High Spectral Resolution Infrared Land Surface Modeling & Retrieval for MURI 28 April 2004 MURI Workshop Madison, WI Bob Knuteson UW-Madison CIMSS

Land Surface Characterization Needed for Atmospheric Remote Sensing MURI Topic Areas: Spectral emissivity maps from MODIS data (Lucy-UH, Seeman-UW). Enhanced Training sets including IR emissivity and Tskin/Tair along with corresponding vertical Temperature/Water Vapor profiles (S. W. Seeman/E. Borbas-UW). Radiative Transfer Theory (Jun Li, Youri Plokhenko, R. Knuteson) Satellite Validation (H. Revercomb, D. Tobin, R. Knuteson)

Radiative Transfer Theory

The Correlation Problem: Surface Temperature (K) Surface Emissivity Slope at 10  m 1%  E  -0.5 K  Ts For broad-band sensors, such as HIRS, GOES, MODIS, errors in the IR emissivity and surface temperature are highly correlated. Solution: High Spectral Resolution Infrared Observations

Infrared Radiative Transfer Equation (lambertian surface) Formal Solution Surface Emission Surface Reflection Analytic Derivative Varies on/off spectral lines !!!

Simulated Radiance ( Using measured emissivity spectrum) Ts = K Bare Soil Vegetation 60%-40% combination

Simulated IR Reflected Radiance Contribution to TOA Radiance Vegetation 60% Veg. Bare Soil Reflected contribution can be large ! Radiance (mW/(m 2 sr cm -1 ))

The value of Ts can be determined from the variance of emissivity as a function of surface temperature !!! Std. Dev. E(Ts) Emissivity vs. Surface Temperature Minimum Intersection

dE dTs The change in emissivity with Ts varies on and off atmospheric absorption lines! E

Satellite Product Validation

Courtesy of A. Trishchenko DOE ARM Southern Great Plains (SGP) Site Land Cover From MODIS Data

9 miles (15 km) Define an Effective Emissivity and Effective Surface Temperature such that The observed radiance is a linear combination of uniform scenes. The Problem of Mixed Scenes

Emissivity Survey ARM SGP site is dominated by two land cover types “grass vegetation” and “bare soil”. In situ UW surface and aircraft measurements can be represented by a linear combination of pure scene types; bare soil and grass.

Aircraft validation measurements are also consistent with a linear combination of vegetation and bare soil. Aircraft S-HIS LSE Wavenumber (cm -1 ) Bare Soil Pure Vegetation S-HIS OBS

AIRS Granule for 16 Nov :24 UTC Brightness temperature across ARM site at 12  m is fairly uniform. Granule ARM SGP Site

B.T. Diff 9-12  m AIRS Observations over the DOE ARM SGP site Notice the East/West gradient in the B.T. Difference.

AIRS emissivity is consistent with a linear combination of pure scene types. This implies a single vegetation fraction can explain most of the variation in the IR spectra over land. Wavenumber (cm -1 ) Pure Vegetation Bare Soil LSE from AIRS Radiance Using UW Research Algorithm Wavenumber (cm -1 ) Research Product

AIRS 12 µm B.T. (K) LST (K) LST is 2 to 4 degrees warmer than 12  m brightness temperature. UW Research Product shows spatial gradient in land temperature. Research Product

LST (K) LSE (9 µm) Emissivity from UW “research” product shows East/West gradient. High emissivity (grass) is cooler than low emissivity (exposed soil). Research Product

Product retrieval is spatially uniform. No East/West gradient! Tsurf (K) IR Emiss (9 µm) AIRS Standard Product Version

B.T. Difference clearly shows East/West spatial gradient ! AIRS Brightness Temperature Observations (9-12  m)

(Lat, Lon)=(36.590, ) Tobin-ARM SGP Best Estimate for 16 Nov :24 UTC AIRS standard retrieval is within the AMSU footprint after CC.

TSurfStd 290.5K TSurfAir 285.7K AIRS standard retrieval misses spectral contrast in 9  m emissivity. AIRS Cloud-Cleared Radiance for 16 Nov :24 UTC *

Tobin-ARM SGP Best Estimate for 16 Nov :24 UTC ARM “best estimate” interpolates sondes before and after launch.

AIRS standard retrieval “agrees” with sonde2 in below 500 mb. AIRS standard retrieval agrees with sonde1 in above 500 mb.

AIRS “standard retrieval” agrees well with ARM Best Estimate Profile. Tobin-ARM SGP Best Estimate for 16 Nov :24 UTC

AIRS “standard level” retrieval looks good near the surface! Tobin-ARM SGP Best Estimate for 16 Nov :24 UTC

AIRS “100 level” retrieval adds more points near the surface.

AIRS/SGP Overpass 19:24 UTC AIRS TSurfAir 285.7K TSurfStd 290.5K AIRS “surface air” temperature is within 1 degree of truth data! Tobin-ARM SGP Best Estimate for 16 Nov :24 UTC AERI B.T. “truth”

Questions Raised What are the benefits and limitations of the online/offline method for separating surface temperature and emissivity? What improvements are needed in radiative transfer models to take advantage of the high spectral surface reflection in operational models? Is the AIRS Cloud-Clearing working over land or is it introducing “noise” into the retrievals? What are the statistics of the validation of AIRS profile retrievals over land? What about near surface air temperature? How can AIRS data best be used to improve the global characterization of infrared spectral emissivity?

Backup Slides

Tskin Tair

ARM Site Land Use Survey Wheat 57% Pasture & Range 25% Bare soil 6% Rubble 4% Dense trees 4% Rubble & wheat mixture 4% Other 4% November 2002; 63 square mile area. Two land cover types dominate: wheat fields and pasture (grassland). (Osborne, 2003)

ARM SGP LST/LSE “Best Estimate” Formulated in April 2001 to supply the surface contribution to the ARM/AIRS validation product developed by D. Tobin.

Simulated Radiance (S-HIS resolution = 1 cm -1 apodized)

On-line channels have a greater rate of change, dE/dTs ! B.T. (K) Simulated Brightness Temperature Spectrum Ts = K Wavenumber (cm -1 ) Emissivity vs. Surface Temperature