validation of surface-only flux algorithms science branch, nasa langley research center validation...

Climate Science Branch, NASA Langley Research Center

Validation of Surface-Only Flux Algorithms

Twenty Fifth CERES-II Science Team MeetingNASA/LaRC

Hampton, VirginiaApril 26, 2016

David P. Kratz1, Shashi K. Gupta2, Anne C. Wilber2, Victor E. Sothcott2

1NASA Langley Research Center2Science Systems and Applications, Inc.

https://ntrs.nasa.gov/search.jsp?R=20160009257 2019-07-06T19:46:52+00:00Z


Background (Part 1)

CERES uses several Surface-Only Flux Algorithms (SOFA) to compute SW and LW surface fluxes in addition to the more precise model used by SARB. The SOFA algorithms include:

LPSA/LPLA:Langley ParameterizedSW/LW Algorithm

SOFA References: SW A: Li et al. (1993): J. Climate, 6, 1764-1772. SW B: Darnell et al. (1992): J. Geophysical Research, 97, 15741-15760. SW B: Gupta et al. (2001): NASA/TP-2001-211272, 31 pp. LW A: Inamdar and Ramanathan (1997): Tellus, 49B, 216-230. LW B: Gupta et al. (1992): J. Appl. Meteor., 31, 1361-1367. LW C: Zhou et al. (2007): J. Geophysical Research, 112, D15102.

SOFA: Kratz et al. (2010): J. Appl. Meteor. Climatology, 49, 164-180. SOFA: Gupta et al. (2010): J. Appl. Meteor. Climatology, 49, 1579-1589.

FLASH: Kratz et al. (2014): J. Appl. Meteor. Climatology, 53, 1059-1079.

Model A Model B Model C Clear Li et al. LPSA --

SW All-Sky -- LPSA -- Clear Inamdar and

Ramanathan LPLA Zhou-Cess

LW All-Sky -- LPLA Zhou-Cess


Background (Part 2)•  The SOFA LW and SW Models are based on rapid, highly parameterized TOA-to-

surface transfer algorithms to derive the surface fluxes.

•  LW Model A (clear-sky only), LW Model B (all-sky) and SW Model A (clear-sky only) were incorporated at the start of the CERES project.

•  SW Model B (all-sky) was adapted for use in the CERES processing shortly before the launch of the CERES instrument on the TRMM satellite.

•  The Edition 2B LW and SW surface flux results underwent extensive validation [Kratz et al., 2010].

•  The ongoing validation process led to improvements to the LW models [Gupta et al., 2010].

•  LW Model C (all-sky) [Zhou et al., 2007] was introduced into the Edition 4 processing to maintain two independent LW algorithms after the broadband LW Channel was chosen to replace the CERES Window Channel for CERES FM-6 and the next generation Radiation Budget Instrument (RBI).

•  LW and SW Models B were incorporated into the FLASHFlux effort to produce a rapidly available Environmental Data Record [Kratz et al., 2014]

Recent and Future Improvements to the Surface-Only Flux AlgorithmsSW Model Improvements: 1) Replacing the ERBE albedo maps with Terra maps greatly improved the SW retrievals, most notably for polar regions. 2) Replacing the original WCP-55 aerosols properties with monthly MATCH/OPAC datasets while also replacing the original Rayleigh molecular scattering formulation with the Bodhaine et al., (1999) model significantly improved SW surface fluxes for clear conditions. 3) To account for the short term aerosol variability we have incorporated daily MATCH aerosol data into Edition 4. 4) Using a revised empirical coefficient in the cloud transmission formula has improved the SW surface fluxes for partly cloudy conditions. 5) Work continues on the improvement of the cloud transmission method for Edition 4 clouds. Use of SSF-82 (cloud layer note) may reduce uncertainties in surface flux retrievals.LW Model Improvements: 1) Constraining the lapse rate to 10K/100hPa (roughly the dry adiabatic lapse rate) improved the derivation of surface fluxes for conditions involving surface temperatures that greatly exceeded the overlying air temperatures, see Gupta et al. (2010). 2) Limiting the inversion strength to -10K/100hPa for the downward flux retrievals provided the best results for cases involving surface temperatures that were much below the overlying air temperatures (strong inversions).SW and LW Model Improvements: 1) The availability of ocean buoy measurements may improve surface flux retrievals by providing validation over oceans.

Parameterized models for fast computation of surface fluxes for both CERES and FLASHFlux


Status of Total Solar Irradiance (TSI) Measurements

The SORCE TIM (Total Irradiance Monitor) began producing TSI data on February 25, 2003. A battery failure on SORCE, however, stopped regular production during the period from July 16, 2013 through March 4, 2014.

