Size distributions from AERONET : Accuracy, Issues, Improvements

O. DubovikO. Dubovik11, B. N. Holben, B. N. Holben11, A. Smirnov, A. Smirnov1,1, T. F. EckT. F. Eck11, , T. LapyonokT. Lapyonok11, A. Sinyuk, A. Sinyuk11, M. Sorokin, M. Sorokin11, D. Tanre, D. Tanre2 2 , ,

P.GoloubP.Goloub22, I. Slutsker, I. Slutsker1 1 and D. Gilesand D. Giles11

1- Goddard Space Flight Center, NASA (AERONET)2- Université de Sci. et Tech. de Lille , France (PHOTON)

ObservationsObservationsNumerical inversion:Numerical inversion:-Accounting for noise-Solving Ill-posed problem- Setting a priori constraints

Forward model:Forward model:-Spectral and angular scattering by particles with different sizes, compositions and shapes- Accounting for multiple scattering in atmosphere

aerosol particle sizes,aerosol particle sizes,refractive index, refractive index,

single scattering single scattering albedoalbedo, etc.

((Dubovik Dubovik and King, JGR, 2000and King, JGR, 2000)

Single scattering: aerosol particles - homogeneous spheres

Particle Size Distribution:0.05 µm ≤ R (22 bins) ≤ 15 µm

Complex Refractive Index at λ = 0.44; 0.67; 0.87; 1.02 µm

0.01

0.02

0.04

0.05

0.07

0.1 1 10

dV/d

lnR

(µm

3 /µm

2 )

Radius (µm)

Multiple scattering:scalar radiative transfer with Lambertianground reflectance solved by DisOrds (Nakajima-Tanaka or Stamnes et al.)

INPUTINPUT of Forward Model

0.00

0.01

0.10

Wavelength (µm)0.44 0.67 0.87 1.02

Imarinary PartImaginary Part

1.35

1.40

1.45

1.50

1.55

1.60

Wavelength (µm)0.44 0.67 0.87 1.02

Real Part

0.01

0.1

1

10

100

1000

0 20 40 60 80 100 120 140

Almucantar Fitting

Inte

nsity

Scattering Angle (degree)

Int(0.44) * 1000

Int(0.67) * 100

Int(0.87) * 10

Int(1.02)

0

0.1

0.2

0.3

0.4

0.5

0.40 0.60 0.80 1.0

Fitting of optical thicknessin retrievals

MeasurementsFitting

Opt

ical

thic

knes

s

Wavelenths (micron)

0

0.05

0.1

0.15

0.2

0.25

0. 1 10

Retireved size distribution

Radius (microns)

µm3 /µ

m2 )

1.35

1.40

1.45

1.50

1.55

1.60

Wavelength ( µm)0.44 0.67 0.87 1.02

Real Part

0.00

0.01

0.10

Wavelength ( µm)0.44 0.67 0.87 1.02

Imarinary PartImaginary Part

Fitting as a retrieval strategy

Accuracy ???

Theoretical limitations

Forward model:- particle shapes: spheres, spheroids (shape retrieval ?)- particles are homogeneous (bi-components ?)- horizontal homogeneous- vertically homogenous aerosol or assumed profile of extinction (?)- assumed surface albedo or assumed BRDF - assumed gaseous absorption

Inversion assumptions:- smoothness constraints on size distribution- smoothness constraints of spectral dependence of ref. Index- log-normal distribution of random errors

Perspectives:- assuming bi-component aerosols- retrieval of BRDF from combination of AERONET with satellite and aircraft observations- retrieval of shape distribution

Measurement limitations

Measurement accuracies:

-- optical thickness: ~ 0.01optical thickness: ~ 0.01--0.020.02

-- skysky--channel calibration: ~ 5 %channel calibration: ~ 5 %

-- azimuth angle pointing: ~ 0.5azimuth angle pointing: ~ 0.500

-- degree of linear polarization:degree of linear polarization: ~ 1~ 1--2 % (?)2 % (?)

-- consistency between polarization and intensity: good (?)consistency between polarization and intensity: good (?)

-- cloud contamination: cloud contamination: almucantar almucantar (good), principle plane (???)(good), principle plane (???)

Geometry:

-- scattering angle coverage: ~ 3scattering angle coverage: ~ 30 0 (1(10 0 -- ?) ?) -- 15015000

-- spectral coverage: ~ 0.34 spectral coverage: ~ 0.34 -- 1.6 1.6 µµmm

Random ERRORS in AERONET Random ERRORS in AERONET retrievalsretrievals

ASSUMPTIONS:- measurements have Normal Noise:

-- optical thickness: optical thickness: σ = σ = 0.01 0.01

-- skysky--radiances: radiances: σ = σ = 5% 5%

0.00

20.00

40.00

60.00

80.00

0.1 1 10

τ(0.44) = 0.05τ(0.44) = 0.2τ(0.44) = 0.5τ(0.44) = 1.0

Erro

rs (%

)

Radius (µm)

CONCLUSIONS:- the retrievals stable- important tendenciesoutlined

Sensitivity to instrumental offsetsSensitivity to instrumental offsets

Offsets were considered in:

-- optical thickness: optical thickness:

-- skysky--channel calibration:channel calibration:

-- azimuth angle pointing: azimuth angle pointing:

-- assumed ground reflectance:assumed ground reflectance:

Aerosol models considered (bi (bi -- modal logmodal log--normal):normal):

-- WaterWater--soluble aerosol for 0.05 ≤ soluble aerosol for 0.05 ≤ ττ(440) ≤ 1;(440) ≤ 1;

-- Desert dust for 0.5 ≤ Desert dust for 0.5 ≤ ττ(440) ≤ 1;(440) ≤ 1;

-- Biomass burning for 0.5 ≤ Biomass burning for 0.5 ≤ ττ(440) ≤ 1;(440) ≤ 1;

∆τ λ( ) = ±0.01; ± 0.02;∆ I λ;Θ( ) I λ;Θ( ) 100% = ± 5%;∆φ = 0.5o ; 1o ;

∆A λ( )/ A λ( ) 100% = ± 30%; ± 50%;

Results summary:

-- ττ(440) ≤ 0.2(440) ≤ 0.2 -- ddVV//ddlnr lnr ((++), ), nn((λλ) () (--), ), kk((λλ) () (--), ), ωω00((λλ) () (--))-- ττ(440) > 0.2(440) > 0.2 -- ddVV//ddlnr lnr ((++), ), nn((λλ) () (++),), kk((λλ) () (++), ), ωω00((λλ) () (++))-- Angular pointing accuracy is critical for Angular pointing accuracy is critical for ddVV//ddlnr lnr of dustof dust

((++) ) CAN BECAN BE retrievedretrieved ((--) ) CAN NOT BECAN NOT BE retrievedretrieved

Sensitivity to forward model limitationsSensitivity to forward model limitationsMixed aerosols (inhomogeneous spherical aerosols):Mixed aerosols (inhomogeneous spherical aerosols):

-- ExternallyExternally mixed (mixed (nn(l) and (l) and kk(l) different for fine and coarse modes)(l) different for fine and coarse modes)

-- InternallyInternally mixed (mixed (nn(l) and (l) and kk(l) different for core and shell) (l) different for core and shell) -- Biomass BurningBiomass Burning

Non-spherical aerosols:-- SpheroidsSpheroids ((prolateprolate, axis ratio 2) , axis ratio 2) -- Desert dustDesert dust

Results summary:

-- ddVV//ddlnr lnr ((++), ), ωω00((λλ) () (++), ), nn((λλ) () (+,+,effectiveeffective), ), kk((λλ) () (+, +, effectiveeffective) )

(+) CAN BE retrieved (-) CAN NOT BE retrieved

Results summary:- dV/dlnr - coarse mode (+), fine mode (+, zenith angle < 25o)- ω0(λ) (+) - full solar almucantar (zenith angle ≥ 50o) - k(λ) (+)- n(440) (-), n(670) (-), n(870) (+/-), n(1020) (+)

