Next Article in Journal
Effect of Photoperiod and Glycerol Supplementation on the Biomass Productivity and Protein Production of Spirulina sp. LEB 18 Cultures
Previous Article in Journal
Comparison of Media Company Performance Efficiency Based on the Search Engine’s Method of Providing News Content (External Links vs. Internal Links)
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:

Estimating Housing Vacancy Rate Using Nightlight and POI: A Case Study of Main Urban Area of Xi’an City, China

College of Geography and Environmental Science, Northwest Normal University, Lanzhou 730070, China
Author to whom correspondence should be addressed.
Appl. Sci. 2022, 12(23), 12328;
Submission received: 2 September 2022 / Revised: 29 November 2022 / Accepted: 30 November 2022 / Published: 2 December 2022
(This article belongs to the Section Environmental Sciences)


Estimating the housing vacancy rate (HVR) has always been a hard-to-break point in the study of housing vacancy. This paper used nighttime light and POI (point of interest) data to estimate the HVR in the main urban area of Xi’an city based on extracting built-up areas. The built-up area was extracted using the threshold method, and the spatial resolution of the results was 130 m (same as Luojia-1). Meanwhile, after removing the non-residential areas from the images, the HVRs for the period 2018–2019 from four nighttime light images were calculated, and the HVR of the main urban area of Xi’an city was estimated using the average method and its spatial patterns were analyzed. The results show that: (1) Luojia-1 has great advantages in estimating urban HVRs. The HVRs calculated by Luojia-1 were characterized by a high resolution and a short calculation time. (2) After estimating the results of the four scenes’ remote sensing images, it was found that the results obtained using the average were closest to the actual vacancy situation, and the spatial distribution of the vacancy could be seen using the minimum values. (3) The overall housing occupancy in Xi’an city was good, and the HVRs were low, but the overall vacancy rate for the edge of the built-up area was high. The government should devote more attention to places with high HVRs.

1. Introduction

Vacant housing refers to a house that has been neither rented nor sold for a period of time. Although reasonable housing vacancies will stabilize the housing market [1], an excessive housing vacancy rate (HVR) will lead to limited urban development and various social problems [2,3,4]. With the rapid development of urbanization in China, from 2002 to 2017, China’s urban residential construction area increased from 17.3 to 24 billion square meters, and during the “13th Five-Year Plan” period, the cumulative built-up construction area reached 1.6 billion square meters, with an average annual growth rate of approximately 54% [5]. The government began to plan new residential areas to solve the problems of high population density and large demand for land in the old city. However, a significant portion of these newly developed units are vacant [6]. The prosperity of the real estate industry has become an important indicator of the degree of development of a city [7]. Some cities vigorously develop the real estate industry and create a development illusion to adapt to the economic development trend in the general environment. The consequence is that a large number of houses are vacant and unable to exert their original value. This development has been accompanied by a series of social and economic problems [8]. At the same time, a large number of new lands have been developed to cope with the growing demand for land resources and population density, and real estate development has become the main way of urban spatial development in China [9].
In China, long-term rapid economic growth and a boom in the real estate sector have masked the phenomenon of housing vacancies which has led to lagging research. Owing to domestic research on housing vacancies being still in its infancy, the relevant data related to housing vacancies are missing, the data are not disclosed, and the real estate industry is not transparent, resulting in the study of housing vacancies being very underdeveloped [10,11]. On the one hand, there are different opinions in society regarding the implementation of the housing vacancy tax, as it lacks strong supporting evidence. On the other hand, there is a lack of research data and methods. The study of housing vacancies is conducive to understanding the spatial patterns and dynamic changes in the current real estate market, calculating the HVR from the perspective of geography [12], and providing a reference for the government to formulate regulatory policies and future development plans for the city.
Over the past decade, there have been much controversy and advocacy in academia regarding housing vacancies. In terms of the definition of vacant housing, Denise and William [13] argue that a vacant house is a home that satisfies the need for unused housing and is planned to be rented or sold. The American Federal Bureau of Statistics defines a vacancy as a house that is uninhabited and not temporarily uninhabited at the time of the statistical survey, and a house that is inhabited or whose inhabitants are temporarily out of the country at the time of the statistical survey is defined as a house in use. The “2017 China Urban Housing Vacancy Analysis Report”, released by the Southwestern University of Finance and Economics, divided vacant houses into two types according to the number of houses owned by housing families: a single suite owned by a family was vacant because of leaving to go to work; multisite family homes that were neither inhabited nor rented [14]. In this study, we defined a housing vacancy as a house that has been unoccupied for a period of time, whether rented or sold, and thus vacant. This definition implies that a housing vacancy is a persistent state rather than a state at the point in time of an investigation [15], emphasizing the continuity of the state.
From the perspective of the development of the real estate, the reasons that affect housing vacancy mainly include market supply and demand factors that lead to a mismatch between housing supply and housing demand [16], functional housing factors that do not allow for normal living due to the presence of defective housing functions [17], and speculative factors that lead to the purchase of multiple houses to maximize profits [18]. However, it is difficult to assess housing vacancies from the same perspective due to the different architectural structures and cultures in China. Therefore, whether a city’s current housing market is saturated and whether it still needs to vigorously develop real estate.
At present, there is very little research on housing vacancy, especially on the grid scale [19,20]. From the research data and methods on housing vacancies, Sahil Gandhi [21] used a linear regression method to analyze the HVR and its influencing factors based on census data. Park and Ciorici [22] selected the logit models, community gardens (CGs), and existing vacant lots, to simulate the different uses of vacant land and to identify the factors that would help convert vacant land into CGs. Gill Armstrong [23] proposed a new quantitative method called the vacancy visual analytics method (VVAM) for identifying vacancies in urban populated areas. After defining and analyzing the vacancy phenomenon, Ge [24] also conducted a correlation analysis with various factors, such as infrastructure and traffic conditions, which represented the HVR of the city. Branas [25] selected a city-wide, randomized controlled trial in the United States to demonstrate that structural disrepair and abandonment are key causes of the negative outcomes regarding people’s safety through fieldwork. Chen [12] calculated the HVR based on NPP-VIIRS using the ratio of urban light intensity (ULI) and the light intensity of non-vacancy (LINV). These studies of urban vacancies expands from a single housing statistic to combined research methods with multidisciplinary and multidata sources, which can facilitate the study of urban vacancies with a more reasonable and clear understanding. In this study, four scene images of Luojia-1 were used to calculate HVR. The average and minimum values of the calculation results were compared. Then we obtained the HVR closest to the real situation. This method is an extension of the calculation method used by our predecessors and leads to more reliable calculation results.
The paucity of available real estate information in China means that many essential data are difficult to obtain. No official data have been released on vacant housing monitoring and evaluation in China whereas academic research on this topic mainly relies on survey data [26], urban nighttime light data [27], and statistics data [28]. The research focuses on the characteristics of urban vacancies at a low resolution. However, there are also certain shortcomings with the existing studies such as a short research cycle. The previous research ignored the dynamic changes in the HVR at different times and different locations [29]. Although some studies have evaluated housing vacancies based on synthesizing monthly and annual data [12,30,31], their methods ignored the impact of roads and nonresidential areas using a single data. The final results for calculating HVR are quite different from the real situation and cannot represent the real housing vacancy situation.
In this paper, the main urban areas of Xi’an city were used to map the HVR at a fine scale. The built-up areas were extracted by constructing the LJ&POI index. The HVR was calculated after removing the impact of roads and nonresidential areas. Finally, through multiple periods of nighttime light data, the spatial distribution results that could represent the vacancies in Xi’an city for a period of time were obtained. In this way, decision-makers can understand the spatial distribution of housing vacancies in Xi’an city and put forward targeted policy recommendations. At the same time, this calculation method improves the accuracy of HVR calculations, enriches the method for calculating HVRs, and provides a quantitative basis for policy formulation in the real estate industry.
The rest of this paper is structured as follows. In Section 2, the study area and data are introduced. Section 3 expounds on the proposed methodology. In Section 4, the results are presented. In Section 5, the discussion is introduced. Finally, the paper is concluded in Section 6.

