The new GSFC slant column density retrieval of
nitrogen dioxide
Marchenko, S., Celarier, E., Lamsal, L., Krotkov, N., Swartz, B.,
Bucsela, E.
19th OMI STM, 31 Aug.- 2 Sept. 2015
The problem: ~30% high bias traceable to the OMI SCD(NO2),
and the SP V3.0 solution
Before (SP v2.1)
Now (planned SP V3.0)
Monthly Pacific (140W- 180W) zonal-mean NO2 columns for March 2010,
appropriately adjusted for the OMI orbital times and SCIAMACHY swath.
The new GSFC DOAS retrieval
1. Adjustment of the instrumental wavelength shifts combined with
iterative removal of the Ring spectral features in multiple ‘microwindows’; 7 windows in the NO2 retrieval range (402-465 nm).
2. Iterative, sequential estimates of SCDs of the trace gases (NO2, H2O,
CHOCHO) in optimized (broad) windows: e.g., 433-459 nm for CHOCHO.
… attempting to ‘orthogonize’ the relevant trace-gas absorptions
and optimize retrievals of the individual gas species.
3. Iterative removal of the instrumental noise; correction of the fixed
spectral patterns
good performance in the SAA region.
---------------------------------------------------------------------------------------------------
The result: an overall reduction of the OMI SCD(NO2) by 10-45%.
Sequential SCD retrieval
OMI reflectances
Iterative wavelength adjustment and Ringpattern removal in ‘micro-windows’
OMI RS-free reflectances
SCD
retrieval in
broad,
optimized
windows
SCD(NO2): 402-465 nm, omitting 441.5-444.0 nm
SCD(H2O): 440-449 nm
SCD(CHOCHO): 433-459 nm
Instrument
noise removal
Flexible wavelength adjustment and Ring removal: ‘micro-windows’
Ring H2O NO2
Diff. Opt. Depth [%]
1
Ring+NO2+…
(open water)
0
Ring+NO2+…
(Beijing)
-1
‘micro –
windows’
OMI: 20 March 2005, orbit #03610
Flexible wavelength adjustment and Ring removal: ‘micro-windows’
0.001 nm wavelength error
𝟎𝟎. 𝟐𝟐 × 𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏 [molec cm-2] SCD(NO2) change
(van Geffen et al. 2015, AMT 8, 1685)
Our SCD(NO2) errors ~ 𝟎𝟎. 𝟖𝟖 × 𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏 [molec cm-2]
I.e., we are [potentially] sensitive to Δλ > 0.002 nm errors
(~1/100 OMI pix).
Flexible wavelength adjustment and Ring removal: ‘micro-windows’
1. In each ‘micro-window’, irradiance spectrum is offset by Δλ, then splined to radiance
wavelengths
reflectance
2. Iterative polynomial (2nd order) smoothing of the reflectances.
3. Linear fit of the reflectance to the Ring spectrum
4. Removal of the RS : 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 =
(R1 , R2)
𝑅𝑅𝑅𝑅𝑅𝑅
𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 [ 𝑅𝑅1 ∗𝑅𝑅𝑅𝑅+ 𝑅𝑅2 ]
5. Cost function = Standard Deviation of the RS-free reflectances.
6. Δλoptimal at min(Cost function) in each window.
7. In each window, application of Δλoptimal shifts to irradiances; final evaluation of RS
amplitudes.
8. RS smoothing (running mean), then removal of the Ring patterns from the reflectances.
Remember that RS amplitudes are wavelength-dependent.
The Ring-pattern amplitudes: ‘micro-windows’
off-nadir FOV
near-nadir FOV
The Ring line-filling scales for two fields of view (rows).
RS[1] - the first micro-window, 402-410 nm; RS[7] - the last micro-window, 451-465 nm.
20 March 2005, orbit #03610
The wavelength corrections: micro-windows
the farthest off-nadir FOV
Orbital exposures:
the closest to nadir FOV
200—599; 600—999; 1000—1399.
The ±1σ bars characterize the spread within the orbital blocks.
