buildings with greater solar radiation can enhance the evapotranspiration of vegetation and water, thus reducing the capacity of absorb heat. For sample 2, SVF inhibits the positive effect of IA on ${\text{EUI}}_{\text{grid}}$. Higher SVF has better ventilation conditions and reduces cooling demand of buildings (Deng et al., 2021).

In narrow urban canyons (large AR), the shadow between buildings can be used to reduce building cooling building use, and reducing building energy consumption. For integrated grid group, AR weakens the positive relationship between NDVI and ${\text{EUI}}_{\text{grid}}$. A large AR with more shadows between buildings can alleviate the transpiration of vegetation and water, which can promote the absorption of heat. Therefore, the positive effect of NDVI on ${\text{EUI}}_{\text{grid}}$ is inhibited.

Based on the findings of this study, building configuration plays an important role in mitigating the effect of LULC and landscape structure on building energy consumption. Therefore, we propose suggestions for reducing building energy consumption. For integrated grids group, BF, BCR and AR can weaken the positive relationship between NDVI and ${\text{EUI}}_{\text{grid}}$. And BCR weakens the positive effect of MNDWI on ${\text{EUI}}_{\text{grid}}$.

Moreover, the building configuration as a mitigation measure to moderate ${\text{EUI}}_{\text{grid}}$ should consider heterogeneity between area with different land cover. In this study, we also provided mitigating ways for different land use with grid groups. For sample 1 (grid group under woodland and water area), BCR inhibits the negative effect of PARA-MN on ${\text{EUI}}_{\text{grid}}$. For sample 2 (Grid group under cultivated and impermeable area), BF promotes the negative effect of NP and R on ${\text{EUI}}_{\text{grid}}$. And SVF inhibits the positive effect of IA on ${\text{EUI}}_{\text{grid}}$. Consequently, the reasonable planning the building configuration of different land covers can better moderate the effect of LULC and landscape structure on ${\text{EUI}}_{\text{grid}}$.

Limitations and future work

In this study, a moderating effect model was applied to explored the relationship between built environment and $\text{EUIgrid},$ whereas there are some limitations of this study. Firstly, due to the inadequacy of built environment metrics systems, the effect of built environment on building energy consumption still needs to substantiate the existing findings. Secondly, in the calculation of building energy consumption, we ignored the interaction between the buildings. Besides, considering the limitation of the study area, we only considered two kinds of grid groups with different land use.

Consequently, further research should be carried out to optimize the built environ metrics systems and expand the spatial ranges. Furthermore, if more accurate building energy consumption can be calculated, the accuracy of the regression model can be improved. This provides a new possibility to explore the relationship between built environment and building energy consumption more accurately. This study supplies further research directions to meet the challenges of the COVID-19 epidemic crisis. In the world, the ongoing infectious disease has generated health-related demands and energy consumption (Geng et al., 2022; Paulson et al., 2021; Yu et al., 2022). The lockdown policies worldwide had a significant impact on global energy consumption during the epidemic, accompanied by a rise in residential energy consumption (Khalil & Fatmi, 2022). The more interesting research direction need to consider the impact on building energy demand during the COVID-19 epidemic.

Conclusions

In the urban planning field, built environment is understood as crucial to reduce building energy consumption. Yet investigations into the comprehensive effect of built environment on building energy consumption are limited, especially the moderating effect. This study analyzes the effects of built environment metrics on ${\text{EUI}}_{\text{grid}}$ using backward regression and hierarchical regression model focused on moderating effect. Moreover, on the basis of the matching relationship between ${\text{EUI}}_{\text{grid}}$ and land use, the samples are divided into two categories for specific analysis.

We find that building configuration plays a significant role in mitigating the effect of LULC and landscape structure on ${\text{EUI}}_{\text{grid}}$. For integrated grids group, BF, BCR and AR can weaken the positive relationship between NDVI and ${\text{EUI}}_{\text{grid}}$. And BCR weakens the positive effect of MNDWI on ${\text{EUI}}_{\text{grid}}$. The moderation effect of building configuration on EUI varies in the same grid group and among different grid groups. For sample 1, BCR inhibits the negative effect of PARA-MN on ${\text{EUI}}_{\text{grid}}$. For sample 2, BF promotes the negative effect of NP and R on ${\text{EUI}}_{\text{grid}}$. And SVF inhibits the positive effect of IA on ${\text{EUI}}_{\text{grid}}$. In Sample 1, BCR plays a vital role in mitigating the effect of landscape structure on ${\text{EUI}}_{\text{grid}}$. In Sample 2, BF and SVF are more severely moderating ${\text{EUI}}_{\text{grid}}$.

Due to change the ventilation conditions of street canyons, solar radiation conditions and shielding between buildings, building configurations have different influences on the surface temperature in different seasons and thus have different influences on building energy consumption. Future research can focus the influence paths of building configuration and various effect in different seasons.

Supplementary Information

Below is the link to the electronic supplementary material.

Data availability

The datasets analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of interest

The authors declared that they have no conflicts of interest to this work. We declare that the manuscript “Influence of built environment on building energy consumption: A case study in Nanjing, China” we do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted.