AMTAtmospheric Measurement TechniquesAMTAtmos. Meas. Tech.1867-8548Copernicus PublicationsGöttingen, Germany10.5194/amt-10-3677-2017Deriving the slit functions from OMI solar observations and its implications for ozone-profile retrievalSunKangkang.sun@cfa.harvard.eduhttps://orcid.org/0000-0002-9930-7509LiuXiongHuangGuanyuhttps://orcid.org/0000-0001-7314-8485González AbadGonzalohttps://orcid.org/0000-0002-8090-6480CaiZhaonanChanceKellyhttps://orcid.org/0000-0002-7339-7577YangKaihttps://orcid.org/0000-0003-0767-2451Harvard–Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA, USADepartment of Atmospheric and Oceanic Science, University of Maryland, College Park, MD, USAKang Sun (kang.sun@cfa.harvard.edu)9October201710103677369524April20175September201726August201731May2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://amt.copernicus.org/articles/10/3677/2017/amt-10-3677-2017.htmlThe full text article is available as a PDF file from https://amt.copernicus.org/articles/10/3677/2017/amt-10-3677-2017.pdf
The Ozone Monitoring Instrument (OMI) has been successfully
measuring the Earth's atmospheric composition since 2004, but the
on-orbit behavior of its slit functions has not been thoroughly
characterized. Preflight measurements of slit functions have been
used as a static input in many OMI retrieval algorithms. This study
derives on-orbit slit functions from the OMI irradiance spectra
assuming various function forms, including standard and super-Gaussian functions and a stretch to the preflight slit
functions. The on-orbit slit functions in the UV bands show U-shaped
cross-track dependences that cannot be fully represented by the
preflight ones. The full widths at half maximum (FWHM) of the
stretched preflight slit functions for detector pixels at large
viewing angles are up to 30 % larger than the nadir
pixels for the UV1 band, 5 % larger for the UV2 band, and
practically flat in the VIS band. Nonetheless, the on-orbit changes
of OMI slit functions are found to be insignificant over time after
accounting for the solar activity, despite of the decaying of
detectors and the occurrence of OMI row anomaly. Applying the
derived on-orbit slit functions to ozone-profile retrieval shows
substantial improvements over the preflight slit functions based on
comparisons with ozonesonde validations.
Introduction
The Dutch–Finnish Ozone Monitoring Instrument (OMI) on board the
NASA Aura satellite has been measuring the direct sunlight and
backscattered sunlight from the Earth since 2004. Spectrally, the
OMI instrument incorporates two 2-D charge-coupled device (CCD)
detectors for the ultraviolet (UV) and visible (VIS) bands. The UV
band is optically separated into the UV1 band
(264–311nm, ∼0.6nm resolution) and the
UV2 band (307–383nm, ∼0.4nm
resolution), and the VIS band covers 349–504nm at
∼0.6nm
resolution . Spatially, the OMI
instrument has a wide, 115∘ field of view, which is
divided into 30 cross-track positions in the UV1 band and 60
cross-track positions in the UV2 and VIS bands. At the nadir
point, the spatial resolution is 13×48km2 for
UV1 and 13×24km2 for UV2/VIS, significantly
finer than the nadir resolutions of existing spaceborne
measurements over similar spectral ranges like GOME, SCIAMACHY,
GOME-2, and OMPS. The spectral and spatial coverages of OMI enable
retrievals of various key constituents of the Earth's atmosphere,
including ozone, nitrogen dioxide, sulfur dioxide, formaldehyde,
BrO, water vapor, and many others, with daily global coverage at
the Equator . In addition to abundances of
atmospheric species, OMI delivers products like cloud
fraction/height, aerosol optical depth, surface UV radiation, and
solar activity
proxies .
Solar irradiance spectra observed by the UV1, UV2, and VIS bands of OMI
(overlapped spectral regions are not shown). The background high-resolution
solar reference spectrum in gray is from . Deep
solar lines used to monitor solar activity are labeled. The colored
horizontal bars indicate the spectral windows within which slit functions are
fitted in Sect. .
The slit function, also called the instrument transfer function, instrumental spectral response function, or
instrument line shape in the literature, is the instrument's
response to a Dirac delta function in the spectral
domain . Therefore,
the OMI observed spectra can be modeled as a convolution of the
OMI slit functions and the native resolution spectra. A thorough
understanding of the OMI slit function is crucial to accurately
modeling the OMI spectra and retrieving atmospheric
constituents/properties. The OMI slit functions vary in both the
spectral dimension (columns) and the spatial dimension (rows, or
cross-track direction) of the 2-D detectors and have been measured
accurately in a preflight
experiment . These preflight slit
functions have been adopted in a wide range of operational OMI
retrieval
algorithms . However,
it is still unclear whether the preflight slit functions adequately
represent the on-orbit slit functions for OMI, given the impact of
the launching process and the contrasts between laboratory and
space conditions. Additionally, the on-orbit slit functions may
evolve over time due to instrumental issues, as observed in some
other satellite
instruments . The
optical degradation of OMI has been markedly small over the
mission, but the row anomaly (RA) appeared in 2007 and has made
significant impact on about one-third of cross-track positions
since January 2009 . The impact of RA on
the OMI slit functions is still poorly understood.
The Smithsonian Astrophysical Observatory (SAO) ozone-profile retrieval algorithm derives on-orbit slit
functions from averaged OMI solar observations in 2005–2007,
assuming standard Gaussian slit function form, which showed better
performance than using the complex preflight slit
functions . This indicates that although no
significant changes in the OMI slit functions have been noted over
the mission, the on-orbit slit functions may differ from the
preflight at the ozone retrieval windows. Therefore, it is
necessary to reevaluate the OMI slit functions using on-orbit
data. To this end, we investigate the temporal variation of the
slit function using more than 10 years of OMI irradiance data,
compare the on-orbit slit functions derived from irradiance with
the preflight ones, and evaluate multiple on-orbit slit function
forms by validating the ozone profiles retrieved using different
slit functions with ozonesonde observations.
Instrument and data analysesOMI instrument and its solar measurements
The OMI instrument is a push-broom grating spectrometer flying in
a 705km sun-synchronous polar orbit with ∼13:45
Equator-crossing time. The two 2-D CCD detectors have 780 columns
(the spectral dimension, along the flight direction) and 576 rows
(the spatial dimension, perpendicular to the flight direction). In
the global mode that is mostly used in Earth observation, the 576
spatial pixels are selectively binned into 60 (UV2/VIS) or 30 (UV1)
rows, corresponding to the same numbers of cross-track positions at
the Earth's surface. Anomalous behavior of some rows started in
June 2007 and became much more significant in January 2009. This
RA permanently affects radiances of
rows 25–42 (out of 1–60) and 54–55 of the UV2/VIS bands and
occasionally affects rows 43–53. The RA rows in UV1 have similar
relative positions, although all UV1 rows are affected in the
northern parts of orbits. The RA is believed to be caused by loose
thermal insulating materials partially blocking OMI's
Earth-observing field of view. Detailed description of the RA can
be found in .
OMI also has an irradiance view port for solar calibration. Solar
spectra, shown in Fig. , are measured once per day
near the northern terminator of an orbit. The direct sunlight is
attenuated by an optical mesh, then reflected by one of the solar
diffusers (quartz volume diffuser, regular aluminum, or backup
aluminum), and finally reflected by a folding mirror to the
remainder of the optical system that is identical to the Earth
radiance measurement. The blocking effects that caused the RA in
the radiance measurement are not observed in the irradiance; the
only noted RA-related impact on irradiance is that the RA rows in
the UV1 band show faster optical degradation due to additional
solar exposure from RA
reflections .
(a) Preflight slit functions at the median wavelength of
the spectral window 343–356nm (349.5nm). The
spectral variations of preflight slit functions over this window are small
(see the second row of Fig. for the range of preflight
slit function width in UV2). (b–d) Derived slit functions for the
spectral window 343–356nm by fitting a stretch factor to the
preflight slit function, super-Gaussian function, and standard Gaussian
function, respectively. The solar spectrum used in the fitting is the average
OMI irradiance from 2005 to 2007. Adding the asymmetric parameter in the
super/standard Gaussian fitting gives little difference. Only half of the UV2
cross-track positions (1–30) are shown as the other half are essentially
a mirror image (see the second row of Fig. ).
Comparisons of the FWHM of on-orbit slit functions derived from the average
OMI irradiance over October 2004–June 2007 with the preflight ones for all
three OMI bands, each divided into four fitting windows as shown in
Fig. . Because the preflight slit functions vary continuously
with wavelength, the ranges of preflight slit function FWHM within the
spectral window are plotted as the gray bands. The super-Gaussian shape
parameter k is shown with vertical axes on the right.
The OMI solar irradiance is also used to monitor the solar activity
through the variations of deep solar lines, which generally get
shallower when the sun is more active. The most used solar activity
proxies are the core-to-wing indices of the Mg II line at
280nm, Ca II K line at 393.4nm, and Ca II H
line at 396.8nm (labeled in Fig. ). The OMI
Mg II index varied by ∼10% between solar cycle
minimum and maximum. Although the Ca II indices are well correlated
with the Mg II index, their relative variations are smaller by
factors of 7–9 . The irradiance spectrum at
other wavelengths has weaker correspondence with the solar cycle,
generally <0.2% between cycle minimum and maximum in
the UV2/VIS bands and 0.5–1% in the UV1
band .
OMI slit functions
The OMI slit functions were determined during the preflight
characterization for each spectral pixel and cross-track
position. The measured slit functions at discrete wavelengths were
then fitted using a combination of a standard Gaussian and
a flat-top function, defined at a wavelength grid Δλ:
Spre(Δλ)=A0exp-Δλ-λ0w02+A1exp-Δλ-λ1w14,
where A0 and A1 are the relative amplitude of the standard and
flat-top Gaussian components, λ0 and λ1 are their
central positions, and w0 and w1 determine their widths. For
the UV1 band, only the standard Gaussian component is necessary
(λ0 and A1 are zero). The accuracies of these preflight
slit functions were demonstrated to be better than 2%
within ±2 FWHM (full widths at half maximum) and ∼3% between ±2 FWHM and ±3 FWHM during the
preflight test .
Information on the on-orbit slit function can be retrieved by
fitting the observed solar irradiance with a high-resolution solar
reference spectrum (shown in Fig. ) and some assumed
slit function forms. Wavelength shift/squeeze terms and
a polynomial baseline are also included in the fitting. The fitting
applies a weighted Levenberg–Marquardt nonlinear least-square
algorithm to minimize the sum of squares of fitting residuals
weighted by OMI spectral uncertainties. The high-resolution
reference spectrum is extended by 5 nm beyond the fitting
window edges to mitigate the edge effect. The wavelength shift
terms derived here are consistent with the spectral calibration
trends using the OMI radiance (Fig. 33 in
). No significantly different trends of
spectral shifts are observed for different symmetric slit function
fits. Two high-resolution solar reference spectra are tested: the
KNMI spectrum 0.025nm resolution and
0.01nm sampling; and the SAO2010
spectrum 0.04nm resolution and 0.01nm
sampling;. They give very similar results
for the derived slit functions. The KNMI spectrum is used in the
following results due to its better radiometric calibration in the
OMI spectral range. This slit function fitting method has been
described in , where multiple analytical and
numerical function forms were tested to represent the slit
functions of the OCO-2 instrument. These function forms, as well as
the functional form used to parameterize the OMI preflight slit
functions (Eq. 1), were also tested for OMI. The (a)symmetric
standard Gaussian, super-Gaussian, and fitting a homogeneous
stretch to the preflight slit functions were found to produce
stable fitting results. The other function forms (stretch/sharpen,
hybrid Gaussians, and the functional form of Eq. 1) were unstable
due to fitting too many parameters and/or the correlation of some
parameters. The stability issue became more significant later in
the OMI mission, when the signal-to-noise ratio (SNR) of solar
spectra was degraded. The Gaussian function family can be
generalized as the asymmetric super-Gaussian
function :
S(Δλ)=exp-Δλw+sgn(Δλ)awk,
where w is the half width at 1/e, k is the shape parameter,
aw is the asymmetry parameter, and sgn() is the sign
function. If k is fixed at 2, the slit function is standard
Gaussian; if aw≠0, the slit function is asymmetric. The
homogeneously stretched preflight slit function is simply
S(Δλ)=SpreΔλ/stretch.
Examples of the preflight slit functions and the fitted function
forms are illustrated in Fig. for the fitting
window 343–356nm in the UV2 band.
Temporal variations of derived standard Gaussian slit function FWHM for all
cross-track positions from September 2004 to May 2016. Three OMI bands
(columns of the plot) are each divided into four fitting windows (rows of the
plot). Spectral coverages of each window are also labeled in the title of
each subplot.
In the slit function fitting algorithm, the slit function is
generally assumed to be constant within the fitting spectral
window. When fitting a stretch to the preflight slit function, the
preflight slit function at the median wavelength is applied to the
entire window. Spectrally resolved stretch of the preflight slit
functions was also tested by stretching spectrally dependent
preflight slit functions over sliding and overlapping windows, but
the results were influenced by radiometric and solar reference
spectrum uncertainties over short wavelength ranges. The ozone
profiles retrieved assuming spectrally constant slit functions also
outperform those retrieved using spectrally resolved slit
functions. Consequently, the derived slit functions are assumed to
be spectrally constant over each fitting window.
Scatter plots between the average slit function FWHM of all non-RA rows (assuming standard Gaussian) and the Mg II index. The columns and rows are the same as Fig. . The linear fitting slopes between the normalized slit function widths and the normalized Mg II index are shown in the title of each subplot.
The same as Fig. but for the RA rows. The ranges of vertical axes are the same as Fig. as well for easier comparison.
Difference between on-orbit and preflight slit functions
The OMI solar irradiance is generally assumed to be stable, and
many retrieval algorithms use the average or the first principal
component of OMI irradiance spectra over multiple years to enhance
the
SNR . Figure
compares the FWHM of preflight slit functions with those of
on-orbit slit functions derived from the average OMI irradiance
over October 2004–June 2007 (before RA) at all three OMI
bands. The ranges of preflight slit function FWHM within each
fitting window are denoted by gray areas. Adding asymmetry factors
in the on-orbit slit function fitting makes little difference in
all fitting windows, indicating that the OMI slit functions are
sufficiently symmetric, as noted by
. For the UV1 band (the first row
of Fig. ), the preflight slit functions are
standard Gaussian, so fitting a stretch to the preflight slit
functions is identical to fitting a standard Gaussian. The super-Gaussian shows unstable cross-track features, favoring the use of
standard Gaussian in the UV1 band. At the low wavelength end
(270–277nm), the on-orbit slit functions are
remarkably broader than the preflight ones toward the edges of
cross-track positions, by up to 30%. Similar effects
are observed at the high wavelength end of the UV1 band
(299–309nm), although much less significant.
