An Overview of Recent ACE-FTS Version 3.5 Validation Studies

An overview of ACE-FTS v3.5
validation studies
Patrick E. Sheese1, Kaley A. Walker1,2, Chris D. Boone2
1University
2University
of Toronto, Toronto, Canada
of Waterloo, Waterloo, Canada
Atmospheric Composition Validation and Evolution, ESA-ESRIN, Frascati, Italy, 20 October 2016
Outline
• ACE-FTS
• Recent validation papers and important findings
• Error budget analysis
• Drift analysis
ACE-FTS
Atmospheric Chemistry Experiment – Fourier Transform Spectrometer
• Canadian satellite SciSat was launched into a circular, highinclination orbit in August 2003
• ACE-FTS and MAESTRO instruments on board
• ACE-FTS is a solar occultation instrument
• High spectral resolution FTS in the 2.2 to 13.3 µm spectral range
• 30+ trace species are retrieved, as well as 20+ subsidiary isotopologues
• Vertical resolution of 3-4 km
• ACE-FTS level 2 version 3.5 data were used in this study
• Complete dataset currently spans 2004-2013
• Data set supplemented with Jan-Apr 2016 data (not yet released)
Validation papers – ACE-FTS/MIPAS/MLS
• ACE-FTS ozone, water vapour, nitrous oxide, nitric acid, and carbon
monoxide profile comparisons with MIPAS and MLS
• Sheese et al., JQSRT, 2016, doi:10.1016/j.jqsrt.2016.06.026
• Compared ACE-FTS O3, H2O, N2O, HNO3, CO with MIPAS (ESA and
IMK-IAA) and MLS
• O3 is within 2% in lower stratosphere, high bias on order of 10-20% above
peak in upper stratosphere
• H2O has dry bias in stratosphere, up to 10%
• CO is much improved over v2.2; MLS summer CO likely should not be
used
Validation papers – ACE-FTS/MIPAS/MLS
Where correlation is greater than 0.8 and standard deviation of relative
difference is less than 50%
Species
O3
H2O
N2 O
HNO3
Altitude
range (km)
10-45
46-60
13-16
17-46
47-70
20-35
MIPAS: 36-44
13-17
18-27
28-38
Mean bias
(%)
+2
0 to +19
-10
-2 to -10
±8
-3
-8
+7
±2
+3 to +19
Coincidence criteria of < 3h, < 350 km
Validation papers – NOy species
• Validation of ACE-FTS version 3.5 NOy species profiles using
correlative satellite measurements
• Sheese et al., AMT (accepted) 2016, doi:10.5194/amt-2016-69, 2016
• Compared ACE-FTS NO, NO2, HNO3, N2O5, ClONO2
• Compare with HALOE, GOMOS, MAESTRO, MIPAS, MLS, OSIRIS, POAM III,
SAGE III, SCIAMACHY, SMILES, and SMR
• Used photochemical box model to scale ACE-FTS local times to other insts
• ACE-FTS NO2 has ~10-15% negative bias above peak in upper
stratosphere
• Evening ACE-FTS N2O5 is quite noisy, only morning data recommended for
use
Validation papers – NOy species
Species
GOMOS, HALOE, MAESTRO,
MIPAS, OSIRIS, POAM III,
SAGE III, SCIAMACHY
NO
HALOE
NO
MIPAS IMKIAA
(Summer only)
𝑟𝑟 > 0.8, 𝜎𝜎 < 50%
Altitude (km)
Mean bias (%)
27-53
-15 to 6
36-52
-9 to 2
17-27
Better than 18
28-41
Better than -15
9-17
Within ±7
18-26
Within ±1
HNO3
MIPAS, MLS, SMR, SMILES
27-35
1 to 20
N2O5
22-34
Better than -7
(Morning only)
35-38
0 to 7
MIPAS
16-24
Better than -20
ClONO2
21-33
Better than -8
Different coincidence criteria for each species
NO2
N2O in mesosphere and lower thermosphere
• Nitrous oxide in the atmosphere: first measurements of a lower
thermospheric source
• Sheese et al., GRL, 2016, doi:10.1002/2015GL067353
• First observations of N2O in the lower thermosphere
• Consistently being produced via energetic particle precipitation
• Is transported down into upper stratosphere during polar winter
• Not actually a validation paper, but we are now starting a project to compare with
WACCM results, with N2O chemistry added to the WACCM (including ion chemistry)
runs
N2O in mesosphere and lower thermosphere
Altitude (km)
January-February
July-August
90
90
80
80
70
70
60
60
50
50
40
40
-80
-40
-60
-20
0
40
20
60
-80
80
-40
-60
-20
40
20
0
60
80
Latitude (deg)
