The Wind Lidar Mission ADM-Aeolus
Data Processing
David Tan
Research Department
ECMWF
Acknowledgements:
ESA (Mission Science & Aeolus project team)
Aeolus Mission Advisory Group
July 2008
Data
Processing for ADM-Aeolus
– LWG Wintergreen
Level-1B/2A/2B
Development
Teams& JCSDA
Slide 1
Contents
 Summary
on 2 Slides
 Background
 Data Processing


Assimilation of Level-2B hlos wind

Simulations of Level-2B hlos wind data

Assimilation impact study
Level-2B processor development

How to make operational Level-2B hlos

Algorithms & rationale

Validation
 Conclusions
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 2
Summary of ECMWF activities for ADM-Aeolus
 Prepared for assimilating L2B hlos wind

2002-04, example for other centres
 Developing

ECMWF is lead institute, 5 sub-contractors

2004-present
 Other
ongoing work/operational phase

MAG, GSOV, Cal/Val, In-orbit commissioning

ECMWF to generate operational L2B/L2C
products, monitor & assimilate Aeolus data,
assess impact on NWP

Maintain, develop & distribute L2B processor

July 2008
Level-2B processor
On behalf of ESA, using NWP-SAF approach
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 3
Status summary: Day-1 system on track
1. Level-2B hlos winds – primary product for assimilation
a. Account for more effects than L1B products
b. Will be generated in several environments
c. Motivated strategy to distribute source code
2. Main algorithm components developed & validated
a. Release 1.33 available – development/beta-testing
b. Documentation and Installation Tests
c. Portable – tested on several Linux platforms
3. Ongoing scientific and technical development
a. Sensitivity to inputs, QC/screening, weighting options
4. Contact points – ESA and/or ECMWF
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 4
Contents
 Summary
on 2 Slides
 Background
 Data Processing
 Conclusions
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 5
Atmospheric Dynamics Mission ADM-Aeolus
ADM-Aeolus with single payload
Atmospheric LAser Doppler INstrument
ALADIN



[H]LOS



July 2008
Observations of Line-of-Sight LOS wind profiles in
troposphere to lower stratosphere up to 30 km
with vertical resolution from 250 m - 2 km
horizontal averages over 50 km every 200 km
(measurements downlinked at 1km scale)
Vertical sampling with 25 range gates can be
varied up to 8 times during one orbit
High requirement on random error of HLOS
<1 m/s (z=0-2 km, for Δz=0.5 km)
<2 m/s (z=2-16 km, for Δz= 1 km),
unknown bias <0.4 m/s and linearity error <0.7 %
of actual wind speed; HLOS: projection on
horizontal of LOS => LOS accuracy = 0.6*HLOS
Operating @ 355 nm with spectrometers for
molecular Rayleigh and aerosol/cloud Mie
backscatter
First wind lidar and first High Spectral Resolution
Lidar HSRL in space to obtain aerosol/cloud
optical properties (backscatter and extinction
coefficients)
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 6
ADM-Aeolus Coverage and Data Availability
 3200 wind profiles per day:
about factor 3 more than radiosondes
 3 hour data availability after observation
(NRT-Service) => 1 data-downlink per orbit;
30 minutes data availability for parts of orbit
(QRT-Service with late start of downlink)
 launch date May 2010 (consolidated launch
date prediction in some months expected)
 mission lifetime 39 months: observations
from 2010-2012
ADM-Aeolus Science Report
(ESA publication SP-1311, 2008)
TELLUS 60A(2), Mar 2008 special issue on
ADM-Aeolus workshop 2006
50 km observations during 6 hour period
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 7
ADM-Aeolus Ground Segment
[ESA-ESRIN]
[ESA-ESOC]
[DLR]
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 10
ADM-Aeolus Data Products
Product
Level 0
Level 1b
Contents
Processor developer
and location
Time ordered source packets with ALADIN measurement &
housekeeping data
MDA (Canada)
Geo-located, calibrated observational data
MDA (Canada)
preliminary HLOS wind profiles (standard atmosphere used
in Rayleigh processing) – not suitable for assimilation
Size in
MByte/orbit
47
Tromsø (Norway)
10-15 (BUFR)
+

spectrometer readouts at “measurement” scale ( 1-5 km )
– input for Level 2a/b processing