Panels in the second row of Fig. show the
results for the UV2 band, divided into four fitting windows. The
on-orbit and preflight slit functions at the first half of the
cross-track positions (1–30) for window 3 are plotted in
Fig. . The cross-track patterns are similar for
all four fitting windows. The preflight slit functions show very
little cross-track variation (<1% in FWHM), whereas
the FWHM of standard Gaussians and the stretched preflight slit
functions show a U-shaped cross-track dependency. The derived slit
function FWHM at large off-track viewing angles are up to
5% broader than the nadir ones. The FWHM of super-Gaussian is only weakly cross-track dependent, but the shape
parameter k has a strong reverse-U-shaped cross-track
dependency. As illustrated more clearly in Fig. ,
the derived on-orbit slit functions are broader toward the larger
off-track viewing angles if only the widths are fitted; when the
slit function shape can also be adjusted in the super-Gaussian
fitting, the broadening is redistributed towards the tails of the
slit functions. The cross-track dependency is significant for all
functional forms used in the fitting, indicating that the changes
observed between on-ground and in-orbit is instrument related, not
due to usage of different fitting functional forms.
Panels in the third row of Fig. show the
results for the VIS band. The preflight slit functions also have
very little cross-track variation; the on-orbit slit functions
capture some cross-track dependency but not as significant as the
UV2 band. Fitting window 1 is an exception, where
a reverse-U-shaped stretch factor and a U-shaped shape parameter
k are present.
We have also included the fitting results using average OMI solar
spectra from January 2007 to January 2009 (insignificant RA, solar
minimum) and from January 2010 to September 2012 (significant RA,
significant solar activity) in Figs. and
in the Appendix. The cross-track features
are very similar during these periods, indicating that these
cross-track dependent features observed on-orbit are not due to
any temporal effects (e.g., RA, solar activities). For the UV1
band and window 1 of the UV2 band, the super-Gaussian results show
unstable cross-track features, likely due to the correlation
between the width and shape parameters, but appear to be generally
invariant over time. Figure further
compares the fitting residuals agglomerated to within each OMI
band using different slit functions at different cross-track
positions and early/late in the mission. The super-Gaussian and
stretched preflight slit functions show smaller fitting residuals
than standard Gaussian for the UV2 and VIS bands, although an
ozone retrieval test using these three slit function forms at
window 1 of UV2 band shows only small differences (see
Sect. 5.3).
Temporal variation of OMI slit functions
The on-orbit variations of OMI slit functions have widely been
deemed insignificant throughout the mission. By deriving on-orbit
slit functions using the daily OMI solar irradiance, it is
possible to verify this assumption. Figure
presents the evolution of derived slit function widths for all
cross-track positions from September 2004 to May 2016, where the
three OMI bands are each divided into four fitting windows (see
Fig. for the window locations), and a standard
Gaussian slit function is derived for each window. The super-Gaussian and stretched preflight fits are also tested at selective
cross-track positions and fitting windows. The fitted stretch
factor and shape factor k of super-Gaussian have essentially the
same temporal trends as the width of standard Gaussian, so only
the widths of derived standard Gaussian are used to represent the
slit function change. Significant temporal variations of the slit
function widths can be observed at windows 1–2 of the UV1 band
(270–287nm, rows 1–2, column 1 of
Fig. ) and window 1 of the VIS band
(380–402nm, row 1, column 3 of
Fig. ).
We assume that the high-resolution solar reference spectrum stays
constant because there have been no high-resolution solar
reference spectra that incorporate the solar cycles
available. However, this assumption does not hold near deep solar
lines and at short wavelengths in the UV at the OMI
resolution . Since
the true solar irradiance is assumed to be invariant, the relative
changes of observed solar lines following the solar cycles will be
interpreted as the slit function change in the fitting. As
a result, the temporal variations seen in
Fig. may result from solar cycles and not
necessarily indicate the real changes of on-orbit slit functions.
The solar activity can be represented by the OMI Mg II
core-to-wing index, defined as the ratio between OMI irradiance at
280nm and the average irradiances at 277 and
283nm. The OMI Mg II index has also been verified
against other ground-based and spaceborne solar
observations . The Mg II index is modulated
by a distinct 11-year solar cycle and a 27-day solar rotation
cycle, whose amplitude also changes due to variations in
faculae. No seasonal variations due to instrument temperature
effects can be found in the Mg II index. Given the complex nature
of solar activity, it is highly unlikely that any potential
factors that may contribute to slit function change (e.g.,
instrument degradation, RA effects, and instrument temperature
variations) are correlated with the Mg II index. To separate the
spurious slit function changes due to insufficient consideration
of solar activity with potential real changes, the derived slit
function widths are plotted against the Mg II index for the
aggregated rows that are not influenced by RA (non-RA rows) and
influenced by RA (RA rows) in
Figs. –, color-coded
by time over 2004–2016. For the non-RA rows, the derived on-orbit
slit function widths show linear correlations with the Mg II index
for most windows throughout the mission
(Fig. ). Window 4 at UV1 is the only
exception, where the derived width abruptly increased by ∼7% in April 2008. This change cannot be explained by
solar activity or major RA events but occurred exactly when the
SNR of the UV1 band started to decrease , so it
is still possibly a secondary RA effect. The slopes between the
relative changes of slit functions widths and relative change of
the Mg II index are also labeled in
Fig. . The slope is the largest
(0.50–0.51, 95% confidence interval) for window 2
of the UV1 band (where the Mg II line is located), meaning that as
the Mg II index varied by 10% from solar cycle minimum
to maximum, the derived slit function width varied by
5%. It is followed by window 1 of the UV1 band
(0.24–0.25) and window 1 of the VIS band (0.13). Window 1
at UV1 has the shortest wavelength, and window 1 at VIS contains
the strong Ca solar lines, so these two windows are also strongly
influenced by solar activity. Most of the derived slit function
changes can be explained by the Mg II index and hence solar
activity variations (see the R2 in the plot), and the residual
variations are mostly random noise that increases with time as the
instrument slowly degrades. Therefore, the apparent temporal
variations in Fig. are due to solar
activity rather than real on-orbit slit function changes.
The locations of the ozonesondes used in this study. The size of circles denotes the number of successful validations for the two-band
case in 2004–2008, and the color denotes mean bias between the tropospheric ozone column from OMI and the ozonesonde convolved with the OMI averaging
kernels.
Number of successful validations at different cross-track positions for the two-band case (blue) and UV2-only case (red) over 2004–2008.
Although no significant RA impact on the derived slit functions
can be observed from Fig. , the impact is
visible when plotting the slit function widths of the RA rows
against the Mg II index (Fig. , in the same
format as Fig. ). Coincidently, both the
solar minimum (also the minimum of the Mg II index) and the major
RA events occurred in 2008–2009. Hence the impact of RA will be
manifested as hysteresis in the scatter plots. In contrast to the
non-RA rows, the derived slit function widths of the RA rows show
different relationship with the Mg II index before and after the
major RA event in January 2009, mostly notable at windows 1–3 of
the UV1 band. Some extra hysteresis is observable but small
(0.3% of slit function width) at window 2 of the UV2
band and window 2 of the VIS band. Overall, the temporal
variations of slit function widths for non-RA rows, if they exist,
are no more than 1% and much smaller than differences
between the preflight and on-orbit slit functions for all three
OMI bands (Fig. ). The RA rows also did not
show any significant temporal variations before the major RA
events in 2009.
Validation of ozone profiles retrieved using different slit functionsSAO OMI ozone-profile retrieval algorithm and validation
The optical path of the solar irradiance and the Earth radiance
measurements are very similar, but it is still unknown
whether the slit functions are identical for the solar and
earthshine spectra. Besides, the derived on-orbit slit functions
are sometimes simplified compared to the preflight ones. Therefore,
it is necessary to apply the on-orbit slit functions derived from
solar irradiance to the earthshine spectra fitting algorithm and
validate the retrieval results. In this study, the SAO OMI
ozone-profile retrieval algorithm is used to test different options
of slit functions. The OMI ozone-profile retrievals have
substantially higher relative accuracy than other OMI products
(e.g., SO2, NO2, HCHO), and there have been
extensive validations for the ozone-profile product. Therefore, the
OMI ozone-profile retrieval is the first choice to test the
on-orbit slit functions.
The SAO OMI ozone-profile retrieval algorithm was described in
detail by with further improvements by
and validated
extensively against ozonesonde and Microwave Limb Sounder (MLS)
observations . In the operational SAO
ozone-profile algorithm, partial ozone columns (in Dobson units;
1DU=2.69×1016moleculescm-2) are
retrieved at 24 layers from the surface to about 60km
using the optimal estimation technique. Stratospheric ozone column
(SOC) and tropospheric ozone column (TOC) are derived from the OMI
ozone profile using the thermal tropopause heights, defined by the
lapse rate criterion , from the National
Center for Environmental Protection (NCEP) reanalysis. The total
degree of freedom for signal (DFS) of the retrieved ozone profile
is 6–7 with 5–7 in the stratosphere and 0–1.5 in the
troposphere. Fitting windows in both UV1 (270–309nm)
and UV2 (311–330nm) are used to retrieve ozone
abundance at different altitudes. Wavelengths around Mg II
(280nm) and Mg I (285nm) lines are not
included in the retrieval. Because of the mismatch of cross-track
positions between UV1 and UV2, UV2 spectra at every two adjacent
cross-track positions are co-added to match the UV1 spatial
resolution. When applying the preflight slit functions, they are
also averaged every two cross-track positions for UV2. Hence the
retrievals are performed at the UV1 spatial resolution. In
addition, OMI radiances are pre-calibrated based on 2-day-average
radiance differences in the tropics between OMI spectra and
spectral simulations using zonal mean MLS ozone profiles at
pressure less than the 215 hPa level and climatological
ozone profiles at pressure greater than the 215 hPa
level. This “soft calibration” significantly reduces the OMI L1B
calibration errors that are dependent on both wavelength and
cross-track positions.
In this study, the ozone profiles are retrieved using four
different options of slit functions: the preflight as well as
standard Gaussian, super-Gaussian, and stretched preflight derived
from OMI irradiance. The soft calibration is turned off to make
fair comparisons between slit functions, because the soft
calibration algorithm is currently only implemented with standard
Gaussian slit functions. The other retrieval options are kept the
same as in the operational algorithm whenever possible. This
slit function comparison is named as the “two-band” case. Another
case is also tested by using only the UV2 window
(311–330nm). In this “UV2-only” case, there are no
compounding factors induced by averaging the adjacent UV2 spectra
and preflight slit functions. The OMI ozone profiles are retrieved
at 60 cross-track positions rather than 30 cross-track positions
for the two-band case. A “UV1-only” case is also tested with
a subset of ozonesondes but found to be insensitive to different
options of slit functions (although the on-orbit/preflight slit
functions are quite different for the UV1 band; see the first row
of Fig. ).
Ozonesonde observations are widely used to validate satellite
ozone-profile
retrievals . We
use the same global ozonesonde dataset described by
but only during 2004–2008. The ozonesonde
profiles extend from the surface up to ∼35km with
vertical resolution of 100–150m, 3–5%
precision, and 5–10%
accuracy . When
compared to the OMI ozone profiles that have much lower vertical
resolution, ozonesonde profiles are first integrated into the
corresponding OMI layers and then degraded to the OMI vertical
resolution using the OMI a priori ozone profile and averaging
kernels (AK). The locations of the ozonesondes are shown in
Fig. , where the size of circles denotes the number
of successful ozonesonde validations for the two-band case in
2004–2008, and the color denotes mean bias between the
TOC retrieved in the two-band case using
standard Gaussian slit functions and the TOC
from ozonesonde with OMI AK applied. A successful validation is
defined when OMI retrievals using different slit functions and the
collocated ozonesonde profile all pass the filtering criteria. The
OMI/ozonesonde data filtering criteria in this study are very
similar to : OMI effective cloud fraction of less
than 0.3; OMI solar zenith angle of less than 75∘;
ozonesondes that reach an altitude of at least 30km and
have data gaps no greater than 3km; ozonesonde
correction factors (CFs), if they exist, in the range of 0.85 to
1.15. We did not apply these CFs because it is not clear that
they should be applied to the ozone profiles, especially for the
troposphere, and CFs are only available for a limited fraction of
ozonesondes . For each filtered ozonesonde
profile, the nearest filtered OMI profile within ±1∘
latitude, ±1∘ longitude, and ±6h is used
for validation on the individual profile basis. The validation in
this work also has the following important differences from
:
used the operational ozone-profile product
that co-added four spatial pixels along the track and hence had
a nadir spatial resolution of 52×48km2; this study
retrieves ozone profiles only at pixels collocated with ozonesonde
stations with standard spatial resolution of the UV1 band
(13×48km2 at nadir) for the two-band case and
resolution of the UV2 band (13×24km2 at nadir) for
the UV2-only case. As the soft calibration is turned off, the
overall biases are slightly larger than the operational SAO ozone-profile product.
As shown in
Sects. –,
the cross-track discrepancies between on-orbit and preflight slit
functions are much more significant than the temporal variations of
on-orbit slit functions over the OMI mission. Hence only the
“pre-RA” period (2004–2008) is included in the validation to
consistently compare all cross-track positions.
only used cross-track positions 4–27 at the
UV1 spatial resolution, whereas in this study all cross-track
positions are included in the validation. The successful
validations are further grouped according to the cross-track
positions of the OMI ozone profiles. Figure
shows the number of successful validations in the two-band case and
UV2-only case over 2004–2008. On average, there are 135 successful
validations per UV1 cross-track position for the two-band case and
74 successful validations per UV2 cross-track position for the
UV2-only case.
Scatter plots of OMI retrieved ozone partial columns vs. the ozonesonde data
for SOC (a), TOC (b), and TOC500 (c). The four
columns represent retrievals using standard Gaussian, super-Gaussian,
preflight, and stretched preflight slit functions. The slope and offset of
the linear regression, correlation coefficient (R), and the mean bias ±1σ (Δ) are shown for each subplot. The linear regressions are
shown as red lines, and the black dash line is the 1:1 line. The number of
successful validations is 4056 (same as for all slit functions; see
Fig. ).
