HTML
--> --> -->Considering the rapid development of large cities and their constantly changing anthropogenic loads on neighboring regions, it is extremely important to monitor the variations of atmospheric compositions in cities and their surroundings. Furthermore, such monitoring in background areas is essential to assess the impact of human activities on atmospheric compositions (IPCC, 2001, 2007, 2013).
CO is a toxic gas and one of the major pollutants of urban air. It is also one of the most important chemically active gases determining the oxidative capacity of the atmosphere (Khalil et al., 1999). CO mainly derives from anthropogenic sources, as well as biomass burning (Duncan et al., 2007; Yurganov et al., 2011). The concentration of CO in the lower troposphere correlates closely with concentrations of other important substances [aerosols, carbon dioxide (CO2), nitrogen oxides etc.] (Wunch et al., 2007; IPCC, 2013; Golitsyn et al., 2015). In the case of photochemical smog, CO [as well as nitrogen dioxide (NO2) and non-methane volatile organic compounds] also largely defines the surface concentration of ozone (Khalil et al., 1999; IPCC, 2013). The lifetime of CO in the atmosphere is sufficiently long (two weeks to three months) to serve as a tracer in studies of atmospheric pollutant transport (Garrett et al., 2010; Vasileva et al., 2011; Golitsyn et al., 2015).
A considerable amount of representative data is needed to monitor changes in atmospheric compositions and climate. Unfortunately, the number of such monitoring stations is quite small in Eurasia, especially in distant and sparsely populated regions like central Siberia, the Arctic, the Far East, and the desert and mountain areas of Asia. On the other hand, the transport of pollutants from industrial activities in the North China Plain region leads to high surface concentrations of CO and aerosols in the mountain areas located 150-200 km to the north of Beijing (Golitsyn et al., 2015). The transport of pollutants from industrial regions in Europe and Southeast Asia can cause a doubling of surface CO concentrations, even in background areas of the Krasnoyarsk region (Vasileva et al., 2011).
Because of the limited number of ground monitoring stations in Eurasia, it is necessary and important to obtain information on atmospheric compositions [e.g., the climatically active gases of CO2 and methane (CH4), aerosol optical thickness, and pollutant species (CO, NO2)] from a large variety of satellite data. However, all satellite observations have obvious disadvantages in terms of their accuracy and feasibility under different weather conditions. The most serious drawbacks of satellite measurements are the uncertainties introduced by cloud and surface albedo (Crevoisier et al., 2003; Deeter et al., 2003) and the low sensitivity of orbital sensors to variations in chemical species in the lower troposphere, especially in heavy polluted atmosphere (Fokeeva et al., 2011; Yurganov et al., 2011; Safronov et al., 2015), as well as their coarse temporal resolutions (Wang and Zhao, 2017). Nevertheless, despite these disadvantages, orbital instruments, such as spectrometers (MOPITT, AIRS, IASI etc.), can provide vast quantities of information on many important gaseous components (Clerbaux et al., 2003, 2009; Worden et al., 2013). Some orbital sensors, such as MOPITT and AIRS, can also provide long-term (>10 yr) observational data (Deeter et al., 2017; Rakitin et al., 2017).
The validation of satellite data using ground-based measurements is essential to ensure their reliability. Many ground-based observational stations have therefore been established. The quality and accuracy of such validation work depends largely on the representativeness of the spatial and temporal variations of the ground-based measurements. To compare ground- and satellite-based observations, the following approaches are usually used:
(1) Comparisons using ground-based data from a large number of stations averaged over a period of 10-30 days. With this approach, the average difference between ground- and satellite-based data is small, though in some areas the differences can reach up to 20%-30%, even for 30-day averages (Garsia et al., 2013).
(2) The retrieval of surface component concentrations from satellite measurements whilst using standard vertical profiles at deliberately selected background stations (Sussmann et al., 2005; Clerbaux et al., 2009; Deeter et al., 2013; Worden et al., 2013). The differences between ground- and satellite-based data in background areas are often small, but in cases of strong air pollution in the lower troposphere the differences can reach an order of magnitude or more (Fokeeva et al., 2011; Yurganov et al., 2011; Crevoisier et al., 2003).
(3) The validation of vertical profiles of atmospheric components obtained from airplane or balloon measurements. Arguably the most informative approach, results of this type have been reported in several papers (e.g., Arshinov et al., 2014; Deeter et al., 2014). However, a limitation of the verification in these studies is that the measurements were only carried out for a few days and/or mostly in unpolluted conditions. Nevertheless, even for background atmosphere many researchers have indicated underestimations of CO concentrations by 8%-30% for IASI in the mixing layer (e.g., Arshinov et al., 2014).
Crucially, there have been far from enough studies on the validation of satellite measurements to provide sufficient quantities of representative information on the concentrations and variations of many trace gases and pollutants (CO, CO2, CH4 etc.) in areas where ground-based measurements are unavailable. The lack of such information leads in particular to uncertainties in obtaining the trends and emissions of atmospheric components (Crevoisier et al., 2003; Fokeeva et al., 2011; Yurganov et al., 2011; Safronov et al., 2012; Sitnov et al., 2013). Besides, orbital instruments are always subject to the issue of "drift" in terms of some of their technical characteristics, like time of operation; thus, improvements in computational algorithms throughout almost the whole process of validation and comparison are needed.
The aims of this study were to: (1) evaluate the quality of modern satellite products of CO column content through comparison with ground-based measurements, and study the possibility of using satellite data to obtain the spatial and temporal distribution of CO and assess CO regional trends; (2) compare CO pollution levels in several Eurasia metropolises in different regions and climatic zones, and with different economic growth rates; and (3) evaluate the temporal trends of total-column carbon monoxide (CO TC) in these regions.
2.1. Satellite products
The latest versions of several satellite CO products [namely, MOPITTv6 Joint, AIRSv6 (standard, data only) and IASI MetOp-A] were used.MOPITT (Measurements of Pollution in the Troposphere) was launched onboard the Terra satellite in late 1999. It records infrared radiation spectra at about 4.7 μm (channel TIR, TermalInfraRed) and 2.3 μm (channel NIR, NearInfraRed) at 1030 LST (Local Standard Time) (Deeter et al., 2003; Drummond et al., 2010; Worden et al., 2013; Buchholz et al., 2017). Specifically, the MOPITT v6 Level 3 Joint (combination of channels TIR/NIR) daily data of CO TC were used in this study, which have a spatial resolution of 1°× 1° corresponding to about 60 km from west to east and 100 km from north to south in the midlatitudes, and subsequent averaging over other domains. MOPITT v5, v6 and v7 are presented in Worden et al. (2010, 2013) and Deeter et al. (2014, 2017).
AIRS (Atmospheric Infrared Sounder) was launched onboard the Aqua satellite on 4 May 2002. This orbital diffraction spectrometer records the spectra of atmospheric absorption of Earth's infrared radiation from 3.75 to 15.4 μm (Aumann et al., 2003; McMillan et al., 2011; Worden et al., 2013; Olsen, 2015) twice a day, covering more than 80% of Earth's surface. Data of primary levels are for cells of about 45× 45 km. We used the data of Level 3 v6 with a resolution of 1°× 1°, and only daytime measurements of CO TC from the ascending orbit (i.e. around 1230-1330 LST for each point), with subsequent averaging over other domains.
IASI (Infrared Atmospheric Sounding Interferometer) is a part of the orbital complexes MetOp-A (launched in 2006) and MetOp-B (2012) (Clerbaux et al., 2009, 2010; August et al., 2012; Worden et al., 2013). IASI was designed to record Earth's spectrum of radiation in the range from 645 to 2760 cm-1 (15.5 and 3.63 μm, respectively), with a spectral resolution of about 0.5 cm-1. The device is used to provide real-time information on temperature and water vapor content in the atmosphere over the entire surface of Earth. Moreover, the vertical profiles of some gases, including CO, CH4 and CO2, can also be obtained. We used the data from IASI MetOp-A, Level 2 (vertical profiles of CO concentrations for cells of about 18× 22 km), with subsequent averaging over domains of different lengths. We calculated the average CO TC in the first half of the day, i.e., during ground-based spectroscopic measurements.
The daily mean CO TC was averaged over the domains 1°× 1° and 5°× 5°, with the center at the corresponding ground-based observational site. Ground-based data were averaged within 1.5 h of the time when the satellite passed over the observation site. Only IASI data in the first half of the day were used; only AIRS data from the ascending orbit were used; and only daytime data of MOPITT were used.
The most complete overview of the abovementioned satellite platforms and instruments is presented in (Worden et al., 2013).
2
2.2. Ground-based measurements
The datasets of six spectrometers were used in this study. The technical parameters of all the ground-based spectrometers at all the sites are presented in Table 1. Note that at the site of Peterhof the spectrometer was changed in 2009, meaning there was a different instrument in 1998-2009 to that in 2009-2014, and these two instruments had different channels and spectral resolutions, as detailed in Table 1.Ground-based data were obtained by using solar diffraction spectrometers of medium resolution (0.2 cm-1), similar to those onboard satellites (Fokeeva et al., 2011; Rakitin et al., 2011; Yurganov et al., 2011; Wang et al., 2014; Golitsyn et al., 2015), at the following sites: OIAP Moscow (OIAP RAS, located in the center of Moscow); ZSS (Zvenigorod, 53 km west of Moscow); ZOTTO (central Siberia); and IAP Beijing (urban Beijing). The locations of the ground stations are given in Table 1. ZSS can be considered as a background station, since the influence of Moscow on the CO column at this station is small (Rakitin et al., 2011; Wang et al., 2014; Golitsyn et al., 2015). All spectrometers used at the abovementioned stations were directly intercalibrated among themselves (based on simultaneous measurements at one point), and indirectly intercalibrated with Fourier spectrometers at The Network for the Detection of Atmospheric Composition Change (NDACC) stations using long-term observations (Yurganov et al., 2010). In addition, the spectrometer at ZSS has been used for the validation of satellite-based measurements and model calculations (Crevoisier et al., 2003; Yurganov et al., 2010; Fokeeva et al., 2011; Yurganov et al., 2011). The error of a single measurement of the spectrometers occasionally reaches 5%-7%; however, in the absence cases of diurnal CO TC variations, the average dispersion is equal to 3%-5% (Rakitin et al., 2011; Yurganov et al., 2010; Wang et al., 2014).
Spectroscopic measurements of the CO TC in the atmosphere at Peterhof [(59.88°N, 29.82°E); 20 m MSL; ～35 km southwest from the center of St. Petersburg] were conducted by the Department of Atmospheric Physics, St. Petersburg State University (Makarova et al., 2004, 2011). During the period 1995-2011, the IR spectra of direct solar radiation in the spectral range 2140-2180 cm-1 were recorded using a classic grating infrared spectrometer with medium spectral resolution (0.4-0.6 cm-1). The CO TC was retrieved from the spectra using software developed by the Department of Atmospheric Physics, St. Petersburg State University. This software implements a method of statistical regularization. The estimate of the random error of a single measurement of CO TC is about 6%-8%, and the daily average standard deviation is 2%-4% (Makarova et al., 2004, 2011) in the absence cases of short-period CO TC variations.
Since 2009, the solar IR spectra have been measured using a Fourier spectrometer of high spectral resolution (IFS125 HR, Bruker, Ettlingen, Germany). Measurements are made with an optical path difference of 180 cm, which corresponds to a spectral resolution of 0.005 cm-1 (Hase et al., 1999).
The determination of CO TC via the high-resolution spectra of direct solar radiation is carried out using the SFIT2 v3.92 software developed for NDACC (Hase et al., 2004; Garcia et al., 2007).
Three spectral micro-windows are recommended for CO TC restoration at NDACC stations: 2057.70-2058.00, 2069.56-2069.76 and 2157.50-2159.15 cm-1 (Senten et al., 2008). The average value of a random relative error for a single measurement of total CO is about 3%, and the variability of the total value of the CO TC during one day is about 1%-2%.
It is important to note that, largely because of the topography and characteristics of the atmospheric circulation in the region, the meteorological conditions in Beijing have specific characteristics compared with those at all the other locations. The presence of mountain ranges (15-50 km to the west and north) on one side, and the relative proximity to the ocean (about 200 km to the east of the city) on the other, leads to frequent changes in wind direction and speed, as well as high horizontal and vertical inhomogeneities of the wind fields (Wang et al., 2014; Golitsyn et al., 2015).
Ground-based CO TC data were used in the form of mean daily values. Measurements at all ground stations were conducted on sunny days. Diurnal values of CO TC were obtained by averaging individual measurements (times of measurement presented in Table 1), i.e., around the satellite overpass times. A linear relationship between ground- and satellite-based CO TC was assumed: K=(U gr-A)/U sat, where U gr and U sat are the CO TC derived from ground- and satellite-based measurements, respectively, and A is a constant.
To explore the impact of PBL parameters on the comparison results, ground- and satellite-based data at various PBL heights were used, calculated at each point using the Global Data Assimilation System (GDAS) with a 1°× 1° resolution and three-hour averages (Hase et al., 1999).
3.1. Total content of CO: long-term trends
The annual mean CO TC at the different sites is shown in Figs. 1a and b. The period (1998-2014) was selected to achieve maximum coincidence among the measurement data at all points, and the best statistical reliability of the measurements. Autumn was chosen as the observational period for all the selected sites because the measurements at Beijing were conducted up to 2012 during this period (October-November), where there were more sunny days. In Moscow, ZSS and Peterhof, the number of sunny days is relatively lower in autumn, so the averaging period was increased to 15 September to 30 November.The relatively high level of CO in Beijing is noteworthy. The highest CO TC rate of decline (3.73% yr-1) in 1998-2014 was found in Moscow. Although CO TC decreased at a rate of 1.14% yr-1 in Beijing, the interannual variation was considerably large. The trend at ZSS can be clearly seen even excluding the data of 1998, which characterized with high level of CO and by the transition from domestic car brands to imported car brands, which have environmental filters fitted to their exhaust systems. The CO TC measurements at ZSS depend slightly on the impact of the Moscow site (Rakitin et al., 2011, Golitsyn et al., 2015); therefore, ZSS can be considered as a rural station. The trends at both the ZSS and Moscow sites during 1998-2014 are negative, even with the increasing numbers of cars in the Moscow metropolis. Conversely, Peterhof is located in a coastal area subject to sea-land circulation, with the dominant winds from the ocean side. Therefore, Peterhof can be considered as a sea background station.
Figure1. Annual CO TC in autumn at the (a) background sites of Zvenigorod and Peterhof (15 September to 30 November) and (b) urban sites of Moscow (15 September to 30 November) and Beijing (1 October to 30 November). The missing results in some years is because of the statistically insufficient coverage of the measurements in the corresponding period (<5 days). The solid lines show the trend derived from average autumn values for 1998-2014, while the dashed lines represent 2007-2014 (black for Beijing and grey for Moscow in both plots). The trend estimates were obtained at the 95% confidence level, and the vertical lines mark the standard deviation in the determination of average values.Decreasing CO trends were apparent at both background sites. The differences in the mean values of CO at ZSS and Peterhof can be explained by the spatial distribution of CO TC (Yurganov et al., 2010; Makarova et al., 2004, 2011), i.e., a decrease with increasing latitude in the Northern Hemisphere (Dianov-Klokov et al., 1989; Yurganov et al., 2010; Worden et al., 2013). Also, the CO trends at both ZSS and Peterhof changed sign during the period 2007-14 (Fig. 1a), when satellite data were available to make comparisons.
In addition, the data of AIRSv6 L3 (diurnal "ascending" data with a resolution of 1°× 1°) for 2003-2014 were used to assess the regional long-term trends and compare them with the previously obtained ground-based estimates. Based on daily average AIRS v6 data, the regional trends of CO TC for the autumn seasons (15 September to 30 November) of 2003-14 and 2007-14 are illustrated in Figs. 2a and b. Clearly, the CO TC in most of Eurasia decreased during the period 2003-14 (blue area), and these changes were confirmed by the estimates based on the data of all the ground-based sites. Notably, after 2007, a slight increase in CO (0.1%-0.5% yr-1) was found in almost all of northern Eurasia (yellow areas). For Siberia and South Asia, this was perhaps because of the impact of wildfires in July-August of 2012 and 2014 (Siberia), and November 2014 (Malaysia), and the subsequent removal of relatively long-lived CO toward the polar region.
Figure2. Trend distribution of the CO TC over Eurasia in autumn (15 September to 30 November), according to AIRS v6 (spatial resolution: 1°× 1°), during (a) 2003-14 and (b) 2007-14. The observational sites are marked by numbers corresponding to Table 1. Green numbers indicate background sites, while red ones indicate the urban sites.These results regarding the trends of CO at all sites are consistent with our earlier results (Makarova et al., 2004, 2011; Yurganov et al., 2010; Wang et al., 2014; Golitsyn et al., 2015), as well as with the results of (Worden et al., 2013) obtained from satellite data of several orbital spectrometers in 2000-11 in various regions.
4.1. MOPITT v6
The satellite product of MOPITT (recent versions, v5 and v6) records the absorption spectra by using the TIR and NIR spectral channels, and the combination of these channels (TIR/NIR; see section 2). Using the NIR channel, according to the developers, increases the sensitivity of the sensor in the lower troposphere. The correlation coefficients (R2) obtained by the developers from validation are very high (～0.9) for all three channels. Nonetheless, comparisons with ground-based measurements are mainly made at background stations (Deeter et al., 2013, 2014). Without questioning the mentioned results (Deeter et al., 2013, 2014), by comparing the average daily CO product, MOPITTv6 Joint L3, with ground-based data, we found an R2 of 0.43 for the ZOTTO background station, and 0.51 for rural Peterhof (averaging 1°× 1°) (Table 2). The CO TC diurnal values of the ground-based spectrometers and satellite sensors are compared in Table 2. The CO TC satellite data from the MOPITT v06 Joint product (1°× 1° domain) are compared with the data from the ground-based spectrometers (at ZSS and Beijing) during 2010-14 in Fig. 3.
Figure3. Comparison of daily mean CO TC derived from the MOPITT v06 Joint data product with the data from the ground-based spectrometers (at ZSS and Beijing, 2010-14). The cases with impacts from natural fires are excluded.2
4.2. AIRS
Steady positive correlations of AIRS data with ground-based CO TC data under background conditions were obtained. The R2 values ranged from 0.42 to 0.66 for daily means (averaged over 1°× 1°), and the slope coefficient of the regression line (K) was close to 1, similar to that in (Rakitin et al., 2015). The best correlation was observed for Peterhof (R2=0.84). In conditions of increased air pollution (at ZSS and ZOTTO in cases of wildfires, and at Beijing), the correlation was quite low (R2=0.32-0.64; averaged over 1°× 1°), especially at Beijing (Table 2 and Fig. 4).From Table 2, under increased air pollution the correlation coefficients averaged over the domain 5°× 5° were lower than those averaged over the domain 1°× 1°. Meanwhile,under background conditions, the correlation coefficients averaged over both 5°× 5° and 1°× 1° were practically identical for both AIRS v6 and other orbiting spectrometers (Table 2). There was a significant increase in K by a factor of 1.5-2.8 under heavy air pollution, implying an underestimation of CO TC when using AIRS v6, as compared with ground-based measurements.
Figure4. Comparison of daily mean satellite CO TC data (AIRS v6 product; 1°× 1° domain) with ground-based spectrometer observations (at ZSS, ZOTTO and Beijing, 2010-14). Events with impacts of natural fires are excluded.A relatively spatially homogeneous CO distribution has been observed in the case of wildfires in the central European part of Russia in summer 2010, and for haze events in Beijing (Safronov et al., 2015). Thus, a high degree of correlation between CO surface concentrations and CO TC (R2 of around 0.8-0.9) was observed in different parts of Moscow and the Moscow region in July-August 2010 (Fokeeva et al., 2011; Golitsyn et al., 2011), as well as a high correlation between soot and submicron aerosol (in various episodes and years) in urban Beijing and at Xinglong——a mountain site 150 km to the north of Beijing and about 1000 m MSL (Golitsyn et al., 2015). Meanwhile, some studies have stated that, under heavy air pollution, major pollutants can be confined to the lower troposphere——a layer several hundred meters in height (Fokeeva et al., 2011; Golitsyn et al., 2011; Yurganov et al., 2011), where satellite spectrometers have low sensitivity. Based on our results, the discrepancy between ground- and satellite-based data at the spatial scale of 5°× 5° can be explained by the spatially inhomogeneous distribution, but this explanation for the scale of 1°× 1° averaging is less applicable. Thus, there is a systematic underestimation of the CO TC by the AIRS instrument (product v6) in polluted lower-troposphere conditions.
2
4.3. IASI MetOp-A
The average CO TC data of IASI MetOp-A are compared with ground-based spectrometer observations at the background stations of ZSS and ZOTTO in Fig. 5. The R2 values range from 0.19 to 0.23, and K ranges from 0.62-0.90, for the 1°× 1° domain (Table 2 and Fig. 5). The seasonal variations of CO TC agree well with the seasonal variations at the ground stations. Under polluted conditions, and with a domain of 1°× 1°, K increased by a factor of 3.5 at ZOTTO (Fig. 5 and Table 2), and the R2 increased too (0.70).2
4.4. Comparison specifics under heavily polluted conditions
For comparison, all ground-based data were selected and averaged for the time intervals noted in Table 1, i.e. the measurement times of the ground-based spectroscopic observations were close to those of AIRS and IASI. The time shift between single ground-based and orbital observations was no more than 1.5 h for AIRS and IASI, and 3 h for MOPITT. Also, for polluted conditions, those days with strong CO TC variation (>10% in magnitude) within the appointed time intervals of the ground-based observations were excluded from the comparison. Under background conditions, the diurnal behavior of CO TC is generally weak, so a small time-shift could be ignored completely.
Figure5. Comparison of diurnal satellite CO TC data (a product of IASI MetOp-A) with data from ground-based spectrometers (at ZSS and ZOTTO, 2010-13). High CO TC values correspond to periods of natural fires.Elevated levels of atmospheric pollution during 2010-14 were observed at ZSS (natural fires in summer 2010), ZOTTO (wildfires in summer 2011 and 2012) and in Beijing (heavy air pollution episodes). The number of valid observational days at ZOTTO and ZSS was relatively low (33 and 26, respectively), while at Beijing it was high (at about 301 days), during 2010-14 (Wang et al., 2014; Golitsyn et al., 2015).
The main feature of the relationships between the ground- and satellite-based data under heavily polluted conditions is an increase in the slope of the regression line K, as compared with that on less polluted days. This feature is inherent for all satellite products related to CO, and the CO TC in such abnormal cases may increase by several times. Industrial emissions and natural fires usually take place in the lower troposphere, mostly in a layer that is around several hundred meters thick. Therefore, air pollution at the scale of several tens of kilometers is relatively uniform (Fokeeva et al., 2011; Golitsyn et al., 2011; Yurganov et al., 2011; Safronov et al., 2015).
From Table 2, no evident difference in the correlation coefficients between diurnal ground- and satellite-based CO TC could be found at the less-polluted sites (ZSS and ZOTTO in the absence of fires, and Peterhof) or under heavily polluted conditions. However, the regression line slope, K=(U gr-A)/U sat in cases of natural fires in central European Russia and Siberia (ZOTTO and ZSS), and during pollution episodes in winter in Beijing, significantly increased (except based on MOPITT v6 Joint for Beijing). For example, K was 1.08 in summer in Beijing (AIRS v6) for domain 1°× 1°, whereas in winter it was 1.63. Unfortunately, seasonal analysis for MOPITT cannot be presented because of the poor statistical reliability of MOPITT measurements.
2
4.5. Intercomparison of satellite products
When comparing the CO TC daily average values of MOPITT v6J and IASI MetOp-A with AIRS v6 data, a positive correlation between satellite data (averaging over 1°× 1°) was obtained for all sensors (R2=0.25-0.6), depending on the location (Table 3). In comparison with other sites, a relatively low slope of K=0.47 for MOPITT v6, as compared with AIRS v6, was obtained for the polluted environment of Beijing (Table 3).For quantitative analysis, the PBL height during the measurements was chosen as an indicator. This parameter was obtained at each site using GDAS meteorological fields [Air Resources Laboratory, 2014 (http://ready.arl.noaa.gov/HYSPLIT.php. http://arlftp.arlhq.noaa.gov/pub/archives/gdas1)]. The results of the comparison, shown in Table 4 and illustrated in Fig. 6, in which cases with low PBL height (<500 m at ZSS and Beijing, and 400 m at Peterhof) were excluded, indicate that the degree of correlation between ground- and satellite-based CO data at all stations increased significantly.
Figure6. Relationship between the diurnal CO TC data from ground-based spectrometers, as well as AIRS v6 (domain 1°× 1°; sites ZSS and Beijing, 2010-14), and the PBL height: (a) for all days during the measurement period; (b) for days with PBL height ≥ 500 m.A general decreasing trend of CO TC was found at the background sites (ZSS and Peterhof), as well as in urban Beijing and Moscow, during 1998-2014. The level of CO TC in Beijing was significantly higher than that in Moscow. Good agreement, with a systematic discrepancy of less than 3%, between the average annual CO TC values measured at ZSS and Peterhof was obtained.
In most cases of less-polluted conditions, a significant correlation (R2=0.43-0.84; averaged over 1°× 1°) between the CO TC data from ground-based measurements and satellite products (MOPITT v6 Joint and AIRS v6) was found, whereas in heavily polluted areas the satellite products of IASI MetOp-A and AIRS v6 underestimated CO TC by a factor of 1.5-2.8.
The best correlation with the ground-based measurements was obtained for the orbital AIRS spectrometer (version v6), both in the background and polluted areas. The correlation between the average daily satellite CO data and ground-based data increased significantly when the PBL height was greater than 400-500 m. This result was obtained for all selected satellite sensors and ground stations. This is very valuable for the evaluation of emission sources by using satellite remote sensing.
