METimage Cloud Top pressure retrieval from oxygen A-band
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
METimage Cloud Top pressure
retrieval from oxygen A-band
Mathieu Compiègne (HYGEOS, Lille, France, mc@hygeos.com),
Laurent Labonnote (LOA, Lille, France),
Nicolas Ferlay (LOA, Lille, France),
Jérôme Riedi (ICARE/LOA, Lille, France),
Phillipe Dubuisson (LOA, Lille, France),
Didier Ramon (HYGEOS, Lille, France)
ESA LPS2019, Milano
1
METimage (aka VII)
●
METimage is a scanning optical imaging
radiometer supporting operational
meteorology and climate study
– Recording a gapless image from Polar orbit
with large swath 2800 km (global coverage
daily).
– Multi-spectral radiometry in 20 channels
(VIS/NIR/SWIR/IR).
– Ground spatial sampling of 500 m at nadir.
Courtesy : Loredana Spezzi, EUMETSAT
●
It will fly onboard EUMETSAT Polar
System Second Generation (Q4 2022
launch)
2
METimage
EUMETSAT Retrieval
L2 retrieval @ EUMETSAT
Geophysical product as EUMETSAT Parameter to
product ID per user requirement product ID be validated
VII-02-CLD Threshold based decision tree including 18 tests
based on distinct cloud features and measurements Cloud detection /Cloud VII-02-CLD Clear/cloudy
across the full VII spectral range. mask (CLD) flag
VII-02-OCA Optimal Cloud Analysis exploiting the full spectral Cloud top VII-02-OCA CTP and CTT
range and simultaneously constraining cloud phase, temperature/pressure VII-02-CTP
COT, CTP, particle effective radius, and related (CTT,CTP,CTH)
uncertainties.
VII-02-CTP 1Dvar approach exploiting measurements Cloud optical thickness VII-02-OCA COT (optional)
(METimCTP) around the O2 A-band and simultaneously (by-product) (COT) VII-02-CTP
constraining CTP, COT and related Cloud particle effective VII-02-OCA CRE (optional)
uncertainties. radius at cloud top (by-
VII-02-WVV 1Dvar approach using 0.9µm water vapor absorption product) (CRE)
band and simultaneously constraining TPW, surface Cloud liquid/ice water VII-02-OCA LWP and IWP
reflectance and related uncertainties. path (by-product) (optional)
(LWP/IWP)
VII-02-WVI 1Dvar approach exploiting thermal IR (in particular, Volcanic ash VII-02-OCA Ash flag
the 6.7 and 7.3µm water vapor absorption bands) Water-vapour total VII-02-WVV TPW
and constraining TPW, surface/atmospheric column (TPW) VII-02-WVI
temperature and related uncertainties.
Polar atmospheric motion VII-02-AMV AMV direction,
VII-02-AMV Cloud/water vapor feature tracking algorithm vectors (AMV) speed, pressure
between successive images. Height assignment from and
OCA CTP. temperature
Courtesy : Loredana Spezzi, EUMETSAT 3
METimCTP overview
●
Optimal estimate framework with Levenberg-Marquardt iterative
scheme
– State vector x=[ log10(COT), CTP ]
– The measurement vector is y=[ I , I763/I752 ]
● I=I670 over land (darker vegetation)
● I=I865 over ocean (darker water, lower aerosol impact)
●
Forward model F(x,b)
– Non retrieved parameters b= (phase, r eff, cloud vertical profile, surface
BRDF, ground level pressure, geometry)
– Only accounts for mono-layer situation
– Pre-computed Look-Up tables due to NRT constraint ( METimCTP retrieval
time / pixel ~ 0.1 ms)
– Computed with ARTDECO (http://www.icare.univ-lille1.fr/projects/artdeco)
4
METimCTP physical basis
● I763/I752 signal is driven by the mean photon path
●
In a perfect world (dotted line) only Cloud Top
Pressure (CTP) matters
●
But due to photon penetration, photon path is
also affected by:
– cloud optical thickness (COT)
– cloud vertical structure
– droplet or ice cristal size
– ice cristal shape
– surface…
●
These parameters must be constrained to
properly model the signal in F(x,b)
5
METimCTP
Cloud vertical structure
●
The vertical structure of the cloud
is a crucial parameter for O2 A-
band signal modelling
– COT→ ~number of scattering
CGT COT, CTP – CGT / extinction profile→~length
COT, CTP
between scattering
CGT
●
METimage characteristics
(spectral and geometry sampling)
does not offer enough information
content to constrain both the CTP
and vertical structure 6
Cloud vertical profile
climatology
●
Cloudsat climatology (year 2010) extinction profiles for 9
ISCCP cloud types:
– produced and used by Carbajal-Henken et al. (2013) for CTP
O2 A-band retrieval with MERIS
●
Geometrical Thickness climatology (year 2008) built from
CloudSat/Caliop data
– ice over land
– ice over ocean
– liquid over land
– liquid over ocean
●
The vertical structure is
implicitly varied in the forward
model as a function of the
state vector x=[ log10(COT),
CTP ] 7
Caltrack dataset
●
Calxtrack extracts products (L1 and L2) from different sensors (CALIOP,
IIR, MODIS, PARASOL, CLOUDSAT...) in coincidence with CALIOP
measurements
●
It is developped at ICARE data center at
Lille University
●
We apply METimCTP to Parasol/Polder
measurements
– 670, 763, 765, 865 nm channels
– We use a single view angle to be in
METimage conditions
●
Validate COT versus MODIS
●
Validate CTP versus Cloudsat/Caliop 8
Retrieval on 2008 full year ● Number of successfully retrieved (Φ/ny
Cloud Top Height result
Mono-layers
10Error on CTOP versus error on CGT
Mono-layers
●
Very strong correlation between the error
on CTOP and error on CGT
– Deviation from this correlation is related to
other sources of error (extinction profile,
effective radius, ice crystal shape, surface
properties... )
– But we clearly see that the vertical structure
constraints is critical for a good CTOP retrieval
●
To use METimage thermal IR channels
should bring constraints on vertical
structure
11Cloud Top Height result
Mono-layers
●
For a given altitude class, the retrieval
is globally better for thicker clouds
– best cases is Stratus and Nimbostratus
– exception for DCC
●
For a given COT class the bias goes
from overestimating the CTOP altitude
for lower clouds to underestimating it
for high cloud (except for DCC!)
●
The standard deviation increases for
higher clouds
12Cloud Top Height result
Multi-layer situations
●
In multi-layer
situation, our retrieval
does not rely to any of
the layers
●
Nothing (cost, number
of iteration, ...) allows
to diagnose the
presence of multi-
layer in this version of
METimCTP
13Cloud Top Height result
Multi-layer situations
●
In multi-layer
situation, our retrieval
does not rely to any of
the layers
●
Nothing (cost, number
of iteration, ...) allows
to diagnose the
presence of multi-
layer in this version of
METimCTP
14Physical basis
for multi-layer detection using WV channels
●
If gas absorption, the TOA signal
results of absorption/scattering
coupling and is sensitive to the
vertical distribution of absorbers
and scatterers
O2, well mixed
● WV vertical distribution ≠ O2
vertical distribution → H2O, altitude
supplementary information content dependent
on cloud vertical distribution
15SWIR WV channel use
in METimCTP
●
Step I : we run METimCTP with the measurment vector
y=[I,I763/I752] to get the retrieved state vector
●
Step II : we recompute
the cost function using the retrieved state and including the WV
channel in the measurement vector as y=[I, I 763/I752, I914/I865] or
y=[I, I763/I752, I1375/I1240]
– 914 nm shows moderate WV absorption
– 1375 nm shows strong WV absorption
●
Challenging to properly account/model the WV profile
16WV cost in
Multi-layer situations
●
Apply on Caltrack data using
Parasol 914 nm channel and
MODIS 1375 nm channel
●
The WV cost function increases
– in ML situation with 910 nm
channel but even more in the
lower cloud situation
– clearly in ML situation for 1375nm
channel use
●
Early conclusion (from first
caltrack analysis and theoretical/
modelling consideration) is that
1375 nm could be more fruitful
for ML detection
17On going
●
Consolidate the analysis with Caltrack data retrieval
– Refine error budget and better characterize METimCTP retrieval characteristics
– Derive a recipe for multi-layer flag from WV cost
●
Update (seasonal and zonal) vertical structure climatology using
DARDAR product
– Improved results for mono-layer situation
●
Longer term: Use the full METimage spectral range (i.e. merge MOCA
and METimCTP) to increase the information content on cloud vertical
structure thanks to thermal infrared
– improve results for mono-layer situation
– better handling of multi-layer situations
18Backup slides
19WV channel modelling
●
In F(x,b) for WV channels, only WVC is
varied Normalized to 1g/cm2
●
To vary the WV profile is forbidden in
LUT approach due to LUT size problem
●
We compute a variance related to V
profile diversity that will be used to
compute the cost function :
● Sσwvprofwvprof= f(WVC, COT, CTP, SZA, VZA,
RAA) is stored as a LUT
20You can also read
