C. Mitrescu1, J. Haynes2, T. Ellis2, T. LEcuyer2, - PowerPoint PPT Presentation

Title: C. Mitrescu1, J. Haynes2, T. Ellis2, T. LEcuyer2,


1
Vertical profiling of cloud structure and light
precipitation using CloudSats 94 GHz Radar

• C. Mitrescu1, J. Haynes2, T. Ellis2, T.
  LEcuyer2,
• S. Miller3, J. Turk1

1Naval Research Laboratory 2Colorado State
University 3Cooperative Institute for Research in
the Atmosphere
89TH AMS Annual Meeting, Phoenix, 2009
2
Outline

• The CloudSat Mission
• CloudSat Profiling
• A Conceptual Model (i.e. the forward model)
• Cloud Microphysical Model
• 94 GHz Reflectivity Model
• Path Integrated Attenuation (PIA) Model
• Surface so Model
• Multiple Scatter Model
• Melting Layer Model
• The Inverse Model
• Validating/Comparing the results
• Summary and Conclusions

3
The CloudSat Mission
Primary Objective To provide, from space, the
first global survey of cloud profiles and cloud
physical properties, with seasonal and
geographical variations needed to evaluate the
way clouds are parameterized in global models,
thereby contributing to weather predictions,
climate and the cloud-climate feedback problem.
Characteristics

• Nadir pointing, 94 GHz radar (3 mm)
• 3.3?s pulse ? 500 m vertical res.
• 1.4 x 2.5 km horizontal res.
• Sensitivity -28 dBZ
• Dynamic Range 80 dB
• Antenna Diameter 1.85 m
• Mass 250 kg
• Power 322 W

4
The A-Train Constellation
CALIPSO 13115
Aqua 130
PARASOL 133
OCO 115
AURA 138

• A-Train constellation afternoon ascending orbits
  ( 130 pm)
• Sun-synchronous orbit (705 km, 98.2º inclination,
  16-day repeat).
• Nearly simultaneous views of the Earth from a
  multitude of sensors (both passive and active)
  covering a wide-range of spectral bands.
• Of particular relevance to CloudSat are the
  MODIS, AMSR, and CERES (on Aqua), the lidar
  aboard CALIPSO, and the POLDER instrument aboard
  PARASOL

5
Cloud Profiling a Conceptual Model
Ideal World
Real World
Zm
Zs
ND
BB
CloudSat Measured Radar Profile

• Microphysical Model
• description of cloud particles/hydrometeors using
  DSDs, as function of height temperature

• Radar Model
• Z of µPhys
• phase/shape
• Bright Band
• Multiple Scat.

True Scene
Cost Function Evaluation
Feed-back
Forward Model
Solution
Inversion Algorithm/Model (optimal estimation)
x G(Z,be)
Z(x,b) ZMIE/DDA PIA MS e
6
Cloud Microphysical Model

• Clouds a collection of ice crystals and/or
  water drops of various sizes
• It is described in terms of a drop size
  distribution (DSD).
• Using mass-diameter and fall velocity-diameter
  power laws we can write
  
• n(D) N0
  exp(- ?D)
• ? a R-b
• N0, a, b depend on cloud phase (ice or water) and
  are assumed fixed.
• R is the rainfall rate (as the unknown variable).

• Phase assignment using model data
• temperature profile (from numerical model runs,
  i.e. ECMWF/NOGAPS) determines the cloud phase at
  each level.
• T lt -20 C only ice phase
• -20 C lt T lt 0 C, both ice and water phase
• T 0 C ( BB) has water and melted ice.
• T gt 0 C, only water phase

ICE
- 20o C
ICE WATER
0o C
WATER
7
94 GHz Reflectivity Model

• Although attenuated by atmosphere, the 94 GHz
  frequency is well suited to profiling clouds and
  light precipitation.
• Mie based (DDA in the near future) efficiency
  calculations (Qx) are used to determine
  scattering/absorption/backscatter coefficients
  (X) as function of temperature, and rainfall rate
  (via the effective diameter) for cloud (ice
  and/or water) and atmospheric model (O2 , N2 ,
  H2O).
• X p/4 ? n(D) QX D2 dD

• To speed-up retrieval calculations, Look-up
  Tables (LUT) are being used.
• Surface return, multiple scatter radiation and
  melting layer properties are explained below.

8
Path Integrated Attenuation (PIA) Model

• Atmospheric attenuation is computed using MPM89
  model (Liebe, 1989), for all scenes clear or
  cloudy. T and RH are provided by numerical model
  output (NOGAPS or ECMWF).
• Path integrated attenuation (PIA) due to cloud
  particles is computed using the extinction
  coefficient for the ice/water combination. When
  the signal is completely attenuated, the
  assumption of a uniform-to-ground layer is
  employed.

PIA 8.686 ? s(z) dz
Melting Layer (bright band) Model

• Maxwell-Garnett dielectric mixing formula
  (Menenghini and Liao, 1996).
• The melting layer depth is fixed at 3 radar
  gates (3 x 240 m) thus assuming that all ice
  crystals are melt below it.
• A combination of water drops and partially melted
  ice crystals is assumed.
• CloudSat and ECMWF/NOGAPS profiles show that the
  bright band position is very well correlated with
  the 0C isotherm (although, inversions and/or
  deep convection may pose some problems).