Level 2a
Tromsø (Norway)
22 (EE XML
Format)
DLR-IMF (Germany)
12
viewing geometry & scene geo-location data
Supplementary product
 Cloud profiles, coverage, cloud top heights
 Aerosol extinction and backscatter profiles, ground
reflectance, optical depth
Level 2b
Meteorologically representative HLOS wind observations
HLOS wind profiles at “observation” scale ( ~ 50 km )
suitable for assimilation - temperature T and pressure p
(Rayleigh-Brillouin) correction applied with ECMWF (or
other) model T and p
Level 2c
Aeolus assisted wind vector product
Vertical wind profiles (u and v component);
Tromsø (Norway)
ECMWF
Reading (UK)
(and other
NWP/research centres)
ECMWF
Reading (UK)
NWP model output after assimilation of Aeolus HLOS wind
July 2008
18
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 12
22
Ongoing ADM-Aeolus Scientific Studies
Title
Team
Consolidation of ADM-Aeolus Ground Processing including
L2A Products
DLR Germany
Development and Production of Aeolus Wind Data Products
ECMWF UK
Météo-France, KNMI, IPSL, PSol
Météo-France, KNMI, IPSL, DLR, DoRIT
ADM-Aeolus Campaigns
DLR Germany
Météo-France, KNMI, IPSL, DWD, MIM
Optimisation of spatial and temporal sampling
KNMI Netherlands
Tropical dynamics and equatorial waves
MISU Sweden
Rayleigh-Brillouin Scattering Experiment
tbd
ESA plans an Announcement of Opportunity AO for ADM-Aeolus scientific use of data
for late 2008 – distinct from the AO for Cal/Val
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 13
Principle of wind measurement with ALADIN
Atmospheric LAser Doppler INstrument
ALADIN


Principle of spectrometer for molecular signal



principle of spectrometer
for aerosol signal
July 2008
Direct-Detection Doppler Lidar at 355 nm
with 2 spectrometers to analyse backscatter
signal from molecules (Rayleigh) and
aerosol/clouds (Mie)
Double edge technique for spectrally broad
molecular return, e.g. NASA GLOW
instrument (Gentry et al. 2000), but
sequential implementation
Fizeau spectrometer for spectrally small
aerosol/cloud return
Uses Accumulation CCD as detector => high
quantum efficiency >0.8 and quasi-photon
counting mode
ALADIN is a High-Spectral Resolution Lidar
HSRL with 3 channels: 2 for molecular
signal, 1 for aerosol/cloud signal => retrieval
of profiles of aerosol/cloud optical properties
possible
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Fig. U. Paffrath
Slide 14
Contents
 Summary
on 2 Slides
 Background
 Data Processing


Assimilation of Level-2B hlos wind

Simulations of Level-2B hlos wind data

Assimilation impact study
Level-2B processor development

How to make operational Level-2B hlos

Algorithms & rationale

Validation
 Conclusions
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 17
ADM-Aeolus Ground Segment
[ESA-ESRIN]
[ESA-ESOC]
L2B HLOS
[DLR]
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 19
L2B data simulated using ECMWF clouds …



90% of Rayleigh
data have accuracy
better than 2 m/s
In priority areas
(filling data gaps in
tropics & over
oceans)
Complemented by
good Mie data from
cloud-tops/cirrus
(5 to 10%)
ECMWF Analysis VT:Friday 7 February 2003 00UTC Surface: **high cloud cover
Yield (data meeting mission
requirements in % terms) at 10 km
Rayleigh channel Level 8
150°W
120°W
90°W
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
100%
75%
60°N
60°N
50%
25%
30°N
30°N
1%
0°
0°
0.9
30°S
30°S
0.8
0.7
60°S
60°S
0.6
ECMWF Analysis VT:Friday 7 February 2003 00UTC Surface: **high cloud cover
Mie channel Level 8
150°W
120°W
90°W
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
150°W
120°W
90°W
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
Tan & Andersson
QJRMS 2005
July 2008
0.5
100%
75%
60°N
60°N
50%
25%
30°N
30°N
1%
0°
0°
1.0
0.9
30°S
30°S
0.8
0.7
60°S

