VARIATIONAL SYNERGISTIC ICE-CLOUD RETRIEVALS

1
VARIATIONAL SYNERGISTIC ICE-CLOUD RETRIEVALS
1st JADE Meeting- ESTEC 2009

• Julien Delanoë, Robin Hogan
• Thorwald Stein
• University of Reading, UK

First JADE Meeting, 22-23 April 2009 ESTEC
2
Radar-Lidar-IRradiometers
1st JADE Meeting- ESTEC 2009
Now
EARTHCARE
A-TRAIN
CPR (Cloud Profiling Radar) ATLID (ATmospheric
LIDar) MSI (Multi-Spectral Imager) BBR (Broad
Band Radiometer)
CloudSat CPR 94GHz CALIPSO Lidar (532, 1064nm)
IIR AQUA radiometers MODIS, AIRS, CERES,
AMSR-E

• Ice cloud retrievals using radar-lidar-radiometer
  synergy
• An opportunity to tackle questions concerning
  role of clouds in climate
• Need to combine all these observations to get an
  optimum estimate of global cloud properties

3
Radar-Lidar-Radiometer
1st JADE Meeting- ESTEC 2009
Radarlidargt Vertical description of clouds
July 2006
Global-mean cloud fraction
CALIPSO lidar
CloudSat radar
Radar and lidar Radar only Lidar only

• Radar Z?D6, lidar b?D2 combination provides
  particle size
• Lidar sensitive to particle concentration, but
  extinguished when cloud to thick
• Radar sensitive to particle size, cant detect
  small crystals

• Radiances ensure that the retrieved profiles can
  be used for radiative transfer studies
• Single channel information on extinction
  near cloud top
• Pair of channels ice particle size
  information near cloud top

4
How to combine them?
1st JADE Meeting- ESTEC 2009
Variational scheme We know the observations
(instrument measurements) and we would like to
know cloud properties visible extinction, Ice
water content, effective radius
Ingredients already developed (Delanoë and Hogan
JGR 2008-2009)
New ray of data define state vector Use
classification to specify variables describing
ice cloud at each gate extinction coefficient
and N0
Radar model
Lidar model Including HSRL channels and multiple
scattering
Radiance model IR channels
Forward model
Not converged
Compare to observations with an a-priori and
measurement errors as a constraint Check for
convergence
Gauss-Newton iteration Derive a new state vector
Converged
Proceed to next ray of data
5
1st JADE Meeting- ESTEC 2009
A-Train
EarthCare
No HSRL available S is assumed constant with
height Or can be assumed linearly varying with
height if radiance used
When HSRL available S can be relaxed!
6
What do we do with A-Train?
1st JADE Meeting- ESTEC 2009
7
Categorisation
1st JADE Meeting- ESTEC 2009
We merge radar, lidar, MODIS data Using CloudSat
and CALIPSO mask ECMWF temperature gt New
categorisation

• Temperature model (ECMWF)
• gt Ice / Liquid water
• Supercooled liquid layers
• Exploit the different response of radar and lidar
  in presence of supercooled liquid water
• -Very strong lidar signal
• -Very weak radar signal
• Within a 300m cloud layer

Cloudsat radar
CALIPSO lidar
Insects Aerosol Rain Supercooled liquid
cloud Warm liquid cloud Ice and supercooled
liquid Ice Clear No ice/rain but possibly
liquid Ground
Preliminary target classification
8
Radar-lidar example
1st JADE Meeting- ESTEC 2009
ice
water
9
Radar-lidar-radiances
1st JADE Meeting- ESTEC 2009
(radiances Sct)
(radiances Svar)
Radarlidar
lidar
Latitude
Radar
a
a
a
Latitude
N0
N0
N0
re
re
re
Latitude
Latitude
Latitude
10
1st JADE Meeting- ESTEC 2009
Frequency of occurrence of IWC vs temperature

• IWC increases with temperature
• but spread over 2 to 3 orders of magnitude at
  low temperatures
• reach 5 orders of magnitude close to 0 C