2. Study Area and Data

Xi’an city is the capital of Shaanxi Province, located in the Guanzhong Plain of the Yellow River Basin. It is an important central city in western China, between 107°4′–109°49′ E and 33°42′–34°45′ N. The main urban area of Xi’an city includes the Weiyang District, Lianhu District, Beilin District, Yanta District, Xincheng District, and Baqiao District (Figure 1), with a total area of 831.87 km2. By the end of 2021, the total population of the main urban area will be approximately 4.9248 million people. From 1984 to 2016, the construction land in the study area increased by 478.62 km2, an overall increase of 3.27 times [32], and the sales area of commercial residential buildings in the main urban area will reach 9.5796 million m2 in 2021 [33].
The Luojia-1 satellite was developed by Wuhan University and successfully launched in June 2018 [34]. It is the world’s first “one-satellite multipurpose” low-orbit, micro–nanoscience experimentation satellite, with remote sensing and navigational functions. The Luojia-1 is equipped with a highly sensitive luminous camera that can obtain images of nighttime lights at a high resolution. It has a spatial resolution of 130 m, a width of 250 km, a spectral bandwidth of 0.319 μm, a dynamic range of 14 bits at night, and a revisit period of 3–5 days. Compared with traditional DMSP/OLS and NPP/VIIRS nighttime light data (Table 1), Luojia-1 data have obvious advantages in terms of data bits and spatial resolution. Therefore, nighttime light data can more clearly reflect the spatial structure and spatial characteristics of a city [35].
DMSP/OLS and NPP-VIIRS are the main data sources in current housing vacancy studies. The revisit time of DMSP/OLS and NPP-VIIRS is short, the data are easy to obtain, and it is usually used in studying large regional scales. However, the resolution of DMSP/OLS and NPP-VIIRS cannot meet the needs of small regional scales. In this study, the data of Luojia-1 were used as the main data source. Four time points (19 August 2018, 23 August 2018, 31 October 2018, and 11 March 2019) data that could completely cover the study area were selected to estimate the HVR of Xi’an city.
POI data is obtained from the Amap Open Platform (, accessed on 29 October 2021). Amap Open Platform is an open-source and free data interface platform, which only needs to register an account to obtain POI for free. POI has a wide range of application scenarios and strong calculation and expression capabilities. There are 21 first-level classifications, 267 s-level classifications, and 869 third-level classifications of Amap POI. This paper uses a total of 635,364 records at three levels of classification in the main urban area of Xi’an city. The POI can reflect a full range of human activity traces and is more representative of secondary data for housing vacancy studies.
The administrative data were obtained from the National Basic Geographic Information Center (, accessed on 21 October 2021).

3. Method

The main framework for calculating the HVR is shown in Figure 2. The detailed steps are as follows:
Pre-processing of the Luojia-1 and POI data. Geometric correction, radiation correction, and mask processing were used to preprocess the Luojia-1 data, and kernel density analysis was performed to obtain the same resolution as the Luojia-1.
Extraction of built-up areas. The LJ&POI index was constructed by using the treated Luojia-1 and POI data, and then the threshold was determined according to the method of Li [36] to extract the built-up area of the main urban area of Xi’an city.
Estimation HVR. It was necessary to remove the influence of roads and nonresidential areas on the basis of built-up areas to obtain the actual DN value of residential areas because the main research object was residential areas. The DN value in the case of a full residential area was determined. The result was calculated from the ratio of the DN value between a residential area and a full residential area.

