the operational modis aerosol products - nasa€¦ · the global aerosol mod08_d3 daily level 3 1...

The Operational MODISAerosol Products

L.A. Remer, Y. J. Kaufman, D. Tanré

And the MODIS aerosol team:D.A. Chu, C. Ichoku, R. Kleidman, I. Koren, R. Levy,

R-R. Li, J.V. Martins, S. Mattoo

We’ve been extremely productive!

RGB AOT

FLUX Fraction fine Rong-Rong Li

CaliforniaWildfires

Oct. 26, 2003

From Terra-MODIS

MOD04

The global aerosol

MOD08_D3

Daily Level 31 degree data

October 26, 2003

Aerosol Optical Thickness

Fine mode fraction

Paul Hubankshttp//:modis-atmos.gsfc.nasa.gov

MOD08_M3

Monthly meanglobal data

1 degree grid

Oct. 2003

Optical Thickness

Fine mode fraction

Pixel CountsPaul Hubankshttp//:modis-atmos.gsfc.nasa.gov

Deriving aerosol properties over land and ocean

But, before we can begin to derive aerosol...

We need to find the right pixels.

���������� ���� � �� ��� ��������������

Cloud Masking

2.00

1.00

0.50

0

Figure 7. MODIS image from SouthAmerica, Sept. 14, 2002 showingheavy smoke in northwest corner.Bottom images show τ 0.66 retrievals.Bottom left uses operational cloudmask that prevents retrievals in heavy smoke. Bottom right usesspatial variability cloud mask thatmasks the cloud but allow retrievalsfor τ 0.66 > 1.50.

Spatial variability with anassociated test using

1.38 µm to find smooth cirrusis proving to be the best

method to separate aerosol from clouds over both

land and ocean.Martins et al. (2002)

Heavy Smoke

Example from the AmazonStandard Cloud mask Martins Cloud mask

Snow Masking

Li et al. (2005)

(ρ0.86- ρ1.24)/(ρ0.86 + ρ1.24)>0.01And T11 < 285K then SNOW

Sediment Masking

Li et al. (2003)

Spatial variabilityCloud maskInternal spectralSnow maskSediment mask

ocean land

now soon now soon

X X

X

X

Now = Collection 004Soon = Collection 005

� � � �� �� ����� ���� ��������������� � ������ ��!�� ��������������

"#� �$�� ��� ���%

cloud

watersnow

10 km

cloud

400 total- 56 water________

344- 24 snow________

320- 55 cloud_______

265-116 “bright”________

149 “good”

Discard brightest 50%and darkest 20% of the 149 good pixels.

44 pixelsRemer et al. (2004)

How to derive aerosol products from satellite(in 3 easy steps…)

1. Create a Look-Up Table with expected aerosol properties

2. Estimate surface reflectance (to separate signal fromthe atmosphere from signal from the ground).

3. Match the satellite-observed reflectances to the outputof the Look-Up Table

Urban/Industrial(0.96)

“smoke”- moderateabsorption (0.90)

Highly absorbing (0.85)Seasonally urban/industrial (0.96)

Highly absorbing (0.85)Seasonally moderate (0.90)

HighlyAbsorbing(0.85)

3 non-dust modelsplus dust

Set by geography andseason

Models are dynamic f(τ)

� ��� �������������� ����� ��� �&����� ��� ��������'������ �����������(�'�������)�� �� ���*

Step 1: Create a LUT with expected aerosol properties… LAND

Step 2: Estimate surface reflectance LAND

CurrentAssume ρs

2.1 = ρm2.1

Assume ρs0.47 = 0.25ρs

2.1Assume ρs

0.66 = 0.50ρs2.1

PlannedAssume ρs

0.47 ~ 0.35ρs2.1

Assume ρs0.66 ~ 0.60ρs

2.1

Step 3: Match the satellite-observed reflectances to the outputof the Look-Up Table LAND

Current:

Individual channel retrievals:0.47 µm and 0.66 µm

Fine model ratio = η =f(ρο

0.66/ρο0.47)

Remer et al. (2004)

Planned:

True inversion: 3 pieces of information =ρm

0.47 ρm0.66 ρm

2.1

will yield three quantities =τ 0.47, η, ρs

2.1

and for a given aerosol modelthe spectral dependencewill automatically yieldτ 0.55 and τ 0.66

