GERB SW fluxes: an update(Clear ocean SW radiances and fluxes in the

reprocessed V999 June 04 data)

CCédric Bertrandédric Bertrand

GIST 24 14-16th December, Imperial College, London, U.K.

RECAP: unfiltering

New spectral response function

New spectral conversion:

LLLL

F

GERBF

SEVIRI

UF

SEVIRIUF

GERB.

where:

dILUF

SW

0

dISL

F

SW

0

= wavelength (m)I= reflected solar radiance (Wm-2.Sr-1.m-1)S=spectral response function (0 S 1)

LSEVIRI = a + b1 L0.6 + b2 L0.8 + b3 L1.6 + c11 L20.6 + c21 L0.8L0.6 + c22 L2

0.8 + c31 L1.6L 0.6 + c32 L1.6 L 0.8 + c33 L2

1.6 !!!! fct of SZA ONLY !!!!!!!! fct of SZA ONLY !!!!

= unfilter factor

V002

RECAP: unfiltering

Database of co-angular CERES and SEVIRI measurements(March, April and July 04 FM2/FM3 ES8 Edition 2 CERES data)

LSEVIRIUF (LCERES

UF = a + b L0.6 + c L0.8 + d L1.6 + e SZA sza 60°) with a, b, c, d and e surface types dependent

LSEVIRIF CAN NOT BE DETERMINED FROM CERES DATA

General case: LSEVIRI

F = a’ + b’ L0.6 + c’ L0.8 + d’ L1.6 + e’ SZA

Lc

LbaSBDART 2

6.06.0

""" (a", b" and c" = f(SZA,VZA,RAA))

V999:

Clear Ocean: LSEVIRIF = LSEVIRI

UF/SBDART

RECAP: angular conversion

),,,(),,,(

),,()(

mvmsmjmvmsmjadm

mvmsmmsm wsCF

wsRLF

Radm = clear ocean wind speed dependent CERES-TRMM BB ADM anisotropic correction factor

If (glint angle) < 25°

),,,()()( mvmsmjsmadmsm wsCFFF

FADM FDUF FUF

FADM FF FDUF FUF

FF

F l.n – F l.n-2:00 (W.m-2)

F l.n – F l.n+2:00 (W.m-2)

ARGARG JUNE 04JUNE 04

F l.n – F l.n-2:00 (W.m-2)FADM

FADM FS3

FG3

FG3

F l.n – F l.n+2:00 (W.m-2)

FS3

3x3 SEVIRI3x3 SEVIRI JUNE 04JUNE 04

RECAP: resolution enhancement (I)

Up-sampling of the filtered GERB radiances from the nominal GERB footprint resolution (i, j) to the 3 x 3 SEVIRI pixels resolution (x, y).

BUT: rather than LfGERB we use CF = Lf

GERB/LfSEVIRI

FIRST:FIRST: Fluxes at the GERB footprint resolution (i, j) are derived from the 3 x 3 SEVIRI pixels (x, y) based flux estimates

FGERB(i, j) = (xy PSF(i, j, x, y) . FS3 (x, y)). CFL(i, j)

CFL(i, j) = correction of SEVIRI by GERB

SECOND:SECOND: improvement of the spatial resolution of GERB fluxes by use of SEVIRI high resolution information

FGERB(i, j) = xy PSF(i, j, x, y) . CFH(x, y) . FS3 (x, y)

RECAP: resolution enhancement (II)

FINALLY:FINALLY: GERB flux at the 3 x 3 SEVIRI pixels resolution is given by FG3(x, y) = CFH(x, y) . FS3

(x, y)

use of LfADM

(x,y) rather than LfSEVIRI(x,y)

in the CFL(i,j) computation.

To reduce the impact of the SEVIRI spectral conversion in the estimation of the GERB SW fluxes at the 3x3 SEVIRI pixels resolution:

SBDART

UF

ADMF

ADMLL

Limitation: while SBDART is a function of the 3 acquisition angles (sza, vza, and raa) it is still dependent of the assumed GERB spectral response function

F l.n – F l.n-2:00 (W.m-2)FS3

FS3 FG3

FG3_ADM_L

FG3_ADM_L

F l.n – F l.n+2:00 (W.m-2)

FG3

3x3 SEVIRI3x3 SEVIRI JUNE 04JUNE 04

LS3/LADM LG3/LADM

05.06.04

SZA RAAVZA

UNFILTERED RADIANCE RATIO

LG3_ADM_L/LADM

-180

-90

FS3/FADM FG3/FADM

05.06.04

SZA RAAVZA

FLUX RATIO

FG3_ADM_L/FADM

-180

-90

ARG: 05-06-04

LUF/LADM FUF/FADM

GERB and SEVIRI images recorded at times temporal matching neededScene types independent

Resolution enhancement limitations:

GERB pixel

3x3 SEVIRI pixels footprint

100% clear

100% cloudy

Clear ocean S3 footprint for which the CFH(x,y) will be cloud contaminated

Conclusions

Problems over clear land surfaces which were not found in the V002 version

Asymmetry problem in the diurnal evolution of the SW fluxes over clear ocean surface still present. mainly due to deficiencies in the SEVIRI spectral modeling but also possible impact of the - GERB spectral response function through its influence on the estimated Lf

SEVIRI and thus on the CFs

- resolution enhancement reduced clear ocean pixels at the GERB nominal spatial resolution ( temporal matching between GERB and SEVIRI)

ARG Comparison

Unscreened

Cloudy pixels

ARG Comparison

Unscreened Cloudy pixels

GERB spectral response function ?

ARG Comparison

UnscreenedCloudy pixels

Direct unfiltering

Surface type dependent: ocean, vegetation and desert surfaces

The normalized unfilter factor is estimated

constaint: X = 0 Y = 1 X = 1 Y = 0

Where a, b, c and d are SZA dependent X is the normalized filtered radiance

cldoc

cldY

)()( 2cXd

cXbaY

LLLLXcldoc

cld

SBDART filtered SW radiance

SB

DA

RT

unf

ilte

r fa

ctor

Mean of the 5% highest values Lcld

cld

Mean of the 5% lowest

valuesLoc

oc

ARG FF, FDUF and FUF computation

RAL G2_L15A files LF

SZA, VZA, RAA

RMIB G2_SEV1_L20S files Surface type Cloud cover, cloud amount, cloud phase LUF, FUF

Assuming that the cloud optical depth was retrieved in the VIS 0.8 SEVIRI channel over ocean surface (less sensitive to the Rayleight scattering)

FF and FDUF

ARG FF, FDUF and FUF computation

WARNING: The GERB PSF is applied in each of the RMIB G2_SEV1_L20S file fields.

Not defined

FFFF FFDUFDUF FFUFUF