NDDA - Hydrometeorology Soil Moisture / Cloud Assimilation Experiments

1
NDDA - HydrometeorologySoil Moisture / Cloud
Assimilation Experiments

• Dr. Andrew S. Jones
• Dr. Tomislava Vukicevic

2
Current Status

• The passive microwave observational operator
  (including the adjoint) is complete
• Microwave Land Surface Model (MWLSM)
• Based on 6 and 10 GHz passive microwave data
• After Njoku (1999) (AMSR land algorithms)
• Applicable to a new generation of passive
  microwave imagers
• AQUAs AMSR-E(launched May 4, 2002 data gt Feb.
  2003)
• ADEOS-IIs AMSR (launched Dec. 14, 2002)
• WindSat (launch gt Jan. 6, 2003)
• NPOESS C1 CMIS ( 2009)

3
Current Status (continued)

• The MWLSM observational operator is the link that
  connects the microwave remote sensing land
  surface physics to the atmospheric/land surface
  prognostic model during the data assimilation
  minimization process
• A much simpler IR land surface observational
  operator has also been constructed
• Related sensitivity studies are underway using
  the recently completed RAMDAS (the CSU/CIRA 4D
  data assimilation system)
• WRF data interfaces in RAMDAS are used to bring
  in conventional data
• Several experiments are in progress

4
Microwave Land Surface Model (MWLSM)Observationa
l Operator Components
5
What We Learned

• Satellite observational operator sensitivities
  can be a strong function of their base states
• This work creates an improved analysis of the
  multivariate physical interactions
• It required derivation of the adjoint in complex
  number space (publication is in progress)
• Complex numbers are not handled by current
  automated adjoint compiler technologies
• Has practical implications for all future
  satellite observational operators involving
  radiative scattering processes
• Cross-sensor data sets should improve results in
  particularly difficult base state
  environments(i.e., sensitivity transitions
  and/or sensitivity inflection points)

6
4DDA Soil Moisture / Cloudcase study (May 2,
1996)
GOES-9 Visible (Satellite Projection)
Time-dependent IR data (after Jones et al.,
1998a,b) Future MW data IR / MW data IR / MW /
VIS data
Simplest method
Soil Moisture with minimal cloud effects
For Improved Clouds
7
4DDA case study (May 2, 1996)
GOES-9 Visible RAMDAS Projection (via DPEAS)
8
4DDA case study (May 2, 1996)
GOES-9 Visible RAMDAS Projection 25 km grid for
testing purposes
9
4DDA case study (May 2, 1996)
High clouds
GOES-9 Chan 4 (IR) RAMDAS Projection 25 km
grid for testing purposes
Clear
Low clouds
10
4DDA Soil Moisture Future Work

• Finish RAMDAS / DPEAS satellite data interface
• Complete initial RAMDAS observational tests at 25
  km, then go to finer model grid
• Obtain microwave (AMSR/WindSat) data sets as they
  become available
• Verification data sets (some preliminary
  candidates on hand, however much will depend on
  the final case study selections)
• Upcoming field campaign info? e.g., SMEX03DoD
  input is desired
• Comparison to traditional soil moisture
  retrievals (AMSR-like methods)

11
(No Transcript)
12
Backup Slides

• Microwave Land Surface Model (MWLSM)
  Observational Operator

13
Land Surface Data Assimilation Process

• Passive Microwave and IR satellite data are
  complimentary surface data sources
• IR data has a unique high temporal diurnal
  temperature signature useful for surface flux
  retrievals
• MW data has a physical connection to the soil
  moisture via the dielectric constant and to key
  vegetation properties
• Together, MW and IR cross-sensor combinations can
  explore temporal data requirements, and mixed
  pixel issues for better use of satellite
  observations within the NWP context
• 14 input variables/model parameters
• 5 primary control variables for optimization
• Soil Moisture, Surface Roughness, Land Surface
  Temp., Veg. Canopy Temperature, and Veg. Water
  Content

14
Forward Model Results (bare soil)
15
Relative Response(bare soil)
16
(bare soil)
17
Analysis in higher dimensional space

• When only perturbations along the soil moisture
  base state are allowed, soil moisture is the only
  contributing variable to the cost function
  minimization
• What happens when all control variables in the
  5-dimensional space are adjusted simultaneously
  with a positive bias, x, along the red vector,
  and projected back-onto the soil moisture basis
  vector?

SM
18
Forward Model Results(vegetated soil)
19
Relative Response (vegetated soil)
We now have multiple cross-over conditions
20
Small Veg./Roughness Effects
large sensitivity to soil moisture
Large Veg./Roughness Effects
reduced sensitivity to soil moisture
no sensitivity to soil moisture
DRY
WET
21
Examples of Experiments Planned

• Experiment Sequence
• Simulation tests / verifications
• IR
• MW
• IR MW
• 2 week mostly clear case study
• 2 week heavy precip event case study
• Cycling experiments to emulate 3DVAR
• Various data denial experiments (IR without MW,
  or in situ observations, etc.)
• Theoretical simulations to clarify physical cause
  and effect (i.e., how long does the remote
  sensing data impact the 4DDA system, and through
  what predominant physical mechanisms?)

22
For more technical info, references

• http//lamar.colostate.edu/asjones/Jones/default.
  htm
• Jones_at_cira.colostate.edu