3.1. Data Processing

Since the absolute radiation correction data of the Luojia-1 nighttime light data are floating-point data, which are not convenient for storage, the ground system enlarged the floating-point radiance data to the 10th power of 10, and the exponential stretch was converted to INT32 storage [37]. Before building the LJ&POI index, it was necessary to perform radiant brightness conversion operations according to the given formula. The formula for calculating the radiant brightness of the Luojia-1 nighttime light data product is, according to the Data and Application Network of the Hubei Province High-Resolution Earth Observation System, as follows:
L i = D N i 3 2 · 10 10
where Li is the radiant value after absolute radiation correction at point i (W/(m2∙sr∙μm)), and DNi is the image grayscale value at point i. The results before and after data correction are shown in Figure 3.
The POI data can accurately reflect information such as business activities and infrastructure in the region. Before constructing the LJ&POI index, the POI data need to be cleaned and analyzed by nuclear density analysis, and the core density distribution map of the POI data in the study area can be obtained.
Nonparametric kernel density estimation (KDE) is a method that does not require prior assumptions regarding the distribution. The data points that are close to the center point in the calculation area have higher weights and vice versa based on the distance from the center point. Therefore, the results obtained were the weighted average density of all points in the study area. The kernel density (Pi) at any point (i) was calculated as follows:
P i = 1 n π R 2 × j = 1 n k j ( 1 D i j 2 R 2 ) 2
where n is the number of data points (j) in the calculation rule area; R is the bandwidth of the calculation rule area; Kj is the weight of data point j; Dij is the Euclidean distance between spatial point i and data point j. In this paper, it was necessary to reasonably select the bandwidth (R) according to the scientific problem, and since this paper needed to use the POI kernel density to extract the built-up areas with local detailed information [38], we selected a 130 m bandwidth (same as Luojia-1). The results are shown in Figure 4.

3.2. Extraction of the Built-Up Areas

In recent years, nighttime light images have been widely used to extract urban built-up areas. However, due to low data resolution and light spillover effects, the extracted built-up areas have low accuracy. Li et al. [36] put forward the NTL&POI composite index, which is based on the nighttime light images of Luojia 1-01 of China and NPP/VIIRS of the United States, and used the threshold method to achieve the extraction of the built-up areas before and after NTL&POI composite index processing. The comparative analysis with the reference built-up areas shows that the NTL&POI composite index reflects the overall shape of the built-up areas, but also retains the details of their boundaries. The landscape pattern index shows that the extraction result of the built-up area has better connectivity and complexity. According to existing research [36], the inherent stability and effectiveness of POI can eliminate images of transient lighting noise and can further improve the accuracy of built-up area extraction. Therefore, POI is used to eliminate the background noise of nighttime lighting images and to weaken the effects of lighting spillover. As there is a positive correlation between the luminance values of the nighttime lighting data and the kernel density values of the POI data and the built-up areas, the mean method was chosen in this article to combine the positive correlation between the two types of data and the built-up areas.
When extracting the built-up areas, the LJ&POI index was established, and the threshold method was introduced to divide the built-up areas and non-built-up areas in the main urban area of Xi’an city, which is used to eliminate the background noise of the night light data and weaken the impact of light overflow [39]. The detailed expression is presented as follows:
L J & P O I i = P i × L i
where LJ&POIi is the LJ&POI index; Pi is the kernel density value at point i; Li is the nighttime light brightness value at point i. The specific results are shown in Figure 5.

3.3. Estimation of the HVR

Removing the influence of nonresidential areas is the first step. The impact of roads and nonresidential areas needs to be removed based on the previously identified built-up areas before performing calculations because the main study object was residential areas. The method of extracting the built-up area is true and reliable, which has been verified by many studies [36]. The built-up area includes residential and non-residential areas.
The average DN for the non-residential areas was calculated. The 100 points of nonresidential areas were identified through OSM and manual visual interpretation to calculate the average DN value of the nonresidential areas (Formula (4)). The true DN values for the residential areas of grids were calculated. We subtracted the DN value of the nonresidential areas from the original DN value of the study area to obtain the DN value of the residential area (Formula (5)). The calculation formulae are as follows:
N f = 1 n j = 1 n N j
N i = L i N f
where Nf is the average DN value of the nonresidential areas; Nj is the DN value of j roads and nonresidential areas; n is the number of roads and nonresidential areas identified (n = 100); Ni is the DN value of i residential areas; Li is the original DN value of i built-up areas.
Followed by the calculation of the fully occupied residential area DN value, it was necessary to determine the threshold DN value when a house was full and to select a fully occupied residential area to calculate the HVR in the main urban area of Xi’an city accurately. The following two conditions need to be met when selecting a fully residential area: (1) a high building density. (2) high DN values. The built-up areas and the non-built-up areas were different due to the building density. It was necessary to select different building proportion areas and then select the grid that met the threshold in these areas. This paper selected a grid with a large DN value as a fully occupied residential area in the filtered grid. The DN value of the fully occupied residential area was calculated as follows:
N F = 1 m k = 1 m N k
where NF is the average DN value of the fully occupied residential area; Nk is the DN value of the k fully occupied residential area grid; m is the number of fully occupied residential area grids selected (m = 10).
Finally, the HVR in the main urban area of Xi’an city can be estimated based on the previously calculated results of the built-up area that removed the impact of roads and nonresidential areas. The average DN value when the houses were full, was calculated as follows:
H V R i = { 1 N i N F × 100 % N i < N F 0 N i N F
where HVRi is the HVR for the i grid. When the DN value of the i grid was greater than the full HVR, it was considered an outlier, and the HVR was 0.

4. Results

4.1. Spatial Distribution of Vacant Housing

The latest four scene images of the Luojia-1, mainly from June 2018 to June 2019, were selected to calculate the HVR of Xi’an city. We selected all images of the Luojia-1 covering the study area during this time period, including 19 August 2018, 23 August 2018, 31 October 2018, and 11 March 2019. The results of the four grid-based HVRs were obtained based on the remote sensing data and methods of spatial analysis (Figure 6).
The HVR results were divided into seven levels (i.e., extremely low: 0–0.05; lower: 0.05–0.1; low: 0.1–0.2; medium: 0.2–0.4; high: 0.4–0.5; higher: 0.5–0.6; extremely high: >0.6) to more intuitively observe the spatial distribution patterns of housing vacancies in the main urban area of Xi’an city, The result indicates that the spatial distribution of the HVRs for the main urban area of Xi’an city was highly consistent with the observation of the nighttime lights, and the results of this paper (Figure 6) show that the grid was used as the basic unit, and the spatial resolution was consistent with the Luojia-1 data (130 m). The results present that, the areas with a low HVR were mainly distributed in Beilin District, Yanta District, and Lianhu District, and most of the HVRs in these three districts were less than 0.1, and the full areas were concentrated. The areas with a high HVR were mainly distributed in Weiyang District, Baqiao District, and Xincheng District and mainly distributed in the marginal areas of these three districts, of which the new urban areas were mostly commercial areas and office areas. On the whole, the distribution of the HVRs was scattered, unlike other areas.