Thus, for the period from July 1, 2013 though October 31, 2013 we used the RMIB composite TSI data provided by S. Dewitte.

The RMIB data, however, requires an offset from the DIARAD VIRGO solar minimum value of ~1363 W/m2 to match the SORCE solar minimum of ~1361 W/m2. Note, for CERES Edition 4 processing all TSI data are offset to match SORCE TSI Version 15. The offset is: V15 – V17 = −0.0049 W/m2.

In the meantime, the TSI Calibration Transfer Experiment (TCTE) was launched into orbit on November 19, 2013 and began producing TSI data on an irregular basis on December 16, 2013, and more recently, on a regular daily basis since January 1, 2015.

The SORCE instrument resumed data production on a daily basis on March 5, 2014. CERES subsequently resumed merging the SORCE TSI data into the CERES processing beginning on November 1, 2014.


TSI composite data from WRC, SORCE(V15) and RMIBfor the Timeframe of CERES Terra, Aqua & NPP

1356

1357

1358

1359

1360

1361

1362

1363

1364

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Tota

l Sol

ar Ir

radi

ance

(Wm

-2)

Year

WRC Composite

SORCE (V15/V17)

RMIB Composite

01Mar2000 - 24Feb2003

25Feb2003 - 30Jun201301Nov2014 - 31Dec201501Jul2013 - 31Oct2014

Total Solar Irradiance for CERES Edition-4 (20000301-20151231)


For CERES Ed4, all TSI data are offset to match SORCE TSI Version 15

Comparison of TSI data [SORCE(V15) versus RMIB] for the 5-year overlap period 1-Mar-2003 to 29-Feb-2008

13561357135813591360136113621363136413651366

2003 2004 2005 2006 2007 2008 2009

Tota

l Sol

ar Ir

radi

ance

(Wm

-2)

SORCE-V15 RMIB-Comp RMIB-Offset

2.0

2.2

2.4

2.6

2.8

3.0

2003 2004 2005 2006 2007 2008 2009

RMIB

-V15

TSI

(Wm

-2)

Year

RMIB-V15

Mean Offset = 2.4447

RMIB - SORCE V15 Offset -- 01Mar 2003 to 29 Feb 2008

Slope = -0.0237 Wm-2/year


Slope of RMIB vs. SORCE is -0.0237 W/m2/y which yields an offset of 1 W/m2 in 42.19 years

This timeframe corresponds to the first 60 months of the SORCE data record

Comparison of SORCE(V15) and RMIB for the timeframe: 1-Mar-2003 to 23-Jun-2015

1356

1358

1360

1362

1364

1366

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Tota

l Sol

ar Ir

radi

ance

(Wm

-2)

RMIB Comp SORCE V15 RMIB Offset

1.8

2.0

2.2

2.4

2.6

2.8

3.0

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Diffe

renc

e (W

m-2

)

Year

RMIB Offset - V15

RMIB Offset vs.SORCE V15 -- 01 Mar 2003 to 23 Jun 2015

Slope = -0.0229 Wm-2/year

Mean (RMIB - SORCE V15) = 2.3653 Wm-2


Slope of RMIB versus SORCE -0.0229 W/m2/y corresponds to first 120 months and yields an offset of 1 W/m2

in 43.67 years

This timeframe corresponds to the first 148 months of the SORCE data record

RMIB offset by +2.3653 W/m2

from SORCE

Comparison of SORCE(V17) and TCTE(V02) Daily/Weekly Total Solar Irradiance Retrievals

1358

1359

1360

1361

1362

1363

1364

Tota

l Sol

ar Ir

radi

ance

(Wm

-2)

TCTE V02 SORCE V17

201312 201403 201406 201409 201412 201503 201506 201509 201512

TSI Comparison: SORCE V17 vs. TCTE V02 (16 Dec 2013 to 31 Dec 2015)


SORCE: 1 value/day, Dec 22, 2013 through Dec 28, 2013, and 1 value/dayMar 5, 2014 through Mar 31, 2016; Absolute Accuracy: ±0.48 W/m2 at 1361 W/m2

TCTE: 1 value/day, Dec 16, 2013 through May 8, 2014, 1 value/week May 11, 2014 through Dec 31, 2014, 1 value/day Jan 1, 2015 through Mar 31, 2016; Absolute Accuracy: ±1.36 W/m2 at 1361 W/m2

SORCE minus TCTE (Jan 1, 2015 to Dec 31, 2015) is -0.4879 ± 0.0599 W/m2 SORCE minus TCTE (Jan 1, 2015 to Mar 31, 2016) is -0.4965 ± 0.0582 W/m2

Comparison of SORCE(V17) and TCTE(V02) Daily Total Solar Irradiance Retrievals

1358

1359

1360

1361

1362

1363

1364

Tota

l Sol

ar Ir

radi

ance

(Wm

-2)

TCTE V02 SORCE V17

201312 201403 201406 201409 201412 201503 201506 201509 201512

TSI Comparison: SORCE V17 vs. TCTE V02 (16 Dec 2013 to 31 Dec 2015)


SORCE: 1 value/day, Dec 22, 2013 through Dec 28, 2013, and 1 value/dayMar 5, 2014 through Mar 31, 2016; Absolute Accuracy: ±0.48 W/m2 at 1361 W/m2

TCTE: 1 value/day, Dec 16, 2013 through May 8, 2014, 1 value/week May 11, 2014 through Dec 31, 2014, 1 value/day Jan 1, 2015 through Mar 31, 2016; Absolute Accuracy: ±1.36 W/m2 at 1361 W/m2

SORCE minus TCTE (Jan 1, 2015 to Dec 31, 2015) is -0.4879 ± 0.0599 W/m2 SORCE minus TCTE (Jan 1, 2015 to Mar 31, 2016) is -0.4965 ± 0.0582 W/m2

TSI composite data from WRC, SORCE(V15) and RMIBfor the Timeframe of CERES Terra, Aqua & NPP

1356

1357

1358

1359

1360

1361

1362

1363

1364

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Tota

l Sol

ar Ir

radi

ance

(Wm

-2)

Year

WRC Composite

SORCE (V15/V17)

RMIB Composite

01Mar2000 - 24Feb2003




For CERES Ed4, all TSI data are offset to match SORCE TSI Version 15

1356

1357

1358

1359

1360

1361

1362

1363

1364

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

Total

Solar

Irrad

iance

(Wm-2 )

Year

WRC Composite

SORCE (V15/V17)

RMIB Composite

01Mar2000 - 24Feb2003



Sunspot Numbers for Solar Cycles 23 & 24


Surface Sites Available for Validation of Terra & Aqua Ed4, and NPP Ed1


Continental

Desert

Coastal

Island

Polar

Ocean

CERES Terra Edition 4 LW Ground Validation (Global)Clear-Sky

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)

Ground Measured DLF (W m-2)

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

N = 21029Bias = 5.1 W m-2

R.E = 15.0 W m -2

N = 21032Bias = -4.7 W m-2

R.E = 14.1 W m -2

N = 20969 Bias = -4.7 W m-2

R.E = 13.6 W m -2

LWA LWB LWC

Combined LW Ground Validation for Terra (4/2000 through 12/2014).


CERES Terra Edition 4 LW Ground Validation (Global)All-Sky

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


LWB

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

LWC

N = 210752Bias = 2.6 W m-2

R.E = 23.1 W m -2

N = 209526Bias = 3.1 W m-2

R.E = 21.5 W m -2

Combined LW Ground Validation for Terra (4/2000 through 12/2014).


CERES Aqua Edition 4 LW Ground Validation (Global)Clear-Sky

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

N = 17678Bias = 6.1 W m-2

R.E = 14.9 W m -2

N = 17682Bias = -4.7 W m-2

R.E = 13.8 W m -2

N = 17596Bias = -4.9 W m-2

R.E = 12.6 W m -2

LWA LWB LWC

Combined LW Ground Validation for Aqua (7/2002 through 12/2014).


CERES Aqua Edition 4 LW Ground Validation (Global)All-Sky

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


LWB

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d4A

Der

ived

DLF

(W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

LWC

N = 185263Bias = 2.4 W m-2

R.E = 23.0 W m -2

N = 184104Bias = 2.8 W m-2

R.E = 21.2 W m -2

Combined LW Ground Validation for Aqua (7/2002 through 12/2014).


CERES NPP Edition 1 LW Ground Validation (Global)Clear-Sky

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d1A

Der

ived

DLF

(W

m-2

)


0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d1A

Der

ived

DLF

(W

m-2

)


0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d1A

Der

ived

DLF

(W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

N = 5402Bias = -0.6 W m-2

R.E = 14.4 W m -2

N = 5407Bias = -7.7 W m-2

R.E = 13.7 W m -2

N = 5337 Bias = -5.9 W m-2

R.E = 12.1 W m -2

LWA LWB LWC

Combined LW Ground Validation for NPP (1/2012 through 12/2015).


