The utilization of satellite data is expanding to observe social and economic changes. In combination with artificial intelligence or machine learning, satellite data are used to observe and analyze changes in countries’ and communities’ socioeconomic conditions [25, 26]. Such data was recently used in the Russo-Ukrainian War to observe the transport of war equipment, the location of tanks, and the movement of combat units [27]. In particular, night-time light (NTL) satellite data are becoming increasingly useful for demonstrating economic progress, external economic shocks, and economic activities [28,29,30,31,32]. Compared to other data on North Korea, the changes in night-time light intensity captured by satellite imagery provide a more objective indicator of the country’s economic activity.

NTL data refers to data generated by the Visible Infrared Imaging Radiometer Suite Day-Night Band aboard the Suomi-NPP satellite of NASA and the National Oceanic and Atmospheric Administration [33]. The Earth Observation Group(EOG) provides monthly and yearly night illumination light images taken by satellites, and the image resolution is 15 arc seconds (approximately 500 m). There are two types of monthly files provided by EOG, vcm, which is a file that deletes the part affected by stray light, and vcmsl, which is a file that corrects it. In this study, we aimed to utilize as much information as possible by using vcmsl files that were corrected for stray light artifacts. In the GeoTiff file, avg_rade9h, which represents light information, is used; the unit of light is nW/cm^2/sr. GIS software was used to extract the light information from the downloaded light image files.

We aimed to observe the changes in North Korea-Russia and North Korea-China trade after the border blockade; data were organized around trade cities located in the border area. Chongjin City and Rason City are North Korea-Russia regions, and Sindo, Sinuiju City, Ryongchon, Manpho City, and Kanggye City are North Korea-China regions. Figure 2 shows the location of each region. Kanggye, while not situated at the border between North Korea and China, serves as a hub for trade with China through its railway connection to Manpo. Similarly, Chongjin, not located at the border between North Korea and Russia, is recognized as a significant trade city, not only for its land-based trade connections through Rason City, but also for its potential for maritime trade. Thus, these cities were categorized and included in the analysis based on the group division.

Location of North Korea-China and North Korea-Russia border trade cities. Note: The above figure is shown using the boundary map of North Korea’s administrative districts provided by the UN, Open Street Map, and World Map, and built into QGIS. The regional boundary map of North Korea utilized a file from the United Nations Office for the Coordination of Humanitarian Affairs (UNOCHA), which can be accessed through the link https://data.humdata.org/dataset/cod-ab-prk

An analysis was conducted targeting locations wherein lights appeared before border closure. In North Korea, electric energy use is sparse; only approximately 26% of the total population has access to electricity [34]. In addition, the aging and inefficient production methods of plants that produce electricity worsens the situation [35]. In this regard, we assumed that the locations of the light observed in North Korea’s trading cities in 2019 are where light is essential; thus, by analyzing the changes in light in these locations, we analyzed the change in trade activities after the border closure. The area where lights occurred was filtered using the 2019 lit mask v2.1, provided by the EOG [36]. This mask file provides the location of cells where light occurred in 2019; therefore, data in cells in areas with no light were treated as zero values. Table 1 shows the number of cells in each area and the statistics of the light intensity.

After the masking process using the lit mask file in 2019, 5,397 cells with light were identified, and 53,970 cells were used, corresponding to 10 months of data collection (October 2019 to July 2020). The number of cells was highest in Pyongyang (2,494), followed by Sinuiju (584) and Chongjin (553).

Compared with other cities, Pyongyang and Sinuiju had higher light intensities. This seems to be influenced by the cities’ roles. Sinuiju is a trading city where transactions between North Korea and China are the most active; it is an area with a transportation and logistics system that can deliver goods to other regions [37]. As Pyongyang is the capital of North Korea, where the country’s key government agencies are located, the lights appear strong.