4.2. Analysis of the Results

In this section, the final results were analyzed. Average and minimum methods were chosen. Regarding the choice of the minimum method, we consider that the HVR is a constantly changing process over a specific time, and the influencing factors affecting housing vacancy are also complex. The HVR will be inflated at different times. Therefore, the minimum value was chosen to observe and analyze the change in housing vacancy. The average method, which is often used in the characterization of the object of study, such as air temperature. The researchers represent the average temperature of the day by calculating the average of each time point. This method was referred to compare it with the minimum method to find the HVR calculation method that is closest to the real situation. The results are as follows:
Figure 7 and Figure 8 show the calculated results of HVR. The spatial distribution of vacant houses can be seen in Figure 7, with clear transition zones. From Figure 8, we can see a clear high value area and low value area, while the part directly between high and low values is not obvious. In the final field verification of the main urban area of Xi’an city, it is found that the HVR results obtained by the average method are closer to the actual results. Therefore, the results were further analyzed in accordance with Figure 8.
The results of HVR were grouped by districts, and the average HVR of each district was obtained, which can be seen in Figure 9.
As the iconic old town of Xi’an city, The average HVR of Beilin District was 0.043. Beilin District is named after the well-known “Forest of Steles in Xi’an”, at home and abroad, which is an urban area with a developed culture and education system, cultural relics and scenic spots, strong scientific and technological projects, and a prosperous market economy. There are more employment opportunities in urban areas and the various basic service facilities are of high quality; thus, the HVR in Beilin District was lower than that in other urban areas.
The average HVR in Yanta District was 0.079. The data show that in 2021, the sales area for commercial housing in Yanta District reached 3.731 million square meters, accounting for 20.1% of the total sales area of Xi’an city. Having the largest GDP in Xi’an city, Yanta District is the jurisdiction where many organs are located as well as the jurisdiction with news media such as television stations and key reception venues. The infrastructure and living facilities in Yanta District are of high quality, and 162 municipal roads and five subway lines have jointly formed an efficient and convenient three-dimensional transportation network. Therefore, the overall HVR in Yanta District was low, and the utilization of housing was in good condition.
The average HVR in Lianhu District was 0.129. As the largest ethnic minority settlement in the city, Lianhu District has 43 ethnic minorities, mainly concentrated on the 2.3 square kilometer Beiyuanmen Historical and Cultural Block within the Ming City Walls. Lianhu District has significant locational advantages, with more than 25,000 commercial outlets, Star Hotel accounts for 30% of Xi’an city, and a high-quality traffic road network. As a key area for transformation and development, its purpose is to form a new pattern of linked developments for various areas in the district. The HVR increased significantly compared with Yanta District and Beilin District, and the housing use situation was better since Lianhu District is in the process of comprehensive transformation from the old city to a new urban area.
The average HVR in Weiyang District was 0.151. Weiyang District is the administrative center of Xi’an city, and the urban form, development mode, and living standards of the masses have changed rapidly toward urbanization. It vigorously supports a large number of historical and cultural attractions with far-reaching influence, such as Weiyang Palace, so that the urban space extends to the banks of the Weihe River. Metro Line 2 was completed and put into operation, and the west copper, the ring road, and the airport pass through the border at high speed, and the occupancy capacity of the city has continuously been enhanced. At the same time, the spatial distribution of housing vacancies in Weiyang District was obvious, and the vacancy rate for housing around the Xi’an North Railway Station and the Xi’an Municipal Government was low, but the vacancy rate for housing around Weiyang Palace, Daming Palace, and other scenic spots was high. Overall, the HVR in Weiyang District was higher and the housing utilization rate was lower.
The average HVR in the Xincheng District was 0.197. The new urban area, as one of three central urban areas of Xi’an city, is the seat of the Shaanxi Provincial Government and some provincial and municipal organs. The new urban area is rich in social resources, with historical relics such as Daming Palace National Heritage Park and the Ming City Wall. At the same time, it is the Xi’an city headquarters of large financial and securities institutions, such as the Xi’an city branch of various banks, and it is the most concentrated area of financial institutions in Shaanxi Province and even Northwest China. As a location for many government agencies in Shaanxi Province, the HVR in the new urban area was relatively high overall, second only to Baqiao District, mainly because of the obvious separation of occupation and housing in Xi’an city, and the new urban area is a functional agglomeration area, rather than a residential agglomeration area; therefore, the HVR was high and the housing use situation was poor.
The average HVR in Baqiao District was 0.244, which was the highest HVR in the main urban area of Xi’an city. As the first development zone named after ecology in China and the first national ecological zone in Northwest China, Baqiao District has unique natural resources, with more than 20,000 acres of natural water surface and ecological wetlands. At the same time, there are large ecological parks, such as the Xi’an Expo Park and Baqiao Ecological Wetland Park, in the jurisdiction. However, the traffic conditions in Baqiao District are worse than those in the other five main urban areas, and there are problems of insufficient commercial facilities and a lack of living facilities in the jurisdiction. Therefore, the infrastructure is not perfect and transportation is not convenient, resulting in a large number of houses in Baqiao District that have not been effectively utilized; therefore, the HVR was much higher than that in other main urban areas.

