The Korea–United States Air
Quality Study (KORUS-AQ) conducted during May–June 2016 offered
the first opportunity to evaluate direct-sun observations of formaldehyde
(HCHO) total column densities with improved Pandora spectrometer instruments.
The measurements highlighted in this work were conducted both in the Seoul
megacity area at the Olympic Park site (37.5232
Diurnal variation in HCHO total column densities followed the same pattern at
both sites, with the minimum daily values typically observed between
6:00 and 7:00 local time, gradually increasing to a maximum between 13:00 and
17:00 before decreasing into the evening. Pandora vertical column densities
were compared with those derived from the DC-8 HCHO in situ measured profiles
augmented with in situ surface concentrations below the lowest altitude of
the DC-8 in proximity to the ground sites. A comparison between 49 column
densities measured by Pandora vs. aircraft-integrated in situ data showed
that Pandora values were larger by 16 % with a constant offset of 0.22 DU
(Dobson units;
Formaldehyde (HCHO) is a key constituent in tropospheric chemical cycling.
Its abundance is dominated by secondary formation through the oxidation of
methane and non-methane hydrocarbons. It is also short lived, undergoing
photolysis or oxidation by OH within a few hours under typical daytime
conditions. As such, HCHO provides an important indicator of the integrated
oxidation of hydrocarbons that contributes to tropospheric ozone production
in the presence of nitrogen oxides. The degradation of HCHO can also
constitute an important secondary source of
The attributes described above make HCHO an important test species in
evaluating our mechanistic understanding of tropospheric oxidation reactions
as well as a valuable proxy for hydrocarbon emissions. Remote sensing of
HCHO promises valuable insight into the emissions and processes driving
tropospheric chemistry. For instance, satellite measurements of HCHO by the
Global Ozone Monitoring Experiment (GOME)
(Fu et al., 2007; Palmer, 2003; Palmer et al., 2006; Shim et al., 2005), SCanning Imaging
Absorption SpectroMeter for Atmospheric CHartographY (SCIAMACHY) (Wittrock et
al., 2006), and Ozone Monitoring Instrument (OMI) (Marais et al., 2012) have
been used to map the isoprene emissions on a global scale. In combination
with remote sensing of
With the promise of both temporal and spatial information for HCHO on the horizon from a constellation of geostationary satellites (Zoogman et al., 2017), other possible uses for satellite observations of HCHO are emerging. Recent work by Schroeder et al. (2016) suggests that column HCHO shows promise as a proxy for surface ozone. Valin et al. (2016) examine the relationship between column HCHO and its dependence on OH production and VOC reactivity, demonstrating the importance of this information to improving satellite-derived emissions estimates for isoprene and other hydrocarbons. These efforts to further develop and improve the use of future satellite observations elevate the need for ground-based remote sensing to validate satellite-measured HCHO columns. Remote-sensing differential optical absorption spectroscopy (DOAS) has been widely used to measure HCHO from ground (Lee et al., 2005; MacDonald et al., 2012; Pikelnaya et al., 2007; Vlemmix et al., 2015), aircraft (Baidar et al., 2013), and satellite (Bauwens et al., 2016; De Smedt et al., 2015) platforms. The uncertainties of the DOAS-derived HCHO columns are impacted by the DOAS fit uncertainty and the uncertainty in the air mass factors. Validation of such measurements is challenging due to air volume sampling differences between different platforms.
In this paper we present HCHO total columns from DOAS measurements of unscattered direct-sun (DS) photons using NASA/GSFC (National Aeronautics and Space Administration/Goddard Space Flight Center) Pandora instruments and in situ measurements over two sites during the Korea–United States Air Quality Study (KORUS-AQ) conducted in May–June 2016 in South Korea.
Pandora instruments are field grade spectroscopic UV–Vis systems (Herman et
al., 2009). They are part of the growing joint NASA- (USA) and European Space
Agency-sponsored Pandonia Global Network (PGN). The main goal of PGN is to
provide consistent ground-based total
Pandoras deployed during KORUS-AQ were retrofitted with new UV grade fused
silica windows with broadband antireflection coating (ARC, 250–700 nm).
This modification from the earlier versions of Pandora (pre-2016) was
necessary to decrease spurious spectral structure in DS spectra.
This new ARC window improved
The rest of the paper is organized in the following sections. Section 2 describes in detail ground-based (Pandora and in situ) and aircraft measurements during the KORUS-AQ 2016 study. Section 3 explains how HCHO vertical column densities are calculated from the in situ measurements (aircraft and surface) for comparison with Pandora column measurements. Section 4 shows the results by comparing HCHO vertical columns from Pandora, surface, and aircraft measurements. Section 5 focuses on conclusions.
KORUS-AQ fielded a multi-perspective suite of observations including both remote sensing and in situ observations of air quality at ground sites across the peninsula and on research aircraft collecting valuable data on conditions aloft. Pandora spectrometers were used to observe total columns of HCHO at five locations, but two sites in particular also included ground-based in situ measurements of HCHO and frequent atmospheric profiling overflights by the NASA DC-8 aircraft with an in situ measurement of HCHO on board.
The first site was located in the Seoul megacity at Olympic Park
(37.5232
The second site was at Mount Taehwa (37.3123
The instrument consists of a small Avantes low-stray-light spectrometer
(280–525 nm with 0.6 nm spectral resolution with 5 times oversampling)
connected to an optical head by a 400
Pandora spectra are automatically collected and submitted to NASA/LuftBlick
servers for centralized uniform processing by the Blick Software Suite
(Cede, 2017). All standard operational Pandora data products are
available at
Pandoras measure unscattered solar photons in a narrow cone (2.1
DOAS fitting parameters used to calculate HCHO
HCHO total vertical column densities are calculated from Pandora measurements
of unscattered sun photons (with visible light blocked by a U340 filter)
using the DOAS technique (Platt and Stutz, 2008). The analysis consists of
the following steps, described in detail by Cede (2017):
Correction of the DS collected spectra (level L0 to L1) for dark
current, charge-coupled device (CCD) nonlinearity, latency effect, pixel response nonuniformity,
filter transmission, instrument temperature sensitivity, stray light,
wavelength shift, etc. Selection of the reference spectrum, ideally, a Pandora-measured
reference spectrum with the smallest possible HCHO absorption and highest
signal-to-noise ratio. In this study all spectra with low measurement noise
collected around local noon ( Calculation of HCHO differential slant column densities ( The fitting window used in this study to calculate HCHO columns is
332–359 nm ( Calculation of the air mass factor for DS observation geometry
(AMF Estimation of HCHO slant column density in the reference spectrum
(SCD Calculation of the HCHO VCD (L2 Pandora
data):
The total error in the Pandora direct-sun HCHO VCD
(
Pandora HCHO total column error budget from direct-sun measurements.
Uncertainties in Selection of a fitting scenario (332–359 nm; see Table 1) can result in an
error of Uncertainties due to the laboratory-measured high-resolution molecular
cross sections ( Uncertainty due to extraneous spectral structure (ESS) in DS spectra (even
with the new ARC window) is harder to evaluate and will be the subject of future
studies. Figure 2b shows an example of common optical depth residuals
calculated by the DOAS fitting algorithm of 4537 cloud and spatial
stray-light-free DS measurements and scaled by DS AMF. Figure 2c illustrates the
effect of this residual spectrum on the retrieval of 0.5 DU (background
levels) of HCHO. Some of this common residual spectrum is potentially due to
ESS. At this point we estimate that the error due to ESS is on the order of
0.025 DU. Uncertainties in the SCD Uncertainty in the DS AMF is less than 1 % at SZA smaller than
80 We estimate that the total error in DS Pandora HCHO total column measurements
during KORUS-AQ is [ Figure 2a demonstrates dependence of the total HCHO error on the measurement
time (AMF) according to Eq. (4). The V shape is mostly due to the error in
SCD
Surface HCHO concentrations were measured at Mt. Taehwa and Olympic Park by tunable infrared laser direct absorption spectroscopy (Li et al., 2013) with quantum cascade lasers at mid-IR wavelengths (QC-TILDAS from Aerodyne Research, Inc.). In situ HCHO measurements were conducted by the US Environmental Protection Agency (EPA) at the Olympic Park research site, and by Aerodyne Research, Inc., at the Mt. Taehwa site.
Light from 1765 cm
Absorption measurements were made relative to a zero-air background gas
obtained from an ultra-high-purity zero-air gas cylinder. Backgrounds were
taken through the same inertial inlet used to measure samples. A 30 s
background (with a 15 s flush time) was taken every 10–15 min. Nitrogen
(
Spectra were averaged for 1 s intervals and fit using a nonlinear least squares fitting algorithm, with parameters based on the HITRAN database (Gordon et al., 2017). One-second HCHO data were averaged to 10 and 60 s averages to improve precision. The Allan deviation (estimate for precision) is 0.100 ppb for 10 s HCHO data and 0.060 ppb for 60 s data. Estimated accuracy is approximately 10 %.
Figure 3 shows time-coincident in situ surface HCHO volume mixing
ratios (vmr) at Olympic Park and
Mt. Taehwa. The average vmr during the campaign at Mt Taehwa was
The Compact Atmospheric Multispecies Spectrometer (CAMS) is a dual-channel
infrared laser absorption spectrometer that provided measurements of HCHO
with 1 s time resolution on the NASA DC-8. A comprehensive description of
CAMS can be found in Richter et al. (2015). Briefly, mid-IR laser light at
3.53
There were a total of 20 local flights of the DC-8 over Korea from 2 May to 10 June 2016. As described earlier, flights included routine overflight of the two sites as well as vertical profiling in their vicinity multiple times per day. Figure 4 shows a summary of all flight trajectories and measured HCHO over and near the two sites. We “assigned” data collected below 3 km to a respective site if the ground distance from the site to the aircraft was less then 15 km (Fig. 4a and d). This resulted in a total of 38 DC-8-measured profiles over Mt. Taehwa and 43 over Olympic Park.
Summary of all DC-8 flights over Olympic Park and Mt. Taehwa:
Most DC-8 measurements directly above the Mt. Taehwa site were done at
an altitude of 0.6–1 km a.s.l., reaching the minimum altitudes in a narrow path
when approaching the site from the west and descending to the east before
conducting the spiral ascent (Fig. 4d). In-line overpasses over Olympic Park
extended to a maximum height of 2–3 km north of the site with a variable
minimum altitude (
Correlation between surface in situ and near-surface DC-8
measurements at Olympic Park
Figure 5 shows the linear correlation between the in situ ground-based
measurements at each site and the aircraft measurements averaged over the
lowest 200 m in proximity to the Olympic Park site and the lowest 600 m a.s.l. near
the Mt. Taehwa site. The total duration of flight time needed to sample the
corresponding vertical distances was between 0.5 and 3 min. This resulted in
ground distance coverage of
The absolute difference between the averaged near-surface DC-8 and in situ
measurements was
To account for the partial column between the surface and the lowest aircraft
altitude, we complement DC-8 profiles with the in situ surface measurements.
Air density at the surface was calculated from the Lufft WS501 measurements
of temperature and pressure at Mt. Taehwa. There were no pressure
measurements at Olympic Park, so we scaled pressure from Osan Air Base to the
Mt. Taehwa altitude. Temperature measurements at Olympic Park did not cover
the entire campaign, so we used temperature from Mt. Taehwa (
Total columns from DC-8 HCHO profiles were determined by numerical
integration of the volume number density from the lowest to the highest
altitudes. Errors in derived DC-8 HCHO total columns are comprised of the
instrumental uncertainties of the measurements, errors in temperature and
pressure profiles, errors due to spatial and temporal heterogeneity of the
HCHO distribution in the sampled air relative to the specific volume over the
site, and errors due to extrapolation to the parts of the atmosphere not
sampled by the aircraft. In this study we approximate errors due to spatial
and temporal heterogeneity of the HCHO distribution by comparing DC-8
measurements within the lowest 200 m for Olympic Park and 470 m for Mt.
Taehwa to the in situ surface mixing ratios (see Fig. 5). This uncertainty
source leads to a potential underestimation of 8 % for Olympic Park and
19 % for Mt. Taehwa. Instrumental errors are random and are on the order of
4 %–6 %. We assume that the uncertainty in the partial column above the DC-8
is 50 %, which translates to about 2.5 % of the total column. We assume
that the uncertainty in the partial column below the DC-8 minimum altitude is
dominated by the uncertainty due to heterogeneity. Another source of error in
the calculated columns over Olympic Park is the potential heterogeneity above
the highest DC-8 altitude above Olympic Park (2–3 km) and Mt. Taehwa.
When all these sources are considered, the total error in derived VCD from the
aircraft measurements is about (
Mixing layer height (MLH) above ground level derived from Vaisala Ceilometer CL51 backscatter profiles (910 nm) over Olympic Park and Mt. Taehwa during KORUS-AQ.
