1. COMBINING STATISTICAL AND PHYSICAL RETRIEVALS FOR AERI SOUNDING
W. F. Feltz, *W. L. Smith, #D. Turner, H. B. Howell, and H. Woolf
Hyperspectral Resolution Workshop 2003
University of Wisconsin CIMSS/SSEC
*NASA LaRC, # Pacific Northwest National Laboratory

2. OVERVIEW Instrument Overview (Description and Locations)
Retrieval Improvements
Hybrid First Guess Retrieval
Vertical resolution and fast model update
Temperature and Moisture Retrieval Statistics
Conclusions

3. AERI SPECIFICATIONS
Spectral Resolution better than 1 cm-1 wavenumber from cm-1 ( um)
Calibrated to 1% ambient radiance (better than 1 K ambient temperature)
Automated and environmentally hardened
Time resolution: minutes (adjustable)
Ground-based and portable

4. AERI SYSTEMS AROUND THE WORLD
UW AERI - 2 (AERIBAGO, SSEC)
DOE AERI - 8 (Kansas/Oklahoma, Alaska, S. Pacific)
U-Miami M-AERI - 3 (Florida)
Bomem AERI - 6 (Italy, California, Maryland, Canada)
U Idaho P-AERI - 1 (Antarctica)

5. DOE ARM SGP AERI LOCATIONS

6. AERI CHANNEL RETRIEVAL SPECTRAL REGIONS

7. AERI Retrieval Flow Chart
Radiosonde
Climatology Regression
Data Base
T,Q Surface to 3 Km
AERI Radiance
Cloud Base Height
(if present)
Hybrid First Guess
Physical Retrieval
NWP Model/Satellite
T,Q Profile 2 Km -
Tropopause
Temperature/
Moisture Profile
Surface Moisture

8. AERIplus Retrieval Description
Two step retrieval, construct optimal first guess than conduct physical retrieval
Statistical retrieval is a combined regression and NWP profile
Physical retrieval used is iterative onion peel technique
Temperature and water vapor mixing ratio profiles retrieved from high resolution AERI spectra up to 3 km
Temporal resolution ~ ten minutes (increasing to < 1 minute)
Vertical resolution improved in lowest kilometer from 100 m to 50 m
Fast model update based on LBLRTM with HITRAN 2000 trans. coeffs
Constraints: AERI retrieval only possible in clear sky or below cloud base to 3 kilometers

9. Statistical Retrieval Description
Synthetic EOF Regression
~1100 clear radiosondes from DOE ARM central facility near Lamont, Oklahoma for
Forward calculations using 60-level fast model based on AER LBLRTM
EOF regression conducted relating each radiosonde – spectrum (562 channels) pair
Regression data set is very robust since a wide variety of weather is experiences in northern Oklahoma throughout the year

10. Hybrid First Guess
Statistical retrieval achieved during clear sky conditions (no update until a clear scene is present)
Statistical retrieval errors grow rapidly with altitude above one kilometer
Satellite profiles and NWP can be used to constrain the first guess through a simple linear interpolation between 1-2 km
This improved first guess constrains the physical retrieval and improves PBL retrieval profile
Currently combination of hourly RUC analysis/AERI statistical retrieval used over United States used to optimize the retrievals

11. Physical Retrieval Description
“Onion peel” retrieval methodology
Physical iterative recursive retrieval solution of the infrared radiative transfer equation
During each iteration the temperature and water vapor mixing ratio profile adjustments are made to minimize the difference between observed and calculated spectra
To correctly account for forward model spectroscopy and regression errors a static bias is used for water vapor (not needed for temperature retrieval spectral regions), currently bias is static but dynamic bias ties to water vapor is being considered with goal to be bias free

12. A Combined Statistical and Satellite First Guess
RMS Temperature Differences
RMS Water Vapor Mixing Ratio Differences

13. AERIplus, RUC-2, GOES Statistics

14. LBLRTM FAST MODEL
New model now uses HITRAN 2000 and vertical resolution doubled in first kilometer from 100 m to 50 m resolution resulting in improved water line residuals and improved temperature structure retrieval
15 m differences due to undersampling vertical levels in fast model
Improved agreement to AERI obs

15. Examples of the updated AERIplus temperature retrieval improvement during the IHOP field experiment. The red profile was calculated with old algorithm while the green profile uses the new LBLRTM based fast model with improved vertical resolution and spectroscopy as compared to radiosonde (thick black line)
Improved retrieval of temperature and water vapor profiles from AERI infrared radiances has been achieved by updated the AERI retrieval algorithm with a more accurate spectroscopic fast model based upon LBLRTM (instead of FASCODE) and increasing the vertical resolution within the first kilometer of atmospheric from 100 meters to 50 meters. An example of the temperature retrieval improvement is indicated from the IHOP field program on 02 June 2002 from Lamont, Oklahoma. AERI derived temperature profiles (red - old algorithm, green - new algorithm) are compared to radiosondes (black) at 1130 and 1430 UTC indicating the new algorithm can better resolve multiple temperature inversions matching the radiosonde profile more accurately.
Double inversion resolved with new algorithm

