Robin Hogan, Julien Delanoe and Nicola Pounder

1
Towards unified retrievals of clouds,
precipitation and aerosols

• Robin Hogan, Julien Delanoe and Nicola Pounder
• University of Reading

2
Introduction

• Most exciting aspect to EarthCARE is synergy by
  design
• A well formulated synergistic algorithm ought to
  always outperform a single-instrument algorithm
• If different species present in the same profile
  then need to retrieve them simultaneously in
  order to interpret measurements that are
  simultaneously sensitive to both (e.g.
  path-integrated attenuation and radiances
  sensitive to whole column)
• In the RATEC project we will begin development of
  a unified retrieval algorithm for clouds,
  precipitation and aerosols
• A variational formulation will weight information
  from all sources (radar, lidar, radiances and
  prior information) according to its error
• This could also serve as 1- and 2-instrument
  algorithms (to insure against instrument
  degradation or failure) by simply removing
  certain observations
• This talk will present the ingredients that have
  been gathered so far...

3
Motivation and classification
Global-mean cloud fraction
Cloudsat radar
CALIPSO lidar

• Radar and lidar
• Radar only
• Lidar only

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
4
Retrieval framework

• Ingredients developed
• Not yet developed

5
State variables ice clouds

• Ice clouds already done by Delanoe and Hogan
  (2008), extended in CASPER to use HSRL lidar
• Variational version of Donovan and Tinel
  radar-lidar algorithms
• Blends seamlessly between regions of cloud
  detected by radar and lidar
• State vector contains these elements to describe
  ice clouds
• Visible extinction coefficient at each gate, a
• Normalized number concentration parameter, N0
• Lidar extinction-to-backscatter ratio, S
• Prior information and other constraints
• Temperature dependence of N0(T) from aircraft
  in-situ data
• Smoothness constraint on the state variables so
  that noisy observations (particularly lidar Mie
  and Rayleigh channels) dont result in noisy
  retrievals
• Prior estimate of S (e.g. 20 sr)
• Microphysical model assumptions, e.g. mass-size
  relationship, infrared scattering properties

6
State variables liquid clouds

• Largely new, but will build on
• Smith Illingworth estimate of LWP from
  path-integrated attenuation
• CloudSat radar MODIS solar channels
• Information from HSRL using multiple-scattering
  forward model
• Possible state variables for liquid clouds
• Liquid water content, LWC (or possibly a) at each
  gate
• Droplet number concentration, constant in each
  contiguous layer (via size information from MSI
  channels, and combination of LWP from
  path-integrated attenuation and optical depth
  from MSI)
• Prior information and other constraints
• Smoothness constraint on profile of LWC
• Prior estimate of number concentration (e.g. from
  sea versus land)
• Assume lidar extinction-to-backscatter ratio is
  constant at 18.5 sr
• LWC gradient at cloud base tends to the known
  adiabatic profile given the temperature and
  pressure

7
State variables precipitation

• New would need to build on results of other
  ESA/JAXA studies
• Key ingredients would be radar multiple-scattering
  model, surface return from ocean, profile of
  attenuated reflectivity (e.g. CloudSat), and
  Doppler velocity in stratiform conditions
• Possible state variables for precipitation
• Rain rate profile, R
• Normalized number concentration, Nw (one value
  per profile)
• Riming factor for snow and for ice above rain
  (one value per profile) invoked in convective
  conditions to account for higher density ice, and
  also in snow (treated as an extension to the
  ice-cloud retrieval)
• Melting-layer thickness scaling factor...
• Prior information and other constraints
• Strong smoothness constraint on profile of rain
  rate
• Estimate of Nw dependent on warm rain (e.g. Sc
  drizzle) or cold rain
• Warning this will be difficult!

8
State variables aerosols

• New would need to build on results of other
  ESA/JAXA studies
• Key ingredients would be HSRL, MSI solar channels
  in the day and optical depth constraint from
  lidar ocean surface return
• Relatively straightforward compared to
  precipitation!
• Possible state variables for aerosols
• Extinction coefficient at 355 nm
• Exinction-to-backscatter ratio (one value per
  layer)
• Mean particle size (one value per layer)?
• Prior information and other constraints
• Extinction-to-backscatter ratio estimate
  dependent on geographical region

9
Forward models active instruments

• Radar
• Microphysics scattering library for cloud
  liquid, ice and precipitation particles, ideally
  based on DDA and T-matrix rather than Mie
• Propagation fast multiple-scattering model is
  available (Hogan and Battaglia 2008) but needs an
  analytic Jacobian model
• Doppler terminal fallspeeds straightforward
  main challenge is to characterize error due to
  vertical wind and non-uniform beam filling
• Surface return requires first pass to
  interpolate between clear skies?
• Lidar
• Microphysics backscatter problem overcome by
  retrieving extinction-to-backscatter ratio, but
  some uncertainty between phase functions
• Propagation fast multiple-scattering forward
  model exists for ice clouds, where we are in the
  small-angle limit, but wide-angle model for
  liquid clouds currently lacks an analytic
  Jacobian model or the ability to represent the
  individual HSRL channels
• Depolarization currently no forward model for
  either single-scatter depolarization, or
  depolarization due to multiple scattering

10
Forward models passive instruments

• Infrared radiances
• Microphysics scattering library for cloud
  liquid, ice and aerosols required
• Propagation two models suitable for use RTTOV
  (used by ECMWF and Met Office data assimilation
  systems) and the Delanoe and Hogan (2008) scheme
• Model inputs note that the error in this model
  is significantly determined by the error in the
  temperature profile
• Solar radiances
• Microphysics scattering library required for
  liquid, ice and aerosols, with uncertainty in the
  asymmetry factor and single-scatter albedo
• Propagation fast Radiant code from Colorado
  State University could be implemented
• Model inputs Need to assume a surface albedo
• Other uncertainties three-dimensional scattering
  effects could be important but very difficult to
  incorporate in a 1D retrieval

11
Examples of wide-angle multiple scattering

• LITE lidar (lltr, footprint1 km)
• CloudSat radar (lgtr)

12
Fast multiple scattering forward model
Hogan and Battaglia (J. Atmos. Sci. 2008)

• New method uses the time-dependent two-stream
  approximation
• Agrees with Monte Carlo but 107 times faster (3
  ms)
• Added to CloudSat simulator

CloudSat-like example
CALIPSO-like example
13
Exploiting multiple scattering
14
Results

• 1D-Var retrievals using Hogan and Battaglia
  forward model (Nicola Pounder)

First 3 optical depths would be seen by HSRL
15
Test dataset ER-2 radars and lidar
10-GHz radar

94-GHz radar

• Can perform 94-GHz radar precipitation retrievals
  (using surface return from the oceans), then
  evaluate them by forward modelling the less
  attenuated 10-GHz radar

16
Next steps

• Within RATEC
• Code up flexible retrieval framework and error
  reporting
• Add various forward models
• Implement ice and liquid cloud capability
• Test on A-train and aircraft datasets
• Provide product description for 3D scene
  construction
• Post RATEC
• Test in ECSIM
• Via collaboration, implement precipitation and
  aerosol components
• Test when in 1- and 2-instrument configurations
  in case of instrument degradation or failure