1.0
60°S
0.6
0.5
150°W
120°W
90°W
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
LIPAS-simulated HLOS data – operational processors later
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 20
… & impact studied via assimilation ensembles
12-hr fc impact
(Tan et al QJRMS 2007)
Spread in zonal wind (U, m/s)
S.Hem
100
Pressure (hPa)
p<0.001
Scaling factor ~ 2 for wind error
Tropics, N. & S. Hem all similar
ADM-Aeolus
Simulated DWL adds value at all
altitudes and in longer-range
forecasts (T+48,T+120)
Differences significant (T-test)
Supported by information content
diagnostics
NoSondes
1000
0.0
0.5
1.0
1.5
Cheaper than OSSEs
Zonal wind (m/s)
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 21
Assimilation of prototype ADM-Aeolus data
2003/4: introduced L2B hlos as new observed quantity in 4d-Var
Prototype Level-2B (LIPAS
simulation, includes
representativeness error)
Observation Processing
Data Flow at ECMWF
Non-IFS processing
“Bufr2ODB”
Convert BUFR to ODB format
Recognize HLOS as new known observable
Observation Screening
IFS “Screening Job”
Check completeness of report, blacklisting
Background Quality Control
Assimilation Algorithm
IFS “4D-VAR”
Implement HLOS in FWD, TL & ADJ Codes
Variational Quality Control
Diagnostic post-processing
“Obstat” etc (Lars Isaksen)
Recognize HLOS for statistics
Rms, bias, histograms
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 23
Analysis
Assimilation of prototype ADM-Aeolus data
2004-: Receive L1B data & L2B processing at NWP centres
Observation Processing
Level-1B data
(67 1-km measurements)
Non-IFS processing
Data Flow at ECMWF
“Bufr2ODB”
Convert BUFR to ODB format
Recognize HLOS as new known observable
Observation Screening
IFS “Screening Job”
Check completeness of report, blacklisting
Background Quality Control
Assimilation Algorithm
IFS “4D-VAR”
Implement HLOS in FWD, TL & ADJ Codes
Variational Quality Control
Diagnostic post-processing
“Obstat” etc (Lars Isaksen)
Recognize HLOS for statistics
Rms, bias, histograms
L2BP (1 50-km observation)
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 24
Analysis
Level-2B processor will run in different environments
ECMWF will supply source code - use as standalone or callable subroutine
Aeolus Ground Segment & Data Flows - schematic view
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 25
Retrievals account for receiver properties …
Tan et al Tellus
60A(2) 2008

Dabas et al same
issue

Mie light reflected
into Rayleigh channel

Rayleigh wind
algorithm includes
correction term
involving scattering
ratio (s)

ADM-Aeolus Optical Receiver - Astrium Satellites
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 26
… and for atmospheric scattering properties
ILIAD – Impact of P & T and backscatter ratio on Rayleigh Responses - Dabas Meteo-France, Flamant IPSL
Response curve
Function of P, T and s
Mie
Rayleigh
(P,T)
Line shape
Depends on
P, T and s
Transmission
through the
Double FP
Light transmitted through TA and TB
 1km-scale

July 2008
Response (function of fD)
spectra are selectively averaged
Account for atmospheric variability - improve SNR
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 27
BRC4
Retrievals validated for idealized broken multi-layer
clouds – E2S simulator + operational processing chain
Cloud5
Cloud4
Cloud3
Cloud2
Specified
cloud
layers
Retrieved
clouds and
aerosol
Specified
wind=50
m/s
Retrieved Rayleigh
winds are accurate
in non-cloudy air
Classify scene (threshold)
then average cloudy/noncloudy regions separately
50 km
July 2008
Retrieved Mie
winds are accurate
in cloud and
aerosol layers
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 28
Realistic scenes simulated

Real scattering
measurements obtained
from the LITE and
Calipso missions
ESA’s software (E2S) is
used to simulate what
ADM-Aeolus would ‘see’