5. Discussion

5.1. Validation of the Estimated HVR

We verified the accuracy of the vacant housing results by investigating the vacancy of housing in the main urban area of Xi’an city. Different communities in the research area were selected to take pictures for verification. The following results were obtained, which can be seen through the pictures (Figure 10).
Judging from the number of lights in the community and the vacancy rate in the area, the housing placement of each community was highly compatible with the research results, and the vacancy conditions under different vacancy rates were different. Huangjia Garden is located in the northwest of the Yanta District. it’s close to the commercial center and the ancient city wall of Xi’an, with a fast flow of people, large population density, and few vacancies. In the northeast of the Yanta District, Mengcun community, due to the nearby Xi’an Jiaotong University, the Xi’an University of Technology, and other institutions of higher learning, as well as tourist attractions, such as Datang Furong Garden, the population concentration was high and the HVR was also low. The HVRs of the Lingguan community, the Country Garden, and the Shirong community, which are located in a marginal urban area, were higher because the communities are in a transition zone between urban and rural areas. Although there were more new houses, the place was sparsely populated and the use of housing was poor. Located in the northwest of Lianhu District, Taixiang Garden, although close to Weiyang Palace, people are more willing to go to the more prosperous and convenient city center because of its proximity to the city center; thus, the phenomenon of housing vacancy was more obvious than that of the Huangjia Garden and the Mengcun community.

5.2. Comparison of the Results

The number of vacant houses was low in the core area and high in the outer suburbs based on the HVR of the main urban area of Xi’an city, which is consistent with the findings of many existing vacant case cities such as Guiyang [40], Yichun [41], and Detroit [42]. The results of this study show that the HVR was the same as the distance from the city center. First, the distance from the city center had the most obvious impact on the HVR, which is consistent with Zhu’s [43] research based on Hangzhou. The closer you get to the city center, the stronger the integrity and sharing of commercial infrastructure in driving and aggregating urban real estate development. Secondly, areas with significant employment locational advantages, good public service facilities, such as education and medical care, and high advantages in subway resources are more likely to meet the commuting needs of residents and enhance their willingness to live there, resulting in ideal housing use and a low HVR [40,44]. In addition, from the results of this study, it can be seen that population density had a certain impact on urban vacancy rates. The POI data represent the degree of population aggregation to a certain extent; however, this is rarely mentioned in existing studies, indicating that population density is also a direction that should be paid more attention to in the study of HVRs. The higher the data resolution, the more accurate the estimation results, and the estimation method should be more novel and innovative.
This paper uses POI as auxiliary data to calculate the HVR in the main urban area of Xi’an city by combining Luojia-1. Compared with the results of estimating the HVR using DMSP/OLS and NPP-VIIRS, the evaluation results of HVRs herein are more accurate because the spatial resolution of Luojia-1 data used in this paper is much higher than that of DMSP/OLS [45,46] and NPP-VIIRS [12,47]. The higher the resolution of the results, the easier for us to see the details of housing vacancies. We can get a true picture of housing use at a small regional scale, which is more conducive to regional planning and development. When using Luojia-1 [31] to calculate the HVR, the impact of non-residential areas such as roads is not taken into account, resulting in a large difference between the results and the real situation. We use auxiliary data (POI) to avoid the impact of non-residential areas. When using Jilin 1-03 [8], although the spatial resolution is better than that of Luojia-1, the data acquisition is much more difficult than other data (including POI), thus it is not suitable for all researchers. Considering the above factors, the results of this paper have certain advantages in terms of resolution and accuracy.

5.3. Highlights and Limitations

Real estate plays an important role in social and economical development, especially in the industry of the national economy. The stable development of the real estate market can directly influence the life quality of residents. The study of housing vacancy has important practical implications. The highlights of this paper are reflected in the following two aspects based on the above background. Firstly, the method improves the accuracy of the calculation results. The used spatial resolution of Luojia-1 (130 m) used herein is much higher than the NPP-VIIRS (500 m) and the DMSP/OLS (1000 m), which has a high reference value for studying small-scale HVR. Secondly, the highlight of the calculation method. We compared the average results of the four phases of the images with the average results after calculating the HVR of the four-phased images separately, combined with the survey data. The result indicated that the average method was closer to the real housing vacancy conditions and provides new ideas and new methods for the calculation and research of the HVR.
This paper proposed a new method to quantitatively evaluate the current situation of housing vacancies in the area using remote sensing data and the POI of night lights. Prior to the use of this methodology to evaluate the HVRs, various methods should be used to verify the accuracy of the data and to compare the data obtained in various ways to determine the applicability and accuracy of the method. Coupled with the complexity of the housing vacancy problem, the research results of this paper also have a certain degree of limitations, The main limitations are as follows:
Although this paper used the highest resolution Luojia-1 nighttime light data (spatial resolution: 130 m) that can be obtained at present, such a spatial resolution was still lower and insufficient for conducting urban internal research. The accuracy is slightly unsatisfactory.
Nighttime light data are susceptible to the urban area light intensity, fire, exhaust gas combustion, and many other background noises. At present, researchers have proposed noise reduction, calibration, desaturation, and other processing methods, but there is no universally accepted method. Although the Luojia-1 data used in this paper greatly reduced the saturation spillover effect, the measurement error will still have an impact on the accuracy of the housing vacancy estimate.
In many cities, especially in megacities, the urban built-up area and the night lighting area do not necessarily coincide strictly, the separation of work and housing is widespread, and there is no strict proportional relationship between the strength of the light and the housing vacancy. The higher the degree of separation between work and housing, the greater the impact on the estimation of the HVRs.
At present, the situation of vacant housing in cities and towns is grim, and in the future, we will explore housing vacancies in more depth and analyze the spatial and temporal changes of vacant housing by obtaining multiperiod vacancy data. A comprehensive influencing factor index system related to vacancy may be constructed and its relationship with the social environment, built environment, and other factors will be explored further. Future work could provide a better understanding of the types of vacancy problems for researchers that are unique to different cities and regions in China.