• Advantage of the algorithm
• Deep ice clouds radar
• Thin ice clouds lidar
• When radar and lidar work well together very good
  confidence in the retrievals
• Obvious complementarity
• radar-lidar

11
Comparison with other products
1st JADE Meeting- ESTEC 2009
Comparison CloudSat IWC IWC-Z-T- Variational
method
IWC vs Temperature
CloudSat
CloudSat
IWC-Z-T
Stein et al. 2009
12
Comparison with other products
1st JADE Meeting- ESTEC 2009
Optical depth MODIS vs radar-lidar variational
method As a function of latitude (2 weeks in July
2006)
20

• MODIS
• Radar-lidar

15
Optical depth
10
5
0
0
-50
50
Latitude
13
Model comparison
1st JADE Meeting- ESTEC 2009
Work in Collaboration with Alejandro
Bodas-Salcedo (Met-Office)
Forecasts Vertical profiles were extracted from
the model along the CloudSat-Calipso tracks at
the closest time to the observations. A-train
data averaged to models grid
A-Train vs UK met-Office
x10-4
x10-4
gt Models capture the trend of the IWC-T
distribution (not the rest)
14
Model comparison
1st JADE Meeting- ESTEC 2009
Work in Collaboration with Richard Forbes (ECMWF)
Forecasts Vertical profiles were extracted from
the model along the CloudSat-Calipso tracks at
the closest time to the observations A-train data
averaged to models grid
A-Train vs ECMWF
x10-4
x10-4
x10-4
gt Models capture the trend of the IWC-T
distribution (not the rest)
15
Future for A-Train
1st JADE Meeting- ESTEC 2009

• We will treat the entire period of
  CloudSat-CALIPSO, however doing this requires
  resources
• Icare (http//www-icare.univ-lille1.fr/)
• They provide various services to support the
  research community in fields related to
  atmospheric research, such as aerosols, clouds,
  radiation, water cycle, and their interactions.
• Production and distribution of remote sensing
  data derived from Earth observation missions from
  CNES, NASA, and EUMETSAT.
• They provide
• Merged files (radar-lidar-radiometers colocated,
  CloudSat track)
• Categorisation (under development)
• Retrievals (using our radar-lidar algorithm)

16
What do we do to prepare EarthCare?
1st JADE Meeting- ESTEC 2009

• CASPER project

17
Variational for EarthCare
1st JADE Meeting- ESTEC 2009
Simulate EarthCare measurements (cf Blind Test
Hogan et al. 2006) Simulated profiles
(LidarRadar) and MSI radiances are taken as truth
CPR
ATLID
S
Visible a
Effective radius
18
ECSIM
1st JADE Meeting- ESTEC 2009
Attenuated lidar backscatter (Mie channel)
Attenuated lidar backscatter (Ray channel)
Measurements
Forward modelled
Reflectivity
a profile
Visible extinction (truth)
Measurements
Retrieved visible extinction
Forward modelled
19
Future work
1st JADE Meeting- ESTEC 2009

• Future work
• More than one month Icare
• Develop algorithms for EarthCARE, HSRL lidar and
  Doppler radar
• Retrieve properties of liquid-water layers,
  drizzle and aerosol (Robin Hogans talk)
• A-train data and validate using in-situ
  underflights

20
Model comparison
1st JADE Meeting- ESTEC 2009
Work in Collaboration with Richard Forbes
(ECMWF) and Alejandro Bodas-Salcedo (Met-Office)
Forecasts Vertical profiles were extracted from
the model along the CloudSat-Calipso tracks at
the closest time to the observations A-train data
averaged to models grid
A-Train vs ECMWF
A-Train vs UK met-Office
x10-4
x10-4
x10-4
x10-4
x10-4
gt Models capture the trend of the IWC-T
distribution (not the rest)
IWC cut-off _at_ 10-4 kg m-3 between 0C -20C
explained by the parameterization of the
ice-to-snow autoconversion rate. When ice water
contents reach 10-4 kg m-3 gt precipitating snow.