The L1B software
retrieves scattering
ratio at the 1 km
measurement resolution
Our input not perfect
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 29
Wind retrieval validated in the presence of
heterogeneous clouds and wind – E2S simulation
Retrievals fairly accurate
Mie
particles
Rayleigh
molecular
Level-1B
Backscatter from Calipso
July 2008
Outliers being examined
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 30
… but only after bugs were fixed in earlier versions
of the L1B processor
Level-1B
BRC #2
Rayleigh
molecular
25
E2S input (238)
2b Rayleigh Cloudy (11)
2b Rayleigh Clear (14)
1b Rayleigh obs (24)
2b Mie Cloudy (12)
2b Mie Clear (14)
1b Mie obs (8)
Altitude [km]
20
15
Mie
particles
10
July 2008
5
0
-100
-80
-60
-40
-20
0
20
HLOS wind [m/s]
40
60
80
100
Retrieved Mie winds revealed
systematic error in L1B input
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 32
Wind retrieval error from ACCD digitization
- theory confirmed by E2S simulation
Photon noise will dominate
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 33
Level-2B hlos error estimates – reqts met
Poli/Dabas
Meteo-France
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 34
Contents
 Summary
on 2 Slides
 Background
 Data Processing
 Conclusions
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 35
Conclusions – Day-1 system on track
1. Level-2B hlos winds – primary product for assimilation
a. Account for more effects than L1B products
b. Will be generated in several environments
c. Motivated strategy to distribute source code
2. Main algorithm components developed & validated
a. Release 1.33 available – development/beta-testing
b. Documentation and Installation Tests
c. Portable – tested on several Linux platforms
3. Ongoing scientific and technical development
a. Sensitivity to inputs, QC/screening, weighting options
4. Contact points – ESA and/or ECMWF
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 36
5.2 Key assimilation operators

Tan 2008 ECMWF Seminar Proceedings

HLOS, TL and AD


H
=
- u sin φ - v cos φ

dH
=
- du sin φ - dv cos φ

dH*
= ( - dy sin φ, - dy cos φ )T

Generalize to layer averages later
Background error

Same as for u and v (assuming isotropy)

Persistence and/or representativeness error

Prototype quality control

July 2008
Adapt local practice for u and v
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 38
5.1 Prototype Level-2C Processing
 Ingestion
 Assimilation
of L1B.bufr into
the assimilation system

of HLOS
observations (L1B/L2B)
L1B obs locations within
ODB (internal
Observation DataBase)