5.4. Policy Implications

In recent years, the combination of remote sensing and big data has provided a new approach to housing vacancy research. As a dynamic and variable phenomenon, it is impossible to evaluate housing vacancy through a single piece of data. The advent of big data fills this gap. The real-time nature of big data allows us to monitor the state of housing and understand the status of housing use at any time. In terms of research methods, big data can be combined with remote sensing to develop a real-time housing vacancy monitoring method to keep track of the state of housing in cities. Transforming scientific research into practical applications.
The conclusions of this study provide a reference for government departments to formulate relevant policies to solve the problem of vacant housing and alleviate housing pressure.
In areas with serious housing vacancy problems, such as the suburbs, it is recommended to limit large-scale real estate development, strictly control the number of houses, and avoid oversupply. However, the vacancy rate in the city center was low, which indicates that there is a high demand for housing in these areas. Authorities should provide more housing for home buyers/renters, ensure the housing supply, and alleviate housing pressure. For example, housing needs can be met by increasing the proportion of plots in urban areas.
We suggest guiding developers to develop rationally and avoid blindly developing large and high-priced houses to pursue profits, resulting in a mismatch between the supply and demand structure.
For areas with high altitudes, due to the fact of remoteness and other reasons, it is recommended to further strengthen the construction of housing infrastructure, provide residents with convenient transportation, mature and perfect facilities, etc., to meet the convenience needs of residents in work and daily life.
Finally, we also suggest curbing speculation by some developers (e.g., overselling) that lead to a large number of vacant homes.

6. Conclusions

In this paper, the HVR in the main urban area of Xi’an city was estimated by using nighttime and POI data, and the housing vacancy status in the main urban area of Xi’an city from 2018 to 2019 was obtained by analyzing data from remote sensing images at different times. The main conclusions and highlights of this study were that the (1) Luojia-1 nighttime light and POI data improved the spatial resolution of the housing vacancy results, and this paper provided new ideas for housing vacancy research; (2) the multiperiod data were processed and calculated and the HVR was obtained by calculating the average, which represented the housing vacancy level in the main urban area of Xi’an city from 2018 to 2019, and we provided a new method to analyze the spatial distribution of housing vacancies.
Housing vacancy is an interesting and appealing concern. The four scene images of Luojia-1 with high resolution and POI were used to estimate HVR in this paper. The combination of multisource remote sensing data with higher spatio-temporal resolution and statistical data to estimate HVR will be conducted in the next step. In addition, machine learning algorithms are also considered to estimate HVR. The method proposed in this paper makes up for the shortcomings of traditional HVR statistics and improves the accuracy of HVR spatial estimation. It is expected to become a new spatial estimation method for monitoring housing vacancy. The estimation results can provide scientific reference for the healthy development of regional real estate, government macro-control, and related research of scholars.

Author Contributions

Conceptualization, P.Y. and J.P.; methodology, P.Y. and J.P.; software, P.Y.; validation, P.Y.; formal analysis, P.Y.; investigation, P.Y.; resources, P.Y.; data curation, P.Y.; writing—original draft preparation, P.Y.; writing—review and editing, P.Y. and J.P.; visualization, P.Y.; supervision, J.P.; project administration, J.P.; funding acquisition, J.P. All authors have read and agreed to the published version of the manuscript.


This study was funded by the National Natural Science Foundation of China (no. 42071216).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data and statistical analysis methods are available upon request from the corresponding author.


The authors thank the editors and anonymous reviewers for their thoughtful and helpful suggestions on improving the manuscript.

Conflicts of Interest

The authors declare no conflict of interest.


