An example of NO₂ retrieval from two independently calibrated Pandoras, which were initially located at the same site (Yeoju, Korea, 37.3385^∘ N, 127.4895^∘ W), are compared in Fig. 2a showing that the differences in C(NO₂) amounts are less than 0.05 DU, even in the presence of thin afternoon clouds (Fig. 2b) that decrease the measured solar irradiance by more than a factor of 2. Though Yeoju is a relatively clean site in Korea (located to the southeast of Seoul, 37.5644^∘ N, 126.934^∘ W), C(NO₂) amounts frequently reach moderately high values (e.g., 1 DU on 3 June 2016), and occasionally even higher (2–3 DU). However, Yeoju has much less C(NO₂) compared to Seoul, less than 30 km distant, where PSI measurements were found to reach over 3 DU (Fig. 3) during the campaign period from mid-April to early June, 2016. Typical C(NO₂) amounts are 0.3 to 0.5 DU in polluted regions.

Figure 3Frequency distributions of C(NO₂) across the KORUS-AQ PSI network: 20 April to 6 June 2016, except as labeled. The axes vary for different sites.

Download

In a manner similar to Fig. 2a, C(NO₂) amounts can show large variability from day-to-day and intra-day, as well as between different sites. The largest amounts of C(NO₂) are in the north (Seoul and Olympic Park) associated with the largest population and industry concentrations, while the southern cities of Busan and Gwangju have smaller amounts of C(NO₂). The smallest C(NO₂) amounts are at Anmyeondo (an island on west coast of Korea 42 km south of Seoul, usually not downwind of Seoul), and Songchon to the east of Seoul.

Figure 4NO₂ time series vs. day of the year (DOY) and diurnal variability (daily vertical extent) at nine Pandora sites. Notice the very high NO₂ amounts in Seoul and nearby Olympic Park. The black curves are approximate weekly least squares running averages. Note that the vertical scales are different for each site to show the daily variability relative to the running average.

Download

Figure 2b shows the effect of thin clouds in terms of reduced measured count rates for a single spectrometer pixel near 500 nm showing a near noon count rate of 1.26×10⁷ counts s⁻¹ followed by a reduced count rate as clouds move in front of the sun. The cloud plus aerosol cover estimate is from the same date 3 June 2016 as the C(NO₂) amounts shown in Fig. 2a. The effect of thin clouds for C(NO₂) retrieval (Fig. 2a) is increased noise (reduced precision) with a very small impact on accuracy. There are two effects on PSI observations to consider in association with thin clouds. First, is multiple scattering within the cloud affecting the optical path and effective air mass factor AMF. This has a very small effect on AMF, as most of the NO₂ is near the surface well below the clouds. Second, is the reduction in the number of measurements during a fixed 20 s measuring period causing a decrease in the signal to noise ratio. The weather during the campaign was occasionally very cloudy, which caused some missing NO₂ and O₃ data. However, most of the cloudy days were light to moderate cloud cover, which permitted C(NO₂) amounts to be determined, but with lower precision compared to clear-sky direct sun measurements (e.g., Fig. 2a and b). When the cloud cover becomes sufficiently thick, precision is reduced (increased point-to-point scatter) and the spectral fitting error increases. A small percentage of data points with high retrieval error, C(NO₂Error) > 0.1 DU, have been removed from the data set.

Figures 3 and 4 summarize all of the Pandora C(NO₂) data obtained during the KORUS-AQ campaign. Figure 3 presents histograms in percent frequency of occurrence for all nine sites. All of the sites located within or downwind of major cities have production of NO_x mainly from transportation and power generation as its major sources. The ratio of transportation NO_x production is a factor of 3 larger compared to all other sources (Kim et al., 2013). Of these sites, the Anmyeondo PSI frequently (40 %) retrieves values of C(NO₂) that are close to the typical stratospheric values of 0.1±0.05 DU. Other sites occasionally have clean days with similarly low values.

Figure 5NO₂ amounts vs. day of the year (DOY) and local time (KST) for six sites as labeled in each panel. Day 120 is 29 April, Day 130 is 9 May, Day 140 is 19 May, Day 150 is 29 May, Day 160 is 8 June, Day 170 is 18 June.

Download

The Seoul site frequently has amounts of C(NO₂) greater than 2 DU. The same is true of Olympic Park, located in the eastern part of the Seoul metropolitan area. For locations increasingly distant from Seoul, the amount of C(NO₂) decreases in response to smaller local emissions, as the short chemical lifetime of NO₂ normally precludes long-distance transport. Compared to Seoul, the two smaller southern cities, Gwangju and Busan, have relatively low levels of C(NO₂) on most days, with the most typical values ranging from 0.3 to 0.5 DU, although high values exceeding 2 DU can occur on rare occasions.

Figure 4 shows the same data as Fig. 3, but in the form of a time series covering the KORUS-AQ period. The daily variation (at least one point every 2 min) is shown in the vertical extent corresponding to each day's data. Figures 3 and 4 show that sites near the Seoul metropolitan (e.g., Olympic Park) area have larger amounts of pollution compared to those further away (Taehwa, Songchon, and Yeoju). Even though average C(NO₂) amounts are much lower at Songchon and Yeoju, there are times when the pollution levels are quite high (C(NO₂) > 2 DU, Figs. 4 and 5). There are days when the amount of C(NO₂) gets close to 4 DU in Seoul, 3 DU in Olympic Park and Busan, and 4 DU for one day in Yeoju (27 April). The southern cities, Busan and Gwangju are much less polluted on average, which results in a much smaller effect on adjacent regions. Busan is located on the southeastern coastline, so that some of its NO₂ pollution dissipates over the ocean, except for occasional days when very high amounts (3 DU) occur. Anmyeondo is quite clean, as it is located on the western coast well south of Seoul. The most frequently occurring C(NO₂) value at Anmyeondo is 0.15–0.2 DU, which means that the measured NO₂ amount are partly from the stratosphere (0.1±0.05 DU) with very little tropospheric or boundary layer NO₂. There are occasional C(NO₂) plumes that could be from industrial activity to the north, and, perhaps, from China. Transport of NO₂ from China occurs episodically in significant amounts (Lee et al., 2014).

Grouping the diurnal variation together from multiple days (Fig. 5) reveals a pattern to NO₂ emissions and accumulation related to the main NO₂ emission sources (automobiles and power generation) for the three largest cities in Korea: Seoul (Pan40), Busan (Pan17), and Gwangju (Pan26). For Seoul, the amounts of C(NO₂) during the morning (1 DU at 10:00 KST; all times are specified as Korean standard time (KST) in this paper) are much lower than later in the afternoon (over 2–3 DU at 16:00) on almost every day with values occasionally reaching as high as 6 DU. Even the relatively low morning values of C(NO₂) represent a significant amount of pollution. The 6 DU C(NO₂) amount in Seoul is unusual, but coincides with the peak values frequently occurring in the late afternoon. C(NO₂) behavior at nearby Olympic Park to the east of Seoul is very similar to Yonsei University in the heart of Seoul, even though Olympic Park's traffic density is lower than Seoul. Olympic Park is close enough to the metropolitan Seoul area for the transport of NO₂ combined with local production from traffic to produce a very similar diurnal pattern. The moderately large city of Busan also has high values of NO₂, occasionally reaching 3 DU in the afternoon. Busan has relatively low values of NO₂ in the morning, having peaks in the mid-afternoon and declining in the late afternoon. Gwangju, located in the southwest, is a smaller city with less pollution (peak values = 1.6 DU) and does not have as distinct an afternoon maximum.

The panels in Fig. 5 for Taehwa Mountain and Anmyeondo show regions outside the Seoul metropolitan area that still show substantial amounts of NO₂. Compared to Seoul, the Taehwa site is a semi-rural location with only a modest amount of car traffic in the immediate area. However, there are major highways about 6 km from the site that are close enough to permit transport of NO₂ to the Taehwa Mountain site. All of the sites showed a tendency to have peak NO₂ occur in the late afternoon. Anmyeondo on the west central coast of Korea shows C(NO₂) amounts that are quite low with occasional plumes arriving from the north or the west (China).

The basic daily pattern of C(NO₂) in urban Korea arises from large amounts of automobile traffic and power plants emitting NO_x (for modern automobiles, roughly 99 % NO and 1 % NO₂). An FTIR analysis of automobile exhaust shows that NO is emitted at 127 ppm, NO₂ at 1.6 ppm, HCHO at 39 ppm, and CH₃OH at 139 ppm as part of the main emissions containing H₂O (144 ppm) and CO₂ (122 ppm) (https://tools.thermofisher.com/content/sfs/brochures/D10248~.pdf, last access: 31 July 2018; see also Walters et al., 2015).

Figure 6Approximately 1 year of daily column C(NO₂) amount data (a, d) and the monthly running average amount (dark plot in panels a and d). The data are from GIST at Gwangju and Amnyeondo. Panels (a) and (d) are the original time series with one data point every 80 s, panels (b) and (e) are the deseasonalized time series. Panels (c) and (f) are an expanded scale of the monthly running averages M(t) of C(NO₂) that are identical to the solid lines in panels (a) and (d). The vertical extent (a, b, d, e) on a given day is the range of diurnal variation from early morning to late afternoon.

Download

NO quickly converts into NO₂ in the presence of ozone and volatile organic compounds VOCs in the atmosphere and can convert back to NO by solar photolysis. KORUS-AQ results frequently show increasing NO₂ during the day with peaks in the afternoon. For these days the measurements imply that the amount of locally produced NO_x and conversion into NO₂ dominates the losses of NO₂ through photolysis and transport out of the region. Other days occasionally show a different behavior, with NO₂ peaks in the morning and a decline thereafter suggesting transport out of the region.

Some of the sites used for the KORUS-AQ campaign (Gwangju and Amnyeondo) had PSIs set up in April 2015, about 1 year before the start of the campaign. Two other sites (Seoul and Busan) have PSI C(NO₂) data starting in 2012. The extended data sets for Seoul and Busan provide the opportunity to estimate 5-year changes in C(NO₂) amount and seasonal dependence.

In Fig. 6, the daily variation over one year at Gwangju and Anmyeondo are evaluated to estimate one year secular trends. The vertical extent in the time series is not noise or uncertainty, but rather the 80 s per data point variability throughout each day (e.g., see Fig. 2). Before calculating linear least squares slopes, the unadjusted time series (grey data points in panels a and d) were deseasonalized (grey data points in panels b and e) by subtracting a function with zero slope derived from a 30 day running average (dark line in panels a and d or the identical curves in panels c and f). The running average curves in panels (a) and (d) are shown with expanded scale in panels (c) and (f) to clearly show the seasonal variation. The “zero slope functions” ZM(t) are obtained by subtracting a linear least squares fit L(t) to monthly running average curves M(t) in panels (c) and (f) to form zero slope functions $\mathrm{ZM}\left(t\right)=M\left(t\right)-L\left(t\right)$. The results ZM(t) are functions that look similar to the M(t) plots in panels (c) and (f), but with zero slopes. The resulting ZM(t) are then subtracted from the respective original time series (grey circles) in panels (a) and (d). The results are the grey circles in panels (b) and (e). Similar monthly running means are shown in panels (b) and (e) that have almost no monthly variations (see Appendix Fig. A1).

The linear trends in Fig. 6b and e suggest that there was an increase in pollution levels in Gwangju and Anmyeondo over the period of observation. The southern city of Gwangju (Pan 26) has higher average C(NO₂) amounts, 0.34±0.19 DU, compared to the relatively clean coastal site Amnyeondo, 0.26±0.14 DU. Gwangju seasonal cycle has a minimum in C(NO₂) amount in September–October and a very broad maximum from December to May. The Gwangju PSI is located away from major city traffic on a university campus (Gwangju Institute of Science and Technology, GIST) so that the average amount of NO₂ (about 0.34 DU) is moderate with some days reaching 1.5 DU. The slopes are statistically significant at the 2-standard deviation level (p<0.05) and imply that C(NO₂) was increasing at a substantial rate. However, the period of observation was too short to estimate multi-year long-term trends. Additional long-term monitoring of these sites would be desirable for air quality purposes.

The PSI on Anmyeondo was located away from a commercial area with moderate traffic and very near the shore of the Yellow Sea at a regional Global Atmosphere Watch (GAW) station. For Amnyeondo there is a clear seasonal cycle similar to that in Gwangju with a minimum in September–October and a broad maximum during the winter–spring months. Amnyeondo had an average amount of 0.25 DU, which is lower than observed at Gwangju.

Figure 7(a) C(NO₂) time series at Yonsei University in Seoul NO₂(grey) and (b) deseasonalized time series. Combined slope $=-\mathrm{0.05}\pm \mathrm{0.001}$ DU yr⁻¹ and mean $=\mathrm{1.2}\pm \mathrm{0.8}$ DU or the decrease is $-\mathrm{4}\pm \mathrm{0.08}$ % yr⁻¹. Seoul has no clear seasonal cycle.

Download

Figures 7 and 8 each contain an approximately 5-year daily time series (grey) for Seoul (Yonsei University) and Busan (Pusan University) and a linear fit to a deseasonalized version of the time series. As the observations at both sites had an extended period of missing data, the slopes were estimated separately for each segment and for the combined time series. Both Seoul and Busan show a steady reduction in NO₂ air pollution with an average reduction of about −4 % per year. A recent paper by Duncan et al. (2016) estimated a decrease in C(NO₂) for Seoul in about a 10×10 km box of about 1.6±1.4 % per year over the 2004 to 2013 period based on a 2014 average C(NO₂) amount of 0.6 DU, or about half of the average value 1.3±0.8 DU observed by the PSI. The larger reduction in C(NO₂) measured by the PSI is caused by a reduction in higher than average afternoon C(NO₂) amounts that are rarely observed by OMI overpass at 13:30 KST. OMI is a polar orbiting push broom hyperspectral instrument (300–500 nm with a spectral resolution of 0.45 nm in the UV and 1 nm in the visible, and a spatial resolution of 13×24 km²) onboard the AURA satellite. The high observed late afternoon values are not restricted to Seoul, but occur for all of the urban areas where the PSI has been deployed. The high late afternoon values do not regularly occur in remote rural areas such as Amnyeondo.

Figure 8(a) Pusan University in Busan C(NO₂) daily time series (grey) and (b) deseasonalized time series with linear trends.

Download

Figure 9(a) Comparisons between the daily values of C(NO₂) for OMI (black) and PSI (red) at Seoul and Busan for a 5-year period. Solid lines show the average seasonal variation (Lowess(0.1)), see also panel (b). Linear interpolation is used where there are missing data points. (b) Comparisons between the seasonal averages for C(NO₂) from OMI (black) and PSI (red) at Seoul and Busan for a 5-year period. The lower panels show the seasonal difference between the PSI and OMI. The individual data points are shown derived from a Lowess(0.1) smoothing, approximately a 3-month running averages of the daily data. Interpolation has been used where there are missing data points.

Download

Seoul and Busan C(NO₂) measurements are remarkable for the large peak amounts that are seen on many days compared to the 1.5 to 2 DU peak values for Gwangju and Amnyeondo. For Yonsei, the peak values range above 5 to 6 DU in the years 2012 to 2015, but decrease somewhat in 2015 to 2016. In 2015–2016, the decrease appears to be large, but is only 0.2 DU relative to a mean of about 1.2 DU. A smaller decrease appears for Busan (Fig. 8) relative to a mean of about 0.6 DU. All of the PSI measurements show very high values of NO₂ during almost every day when measurements were possible. As the NO₂ concentrations represented by these large column amounts are probably in the boundary layer near the sources of NO₂, there is a strong effect on local air quality.

Seoul and Busan have 5-year PSI data records (Fig. 9a and b), and Gwangju has a 1-year data record (Figs. 6 and 10) spanning the KORUS-AQ campaign. The PSI C(NO₂) can be matched in time (±8 min) with the overpass time from OMI onboard the AURA satellite (mid-day overpass times $\mathrm{13}:\mathrm{30}\pm \mathrm{90}$ min). Figure 9a shows the C(NO₂) daily variation at the OMI overpass time with far more high values of C(NO₂) from the PSI than observed by OMI. The solid lines represent the seasonal dependence, which are shown separately in Fig. 9b along with the C(NO₂) differences, PSI - OMI. The result is that the average PSI values are double those observed by OMI's large FOV (OMI Version 03: https://avdc.gsfc.nasa.gov/index.php?site=666843934&id=13, last access: 31 July 2018).

The seasonal dependence (Fig. 9b) of C(NO₂) from OMI for both Seoul and Busan is fairly regular, with maxima in January of each year and minima in July–August. The seasonal behavior of C(NO₂) obtained from the PSI in Seoul varies with high values extending from January into the summer months and with minima varying from August in 2012, September–October in 2013, missing in 2014, July in 2015, and June in 2016. For Busan, the maxima occur in the spring for 2013 and 2014, October for 2015, and in the spring for 2016. The minima are also variable. The difference between OMI and PSI retrievals depends on local conditions for PSI and on an area average for OMI.

Figure 10C(NO₂) time series from Pandora (red) and OMI (black) for GIST University in Gwangju Korea and their differences. The comparison is formed from time coincidences between Pandora and OMI.

Download

Figure 9b shows that the PSI has a mean difference compared to OMI in Busan of 0.35 DU and peak values (up to 2.5 DU at 13:30 and 4 DU in the late afternoon). The differences are important when considering pollution effects on human health (Krafta et al., 2005; Latza et al., 2009). Even larger differences are observed in Seoul, where the mean difference is 0.58 DU between Pandora and OMI at the satellite overpass time. The results from PSI suggest that local ground-based monitoring of pollution is important for estimating their impact on human health, particularly as amounts of C(NO₂) occurring later in the afternoon exceed the amounts at the time of the satellite overpass.

A comparison with Lowess(0.1) fits (Locally Weighted least squares fit to 0.1 of the data points; Cleveland, 1981) to the matched Pandora vs. OMI overpass data (about 3-month averages) shows that PSI C(NO₂) is larger than OMI measured C(NO₂) mostly because of its much smaller 2^∘ field of view (a circle of 35 m diameter at 1 km altitude) compared to OMI's FOV of 13×24 km² at nadir, which may encompass areas outside of the city or the adjacent ocean areas. For example, the center of Seoul is about 48 km from the Yellow Sea, while the OMI overpass file lists FOV center distances of over 60 km from Seoul. Another possible reason for the differences is that OMI C(NO₂) retrievals use NO₂ vertical profile shape factors from the low resolution ($\sim \mathrm{110}\times \mathrm{110}$ km) Global Model Initiative (GMI) model simulation to calculate air mass factors that are used to determine observed tropospheric NO₂ vertical columns, while much finer resolution profiles are needed to more accurately represent highly polluted urban areas such as Seoul. Increases in OMI retrieved tropospheric column NO₂ up to 160 % are found when using model derived 1.33×1.33 km² profile shape factors (Goldberg et al., 2017). The effect of moderate amounts of cloud or aerosol have little effect on the PSI direct -sun spectral fitting retrieval of C(NO₂) as shown in Fig. 2. OMI and MAXDOAS retrievals are sensitive to the presence of aerosols and clouds (Kanaya et al., 2014), which may contribute to the underestimate of C(NO₂) by OMI even after corrections are made for retrieved aerosol and cloud amounts (Chimot et al., 2016).

The implications for assessing clean air indices suggest that OMI underestimates the human health effect from trace gases such as NO₂, especially in highly populated urban areas. Figure 5 gives a much clearer picture of the degree of pollution than is possible with just the 13:30 OMI comparison measurements, as the late afternoon is the time of maximum pollution.

The city of Gwangju is much smaller than Busan, with less industrial activity, especially automobiles. PSI observations at GIST show much closer agreement with OMI (Fig. 10), especially as GIST is located within the city boundaries, but in an area with much less concentrated industrial activity compared to the center of Gwangju. The large OMI FOV over a relatively clean area reduce the OMI difference in measured NO₂ amount compared to the PSI C(NO₂) amounts. OMI still measures less than the PSI (0.12±0.15 DU), but the mean difference is not statistically significant. However, OMI clearly misses the high values of C(NO₂) that are present in the PSI observations.

Comparison with 4STAR DC-8 overpass data

C(NO₂) results were obtained by the Spectrometer for Sun-tracking Sky-Scanning Atmospheric Research (4STAR) flown on-board the DC-8 during KORUS-AQ and compared with the PSI (Fig. 11). The 4STAR is an airborne sunphotometer, capable of measuring total C(NO₂), C(O₃), water vapor and AOD columns in its direct-sun mode (Segal-Rozenhaimer et al., 2014; Shinozuka et al., 2013), which is similar to the mode used by the PSI network.

A detailed description of 4STAR is given in Dunagan et al. (2013). In brief, the instrument has two structurally rigid grating array spectrometers that are combined to yield continuous spectra between 300 and 700 nm. The instrument sampling rate is 1 Hz, and the nominal integration time used for C(NO₂) retrievals is 50 ms (with six spectra averaged per one sampling period). Dark counts are measured every 20 min using a shutter mechanism. The 4STAR light collection system has fiber optic bundle foreoptics that is connected to the spectrometers. A two axis motion control system with analog feedback provides active tracking of the solar disk. The instrument full field of view (FOV) is ∼1.25^∘. C(NO₂) is retrieved following a method described in Segal-Rozenhaimer et al. (2014), but using the 460–490 nm spectral range. A series of 4STAR columnar NO₂ values above aircraft (for legs below 300 m) taken from DC-8 “missed approach” maneuvers flying over Olympic Park PSI station, within a radius of 5 km, are shown in Fig. 11. There is a relatively good correlation (R²=0.7), with a slight positive bias of 4STAR compared with the PSI values. This might result from higher noise effects (i.e., small amount of spectra averages) for 4STAR during the fast change of altitude when the aircraft performs its “missed approach” overpasses over the PSI stations. Relaxing the altitude constraint to include legs below 500 m showed good agreement with the PSI station at Taehwa Mountain, but with an overall lower correlation coefficient (R²=0.54), which is expected due to averaging of larger vertical range. As with PSI, 4STAR shows better agreement with OMI C(NO₂) for low values of C(NO₂), but considerable differences over polluted areas (Segal-Rozenhaimer et al., 2018), when 4STAR C(NO₂) values are averaged within each of the OMI pixels corresponding to the flight path for each of the days.

Figure 11A correlation plot of C(NO₂) from 4STAR onboard the DC-8 compared to the C(NO₂) amount measured by the PSI at Olympic Park on nine different days. The solid black line is the 1 : 1 line drawn for reference. The dashed line represents the data linear fit, with a slope of 1.05, and a correlation coefficient r²=0.7, as shown on the plot.

Download

Figure 12C(HCHO) from PSI at Olympic Park for 6 days in June 2016. C(HCHO) on 2 June 2016 has a peak value of 2.3 DU at 14:30 KST.

Download

PSI makes two sets of direct-sun measurements every 80 s. One set is for measurements in the visible range (380–525 nm used for NO₂) and the other is for the UV range (290–380 nm with a filter, U340, which blocks visible light). Formaldehyde is derived from the same set of spectral measurements used for ozone (i.e., with a U340 blocking filter), but using the spectral range 332–359 nm. Sources of error in the C(HCHO) retrieval arise from the selection of the fitting window and the amount of C(HCHO) remaining in the reference spectrum after application of the modified Langley estimation (MLE) method of calibration (Herman et al., 2009; Spinei et al., 2018). The MLE extrapolation to zero C(HCHO) could have an offset error of 0.1 to 0.2 DU. Selecting different fitting windows can also cause the C(HCHO) retrievals to differ. For example, a wider alternate fitting window, 324–360 nm, retrieves HCHO values that are about 8 % higher because of different amounts of interference from overlapping absorption by O₃, NO₂, and BrO at the spectral resolution of 0.5 to 0.6 nm currently in use. Absolute offset errors do not affect the retrieval precision (relative column amounts), which is approximately 0.1 DU. A detailed analysis of the algorithms and uncertainties is discussed by Spinei et al. (2018).

The Olympic Park area has much more vegetation than central Seoul for the production of isoprene (http://www.olympicpark.co.kr, last access: 31 July 2018), which is a significant source of the chemicals needed for formaldehyde production in the atmosphere (Luecken et al., 2012). Observations from PSI show that C(HCHO) starts out every day at low levels 0.6 DU at about 08:00 and increases to over 2 DU until 18:00 (Figs. 12 and 13). Most HCHO arises from photochemical production, while a significant fraction is chemically derived from automotive emissions in densely populated urban areas (Friedfeld et al., 2002; Garcia et al., 2006; Lei et al., 2009; Liteplo et al., 2010). Regardless of the precursor source, HCHO forms in the atmosphere primarily though photochemistry, which causes HCHO to usually be at a minimum early in the day, increase into the afternoon, and decline towards evening. The PSI C(HCHO) observations (Figs. 12 and 13) support this pattern of daily variation.

A summary of the daily time dependence of C(HCHO) at Olympic Park during the entire KORUS-AQ campaign is shown in Fig. 13. As in Fig. 12, minimum values are observed in the morning (06:00–08:00) before the chemical and direct sources of HCHO are significant. There is strong buildup during the day that reached a maximum between 15:00 to 16:00, and then diminished towards sunset. As with NO₂, the daily pattern of late afternoon peaking of HCHO amounts presents a problem for polar orbiting satellite observations (e.g., OMI observations at 13:30) assessing air quality.

Figure 13Pandora measured formaldehyde amounts vs. day of the year and local time for 29 April 2016 to 11 June 2016 in Olympic Park.

Download

Figure 14 shows two altitude profiles acquired by the Compact Atmospheric Multispecies Spectrometer (CAMS) (Richter et al., 2015) onboard the DC-8 aircraft as it spiraled over the Olympic Park area on 4 May 2016 in the morning and at midday. Quoting from Richter et al. (2015):

CAMS is a multi-species spectrometer configured for the simultaneous detection of ethane (C₂H₆) and formaldehyde (CH₂O). The spectrometer utilizes a tunable, fiber optically pumped difference frequency generation laser source in combination with a Herriott type multi-pass absorption cell with an effective path length of 89.6 m.

The morning integrated amount on 4 May was 1.02×10¹⁶ molecules cm⁻² (0.38 DU) and the afternoon amount was 6.95×10¹⁵ molecules cm⁻² (0.26 DU), both substantially less than the PSI measured values of 0.48 and 0.42 DU, respectively. There were no surface measurements of HCHO mixing ratio on 4 May at Olympic Park. On 2 June at 11:40 there was a surface measurement 3.94 ppb. Including the surface measurement in the profile integral yields Integ(0.026, 7.2 km) = 0.55 DU, while PSI measured 1.2 DU, which is consistent with the differences shown in Fig. 14. The notation in Fig. 14 is $\mathrm{Integ}(z\mathrm{1},z\mathrm{2})=\underset{z\mathrm{1}}{\overset{z\mathrm{2}}{\int }}\mathrm{HCHO}\left(z\right)\mathrm{d}z$ for the altitudes z1 to z2, where z is in km.

Figure 14HCHO altitude profile measured by CAMS onboard the DC-8 on 4 May at 07:54 KST (a) and 11:54 KST (b) over Olympic Park, Korea. (c) PSI measurements of total column HCHO. Vertical bars (c) mark the DC-8 flight duration for the profiles yielding altitude integrated column amounts of 0.38 and 0.26 DU.

Download

Figure 15DC-8 CAMS HCHO mixing ratio measurements over Olympic Park on 4 June. The continuous blue profiles show the 1 s HCHO data while the black points with error bars show the 10 s average and standard deviation of this data at points of closest approach above the Olympic Park site.

Download

The profiles used data for lower altitudes obtained from aircraft “missed approach” maneuvers at a nearby Seoul Airbase, 8.5 km from Olympic Park, (Fig. 15). When available, a single surface altitude point was added using ground-based volume mixing ratio measurements obtained from US Environmental Protection Agency measurements using quantum cascade laser instruments (Fried et al., 2008; Hottle et al., 2009; Richter et al., 2015; Spinei et al., 2018 and references therein). The DC-8 minimum altitude exactly over Olympic Park was typically around 0.4 km above the surface (black circles Fig. 15). Large vertical CAMS retrieved HCHO gradients were observed as the DC-8 descended to lower altitudes over Seoul Airbase. A comparison of 10 s DC-8 HCHO averages at the points of closest spatial approach to the Olympic Park (black circles) site on 4 June, for example, to peak HCHO measurements during missed approaches at the nearby Seoul Airbase (20–40 m above the ground) revealed ratios in the observed HCHO (black circles) ranging between 75 % to 83 % of the maximum values near the surface. As Olympic Park DC-8 overpasses miss significant near-surface HCHO amounts, the profiles shown in Figs. 14 and 16 incorporate the HCHO amounts down to the surface at an altitude of 0.026 km a.s.l (above sea level) derived from the “missed approach” at Seoul airbase. HCHO measurements above the maximum altitude over Olympic Park (see Figs. 14 and 16) were taken from the closest time over the Taewha Mountain site, 28 km from Olympic Park. The assumption is that the horizontal gradients above 2.2 km (Fig. 15) can be neglected.

After conversion from mixing ratio to molecules cm⁻³ using the measured atmospheric density, the resulting profile data were integrated from the minimum (0.026 km asl, Table 1) to the maximum heights indicated in Fig. 14. The result is 0.38 DU at 07:54 and 0.26 DU at 11:54 compared to the measurements from the Pandora instrument 0.48 and 0.38 DU The derived vertical HCHO columns from the DC-8 data in Fig. 14a and b are 79 % of PSI measured C(HCHO) in the morning and 68 % of PCI C(HCHO) at midday (Fig. 14c).

A similar comparison is shown in Fig. 16 for 5 June 2016 where the amount of C(HCHO) is much larger than on 4 May. Integration of the measured profiles yields column densities of 0.60 and 0.82 DU at 08:30 and 15:21 h. For this case, at both times the CAMS DC-8 values are about 77 % and 63 % of the PSI measured column amounts, 0.78 and 1.3 DU. For both cases in Figs. 14 and 15 the 23 % to 37 % differences are outside of the expected error from PSI fitting window selection and from residual HCHO included in the MLE calibration method.

Figure 16HCHO altitude profiles measured by CAMS onboard the DC-8 on 5 June at 08:30 KST (a) and 15:21 KST (b) over Olympic Park, Korea. (c) PSI measurements of total column HCHO. Vertical bars mark the DC-8 flight duration for the profiles yielding column amounts of 0.60 and 0.82 DU.

Download

Another Olympic Park case on 9 June 2016 shows CAMS = 0.79 vs. PSI = 1 DU at 08:06, CAMS = 0.74 vs. PSI = 1.3 DU at 12:12, and CAMS = 1.13 vs. PSI = 1.9 DU, or the CAMS measurements on the DC-8 are 79 % and 57 % less than the PSI total column HCHO. All of the remaining comparisons of CAMS DC-8 profile results with PSI C(HCHO) show similar results. The reasons for the disagreement between C(HCHO) measured by direct sun observations (PSI) and the integrated column density from aircraft measurements of HCHO VMR are not known. Contributions to the differences include the selection of the PSI wavelength window (332–359 nm) and possible interference from overlapping NO₂ and O₃ absorption that are not properly included, and, more likely, the use of CAMS measured volume mixing ratios at the lowest altitudes from the nearby Seoul airbase, 8.5 km from Olympic Park, where spatial variation may affect the calculation of C(HCHO). The use of Taehwa Mountain data for higher altitudes over Olympic Park contributes 25 % for 3 of the above cases and 50 % for 4 May 2016 at 07:54 (Fig. 14a). This is probably not the reason for the disagreement between CAMS and PSI, as the percent underestimate for CAMS over Taewha is about the same magnitude (Table 2) as over Olympic Park.

Table 2Taehwa Mountain. DC-8 compared to PSI measurements in Fig. 18.

Download Print Version | Download XLSX

PSI measurements show that Olympic Park produces more HCHO almost every day than observed at the Yonsei University in Seoul and Taehwa Mountain sites (Figs. 12, 17, 18). The hourly variations observed during the KORUS-AQ campaign at the Yonsei University in Seoul and at Taehwa Mountain sites are similar to Olympic Park even though most of the HCHO is locally produced by photochemistry, but has a relatively short lifetime of a few hours in polluted air where there is significant ozone and OH. However, at typical wind speeds of 10–20 km h⁻¹ and a chemical lifetime of 2.5 h (Dufour et al., 2009), HCHO can be transported about 25–50 km, which is far enough for some transport of HCHO between the PSI sites at Yonsei, Olympic Park, and Taewha Mountain DC-8 CAMS results over the Taehwa Mountain site compared to PSI are given in Table 2 with differences similar to Olympic Park.

Figure 19a and b summarizes all of the C(HCHO) data obtained during KORUS-AQ at the five sites. The graphs on the left show all of the data points (light gray circles) as a function of the local time and a Lowess(0.1) fit to the data showing the average hourly behavior. The spread of the data about the Lowess(0.1) fit represents the day-to-day variation at a given local time. On average, Mt. Taehwa tends to increase throughout each day, while Yonsei and Olympic Park show maxima at 14:00 and 15:30 KST, respectively. Similarly, in Fig. 18b Yeogju increases during the day having a maximum at 17:42 KST while Anmyeondo has a broad peak with maxima at 12:00 and 13:42 KST.

Figure 17Total column HCHO from Pandora Yonsei University, Seoul for 6 days in June 2016. C(HCHO) on 2 June 2016 has a peak value of 1.2 DU at 13:30.

Download

Figure 18Total column HCHO from Pandora Taehwa Mountain for 6 days in June 2016. C(HCHO) on 2 June 2016 has a peak value of 1.2 DU at 12:45. Δ are DC-8 CAMS measurements on 10 June.

Download

Figure 19(a) Summary of total column HCHO for the stated dates during the KORUS-AQ campaign. The solid line is a Lowess(0.1) fit to the data. The sharp cutoffs in panels (a), (b), and (c) were caused obstructions of the direct sun from the PSI FOV in the afternoon. (b) Summary of total column HCHO for the stated dates during the KORUS-AQ campaign. Panels (a) and (b) represent the daily variation at a given local time. The solid line is a Lowess(0.1) fit to the data. Panels (c) and (d) show the frequency of occurrence (%) for different amounts of C(HCHO).

Download

Figure 20The springtime change in C(HCHO) over about a 40 day period depending on the site. The “vertical bars” are the diurnal variation within each day of data. The thicker red curve is a Lowess(0.3) fit to the data, while the thin red line is a linear least squares fit. The Lowess(0.3) fit is approximately a 10-day local least-squares average.

Download

The histograms on the right side of Fig. 19 represent the percent frequency of occurrence of C(HCHO) in 0.1 DU bins. C(HCHO) at Mt. Taehwa and Seoul rarely exceeds 1.5 DU compared to Olympic Park where C(HCHO) > 2 DU for a significant fraction of time. The most frequent values are 0.6 DU for Seoul, 0.9 DU for Mt. Taehwa, and over 1 DU for Olympic Park. Olympic Park also has a broader distribution towards higher values of C(HCHO) than other sites.

The general intra-day C(HCHO) time dependence and C(HCHO) percent occurrence are shown for two additional sites (Fig. 19b), Yeogju and Amnyeondo. Yeogju shows an increase in C(HCHO) from morning to a peak value of 0.85 DU at 14:42 KST, which then declines after 16:00 KST. In contrast, Amnyeondo is almost symmetric with the sun position, having a maximum of about 0.77 DU near 12:00 and 13:42 KST.

The average change in C(HCHO) during the spring campaign at the five sites is summarized in Fig. 20. Of the sites, Olympic Park showed the largest change rate, 58 % month⁻¹ followed by Amnyeondo at 50 % month⁻¹, then Taehwa (33 % month⁻¹), Yonsei Seoul (25 % month⁻¹), and Yeogju (−13 % month⁻¹). Amnyeondo tends to have lower C(HCHO) amounts because of its relatively isolated coastal location. These 2-month trends include seasonal increases during the campaign months May and June, 2016.

It is difficult to compare PSI C(HCHO) with OMI for the KORUS-AQ period, as OMI overpass C(HCHO) data for 2016 have some missing days (Fig. 21). For days with matching data points over Seoul, PSI C(HCHO) (approximately 0.8 DU) is almost always larger than the OMI values (0.2 DU) plus a few very high PSI values and two high OMI values. The general day-to-day variations are similar.

OMI Formaldehyde HCHO Version 03 data are available from: https://avdc.gsfc.nasa.gov/index.php?site=1113974256&id=81 (last access: 31 July 2018); OMI Nitrogen Dioxide NO₂ Version 03 data are available from: https://avdc.gsfc.nasa.gov/index.php?site=666843934&id=13 (last access: 31 July 2018); Pandora KORUS-AQ NO₂ and HCHO data are available from: https://avdc.gsfc.nasa.gov/pub/DSCOVR/Pandora/DATA/KORUS-AQ/ (last access: 31 July 2018). The OMI data were prepared by the Goddard Space Flight Center OMI processing team located in Greenbelt, Maryland, USA. The Pandora KORUS-AQ data were prepared by Elena Spinei and by Jay Herman.

JH wrote most of the paper and performed the analysis and comparisons with the DC-8 aircraft measurements. ES derived the formaldehyde altitude profiles suitable for comparison with Pandora data. AF obtained the HCHO profile data from the DC-8 CAMS instrument. JhK provided support for the installation of Pandora instruments in Korea. JaK provided support for the Pandora located in Busan. WK provided support in installing the Pandoras and analyzing the raw data. AC provided calibration and data analysis support. NA provided Pandora setup in Korea and provided the maintenance of calibration. MSR provided the 4STAR NO₂ data from the DC-8 flights and the comparison with Pandora.

The authors declare that they have no conflict of interest.