Urban Resilience of Large Public Health Events Based on NPP-VIIRS Nighttime Light Images: A Case Study of 35 Large Cities in China

Abstract

The COVID-19 outbreak directly and severely threatens global public health. Non-drug interventions in response to the COVID-19 pandemic have significantly altered urban socioeconomic activity. Understanding the different levels of city resilience to the impact of COVID-19 on urban human activities is essential. In this paper, 35 large cities in China were selected as research areas, and based on NPP-VIIRS night light images, the spatial pattern changes in human activities during the epidemic period from the end of December 2019 to December 2022 were explored. The results are as follows: (1) In the first two months of the epidemic, the luminous value of large cities showed an extensive range of decline, and the decline in different urban functional places was different. (2) There is a significant positive correlation between the urban population and the luminous change value. The closer the relationship between urban places and human activities, the stronger the correlation between the population and the luminous change value of urban places. (3) In the middle and later stages of the epidemic, the night light value of all cities showed an upward trend, but there was a difference. (4) The increase in the number of confirmed cases in the middle and later stages of the epidemic could hardly lead to a significant decrease in the value of night light on a monthly scale unless the city had a relatively large area and a relatively strict lockdown policy in that month. This study will help inform future strategies and decisions to effectively combat epidemics and the construction of resilient cities.

1. Introduction

The occurrence of major public health emergencies has profound impacts on urban populations, disrupting both economic and daily activities. The COVID-19 pandemic serves as a salient example, significantly affecting cities worldwide. Economic activities have been curtailed, with businesses facing closures and supply chains experiencing unprecedented disruptions. Simultaneously, daily life has been upended, as social distancing measures, lockdowns, and other restrictions have altered how individuals work, socialize, and access essential services. Given these extensive disruptions, it becomes crucial to evaluate urban resilience in the face of such public health crises, which is closely related to the construction of sustainable cities. Understanding how cities can withstand, adapt to, and recover from these shocks is essential for developing strategies that enhance urban sustainability and mitigate the adverse effects of future emergencies.

Numerous studies have assessed the multifaceted impact of COVID-19 on diverse environmental and socioeconomic aspects, including epidemiology [1,2], air pollution [3,4], economic growth [5], tourism [6,7], electricity consumption [8], and human activities [9,10,11]. Nighttime light (NTL), detectable through remote sensing, closely correlates with human activities, reflecting urban and rural illumination, including sources like fishing boats, natural gas combustion, and forest fires. Widely applied across disciplines, NTL data efficiently assess long-term socioeconomic parameters (e.g., GDP, human population, power consumption, poverty) [12,13,14,15,16], quantify urbanization and sprawl [17,18,19], and monitor impacts of unforeseen events (e.g., war, natural disasters, power outages) [20,21,22]. Considering that NTL data can analyze economic or social activities, and VIIRS can detect NTL with higher spatial and radiological resolution than DMSP-OLS, many studies have used NPP-VIIRS data to track changes in urban activity during the COVID-19 epidemic. For example, March et al. investigated the worldwide alterations in maritime traffic patterns throughout the initial six months of 2020, utilizing the Satellite Automatic Identification System (S-AIS). Their findings revealed that, concomitant with the declaration of the COVID-19 pandemic in March 2020, approximately 52.2% of designated geographic cells exhibited a decline in traffic density. Subsequently, this general diminution in maritime traffic density persisted throughout the subsequent three months of the study, with the highest reductions observed in April, reaching a peak of 54.8% in the designated cells [23]. Some studies have taken into account changes in carbon emissions during COVID-19 [24,25,26]. Liu et al. estimated the daily CO₂ emissions from different sectors at the national level in the first half of 2020. The results show that global CO₂ emissions plunged by 8.8% (−1551 Mt CO₂) compared to the same period in 2019 [27]. Chen et al. analyzed the water quality of the Haihe River Basin during the COVID-19 control period using Sentinel-2 visible and near-infrared band reflectance, along with the Normalized Difference Turbidity Index (NDTI) [28]. Du et al. develops a rapid accounting method for local residential electricity power consumption (EPC) using nighttime light data, allowing us to examine the changes in residential EPC before and after the COVID-19 pandemic [29]. Yan et al. gathers points of interest (POI) data and NTL remote sensing data (RSD) to spatialize nighttime economic activities, aiming to provide a reference for effective regional and urban economic planning [30].

However, there is a lack of research exploring urban resilience and patterns in response to COVID-19. In the case of similar economic levels, human activities in cities tend to show specific rules, and the differences in administrative regions, population numbers, and other factors will have a specific impact on the performance of human activities [31,32]. Therefore, based on NPP-VIIRS data, this paper constructs a month-by-month time series of night lights in 35 large cities in China from December 2019 to December 2022 and classifies these cities into six administrative regions in China. The changes in the NTL of the night light index were explored under the influence of factors such as economic level, population size, geographical location during the epidemic period, and the impact of epidemic containment policies on human activities. Urban resilience and patterns during the COVID-19 period are further explored. To test the reliability of the comparative analysis, we compared the changing pattern of the time series from December 2016 to December 2019. In addition, we also explored the differences in NTL changes in different functional locations within cities. We constructed a dataset of changes in different functional locations of 35 large cities in China which will be useful for future research. This paper can guide economic loss assessment and prevention and control policy formulation of similar large-scale public health events in the future.

2. Study Area and Dataset

In this study, 35 big cities in China are selected as the study area, which are the hot spots of urbanization in China today and before 2035 in terms of economic level, urbanization degree, and population size. Considering that the division of urban agglomerations and administrative districts may have similarities and differences in the performance of large cities, the selection of large cities in the study area took these two influencing factors into account. The selected cities are distributed in the administrative regions of China, namely, Northeast China, North China, Northwest China, South-central China, Southwest China and Eastern China. Cities in China’s large urban agglomerations are also included in the selected category, including the Northwest urban agglomeration, the Northeast urban agglomeration, the Beijing–Tianjin–Hebei urban agglomeration, the Sichuan–Chongqing urban agglomeration, the Yangtze River Delta urban agglomeration and the Pearl River Delta urban agglomeration. Tracking human activity in these cities during the pandemic is critical as they represent the main body and highest level of human activity within cities across China. Figure 1 shows the geographic location, administrative divisions, urban agglomeration distribution and population size of 35 cities selected in this study.

The Version 1 monthly averaged NPP-VIIRS NTL images, including 72 months in total from December 2016 to December 2022, were used to track changes in human activity during the pandemic. This dataset constitutes a monthly composite product, offering an aggregate representation of the average radiance derived from all eligible, cloud-free observations within the temporal span of a single month. The NPP-VIIRS image was resampled to a resolution of 500 m × 500 m, and a Lambert projection was applied. The threshold method is used to preprocess the maximum outliers. We believe that the maximum NTL value in China’s city-wide can only appear in megacities such as Beijing, Shanghai, and Shenzhen. Considering that a single megacity may be affected by lockdown measures during the epidemic and have low NTL values, we selected ten large cities, such as Beijing, Shanghai, and Shenzhen, as candidates for the maximum NTL value. The maximum night light value of these ten cities in each month is used as the threshold value, and night light values higher than this value are suppressed. At the same time, it is worth noting that the quality of NPP-VIIRS images will be seriously affected by stray light in mid-high latitudes in May, June, and July every year, resulting in its unavailability. Therefore, considering the above factor and facilitating the alignment between cities, we have uniformly excluded the image data of each city in May, June, and July. In addition, urban population data, economic data, and data on the number of confirmed cases were downloaded for use in the city statistical yearbook and official website.

3. Methodology

Difference analysis was used in the early phase of the outbreak (January and February 2020) analysis, where we differentially processed the night light images of each city in December 2019 and January 2020, and January 2020 and February 2020. It counted the places in the city with significant changes. All the places are divided into six categories according to different functions: transportation junctions, commercial blocks, industrial parks, district communities, infrastructure constructions, and tourist attractions.

A time series based on a linear model was used to analyze the subsequent phases of the outbreak. We constructed a time series from December 2016 to December 2022 based on monthly images of night lights. The period from December 2019 to December 2022 is the three-year period in which COVID-19 exists in China, after which the prevention and control measures against the virus were relaxed. To ensure that the trend of the night light value time series is analyzed, the data from December 2016 to December 2019 are added to the time series. We filter outliers based on the average luminous values for each city, and any month with luminous values greater than one standard deviation of the mean will be excluded and not used to construct the time series. The linear model is adopted, and the data of this month and the two months before the data of this month are used as independent variables to perform linear regression on the data of this month, and then the predicted value of the model and the probability of the actual value occurring based on the model are obtained. Figure 2 shows technique process of this study.

4. Results

4.1. The Change in Night Light Value in the Early Stage of the Epidemic

4.1.1. Overall Change in the Value of City Light at Night

During the first two months of the COVID-19 epidemic, many pixels in the selected cities were observed to have decreased night light values.

Table 1. Changes in NTL in 35 cities at the beginning of the COVID-19 outbreak.

Cities	December 2019 to January 2020	January 2020 to February 2020
Beijing	9.89%	23.35%
Cangzhou	2.37%	9.76%
Changsha	7.87%	8.22%
Chengdu	19.12%	7.95%
Dalian	2.12%	6.67%
Dongguan	37.46%	26.56%
Foshan	21.66%	21.86%
Guangzhou	17.94%	19.42%
Hefei	9.54%	3.38%
Jinan	7.96%	11.55%
Lanzhou	2.71%	2.09%
Nanjing	14.87%	19.26%
Nantong	7.59%	8.98%
Ningbo	16.68%	9.44%
Qingdao	7.02%	11.35%
Shanghai	35.65%	29.79%
Shenyang	9.34%	11.90%
Shenzhen	38.61%	31.65%
Shijiazhuang	2.95%	9.04%
Suzhou	24.96%	17.70%
Tangshan	3.29%	9.59%
Tianjin	5.92%	28.03%
Wenzhou	9.71%	9.54%
Wuhan	17.41%	7.34%
Wulumuqi	6.94%	7.48%
Wuxi	19.22%	12.57%
Xian	11.22%	4.85%
Xuzhou	5.31%	5.92%
Yantai	3.69%	4.72%
Zhengzhou	14.52%	15.41%
Huhehaote	2.80%	7.51%
Hangzhou	10.65%	5.28%
Changchun	2.78%	15.26%
Chongqing	3.24%	2.35%
Haerbin	1.45%	5.60%

shows the percentage of pixels with reduced night light in 35 cities from December 2019 to January 2020 and from January 2020 to February 2020. In the first month of COVID-19, most cities experienced a large area of decline in NTL values. In Shanghai, Shenzhen, and other cities, the pixel value of the night light decline accounts for more than 35%. In the Yangtze River Delta city cluster and the Pearl River Delta city cluster, the proportion of pixels with decreased night light is higher, more than 10%, while in the Beijing–Tianjin–Hebei city cluster, the northwest city cluster and the northeast city cluster, the proportion of pixels with decreased night light is relatively small; most of them are less than 10%. As municipalities directly under the central government, Beijing, Chongqing, and Tianjin have large urban administrative areas, and the night light images contain many non-urban pixels. As a result, although many pixels of NTL value decreased in these three cities, the overall proportion was relatively insignificant. Some cities have experienced a “lag” due to the COVID-19 epidemic. In the second month after COVID-19 began, the proportion of changing pixels in some cities was much higher than in the first month, such as Beijing and Tianjin. In addition, from the perspective of the situation of the two months, we found that the more developed cities and regions were, the more pixels in their night lights images were affected.

4.1.2. The Change in Different Functional Places in the City

In the city, the correlation and interaction between different places and human activities are different because of the different functions of different places and the different forms of human activities. Therefore, this study divides functional places within cities into six categories: transportation junctions, commercial blocks, residential quarters, industrial parks, infrastructure constructions, and tourist attractions. Figure 3 shows the specific locations of the changes in these six types of venues in Wuhan, Beijing, and Hangzhou from December 2019 to January 2019. It can be seen that these venues have been significantly affected, and the luminous value has shown a significant decline. We extracted the image difference values of 35 large cities in December 2019 and January 2020, as well as January 2020 and February 2020, and counted the data of the sites that were significantly affected in the six types of places, as shown in Figure 4.

Transportation junctions and tourist attractions showed the most significant decrease in night light value, and under the epidemic containment policy, traffic was significantly affected, and the traffic flow between different cities was significantly reduced. At the same time, the tourism industry was subjected to unprecedented restrictions. Under the stringent containment policy in the early stage of the epidemic, tourism-related activities tended to stall, and the value of night lights were significantly reduced. Commercial blocks and industrial parks are closely related to economic activities. Although commercial activities and industrial production are suppressed to a certain extent under the containment policy, they are not as affected as the transportation and tourism industries. Urban infrastructure constructions and residential quarters, closely related to basic daily life, will be affected by the containment of the epidemic. However, the extent of the impact will not be considerable.

It is worth noting that some places still had increased night light brightness values during the first two months of the epidemic. Figure 3 shows the locations of Wuhan and Hangzhou with significant increases in night light values (ΔNTL > 100) in January 2020 compared to December 2019. Through data collection and comparison of satellite images, we know that there are frequent and active human activities in these locations. For example, Figure 3b shows the night light changes at Leishenshan Hospital in Wuhan. The hospital is an emergency epidemic prevention hospital built for epidemic prevention and control in 2020, and it took 12 days from design to completion.

We analyzed the NTL changes of 6 functional places in 35 cities in the first 2 months of the epidemic, as shown in Figure 5. In the first and second months of the outbreak, the NTL values of transportation junctions showed a more comprehensive range of changes compared with other functional sites. The reduction in the brightness value of the night light is mainly between 20 and 60. Unlike other functional places, the human activities represented by transportation junctions are more inclined to communication between cities. In order to prevent the spread of COVID-19, the local government has implemented a strict lockdown policy, reducing communication between cities. Wuhan, the city with the most infected people, adopted a stringent policy of “shutting down the city”. Tourist attractions were also significantly affected, but the reduction in the night light value is between 15 and 40, which is different from the 20 to 40 of other functional places. This may be because tourist attractions’ type of night light source differs from other available places. The COVID-19 epidemic had a lag effect on the night light value of residential quarters, and there was a significant decrease in the night light value of residential neighborhoods in the second month after the epidemic’s beginning. Compared with January 2020, the reduction range and average value of night light brightness in residential quarters were more extensive. Compared with other functional places, the decreased range and mean change of night light values in residential quarters in February 2020 were much higher. In 2020, the Chinese Lunar New Year fell in January, during which the night light value of residential quarters was more significant, which is one of the factors causing the large difference in night light value between January and February. Most of the NTL values in commercial blocks vary between 20 and 40, with little difference between January and February. The night light value changes in February 2020 are more comprehensive and higher for infrastructure constructions and industrial parks. To prevent the spread of COVID-19, strict prevention and control policies were implemented. Residents’ living activities related to urban infrastructure constructions and production activities related to enterprises were limited to a certain extent, and the corresponding reduction range of night light brightness value is more significant.

We further explored the positive correlation between the night light decline value of different functional places in the city and the indicators related to human activities, such as population, GDP, and built-up area, in the first two months of the COVID-19 outbreak, as shown in Figure 6. There was a strong positive correlation between the population value and the change value of night light in each functional place. Tourist attractions, commercial blocks, and transportation junctions have the most significant positive correlation with the population, and human activities in these places have precisely the strongest positive correlation with the population. In January 2020, the luminous change values of the six functional places showed a strong correlation with the built-up area, but in February 2020, the positive correlation became weak, and the correlation between the luminous change values of tourist attractions and residential quarters and the built-up area became negative. In January 2020, the luminous change value of six functional places showed a weak correlation with urban GDP. However, in February 2020, the weak positive correlation became weaker, and the correlation between the luminous change value of tourist attractions, residential quarters, industrial parks, and infrastructure constructions and the built-up area became a negative correlation. This shows that these four types of venues in cities with larger GDPs were less affected by the subsequent night lights in the early stage of the epidemic.

