Transcript Time Series - IUP

SCILOV-10
Validation of SCIAMACHY limb
operational NO2 product
F. Azam, K. Weigel, Ralf Bauer, A. Rozanov, M. Weber,
H. Bovensmann and J. P. Burrows
ESA/ESRIN, Frascati, Italy
27-02-2014
Contents:
 SCIAMACHY ESA vs IUP (datasets)
 Validation Strategy
 Validations:
ESA/DLR - IUP Inter-comparisons
Validation with occultation instruments
Validation with limb instrument OSIRIS
 Conclusion/Outlook
1
ESA /DLR vs IUP NO2: main retrieval differences
ESA/DLR
Processors:
Pre-processing
Spectral range
Regularization
Speed Optimized
no
420-470 nm
Optimal regularization
parameter using L- curve
method
IUP
Precision optimized
yes
auxiliary spectral fits for each
tangent height independently,
(improves quality of spectra)
420-450 nm
weak statistical
regularization
(smoothness constrain)
See for details; Rozanov et al, Atmos. Meas. Tech., 4, 1319–1359, 2011
* L-curve: too strong regularization with deteriorating vertical resolution and large smoothing errors
2
Validation Strategy
SCIAMACHY limb coverage:
Profiles/day:1500 profiles for Aug 2002 - Apr 2012: above 4.5 million
(a single measurement is performed for four azimuths of each limb state)
Data versions: ESA/DLR NO2 version 5.02 and IUP version 3.1
Sub-sampling:
ESA SCIAMACHY Sub-sampling (allows for faster computation):
 Distance between two profiles is set larger than 5000 km
 A profile is not allowed in the same 5° latitude band as any of 26 profiles
before
 Each latitude band is limited to 20% more profiles than the average
3
Validation Strategy
Subsampling results in 3% of the entire datasets well distributed over all
latitudes, longitudes and time
ESA – IUP Collocation criteria: time = 0.001 h, distance = 60 km
Reason: although same measurements but the determination of the horizontal
positions of the tangent point is different for both datasets.
4
Statistics
Standard deviation of the bias corrected difference
Standard error of the mean (too small to be visible on plots)
Dupuy et al. (2009)
Mean relative difference
Results will be shown as profiles, annual cycles and time series comparisons
for 30° latitude bins.
5
ESA/DLR v5.02 – IUP v3.1
ESA/DLR vs IUP Profile comparisons
mean
relative
differences
Tropics
Near global
NH mid lat.
SH mid lat.
NH high lat.
SH high lat.
6
Annual cycle
Annual cycles comparisons for high lat. show large differences in winter month
(low NO2 concentrations)
60N to 90N
90S to 60S
7
Time series
30S to 30N
60N to 90N
90S to 60S
7
ESA/DLR v5.02 – Occultation instruments
 Occultation instruments and their NO2 datasets:
ACE-FTS: version 3.0
(2002-2010)
HALOE: version 19
(2002-2005)
SAGE II: version 6.2
(2002-2005)
Note: SAGE II sunrise (SR) and sunset (SS) events are separately compared
sunrise events suffer from instrumental problem (Bauer et al., 2012). The
sunset events have a better quality but are mainly restricted to the northern mid
and high latitudes
ESA – Occultation instruemnts Collocation criteria:
time = 6 h, distance = 1000 km
9
Comparisons with occultation instruments
To Consider:
1)- Strong diurnal variation of NO2: pronounced at sunrise and sunset,
SZA ≥ 90 largely effected
a)- Changing illumination
b)- Different instruments
conditions (SZA) along the
line of sight
measure at different SZA
diurnal effect error
Photochemical correction
needed
usually larger for occultation
instruments below 25 km
(e.g. Bauer et al., 2012
McLinden et al., 2006)
10
Photochemical Correction Application:
 Look up table with precalculated diurnal cycles provided by Chris McLinden,
model from the University of California (McLinden et al, 2000, Prather 1992)
 2 km vertical resolution and 2.5° latitude grid size.
 For each collocation pair, matching geolocations and SZAs read in lookup-table
 Calculation of scaling factors by dividing profile from look up table at
SCIAMACHY SZA by the profile corresponding to the SZA of the other
instrument
 Scaling factors applied to the NO2 profile of the other instrument
Estimated errors introduced by this photochemical correction ~ 20%
Bracher et al. 2005
11
To Consider:
2)- Avoid comparison of profiles at different vortex conditions and air
masses
 Strong horizontal gradients at the edge of the polar vortex
 Modified potential vorticity (MPV) calculated from ECMWF-Interim
 Profiles polewards of 35° latitude are excluded, if:
 the MPV is > 30 and < 40 PVU (vortex edge)
 their MPV differs by more than 3 PVU
12
ESA/DLR vs Occultation profile comparisons
13
ESA/DLR vs Occultation profile comparisons
14
ESA/DLR v5.02 – OSIRIS
OSIRIS version 3.0 is used.
OSIRIS coverage is near global with the exception of the winter hemispheres
ESA –OSIRIS Collocation criteria:
time = 12 h, distance = 1000 km
(OSIRIS performs measurements in a sun-synchronous orbit with an equator
crossing time of the ascending node at 18:00 local solar time)
Note:
Based on OSIRIS comparisons with other instruments, the precision of OSIRIS
NO2 measurements is observed to be16% for 15–25 km and 6% between 25
and 35 km
Reference: (http://osirus.usask.ca/?q=node/245)
15
ESA/DLR vs OSIRIS profile comparisons
mean
relative
differences
Tropics
Near global
NH mid lat.
SH mid lat.
NH high lat.
SH high lat.
6
Time series
30S to 30N
60N to 90N
90S to 60S
7
Summary time series plots
Time series
Tropics
18
Time series
NH high lat.
19
Time series
SH high lat.
20
Conclusions
ESA-IUP
Large differences observed in the high latitude winter, elsewhere agreement
within a few percents
Retrieval differences (regularization) probable cause of the differences
ESA-Occultation
25–35 km:ACE-FTS ~10% , HALOE ~15% and SAGE II ~20% on the average
20 -25 km:- differences with the instruments may approach 50%.
Below 20 km and above 40 km: large biases observed
Diurnal effect error a potential dominant source of differences below 25
ESA-OSIRIS
Good agreement within the precision limit of OSIRIS
21
Outlook/Recommendations
For ESA/DLR limb NO2, retireval should be precision optimized, studies on the
choice of photochemical correction should be part of future validation activities
23
Extra Slides
Time series
NH mid.lat
Time series
SH mid lat.
Results
Profile comparisons:
mean relative differences plots with the standard deviation of the bias corrected
differences for 20-35 km
Annual cycles:
annual cycle vs altitude plots as monthly mean absolute amounts, monthy
mean percental difference and the monthly mean percental differences for
selected altitudes (20, 24, 27 and 31 km)
Time Series:
compared for 20–35 km on a monthly grid. For the selected altitudes (20, 24, 27
and 31 km), comparisons carried out on 30 days running averages if more than
10 collocations are found
Photochemical correction effect:
Mean latitude-altitude cross section (monthly averages in 10° bins)
MPV criteria effect:
Mean latitude-altitude cross section (monthly averages in 10° bins)