9
Surface so Model

• Surface reflection (so) is very strong for most
  of the surface types. However, soil moisture,
  canopy cover, or surface roughness can strongly
  influence the return signal.

IGBP-based so histograms

• The International Geosphere-Biosphere Programme
  (IGBP) surface type is used to construct a
  clear-sky so data base. For cloudy-sky, the
  difference in so due to cloud attenuation (PIA),
  that is used as a constraint in the retrieval
  algorithm.

IGBP surface type
10
Multiple Scatter (MS) Model

• MS arises from CPRs large FOV ?z combined with
  a non-negligible single scatter albedo (Mitrescu,
  2005)
• preliminary simulations show that MS contribution
  reaches 2 of the signal for areas where Z10
  dBZ, and lags over 3-4 gates it builds up
  rapidly.
• due to a quasi-isotropic phase function, all
  scattering directions contribute to the MS
  signal.
• second order scatter easy to evaluate (near
  real-time) a 10-stream model.
• Antenna gain is accounted for.
• Computationally very intensive (_at_ 40 m resolution
  gt 6x6 increase in resolution)
• third and above scatter require intensive
  calculations, thus cannot be evaluated
  operationally.
• MS contribution hitting the ground must be
  evaluated.

backscatter vs. multiple scatter
11
The Inverse Model

• Optimal estimation technique find the solution
  to the problem by minimizing the Cost Function
  (LEcuyer and Stephens, 2002)

J S (Zobs Zmod)T Sz-1 (Zobs Zmod)

• although slow when compared to analytical
  methods, it has the flexibility advantage
• it uses the forward model itself (ideal for most
  non-linear problems) no need to construct an
  inverse model
• ease in adding measurements and constraints, or
  changing the forward model
• can be combined with analytical methods.
• it has built-in error estimates however,
  computing all parameter errors can be
  computationally very demanding.

12
Validating/Comparing the Results

• TRMM (whenever it intersects CloudSat)
• Ground Radar Networks (NexRad)
• AMSR-E precipitation estimates (as part of the
  A-Train)
• Blended multi-sensor methods
• In-situ (field campaigns) measurements
• Numerical Model Data Output

13
Comparison Example 1 31 July 2006 1919 UTC

• NEXRAD S-band radar located in New Orleans (KLIX)
  using
• R 0.017 Z0.714 ( )
• Matrosov (2007), applies a k-R ( )
  retrieval.
• Haynes et al (2008) use CloudSat derived PIA
  alone to infer an uniform cloud layer rain rate
  ( ).
• AMSR-E precipitation data along CloudSat track (
  )
• The profiling algorithm also provides vertical
  structures of LWC/IWC (and Do) regardless of the
  BB (Matrosov) and/or surface (Haynes et al) (
  ).

ocean
land
14
Comparison Example 2 07 Sept 07 2006 1840 UTC
NEXRAD KCLX (Charleston, SC)
GOES 12
15
Comparison against in-situ data

• The frequency of precipitation occurrence over
  the global oceans for 20062007 period using
  CloudSat is compared to ship-based (ICOADS) and
  island-based (GSOD) data.
• The comparison to ship-based data reveal that
  CloudSat results are consistent with ship
  observations well into the high latitudes and
  appear to capture the seasonal cycle of
  precipitation well.
• The comparison to island data also shows good
  qualitative agreement, although the spatial scale
  mismatch complicates the efficacy of such
  comparisons.
• CloudSat is shown to be a viable platform for
  obtaining quality satellite-based precipitation
  frequency measurements, but this should still be
  regarded as a work in progress (Ellis et al, to
  be submitted).

16
CloudSat 2007 light precipitation rates
mean values
standard deviations
17
CloudSat 2007 event frequency
light precipitation
heavy precipitation
18
AMSRE 2007 precipitation rates frequency
mean values
events
19
Summary and Conclusions

• Part of the A-Train, CloudSats 94 GHz radar is
  aimed at measuring clouds and light precipitation
  on a global scale.
• Present work introduces the basic formulation for
  the CloudSat precipitation profiling algorithm.
  Although still a work in progress, preliminary
  results are encouraging. Here are some thoughts
  about it
• Strengths
• Retrieval framework, flexible enough to
  accommodate any changes in the forward model
  provides a suite of Q.C. and error diagnostics.
• CloudSats CPR offers higher spatial resolution
  than other sensors that directly measure
  precipitation.
• Sensitivity to both clouds and rainfall
  facilitates studying the microphysical processes
  at work.
• Weaknesses
• Strong attenuation at 94 GHz limits retrievals to
  light precipitation.
• Single-frequency method limits information
  regarding the dielectric properties of the
  melting layer and restricts DSD assumptions.
• CPR is nadir-pointing providing only a 2D slice
  of the real world.
• Future work plans
• DDA calculations for ice crystals,
• add a separate DSD function describing cloud
  particles that dont precipitate,
• include MS effects (probably as
  parameterization),
• continue validation/comparison work