Use of AMSR-E Land Parameter Modeling and Retrievals for SMAP Algorithm Development

1
Use of AMSR-E Land Parameter Modeling and
Retrievals for SMAP Algorithm Development
Steven Chan Eni Njoku
Joint AMSR Science Team Meeting Telluride,
Colorado July 14-16, 2008
2
The SMAP Algorithm Development Testbed
Purposes

• An end-to-end Observing System Simulation
  Experiment (OSSE) for SMAP

• An L-band mission simulator (simulated TBs to be
  as realistic as possible)

• Modules capable of testing multiple algorithms
  subject to the same inputs

Features

• Realistic orbital and instrument sampling

• Global and continental scales

• Annual cycle(s)

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
3
The SMAP Algorithm Development Testbed
Anticipated Benefits

• Investigate relative strengths/weaknesses of
  different retrieval algorithms

• Impacts of pursuing different science/instrument/m
  ission trades

• Impacts of ancillary data uncertainties

• Realistic orbital and instrument sampling

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
4
Major Testbed Components
Excerpted from SMAP Algorithm Development Testbed
for MCR in Jun 2008

• Orbital and instrument sampling

• Land surface model (LSM) input

• Forward microwave models (i.e. radiometer and
  radar)

• Environmental effects (e.g. Faraday rotation,
  galactic radiation)

• Instrument effects (e.g. instrument precision,
  calibration errors)

• Inverse models (e.g. radiometer, radar, and
  combined radar-radiometer)

• Error analysis

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
5
Testbed Flowchart
LSM Input Parameters
Environmental Effects
Instrument Effects
Surface Temp
Soil Texture
Orbital/ Instrument Sampling
Forward Models (radiometer/radar)
TB,s
Vegetation
? ? ?
Inverse Models (radiometer/radar)
Truth Soil Moisture
Retrieved Soil Moisture
Error Analysis
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
6
Orbital and Instrument Sampling
SMAP Ground Tracks
Altitude 670km Sampling Period 42ms
Antenna 6m Incidence 40
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
7
Orbital and Instrument Sampling (Radiometer)
SMAP Instrument Sampling (boresight and antenna
beam pattern)
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
8
Major Testbed Components
Excerpted from SMAP Algorithm Development Testbed
for MCR in Jun 2008

• Orbital and instrument sampling

• Land surface model (LSM) input

• Forward microwave models (i.e. radiometer and
  radar)

• Environmental effects (e.g. Faraday rotation,
  galactic radiation)

• Instrument effects (e.g. instrument precision,
  calibration errors)

• Inverse models (e.g. radiometer, radar, and
  combined radar-radiometer)

• Error analysis

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
9
Land Surface Model Input Parameters
Current scheme Use geophysical data fields from
GLDAS
Soil Moisture
Vegetation Water Content
Surface Temperature
Soil Temperature
Sand Fraction
Clay Fraction
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
10
Advantages

• Convenience one-stop portal of geophysical input
  data fields

• Fields are consistent with the underlying land
  model that generates them

Potential Limitations

• Antenna sampling not well addressed cell size
  (above) gtgt SMAP footprint

• Simulated TBs may have QC issues due to
  unrealistic inputs

• Inputs may introduce unreal spatial and temporal
  correlations

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
11
Major Testbed Components
Excerpted from SMAP Algorithm Development Testbed
for MCR in Jun 2008

• Orbital and instrument sampling

• Land surface model (LSM) input

• Forward microwave models (i.e. radiometer and
  radar)

• Environmental effects (e.g. Faraday rotation,
  galactic radiation)

• Instrument effects (e.g. instrument precision,
  calibration errors)

• Inverse models (e.g. radiometer, radar, and
  combined radar-radiometer)

• Error analysis

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
12
First Step Towards More Realistic Inputs/Outputs
Proposed scheme Use AMSR-E data fields ( )
to first optimize forward model and then extend
its frequency dependence to L-band frequency
AMSR-E Soil Moisture
Forward Model at 6.9 GHz
Forward Model at 10.7 GHz
Forward Model at 18.7 GHz
Simulated TBs at 6.9 GHz
Simulated TBs at 10.7 GHz
Simulated TBs at 18.7 GHz
AMSR-E TBs at 6.9 GHz
AMSR-E TBs at 10.7 GHz
AMSR-E TBs at 18.7 GHz
Optimal Model Parameters at 6.9 GHz
Optimal Model Parameters at 10.7 GHz
Optimal Model Parameters at 18.7 GHz
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
13
First Step Towards More Realistic Inputs/Outputs
Radiative Transfer Forward Model Ts surface
temperature, rs Fresnel reflectivities, t
vegetation opacity ? single-scattering albedo,
h surface roughness, Q polarization mixing ratio

• Use AMSR-E soil moisture and AMSR-E TBs to
  determine optimal model parameters at a given
  frequency

Model Parameter (e.g. ?)

• Repeat the above procedure for other frequencies

10.7
18.7
6.9
1.4

• Explore frequency dependence of optimal model
  parameters and extend (extrapolate) their values
  to L-band frequency

20
15
10
5
0
Frequency (GHz)
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
14
Simulated L-band TB (1 Day)
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
15
Simulated L-band TB (2 Days)
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
16
Simulated L-band TB (3 Days)
Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008
17
Summary

• SMAP Testbed as an end-to-end mission simulator

• Simulated L-band observations to be as realistic
  as possible

• AMSR-E data fields could be used to optimize
  forward model and extend its frequency dependence
  to L-band frequency, thus assisting the SMAP
  Testbed to generate realistic L-band observations

Steven K. Chan ? Joint AMSR Science Team Meeting
? Telluride, Colorado ? Jul 14-16, 2008