Retrieval accuracy and limitationsSensitivity tests byDubovik et al. 2000

Real Part Imaginary PartSSA

τ(0.44) ≤ 0.20.05

80-100%0.05-0.07

τ(0.44) ≥ 0.50.02550%0.03

Size Distribution:

0.00

20.00

40.00

60.00

80.00

0.1 1 10

τ(0.44) = 0.05τ(0.44) = 0.2τ(0.44) = 0.5τ(0.44) = 1.0

Err

ors

(%)

Radius (µm)

bias ∆τ = ± 0.01Effective

Random errors Nonsphericitybiases

0

0.05

0.1

0.15

0.1 1 10

aureolefull almucantar

dV/d

lnR

(µm

3/µ

m2 )

Radius (µm)

1.30

1.35

1.40

1.45

1.50

1.55

1.60

Wavelength (µm)0.44 0.67 0.87 1.02

Real Part

wide angularcoverage

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.02

0.06

0.16

0.46

1.31

3.72

10.6

0.04

0.1

0.25

0.64

1.65

4.25

10.9

0.04

0.1

0.25

0.64

1.65

4.25

10.9

realretr/ext 1retr/ext2retr/ext3realfull<75<40<30realfull<75<40<30

external mixtures

internal mixture 1

internal mixture 2

error effects of particles mixing for different agular ranges: full almucantar, up to 75, 40 or 30 degrees

radius, µm

Almucantar: ττ((λλ), I(), I(λ,Θλ,Θ))λ= 0.38, 0.44, 0.5, 0.67,

0.87, 1.02, 1.64, µm

AERONET inversion scenarios

Principal Plane: ττ((λλ), I(), I(λ,Θλ,Θ))λ = 0.38, 0.44, 0.5, 0.67,

0.87, 1.02, 1.64, µm

Polarized Principal Plane: ττ((λλ), I(), I(λ,Θλ,Θ),P(),P(λ,Θλ,Θ))λ = 0.87µm

spheres

spheroids

Inversion Products: dVdV//dlndln(r(rii) )

n(n(λ)λ)k(k(λ)λ)

BRDFBRDFerrorserrors

satellite, aircraft, etc.

ω0(λ)P11(λ), P12(λ), ...

fine & coarsefine & coarsefluxes, …, …

AERONETAERONET model of aerosolmodel of aerosolspherical:spherical:

Randomly orientedRandomly orientedspheroids :spheroids :

((Mishchenko Mishchenko et al., 1997)et al., 1997)

spheroidspheroid kernels data basekernels data basefor for operational modeling !!!operational modeling !!!

Basic Model byBasic Model by Mishchenko Mishchenko et al. et al. 1997:1997:

randomly oriented homogeneous spheroids

ω(ε) - size independent shape distribution

τ λ( ),F11,...,F44 ≈ K ip ...;n;k( )i;p( )∑ ω p V ri( )

K - pre-computed kernel matrices:Input: n and k

Input: ωp (Np =11),

V(ri) (Ni =22 -26)

Output: τ(λ), ω0(λ), F11(Θ), F12(Θ),F22(Θ),F33(Θ),F34(Θ),F44(Θ)

Time:Time: < < one sec.one sec.Accuracy:Accuracy: < < 11--3 %3 %

Range of applicability:Range of applicability:0.15 ≤ 20.15 ≤ 2ππr/r/λ λ ≤ ≤ 280 280 (26 bins)(26 bins)

0.4 ≤ 0.4 ≤ ε ε ≤ 2.4 ≤ 2.4 (11 bins)(11 bins)1.33 ≤ n ≤ 1.6 1.33 ≤ n ≤ 1.6

0.0005 ≤ k ≤ 0.50.0005 ≤ k ≤ 0.50.1

1

10

100

0 40 80 120 160

Phase Functions (0.67 µm)

SpheresSpheroids - ω

1(ε)