  1. Wuyts, W.; Sedlitzky, R.; Morita, M.; Tanikawa, H. Understanding and managing vacant houses in support of a material stock-type society—The case of Kitakyushu, Japan. Sustainability 2020, 12, 5363. [Google Scholar] [CrossRef]
  2. Kelling, G.L.; Wilson, J.Q. Broken Windows: The Police and Neighborhood Safety; The Atlantic Daily: Washington, DC, USA, 1982; Volume 249, pp. 29–38. [Google Scholar]
  3. Han, H.-S. The impact of abandoned properties on nearby property values. Hous. Policy Debate 2014, 24, 311–334. [Google Scholar] [CrossRef]
  4. Kato, H. Total Fertility Rate, Economic–Social Conditions, and Public Policies in OECD Countries. In Macro-Econometric Analysis on Determinants of Fertility Behavior; Springer: Berlin/Heidelberg, Germany, 2021; pp. 51–76. [Google Scholar]
  5. Xu, H.; Rong, C. Changes in China’s Housing Demand and Suggestions for Countermeasures in the “14th Five-Year Plan” Period. Macroeconomics 2021, 8, 151–167. [Google Scholar]
  6. Jin, X.; Long, Y.; Sun, W.; Lu, Y.; Yang, X.; Tang, J. Evaluating cities’ vitality and identifying ghost cities in China with emerging geographical data. Cities 2017, 63, 98–109. [Google Scholar] [CrossRef]
  7. Han, X.; Zhang, M.; Hu, Y.; Huang, Y. Study on the Digital Transformation Capability of Cost Consultation Enterprises Based on Maturity Model. Sustainability 2022, 14, 10038. [Google Scholar] [CrossRef]
  8. Du, M.; Wang, L.; Zou, S.; Shi, C. Modeling the census tract level housing vacancy rate with the Jilin1-03 satellite and other geospatial data. Remote Sens. 2018, 10, 1920. [Google Scholar] [CrossRef] [Green Version]
  9. Lu, D. Urbanization Process and Spatial Sprawl in China. Urban Plan. Forum 2007, 4, 47–52. [Google Scholar]
  10. Zhao, S.; Li, W.; Zhao, K.; Zhang, P. Change Characteristics and Multilevel Influencing Factors of Real Estate Inventory—Case Studies from 35 Key Cities in China. Land 2021, 10, 928. [Google Scholar] [CrossRef]
  11. Yin, L.; Silverman, R.M. Housing abandonment and demolition: Exploring the use of micro-level and multi-year models. ISPRS Int. J. Geo-Inf. 2015, 4, 1184–1200. [Google Scholar] [CrossRef] [Green Version]
  12. Chen, Z.; Yu, B.; Hu, Y.; Huang, C.; Shi, K.; Wu, J. Estimating house vacancy rate in metropolitan areas using NPP-VIIRS nighttime light composite data. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2015, 8, 2188–2197. [Google Scholar] [CrossRef]
  13. Di Pasquale, D.; Wheaton, W.C. Urban Economics and Real Estate Markets; Prentice Hall: Englewood Cliffs, NJ, USA, 1996; Volume 23. [Google Scholar]
  14. Survey and Research Center for China Household Finance. Urban Housing Vacancy Rate and Housing Market Development Trend. Available online: (accessed on 12 April 2022).
  15. Lee, J.; Newman, G.; Lee, C. Predicting Detached Housing Vacancy: A Multilevel Analysis. Sustainability 2022, 14, 922. [Google Scholar] [CrossRef]
  16. Zhang, C.; Jia, S.; Yang, R. Housing affordability and housing vacancy in China: The role of income inequality. J. Hous. Econ. 2016, 33, 4–14. [Google Scholar] [CrossRef]
  17. Radzimski, A. Changing policy responses to shrinkage: The case of dealing with housing vacancies in Eastern Germany. Cities 2016, 50, 197–205. [Google Scholar] [CrossRef]
  18. Wang, H. Stickiness of rental rate and housing vacancy rate. Econ. Lett. 2020, 195, 109487. [Google Scholar] [CrossRef]
  19. Yoo, H.; Kwon, Y. Different Factors Affecting Vacant Housing According to Regional Characteristics in South Korea. Sustainability 2019, 11, 6913. [Google Scholar] [CrossRef] [Green Version]
  20. Hino, K.; Mizutani, K.; Asami, Y.; Baba, H.; Ishii, N. Attitudes of parents and children toward housing inheritance in a Tokyo suburb. J. Asian Archit. Build. Eng. 2022, 21, 2131–2140. [Google Scholar] [CrossRef]
  21. Gandhi, S.; Green, R.K.; Patranabis, S. Insecure property rights and the housing market: Explaining India’s housing vacancy paradox. J. Urban Econ. 2022, 131, 103490. [Google Scholar] [CrossRef]
  22. Park, I.K.; Ciorici, P. Determinants of vacant lot conversion into community gardens: Evidence from Philadelphia. Int. J. Urban Sci. 2013, 17, 385–398. [Google Scholar] [CrossRef]
  23. Armstrong, G.; Soebarto, V.; Zuo, J. Vacancy Visual Analytics Method: Evaluating adaptive reuse as an urban regeneration strategy through understanding vacancy. Cities 2021, 115, 103220. [Google Scholar] [CrossRef]
  24. Ge, W.; Yang, H.; Zhu, X.; Ma, M.; Yang, Y. Ghost City Extraction and Rate Estimation in China Based on NPP-VIIRS Night-Time Light Data. ISPRS Int. J. Geo-Inf. 2018, 7, 219. [Google Scholar] [CrossRef] [Green Version]
  25. Branas, C.C.; South, E.; Kondo, M.C.; Hohl, B.C.; Bourgois, P.; Wiebe, D.J.; MacDonald, J.M. Citywide cluster randomized trial to restore blighted vacant land and its effects on violence, crime, and fear. Proc. Natl. Acad. Sci. USA 2018, 115, 2946–2951. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  26. Gabriel, S.A.; Nothaft, F.E. Rental housing markets, the incidence and duration of vacancy, and the natural vacancy rate. J. Urban Econ. 2001, 49, 121–149. [Google Scholar] [CrossRef]
  27. Pan, J.; Dong, L. Spatial identification of housing vacancy in China. Chin. Geogr. Sci. 2021, 31, 359–375. [Google Scholar] [CrossRef]
  28. Lu, H.; Lv, J. The research of Wuhan commercial housing vacancy rate. In Proceedings of the 2012 International Conference on Information Management, Innovation Management and Industrial Engineering, Sanya, China, 20–21 October 2012; pp. 1–4. [Google Scholar]
  29. Kabisch, N.; Haase, D.; Haase, A. Evolving Reurbanisation? Spatio-temporal Dynamics as Exemplified by the East German City of Leipzig. Urban Stud. 2010, 47, 967–990. [Google Scholar] [CrossRef]
  30. Zheng, Q.; Deng, J.; Jiang, R.; Wang, K.; Xue, X.; Lin, Y.; Huang, Z.; Shen, Z.; Li, J.; Shahtahmassebi, A.R. Monitoring and assessing “ghost cities” in Northeast China from the view of nighttime light remote sensing data. Habitat Int. 2017, 70, 34–42. [Google Scholar] [CrossRef]