(a) Median biases of OMI profiles retrieved using four different
slit functions compared to the ozonesondes. The median difference between
a priori profile and the ozonesonde data is also shown. (b) Standard
deviation of the differences between different retrievals and ozonesonde
data. The a priori error is shown in (b).
(a) Medians of difference profiles between OMI and ozonesondes at
each cross-track position. (b) Distributions of fitting residual RMS
at each cross-track position for the UV1 band. (c) Same as the second
row but for the UV2 band. The mean and standard derivation of all residual RMS
are labeled in each subplot at (b and c).
Similar to Fig. but for the UV2-only case. Only SOC and TOC are shown. The number of successful validations is 4461 (same as for all slit functions, see Fig. ).
Same to Fig. but for the UV2-only case.
Similar to Fig. but for the UV2 band only. Note there are 60 rather than 30 cross-track positions.
Impact of slit functions on ozone-profile retrieval: the two-band case
Figure compares the SOC (first row), TOC (second
row), and the TOC from surface to
500hPa (TOC500, third row) of OMI retrieved using four
different options of slit functions with the ozonesonde data over
2004–2008. The standard/super-Gaussian and the stretched preflight
slit functions are derived using the mean solar irradiance over
2005–2007, and the slit functions are assumed to be invariant
over time. For the SOC, the correlation coefficients between OMI
and ozonesondes are similar for all slit functions, whereas the
stretched preflight and super-Gaussian options show smaller mean
absolute biases (-0.11 and -0.19DU) than the standard
Gaussian option (0.60DU), relative to the ozonesonde
data. The preflight option shows the largest mean absolute bias
(-2.01DU) and the largest standard deviation
(13.48DU). For the TOC, the standard Gaussian retrieval
shows the lowest mean bias (1.89DU), followed by
stretched preflight (2.59DU) and the super-Gaussian
(2.69DU). The standard and super-Gaussian retrievals
show better correlation coefficient (0.85), followed by the
stretched preflight (0.84). When comparing the TOC500, the
relationships between different slit functions are similar to
TOC. For the mean absolute biases, standard
Gaussian < stretched preflight ≈ super-Gaussian < preflight. However, the correlation coefficient for the super-Gaussian becomes slightly higher than the standard Gaussian.
As shown by Fig. , the ozone partial columns
retrieved using the preflight slit functions consistently show the
lowest performance compared to the derived on-orbit slit
functions. Figure illustrates the
vertical distributions of biases between OMI and the ozonesonde
profiles. The left panel shows the median biases between OMI
profiles retrieved using four different slit functions and the
ozonesondes; the right panel shows the standard deviation of the
differences between different retrievals and ozonesonde data. The
median bias of the a priori and the a priori error are also
plotted for reference. The retrieval using preflight slit
functions shows significant positive bias in the stratosphere,
negative bias at upper troposphere–lower stratosphere, and
positive bias in the lower troposphere. The retrieval using
preflight slit functions also has much larger variations. The
median biases show different vertical distributions between
standard Gaussian and super-Gaussian/stretched preflight, with no
great altitude-dependent variation for the standard Gaussian and
almost the same larger oscillations for the latter two. Unlike the
median biases, the standard deviations of biases are very similar
for retrievals using these three on-orbit slit function options.
The bias profiles between OMI and ozonesonde are further grouped
according to the OMI cross-track positions, and the medians of
bias profiles at each cross-track position are shown in the first
row of Fig. . The ozone profiles retrieved using
the preflight slit functions show substantial cross-track
dependent biases; the bias profile shown in
Fig. can be largely attributed to the biases
near the edges of the cross-track positions. The second and third
rows of Fig. present the fitting residual
root mean square (RMS) of UV1 and UV2 fitting windows,
respectively. Although the on-orbit slit functions show the
largest difference from the preflight at the UV1 band
(Fig. ), the fitting residuals at the UV1 band
using different slit functions are very similar; the residual RMS
values using the preflight slit functions are only marginally
larger than retrievals using on-orbit slit functions. The
insensitivity of fitting residuals to slit functions at the UV1
band is likely due to the smoothness of the ozone absorption
cross sections. The residual RMS at the UV2 band for the retrieval
using the preflight slit functions, however, shows a distinct
U-shaped cross-track distribution, very different from the other
retrievals. Apparently, the preflight slit functions are not
accurate near the edges of the cross-track positions in this
two-band test case. It should be emphasized that in this case the
UV2 radiance spectra and preflight slit functions are co-added
from 60 to 30 cross-track positions. The co-adding of adjacent
radiance spectra defined at different wavelength grids introduces
some effective broadening of slit functions, which may partially
explain the mismatch of slit functions for the preflight retrieval
at UV2. Therefore, the UV2-only case is also tested to avoid
co-adding and test the UV2 slit functions only.
Impact of slit functions on ozone-profile retrieval: the UV2-only case
When only using the UV2 window in the ozone-profile retrieval, the
DFS is reduced from 6–7 to 1–2 due to the loss of most
stratospheric information content from UV1.
Figure compares the SOC (first row) and TOC
(second row) from OMI with ozonesondes. Given the low DFS, only the
SOC and the TOC are validated. Similar to the two-band case, the
retrieval using the preflight slit functions shows the largest
absolute biases and lowest correlation with ozonesondes for both
TOC and SOC. The standard Gaussian retrieval consistently shows
higher correlation coefficients and lower standard deviations of
biases compared to the other two forms of derived on-orbit slit
functions that are supposed to better represent the true slit
function shapes in UV2 (see Figs. and
). The only exception is that the
stretched preflight retrieval shows marginally better correlation
coefficient than the standard Gaussian for the TOC.
Figure compares the medians and standard
deviations of bias profiles from different retrievals, similar to
Fig. . The retrieval using the preflight slit
functions also shows the largest biases and standard
deviations. The three derived on-orbit slit functions show
different altitude dependency in the median biases with the bias of
stretched preflight similar to super-Gaussian in shape but smaller
in absolute value. The standard Gaussian shows the lowest standard
derivations of the bias profiles, followed by stretched preflight
and super-Gaussian.
Figure shows the cross-track distributions of
median bias profiles and fitting residual RMS for ozone-profile
retrievals using the four different slit functions, similar to
Fig. but for UV2 only. The preflight retrieval
again stands out with significant cross-track dependent biases,
larger towards the edges of the cross-track positions. The RMS of
the retrieval using preflight slit functions also show U-shaped
cross-track dependency, but smaller than those in
Fig. due to no broadening of effective slit
functions from co-adding. These large RMS at both edges are
mitigated by fitting a stretch to the preflight slit functions. The
mean RMS of using super-Gaussian and stretched preflight slit
functions are slightly better than using the standard Gaussian,
indicating that generally a broad-top slit function can better
model the OMI spectra. However, the retrieval using standard
Gaussian slit functions shows the smallest variations of biases and
variations of residual RMS, which is not fully understood. It is
possible that the slit function in radiance measurements cannot be
fully represented by the on-orbit slit functions derived from the
solar irradiance due to scene
inhomogeneity ,
unaccounted stray lights, or intra-orbit slit function changes
(observed in GOME-2 by Beirle et al., 2017) in earthshine spectra.
Conclusions
The accurate characterization of slit functions is essential for
the spectral calibration of spaceborne grating spectrometers and
the retrieval of the Earth's atmospheric constituents. We derive
on-orbit slit functions by fitting the OMI irradiance spectra with
a high-resolution solar reference spectrum and various assumptions
on slit function forms, including standard and super-Gaussian
functions and a homogeneous stretch to the preflight slit
functions. The on-orbit slit functions derived from multi-year
averaged OMI solar irradiance show U-shaped cross-track dependences
at the UV bands that cannot be fully represented by the preflight
slit functions. The FWHM of the stretched preflight slit functions
of detector pixels at large viewing angles are up to
30% larger than the nadir ones for the UV1 band and
5% larger for the UV2 band. When fitting super-Gaussian
slit functions in the UV2 band, the cross-track variations of FWHM
are much smaller, but the cross-track variations of the shape
parameter k are significant. No significant discrepancy is found
in the VIS band except for the first fitting window
(380–402 nm), where a moderate reverse-U-shaped
cross-track dependency of derived slit function width is present.
The derived on-orbit slit functions using daily OMI solar
irradiance from 2004 to 2016 show little temporal variations after
taking account of the impact of solar activity. Overall, these
temporal variations for non-RA rows, if they exist, are no more
than 1% and much smaller than differences between the
preflight and on-orbit slit functions for all three OMI bands. The
RA rows also did not show any significant temporal variations
before the major RA events in 2009.
Considering the insignificant temporal variations of on-orbit slit
functions, the slit functions derived from the multi-year average
OMI solar irradiance are applied in the SAO ozone-profile
retrieval, and the results are compared with the retrieval using
the preflight slit functions. Two cases are tested: one is to keep
the same options of the operational algorithm whenever possible,
and the other one is to use only the UV2 band to avoid co-adding the
UV2 spectra and eliminate the interaction between UV1 and UV2. The
retrievals using derived on-orbit slit functions consistently show
smaller biases and better correlations with the ozonesonde
validations in both cases. Although the on-orbit slit functions
show larger difference from the preflight in the UV1 band, the
impact of slit functions on the ozone-profile retrieval is
dominated by the UV2 band due to the more complicated structure of
ozone absorption cross section in UV2. The UV-2-only test case has
direct implications for other OMI products that use the UV2 band,
such as SO2, HCHO, and BrO. The on-orbit slit functions of
the OMI VIS band also have cross-track discrepancies compared to
the preflight, although less significant. Future comparisons of
retrievals using different slit functions will be performed in the
VIS band with well-validated algorithms, such as the water vapor
retrieval .
It is challenging to characterize the slit functions of 2-D
detectors of OMI-like instruments, as the slit functions vary in
both the wavelength (the column dimension) and cross-track viewing
dimension (the row dimension) and, more critically, may vary over
time. Future work will involve characterizing the differences
between the slit functions derived from solar irradiance and the
slit functions of earthshine radiance. These differences may be
caused by scene heterogeneity, differences in stray light between
irradiance and radiance, and intra-orbit instrumental changes. It
is possible to linearize the slit function fitting by constructing
“pseudo-absorbers” based on derivatives of slit functions and
including them in the radiance
fitting . Accurate knowledge of the
on-orbit slit functions, as demonstrated in this work, will also be
important for the near-future missions that have more spatial
pixels and higher retrieval targets than OMI (TROPOMI/Sentinel-5P,
Sentinel-5, Sentinel-4, GEMS, and TEMPO).
The results in this study are based on OMI L1B solar
irradiance products, which are publicly available at
https://aura.gesdisc.eosdis.nasa.gov/data/Aura_OMI_Level1/OML1BIRR.003/.
Additional comparisons of different slit function forms
Figures and repeat the
analysis of Fig. using alternative solar spectra
averaged over January 2007–January 2009 and
January 2010–September 2012,
respectively. Figure shows the slit
function fitting residuals using OMI solar irradiance agglomerated
within each OMI band at different cross-track positions. The 1-year
averaged solar spectra in 2005 and in 2015 are used in the plot.
Similar to Fig. but using averaged solar spectra from January 2007 to January 2009.
Similar to Fig. but using averaged solar spectra from January 2010 to September 2012.
The residual RMS when fitting different slit functions over the OMI UV1, UV2, and VIS bands.
The lines without circles are fitting results using solar spectra averaged over 2005;
the lines with circles are fitting results using solar spectra averaged over 2015.
The authors declare that they have no conflict of
interest.
This article is part of the special issue “Ten years of Ozone
Monitoring Instrument (OMI) observations (ACP/AMT inter-journal SI)”. It is
not associated with a conference.
Acknowledgements
This paper is supported by NASA's Atmospheric Composition: Aura Science Team
program (sponsor contract numbers NNX14AF16G and NNX14AF56G). We acknowledge
Sergey Marchenko and Matthew DeLand at NASA and Science Systems and
Applications, Inc. for making the solar proxy index available at
https://sbuv2.gsfc.nasa.gov/solar/omi/. We thank Sergy Marchenko for
discussions on the OMI solar data products. We thank Caroline Nowlan,
Christopher Chan Miller, and Huiqun Wang at the Smithsonian Astrophysical
Observatory (SAO); Steffen Beirle at the Max Planck Institute for Chemistry
(MPI-C); and Can Li at NASA GSFC for helpful discussions on slit function
parameterizations. The OMI International Science Team is acknowledged for
making OMI L1B data available at
https://aura.gesdisc.eosdis.nasa.gov/data/Aura_OMI_Level1/OML1BIRR.003/.
We also thank the ozonesonde providers and their funding agencies for making
the ozonesonde measurements, as well as the Aura Validation Data Center
(AVDC), World Ozone and Ultraviolet Radiation Data Center (WOUDC), and the
Southern Hemisphere ADditional OZonesonde (SHADOZ) for archiving the
ozonesonde data.
Edited by: Joanna Joiner
Reviewed by: Ruediger Lang and one anonymous referee
ReferencesBak, J., Liu, X., Wei, J. C., Pan, L. L., Chance, K., and Kim, J. H.: Improvement of OMI ozone profile retrievals in the upper troposphere and lower stratosphere by the use of a tropopause-based ozone profile climatology, Atmos. Meas. Tech., 6, 2239–2254, 10.5194/amt-6-2239-2013, 2013.Bak, J., Liu, X., Kim, J. H., Deland, M. T., and Chance, K.: Improvement of OMI ozone profile retrievals by simultaneously fitting polar mesospheric clouds, Atmos. Meas. Tech., 9, 4521–4531, 10.5194/amt-9-4521-2016, 2016.Beirle, S., Lampel, J., Lerot, C., Sihler, H., and Wagner, T.: Parameterizing the instrumental spectral response function and its changes by a super-Gaussian and its derivatives, Atmos. Meas. Tech., 10, 581–598, 10.5194/amt-10-581-2017, 2017.
Chance, K. and Kurucz, R.: An improved high-resolution solar reference spectrum for Earth's atmosphere measurements in the ultraviolet, visible, and near infrared, J. Quant. Spectrosc. Ra., 111, 1289–1295, 2010.Chan Miller, C., Gonzalez Abad, G., Wang, H., Liu, X., Kurosu, T., Jacob, D. J., and Chance, K.: Glyoxal retrieval from the Ozone Monitoring Instrument, Atmos. Meas. Tech., 7, 3891–3907, 10.5194/amt-7-3891-2014, 2014.DeLand, M. and Marchenko, S.: The solar chromospheric Ca and Mg indices from Aura OMI, J. Geophys. Res.-Atmos., 118, 3415–3423, 10.1002/jgrd.50310, 2013.Deshler, T., Mercer, J. L., Smit, H. G. J., Stubi, R., Levrat, G., Johnson, B. J., Oltmans, S. J., Kivi, R., Thompson, A. M., Witte, J., Davies, J., Schmidlin, F. J., Brothers, G., and Sasaki, T.: Atmospheric comparison of electrochemical cell ozonesondes from different manufacturers, and with different cathode solution strengths:
The Balloon Experiment on Standards for Ozonesondes, J. Geophys. Res., 113, D4, 10.1029/2007JD008975, 2008.De Smedt, I., Van Roozendael, M., Stavrakou, T., Müller, J.-F., Lerot, C., Theys, N., Valks, P., Hao, N., and van der A, R.: Improved retrieval of global tropospheric formaldehyde columns from GOME-2/MetOp-A addressing noise reduction and instrumental degradation issues, Atmos. Meas. Tech., 5, 2933–2949, 10.5194/amt-5-2933-2012, 2012.
Dirksen, R., Dobber, M., Voors, R., and Levelt, P.: Prelaunch characterization of the Ozone Monitoring Instrument transfer function in the spectral domain, Appl. Optics, 45, 3972–3981, 2006.
Dobber, M., Voors, R., Dirksen, R., Kleipool, Q., and Levelt, P.: The high-resolution solar reference spectrum between 250 and 550 nm and its application to measurements with the Ozone Monitoring Instrument, Sol. Phys., 249, 281–291, 2008.Froidevaux, L., Jiang, Y. B., Lambert, A., Livesey, N. J., Read, W. G., Waters, J. W., Browell, E. V., Hair, J. W., Avery, M. A., McGee, T. J., Twigg, L. W., Sumnicht, G. K., Jucks, K. W., Margitan, J. J., Sen, B., Stachnik, R. A., Toon, G. C., Bernath, P. F., Boone, C. D., Walker, K. A., Filipiak, M. J., Harwood, R. S., Fuller, R. A., Manney, G. L., Schwartz, M. J., Daffer, W. H., Drouin, B. J., Cofield, R. E., Cuddy, D. T., Jarnot, R. F., Knosp, B. W., Perun, V. S., Snyder, W. V., Stek, P. C., Thurstans, R. P., and Wagner, P. A.: Validation of aura microwave
limb sounder stratospheric ozone measurements, J. Geophys. Res.-Atmos., 113, D15, 10.1029/2007JD008771, 2008.González Abad, G., Liu, X., Chance, K., Wang, H., Kurosu, T. P., and Suleiman, R.: Updated Smithsonian Astrophysical Observatory Ozone Monitoring Instrument (SAO OMI) formaldehyde retrieval, Atmos. Meas. Tech., 8, 19–32, 10.5194/amt-8-19-2015, 2015.Huang, G., Liu, X., Chance, K., Yang, K., Bhartia, P. K., Cai, Z., Allaart,
M., Ancellet, G., Calpini, B., Coetzee, G. J. R., Cuevas-Agulló, E.,
Cupeiro, M., De Backer, H., Dubey, M. K., Fuelberg, H. E., Fujiwara, M.,
Godin-Beekmann, S., Hall, T. J., Johnson, B., Joseph, E., Kivi, R., Kois, B.,
Komala, N., König-Langlo, G., Laneve, G., Leblanc, T., Marchand, M.,
Minschwaner, K. R., Morris, G., Newchurch, M. J., Ogino, S.-Y., Ohkawara, N.,
Piters, A. J. M., Posny, F., Querel, R., Scheele, R., Schmidlin, F. J.,
Schnell, R. C., Schrems, O., Selkirk, H., Shiotani, M., Skrivánková,
P., Stübi, R., Taha, G., Tarasick, D. W., Thompson, A. M., Thouret, V.,
Tully, M. B., Van Malderen, R., Vömel, H., von der Gathen, P., Witte, J.
C., and Yela, M.: Validation of 10-year SAO OMI Ozone Profile (PROFOZ)
product using ozonesonde observations, Atmos. Meas. Tech., 10, 2455–2475,
10.5194/amt-10-2455-2017, 2017a.Huang, G., Liu, X., Chance, K., Yang, K., and Cai, Z.: Validation of 10-year
SAO OMI Ozone Profile (PROFOZ) Product Using Aura MLS Measurements, Atmos.
Meas. Tech. Discuss., 10.5194/amt-2017-92, in review, 2017b.Jia, J., Rozanov, A., Ladstätter-Weißenmayer, A., and Burrows, J. P.: Global validation of SCIAMACHY limb ozone data (versions 2.9 and 3.0, IUP Bremen) using ozonesonde measurements, Atmos. Meas. Tech., 8, 3369–3383, 10.5194/amt-8-3369-2015, 2015.Jiang, Y. B., Froidevaux, L., Lambert, A., Livesey, N. J., Read, W. G., Waters, J. W., Bojkov, B., Leblanc, T., McDermid, I. S., Godin-Beekmann, S., Filipiak, M. J., Harwood, R. S., Fuller, R. A., Daffer, W. H., Drouin, B. J., Cofield, R. E., Cuddy, D. T., Jarnot, R. F., Knosp, B. W., Perun, V. S., Schwartz, M. J., Snyder, W. V., Stek, P. C., Thurstans, R. P., Wagner, P. A., Allaart, M., Andersen, S. B., Bodeker, G., Calpini, B., Claude, H., Coetzee, G., Davies, J., De Backer, H., Dier, H., Fujiwara, M., Johnson, B., Kelder, H., Leme, N. P., König-Langlo, G., Kyro, E., Laneve, G., Fook, L. S., Merrill, J., Morris, G., Newchurch, M., Oltmans, S., Parrondos, M. C., Posny, F., Schmidlin, F., Skrivankova, P., Stubi, R., Tarasick, D., Thompson, A., Thouret, V., Viatte, P., Vömel, H., von Der Gathen, P., Yela, M., and Zablocki, G.:
Validation of Aura Microwave Limb Sounder Ozone by ozonesonde and lidar measurements, J. Geophys. Res.-Atmos., 112, D15S20, 10.1029/2007JD008776, 2007.Johnson, B. J., Oltmans, S. J., Vömel, H., Deshler, T., and Kroger, C.: ECC ozonesonde pump efficiency measurements and tests on the sensitivity to ozone of buffered and unbuffered ECC sensor cathode solutions, J. Geophys. Res., 107, 4393, 10.1029/2001JD000557, 2002.
Komhyr, W.: Operations on handbook-Ozone measurements to 40-km altitude with model 4A electrochemical concentration cell (ECC) ozonesondes, NOAA Tech. Memo. ERLARL-149, Air Resour. Lab., Boulder, CO, 1986.
Komhyr, W., Connor, B., McDermid, I., McGee, T., Parrish, A., and Margitan, J.: Comparison of STOIC 1989 ground-based lidar, microwave spectrometer, and Dobson spectrophotometer Umkehr ozone profiles with ozone profiles from balloon-borne electrochemical concentration cell ozonesondes, J. Geophys. Res., 100, 9273–9282, 1995.Kroon, M., de Haan, J. F., Veefkind, J. P., Froidevaux, L., Wang, R., Kivi, R., and Hakkarainen, J. J.: Validation of operational ozone profiles from the Ozone Monitoring Instrument, J. Geophys. Res., 116, D18305, 10.1029/2010jd015100, 2011.