20 March 2005, orbit #03610
RMS of spectral residuals
Single-window
(402-465 nm)
shift-and-squeeze
‘Micro-windows’
20 March 2005, orbit #03610
Spectral cross-talk
Cross-correlation: NO2 absorption and the
Ring line-filling
At the OMI 0.63 nm spectral resolution:
Ring
Ring
NO2
H2O
CHO
CHO
NO2
H2O
CHO
CHO
-0.034
0.005
-0.040
-0.389
-0.089
-0.117
Spectral cross-talk
At the OMI 0.63 nm spectral resolution:
1. NO2 absorptions are typically ~5-7 weaker than RS. CHOCHO is ~5 weaker than
NO2. H2O is ~comparable to RS.
2. Wavelength errors may increase the correlation between NO2 and RS.
3. ~30% RS-amplitude error may lead to ~10% SCD(NO2) bias.
4. One should consider: the relative strengths of the trace-gas absorptions and RS
patterns; the wavelength dependence of the RS and trace-gas spectra; the
spectral resolution.
RS may influence the NO2 retrievals.
NO2 may affect the H2O and CHOCHO retrievals.
Sequential
SCD retrieval
The new SCD(O4) retrieval
1. Temperature-dependent O4 cross-sections from Thalman & Volkamer
(2013) – 5 temperatures btw. 203 - 293 K.
2. Removal of the O3, NO2 and H2O absorptions prior to the O4 retrieval.
See also Yang’s et al. poster.
A few more things to consider:
striping;
saturation, affecting up to ~0.3% of SCD(NO2) in some
cases (spring and fall seasons in the Sun-glint areas);
Solar reference spectrum (fixed vs. dynamic): potentially
small impact on NO2 @ OMI resolution and S/N.
σ=5.0% (as is)
σ=3.3% (single-orbit de-striping)
NO2
diurnal
cycle
20 %
20 March 2005, orbit #03610
Spring2005
NO2
diurnal
rates
(Dirksen
et al.
2011)
Radiances: o#03597 (19 March, 2005), iTime=803
SATURATION
SATURATION
* Rows
#18-38 are marked as potentially saturated
Grey line:
Black line:
long-term (5 years) Solar variability (OMI resolution)
short-term (27-day) Solar changes (OMI resolution)
Marchenko, S. & DeLand, M., 2014, Astrophys. J., 789, 117
The presented approach practically eliminates the ~30% OMI SCD(NO2) bias.
The new GSCF and improved KNMI SCD(NO2) * retrievals agree
to within ~ 5-10%
The new GSFC SCD(NO2) should be available by ~Dec. 2015. You may assess
Y2005 results at GSFC-TLCF:
/omi/live/dd/70003/OMNO2SCD/2005/
See more details and results in:
•
Marchenko, S.V., Krotkov, N.A., Lamsal, L.N., Celarier, E.A., W. H. Swartz, W.H., Bucsela, E.J.,
2015, "Revising the slant-column density retrieval of nitrogen dioxide observed by the Ozone
Monitoring Instrument", JGR-Atmospheres, 120, Issue 11, pp. 5670-5692.
* Improved KNMI NO2 data courtesy Jos van Geffen et al.
Backup
Left panel: Bremen (courtesy Andreas Richter) and Goddard SCD(NO2) retrievals for
different cloud conditions: OMI orbit #03610 from 20 March 2005.
Right panel: the (Goddard / Bremen) ratios (black line) and (Goddard - Bremen) differences
(orange line).
Spectral cross-talk: typical absorption spectra @ OMI resolution
Slant Column Density calculation
1st Trip
Averaged
OMI
irradiances
OMI
radiances
Slant
columns
(1st
approx.)
Fixedpattern
residuals in
reflectance
s
2nd Trip
Slant
columns
(Final
approx.)
Slant
column
uncertaint
ies
SCD calculation
1st Pass
1st Trip
1st Pass
Noise removal
Calculation of fixed-pattern residuals
2nd Pass
Slant column
estimation
Low-pass filt.
1st Pass
Noise removal
RRS signal
removal
Spike
removal
2nd Pass
2nd
Trip
𝜆𝜆 offset
2nd Pass
Slant Column
estimation
Correlation
Check
SCD
Uncertainties