Corresponding analysis
increments (Z100)
Diff in RMS of an-Incr: RMS(an_erhg - bg_erhg) - RMS(an_ercp - bg_ercp)
Lev=100, Par=Z, anDate=20041002-20041002 0Z, Ana Step=0, Fc Step=6
NH=0.93 SH= 2.92 Trop= 1.35 Eur=-0.64 NAmer= 0.12 NAtl= 1.58 NPac= -0.29
120°W
60°W
0°
60°E
120°E
2.5
1600
2
1600
60°N
1600
1640
60°N
0
164
30°N
1.5
1
30°N
0.5
0.25
0°
30°S
60°S
0°
0.1
-0.1
-0.25
1640
1640
1600
1560
1520
1480
1640
1560
1520
-0.5
30°S
-1
1600
1480
1560
60°S
1520
0.5
1480
-0.5
-1.5
-2
-2.5
120°W
July 2008
60°W
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
0°
Slide 53
60°E
120°E
2a-4. Other NWP configurations
L2C Product
Meteorological
(optional)
products
Assimilation
(L2C Processor)
Other observations
Possibility to modify
algorithm source
code.
L1B Product
July 2008
L2B Product
L2B Processor
Auxiliary Data
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Forecast Model
Integrated or Standalone,
other options possible.
Processing
Parameters
Slide 54
2a-1. ECMWF “operational” configuration
L2C Product
Meteorological
products
Assimilation
(L2C Processor)
Other observations
L2B Product
L2B Processor
L1B Product
July 2008
Auxiliary Data
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Forecast Model
Integrated with
assimilation system
Processing
Parameters
Slide 58
2a-2. ESA-LTA late- and re-processing
L2B Product
L2B Processor
L1B Product
July 2008
Auxiliary Data
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Standalone configuration
Processing
Parameters
Slide 59
2a-3. Research/general scientific use
Other observations
Possibility to modify
algorithm source
code.
L1B Product
July 2008
L2B Product
L2B Processor
Auxiliary Data
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Forecast Model
Standalone configuration.
Processing
Parameters
Slide 60
1.3 Integration of Aeolus L2BP at ECMWF
L1B preprocessing (if required)
L1B/2B/2C
IFS(ODB) allocation
AuxMet/L2B
in Screening
L2C
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 61
1.3 ECMWF operational schedule
Processing of L1B 09-21Z starts at 02Z (D+1) “dcda-12utc”
Aux
L1B
L2B
L2C
Orbit from 20:00--21:40Z
is split over two
assimilation cycles
AP/L2BP
within
Screening
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 62
2b. Why distribute L2BP Source Code?
 Distribution of executable binaries only permits
─ limited number of computing platforms
─ different settings in processing parameters input file
─ thresholds for QC, cloud detection
─ different auxiliary inputs
─ option to use own meteorological data (T & p) in
place of ECMWF aux met data (available from LTA)
 Provide maximum flexibility for other centres/institutes
to generate their own products
─ different operational schedule/assimilation strategy
─ scope to improve algorithms
─ feed into new releases of the operational processor
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 63
3a. How it works – Tan et al Tellus A 2008
 Rayleigh channel HLOS retrieval – Dabas et al, Tellus A
─ R = (A-B) / (A+B) and HLOS = F-1 (R;T,p,s)
─ T and p are auxiliary inputs
─ correction for Mie contamination, using estimate of
scattering ratio s
 Mie channel HLOS retrieval
─ peak-finding algorithm (4-parameter fit as per L1B)
 Retrieval inputs are scene-weighted
─ ACCD = Σ ACCDm Wm, Wm between 0 and 1
 Error estimate provided for every Rayleigh & Mie hlos
─ dominant contributions are SNR in each channel
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 64
3b. Level-2B input screening & feature finding
5
3
Poli/Dabas
Meteo-France
calcu- 3: Level-2b classi cat ion map for each measurement bin (t op) and reFigure 1: Scat t ering rat io input t o t he E2S (t op) and scat t ering rat ioFigure
lat ed by t he level-1b processor (bot t om)
sult of t he L1B Input Screening process for each measurement bin (bot t om),
July 2008
Data Processing for ADM-Aeolus – LWGfor
Wintergreen
& JCSDA
Rayleigh
Slide 66
4
3b. Level-2B hlos wind retrievals
5
BRC #2
20
18
E2S input (206 excld:0)
2b Rayleigh Cloudy (4 excld:0)
2b Rayleigh Clear (16 excld:0)
1b Rayleigh obs (24 excld:0)
2b Mie Cloudy (3 excld:3)
2b Mie Clear (2 excld:2)
1b Mie obs (2 excld:0)
16
Altitude [km]
14
12
10
8
6
4
2
Poli/Dabas
0
Meteo-France
- 100
- 80
- 60
- 40
- 20
0
20
HLOS wind [m/s]
40
60
80
100
Figure 2: Level-2b scat t ering rat io used in t he Rayleigh inversions (calculat
ed 3: Level-2b classi cat ion map for each measurement bin (t op) and reFigure
by weight ing over t he level-1b scat t ering rat ios) for t he clear and sult
cloudy
of t he L1B Input Screening process for each measurement bin (bot t om),
observat
Julyions
2008 Data Processing for ADM-Aeolus – LWGfor
Wintergreen
& JCSDA
Slide 67
Rayleigh
Figure 7: Retrievals for the 2nd BRC
3c. Future work
 Quality Indicators
─ Highlighting doubtful L2B retrievals
─ More complicated atmospheric scenes from
simulations + Airborne Demonstrator
 Advanced feature-finding/optical retrievals
─ Methods based on NWP T & p introduce error
correlations
 Modified measurement weights
─ More weight to measurements with high SNR?
 Height assignment
─ In situations with aerosol and vertical shear
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 69
4. Distribution of L2BP software
 Software releases issued by ECMWF/ESA
─ Details & timings to be determined
─ Probably via registration with ECMWF and/or ESA
─ Source code and scripts for installation
─ Fortran90, some C support
─ Developed/tested under several compilers
─ Suite of unit tests with expected test output
─ Documentation
─ Software Release Note
─ Software Users’ Manual
─ Definitions of file formats (IODD), ATBD, etc.
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 70
Conclusions
 Expectations for ADM-Aeolus are high
─ On track for producing major benefits in NWP
─ Meeting the mission requirements for vertical
resolution & accuracy
─ Extending to stratosphere, re-analysis
─ Our software available to NWP/science community
─ Combine with other observations
─ Height assignment for AMVs
─ Complement other cloud/aerosol missions
─ Related research
─ Background error specification
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 71
1

Baseline L2BP Algorithm
Purpose of L2BP

Produce L2B data from L1B data and aux met data
 50
km observations from ~ 1 km measurements
 Error


estimates and quality indicators

Temperature and pressure corrections via met data

Scene classification and selective averaging
Design a portable source code for three processing modes

Integration at many met centres

Reprocessing @ ESA (ECMWF-supplied met data)

Testing in a range of environments

Simple to use, yet flexible to permit extensions
Auxiliary processing – prepares met data as L2BP input
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 75
Scene classification influences L2B output
24 Mie or Rayleigh height bins
P L1 measurements
L2B Profiles
L2bP
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 76
1.4 Baseline architecture – L2BP

Auxiliary L2B processing (centre-dependent)



July 2008
Profiles of temperature and pressure vs height

At requested locations, full model vertical resolution

L2BP will perform conversion to WGS84 coords
Extract from “first-guess fields” during “screening”

Nearest time (within 15 mins at ECMWF)

At ECMWF, vertical profiles and not slanted

Currently one profile per observation
Pre-processing step

Standardize input for primary L2B processing

Align met data with L1B measurements in horizontal

Could be achieved via extrapolation or interpolation
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 77
1.5 Locations for computing aux met data

Obtain from geolocation information in real L1B data

Offset from the sub-satellite track

Example shows 30 mins x 50 km spacing along-track
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 78
1.4 L2BP - auxiliary pre-processing
 Collocation implemented, suitable for 1 met locn per BRC

Sensitivity study to guide extensions, eg interpolation code
P L1 measurement locations
Aux met height bins, 1 per model level
N_met raw aux met profiles
PreProc
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 79
1.5

L2BP - primary processing
Primary processing (HLOS retrieval)

L1B product validation (mainly in Consolidation Phase)
 Signal classification (+ further code from L2A study)
 Assign weights to signals (+ further development)

Apply weights to a general parameter
 lat & lon = L2B centre-of-gravity

temperature & pressure = Tref & Pref

HLOS temperature & pressure corrections

Error estimates, quality indices

Output in EE format
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 80
2

Future work
Key inputs from other activities

L1B test datasets

Cloud detection and scene classification


Details of temperature & pressure correction scheme


ILIAD results & implementation (e.g. lookup table)
Algorithms for


algorithms/codes based on L2AP
HLOS error estimates & Quality indicators
Check suitability of interfaces for many met centres

July 2008
Basic concept ~ screening of radiosonde observations
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 81
Facts and figures for ADM-Aeolus

ESA point of contact – Dr Paul Ingmann

Mission Experts Division, ESA/ESTEC, The Netherlands
Orbit
Sun-synchronous
Dawn-dusk
97 
408 km
1100 kg
450 kg
Nd:YAG,
frequency tripled
to 355 nm
150 mJ
100 Hz
10 s every 28 s
- inclination & altitude
Mass - total & “ALADIN” lidar component
Transmitter
- laser type & pulse energy
- pulse repetition freq. & duty cycle
Receiver - telescope diameter
- spectrometers
Average power demand
Launch date & mission lifetime
July 2008
1.5 m
Fizeau (Mie)
Dual edge etalon
(Rayleigh)
1400 W
2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 82
3 years
1


