1. Use of ATOVS at DAO Joanna Joiner, Donald Frank, Arlindo da Silva,
Emily Liu, Clark Weaver
Data Assimilation Office, NASA/GSFC
ITSC-12

2. Joanna Joiner, DAO ITSC-12
Outline
Introduction: DAOTOVS 1DVAR assimilation
Assessment of cloud- and land-affected data in DAS
Use of OPTRAN
OSSE simulations
Use of TOVS for land-surface analysis/assimilation
Off-line skin temperature analysis (including bias correction)
Off-line skin temperature assimilation
AIRS dynamic channel selection based on cloud height
Incorporate effects of aerosol
Summary and Future Plans
Joanna Joiner, DAO ITSC-12

3. Joanna Joiner, DAO ITSC-12
DAOTOVS attributes
DAOTOVS 1D-Var Assimilation of Radiances:
Uses Level 1b data
HIRS, MSU, SSU and AMSU-A radiances
Variational cloud-clearing (Joiner and Rokke, 2000)
Eigenvector FOV determination (AIRS ATBD)
Physically-based systematic error correction
GLATOVS, MIT -> OPTRAN (JCSDA)
Running in operational GEOS-DAS and next-generation Finite-volume DAS (fvDAS) currently running in parallel
Joanna Joiner, DAO ITSC-12

4. Joanna Joiner, DAO ITSC-12
fvDAS Data Flow
(PED coeff)
Joanna Joiner, DAO ITSC-12

5. DAOTOVS: What makes it different?
Uses cloud- and land-affected data (CERES land-emissivity data set based on satellite/laboratory measurements).
Uses all channels except HIRS 16, 17 AMSU 1,2 (IR bi-directional reflectance, mw emissivity in 1DVAR state vector)
Variational cloud-clearing (done simultaneously with retrieval); allows for internal quality control, consistency
Tuning using collocated radiosondes (not background). Updated daily via Kalman filter.
Errors in assimilation system include separate components with and without vertical/horizontal correlations
Joanna Joiner, DAO ITSC-12

6. How many cloud formations are seen in NOAA-K data? Answer: ~2
Look at eigenvectors of 3x3 array of HIRS pixels
R1-Rn
~95% of cases explained by two modes (cloud-formations)
Joanna Joiner, DAO ITSC-12

7. Joanna Joiner, DAO ITSC-12
O-F Statistics
Fit to Rawinsondes
Obs – 6h Forecast
Bias (spatial RMS, time mean)
Standard Deviations NW NE
Tropics SW SE
Joanna Joiner, DAO ITSC-12

8. red: DAOTOVS w/cloud-cleared, blue: DAOTOVS, no cloudy
Cloud clearing has positive impact on 6 hour forecast, verified with radiosondes in finite-volume DAS
green: NESDIS TOVS,
red: DAOTOVS w/cloud-cleared,
blue: DAOTOVS, no cloudy
Joanna Joiner, DAO ITSC-12

9. Joanna Joiner, DAO ITSC-12
Forecast experiments, RMS error 500 hPa height red: cloud-cleared, blue: no cloudy
Joanna Joiner, DAO ITSC-12

10. Impact of land-affected data (red-includes land, blue-no land)
Joanna Joiner, DAO ITSC-12

11. OPTRAN significantly reduces ATOVS radiance biases
note: a) scale b) large reduction in channel 1 and 12 biases
OPTRAN
GLATOVS
Joanna Joiner, DAO ITSC-12

12. Observing System Simulation Experiments (OSSE)
Use fvCCM/Optran to simulate cloudy radiance
Use GEOSDAS/ GLATOVS for assimilation
Model has reasonable simulation of cloud/upper tropospheric humidity (use maximum overlap assumption)
Joanna Joiner, DAO ITSC-12

13. Skin temperature biases over land
*The problem:
Skin temperature biases over land
(especially desert) causing clear-sky Outgoing Longwave Radiation (OLR) biases as compare with CERES;
*Problem caused by emissivity used in land-surface model (LSM) and inconsistent definition of ground temperature

14. Control
fvDAS
Ts Bias
ECMWF
|top|-|mid|

15. Unbiased Analysis Equation
Joanna Joiner, DAO ITSC-12

16. Ts Bias and Anal. Increments
Joanna Joiner, DAO ITSC-12

17. Joanna Joiner, DAO ITSC-12

18. New
fvDAS
Ts Bias
Control
fvDAS
Ts Bias
|top|-|mid|

19. More TOVS marked “clear” by internal 1DVAR QC
Red in bottom panel means more TOVS 1DVAR passes internal cloud checks
And determined to be “clear”
Joanna Joiner, DAO ITSC-12

21. AIRS initial channel selection
Joanna Joiner, DAO ITSC-12

22. Channel selection based on retrieved cloud height
Cloud: 50% at 200 hPa
Yellow: Clear-Cloudy
Green: Add noise, background errors
17 channels unaffected by cloud
Joanna Joiner, DAO ITSC-12

23. Channel selection based on retrieved cloud height
Cloud: 10% at 700 hPa
Yellow: Clear-Cloudy
Green: Add noise, background errors
77 channels unaffected by cloud
(If retrieve pressure of 525, get 58 channels)
Joanna Joiner, DAO ITSC-12

24. Using model-simulated aerosol in DAOTOVS (Weaver poster)
Top: O-F HIRS 8 no dust in calculations
Bottom: O-F HIRS 8 dust from transport model included in radiative transfer
Joanna Joiner, DAO ITSC-12

25. Summary and Future Work
Cloud- and land-affected data has positive impact on forecasts (6hrs-5 days)
OPTRAN reduces biases, but little overall impact due to tuning
OSSE simulations show reasonable model cloud
TOVS Ts analysis (including bias correction) improves OLR, clear-scene identification over land
AIRS channel selection good for cloudy situations (sharp weighting functions); Dynamic channel selection in cloudy scenes, cloud slicing-like approaches worthwhile
Aerosol effects are significant (see Weaver poster)
Joanna Joiner, DAO ITSC-12

26. Joanna Joiner, DAO ITSC-12
In the future…
GOES sounder (JCSDA)
AMSU-B
Analyze pseudo-relative humidity instead of ln(q) in 1DVAR
Partial eigen-value decomposition/radiance assimilation
AIRS – more from Don Frank
Joanna Joiner, DAO ITSC-12