4.2. Changes in Night Light Value in the Middle and Late Period of the Epidemic

4.2.1. Overall Trend of Night Light Value under Time Series

As of 8 May, the repercussions of the pandemic on the nocturnal luminosity of Chinese urban areas exhibited a mitigated trajectory consequent to the relaxation of lockdown measures and the resumption of select economic pursuits. Nevertheless, discernible differentials endure among cities, characterized by sustained diminution in nocturnal luminosity in certain megacities where there is a pronounced escalation in confirmed cases of COVID-19. Based on monthly images, we constructed a long-time series of the total value of night lights in each city. We constructed a time series trend based on a linear model, as shown in Figure 7. We found that, on an annual scale, the COVID-19 pandemic did not significantly impact the total value of urban night light. In the cities studied, the total value of night light in already developed megacities shows a slow growth trend or tends to flatten out. These cities include Beijing, Tianjin, Suzhou, Chongqing, and Shanghai, and the total value of night lights in these cities exceeds 350,000. In recent years, the total value of night lights in large and dynamic cities has also shown a rapid growth trend. In the six years from 2017 to 2022, the total value of night lights in Shenzhen has increased by 40.49%, and in the three years from 2020 to 2022 that the COVID-19 pandemic lasted, the total value of night lights in Shenzhen also increased by 21.60%. We observed little increase or even a decrease in the full value of night light in some cities, almost all in the northwest and northeast regions.

4.2.2. Correlation between NTL Values and Human Activities on an Annual Scale

We counted the total value of NTL in 35 cities on a monthly scale and excluded the months with outliers. The NTL values of effective months in each year were added and averaged to obtain the annual average NTL total value of the city in that year. We also counted GDP, population, and built-up area data for each city year. The data used are from the city statistical yearbook published by the city government. We conduct a correlation analysis between the obtained results and the annual NTL values of the cities. Figure 8 shows their Spearman correlation coefficient, and

Table 2. Spearman correlation coefficient between the total value of night light and GDP, population and built-up area.

Year	2017	2018	2019	2020	2021	2022
GDP	0.717 **	0.661 **	0.781 **	0.772 **	0.782 **	0.741 **
Built-up Area	0.706 **	0.681 **	0.768 **	0.71 **	0.73 **	0.759 **
Population	0.429 *	0.423 *	0.473 **	0.749 **	0.659 **	0.728 **

* p < 0.05, ** p < 0.01.

shows the detailed correlation data. We found that the correlations of NTL values with population, GDP, and built-up area were higher in the three years when COVID-19 was ongoing compared to the three years before COVID-19. In addition, we also find that the total value of urban NTL has the highest correlation with the city’s GDP. In the three years of the COVID-19 epidemic, the three-year mean value of the correlation between the two is higher than 0.760, which is a significant correlation. The correlation between urban built-up areas and the total value of urban NTL is weaker than the correlation between GDP and the total value of urban NTL. However, the three-year average is also higher than 0.735. Population is even less correlated with the total value of the city’s NTL, with a three-year mean of 0.712. The above results show that the COVID-19 epidemic has not significantly impacted the city’s annual development and has not affected its development trend. In addition, the analysis of the correlation between the total value of urban NTL and GDP and other indicators of human activities shows that the NTL data can reflect the intensity of urban human activities and their changing trend to a certain extent.

5. Discussion

5.1. Regional Difference and Urban Agglomeration Difference

In the middle and later stages of the COVID-19 epidemic, we observed that the changing trend of the total value of night light in urban agglomerations showed a certain regularity. Most of the total value of urban night light in the Yangtze River Delta and the Pearl River Delta city cluster showed a rapid growth trend. In contrast, the total value of urban night light in the Beijing–Tianjin–Hebei city cluster grew slowly. The total value of urban light at night has hardly increased for the northwest urban agglomeration and the northeast urban agglomeration. For the northeast city cluster, Changchun and Harbin are located in the north, and the total value of night light tends to be flat. In contrast, Shenyang and Dalian are located in Liaoning Province. The geographical location is south and near the sea, and the total value of night light increases somewhat, but the growth rate is slow. Therefore, in the recovery stage after the impact of large-scale public health events, the urban resilience of different cities within urban agglomerations shows certain similarities, while there are certain differences between different urban agglomerations. Future studies could explore this difference in more depth.

5.2. The Impact of the Containment Policy

In the initial stage of the COVID-19 epidemic, all cities selected in this study implemented strict containment policies, while in the middle and later stages of the COVID-19 epidemic, due to the remission of the COVID-19 epidemic and the different numbers of infected people in different cities, the containment policies varied from city to city. We collected the number of diagnosed cases per month in 35 large cities in the study area during the three years and explored the relationship between this and the total monthly luminous value of the cities. In the early stage of COVID-19, there was a significant correlation between the number of COVID-19 cases confirmed and the reduction in the brightness value of the night light; strict containment policies are a key factor in cities’ resistance to the impact of COVID-19. In the middle and later stages of the epidemic, there was no significant correlation between the increase in the number of confirmed cases in cities and the change in the total luminous value of cities in the same month. Appropriate containment policies have played a positive role in the adaptation and recovery phase of the middle and late stages of COVID-19 instead. In contrast, large areas and strict containment policies in cities are more likely to cause a decline in the total value of urban luminous value.

5.3. Limitations

This study is from December 2019 to December 2022, which is long and relatively recent. Currently, only free NPP-VIIRS night light images can meet the requirements of this research. However, the spatial resolution of 500 m is rough and does not reflect noctilucent changes at finer scales. Although China’s domestic satellites, such as Jilin-1, can provide spatial resolution images of the research area of better than 100 m, this satellite is a commercial satellite, and the price of high-resolution noctilucent images to meet the needs of this research is astronomical. The monthly composite image used in this study can reflect the variation of noctilucent on the lunar scale. However, the duration between the daily and monthly scales will be neutralized.

Most of the statistical data in this study came from the statistical yearbooks of various cities, but there were some missing items in some cities. We used the statistical bulletin of the official departments of cities to supplement the missing items. Such data may be subject to some bias.

In future studies, nighttime light data with higher temporal and spatial resolution, such as Jilin-1 nighttime light data and SDG-1 low-light band data, can provide more detailed information on urban resilience during large-scale public health events.

6. Conclusions

This paper selected 35 large cities in China as research areas. Based on NPP-VIIRS monthly night light time series images, we explored the changes in the luminous brightness values of different functional places in cities in the early stage of the COVID-19 outbreak and the changing trend of luminous values of each city in the middle and later stages of COVID-19 outbreak. Furthermore, the correlation between the change in luminous value and human, economic, geographical, and policy factors is analyzed. The results are as follows:

(1) In the first two months of the epidemic, the luminous value of large cities showed an extensive range of decline, and the decline of different urban functional places was different. The luminous value of transportation hubs was affected the fastest.

(2) There is a significant positive correlation between urban population and luminous change value. The closer the relationship between urban places and human activities, the stronger the correlation between population and luminous change value of urban places.

(3) In the middle and later stages of the epidemic, the night light value of all cities showed an upward trend, but there was a difference. The mainly developed megacities showed a slower growth rate of night light value. In contrast, the newly emerging large cities with rapid development in recent years showed a more rapid increase in night light brightness value.

(4) The increase in the number of confirmed cases in the middle and later stages of the epidemic could hardly lead to a significant decrease in the value of night light on a monthly scale unless the city had a relatively large area and a relatively strict lockdown policy in that month.

This paper proves the reliability of night light remote sensing images in revealing human activities, and the conclusions obtained can provide reference and guidance for the management decisions of large-scale public health events in large cities and the construction of resilient cities.