CERES NPP Edition 1 LW Ground Validation (Global)All-Sky

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d1A

Der

ived

DLF

(W

m-2

)


LWB

0

100

200

300

400

500

600

0 100 200 300 400 500 600

CE

RE

S E

d1A

Der

ived

DLF

(W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

LWC

N = 90272Bias = -1.4 W m-2

R.E = 21.9 W m -2

N = 89547Bias = 1.2 W m-2

R.E = 20.9 W m -2

Combined LW Ground Validation for NPP (1/2012 through 12/2015).


LW Surface Flux Results

The Terra, Aqua and NPP LW fluxes derived for Clear Sky conditions show low biases ≈±6 W/m2, and relatively low uncertainties ≈12 to 15 W/m2.

The Terra, Aqua and NPP LW fluxes derived for Cloudy Sky conditions show low biases ≈±3 W/m2, and moderate uncertainties ≈20 to 23 W/m2.


CERES Terra Edition 4 SW Ground Validation Global Clear-Sky

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)

Ground Measured DSF (W m-2)

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

SWA SWB

N = 8239Bias = -11.0 W m-2

R.E. = 41.7 W m-2

N = 8519Bias = -3.4 W m-2

R.E. = 35.5 W m-2

Combined SW Ground Validation for Terra (4/2000 through 12/2014).


CERES Terra Edition 4 SW Ground Validation (Global)All-Sky

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


1 2 - 50 51 - 100 101 - 500 501 -1000 >1000

SWB

N = 99111Bias = -7.9 W m-2

R.E. = 93.0 W m-2

SWB Ground Validation for Terra (4/2000 through 12/2014).


CERES Aqua Edition 4 SW Ground Validation Global Clear-Sky

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


1 2 - 25 26 - 100 101 - 600 601 - 800 > 800

SWA SWB

N = 6417Bias = -3.6 W m-2

R.E. = 39.2 W m-2

N = 6592Bias = -2.4 W m-2

R.E. = 30.8 W m-2

Combined SW Ground Validation for Aqua (7/2002 through 12/2014).


CERES Aqua Edition 4 SW Ground Validation (Global) All-Sky

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


1 2 - 50 51 - 100 101 - 500 501 -1000 >1000

SWB

N = 86632Bias = -5.3 W m-2

R.E. = 92.1 W m-2

SWB Ground Validation for Aqua (7/2002 through 1/2014).


CERES NPP Edition 1 SW Ground Validation (Global) Clear-Sky

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


1 2 - 5 6 - 20 21 - 40 41 - 50 > 50

SWA SWB

N = 2141Bias = -5.0 W m-2

R.E. = 31.2 W m-2

N = 2180Bias = -0.4 W m-2

R.E. = 25.5 W m-2

Combined SW Ground Validation for NPP (1/2012 through 12/2015).


CERES NPP Edition 1 SW Ground Validation (Global)All-Sky

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200

CE

RE

S E

d1A

Der

ived

DS

F (W

m-2

)


1 2 - 20 21 - 100 101 - 200 201 - 400 > 400

SWB

N = 42052Bias = -4.7 W m-2

R.E. = 91.7 W m-2

SWB Ground Validation for NPP (1/2012 through 1/2015).


SW Surface Flux Results

The Terra, Aqua and NPP SW fluxes derived for Clear Sky conditions show relatively low biases ≈0 to -10 W/m2, and moderate uncertainties ≈25 to 35 W/m2.

The Terra, Aqua and NPP SW fluxes derived for Cloudy Sky conditions show relatively low biases ≈-5 to -10 W/m2, but large uncertainties ≈90 W/m2.

Our current effort is focused on finding methods to reduce the large uncertainties in the retrieved SW Cloudy Sky surface fluxes.


Conclusions for SOFA Ed4 algorithms

Validation studies have demonstrated that revisions to both the LW algorithms and the SW algorithms (for clear to partly cloudy conditions) appear to be working well, though further revisions to the cloud transmission method and/or overcast albedo method are needed for SW Model B. Current attention is focused on the cloud transmission data, and whether use of SSF-82 (cloud layer note) could be used to refine the results.

An analysis of the LW and SW surface only flux algorithm results using the Edition 4 inputs, especially those from the Clouds Subsystem, has indicated improved accuracies for most locations.

The Terra, Aqua and NPP flux retrievals show the anticipated results.


validation of surface-only flux algorithms science branch, nasa langley research center validation...

Documents