No longer assumeρs

2.1 = ρm2.1

The Ocean Algorithm

Choice of 4 fine modesand 5 coarse modes

And 5 coarse modes

In order to minimize(ρmeas - ρLUT) over 6 wavelengths

0.0001

0.001

0.01

0.4 0.5 0.6 0.8 1 2

dry smoke reff

=0.10 µm

urban reff

=0.20 µm

wet reff

=0.25 µm

salt reff

=1 µm

dust reff

=1 µm

dust reff

=2.5 µm

Aer

osol

refl

ecta

nce

wavelength (µm)Remer et al. (2004)

+ ��� ����

oceanlandboth

AERONET sites

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

O CEAN660 nm

N = 2052

100 points50 points25 points15 points

MO

DIS

AO

T (

660

nm)

AERONET AOT (660 nm)

y = 0.008 + 0.95 x R = 0.92

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

LAND660 nm

N = 5906

300 points150 points75 points32 points

MO

DIS

AO

T (

660

nm)

AERONET AOT (660 nm)

y = 0.059 + 0.70 x R = 0.68

66% of MODIS aerosol retrievalsover ocean fall within expected uncertainty

71% of MODIS aerosol retrievalsover land fall within expected uncertainty

MODIS aerosol validation 2000-2002

ocean

land

Remer et al. (2004)

Ichoku et al. (2004)3 year validation data set Ichoku et al. (2004)3 year validation data set

Global Mean (MODIS minus AERONET) AOT550 difference

-0.2

-0.1

0

0.1

0.2

0.3

0.4

5 10 15 20 25 30 35 40 45 50 55 60 65

0 5 10 15 20 25 30 35 40 45 50 55 60Sensor Zenith Angle Bin Ranges (degrees)

Mea

n (M

OD

IS -

AE

RO

NE

T) A

OT5

50

Land_T003 Ocean_T003Land_T004 Ocean_T004Land_A003 Ocean_A003

AERONET Fine FractionO’Neill’s method from sun data

MO

DIS

Fin

e Fr

actio

n

Individualretrievals

Monthly means

Dubovik inversion

Over Ocean: Validating Size ParametersNon-sphericity of dust causes MODIS to under estimate size

(over estimate fine fraction.)

Remer et al. (2004) Kleidman et al. (2005)

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.1 0.2 0.3 0.4 0.5 0.6

16 quadrants of the Earth

Anmyon, KoreaE Asia

Capo Verde

Europe

WASHINGTONIndia

Pacific Ocean

S Africa

Aerosol optical thickness

more dust

more pollution

baselineoceanic aerosol

Yoram Kaufman

Monthly mean fine fraction 2001

τ ~ 0.2

τ ~ 0.1

The switch from Side Bto Side A electronics inJune 2001, creates a smallcalibration shift that affectsaerosol size parametersbut not optical thickness.

The effect is magnified whenwhen τ is low.

Applications

Direct and indirect radiative effects and forcing, Aerosol effects on clouds,

Air quality,Quantitative estimate of dust transport over ocean,

Estimates of biomass burning emissions at fire sources

Semi-direct Forcing(Heavy smoke in the Amazon

kills the clouds)

Koren et al. (2004) in Science

Clouds only (Solar)

Clouds, IR

Smoke only

Total – Semi direct effect

July 2002

AOT < 0.2

AOT > 0.2

Water Cloud fraction: AOT < 0.2

Water Cloud fraction: AOT > 0.2

Koren, Kaufman, Rosenfeld, Remer, Rudich

But over the Atlantic aerosols appear to increase cloudiness!

Future plans1. Dust nonsphericity2. True inversion for land retrievals3. Include polarization over land4. Evaluating and updating aerosol models and surface

assumptions.5. Better masking

Collection 0051. Spectral snow mask2. Land Cloud Mask3. Remove Flux products4. Corrected mistakes in the land retrieval over bright surfaces

Collection 005 and Beyond….

Acknowledgements

MODIS Aerosol Team: D.A. Chu, C. Ichoku, R. Kleidman, I. Koren, R. Levy, R-R. Li, J.V. Martins, S. Mattoo, Z. Ahmad

AERONET: B. Holben, O. Dubovik, T. Eck, D. Giles, I. Slutsker, A. Smirnov

AERONET PIs: (SIMBIOS) C. McLain, G. Fergion, C. Pietras

MODIS Atmospheres: M. King, P. Menzel , B-C. Gao, S. Ackerman, R. Frey , M. Gray,L. Gumley, P. Hubanks, R. Hucek, E. Moody, W. Ridgway, K. Strabala

MODIS Land/ Univ. of Maryland: E. Vermote

NOAA/NESDIS/ORA: A. Ignatov, X. Zhao