Latitude (deg)
January-March Arctic 7-day mean values
90
80
Altitude (km)
70
60
50
40
M
J05
F
M
J06
F
M
J07
F
M
J08
F
M
J09
Month and year
F
M
J10
F
M
J11
F
M
J12
F
M
J13
F
M
Error budget
• ACE-FTS currently does not have a comprehensive error budget
• For a sample of ACE-FTS occultations, perturb different variables by
their expected uncertainty
• Allow errors to propagate through retrieval
• Calculate 2σ variation of differences from v3.5 retrievals
• Preliminary results for O3, H2O, NO2, and CH4 using 100 sample occultations
• Calculating useful averaging kernels analytically is not possible for
individual ACE-FTS occultations
• Calculated numerically by using synthetic spectra from “true” state profiles
• Perturb true state and calculate difference in retrieval
Error propagation and numeric averaging
kernels
• Inverse instrument signal to noise at
each wavenumber
Measurement Spectroscopic Tangent height
90
O3 synth avg Ker
• Spectroscopic error
• Tangent height error
• Assumed max of ±0.5 km (too large)
• A priori error
• Essentially 0% for species examined
• Pencil beam error
• Error from using single ray in forward
model line of sight. Compare with runs
using 7 rays. Typically on order of 5%
• Still need: p/T error, instrument line
shape error
80
90
70
Altitude (km)
• Line strength and position uncertainty
from HITRAN 2004
100
80
60
Altitude (km)
• Measurement error
2σ propagated error for O3
50
40
30
70
60
50
40
30
20
20
10
10
-20
-10
0
%
10
20
-0.2
0
0.2
0.4
Ak
0.6
0.8
1
Drift analysis
• All coincidences from 2004-2013 and 2016
• Criteria of within ±2 h, ±250 km
• Here using global data
• Daily means of relative differences (ACE-FTS – INST)
• Take linear fit (iterative reweighted least squares), 95%
confidence in slope as error bounds
• Weighted average of values use weights of INST inverse-squared
standard error multiplied by the ACE-FTS to INST correlation
coefficient, i.e.,
𝑟𝑟
• 𝑊𝑊𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 = 2
𝜎𝜎𝑠𝑠
Drift for O3
• Comparisons
with data from
• MIPAS
• ESA v7
• IMK-IAA v5R
• Non-zero
negative drift in
17-43 km, on
order of 2% dec-1
• Non-zero positive
drift near 55 km,
on order of 6%
dec-1
70
70
70
60
60
60
60
50
50
50
50
40
40
40
40
MIPAS ESA
Altitude (km)
• MLS v4.2
• OSIRIS v5.17
70
MIPAS IMK-IAA
MLS
OSIRIS
30
30
30
30
20
20
20
20
10
10
10
10
-10
10
0
-
Drift (% dec
-10
10
0
-
1
)
Drift (% dec
1
)
-20
0
Rel diff (%)
20
Weighted avg
20
0
1
of rel diff (%)
40
Drift for H2O
• Comparisons
with data from
• MIPAS
• MLS
• May be driven
by MLS positive
drift
• Non-zero positive
drift in UTLS
80
80
80
70
70
70
70
60
60
60
60
50
50
50
50
40
40
40
40
30
30
30
30
20
20
20
20
10
10
10
10
MIPAS ESA
Altitude (km)
• Non-zero
negative drift in
28-45 km, on
order of 5% dec-1
80
MIPAS IMK-IAA
MLS
-20
20
0
-
Drift (% dec
-10
10
0
-
1
)
Drift (% dec
1
)
-20
0
20
Rel diff (%)
40
60
Weighted avg
50
0
1
of rel diff (%)
100
Summary
• Validation is fun
• 2 papers published this year on v3.5 validation
• Species common to ACE-FTS, MIPAS, and MLS (O3, H2O, N2O, HNO3, CO)
• NOy species (NO, NO2, HNO3, N2O5, ClONO2)
• Also an N2O paper,
• ACE-FTS N2O will be compared to modified WACCM runs
• More comprehensive error budget calculations are in the works
• Started for O3, NO2, H2O, CH4
• Preliminary results for measurement, spectroscopic, tangent height, a priori, and pencil
beam error
• Still need to calculate p/T, and ILS errors
• Project temporarily on hold due to hardware failures in Waterloo
• Non-zero drift found in:
• O3 – negative drift on order of 2-3%/dec near 15-40 km
• H2O – negative drift on order of 5%/dec near 30-45 km – May be due to MLS positive drift
• Eventually, version 4 will come out, and we’ll get to do it all over again!
Thanks!
The extra bits
Altitude (km)
MIPAS ESA
70
70
70
60
60
60
50
50
50
40
40
40
30
30
30
20
20
20
10
10
10
-10
0
Drift (%/decade)
10
-50
0
Rel diff (%)
50
0
50
1
of rel diff (%)
100
Altitude (km)
MIPAS IMK-IAA
90
90
90
80
80
80
70
70
70
60
60
60
50
50
50
40
40
40
30
30
30
20
20
20
10
10
10
-10
0
Drift (%/decade)
10
-50
0
Rel diff (%)
50
0
50
1
of rel diff (%)
100
Altitude (km)
OSIRIS
55
55
55
50
50
50
45
45
45
40
40
40
35
35
35
30
30
30
25
25
25
20
20
20
15
15
15
10
10
10
-10
0
Drift (%/decade)
10
-50
0
Rel diff (%)
50
0
50
1
of rel diff (%)
100
Altitude (km)
MLS
70
70
70
60
60
60
50
50
50
40
40
40
30
30
30
20
20
20
10
10
-10
0
Drift (%/decade)
10
10
-50
0
Rel diff (%)
50
0
50
1
of rel diff (%)
100
Error propagation – 100
occultations
Mean diff
2σ
50
80
45
70
40
Altitude (km)
Altitude (km)
NO2
90
60
50
40
35
30
25
30
20
20
15
10
10
-10
-20
0
%
H2O
20
10
-20
-10
0
%
CH4
10
20
-20
-10
0
%
10
20
100
70
60
Altitude (km)
80
Altitude (km)
Measurement error
Inverse instrument
signal to noise at
each wavenumber
• Typically on the
order of 1-5%,
greater at upper
altitude limits
where there is
less signal.
Typically less than
the v3.5
statistical fitting
error
O3
60
40
50
40
30
20
20
10
-30
-20
-10
0
10
% 2016
ACE Science Team Meeting, 17 May
20
30
Error propagation – 100
occultations
Mean diff
2σ
50
80
45
70
40
Altitude (km)
Altitude (km)
90
60
50
40
35
30
25
30
20
20
15
10
10
-10
-20
0
%
H2O
10
20
-20
-10
0
%
CH4
10
20
-20
-10
0
%
10
20
100
70
60
Altitude (km)
80
Altitude (km)
Spectroscopic error
Line strength and
position uncertainty
from HITRAN 2004
• Typically on the
order of 1-5% in
stratosphere, 520% in upper
troposphere.
Typically less
than the v3.5
statistical fitting
error
NO2
O3
60
40
50
40
30
20
20
10
-30
-20
-10
0
%
10
ACE Science Team Meeting, 17 May 2016
20
30
Error propagation – 100
occultations
Mean diff
2σ
50
80
45
70
40
Altitude (km)
Altitude (km)
90
60
50
40
35
30
25
30
20
20
15
10
10
-20
-10
0
%
H2O
10
20
-20
-10
0
%
CH4
10
20
-20
-10
0
%
10
20
100
70
60
Altitude (km)
80
Altitude (km)
Tangent height
error
Assumed max of
±0.5 km
• typically
result in 2σ
variation in
VMRs of ~1020%
• 0.5 km is too
large
NO2
O3
60
40
50
40
30
20
20
10
-30
-20
-10
0
10
% 2016
ACE Science Team Meeting, 17 May
20
30
no2
50
80
45
70
40
Altitude (km)
90
60
50
40
35
30
25
30
20
20
15
10
10
-20
-10
0
%
10
20
-20
-10
0
%
ch4
10
20
-20
-10
0
%
10
20
h2o
100
70
60
Altitude (km)
80
Altitude (km)
Pencil beam
error
Altitude (km)
o3
60
40
50
40
30
20
20
10
-20
-10
0
%
10
20
Levenberg-Marquardt least squares fitting
𝐱𝐱 𝑖𝑖+1 =
𝐱𝐱𝑖𝑖 + 𝐊𝐊 𝑇𝑇𝑖𝑖 𝐒𝐒𝑦𝑦−1 𝐊𝐊 𝑖𝑖 + 𝜆𝜆𝑖𝑖 𝐃𝐃𝑖𝑖
𝜕𝜕
𝜕𝜕
𝐆𝐆 =
𝐱𝐱 𝑖𝑖+1 =
𝐱𝐱 𝑖𝑖 + 𝐊𝐊 𝑇𝑇𝑖𝑖 𝐒𝐒𝑦𝑦−1 𝐊𝐊 𝑖𝑖 + 𝜆𝜆𝑖𝑖 𝐃𝐃𝑖𝑖
𝜕𝜕𝐲𝐲
𝜕𝜕𝐲𝐲
= 𝐂𝐂𝑖𝑖+1
𝐂𝐂𝑖𝑖+1 = 𝐂𝐂𝑖𝑖 + 𝐌𝐌𝑖𝑖 𝐊𝐊 𝑇𝑇𝑖𝑖 𝐒𝐒𝑦𝑦−1 (𝐈𝐈 − 𝐊𝐊 𝑖𝑖 𝐂𝐂𝑖𝑖 )
𝐊𝐊 𝑖𝑖 𝐂𝐂𝑖𝑖+1 = 𝐈𝐈
𝜕𝜕𝐱𝐱�
∴ 𝐆𝐆 =
= 𝐂𝐂𝑖𝑖+1 → 𝐊𝐊 −1
𝑖𝑖
𝜕𝜕𝐲𝐲
−1
𝐊𝐊 𝑇𝑇𝑖𝑖 𝐒𝐒𝑦𝑦−1 𝐲𝐲 − 𝐊𝐊 𝑖𝑖 𝐱𝐱 𝑖𝑖
−1
𝐊𝐊 𝑇𝑇𝑖𝑖 𝐒𝐒𝑦𝑦−1 𝐲𝐲 − 𝐊𝐊 𝑖𝑖 𝐱𝐱 𝑖𝑖
= 𝐌𝐌𝑖𝑖
in well behaved retrieval, as 𝑖𝑖 → ∞, 𝐂𝐂𝑖𝑖 → 𝐂𝐂𝑖𝑖+1
∴ 𝐀𝐀 = 𝐆𝐆𝐆𝐆 → 𝐊𝐊 −1
𝑖𝑖 𝐊𝐊 𝑖𝑖 = 𝐈𝐈
WTF is an Ak?
OE synth
Altitude (km)
LS synth
100
100
90
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
-0.2
0
0.2
0.4
Ak
0.6
0.8
1
10
These are similar to each
other, however they
require using simulated
data and are much more
computationally expensive
to produce
-0.2
0
0.2
Altitude (km)
100
90
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
-0.2
0
0.2
0.4
Ak
0.6
0.8
1
OE
LS
100
10
0.4
Ak
0.6
0.8
1
10
Averaging kernels, as
typically calculated, do
not represent the
sensitivity of the
retrieval to the true state
(as numerically
calculated)
-0.2
0
0.2
0.4
Ak
ACE Science Team Meeting, 17 May 2016
0.6
0.8
1
WTF is an Ak?