  31. Tan, Z.; Wei, D.; Yin, Z. Housing Vacancy Rate in Major Cities in China: Perspectives from Nighttime Light Data. Complexity 2020, 2020, 5104578. [Google Scholar] [CrossRef]
  32. Ge, Y.; Han, L.; Zhao, Y.; Ao, Y.; Ding, J.; Zhu, Y.; Liu, B. Spatiotemporal analysis of urban expansion in Xi’an from 1984 to 2016. Chin. J. Ecol. 2019, 5, 1491–1499. [Google Scholar]
  33. Governor of Xi’an Building. 2021 Xi’an Real Estate Market Analysis Report. Available online: (accessed on 20 August 2022).
  34. Guan, Z.; Zhang, G.; Jiang, Y.; Shen, X.; Li, Z. Luojia-1 Nightlight Image Registration Based on Sparse Lights. Remote Sens. 2022, 14, 2372. [Google Scholar] [CrossRef]
  35. Ma, M.; Lang, Q.; Yang, H.; Shi, K.; Ge, W. Identification of polycentric cities in China based on NPP-VIIRS nighttime light data. Remote Sens. 2020, 12, 3248. [Google Scholar] [CrossRef]
  36. LI, F.; YAN, Q.; ZOU, Y.; LIU, B. Extraction Accuracy of Urban Built-up Area Based on Nighttime Light Data and POI: A Case Study of Luojia 1-01 and NPP/VIIRS Nighttime Light Images. Geomat. Inf. Sci. Wuhan Univ. 2021, 46, 825–835. [Google Scholar]
  37. Wu, J.; Zhang, Z.; Yang, X.; Li, X. Analyzing Pixel-Level Relationships between Luojia 1-01 Nighttime Light and Urban Surface Features by Separating the Pixel Blooming Effect. Remote Sens. 2021, 13, 4838. [Google Scholar] [CrossRef]
  38. Sheather, S.J.; Jones, M.C. A reliable data-based bandwidth selection method for kernel density estimation. J. R. Stat. Soc. Ser. B 1991, 53, 683–690. [Google Scholar] [CrossRef]
  39. Henderson, M.; Yeh, E.T.; Gong, P.; Elvidge, C.; Baugh, K. Validation of urban boundaries derived from global night-time satellite imagery. Int. J. Remote Sens. 2003, 24, 595–609. [Google Scholar] [CrossRef]
  40. Shi, L.; Wurm, M.; Huang, X.; Zhong, T.; Leichtle, T.; Taubenböck, H. Estimating housing vacancy rates at block level: The example of Guiyang, China. Landsc. Urban Plan. 2022, 224, 104431. [Google Scholar] [CrossRef]
  41. Liu, Y.; Sun, Y.; Sun, H.; Fu, H. Spatial-temporal differentiation and influence mechanism of housing vacancy in shrinking cities: Based on the perspective of residential electricity consumption. Sci. Geogr. Sin 2021, 41, 2087–2095. [Google Scholar]
  42. Bentley, G.C.; McCutcheon, P.; Cromley, R.G.; Hanink, D.M. Race, class, unemployment, and housing vacancies in Detroit: An empirical analysis. Urban Geogr. 2016, 37, 785–800. [Google Scholar] [CrossRef]
  43. Zhu, J.; Ye, Y. Study on the characteristics of housing vacancy—Case study of the housing vacancy rate in hangzhou. Bull. Sci. Technol. 2011, 27, 142–147. [Google Scholar] [CrossRef]
  44. Williams, S.; Xu, W.; Tan, S.B.; Foster, M.J.; Chen, C. Ghost cities of China: Identifying urban vacancy through social media data. Cities 2019, 94, 275–285. [Google Scholar] [CrossRef]
  45. Yao, Y.; Li, Y. House vacancy at urban areas in China with nocturnal light data of DMSP-OLS. In Proceedings of the 2011 IEEE International Conference on Spatial Data Mining and Geographical Knowledge Services, Fuzhou, China, 29 June–1 July 2011; pp. 457–462. [Google Scholar]
  46. Zhang, L.; Qu, G.; Wang, W. Estimating land development time lags in China using DMSP/OLS nighttime light image. Remote Sens. 2015, 7, 882–904. [Google Scholar] [CrossRef] [Green Version]
  47. Wang, L.; Fan, H.; Wang, Y. An estimation of housing vacancy rate using NPP-VIIRS night-time light data and OpenStreetMap data. Int. J. Remote Sens. 2019, 40, 8566–8588. [Google Scholar] [CrossRef]
Figure 1. Location of the study area.
Figure 1. Location of the study area.
Applsci 12 12328 g001
Figure 2. Workflow diagram of the estimation of the HVR.
Figure 2. Workflow diagram of the estimation of the HVR.
Applsci 12 12328 g002
Figure 3. Comparison of (a) before and (b) after image correction.
Figure 3. Comparison of (a) before and (b) after image correction.
Applsci 12 12328 g003
Figure 4. Kernel Density of POI.
Figure 4. Kernel Density of POI.
Applsci 12 12328 g004
Figure 5. Spatial distribution of the LJ&POI index.
Figure 5. Spatial distribution of the LJ&POI index.
Applsci 12 12328 g005
Figure 6. House vacancy rate at different times: (a) 19 August 2018; (b) 23 August 2018; (c) 31 October 2018; (d) 11 March 2019.
Figure 6. House vacancy rate at different times: (a) 19 August 2018; (b) 23 August 2018; (c) 31 October 2018; (d) 11 March 2019.
Applsci 12 12328 g006
Figure 7. Average house vacancy rates in Xi’an city.
Figure 7. Average house vacancy rates in Xi’an city.
Applsci 12 12328 g007
Figure 8. Minimum values of the house vacancy rates in Xi’an city.
Figure 8. Minimum values of the house vacancy rates in Xi’an city.
Applsci 12 12328 g008
Figure 9. House vacancy rates at the district scale.
Figure 9. House vacancy rates at the district scale.
Applsci 12 12328 g009
Figure 10. Results validation in the community.
Figure 10. Results validation in the community.
Applsci 12 12328 g010
Table 1. Comparison of different nighttime light data parameters.
Table 1. Comparison of different nighttime light data parameters.
Data SourceCountryNumber of Data BitsSpatial Resolution/mTime in OrbitRevisit CycleWidth/km
NPP/VIIRSAmerica65002011–present12 h3000
DMSP/OLSAmerica1410001992–201312 h3000
Luojia-1China151302018–present3–5 d250
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Yang, P.; Pan, J. Estimating Housing Vacancy Rate Using Nightlight and POI: A Case Study of Main Urban Area of Xi’an City, China. Appl. Sci. 2022, 12, 12328.

AMA Style

Yang P, Pan J. Estimating Housing Vacancy Rate Using Nightlight and POI: A Case Study of Main Urban Area of Xi’an City, China. Applied Sciences. 2022; 12(23):12328.

Chicago/Turabian Style

Yang, Pengfei, and Jinghu Pan. 2022. "Estimating Housing Vacancy Rate Using Nightlight and POI: A Case Study of Main Urban Area of Xi’an City, China" Applied Sciences 12, no. 23: 12328.

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop