Measurement report: An assessment of the impact of a nationwide lockdown on air pollution – a remote sensing perspective over India

. The nationwide lockdown was imposed over India from 25 th March to 31 st May2020 with varied relaxations from phase-I to phase-IVto contain the spread of COVID-19. Thus emissions from industrial and transport sectorswere halted during lockdown (LD)which resulted in a significant reduction of anthropogenic pollutants. The first twolockdown phases werestrictly followed(phase-I and phase-II) and hence are considered 20 as total lockdown (TLD) in this study. Satellite-based tropospheric columnar nitrogen dioxide (TCN) from the years 2015 to 2020, tropospheric columnar carbon monoxide (TCC) during 2019-2020 and aerosol optical depth (AOD 550 ) from the years 2014 to 2020 during phase-I and phase-II LDand pre-LD periods were investigated with observations from Aura/OMI, Sentinel-5P/TROPOMI, and Aqua-Terra/ MODIS satellite sensors.To quantify lockdown induced changesin TCN, TCC, and AOD and Terra) from the 6-years (2014-2019) mean AOD 550 levels, with a significantreduction (Aqua/MODIS 35 28%)observed overthe Indo-Gangetic plains (IGP) region with a p-value of <<0.05.However, an increase in AOD 550 levels (25% for Terra/MODIS, 15% for Aqua/ MODIS) was also observed over Central Indiaduring LD compared to the preceding year and found significant with a p-value of 0.03.This study also reports the rate of change of TCNlevels and AOD 550 along with statistical metrics during the LD period. in a parameter a of acceleration of the de-trended TCN daily inter-annual variability in TCN.Thus, TCN variability the lockdown-imposed TCN the short-term climatological mean phase-I. Causative factors for this decrease over CI w.r.t phase-I are due to low RH and high wind speed at 700 hPa and 850 hPa over this region. In a Nutshell, this study demonstrates the lockdown induced Terra/MODIS AOD 550 changes over the IGP and CI during total LD period shows a significant change with p-value of 0.01 (99 % confidence interval)with a decrease of 20 % over IGP and 0.03 (97 % confidence interval)with an increase of 25 % over CI when compared to equivalent period in 2019.

activities (21% in India), thermal power plants (28% in India), biomass burning (19% in India)whereas the natural sources of NOx are soils and lightning (Biswal et al., 2020b). Thus, hotspots region of NO2are thermal 65 power plants, urban cities, and industrial regions. In addition to NO2, carbon monoxide (CO) is also an important trace gas in the troposphere and is the main precursor of secondary pollutant ozone in NOx rich environment. Though CO is not a direct greenhouse gas, it has a global warming potential because of its effects on the lifetime of several greenhouse gases. The natural and anthropogenic sources of CO are forest fire,biofuel burning, volcanic activities, and incomplete combustion of fossil fuels, oil, coal, woods, natural gas, and 70 oxidation of hydrocarbons. However, significant amount of contribution to CO is from the anthropogenic emissionsonly (Beig et al., 2020). Harmful effects of CO are dizziness, headaches, stomach-ache, confusion, tiredness.CO is tracer of air pollution due to its lifetimeof about ~1-2 months (Filonchyk et al., 2020).
Natural and anthropogenic activities are responsible for aerosols in the atmosphere. Anthropogenic 75 activities over South Asia have caused considerable changes in aerosol composition and loading. Fine mode aerosols (PM2.5) are mainly from gas to particle conversions which are from biogenic and anthropogenic emissions. Coarse mode aerosols (particles with diameter larger than 10 µm) arise from natural sources namely deserts, oceans, volcanoes and biosphere with less contribution from anthropogenic activities. Over the ocean surface, the natural global aerosol mass is controlled by sulphate, sea salt, and dustaerosols (David et al., 80 2018).Further, Aerosols also affect the earth-atmospheric radiation budget directly in scattering and absorption of incoming solar radiation and indirectlyasclouds formation and precipitation (Ramachandran and Kedia, 2013). Thus, aerosols can influence the Indian monsoon (David et al., 2018). Earlier studies indicate that vehicular (Mahalakshmi et al., 2014(Mahalakshmi et al., , 2015 emissions, industrial, and thermal power plant emissions (Ramachandran et al., 2013) contribute significantly to atmospheric pollution, including gaseous pollutants. The 85 ambient air quality is largely determined by the concentration of trace gases and particulate matter in the atmosphere (Nishanth et al., 2014). Increase in the concentration levels of trace gases and particulate matter has been a challenging environmental issue in urban and industrial areas.Numerous studies have been attempted across the globe to understand the air pollution concentrations during lockdown period and results indicate varied range of percentage reductions in pollutant concentrations. These studies are based on ground 90 based measurements alone (Mohato and Ghosh, 2020; Mor et al., 2020) or satellite data alone (Biswal et al., 2020a;Xu et al., 2020) orwith a combinationof ground and satellite (Ratnam et al, 2020;Biswal et al., 2020b;Singh and Chauhan, 2020). Biswal et al. (2020aBiswal et al. ( & 2020b reported lockdown induced changes in tropospheric NO2 variability over the urban and rural regions of Indian sub-continent with marked reduction of 30-50 % over https://doi.org/10.5194/acp-2020-1227 Preprint. Discussion started: 11 February 2021 c Author(s) 2021. CC BY 4.0 License. the urban and megacities. This change was mainly attributed to the reduced traffic emissions due to restriction 95 on travel.In contrast to the above, increase inlevels of air pollutants during lockdown are also noticed at certain regions,which are associated with natural emissions (dust storms, forest fires) and prevailing meteorological conditions. During India's phase-I of lockdown (25 th March, 2020 to 7 th April, 2020), Ratnam et al. (2020) showed a decrease of AOD550 over IGP region and a drastic increase over the central India, which were mainly due to absence of anthropogenic activities and dominance of natural sources respectively.However the above 100 said studieswere not performed detailed statistical analysis to indicate the observed changes are significantly lower than what could be expected due to inter-annual variability.
Objective of the present study is to understand the air quality quantitatively over the Indian region under the control measures related to COVID-19 restrictions in the country. Thus, the present study examined 105 the spatio-temporal variations of remotely sensed Tropospheric columnar NO2 (TCN), Tropospheric columnar CO (TCC) and Aerosol Optical Depth (AOD550)during LD and pre-LD and compared with preceding year (2019) and short-term mean (2014)(2015)(2016)(2017)(2018)(2019)(2020). With these three air pollutants, we reported lockdown induced changes over the Indian region as a whole, hotspots (usual predominant sources) and urban regions along with the statistical analysis(using de-trended data).However, no one attempted to study the TCC variability during 110 lockdown over the Indian region using satellite measurements. To distinguish natural and anthropogenic emissions, we made an attempt to correlate the subsequent changes associated with meteorology, long range transport as well as forest fires.

Data
Satellite measured air pollutants data offer reliable, un-interrupted observations with high spatial and 115 temporal coverage than ground-based measurements which are point observations. Thus, the TCN observations from the Sentinel-5P/Tropospheric Monitoring Instrument (TROPOMI) and Aura/Ozone Monitoring Instrument (OMI), TCC data from high spatial resolution TROPOMI, AOD550 data from Moderate Resolution Imaging Spectroradiometer (MODIS) Terra/Aqua sensors are used in the present study. The brief details of these sensors are given in Table 1. The TROPOMI was launched on 13 th October 2017 as the single payload on-board the 120 Sentinel-5 Precursor (S5P) satellite of the European Space Agency (ESA). TROPOMI is a push-broom imaging spectrometer flying in sun-synchronous orbit at 824 km altitude and is designed to retrieve the concentration of several atmospheric constituents, which include TCN, TCC, SO2 etc. It was developed jointly by ESA and Royal Netherlands Meteorological Institute (KNMI), which is the most advanced multispectral imaging spectrometer https://doi.org/10.5194/acp-2020-1227 Preprint.   Li et al. (2020). Over land, the previous studies reported that MODIS derived AOD uncertainty with respect to the Aerosol Network (AERONET) is ± 0.05 ± 0.20 × AODAERONET Levy et al. 2013). Details of MODIS AOD retrieved algorithm for collection 6.1 and its validation can be found in Hsu et al., (2019)  Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis which gives hourly data at different pressure levels (700 hPa and 850 hPa) was also used in the present study.Similarly, relative humidity from ECMWF for the respective pressure levels is also used. 145

Methods
In the present study we attempted to assess the impact of lockdown on air quality over India by examining remotely sensed daily concentrations of TCN, TCC and AOD550for the period of 01 st January,2014 to October, 2020. Further, daily concentrations of the above said parameters were de-trended during the study period to subside the inter-annual variability.Hence, phase wisechanges in TCN, TCC levels and AOD550could 150 be attributed to LD induced changes. Thus, the present study focused on the air pollution over the Indian region, its individual states, and state capitals during the strict lockdown period (phase-I and phase-II). Analysis of satellite-based observations of TCN from the years 2015 to 2020, TCC during 2019-2020 and AOD550from 2014-2020 was carried out for lockdown period (phase-I and phase-II)as well as pre-lockdown period. Shortterm climatological means during pre-LD, phase-I, phase-II was computed for TCN from 2015 to 2020 and 155 AOD550from 2014 to 2020to assess the temporal changes of pollutants in the atmosphere. We have focused our analysis for the first two phases of lockdown in which the industrial and transport sectors were brought to a near standstill. Figure 1 shows the data processing and execution strategy,which was followed in this study. The detailed methodology used in this study is as follows. Python programming language is used to analyse 160 TCN,TCC and AOD550 variables during the study period as discussed in Figure 1

205
Where N is total number of qualified pixels over the Indian region and μ b is mean bias. At each pixel, weekly bias (b) of TCN is estimated from the weekly mean TCN during 2020 lockdown period w.r.t 2019 and 2015-2019 period as shown in equations (1-2). For prominent change detection, additionally 1SD deviation filter was 210 applied on the bias values of TCN and AOD550. Therefore, if bias is greater than 1SD then the featuring pixels are classified as positive and if less than -1SD then they are considered as negative pixels. Pixels within ±1SD are omitted to avoid minimalistic feature changes and for better characterization. Subsequently, we computed the percentage of positive (Increased area) and negative pixels (decreased area) using following equations (5-6).
The same equations were applied on AOD550 during 2014-2020 study periods. Further to understand LD induced changes in TCN and AOD550quantitatively daily mean values are de-trended using yearly data, which accounts for inter-annual variability in TCN and AOD550 respectively. De-trended values of TCN and AOD550 are generated by subtracting the linear regression estimated values from the daily means of TCN and AOD550. To 220 study the lockdown induced changes with significant levels, a paired t-test (Freedman et al., 2007) was implemented on the de-trended TCN and AOD550 data during respective study period. The t-test follows a Student's t-distribution under the null hypothesis of H0 with the means (µ) of two populations are equal (µ1= µ2) with alternative hypothesis Ha:µ1≠ µ2.To reject or accept the null hypothesis, p-value was used in this study. The hypothesis H0 is rejected when a p-value is less than 0.05 and accepted if p-value is greater than 0.05 (5 % 225 significance level).

Results and Discussion
Present study analysed the satellite based tropospheric columnar NO2 (TCN), TCC and AOD550 data to assess the lockdown induced changes over the Indian region.

Effect of Lockdown (LD) on TCN
The spatio-temporal variability in TCN concentrations during pre-lockdown and lockdown period (phase-I and phase-II) were analysed for the years 2019 and 2020 using high spatial resolution (Table 1)  India has a significant number of major power plants and refineries with associated industries. During phase-I LD in 2020, a reduction of TCN levels is observed over this region, which is due to shutdown of industries and urban activities such as transport and small-scale industries. However, country's high TCN levels are noticed in the eastern region with less spread indicating the active role of power plant industries located in this region. 255 Simultaneously, the National capital region (NCR) shows marked reduction by about ~70 % during phase-I LD.
The major source of emissions in the NCR are dominated by heavy traffic, densely located industries, and industries of steel, cement and sugar (Ghude et al., 2008). As mentioned above, India's strict lockdown permitted only essential services therefore remaining all activities were halted during phase-I period. Thus, https://doi.org/10.5194/acp-2020-1227 Preprint. Discussion started: 11 February 2021 c Author(s) 2021. CC BY 4.0 License. resulted in low levels of TCN over the hotspot regions except in eastern region due to continuous operation of 260 power plants (Biswal et al., 2020b).
To sustain the Indian economy, activities namely agricultural practices and associated activities (crop residue burning) are given permission during phase-II along with phase-I restrictions. Figure 2c shows TCN levels and their difference map during phase-II LD.With respect to the same period of phase-II in 2019, the TCN levels over the country decreased by 13 % with mean TCN of 1.55×10 15 molecules cm -2 and 1.75×10 15 265 molecules cm -2 in 2020 and 2019. Thus, continued decrease of TCN levels are recorded over the hotspot regions. However, an increase of TCN also observed over the neighbouring regions of eastern region clearly indicating the dispersion of TCN. In contrast to earlier, an increase in TCN levels over north-east region could be due to seasonal biomass burning in this region.Thus, the mean TCN levels over the entire country is 1.54×10 15 molecules cm -2 during total LD (phase-I and phase-II together) with a reduction of 18 % compared to 270 respective period in 2019 as well as with respect to 5 years mean TCN .
Overall, southern part of India reported less TCN values as compared to eastern and NCR regions Arabian Seaover this region (Ramachandran et al., 2013). Thus, reduced TCN levels over southern part of India irrespective of LD were observed due to above facts. Weekly variations of TCN were also shown in Figure 2d to assess the extent of source emission during the lockdown period.Therefore, present study depicted the possible driving factors of TCN values during pre-LD, phase-I and phase-II using high resolution spatial data from satellite. interval during total lockdown period.However, it is also noticed that a decrease of TCN during prior and post 300 lockdown periods, which is further tested statistically and found insignificant change with a p-values of 0.08 and 0.24 respectively. Further statistical significance of TCN variability across hotspot, cold spot regions and also over the major cities where TCN dropped (↓) drastically (except NE) along their percent drop during total LD when compared to 5 years mean TCN levers were summarized in Table 2. It clearly shows the TCN levels over the IGP (22% ↓), eastern region (29% ↓), and major cities (New Delhi 54% ↓) declinedsignificantly compared to 305 preceding 5 years mean TCN levels during the total LD period. Change in TCN during the study period is also associated with the inter-annual and seasonal variability besides its dominant anthropogenic sources. Figure 3c shows annual means of TCN in pre-LD, phase-I and phase-II LD during 2015-2020 period.It clearly depicts inter-annual variability in TCN between the years at each time scale along with the lockdown-imposed changes.
Between the time scales during the study period, a clear seasonality in TCN levels is also observed in Figure 3c. 310 The horizontal bar plots in Figure 3d showsthe rate of change (RoC) in TCN levels in 2020 against the mean during 2015-2019 indicating the impact of lockdown on TCN concentrations over Indian region.The RoC is extremely important in weather and climatological studies because it allows understanding and predicting the trends/patterns in climatic parameters. RoC is used to describe the percentage change in a parameter over a 315 defined time period and it represents the rate of acceleration of the parameter. To compute the RoC in this study, we have used de-trended TCN daily values, which accounts inter-annual variability in TCN.Thus, the present RoC depicts the TCN variability due to the lockdown-imposed changes alone. There is a clearly observable lowering in TCN levels relative to the short-term climatological mean by 12% for the pre-lockdown period as

Effect of LD on TCC
The mean TCC levels over Indian region during the pre-lockdown and LD periods were studied to assess the  during lockdown which is major contribution to CO from rural areas and some parts of urban region (slums).In India, 72 % of the populations live in rural and urban slums and most of them are continued to usehousehold biofuel for cookingunder lockdown (Verma et al., 2018;Beig et al., 2020).
However, the mean TCC levels as shown in Figure 5c are higher during the phase-II of lockdown. Over the entire country, the mean TCC value during phase-II is 2.38 ×10 18 molecules cm -2 in comparison to 2019 390 mean value of 2.32×10 18 molecules cm -2 . In phase-II of LD, the TCC levels are decreased in NE region, which is strongly attributed to the reduced active fire activity in this region as shown in Figure 5c. Except in NE region, consistent increase of TCC levels is observed during phase-II. Since agriculture farming industry is exempted in the phase-II LD and observed active fire counts in the central India, thus observed enhancement in the TCC levels during phase-II. An increase or decrease of TCC levels in the atmosphere is mainly dominated 395 significantly by anthropogenicactivitiescompared to natural emissions (Kanchana et al., 2020)

Effect of LD on AOD550 440
We have used Terra-Aqua/MODIS derived AOD550during 2014-2020 for the months of January to July to understand the lockdown-imposed changes. Terra/MODIS AOD550 represents the footprint for 10:30 and Aqua/MODIS AOD550for the 13:30 local time. As we observed similar spatial variation of AOD550 from both Terra-Aqua/MODIS, only Aqua/MODIS derived AOD550 is shown here (Figure 6). AOD550 levels over the Indian region for 2019, 2020 and the difference in AOD550 for both years for pre-lockdown period is depicted in 445 Figure 6a. During this period, the AOD550 levels for 2020 over the IGP region (~21% of the Indian Territory landmass) is more compared to rest of the regions ofIndia which is expected throughout the year. This is mainly because of its orographic effect and densely populated (accommodating ~40% of the Indian population). The main anthropogenic sources over IGP region arecoal-based power plants and industries, crop residue and forest fires and household cooking which contribute to high AOD in this region.Thus, the IGP is known as first 450 hotspot for anthropogenic aerosol emission in South Asia.During phase-I of LD as shown in Figure 6b, aerosol loading over the IGP region attained its baseline concentration (~45% dropw.r.t. 2019 of the same period) due to the strict implementation of LD. This region is densely populated and congested industrial activities, which were shut down during this period resulted a nearly AOD free atmosphere. This indicates absence of anthropogenic activities due to mobility restrictions. Further, prevailing meteorology over IGP (high wind speed 455 and low relative humidity at 850 hPa and 700 hPa)is also favoured for decrease in AOD550 during phase-I LD.
Despite the strict LD in the country, unexpected increase in AOD550 is observed by ~28 % compared to preceding year of the same period over Central India (CI) which ispredominantly dominated by dust storms (Ratnam et al., 2020) through long range transport and prevailing meteorology (Pandey et al., 2020). Thus, to understand the prevailing meteorology over CI,phase wise relativehumidity and wind speed at pressure levels 460 850 hPa and 700 hPa respectively were analysed as shown in Figures 7a-b. During phase-I and phase-II, majority of the winds over CI dominated by westerly (calm winds) with high relative humidity. Under this prevailing meteorology, calm winds contribute to slow dispersion and high RH modulates the aerosol chemistry and hygroscopic growth mechanism (Pandey et al., 2020). As a result, the increase of AOD550over CI is observed.Further, high AOD550 over NE regions also observed because ofhigh active forest fire counts (Figure  465 5d)compared to 2019 LD period. Figure 6c shows AOD550 during phase-II of India's LD in 2020 against AOD550 in 2019 of same period.During this phase, an increase inAOD550 (~ 3%)over IGP was observed. Over CI, a reduction of AOD550(~ 18%) was observed compared to phase-I of LD andnot much change (~1%)when compared to respective period in 2019which clearly depicts reversal of meteorology in phase-II with respect to https://doi.org/10.5194/acp-2020-1227 Preprint. Discussion started: 11 February 2021 c Author(s) 2021. CC BY 4.0 License. phase-I. Causative factors for this decrease over CI w.r.t phase-I are due to low RH and high wind speed at 700 470 hPa and 850 hPa over this region.
In a Nutshell, this study demonstrates the lockdown induced Terra/MODIS AOD550 changes over the IGP and CI during total LD period shows a significant change with p-value of 0.01 (99 % confidence interval)with a decrease of 20 % over IGP and 0.03 (97 % confidence interval)with an increase of 25 % over CI

Short-term climatological variation of AOD550 due to lockdown
Aerosol optical depth is one of the important short-term climatic forcing agents along with long lived greenhouse gases namely carbon dioxide (CO2), methane (CH4), water vapor (H2O) and nitrous oxide (N2O). A 7-day smoothing average filter was applied on AOD550 time series data as discussed in section 4.1.1. continued effect of lockdownas phase-III and IV and scavenging effect during monsoon season. Due to increase of precipitation in the active summer monsoon (June-July) season, lowering of aerosols is expected (Boucher et al., 2013). Thus, the continued lockdown and active monsoon improved the air quality beyond strict lockdown period as shown in  The annual mean AOD550over the Indian region in each phase is shown as vertical bars in Figures8e In similar manner, the AOD550 measured by the Terra-Aqua/MODIS also shows strong reduction over IGP region during the total LD (top right Figures 10b-c). However, unexpected increasing effect is noticed in the CI states with respect to 6 years mean of respect AOD550 during phase-I. Similar results are also observed when compared to preceding (2019) year mean AOD550 which is discussed earlier section in detail manner. Further, it is observed that, the negative RoC of AOD550 over IGP region during phase-I is more prominent compared to 615 phase-II RoC.It is also noticed further that the RoC of AOD550 computed from the Terra-Aqua/MODIS showing similar trends during the total lockdown period with small difference in the amplitudes. This difference of amplitude between these two sensors could be due to difference in overpass time, which changes atmospheric dynamics such planetary boundary layer height, solar zenith angle and prevailing meteorology. An average of Terra/Aqua MODIS derived RoC of AOD550 show strong reduction in the western part of India mainly 620 Rajasthan (-36 %) and Gujarat (-31 %) respectively during the total LD period (Ranjan et al., 2020). Therefore, in a nutshell an analysis of RoC depicts regional variability of air pollutants during the total LD period in 2020 w.r.t to short-term (5-6 years) mean.

Conclusions
The present study was carried out an analysis on air pollution in connection with the world's largest lockdown 635 imposed by Government of India to contain the spread of COVID-19. The lockdown was extended as 4 lockdowns with strict lockdown from the phase-I to several relaxations in the phase-IV. However, the lockdown was near total only in phase-I and II, with the total shutdown of industrial and transport sectors. Thus, we have only considered first two phases in the present studyas total lockdown. We used satellite-based observations of tropospheric TCN, TCC and AOD550pollutant concentrations analysed during the period of lockdown and prior 640 to LD against the same period of the preceding year (2019) and also against the short-term mean (2014-2019) for about 6 years.
Following are the major findings from the present study  Due to India's strict LD, the TCN levels are dropped significantly to 18 % across the country compared to preceding year with a p-value of 0.0007(confidence interval of 99.93 %). 645  Further, analysis is emphasised over the TCN hotspot regions of the Indian sub-continent and observed reduction of (29%) TCN during the total LD period with higher confidence interval.
 The TCN levels with respect to short-term climatological mean are markedly dropped over the urban locations namely New Delhi (-54%), Bangalore (-43 %), Chennai (-41 %), Mumbai (-35 %) and Hyderabad (-30 %) respectively with high confidence interval about 99.90 %. 650  However, during the total LD, an unexpected increase of TCN levels are recorded over NE region, which is directly correlated with the seasonal biomass burning in this region. This increase is also evaluated statistically against 5-year mean TCN and found insignificant with p-value of 0.19.
 The TCC levels are decreased during the phase-I over IGP, north and south regions which could be due to the absence of transportation and shutdown of industries. Although, variability in the TCC levels 655 werenoticed during the total LD period it was tested statistically and found insignificant.Observed high tropospheric CO levels in the NE region during phase-I LD period, which is mainly attributed to the active fire counts in this region. Also observed low TCC levels in the NE region during phase-II due to the diminished effect of fire counts.
 Since IGP region is densely populated and clustered industries, which were shut down during phase-I 660 of India's LD, the AOD550 levels are attained to near baseline in this region (AOD mean value=0.2).
This drastic decrease of AOD550 in the IGP region statistically evaluated and found very significant with a p-value of 0.008 with preceding year (45% decrease) and 50 % reduction against 6-year mean with a p-value <<0.05. https://doi.org/10.5194/acp-2020-1227 Preprint. Discussion started: 11 February 2021 c Author(s) 2021. CC BY 4.0 License.
 Despite the country's LD, the AOD550 levels are high over the CI, which were predominantly 665 dominated by the transportation of dust storms and prevailing meteorology. Also observed high AOD550 over NE and is associated with active fire counts. However, this increase is significant in the CI with a p-value 0.03 and insignificant in the NE region with a p-value of 0.33, which indicates insignificant change due to LD.
 The LD induced changes in AOD550 measured by the Terra-Aqua/MODIS show a significant change 670 over the Indian region with very high confidence against 6-year short-term climatological mean. This variability helps to improve the regional air quality.
 Further, an analysis of RoCwas carried out to depict the regional variability of air pollutants during the total LD period in 2020 w.r.t to short-term climatological mean.
Therefore, this study successfully demonstrates the satellite based TCN, TCC and AOD550 changes due to the 675 India's lockdown during 2020 and compared against preceding year (2019) and also against the short-term mean picture of 2014-2019.

Code/Data Availability
The satellite and reanalysis data used in the present study are freely available and can be downloaded as 680 summarized in Table 1 with user's credentials.