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%