Spheroids - ω1(ε)

Scattring Angle (degree)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.1 1 10

dV/d

lnR

(µm

3 /µm

2 )

Radius ( µm)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.1 1 10

dV/d

lnR

(µm

3 /µm

2 )

Radius ( µm)

Spheres Spheroids

Cape Verde (2001) dustCape Verde (2001) dustSize distributionsSize distributions

(110 cases; τ(1020) ≥ 0.3; α ≤ 0.6)

9 groups: τ = 0.39, 0.44, 0.48, 0.50, 0.52, 0.57, 0.60, 0.62,0.71

Comparison of “laboratory” Phase Functionwith typical AERONET retrieval

0.1

1

10

100

0 40 80 120 160

MeasurmentsAERONET retrieval

AScattering Angle (degrees)

Phas

e Fu

nctio

n (0

.44

µm)

Retrieval using combinations of up-looking Ground-based and down-looking satellite

observations

Retrieved:

Aerosol Properties:- size distribution- real ref. ind. - imag. ref. ind(AERONET sky channels)

Surface Parameters:-BRDF (MISR channels)-Albedo (MODIS IR channels)

AERONET-MISR

0.01

0.1

1

10

100

1000

104

105

0 20 40 60 80 100

Mongu, August 9, 2003

440 fit440 measurements670 fit*10670 measurements*10870 fit*100870 measurements*1001020 fit *10001020 measurements*1000

Ang

ular

Dis

trib

utio

n of

Rad

iatio

n (A

ERO

NET

)

Scattering Angle (degrees)

(residual=1.3%)

August 9, 2003τ(0.44) ~ 0.3

0.05

0.1

0.15

0.2

0.25

0.3

-80 -60 -40 -20 0 20 40 60 80

Mongu, August 9, 2003

440 fit440 measurements558 fit558 measurements670 fit670 measurements870 fit870 measurements

Ang

ular

Dis

trib

utio

n of

Rad

iatio

n (M

ISR

)

View Angle (degrees)

(residual=3%)

Simultaneous fitting

AERONET- POLDER June 24, 2003τ(0.44) ~ 0.26

Simultaneous fitting

0.01

0.1

1

10

100

1000

104

105

0 20 40 60 80 100

Mongu, June 24, 2003

440 fit440 measurements670 fit*10670 measurements*10870 fit*100870 measurements*1001020 fit *10001020 measurements*1000

Ang

ular

Dis

trib

utio

n of

Rad

iatio

n (A

ERO

NET

)

Scattering Angle (degrees)

(residual=1.5%)

0.05

0.1

0.15

0.2

0.25

80 100 120 140 160 180

Mongu, June 24, 2003

440 fit440 measurements670 fit 670 measurements870 fit870 measurements

Ang

ular

Dis

trib

utio

n of

Rad

iatio

n (P

OLD

ER)

Scattering Angle (degrees)

(residual=2.18%)

Comparisons of Surface Retrievals

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

MISR (August 9, 2003)POLDER (June 24, 2003)assumed

Surf

ace

Alb

edo

Wavelength (µm)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

-80 -60 -40 -20 0 20 40 60 80

MISR 0.44, Aug 9 2003, t=0.3, sza=40.7MISR_0.67MISR_0.87POLDER 0.44, June 24, 2003, =0.26, sza=46.97POLDER 0.67POLDER 0.87

BR

DF

View Angle (degrees)

POLDER: June 24, 2003τ(0.44) ~ 0.26 , SZA=470

MISR: August 9, 2003τ(0.44) ~ 0.3, SZA=400

Surface Effect on Retrievals of the Refractive Index (low aerosol loading)

POLDER: June 24, 2003τ(0.44) ~ 0.26 , SZA=470

MISR: August 9, 2003τ(0.44) ~ 0.3, SZA=400

1.36

1.4

1.44

1.48

1.52

1.56

1.6

1.64

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

Mongu, August 9, 2003

almucantor + MISRalmucantor,surface albedo retrievedalmucantor,surface albedo assumed

Rea

l Par

t of R

efra

ctiv

e In

dex

Wavelength (µm)

1.35

1.4

1.45

1.5

1.55

1.6

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2

Mongu, June 24, 2003

almucantor + POLDERalmucantor,surface albedo retrievedalmucantor,surface albedo assumed

Rea

l Par

t of R

efra

ctiv

e In

dex

Wavelength (µm)

Surface Effect on the Retrievals of the Size Distribution (low aerosol loading)

POLDER: June 24, 2003τ(0.44) ~ 0.26 , SZA=470

MISR: August 9, 2003τ(0.44) ~ 0.3, SZA=400

0

0.01

0.02

0.03

0.04

0.05

0.1 1 10

Mongu,August 9, 2003

almucantor + MISRalmucantor,surface albedo retrievedalmucantor,surface albedo assumed

dV/d

lnR

(µm

3 /µm

2 )

Particle Radius (µm)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.1 1 10

Mongu, June 24, 2003

almucantor + POLDER,res=1.5%almucantor,surface albedo retrieved,res=1.56%almucantor,surface albedo assumed,res=1.47%

dV/d

lnR

(µm

3 /µm

2 )

Particle Radius (µm)

Comparisons of Surface Retrievals (Albedo)

POLDER: September 27, 2003τ(0.44) ~ 0.24 , SZA=340

0

0.1

0.2

0.3

0.4

0.5

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

Solar Village,June 7,2003

spheroid model, retrievedsphere model, retrievedassumed

Surf

ace

Alb

edo

Wavelength (µm)

POLDER: June 7, 2003τ(0.44) ~ 0.67 , SZA=230

MISR: September 27, 2003τ(0.44) ~ 0.24 , SZA=310

0

0.1

0.2

0.3

0.4

0.5

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

spheroid model, POLDERsphere model, POLDERassumedspheroid model, MISRsphere model, MISR

Surf

ace

Alb

edo

Wavelength (µm)

Solar Village, September 27, 2003

Surface Effect on the Retrievals of Size Distribution

POLDER: September 27, 2003τ(0.44) ~ 0.24 , SZA=700

POLDER: June 7, 2003τ(0.44) ~ 0.67 , SZA=700

(Principle Plane with corrected surface)

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.01 0.1 1 10

Solar Village,September 27,2003

spheroid model, retrieved surface albedo spheroid modelsurface albedo assumedsphere model, retrieved surface albedo sphere modelsurface albedo assumed

dV/d

lnR

(µm

3 /µm

2 )

Particle Radius (µm)

0

0.05

0.1

0.15

0.2

0.25

0.01 0.1 1 10

Solar Village,June 7,2003

spheroid model, retrieved surface albedo spheroid modelsurface albedo assumedsphere model, retrieved surface albedo sphere modelsurface albedo assumed

dV/d

lnR

(µm

3 /µm

2 )

Particle Radius (µm)

Fitting Accuracy of Radiances Spheroids and Spheres (principle plane)

POLDER: September 27, 2003τ(0.44) ~ 0.24 , SZA=340

0.01

0.1

1

10

100

1000

104

0 20 40 60 80 100 120

Solar Village,September 27, 2003

440 fit440 measurements670 fit*10670 measurements*10870 fit*100870 measurements*1001020 fit *10001020 measurements*1000

Ang

ular

Dis

trib

utio

n of

Rad

iatio

n (A

ERO

NET

)

Scattering Angle (degrees)

Spheroidal model (residual=2.8%)

0.01

0.1

1

10

100

1000

104

0 20 40 60 80 100 120

Solar Village,September 27, 2003

440 fit440 measurements670 fit*10670 measurements*10870 fit*100870 measurements*1001020 fit *10001020 measurements*1000

Ang

ular

Dis

trib

utio

n of

Rad

iatio

n (A

ERO

NET

)

Scattering Angle (degrees)

Spherical model (residual=5.5%)

2.8% 5.5%Satellite: ~ 1-3%