As North Korea does not have official data on economic indicators, it is necessary to compare and analyze various data from other sources. Accordingly, we explained the changes post-border closure through a comparative analysis involving trade data and the changes in the lights of trade cities in North Korea. We used trade data between North Korea and China disclosed by the General Administration of the Customs People’s Republic of China [38]. Data provided by the UN Comtrade and the Korea International Trade Association (KITA) were used as trade data between North Korea and Russia. Figure 3 shows the change in trade value between North Korea-China and North Korea-Russia.

North Korea’s trade value with China and Russia. Note: The graph above was created using data from the General Administration of the Customs People’s Republic of China, UN Comtrade, and KITA

North Korea’s trade continued to deteriorate in 2016 and 2017 as the UN’s economic sanctions tightened [39]. Although trade has gradually recovered since 2018, it experienced a sharp drop when border closure was implemented. Subsequently, North Korea’s trade with China decreased significantly, while its trade with Russia increased. Although the trade value increased as the number of confirmed COVID-19 cases in China decreased, it was considerably lower than the trade value before the border closure.

After the declaration of the COVID-19 pandemic, North Korea recommended restricting the movement of residents nationwide and implemented border closures. In other words, overall economic activities within North Korea are being influenced by orders for movement control. This study aims to observe changes in North Korea’s trade activities resulting from the border closure. To observe the additional impact on specific regions due to the border closure, beyond changes in economic activities resulting from movement control, a comparative analysis method between regions is necessary. To analyze these additional impacts, the assumption is required that the effects of movement control are uniformly felt across all regions of North Korea. Given the absolute authority and strong enforcement of orders by North Korea’s leadership during the COVID-19 pandemic, this study deems the application of such an assumption appropriate. Therefore, in this study, the difference-in-differences (DID) is used for regional comparative analysis to analyze the impact of the border closure on economic activities in the border regions.

To confirm the changes in the movements of North Korean trade cities located along the border after the border closure, we compared the changes in night-time lights before and after the official border closure declaration by the North Korean government. On January 22, 2020, North Korea closed the border; therefore, we chose January 2020 as the start of the border closure period. STATA 17 was used for analysis [40]. The light change was estimated using the following difference-in-difference (DID) regression model.

$$_=_+\sum\nolimits_\sum\nolimits____+_+_+_,$$

where \(_\) represents logarithm values after adding 1 to the light value at cell \(i\), in area \(g\), and month \(t\). As there are many values less than 1 in the light intensity of the cell unit, taking the logarithm directly of the light intensity of the cell unit results in many negative values. Therefore, 1 was added to the light value and logarithm was taken. \(Area\) is a dummy variable that takes the value of 1 if the cell is in the corresponding region group and 0 otherwise; it is divided into four groups: North Korea-China Trade City, North Korea-Russia Trade City, Pyongyang, and Sunchon City. \(Month\) is a dummy variable that is assigned a value of 1 if the data belongs to that month, and 0 otherwise. Sunchon City was selected as the control group because it is located in an inland area (not a border area), and it showed little change in lighting before and after border closure. As North Korea decided to close the border in January 2020, the event time point (reference month) was set at t = 4 in this study. \(_\) is the time fixed effect that controls for common time trends. \(_\) is the cell-fixed effect, which controls for time-invariant heterogeneity in light intensity within each cell. \(_\) is an error term. \(_\) indicates differences in the logarithmic value of (light intensity + 1) in area \(g\) relative to those in Sunchon City in month \(t\) and compared with the difference between the two area groups in the reference month. The purpose of this study is to analyze changes in economic activity by observing the difference in night-time lights between cities in the border areas affected by the border closure policy and those that are not. It is difficult to directly prove a significant relationship between the total trade volume of North Korea and the night-time lights in border cities, and due to the limitation of the lack of additional data that can affect the night-time lights in border cities, this study could not include other explanatory variables. To overcome these limitations, this study compared the direction of changes in the flow of night-time lights with known changes in North Korea’s trade.