Spatial and temporal variations of the PM2.5 concentrations in Hanoi metropolitan area, Vietnam, during the COVID-19 lockdown

ABSTRACT The Vietnamese government has issued several directives of lockdown in Hanoi, the capital city, for a month (1–30 April 2020) to prevent the human-to-human transmission of COVID-19. This action has affected air pollution due to a decline in transportation. Therefore, this study investigates spatial and temporal change in the PM2.5 concentrations during the first 4 months of 2020 in Hanoi metropolitan area. Spatial distribution maps of the PM2.5 concentration in Hanoi were provided for the first time. The average PM2.5 concentrations at the 22 air monitoring stations were strongly correlated with the population. April had a significantly lower level of PM2.5 than the other 3 months. In particular, the concentrations of PM2.5 and NO2 decreased by 12 and 54%, respectively, between March and April, especially at areas for commercial activities. In April, a higher level of PM2.5 was recorded between Tuesday–Thursday, which is a reverse trend with that in March. Furthermore, Monday and Friday did not show rush hour peaks for the PM2.5 levels in April. A decrease in the PM2.5 concentrations was partly influenced by the long-range transport from the outside to Hanoi. This study implies that a reduction in traffic volumes and public activities may possibly improve the air quality.


Introduction
Coronavirus is well known as SARS-CoV 2 or COVID-19.The first patient was found in Wuhan, Hubei, China, in late December 2019, and they are now spreading widely around the world.The virus is one of the significant pathogens that affect human respiratory system [1].Older people and people with underlying diseases have higher mortality rates than young and healthy people.Vietnam is at high risk of  This study investigates the spatiotemporal variation in the PM 2.5 concentrations in Hanoi metropolitan area, Vietnam prior to and during the lockdown required period due to the COVID-19 pandemic in the first 4 months of 2020.In addition, the first spatial distribution maps of the PM 2.5 concentration in the headquarters of Hanoi were also provided.The air parcel movement prior to and during the lockdown was also studied for effects of longrange atmospheric transport in Hanoi.

Study area
According to the statistics of Hanoi population by the General Statistics Office of Vietnam (https://gso.gov.vn) in 2019, Tu Liem, Hoang Mai, Ha Dong and Dong Da districts were classified with the highest population (Figure 1).The four districts are constructed with a high load of offices and residential complexes.In contrast, Tay Ho and Hoan Kiem districts cover West Lake and Hoan Kiem Lake, respectively, with tourism sites (e.g. the Old Quarter); therefore, the total number of residents in these two districts are the lowest.As the capital city, Hanoi is one of largest cities in Vietnam, which include the headquarters of government, education and economy.Hence, people from other cities/provinces tend to migrate to Hanoi to work and study.The COVID-19 pandemic outbreak occurred in the late of March 2020 in Vietnam.The Vietnamese government issued a directive of the lockdown during 1-30 April 2020 to prevent and control the widespread of COVID-19 [2].Therefore, a huge number of people left Hanoi and came back to their hometowns, leading to a solitary capital city.

Data collection
To investigate the air pollution during the COVID-19 lockdown in Hanoi, hourly data for the PM 2.5 concentration during the first 4 months of 2020 were obtained from PAM Air, Vietnam.A network of 22 air monitoring stations has been established by the PAM Air project of D&L Corp., Vietnam since 2019 to monitor the ambient concentration of PM 2.5 in Hanoi metropolitan area.All the air monitoring stations are at urban and residential areas.The full names and specific locations of the 22 stations are summarised in Table 1S (in the Supporting Materials).A PAM Air device (PAS-OA318, D&L Corp., Vietnam) using lightscattering technology was installed at each station to hourly record the concentration of PM 2.5 .All devices are calibrated against TSI DustTrak II after final assembling process.The accuracy of this instrument has been validated and allowed to apply for mega cities in Vietnam.The real-time results are used for Air Quality Index calculation and published online via the official website of PAM Air (https://pamair.org).The sensor device can measure a range of 0-999 µg/m 3 for PM 2.5 with the lowest detectable particulate size of 0.3 µm within less than 10 s under temperature of 0-50 °C and humidity of 5-95% (noncondensing).In addition, ambient temperature (range: 0-65 °C, accuracy: ±0.5 °C and response time: <30 s) and relative humidity (range: 0-100%, accuracy: ±4.5% and response time: <8 s) are also measured by the device.Further information on the device can be found via the official website of PAM Air.
Previous study on air pollution of Hanoi has indicated that the major source of PM 2.5 was vehicular emissions over the entire year [6].Furthermore, nitrogen dioxide (NO 2 ) has been reported to be produced mostly from traffic sources in Hanoi [20].Due to the unavailability of data for vehicle count on roads, daily data for the maximum concentrations of NO 2 (105.84°E, 21.02 °N) were obtained from the World Air Quality Index-WAQI (https://aqicn.org)to evaluate the impact of traffic source on atmosphere of Hanoi during the first 4 months of 2020.Data from WAQI are provided for 380 major cities worldwide and based on air quality monitoring data from each local government.Although data sources of PM 2.5 and NO 2 are different, the monitoring stations for both PM 2.5 and NO 2 are very close to each other and within the metropolitan area.Therefore, data for the concentrations of PM 2.5 and NO 2 may be combined and interpreted together in this study.
Data for ambient temperature and relative humidity were measured spontaneously with the PM 2.5 concentration and also obtained from PAM Air for all 22 stations.In addition, wind data (speed and direction) were downloaded from the National Oceanic and Atmospheric Administration (NOAA, USA) at Noi Bai International Airport (station ID: 488200-99999, 105.81 °E, 21.22 °N) using the worldmet package [21] in RStudio 1.3.1 (https://rstudio.com)based on R programming software 4.0.3(https://www.r-project.org/).Hanoi is located in the Red river delta, one of the largest river deltas of Vietnam, with a flat topography without mountains.Therefore, the wind data at Noi Bai International Airport can be representative for entire Hanoi.All the data for PM 2.5 , NO 2 , and meteorological conditions in Hanoi during the first 4 months of 2020 were combined and adjusted regarding timescale with full measured values for all parameters.

Backward air trajectory and cluster analysis
The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT 5) (https://www.ready.noaa.gov/HYSPLIT.php)using the reanalysis meteorological data (ftp://arlftp.arlhq.noaa.gov/pub/archives/)was applied to simulate the 120-h backward air trajectory arriving in Hanoi (105.82 °E, 21.02 °N).The starting height was 500 m above ground level.The trajectory was obtained hourly for 24 h from 00:00 (UTC+7).A total of 744 (31 d × 24 h) and 720 (30 d × 24 h) trajectories were used for cluster analysis in March and April 2020, respectively.Cluster analysis of trajectories was also performed using HYSPLIT 5 to demonstrate the mean movement of air parcels arriving in Hanoi.The number of clusters was determined based on the variation in the total spatial variance (TSV).A suitable number of clusters was identified with a dramatic change in the TSV [22].

Statistics
All the statistical analysis including descriptive statistics, Shapiro-Wilk normality test, Spearman's correlation, t-test and Mann-Whitney rank sum test were performed using the stats package [23] in RStudio.The temporal variation in the PM 2.5 concentration was investigated using the openair package [24] also in RStudio.

Meteorological condition
Hanoi is affected by the Southeast Asia monsoon regime with highly humid and tropical climate.The northeast monsoon plays a vital role in winter (December, January, and February), whereas summer (June, July, and August) is significantly influenced by the southeast monsoon.According to a previous study [7], atmospheric conditions of Hanoi from January to April are normally poor in dispersion and affected by subsidence temperature inversions.When ambient temperature starts to increase in March and April, northeasterly winds from the inland China in January and February occur less frequently and are replaced by southeasterly winds from the ocean [7].As a result, the air becomes warmer with high moisture.These weather conditions are in line with the meteorological data in the present study.Wind speeds did not seem to change much with an average value of approximately 3 m/s for each month, whereas a light shift from the northeast (January-February) to southeast (March-April) prevailing wind was observed (Figure 1S in the Supporting Materials).Although there is a statistically significant difference among the 4 months (p< 0.01, Mann-Whitney rank sum test) (Table 2S), the meteorological conditions can be categorised into two periods: January-February (winter) and March-April (spring).During the first period, the weather was characterised with relatively low ambient temperature (20 °C) and relatively low relative humidity (79%) (Figure 2).In March-April, the ambient air was warmer (25 °C) and more humid (82%).According to previous studies, an increase in ambient temperature may greatly affect the volatility of solid air pollutants [25], leading to a decrease in particle concentration [26,27].There is also a relationship between relative humidity and absorption of air pollutants into particulate water [28,29].Furthermore, during the COVID-19 lockdown, meteorological conditions have been proven to significantly influence the PM 2.5 variation [30].Therefore, the PM 2.5 concentrations during the COVID-19 lockdown in Hanoi will be evaluated mainly between March (prior to the lockdown) and April (during the lockdown) 2020 to minimize the seasonal effects.

Monthly variation
The average maximum concentrations of NO 2 and the average concentrations of PM 2.5 in Hanoi during the study period are shown in Figure 3. Generally, February showed the highest concentrations of both NO 2 (35 ± 17 µg/m 3 ) and PM 2.5 (62 ± 43 µg/m 3 ), whereas the lowest levels were observed in April for NO 2 (13 ± 10 µg/m 3 ) and PM 2.5 (43 ± 21 µg/ m 3 ).There are statistically significant differences between data in April and those in other 3 months (p< 0.01, Mann-Whitney rank sum test) (Table 2S).Over the first 4 months of 2020, the average PM 2.5 concentration was 51 ± 32 µg/m 3 .This result is consistent with those in previous studies [7,31].During the lockdown, the air pollutants in April decreased by 54% and 12% for NO 2 and PM 2.5 , respectively, indicating that lockdown effects on the air quality were evident in Hanoi.The decrease of NO 2 in Hanoi is much higher than that in Rio de Janeiro, Brazil (2%) [32], and similar to those in Barcelona, Spain (50%) [33] and Milan, Italy (47%) [34].As mentioned in Section 1, traffic emission is the major source of air pollutants in Hanoi.Therefore, this result implies that the COVID-19 lockdown may incredibly reduce the NO 2 emission due to a limitation of vehicles, despite an unavailability of the vehicle count on roads.Regarding PM 2.5 , the reduction in Hanoi is higher than that in Sao Paulo, Brazil (3.6%) [12], comparative to that in Malaysian cities [35], and lower than that in Milan, Italy (40%) [34].
The distribution maps of PM 2.5 in Hanoi metropolitan area during the first 4 months of 2020 were also drawn (Figure 4) to investigate spatial variation of the PM 2.5 concentrations.Throughout the first 4 months of 2020, the average concentrations of PM 2.5 at the 22 air monitoring stations were significantly positively correlated with the population (Spearman's correlation, r = 0.249, p< 0.05).Regarding monthly variation, a sharp decline in spatial levels of PM 2.  the decrease in traffic emissions and/or PM 2.5 at these three stations is less than that at stations 11 and 20.Stations 6, 10 and 21, which are governmental office and residential areas, are the hotspots of PM 2.5 .A reduction of ~5 µg/m 3 in PM 2.5 between March and April at these 3 stations are negligible.

Daily and hourly variation
Stations 11 (Doi Can) and 20 (Ba Trieu), with the most reduction in PM 2.5 , were selected to investigate the temporal variation of the PM 2.5 concentrations between March (prior to the lockdown) and April (during the lockdown), 2020.Additionally, station 22 (Hang Quat) was also used to compare the temporal variation in the PM 2.5 levels among stations.Figure S2 demonstrates that the PM 2.5 reduction between March and April was apparent regardless of sampling station type.In particular, a decrease in the PM 2.5 concentration by 4.3% was observed at a tourism site with handicrafts (Hang Quat), whereas those at urban sites with high load of stores, restaurants, and offices were 33% (Ba Trieu) and 44% (Doi Can).
Figure 5 illustrates the change in daily and hourly variations in the PM 2.5 concentrations prior to (March) and during the lockdown (April) at Doi Can, Ba Trieu, and Hang Quat.In March, a higher level of PM 2.5 was observed from Friday to Monday (of the next week).As mentioned in Section 2.1, Hanoi is a home to a large number of people from all provinces/ cities in Vietnam.Therefore, those people normally visit their hometowns on Friday and Saturday, then return to Hanoi early on Monday to start working.This pattern of PM 2.5 (Figures 5 and 3S) is different from that in Helsinki, Finland [36] and New York, USA [37], which partly suggests a difference in resident behaviour between developing and developed countries.In developed countries, people may tend to stay at home during weekends, leading to a reduction in vehicular movement on roads.Regarding hourly variation, PM 2.5 at Doi Can showed an unclear trend with a fluctuation around 43 ± 26 µg/m 3 (Figures 5(a) and 3Sa).On the contrary, PM 2.5 at Ba Trieu and Hang Quan had a clear bimodal distribution with two peaks during morning rush hour (06:00) and late at night (20:00-21:00) (Figure 5(b, c) and 3Sb, 3Sc).A similar diurnal variation in the PM 2.5 levels is also reported in Nanjing, China [38].In the afternoon, the development of convection due to increasing ambient temperature may increase the height of planetary boundary layer, leading to a favourable condition for both vertical and horizontal dispersion of air pollutants; hence, the concentration of PM 2.5 near ground level is the lowest within a day [36,39].
During the COVID-19 lockdown in April, daily and hourly variations in the PM 2.5 concentrations showed a different pattern at these 3 stations.At Doi Can (Figures 5(a), 2S and 3Sa), the concentration of PM 2.5 seemed to be relatively constant every day (24 ± 4 µg/m 3 ).However, a small peak of PM 2.5 appeared right before the afternoon rush hour (15:00-16:00).On the other hand, the PM 2.5 variations at Ba Trieu and Hang Quat were similar (Figure 5(b, c) and 3Sb, 3Sc).In particular, the PM 2.5 concentrations were higher during Tuesday-Thursday than Friday-Monday, which is an opposite trend in March.Although the average concentrations of PM 2.5 at Ba Trieu and Hang Quat were lower in April (35 ± 19 and 45 ± 18 µg/m 3 ) compared with those in March (52 ± 33 and 47 ± 20 µg/m 3 ) (Figure S2), respectively, there was no change in hourly variation of the PM 2.5 concentrations.This change in the PM 2.5 variation at these 3 stations partly indicates that the residents in Hanoi tended to spend time at home and not visiting hometowns at weekends during the implementation period of the COVID-19 lockdown.

Air parcel movement
Previous studies in Hanoi have indicated that the high concentrations of PM 2.5 are significantly affected by wind speed and long-range atmospheric transport [7,8].A higher load of air pollutants is mainly transported to Hanoi with air parcels originating from the northnortheastern region other than those from the southwestern region [8].To investigate the movement pathway of air parcels arriving in Hanoi prior to (March) and during the COVID-19 lockdown (April), 120-h backward air trajectories starting hourly at 500 m above ground level were simulated using HYSPLIT 5. A survey on the change in the TSV shows that all the trajectories can be represented by 5 clusters in both March and April (Figure S4).Overall, 3 of 5 clusters are relatively similar between March and April (Figure 6).The major pathway of air parcels in March (M1: 32%) and April (A2: 43%) was from the east-northeastern region passing through the China Sea to the East Sea and several southeastern provinces of China (Guangdong, Fujian, and Zhejiang) (Figures 6 and 5S).This result is consistent with that reported in a previous study [8].The other 2 similar clusters between March and April originated from the Philippines, the South China Sea, and the East Sea (M5: 26% and A1: 13%) and from Cambodia, Thailand, and Laos PDR (M3: 21% and A5: 14%) (Figure 6).Furthermore, the heights of M3 and A5 were both very low (<500 m).The major difference between March and April is height and direction of the other 2 clusters.Except for M3 and A5, the other 4 cluster in March (M1, M2, M4 and M5) were under 500 m and lower than those in April (A1, A2, A3 and A4) with a height of ~1000 m (Figure 6); especially, A4 originated at ~6000 m.In March, M4 passing through the western region (Northeastern India, Myanmar, and Laos PDR) and M2 originating from the northeastern region (China-Hubei, Hunan, and Guangxi) contributed by 7% and 14% as minor clusters, respectively (Figure 6(a)).Meanwhile in April, the cluster from the western region disappeared and the cluster from the northeastern region moved towards the north and was divided into 2 clusters (A3: 23% and A4: 6%) (Figure 6(b)).In particular, A3 (China-Hubei, Chongqing, Guizhou, and Guangxi) was much shorter and lower than A4 (Mongolia and China-Inner Mongolia, Shaanxi, Hubei, Chongqing, Guizhou, and Guangxi).Higher levels of particles are frequently observed within the lower troposphere due to gravitational forces.Another published study in Hanoi have indicated that the concentrations of air pollutants near the Earth surface were controlled by the planetary boundary layer [40].Air parcels moving above the planetary boundary layer may be prevented to transport the air pollutants from distance sources to receptor site.Therefore, air parcels arriving in Hanoi from higher troposphere may partly explain for a lower concentration of PM 2.5 during the lockdown in April.

Conclusion
In this study, a change in the PM 2.5 concentrations during the COVID-19 lockdown (April 2020) in Hanoi, Vietnam was evaluated.The results indicate that the PM 2.5 concentrations decreased compared with those right before the lockdown (March 2020).Daily variation in the PM 2.5 concentrations shifted from high on Friday-Monday (March) to high on Tuesday-Thursday (April) with almost no change in hourly distribution during the lockdown.Peaks of the PM 2.5 level in rush hours disappeared on Monday and Friday in April.Effect of prevention measures to control COVID-19 spread in community, e.g.restrictions of vehicles and public activities, was evident on the air quality by a significant reduction of NO 2 .Additionally, long-range transport from the upper troposphere of neighbouring countries also affected the decrease in the PM 2.5 concentrations in Hanoi.Further studies are necessary to deeply investigate and differentiate effects of COVID-19 lockdown and other driving factors, e.g.meteorological conditions or long-range transport, on variations of the PM 2.5 concentrations.This study also infers that controlling vehicles on road may reduce the PM 2.5 pollution and can be considered a suggestion for sustainable development.

Figure 1 .
Figure 1.Location of the 22 air monitoring stations and population (in 2019) of the 12 central districts in Hanoi, Vietnam.

Figure 2 . 3 )Figure 3 .
Figure 2. Box plots for the average (a) ambient temperature and (b) relative humidity at the 22 air monitoring stations during the first 4 months of 2020 in Hanoi, Vietnam.
5 is obvious in Ba Dinh (stations 11 and 7), Hoan Kiem (stations 17 and 22), and Hai Ba Trung (station 20) districts in April compared with those in other 3 months.Stations 11 and 20 are concentrated with a high load of commercial offices.During the lockdown, these offices were closed, and working-at-home policy was applied, leading to a decrease in the PM 2.5 concentrations from 43 and 52 µg/m 3 in March to 24 and 35 µg/m 3 in April at stations 11 and 20, respectively.Stations 7, 17 and 22 are located in the Old Quarter, one of the largest tourism areas of Hanoi, also showed slight reduction of thePM 2.5 levels from 46, 51, and 47 µg/m 3 in March to 41, 45, and 45 µg/m 3 in April, respectively.Normally, vehicular movement is restricted by the local government during night-time for the Night Market and the Walking Street inside the Old Quarter; therefore,

Figure 4 .
Figure 4. Spatial distributions of the PM 2.5 concentrations in the 12 central districts of Hanoi, Vietnam during the first 4 months of 2020.Four contour maps were drawn using ArcGIS 10.8 (ESRI Inc., USA) and inverse distance weighting (IDW) was applied as an interpolation method.

Figure 5 .
Figure 5. Daily and hourly variations in the PM 2.5 concentration (mean and 95% confidence interval) during March and April at (a) Doi Can, (b) Ba Trieu, and (c) Hang Quat stations in Hanoi, Vietnam.

Figure 6 .
Figure 6.Clusters of backward air trajectories arriving in Hanoi, Vietnam (black star) during (a) March and (b) April.