Effects of COVID-Induced Mobility Restrictions and Weather Conditions on Air Quality in Hungary

Abstract

Similarly to other countries, the first wave of the COVID pandemic induced a collapse of mobility in Hungary during the spring of 2020. From the environmental perspective, the obtained road traffic reduction of 20–50% could be regarded as an undesired traffic regulation experiment. Air quality impacts within Hungary were evaluated based on data from 52 monitoring sites measuring concentrations of pollutants NO_x, O_3, and PM₁₀. Air pollution during the lockdown was compared to the same period (February–June) in the reference years 2014–2019. The large spatial heterogeneity of the air quality response was explored. The emission reduction coincided with the extreme weather of 2020, characterized by unusually warm pre-lockdown February and spring drought. The anomalously low pre-lockdown air pollution was further reduced (NO_x) or increased (PM₁₀) during the restrictions. Compared to the previous years, NO_x concentrations during the curfew were found to differ between −4.1 and +0.2 standard deviations (median −1.55 SD), or −45% and +3% (median −18%) among different monitoring locations. Ozone concentrations were unusually high due to both weather and chemical reasons (median +11% or +0.8 SD), while the PM₁₀ response was modest and largely weather-driven (median +7% or +0.4 SD).

1. Introduction

The global response to the COVID pandemic brought an undesired air pollution reduction experiment throughout the world. The scientific potential of this unfortunate environmental experiment has been exploited worldwide by intensive research on the COVID-air quality (COVID-AQ) response. Early reports of urban NO_x concentration reduction, reaching 40–50% compared to the previous years, have been published from the first COVID hotspots, such as Wuhan [1], Milan [2], and Madrid [3]. Meanwhile, a general increase in ozone levels has also become evident, attributable to the “weekend effect”, or, in this case, the more pronounced “lockdown effect” of NO_x–O₃ chemistry [4,5]. After the first wave of the pandemic, air pollution reduction has been reported from numerous locations worldwide [6,7,8,9,10,11], and air quality was soon identified as a mediating factor between mobility reduction and COVID cases [12,13,14]. However, systematic investigation of global measurements revealed that the air quality improvement was not necessarily observable for every location in their respective lockdown period, and, notably, the Central-Eastern European region suffered a major increase in PM_2.5 concentrations during the lockdown [8]. Furthermore, a large-scale transport episode was demonstrated, causing adverse changes in air pollution in the lockdown period in Hubei, China [15]. Especially particulate matter (PM) concentrations showed a very heterogeneous global and regional response, ranging from a decrease of 40–60% [16] through near-stagnation [7,17,18,19] to 20–30% increase [20]. In general, China and India experienced major improvement in both NO_x and PM levels [16], while the European and American response was more modest and restricted to NO_x at most locations [4,6,7,10,18,20,21]. The lack of general PM reduction in Europe was attributed to the larger contribution of non-traffic emissions, including seasonal agricultural activities, and also to weather impacts [21], while meteorology was judged to be only a secondary factor in China [9].

Three main approaches exist for the selection of the reference period: (a) the pre-lockdown period [3,7,8,16], (b) the lockdown period in the previous years [6,10,17,20], and (c) model estimates for the lockdown period [20,21,22,23,24,25]. Other efforts to overcome these issues include weather analogies [2,26], comparing changes in air quality to changes in boundary layer height [10], and time series breakpoint detection [18,25]. The optimal method for each case depends on the available data and models as well as the regional characteristics. This, on the other hand, challenges the comparison of different results and the synthesis of a global perspective.

This study aims to contribute to the exploration of the COVID-AQ response from the Hungarian perspective. Previous research [10] showed a decrease in NO_x concentrations in the city area of the Hungarian capital city, Budapest, while PM levels were not considerably affected and ozone concentration slightly increased. This study explores the regional heterogeneity of the COVID-AQ response by using air quality data from urban, suburban, and rural monitoring sites in Hungary.

2. Background and Data

2.1. The Study Region

The basin formed by the Carpathian Mountains creates adverse conditions for air pollution, and stagnation episodes often occur, especially in the winter [27]. Concentration guidelines defined by the European Environmental Agency (EEA) are regularly exceeded for NO_x and PM₁₀ [28], and Budapest was found to be among the most affected European cities by air pollution health risks [29]. In 2018, 34% of annual domestic NO_x emissions were attributable to passenger cars, 6% to other transportation, and 26% to agriculture [28]. However, the contribution to primary PM_2.5 emissions is only 6% from transportation and 3% from agriculture, faded by the dominant residential emissions (82%). Furthermore, a study has shown that approximately ¾ of annual PM air pollution in Hungary is of transboundary origin [30], although serious winter air pollution episodes are caused by local sources [31]. It must be noted that the seasonality of the presented ratios is significant. Residential heating emissions dominate the winter. Agricultural emissions peak in early spring (through nitrogen fertilizers and dust suspension from bare soils), as well as during the late autumn burnings.

2.2. Data and Method

Air pollution monitoring data was obtained from the Hungarian Air Quality Monitoring Network (OLM) for the period 2014–2020. Observations from 52 fixed monitoring sites were available, 12 of which were located in the capital city of Budapest (Figure 1). Daily mean concentrations of PM₁₀, NO_x, and O₃ were generally available for investigation. All gas analyzers and dust monitors used in the monitoring network are based on reference methods or standards. PM₁₀ is measured at each measuring site by Grimm EDM 180 (optical), or Environnement SA MP101M (beta attenuation), or Thermo Scientific FH62 C14 (beta attenuation) sensor. NO_x measurements are carried out Thermo Scientific 42i, or Environnement SA AC32M, or Teledyne API T200 sensor, while O₃ is measured by Thermo Scientific 49i, or Environnement SA O3 42M, or Teledyne API T400 sensor at each site. Maintenance of all sensors is provided by the Air Quality Reference Centre of Hungarian Meteorological Service, which includes periodic calibrations in reference laboratory based on EU directives and Hungarian standards (for more details, see: http://levegominoseg.hu, accessed on 18 April 2021). The air quality monitoring network of OLM covers the majority of major Hungarian towns. Station IDs referred further in the text are shown in Figure 1, while their names and other characteristics are enlisted in Appendix A.

Due to the strong seasonality of air pollution in Hungary, air pollutant concentrations in 2020 were compared to those obtained in the same period of 2014–2019. The difference between the mean concentration of 2020 and the respective means in previous years was expressed in terms of the z value, i.e., relative to the sample standard deviation of 2014–2019.

$$z=\frac{{\mu }_{2020}-{\mu }_{2014-2019}}{{\sigma }_{2014-2019}}$$

(1)

where μ is the mean concentration for the respective period and σ is the sample standard deviation (SD) of the mean concentrations among the 6 previous years. If air pollutant concentrations of the same period over consecutive years were normally distributed, z = −1.96 would mark a significant decrease in air pollution at a 95% level, compared to the previous years, while z = −0.67 would sign the 50% significance level.

2.3. Traffic Reduction in the Study Period

Two periods were defined for the investigation: 12 March–16 June and 28 March–4 May, named “emergency” and “curfew”, respectively (Figure 2). A general estimate for countrywide traffic reduction can be obtained by the fuel consumption, reduced by 18–19% in April–June compared to the same quarter of 2019 (data from the Hungarian Petroleum Association). The Google Community Mobility Reports (https://www.google.com/covid19/mobility/, accessed on 18 April 2021) obtained a mean mobility reduction of 50% and 40% at transit stations in the curfew and emergency period, respectively. For more accurate fixed-point observations, camera-based traffic density monitoring data was obtained from the Budapest road network operator (Figure 2 and

Table 1. COVID-related traffic reduction at different road cross-sections of Budapest.

Period	Suburban (Árpád híd)	Urban (Erzsébet híd +Petőfi híd)	City (Szabadság híd)
Public emergency (12 March–16 June)	–26% 1	–38% 1	–27% 1
	–26% 2	–36% 2	–30% 2
Curfew restrictions (28 March–4 May)	–38% 1	–49% 1	–44% 1
	–37% 2	–48% 2	–47% 2
Lowest 7-day mean	–49% 1,3 –41% 1,4	–58% 1,3 –53% 1,4	–52% 1,3 –48% 1,4

¹ related to the mean of the pre-lockdown period (February 2020). ² related to the same period in 2019. ³ 11–17 April (the Easter holiday period with approx. 20% traffic reduction in a normal year). ⁴ 29 March–4 April (lowest 7-day mean traffic in 2020 excluding the Easter holiday). Data from Budapest Public Roads Company (Budapest Közút).

). The public emergency initiated a rapid collapse of mobility in mid-March and ended with a slow, gradual rearrangement during May and early June. The urban road traffic stabilized at 50–70% of pre-lockdown intensity for the curfew period (Figure 2 and Table 1). Dates marked in Figure 1: 4-March: first COVID case in Hungary; 12-March: public emergency, partial lockdown affecting schools, tourism, culture and administration, stay-at-home campaign; 28-March: general curfew restrictions; 4-April: lifting parking fees to unburden private car usage for essential travels; 12-April: Easter holiday; 4-May: gradual reopening begins in Hungary; 16-May: gradual reopening begins in Budapest; 16-June: public emergency withdrawn.

2.4. Meteorological Conditions in the Study Period

The first wave of the pandemic coincided with a spring drought affecting the Carpathian Basin. The pre-lockdown in February 2020 was extremely warm with a monthly mean temperature of 5 °C (climatological mean 0.5 °C between 1981–2010). Freezing temperatures were caused by late March cold fronts (21 March, 30 March) transporting very dry airmass of polar origin. On 26 March, the easterly current transported dust to the southern part of Hungary from the deserted region near the Aral Sea [32]. Precipitation in April was limited to a few days and remained under 20 mm for most of the country or under 5 mm in the northwestern region (climatological mean 35–40 mm). No precipitation occurred in Hungary between 31 March and 11 April, almost exactly overlapping with the first two weeks of the curfew. In May, several cold fronts brought lower temperature (monthly mean temperature 14 °C, climatological mean 16 °C) but caused only a small amount (20–50 mm) of precipitation. Meanwhile, precipitation during May in the central region of Budapest remained under 20 mm (climatological mean 60 mm). Saharan dust arrived above Hungary between 13–15 May. (Source of the presented data is the Hungarian Meteorological Service).

3. Results

3.1. Evolution of Air Quality during the Lockdown

To assess the regional background air pollution impacted by the domestic as well as other European restrictions, the time series of NO, NO₂, O_3, and PM₁₀ pollutants are first presented at three rural sites near the country border (Figure 3). NO₂ levels were lower than usual; however, an important observation is that this anomaly had also been present for weeks pre-lockdown. Unusually warm late-winter induced clean air in terms of both NO₂ and PM₁₀, while, accordingly, late-winter ozone concentrations were among the highest in recent years. In the first weeks of the lockdown, NO₂ background slightly declined, while ozone and especially PM₁₀ concentrations rose. The PM₁₀ concentration time series are extraordinary such that concentrations in April (during the curfew) were higher than in the heating season February, an extreme behavior caused by the unusually warm and dry period during these months. Peaks of a Central Asian dust transport event, coinciding with the onset of the curfew period, were also identifiable in the PM₁₀ time series. In the gradual reopening phase of May, PM₁₀ background concentrations reduced to (but not below) the multi-year range, while NO₂ and O₃ levels remained lower and higher than usual, respectively. Unique photochemical behavior was found at station #41 (Sarród), marked by elevated NO levels and corresponding low ozone concentrations. This site, although defined as a background station located in a natural environment, might have been affected by the plume of the city of Vienna (Austria), located 75 km to the northwest. High ozone concentrations can be partly attributed to the lack of NO titration at site #23 and in May at site #45. However, site #45 experienced higher than usual concentrations of both NO and O₃ during the curfew indicating high photochemical activity. While this can be explained by the extremely dry and sunny weather in April, the different NO responses among rural sites underline the heterogeneous photochemical response. The inverse response of PM₁₀ levels to the lockdown was clearly weather-driven. Both background O₃ and PM₁₀ levels were higher during the lockdown than before and after, while NO₂ maintained its low late-winter background without any major response on restrictions.

The urban impact of traffic reduction is demonstrated through the example of Pécs, a town of 145,000 inhabitants in Southern Hungary with three air quality monitoring sites (Figure 4, sites #34–36 in Figure 2). The sharp decline of NO_x concentrations at the onset of the public emergency marks the airmass change caused by a cold front, but a gradual degradation of NO₂ concentrations was observable during the curfew. Comparison of sites #34–36 provides insight into urban photochemistry. The roadside location (#34) experienced a major reduction of NO levels at the onset of the restrictions, coupled with the corresponding increase in ozone concentrations. Meanwhile, at the urban background site (#36), both NO₂ and ozone levels were lower than usual, with a relatively high NO/NO₂ ratio. The magnitude of inverse chemical response of ozone was not uniform: O₃ concentrations during the curfew were high in the city center as well as in the suburbs, while the urban background ozone level was lower than usual. The urban NO_x–O₃ time series was clearly different from the rural values, reflecting changes in local emissions, but the behavior of PM₁₀ mirrored the background (Figure 2 and Figure 3). The eastern dust transport in the first days of the curfew was especially strong in this southern area of the country, causing nonattainment of the 50 µg/m³ EEA guideline coinciding with the annual minimum of traffic density.

3.2. Comparison of Nitrogen Oxide and Ozone Levels to the Reference Years

The comparison of nitrogen oxide (NO_x) and ozone (O₃) concentrations of 2020 compared to the same period in 2014–2019 at all available locations are shown in Figure 5 and Figure 6 and in Appendix B and Appendix C. At towns with multiple monitoring sites, urban/suburban responses could be distinguished. Roadside NO_x concentration reductions between 21–33% or 1.6–4.1 SD in the curfew period were observed at roadside city locations (#14, #28, #22, #36, Figure 4), coupled with the decrease in NO/NO₂ ratio and increasing ozone concentrations. This shows that the primary impact of traffic reductions was similar among the four towns. However, the response of urban/suburban background pollution was far from being uniform. Among suburban sites, there were examples of significantly decreasing (#29, #36), increasing (#16, #35), and stagnant (#21) ozone concentrations, while corresponding suburban NO_x changes ranged between −45% and +3%. This highlights the different regional chemical responses to traffic reduction. Other cities in Hungary with a single monitoring site have also shown different characteristics: significant NO_x reduction with a corresponding increase in ozone concentration, significant NO_x reduction with no increase in ozone levels, and no NO_x reduction. NO/NO₂ ratio generally increased or stagnated throughout the country, except for roadside and a few urban stations.

In the capital city, Budapest, with 1.7 million inhabitants and dense traffic, the roadside city location (#12) experienced a 44% = 2.5 SD reduction of NO_x concentrations during the curfew period with nearly stagnant ozone levels. Other urban and suburban background sites showed a 0–18% decrease in NO_x pollution. Meanwhile, a considerable (4–63%) increase in urban/suburban background ozone pollution was experienced as a coupled effect of traffic reduction and weather. As expected, the amplitude of the response in both directions was much higher in Budapest than in the rest of the country: the temporal mean, spatial median NO_x concentration in the curfew period decreased from 37 to 30 µg/m³ (−28%) at locations in Budapest, while only from 22 to 21 µg/m³ (−7%) in the rest of the country, compared to the 2014–2019 reference years. Meanwhile, spatial median ozone concentration increased from 50 to 64 µg/m³ in Budapest (+24%), but only from 61 to 67 µg/m³ (+8%) in the rest of the country.

In conclusion, the NO_x–O₃ response to traffic reduction seems to be strongly region-specific. Ozone reduction was mostly achieved in Western Hungary despite the positive weather influence, while most of the country and most notably the capital city Budapest experienced unusually high ozone levels during the curfew with large regional and intra-urban variability.

3.3. Comparison of Particulate Matter Pollution to the Reference Years

As a superposition of emission and weather anomalies, the COVID-PM₁₀ response was modest and very heterogeneous. The comparison of PM₁₀ concentrations of 2020 compared to the same period in 2014–2019 are shown in Figure 5 and Figure 6, and in Appendix D. In general, the spatial variability of PM₁₀ is lower, but the inter-annual variability is higher than those of gaseous pollutants, and the overall source contribution of traffic is low. Thus, the spatial structure of the response is more relevant, and no considerable difference of PM₁₀ concentrations from the reference period 2014–2019 was found. It must be noted, however, that no significant difference from previous years means increasing concentrations compared to the pre-lockdown period because of the unusually clean pre-lockdown weeks presented in Section 3.1.

The capital city, Budapest, had 11 monitoring sites available for PM₁₀ comparison. Two roadside sites (#10 and #12) gained a PM₁₀ concentration reduction of 21% and 12%, respectively. Although these values differ with only 0.7 SD from the reference mean, such reduction is unseen at most locations throughout the country and also at roadside monitoring sites in smaller towns. Suburban monitoring sites in Budapest showed a wide range of PM₁₀ responses between −18%–+28% or −0.9–+1.7 SD in the curfew period, showing that intra-urban variability of particulate pollution was large. However, it is spectacular that of the 11 monitoring sites in Budapest, six had lower PM₁₀ concentrations in the curfew period than the mean of the same period in 2014–2019. In the rest of the country, only 8 of 34 locations had reduced PM₁₀ pollution. On the other hand, the temporal mean, spatial median PM₁₀ concentration among monitoring sites within Budapest was 24 µg/m³ in 2014–2019 and 29 µg/m³ in 2020. For the rest of the country, the spatial median of 22 µg/m³ in the reference period increased to only 23 µg/m³ in 2020. The overall conclusion is ambiguous: in terms of PM₁₀ pollution, the capital city, Budapest was the only location where traffic reduction could dominate over adverse weather to reduce PM₁₀ pollution at some (roadside) sites; however, the whole of Budapest became more polluted relative to smaller towns than in usual years.

4. Discussion and Conclusions

In the first half of 2020, two parallel extremes affected Hungarian air quality: a weather extreme manifesting in an unusually warm late winter and a spring drought, and an emission extreme caused by the COVID pandemic.

The general response to these acting factors in the spring of 2020 was a reduction of NO_x concentrations beginning weeks before the lockdown and reaching an amplitude comparable to the decrease in traffic during the pandemic. A median reduction of 16% was experienced for monitoring sites in Budapest and 21% for the rest of the country during the curfew, compared to the same period in previous years. The maximum roadside reduction in the city of Budapest reached 44% in the curfew period, which fits well within the 32–50% range obtained for U.K. roadsides [18], although it lags behind the reductions of 50–65% in the Mediterranean [3,4]. Ozone concentrations were anomalously high because of both weather and chemical reasons. A median increase of 24% in Budapest and 8% in the rest of the country was found during the curfew, compared to previous years. Similarly, a 20% increase was found in the U.K. [18] and 2–27% in the Mediterranean [4]. It must be noted, however, that obtained air pollution reductions should not be strictly compared to other reports without accounting for weather impacts and different methodology [33].

PM₁₀ concentrations were mainly weather-driven, with a strong negative anomaly in the pre-lockdown period, followed by a sharp increase at the onset of the pandemic. Although the moderation of this increase might be attributable to the traffic reduction at a few roadside locations, no overall pandemic-related changes in PM₁₀ concentrations could be demonstrated. The lack of considerable reduction in particulate concentrations was reported from the U.K. [18] and generally from Europe [21]. The latter study suspected heating and agricultural emissions behind this finding. Rodríguez-Urregos have already noticed that Budapest, as well as nearby capitals Bratislava, Vienna, and Prague, had surprisingly increasing PM_2.5 concentrations during the lockdown due to an unknown reason [8]. The reason can be identified in the drought affecting this region during the curfew, as well as the unusually clean pre-lockdown period used as a reference for their comparison. Large-scale transport events also occurred, most notably the major dust transport episode on 23–27 March 2020 [32], causing poor air quality throughout the Balkans and Southern Hungary. Shen et al. provided a similar example from China for a dominant dust transport event coinciding with the lockdown [15]. Our findings confirm the results of a previous study [10] on COVID-related air quality changes in the city of Budapest, reporting a 22–39% reduction of roadside NO₂, 11–80% increase in ozone levels and no considerable change in particulate concentrations.

Recognizing the serious weather dependence of short-term changes in air quality, Gkatzelis et al. highlighted the importance of weather correction when estimating COVID-AQ response to obtain comparable results among different locations and periods [33]. Meanwhile, they found that the majority of reviewed research did not apply a quantitative correction for meteorology due to limitations of available data and models, while others applied a wide variety of methods [33], including a range of statistical [22] and dynamical [21] models. Reliable statistical weather correction in the Hungarian case is challenged by the extreme weather of 2020 (February was the 3rd warmest, April and May were the 4th and 14th driest since 1901 in Hungary, respectively, according to the Hungarian Meteorological Service), while available air quality model results carry considerable uncertainty [27].

In the heating season, the daily mean temperature is considered to be a good predictor of PM₁₀ levels [27,30] through its correlation with heating emissions. Furthermore, it is a good indicator of air stagnation episodes, characterized by low (winter) or high (summer) temperature anomalies. However, with the increasing role of large-scale transport [31] and seasonal agricultural emissions, the statistical relationship between meteorological variables and air quality decays toward the summer. Spearman correlations (s) between daily mean temperature and air pollutant concentrations are presented in Figures S1–S3 and Tables S1–S3 of Supplementary Material. The daily mean temperature was well correlated with PM₁₀ levels during the curfew period during the reference years 2014–2019 (s between 0.4–0.7), but this correlation nearly disappeared in 2020 (Figure S3). This shows that while the curfew period is normally considered to be within the heating season, the unusually warm spring of 2020 eliminated the importance of heating emissions. The emergency period, including the warm months of May and part of June, showed a weak correlation with temperature. The positive correlation of ozone concentration with temperature was stronger in the emergency period, including early summer months (Figure S2). Meanwhile, in the curfew period, the correlation between temperature and ozone was much stronger in 2020 than in previous years, indicating the weather forcing behind elevated ozone levels.

A recent study found no significant trend of air pollutant gases in Budapest during the last decade, while particulate pollution declined 4–5%/year [34]. Thus, no long-term trend can be identified behind the changes that occurred in 2020. For reference, linear trends fitted on the mean concentration of the curfew period during 2014–2019 are presented in Appendix B, Appendix C and Appendix D for each location; however, these trends are based on only 6 years of data; thus, they are highly uncertain and barely significant.

An important notice for future studies is that the extreme weather in February 2020 with normal emissions (pre-lockdown) caused a more spectacular air quality improvement than the subsequent emission reduction with adverse weather. In fact, the half-year (January–June 2020) mean PM₁₀ pollution in Hungary was significantly lower in 2020 than in the previous years at most locations, but this was entirely attributable to the pre-lockdown winter period, while elevated PM₁₀ levels occurred during the curfew. Furthermore, the observed negative NO_x anomaly had also been present for weeks pre-lockdown. In future studies, it will be necessary to consider the anomalously clean pre-lockdown winter when using it as a reference period or to calculate annual means to assess the air quality impact of the pandemic.