can aeronet help with monitoring clouds?

May 10, 2004 Aeronet workshop

Can AERONET help with monitoring clouds?

Alexander MarshakNASA/GSFC

Thanks to:

Y. Knyazikhin, K. Evans, W. Wiscombe,

I. Slutsker and B. Holben

Supported by:

NASA Radiation Science Program


AERONETAerosol optical depth (from Cordoba-CETT, Argentina)

22 March, 2000

Clouds


Outline

• Overview of cloud optical depth retrievals from ground-based radiometers;

• Main part: • Cloud optical depth retrievals from Cimel’s radiances,

examples, comparison with other surface retrievals at the ARM site;

• Surface inhomogeneity, sensitivity study;• Local climatology of several Aeronet sites• Testing with simulated clouds;

• Summary/Conclusion


Ground-based retrieval

downward radiance downward

flux

upward radiance

(flux)

Common approach is to use downward fluxes:

• broadband pyranometers (Leontieva & Stamnes, 1994; Boers, 1997)

• narrowband radiometers (Min and Harrison, 1996, Min et al., 2003)

from Barker and Marshak, JAS 2001

computed by DISORT:g=0.85, 0=1, surf=0.2


Clouds with Low Optical (Water) Depth “CLOWD”

– Optical depth is the most fundamental cloud optical property

– Solar radiation is most sensitive to changes in cloud optical depth at low optical depths

– Over 50% of the warm liquid water clouds at the SGP site have LWP < 100 g m-2

– MWR’s uncertainty is 20-30 g m-2 (i.e., errors of 20% to over 100%)

– Aerosol indirect effect research needs accurate measurements of LWP and effective radius

ARM IS UNABLE TO ADEQUATELY OBSERVE OVER HALF OF THE CLOUDS OVERHEAD !!

Courtesy of Dave Turner, PNNLPresentation at the ARM STM


Intercomparison between dif. retrievals for 14 March 2000

(Variable Thickness Stratus Case)

Results from 14 Mar in the ARM 2000 Cloud IOP at the ARM SGP site, a day when the cloud was particularly stratiform and uniform

Raman lidar backscatter

Radar reflectivity

comparisons among many volunteered methods for retrieving the low LWP

20:44

courtesy of Dave Turner, PNNLPresentation at the ARM STM


Ground-based retrieval;zenith radiance from simulated 3D clouds

0

0.2

0.4

0.6

0.8

1

1.2

0 10 20 30 40 50 60 70

3D NIR1D NIR

optical depth


Objectives

from Michalski et al., 2002

Pilewskie’s SSFR within 0.0001 rad (640 m) from the

CART site flying at 300 m on April 5, 2000

X

X NIR

REDX BLUE

to exploit the sharp spectral contrast in vegetated surface reflect. across 0.7 µm to retrieve cloud properties from

the Cimel’s measurements


Cimel radiance measurements (GSFC, Bld. 33): four channels (440, 670, 870, and 1020 nm)

0.0

5.0

10.0

15.0

20.0

25.0

30.0

"18:59:39" "19:07:09" "19:14:39"

1020 nm870 nm670 nm440 nm

normalized radiance (arbitrary units)

time (in GMT)

24 May 1999

Clear sky

BrokenClouds

OvercastSky

(c) Transition between (a) and (b):• sharp changes in I around cloud edges;• the “order” of I between all four channels rapidly changes from cloudy to clear and back.

(b) Surface and Clouds dominate:I440 ≈ I670 < I870 ≈ I1020

Cloud optical properties can be retrieved

(a) Atmosphere dominates:I440 > I670 > I870 > I1020

Aerosol optical properties can be retrieved


Two-channel cloud retrievalsSatellite retrieval of and re

from Nakajima-King, JAS 1990

IREDINIRclearcloudyABCα

Surface retrieval and Ac

from Marshak et al., JAS 2004

points A & B are assumed to have the same cloud optical depth but different cloud fraction Ac


2D Look-Up TablesNIR vs. RED plane

IRED= IRED (, Ac)INIR = INIR (, Ac)

is cloud optical depthAc is “effective” cloud fraction

0

0.2

0.4

0.6

0.8

1

0 0.1 0.2 0.3 0.4 0.5 0.6

RED

Ac

0.0

0.2

0.4

1.0

0.6

0.8

10060

40 3020

10

1

3

5

0

0.2

0.4

0.6

0.8

0 0.2 0.4 0.6 0.8 1 1.2 1.4

NIR+RED

Ac

0.0

0.2

0.4

1.0

0.6

0.8

100

60

40

30

20

10

3 51

clear

cloudy


Where are Cimel data points?

0

0.02

0.04

0.06

0.08

0.1

0.12

0 0.2 0.4 0.6 0.8 1 1.2

NIR+RED

1.0

0.9

0.8

0.7

0.6

70

4020 15 10

30

2

4

Ac

RED

=0.092

NIR

=0.289

=62SZA 0

July 28, 2002 ARM CART siteCimel measurements are taken around

13:45, 13:58 and 14:11 UTTSI is from 14:00 UT


Retrieval examples

August 8, 2002;18:00 UT CART site

Cloud optical depth retrieved from:• Cimel• MWR (Microwave Radiometer) assuming re = 7 m

• MFRSR (Multi-Filter Rotating Shadowband

Radiometer)


Retrievals

SZA=16.3Time: 17:50

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0 0.2 0.4 0.6 0.8 1 1.2

LUTCimel

NIR+RED

SZA=530+/-1

0

time:14.5-14.6 UMT

30 25 20 15 10

0.5

35

Ac 0.6

1.0

0.7

( 2)b

0.8

0.9

53

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0 0.5 1 1.5 2 2.5 3

NIR+RED

SZA=170+/-1

0

time:17.7-17.9 UMT

302520

15

10

3

0.5

100

Ac

0.7

1.0

0.8

0.9

0.6

( 2)c

SZA=52.3Time: 14:36

MFRSR data is courtesy of Q. Min

0

20

40

60

80

100

13 14 15 16 17 18 19

July 3, 2002

MWRCimelMFRSRSZA

Time(UT)


Independent retrieval from BLUE and RED SZA: 29

Time: 20:36

QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.

SZA: 19Time: 19:32

SZA: 64Time: 14:00

QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.

SZA: 30Time: 16:24

QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.SZA: 47Time: 15:17

0

20

40

60

80

100

0 20 40 60 80 100

7.287.03

BLUE vs NIR

MODIS surface albedo:BLUE: 0.044RED: 0.092NIR: 0.289

Scatter-plot of RED_vs_NIR

againstBLUE_vs_NIR

retrievals

0

20

40

60

80

14 16 18 20 22

- retrievals using both blue and red channels 28, 2002July

MWR_07_44_87tau_07_67_87tau

MFRSRSZA

( )Time UT


Local Climatology(broadleaf cropland: Bondville, IL)

0

100

200

300

400

500

0 10 20 30 40 50 60 70 80

Bondville (IL)July-August, 2001

optical depth,

1 2 3 4 5

1

2

3

4

5

1 km pixels

1 km pixels

0.1 0.2 0.3 0.4 0.5 0.6Albedo (670 nm)

A2001_193_670_Bondville

1 2 3 4 5

1

2

3

4

5

1 km pixels

1 km pixels

0.1 0.2 0.3 0.4 0.5 0.6Albedo (870 nm)


1 2 3 4 5

1

2

3

4

5

1 km pixels

1 k mpixels

0.0 0.1 0.2 0.3 0.4 0.5 0.6Albedo (670 nm)


1 2 3 4 5

1

2

3

4

5

1 km pixels

1 km pixels

0.1 0.2 0.3 0.4 0.5 0.6Albedo (870 nm)


MODIS srf. albedo670 nm

MODIS srf. albedo870 nm


Local Climatology(Santa Barbara, CA: 2003) Version 0 cloud optical depth product

0

100

200

300

400

500

0 20 40 60 80 100

UCSB

tau

0

500

1000

1500

2000

2500

3000

3500

4000

0 0.2 0.4 0.6 0.8 1

UCSB

Ac

0

20

40

60

80

100

0 20 40 60 80 100

UCSB, 2003

y = 0.7 + 0.98x R= 0.996

RED/NIR

Cloud optical depth

“Effective”cloud fraction

RED&NIR vs. BLUE&NIR retrievals


Local climatologyARM CART cite

Version 0 optical depth product

0

200

400

600

800

1000

1200

20 40 60 80 100

CART site2001-2003

tau

Cloud optical depth

0

500

1000

1500

2000

2500

3000

0 0.2 0.4 0.6 0.8 1

CART site2001-2003

Ac

"Effective" cloud fraction

0

20

40

60

80

100

0 20 40 60 80 100

ARM CART site2001-2003

y = 0.19 + 0.90x R= 0.986

RED/NIR

RED&NIR vs. BLUE&NIR retrievals


Seasonal applicability

0

0.2

0.4

0.6

0.8

1

2001.5 2002 2002.5 2003 2003.5 2004

ARM CART

1 pixel3x3 avg

Date (decimal years)

0

0.1

0.2

0.3

0.4

0.5

0.6

2001.5 2002 2002.5 2003 2003.5 2004

AlbedoBondville, IL

470nm

648nm

858nm

470av

648av

858av

Date (years)

0

0.2

0.4

0.6

0.8

1

2001.5 2002 2002.5 2003 2003.5 2004

Bondville, IL

1 pixel3x3 avg

Date (decimal years)

0

0.1

0.2

0.3

0.4

0.5

2001.5 2002 2002.5 2003 2003.5 2004

AlbedoARM site470nm

648nm

858nm

470av

648av

858av

Date (years)

ARM CART site, OK Bondville, IL

Surface albedo

NDVI


How good is the retrieval?Sensitivity to Surface Albedo

July 3, 2002, ARM CART site

0.05 0.10 0.15

0.25

0.30

0.35

RED albedo

NIR albedo

-10 -5 0 5 10 15 20 25 30delta tau in % (0.1 RED, 0.3 NIR ref.)

• If the uncertainties in surface albedo have the same sign, the algorithm performs well.

• If the NIR albedo is overestim. but the RED albedo is underestim., errors in the retrieved are not severe.

• In the opp. case, the algorithm underestim. multiple refl. in the bright band and greatly overestim. .

0.05 0.10 0.15

0.35

0.30

0.25

RED albedo

NIR albedo

-5 -3 0 3 5 8 10delta tau (0.1 RED, 0.3 NIR ref.)

+ +

<-*> <(-*)/*> in %

0

20

40

60

80

100

13 14 15 16 17 18 19

July 3, 2002

MWRCimelMFRSRSZA

Time(UT)

courtesy of A. Trishchenko

from LANDSAT


How good is the retrieval?Simulations using stochastic cloud models

0

0.2

0.4

0.6

0.8

1

1.2

0 10 20 30 40 50 60 70

3D (red)

1D (red)

optical depth

3D (nir)

1D (nir)


Resultsscatter plots of “retr.” vs. “true”

Total STATISTICS (Ac=81%)

(true) (retr)mean 13.0 13.1std 10.5 10.8

pixel-by-pixel error: 3.0


Cumulative error histograms

0

0.2

0.4

0.6

0.8

1

0 5 10 15 20

errors in 3

75% pixels have errors

< 3


SummaryThe Main Ideas

• Cimel measures zenith radiance when the Sun is blocked by clouds;

• use two wavelengths:

one in RED (670 nm) [or in BLUE (440 nm)], where vegetation albedo is low, and

one in NIR (870 nm), where vegetation albedo is high

• retrieve cloud optical depth (and cloud fraction) using the NIR vs. RED plane

The Results (so far)

• looks promising; it largely removes 3D effects;

• it is not the final answer but a big improvement against single-wavelength retrievals;

• can fill (cloud) gaps in AERONET aerosol optical depth retrievals and estimate (effective) cloud fraction;

• version 0 cloud optical depth product will be soon available for distribution


Can AERONET help with monitoring clouds?

Alexander MarshakNASA/GSFC

Thanks to:

Y. Knyazikhin, K. Evans, W. Wiscombe,

I. Slutsker and B. Holben

Supported by:

NASA Radiation Science Program


Dates with the cloud mode data

Bondville: July 2001 - presentCartel: July 2001 - May 2002Egbert: July 2001 - Jan. 2002GSFC: Dec. 2001 - Dec. 2002Harvard Forest: July 2001 - Dec. 2001SGP CART: Oct. 2001 - March 2004Shirahama: Oct. 2001 - May 2002UCSB: May 2003 - present


NDCINormalized Difference Cloud Index

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

"18:59:39" "19:07:09" "19:14:39"

NDCI

NDCI

time

24 May 1999

Clear skyBroken cloudsCloud

€

NDCI =I(nir) − I(red)

I(nir) + I(red)

0.0

5.0

10.0

15.0

20.0

25.0

30.0

"18:59:39" "19:07:09" "19:14:39"

1020 nm870 nm670 nm440 nm

normalized radiance (arbitrary units)

time (in GMT)

24 May 1999

Clear sky

BrokenClouds

OvercastSky


1D Calculations

€

NDCI =I(nir) - I(red)

I(nir) +I(red)=

1- tanα

1+tanα

where

tanα = I(red) I(nir) on NIR vs.RED plane

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.05

0.1

0.15

0.2

0.25

0 5 10 15 20 25 30 35 40

zenith radiance NDCI

optical depth

I(vis)

I(nir)NDCI

€

vis ≡ red = 0.67μm nir = 0.87μm


0.0

5.0

10.0

15.0

20.0

25.0

11:06:40 16:40:00 22:13:20

Argentine (Cordoba-CETT)3/22/00

1020870670440

normalized radiance

time(GMT)

7.0

8.0

9.0

10.0

11.0

12.0

13.0

14.0

13:59:40 14:01:20

Argentine (Cordoba-CETT)3/22/00

1020870670440

normalized radiance

time (GMT)

Cloud observationsNormalized radiance, NDCI and cloud optical depth

zoom

1D retrieval

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.11

13:59:20 14:00:40

NDCI

ndci

time (GMT)

0

5

10

15

14:00:00 14:01:00

cloud optical depthretrieved from NDCI

optical depth

time (GMT)


How good is the retrieval?Surface Albedo at the ARM SGP site

from Li et al., JGR 2002

+

from Michalski et al., 2002

Pilewskie’s SSFR within 0.0001 rad (640 m) from the

CART site flying at 300 m on April 5,

2000

courtesy of A. Trishchenko

from LANDSAT


Local Climatology( a few other locations)

0

200

400

600

800

1000

0 20 40 60 80 100

Egbert, CAN

tau

0

200

400

600

800

1000

1200

1400

0 20 40 60 80 100

Cartel, CAN

tau 0

500

1000

1500

2000

2500

3000

3500

0 0.2 0.4 0.6 0.8 1

Cartel

tau

0

500

1000

1500

2000

2500

3000

3500

0 0.2 0.4 0.6 0.8 1

Egbert, CAN

Ac

optical depth(effective)

cloud fraction

Egbert (2001)

Cartel (2001-2002)


Encouraging?

LWP comparisons for March 14th and 15th

Comparison of the Miller microbase algorithm for LWP (based on combining microwave and radar data) against all other volunteered results, for two days in the ARM 2000 Cloud IOP at the ARM SGP site. The cloud was very uniform and stratiform on both days.

20:44

courtesy of Dave Turner, PNNL

can aeronet help with monitoring clouds?

Documents

cloud properties

optical depthac

cloud iop

low optical depthsover

surface retrievals

arm sgp site

arm sitesurface inhomogeneity

ired t