Given the broader availability of Pandora observations and surface HCHO measurements without the benefit of complementary airborne sampling, we also developed estimates for column densities depending only on in situ surface measurements and information on mixing layer height (MLH) derived from Vaisala Ceilometer CL51 backscatter profiles at 910 nm (Knepp et al., 2017). The main assumption is that most of the HCHO column is located in the well-mixed layer. Figure 6 shows MLH derived from the backscatter profiles at Mt. Taehwa and Olympic Park. The estimated MLH diurnal changes are very similar at both sites. The minimum MLH (300–500 m) is during the night and early morning hours (22:00–8:00). Planetary boundary layer growth typically starts around 7:00–8:00 in the morning and reaches its maximum (1.5–2 km) around 15:00–16:00 local time. On some days, however, the estimated MLH peaks later (around 18:00) and is significantly higher (around 3 km). Measured MLHs, however, are somewhat lower at Mt. Taehwa compared to Olympic Park in the morning and late afternoon. Diurnal changes in ceilometer-measured MLH have the same trend as the diurnal changes in the vertical distribution of HCHO measured from the aircraft (see Fig. 4c, f and Sect. 2.3) confirming our assumption.
To estimate the total column from in situ surface concentrations, we (1) filtered and averaged the MLH data for both sites to generate “measured” MLH and (2) created a median MLH as a function of local time of day from all measurements. A median MLH was used to test the hypothesis of whether a “generic seasonal” estimation of MLH can be applied to relate in situ surface and column HCHO measurements.
Ceilometer-measured MLH can exhibit sporadic variations that are not related
to the true changes of MLH. We have examined effect of several filtering
schemes on the total columns: (1) no filtering with 5 min averaging of raw
MLH; (2) running median (
We calculated total columns from in situ measurements (ground-up VCD) using
four different profile shapes: (1) a uniform HCHO mixing ratio up to the median
MLH with a free-tropospheric mixing ratio of 0.23 ppb from the MLH to the
average tropopause height of 12.77 km; (2) same as (1) but using the
“measured” MLH; (3) a uniform HCHO mixing ratio up to the median MLH with a
free-tropospheric mixing ratio that exponentially decreases above the MLH to
0.23 ppb within
Free-tropospheric vmr of
A single temperature and pressure profile for the whole campaign was generated from all available radiosonde measurements. This profile was scaled to account for surface temperature and pressure changes during the campaign within the MLH.
Vertical column densities at Olympic Park during KORUS-AQ derived
from Pandora direct-sun measurements (
Vertical column densities at Mt. Taehwa during KORUS-AQ derived
from Pandora direct-sun measurements (
Total columns from Pandora direct-sun measurements, DC-8 aircraft profiles, and surface
measurements (using four profile shape assumptions) are shown in Fig. 7 for
Olympic Park and Fig. 8 for Mt. Taehwa. Days with no or limited data (e.g.,
cloud-screened Pandora data) were excluded from these figures. Diurnal
changes for all of the column estimations show similar trends with minimum
VCD typically in early morning and maximum VCD around 14:00–16:00 local time.
Figures 7 and 8 show the effect of the assumed profile shapes on the derived
“ground-up” columns. As expected, the profile shapes (2 – grey) and (4 – light
blue) that use measured MLH result in the largest VCD when MLH is larger than
the median values. This is obvious on 19 May 2016 (Olympic Park), when
measured MLH in the afternoon was 3 km compared to a median MLH of 1.5 km
(Fig. 9). Considering that exponential function addition to the box shape is
limited to 4 km (or
Vaisala Ceilometer CL51 backscatter measurements (910 nm) on 19 and 20 May 2016 at Olympic Park, during KORUS-AQ. Black dots represent MLH used in this study.
Ground-up and Pandora columns both exhibit similar HCHO changes on a
smaller scale (e.g., 20 May 2016 around 18:00 at Olympic Park). The absolute
values, however, are different. In addition, Pandora total columns tend to have
a smaller rate of change between 6:00 and 10:00 in the morning compared to the
ground-up columns at both sites. This could be an indication of
underestimation of Pandora SCD
DC-8-integrated columns tend to be within the variability of the
ground-up columns from the four profile shapes and are typically smaller
than the Pandora measurements. Figure 10a shows linear regression between
Pandora and DC-8 HCHO columns at the two sites, with the slope equal to
Correlation between HCHO columns for
Linear correlation between HCHO total columns from Pandora
direct-sun measurements and columns calculated from in situ surface concentration
measurements based on different profile shape assumptions (
Figure 10b shows linear regression analysis results for ground-up columns
best agreeing with the DC-8 columns (box profile shape with measured MLH,
2). This profile shape has a linear regression correlation
Based on the DC-8-measured HCHO profile discussion and diurnal changes in the ceilometer-determined MLH, we do not expect any meaningful correlation between the Pandora total columns and in situ surface concentrations. Indeed, Fig. 11a and c show a general correspondence between surface HCHO measurements at Olympic Park and Mt. Taehwa and Pandora column measurements, but the relationship is too diffuse to allow surface values to be derived from column measurements or vice versa.
Pandora-measured HCHO vertical column densities vs. surface in situ mixing ratios and columns calculated from the surface vmr at Olympic Park and Mt. Taehwa during KORUS-AQ (May–June 2016).
Linear regression analysis in Fig. 11 was done between Pandora (
Linear correlation between HCHO total columns from Pandora
direct-sun measurements (
Table 4 summarizes linear regression results for all profile shapes. Standard
deviations in Table 4 for the profile shapes with measured MLH represent the
effect of different filtering of MLH data. In general, the effect of MLH
filtering is very small. For columns derived from the box and exponential
profile shape with the measured MLH and only 5 min MLH averaging, the
correlation with Pandora columns (
The correlation between Pandora and ground-up columns at Mt. Taehwa is
worse than at Olympic Park since there were fewer full-day Pandora
measurements at Mt. Taehwa because of instrumental issues early in the
campaign. There were several days that had only morning Pandora measurements.
During morning hours measured MLH was relatively shallow (
We have presented a first evaluation of Pandora total column HCHO measurements collected in continuous direct-sun-observation mode during the KORUS-AQ 2016 field study. The total column measurements were compared to the integrated DC-8 in situ profile measurements and in situ scaled columns assuming different profile shapes.
The following observations were made.
The largest sources of uncertainty in Pandora HCHO DS column measurements
are from the following:
Systematic errors due to selection of the fitting window and choice of
the cross sections. The combined error is on the order of Estimation of SCD The statistical HCHO total column errors were DC-8 in situ profile measurements were done over limited altitude ranges.
On average the DC-8-integrated columns were complemented with (a) about DC-8 in situ profile measurements ( DC-8 measurements in the lowest 200 and 470 m above Olympic Park and
Mt. Taehwa were on average 8 % and 19 % lower than the time-coincident
surface concentrations at the corresponding sites indicating spatial
(vertical and horizontal) heterogeneity of HCHO distribution within
15–20 km. Pandora HCHO total columns were on average 16 % larger than
DC-8-integrated profiles with an offset bias of 0.22 DU and correlation
coefficient ( DC-8-measured morning HCHO profiles and profiles with low mixing ratios
had an exponential function shape. Profiles during mid-afternoon can be
described by a uniform value in the mixed layer with exponential decay to a
minimum free-tropospheric concentration around 4 km (0.23 ppb). Based on DC-8 profile shape and ceilometer backscatter estimation of MLH
we calculated total columns from in situ measurements (ground-up VCD)
using four different profile shapes: (1) a uniform HCHO vmr up to the median MLH
with a free-tropospheric mixing ratio of 0.23 ppb from the MLH to the
average tropopause height of 12.77 km; (2) same as (1) but using the
measured MLH; (3) a uniform HCHO mixing ratio up to the median MLH with a
free-tropospheric mixing ratio that exponentially decreases above the MLH to
0.23 ppb within Comparison between Pandora and ground-up columns over Olympic Park
suggested that profile shape (4) with measured MLH and exponential decay
produced the best agreement (slope Pandora HCHO columns and ground-up columns disagree the most early in
the morning, when MLHs are very shallow, and the ceilometers detect elevated
residual layers. This disagreement is likely due the tested shapes not
adequately capturing the elevated layers during these conditions
(aerosol-driven MLH is not representative of HCHO distribution when elevated layers
are present). Based on DC-8 and ground-up comparison, Pandoras were able to capture
diurnal variation of HCHO column with some positive bias. This makes Pandora
an excellent validation instrument for TEMPO (Tropospheric Emissions: Monitoring of Pollution).
All data used in this work can be downloaded from NASA Data
Archive:
The authors declare that they have no conflict of interest.
The authors would like to acknowledge the National Institute of Environmental Research (NIER) for its tremendous effort during KOUR-AQ. In particular we would like to thank NIER for its effort in establishing the Olympic Park research site and their assistance with logistics for measurements at both Olympic Park and Mt. Taewha Forest. This work was supported under the NASA Tropospheric Chemistry Program and the EPA Air, Climate, and Energy Research Program. Although this paper has been reviewed by the EPA and approved for publication, it does not necessarily reflect EPA policies or views. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.Edited by: Hendrik Fuchs Reviewed by: three anonymous referees