• As typically calculated in a retrieval
• It is NOT the sensitivity to the true state, where a value of 1 means only
sensitive to the true state and 0 means not at all
• It is the sensitivity to the retrieval at the previous iteration, where a value
of 1 means no sensitivity to the constraint and 0 means you’re only
getting back the constraint
Highlights: ACE/MIPAS/MLS CO
MIPAS__IMK
70
70
Altitude (km)
Altitude (km)
60
50
40
ACE NH winter
IMK NH winter
ACE SH winter
IMK SH winter
ACE NH summer
IMK NH summer
ACE SH summer
IMK SH summer
30
20
-8
10
-7
10
VMR (ppv)
-6
10
70
70
70
60
60
60
50
50
50
40
40
40
40
30
30
30
30
20
20
20
20
60
50
NH winter
SH winter
NH summer
SH summer
0
500 1000
Coincidences
0
0.5
Correlation coeff
1
-50
0
50
Rel diff (%)
0
50
100
1 σ of rel diff (%)
-5
10
MLS
70
Altitude (km)
Altitude (km)
60
50
40
ACE NH winter
MLS NH winter
ACE SH winter
MLS SH winter
ACE NH summer
MLS NH summer
ACE SH summer
MLS SH summer
30
20
-8
10
-7
10
VMR (ppv)
-6
10
70
70
70
70
60
60
60
60
50
50
50
50
40
40
40
40
30
30
30
30
20
20
20
20
0
200
400
Coincidences
0
0.5
Correlation coeff
-5
10
ACE Science Team Meeting, 17 May 2016
1
-50
0
Rel diff (%)
50
0
NH winter
SH winter
NH summer
SH summer
50
100
1 σ of rel diff (%)
Altitude (km)
70
70
70
70
60
60
60
60
50
50
50
50
40
40
40
40
30
30
30
30
20
20
20
20
0
200
400
Coincidences
0
0.5
Correlation coeff
1
-50
0
Rel diff (%)
50
0
NH winter
SH winter
NH summer
SH summer
50
100
1 σ of rel diff (%)
Altitude (km)
MIPAS__IMK
MLS
70
70
60
60
50
50
40
40
ACE NH winter
IMK NH winter
ACE SH winter
IMK SH winter
ACE NH summer
IMK NH summer
ACE SH summer
IMK SH summer
30
20
-8
10
-7
10
VMR (ppv)
-6
10
ACE NH winter
MLS NH winter
ACE SH winter
MLS SH winter
ACE NH summer
MLS NH summer
ACE SH summer
MLS SH summer
30
20
-5
10
-8
10
-7
10
VMR (ppv)
-6
10
-5
10
Global N2O
N2O
55
1487
50
13172
45
17759
40
17759
35
17759
30
17759
25
17759
20
5881
10
0
VMR (ppv)
20
10
Altitude (km )
Altitude (km)
MLS
ACE-FTS – MLS
55
55
55
55
50
50
50
50
45
45
45
45
40
40
40
40
35
35
35
35
30
30
30
30
25
25
25
25
20
20
20
20
0
0.1
0.2
-20
0
20
-50
0
50
0
50
-8
Linear trend correlation coeff
Drift (%/decade)
Rel diff (%)
1
of rel diff (%)
• Agree within ±3% below 26 km; ACE-FTS is ~10% smaller near 28-35 km
• Positive drift of ~5 ppbv/decade near 23 km
100
Global N2O
ACE-FTS – MLS v3
NOTE: Lower altitude
limits are different
between v3 and v4!
Altitude (km)
N2O
50
50
50
50
40
40
40
40
30
30
30
30
20
20
20
20
0
0.02
Linear trend correlation coeff
0.04
0.06
-20
0
20
-50
Drift (%/decade)
0
Rel diff (%)
50
50
0
1
of rel diff (%)
• No significant drift is found when comparing ACE-FTS and MLS v3 N2O
• v3 N2O uses the 640 GHz channel, v4 uses 190 GHz
100