Kurosu, T. P., Chance, K., and Sioris, C. E.: Preliminary results for
HCHO and BrO from the EOS-aura ozone monitoring instrument, in: Fourth
International Asia-Pacific Environmental Remote Sensing Symposium 2004:
Remote Sensing of the Atmosphere, Ocean, Environment, and Space, 116–123,
International Society for Optics and Photonics, 2004.Levelt, P. F., van den Oord, G. H. J., Dobber, M. R., Malkki, A., Visser, H., de Vries, J., Stammes, P., Lundell, J. O. V., and Saari, H.: The ozone monitoring instrument, IEEE T. Geosci. Remote, 44, 1093–1101, 10.1109/TGRS.2006.872333, 2006.Li, C., Krotkov, N. A., Carn, S., Zhang, Y., Spurr, R. J. D., and Joiner, J.: New-generation NASA Aura Ozone Monitoring Instrument (OMI) volcanic SO2 dataset: algorithm description, initial results, and continuation with the Suomi-NPP Ozone Mapping and Profiler Suite (OMPS), Atmos. Meas. Tech., 10, 445–458, 10.5194/amt-10-445-2017, 2017.Liu, X., Bhartia, P. K., Chance, K., Spurr, R. J. D., and Kurosu, T. P.: Ozone profile retrievals from the Ozone Monitoring Instrument, Atmos. Chem. Phys., 10, 2521–2537, 10.5194/acp-10-2521-2010, 2010.Marchenko, S. and DeLand, M.: Solar spectral irradiance changes during cycle 24, Astrophys. J., 789, 10.1088/0004-637X/789/2/117, 2014.Marchenko, S. V., DeLand, M. T., and Lean, J. L.: Solar spectral irradiance variability in cycle 24: observations and models, J. Space Weather Spac., 6, A40, 10.1051/swsc/2016036, 2016.Miles, G. M., Siddans, R., Kerridge, B. J., Latter, B. G., and Richards, N. A. D.: Tropospheric ozone and ozone profiles retrieved from GOME-2 and their validation, Atmos. Meas. Tech., 8, 385–398, 10.5194/amt-8-385-2015, 2015.Morris, G. A., Labow, G., Akimoto, H., Takigawa, M., Fujiwara, M., Hasebe, F., Hirokawa, J., and Koide, T.: On the use of the correction factor with Japanese ozonesonde data, Atmos. Chem. Phys., 13, 1243–1260, 10.5194/acp-13-1243-2013, 2013.Noël, S., Bramstedt, K., Bovensmann, H., Gerilowski, K., Burrows, J. P., Standfuss, C., Dufour, E., and Veihelmann, B.: Quantification and mitigation of the impact of scene inhomogeneity on Sentinel-4 UVN UV-VIS retrievals, Atmos. Meas. Tech., 5, 1319–1331, 10.5194/amt-5-1319-2012, 2012.Schenkeveld, V. M. E., Jaross, G., Marchenko, S., Haffner, D., Kleipool, Q. L., Rozemeijer, N. C., Veefkind, J. P., and Levelt, P. F.: In-flight performance of the Ozone Monitoring Instrument, Atmos. Meas. Tech., 10, 1957–1986, 10.5194/amt-10-1957-2017, 2017.Smit, H. G. J., Straeter, W., Johnson, B. J., Oltmans, S. J., Davies, J., Tarasick, D. W., Hoegger, B., Stubi, R., Schmidlin, F. J., Northam, T., Thompson, A. M., Witte, J. C., Boyd, I., and Posny, F.: Assessment of the performance of ECC-ozonesondes under quasi-flight conditions in the environmental simulation chamber: Insights from the Juelich Ozone Sonde Intercomparison Experiment (JOSIE), J. Geophys. Res.-Atmos., 112, d19306, 10.1029/2006JD007308, 2007.Sun, K., Liu, X., Nowlan, C. R., Cai, Z., Chance, K., Frankenberg, C., Lee, R. A. M., Pollock, R., Rosenberg, R., and Crisp, D.: Characterization of the OCO-2 instrument line shape functions using on-orbit solar measurements, Atmos. Meas. Tech., 10, 939–953, 10.5194/amt-10-939-2017, 2017.
Tanskanen, A., Krotkov, N. A., Herman, J. R., and Arola, A.: Surface ultraviolet irradiance from OMI, IEEE T. Geosci. Remote, 44, 1267–1271, 2006.Torres, O., Tanskanen, A., Veihelmann, B., Ahn, C., Braak, R., Bhartia, P. K., Veefkind, P., and Levelt, P.: Aerosols and surface UV products from Ozone Monitoring Instrument observations: an overview, J. Geophys. Res.-Atmos., 112, d24S47, 10.1029/2007JD008809, 2007.van Geffen, J. H. G. M., Boersma, K. F., Van Roozendael, M., Hendrick, F., Mahieu, E., De Smedt, I., Sneep, M., and Veefkind, J. P.: Improved spectral fitting of nitrogen dioxide from OMI in the 405–465 nm window, Atmos. Meas. Tech., 8, 1685–1699, 10.5194/amt-8-1685-2015, 2015.Vasilkov, A., Joiner, J., Spurr, R., Bhartia, P. K., Levelt, P., and Stephens, G.: Evaluation of the OMI cloud pressures derived from rotational Raman scattering by comparisons with other satellite data and radiative transfer simulations, J. Geophys. Res.-Atmos., 113, d15S19, 10.1029/2007JD008689, 2008.
Veefkind, J. P., de Haan, J. F., Brinksma, E. J., Kroon, M., and Levelt, P. F.: Total ozone from the Ozone Monitoring Instrument (OMI) using the DOAS technique, IEEE T. Geosci. Remote, 44, 1239–1244, 2006.
Voors, R., Dobber, M., Dirksen, R., and Levelt, P.: Method of calibration to correct for cloud-induced wavelength shifts in the Aura satellite's Ozone Monitoring Instrument, Appl. Optics, 45, 3652–3658, 2006.Wang, H., Liu, X., Chance, K., González Abad, G., and Chan Miller, C.: Water vapor retrieval from OMI visible spectra, Atmos. Meas. Tech., 7, 1901–1913, 10.5194/amt-7-1901-2014, 2014.
Wang, H., Gonzalez Abad, G., Liu, X., and Chance, K.: Validation and update of OMI Total Column Water Vapor product, Atmos. Chem. Phys., 16, 11379–11393, 10.5194/acp-16-11379-2016, 2016.Wang, L., Newchurch, M. J., Biazar, A., Liu, X., Kuang, S., Khan, M., and Chance, K.: Evaluating AURA/OMI ozone profiles using ozonesonde data and EPA surface measurements for August 2006, Atmos. Environ., 45, 5523–5530, 10.1016/j.atmosenv.2011.06.012, 2011.
WMO: Definition of the tropopause, WMO Bulletin, 6, 136 pp., 1957.Worden, H. M., Logan, J. A., Worden, J. R., Beer, R., Bowman, K., Clough, S. A., Eldering, A., Fisher, B. M., Gunson, M. R., Herman, R. L., Kulawik, S. S., Lampel, M. C., Luo, M., Megretskaia, I. A., Osterman, G. B., and Shephard, M. W.: Comparisons of Tropospheric Emission Spectrometer (TES) ozone profiles to
ozonesondes: Methods and initial results, J. Geophys. Res., 112, D03309, 10.1029/2006jd007258, 2007.