Baseline L2BP Algorithm
Baseline architecture

HLOS retrieval - TN2.2, Fig 2

Generation of aux met data – TN2.2, Fig 1
L1B BRCs processed independently (& possibly in parallel)


L1B data arriving within met centre operational schedule


No communication of intermediate L2BP results
Met centre produces aux met data, L2B (and L2C)
L1B data missing the ECMWF schedule

ECMWF produces aux met data


July 2008
at locations inferred from predicted flight tracks
L2B possible via re-processing
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 83
1.1 L2BP – Portability considerations

Common design accommodating three processing modes
Met Centres
Operational (ECMWF)
Re-proc (ESRIN)
L1B data (input) in EE format
Received in
Received in NRT (~5h)
LTA/reprocessing
(or predicted orbit locations)
Q/NRT (30m-3h)
Auxiliary meteorological input
Self-generated
Self-generated & sent
Oper available
to LTA
(via LTA)
(T & p profiles, EE/BUFR)
Primary L2BP code
Oper available
Oper
Oper available
Auxiliary parameter input files
Oper available
Oper
Oper available
L2B data output in EE format
Yes
Yes
Yes
L1B/L2B data in BUFR format
EE2BUFR
EE2BUFR
Not required
(for assimilation purposes)
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 84
The ILIAD Study


Why the ILIAD study ?

The L1 processing scheme proposed by the industry for
Rayleigh winds does not take into account the impact of the
pressure and the potential presence of Mie scattering.

Preliminary studies conducted by DLR (O. Reitebuch) and
ESA (M. Endemann) suggested the impact of both exceed
requirements on data quality.
Objectives


Find a correction scheme.
Study Team.

July 2008
IPSL/LMD (P. Flamant, C. Loth), IPSL/SA (A. Garnier),
ONERA/DOTA (A. Dolfi-Bouteyre), HOVEMERE (D. Rees),
MF/CNRM (A. Dabas, M. L. Denneulin)
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 85
ILIAD - Impact of P, T and backscatter ratio
on Rayleigh Responses
Response curve
Function of P, T and r
Mie
Rayleigh
(P,T)
Line shape
Depends on
P, T and r
Transmission
through the
Double FP
Light transmitted through TA and TB
July 2008
Response (function of fD)
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 86
ILIAD - Baseline Inversion Scheme
Aux. data
Ifd;[P, T, r])  IR fd;[P, T])  r  1)IM fd )
Model
P,T
Calibration
TA(), TB()
Line shape
Tenti
+Gauss.
L1Bp
r̂
NA,B fD; [P, T, r]) 
 TA,B f )If  fD;[P, T, r])df
NA, NB
USR
Output
v̂ r ,
v̂R P, T, r) 
dv̂ r dv̂ r
,
dT dP

RR(fD,[P,T,r])
Inversion
)
 1
RR R̂R; [P, T, r]
2
RR fd, [P, T, r]) 
R̂R
NA fd, [P, T, r])  NB fd, [P, T, r])
NA fd, [P, T, r])  NB fd, [P, T, r])
Input
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 87
ILIAD - Simplified correction scheme
Based on a simplification of baseline inversion.
Two-step appraoch:
1.
Inverse response RR as if there were no Mie.

Method: Look-up in the 3D matrix Fd (i,j,k) giving the inverse frequency
(or velocity) for pressures Pi=P0+iDP, Tj=T0+jDT and Rk=R0+kDR

Output parameters:

Vr(Pmod,Tmod,r=1) where Pmod and Tmod are the pressure and temperature
inside the sensing volume as predicted by the NWP model.

dvr/dP, dvr/dT and dvr/dR, that is, the first order derivative of vr with
respect to P, T and the response RR.
2.
Correct from Mie contamination.

July 2008
Method: First order, linear correction based on the estimation of dvr/dr
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 88
ILIAD - Practical implementation
ISR
Rayleigh-Brillouin
spectra
R̂ R
TA(fm), TB(fm)
Df=25 MHz
Pmod, Tmod
r̂
IR fm; [Pi, Tj ])
Df  25 MHz
DP  50 hPa
DT  1 K
Global aux. File
Processed once before launch
~8 Mb
July 2008
Fd Pi, Tj, Rk )
INVERSION
DR  0.01
Look-up table
Aux. File
Processed every time
an ISR is carried out
~2 Mb
v̂r Pmod, Tmod, ˆ
r)
v̂r
P
v̂r
T
Part of L2Bp
Applied to all Rayleigh responses
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 89
Aeolus satellite layout
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 90
ALADIN Structure and
Optical Structural Thermal Model
ALADIN structure has been
completed for OSTM and tested.
Mass-dummies have been
integrated for OSTM:
Power Laser Heads (PLH),
Reference Laser Heads (RLH),
and
Optical Bench Assembly (OBA)
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 91
ALADIN OSTM
Laser Cooling
Radiator
TransmitReceive
Telescope
Platform
simulator
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 92
ALADIN Laser Cooling System
Laser Radiator
Heatpipes
OBA
Redundant
PLH
July 2008
Nominal
PLH
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 93
ALADIN
OSTM
Before shipment to CSL (Liege) for installation in
vacuum chamber and full thermal vacuum testing
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 94
Data simulations for ADM-Aeolus
Yield (%age of data meeting
mission requirements) at 5 & 1 km
ECMWF Analysis VT:Friday 7 February 2003 00UTC Surface: **high cloud cover

5 km: 75% of
Rayleigh have
accuracy < 2 m/s
(also 15% Mie not
shown)
150°W
120°W
90°W
Rayleigh channel Level 13
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
100%
75%
60°N
60°N
50%
25%
30°N
30°N
1%
0°
0°
0.9
30°S
30°S
0.8
0.7
60°S

1 km: 66% of Mie
have accuracy < 1
m/s (aerosol &
cloud returns)
60°S
0.6
Adequate
transmission
through overlying
cloud
July 2008
0.5
ECMWF Analysis VT:Friday 7 February 2003 00UTC Surface: **low cloud cover
Mie channel Level 18
150°W
120°W
90°W
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
150°W
120°W
90°W
60°W
30°W
0°
30°E
60°E
90°E
120°E
150°E
100%
75%
60°N
60°N
50%
25%
30°N
30°N
1%
0°

1.0
0°
1.0
0.9
30°S
30°S
0.8
0.7
60°S
60°S
0.6
0.5
150°W
120°W
90°W
60°W
30°W
0°
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
30°E
60°E
Slide 95
90°E
120°E
150°E
ADM-Aeolus data simulations - comparison with radiosondes/mission spec

Aeolus median like obs error assigned operationally to radiosondes

Aeolus HLOS observations expected to receive appreciable weight
Rayleigh
Mie
Without representativity
(cf mission spec – dash-dot)
July 2008
With cross-track representativity
(cf radiosonde – dashed)
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 96
ADM-Aeolus data simulations – Effects of model cloud cover (2)

Mid-latitude example

QC implications, Task 2

Tails of Rayleigh error
distributions
underestimated, median
barely changed
Model
Observed
cloud
cover
(midlats)
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 97
Assimilation of prototype ADM-Aeolus data
Quality Control for Aeolus data

Most QC parameters taken from conventional wind obs


Background errors & quality control thresholds (BgQC+VarQC)
Aeolus-specific Background Quality Control (recommended option)

Capping of observation error in bg departure classification
Set B = (obs-bg) / ES(obs-bg), accept obs iff abs(B) < 4.
In standard BgQC for Aeolus, ES = (σo2 + σb2)1/2.
Aeolus option: ES = (so2 + σb2)1/2, where so = min(σo, 2.5 ms-1)

Testing with LITE period, LIPAS-simulated Level-2B data

Gaussian + non-Gaussian errors (instrument bias, input wind bias)

Operational model (Cy26r1) at full/reduced resolution, ERA40/NoSSMI
July 2008
Data Processing for ADM-Aeolus – LWG Wintergreen & JCSDA
Slide 98