16. IHOP AERIplus Retrieval RMS Statistics

17. AERIplus and Radiosonde Comparisons
Hillsboro
3 June 0230 UTC
Vici
Lamont
Morris
Purcell

18. AERIplus Cross Section Examples

19. West – East AERI water vapor cross sections for IHOP

20. AERI Retrieved IHOP water vapor time height cross sections from 12 June 2002 indicating rapid water vapor oscillations also indicated by GPS retrieved integrated water vapor
The AERI instrument was deployed from May 13 - June 25, 2002 near Balko, Oklahoma (western Oklahoma panhandle) for the International H2O Program (IHOP) field campaign to measure water variability and how this relates to convective initiation and atmospheric boundary layer development. A rapid (30%) oscillation was retrieved from the AERI infrared radiances and this was also indicated from GPS retrieved integrated water vapor. This is the largest water variability in less than a 24 hour period retrieved from ten minute AERI radiances.
Total Precipitable Water
RUC-2
Stan Benjamin ?

21. Real-time Retrievals
Six AERI systems (including MSN) are currently processing PBL retrievals in near real-time, allowing “on the fly” validation:

22. Example: Rapid-scan AERI data
In the second comparison method, the radiative parameterization from the MM5 is modified to produce radiances in the same spectral range as that observed by the AERI. The AERI radiances are then averaged over the same discreet spectral bands, and compared with radiances from the MM5.
Band 4 from MM5 lies in the same spectral range used to make AERI temperature retrievals, and gives additional information as to the performance of the model with respect to near-surface temperature.
Band 6 from MM5 lies primarily in the atmospheric window region, and can be used to determine whether the model reproduced clouds accurately.
In the three plots shown, it is clear that the MM5 reproduced the near-surface temperatures fairly accurately, while it failed to adequately reproduce the clouds.
The MM5 simulation used in the above comparison was run at 20 km horizontal resolution, and was unable to resolve the small-scale processes that produced the fair-weather cumulus field. As a result, the MM5 produced a near-uniform layer of low cloud, which may be seen in the consistently high band 6 brightness temperatures.

23. Future Research Plans
High temporal resolution (< 1 minute) AERI radiance measurements were collected at Crystal and Texas 2002 to test improvement in gathering cloud property information
Rapid sampling, resulting in the ability to “see” the upper level cloud through holes in the lower cloud deck
Implement AER OSS fast model based on LBLRTM into AERI retrieval algorithm
Test new physical retrieval methodology will be tested to allow error bars and error covariance matrixes to be calculated for mesoscale model assimilation
A combined AIRS/IASI/CrIS retrieval with AERI system?
New research in determination of cloud properties (starting with cloud phase) using emissivity observations derived from AERI data (D. Turner).
One current on-going study involves validating the PSU/NCAR MM5 atmospheric numerical model against AERI data in real-time.
In the first of the two comparison methods, AERI retrievals are compared with time-height cross-sections of temperature and mixing ratio from MM5. The above figure depicts one example of this validation technique from the 2002 IHOP field experiment valid for the time period beginning 0000 UTC 21 June 2002 and ending 0000 UTC 22 June 2002.
The time-height cross-sections of temperature plotted in the upper panels indicate relatively close agreement between AERI retrievals and MM5 model output during this time period. Note the strong diurnal signature in the near-surface temperatures.
Plots of mixing ratio, in the lower panels, depict high time-resolution fluctuations of moisture in the boundary layer. On this particular day, widespread fair-weather cumulus was observed over much of Oklahoma. Passage of convective rolls in the boundary layer may be seen in the high time-resolution AERI retrievals, and are relatively poorly represented in the MM5 model output.

24. AERI Information: References:
Feltz, W. F., W. L. Smith, R. O. Knuteson, H. E. Revercomb, H. M. Woolf, and H. B. Howell, 1998: Meteorological Applications of Temperature and Water Vapor Retrievals from the Ground-Based Atmospheric Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 37,
Feltz, W. F., J. R. Mecikalski, 2002: Monitoring High Temporal Resolution Convective Stability Indices Using the Ground-based Atmospheric Emitted Radiance Interferometer (AERI) During the 3 May 1999 Oklahoma/Kansas Tornado Outbreak. Wea. Forecasting, 17,
Feltz, W. F., H. B. Howell, R. O. Knuteson, H. M. Woolf, and H E. Revercomb, 2003: Near Continuous Profiling of Temperature, Moisture, and Atmospheric Stability using the Atmospheric Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 42,
Smith, W. L., W. F. Feltz, R. O. Knuteson, H. E. Revercomb, H. B. Howell, and H. M. Woolf, 1999: The Retrieval of Planetary Boundary Layer Structure using Ground-based infrared spectral radiance measurements. J. Atmos. Oceanic Technol., 16,
Turner, D. D., W. F. Feltz, R. Ferrare, 2000: Continuous Water Vapor Profiles from Operational Ground-based Active and Passive Sensors. Bull. Amer. Soc., 81,
Turner D. D., S. A. Ackerman, B. A. Baum, H. E. Revercomb, and P. Yang, 2003: Cloud phase determination using ground-based AERI observations at SHEBA. Journal of Applied Meteorology, 42,
AERI HOMEPAGE:
REAL-TIME AERIplus: