calipso-inferred aerosol direct radiative e ects: bias ... aerosol direct radiative e ects: bias...
TRANSCRIPT
CALIPSO-inferred aerosol direct radiative effects:CALIPSO-inferred aerosol direct radiative effects:bias estimates using ground-based Raman lidarsbias estimates using ground-based Raman lidars
Tyler Thorsen1,2 and Qiang Fu2Tyler Thorsen1,2 and Qiang Fu2
1NASA Postdoctoral Program1NASA Postdoctoral Program2University of Washington2University of Washington
LANCE Rapid Response MODIS images: Aug 22, 2015
https://ntrs.nasa.gov/search.jsp?R=20160007832 2018-06-17T13:09:00+00:00Z
Introduction Method Lidar ratio Sensitivity
Aerosol direct radiative effect (DRE)
• The change in radiative flux caused by the presence of aerosols(both natural and anthropogenic)
• How aerosol affects the Earth’s radiation balance in the present climate• Estimation of aerosol radiative forcing (i.e. anthropogenic aerosols)
(Bellouin et al. Nature 2005, Kaufman GRL 2005, Su et al. JGR 2013)
CALIPSO aerosol DRE bias estimates (2/11)
Introduction Method Lidar ratio Sensitivity
Satellite estimates of aerosol DRE
• Many estimates of the shortwave (SW) aerosol DRE have been made using passiveremote sensors (Yu et al. ACP 2006 and references therein)
• Longwave aerosol DRE is usually much smaller• Mostly MODIS-based
• The global-mean SW aerosol DRE at the TOA is about −5.0 Wm−2
• The presence of aerosols increases the amount of reflected SW by 5.0 Wm−2
CALIPSO aerosol DRE bias estimates (3/11)
Introduction Method Lidar ratio Sensitivity
Satellite estimates of aerosol DRE
• Many estimates of the shortwave (SW) aerosol DRE have been made using passiveremote sensors (Yu et al. ACP 2006 and references therein)
• Longwave aerosol DRE is usually much smaller• Mostly MODIS-based
• The global-mean SW aerosol DRE at the TOA is about −5.0 Wm−2
• The presence of aerosols increases the amount of reflected SW by 5.0 Wm−2
CALIPSO aerosol DRE bias estimates (3/11)
Introduction Method Lidar ratio Sensitivity
“Global” estimates of aerosol DRE from passive sensors“Global” estimates of aerosol DRE from passive sensors
Often limited to daytime cloud-free oceanOften limited to daytime cloud-free ocean
Over land?Over land?
Over cloud?Over cloud?
Contamination by undetected cloud / cloud edgesContamination by undetected cloud / cloud edges
No vertical informationNo vertical information
CALIPSO aerosol DRE bias estimates (4/11)
Introduction Method Lidar ratio Sensitivity
“Global” estimates of aerosol DRE from passive sensors“Global” estimates of aerosol DRE from passive sensors
Often limited to daytime cloud-free oceanOften limited to daytime cloud-free ocean
Over land?Over land?
Over cloud?Over cloud?
Contamination by undetected cloud / cloud edgesContamination by undetected cloud / cloud edges
No vertical informationNo vertical information
CALIPSO aerosol DRE bias estimates (4/11)
Introduction Method Lidar ratio Sensitivity
“Global” estimates of aerosol DRE from passive sensors“Global” estimates of aerosol DRE from passive sensors
Often limited to daytime cloud-free oceanOften limited to daytime cloud-free ocean
Over land?Over land?
Over cloud?Over cloud?
Contamination by undetected cloud / cloud edgesContamination by undetected cloud / cloud edges
No vertical informationNo vertical information
CALIPSO aerosol DRE bias estimates (4/11)
Introduction Method Lidar ratio Sensitivity
“Global” estimates of aerosol DRE from passive sensors“Global” estimates of aerosol DRE from passive sensors
Often limited to daytime cloud-free oceanOften limited to daytime cloud-free ocean
Over land?Over land?
Over cloud?Over cloud?
Contamination by undetected cloud / cloud edgesContamination by undetected cloud / cloud edges
No vertical informationNo vertical information
CALIPSO aerosol DRE bias estimates (4/11)
Introduction Method Lidar ratio Sensitivity
“Global” estimates of aerosol DRE from passive sensors“Global” estimates of aerosol DRE from passive sensors
Often limited to daytime cloud-free oceanOften limited to daytime cloud-free ocean
Over land?Over land?
Over cloud?Over cloud?
Contamination by undetected cloud / cloud edgesContamination by undetected cloud / cloud edges
No vertical informationNo vertical information
CALIPSO aerosol DRE bias estimates (4/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO
• Vertically-resolved aerosol properties over allsurface types during both day and night
• Easier to separate cloud from aerosol in thesame profile
• Recent studies have made new estimates of theglobal-mean aerosol DRE using CALIPSO:
Clear-sky ocean All-sky global
Passive sensor-based −5.0 Wm−2 N/A
(Yu et al. ACP 2006)
CALIPSO-based −3.21 Wm−2 −0.61 Wm−2
(Oikawa et al. JGR 2013)
CALIPSO-based −2.6 Wm−2 −1.9 Wm−2
(Matus et al. JCLIM 2015)
Why are CALIPSO-based estimates significantly smaller in magnitude than the passivesensor-based ones?
CALIPSO aerosol DRE bias estimates (5/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO
• Vertically-resolved aerosol properties over allsurface types during both day and night
• Easier to separate cloud from aerosol in thesame profile
• Recent studies have made new estimates of theglobal-mean aerosol DRE using CALIPSO:
Clear-sky ocean All-sky global
Passive sensor-based −5.0 Wm−2 N/A
(Yu et al. ACP 2006)
CALIPSO-based −3.21 Wm−2 −0.61 Wm−2
(Oikawa et al. JGR 2013)
CALIPSO-based −2.6 Wm−2 −1.9 Wm−2
(Matus et al. JCLIM 2015)
Why are CALIPSO-based estimates significantly smaller in magnitude than the passivesensor-based ones?
CALIPSO aerosol DRE bias estimates (5/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO
• Vertically-resolved aerosol properties over allsurface types during both day and night
• Easier to separate cloud from aerosol in thesame profile
• Recent studies have made new estimates of theglobal-mean aerosol DRE using CALIPSO:
Clear-sky ocean All-sky global
Passive sensor-based −5.0 Wm−2 N/A
(Yu et al. ACP 2006)
CALIPSO-based −3.21 Wm−2 −0.61 Wm−2
(Oikawa et al. JGR 2013)
CALIPSO-based −2.6 Wm−2 −1.9 Wm−2
(Matus et al. JCLIM 2015)
Why are CALIPSO-based estimates significantly smaller in magnitude than the passivesensor-based ones?
CALIPSO aerosol DRE bias estimates (5/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO
1 Radiative flux → aerosol extinction →assumed lidar ratio (ratio of extinction-to-backscatter)
2 Is all radiatively-significant aerosoldetected? (Kacenelenbogen et al. 2014, Rogers et al. 2014,
Thorsen et al. 2015)
ARM Raman lidars (RL)
SGP
TWP Darwin
1 Direct extinction measurements(no critical assumptions)
2 Strong signals from aerosols (it’s closer)
CALIPSO aerosol DRE bias estimates (6/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO
1 Radiative flux → aerosol extinction →assumed lidar ratio (ratio of extinction-to-backscatter)
2 Is all radiatively-significant aerosoldetected? (Kacenelenbogen et al. 2014, Rogers et al. 2014,
Thorsen et al. 2015)
ARM Raman lidars (RL)
SGP
TWP Darwin
1 Direct extinction measurements(no critical assumptions)
2 Strong signals from aerosols (it’s closer)
CALIPSO aerosol DRE bias estimates (6/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO
1 Radiative flux → aerosol extinction →assumed lidar ratio (ratio of extinction-to-backscatter)
2 Is all radiatively-significant aerosoldetected? (Kacenelenbogen et al. 2014, Rogers et al. 2014,
Thorsen et al. 2015)
ARM Raman lidars (RL)
SGP
TWP Darwin
1 Direct extinction measurements(no critical assumptions)
2 Strong signals from aerosols (it’s closer)
CALIPSO aerosol DRE bias estimates (6/11)
Introduction Method Lidar ratio Sensitivity
Methodology
• Collocate (±200 km, ±2 hr) CALIPSO aerosol products (VFM, ALay) and ARMRL-FEX product over a 5 year period at SGP, 4 year period at TWP
• Calculate aerosol DRE using the NASA Langley Fu-Liou radiative transfer model:
DRE (TOA) = [F ↓(TOA)− F ↑(TOA)]aerosol − [F ↓(TOA)− F ↑(TOA)]no aerosol
DRE (SFC ) = [F ↓(SFC )− F ↑(SFC )]aerosol − [F ↓(SFC )− F ↑(SFC )]no aerosol
• *Modify RL retrievals to mimic CALIPSO to test the effect of¶ lidar ratio assumptions and· detection sensitivity
*Avoiding using the CALIPSO data directly because of wavelength difference between thelidars
¶ About +10% bias in the aerosol DRE due to the lidar ratio
CALIPSO aerosol DRE bias estimates (7/11)
Introduction Method Lidar ratio Sensitivity
Detection sensitivity
TWP(a)
Solid: allDashed: night
Dotted: day
RL-FEXCALIPSO
Aerosol occurrence (transparent profiles)0 0.2 0.4 0.6 0.8 1
Hei
ght [
km]
0
1
2
3
4
5
6
7
8
9
10SGP
(b)
0 0.2 0.4 0.6 0.8 1
Is this undetected aerosol radiatively-significant?
CALIPSO aerosol DRE bias estimates (8/11)
Introduction Method Lidar ratio Sensitivity
Detection sensitivity
TWP(a)
Solid: allDashed: night
Dotted: day
RL-FEXCALIPSO
Aerosol occurrence (transparent profiles)0 0.2 0.4 0.6 0.8 1
Hei
ght [
km]
0
1
2
3
4
5
6
7
8
9
10SGP
(b)
0 0.2 0.4 0.6 0.8 1
Is this undetected aerosol radiatively-significant?
CALIPSO aerosol DRE bias estimates (8/11)
Introduction Method Lidar ratio Sensitivity
Detection sensitivity
TWP(a)
Solid: allDashed: night
Dotted: day
RL-FEXCALIPSO
Aerosol occurrence (transparent profiles)0 0.2 0.4 0.6 0.8 1
Hei
ght [
km]
0
1
2
3
4
5
6
7
8
9
10SGP
(b)
0 0.2 0.4 0.6 0.8 1
Is this undetected aerosol radiatively-significant?
CALIPSO aerosol DRE bias estimates (8/11)
Introduction Method Lidar ratio Sensitivity
Effect of detection sensitivity
• Method to force RL aerosol occurrenceprofile to match CALIPSO’s byremoving aerosol in each collocatedoverpass.
• “RL-RM”: RL degraded to CALIPSO’ssensitivity
RL RL-RM
Aero
sol
DR
E [
Wm
-2]
-10
-8
-6
-4
-2
0
2
-6.87-6.87
τ=0.223
-4.87
∆=2.00
(-29%)
-4.87
∆=2.00
(-29%)
τ=0.161
TOA SW
TWP(a)
RL RL-RM
-10
-8
-6
-4
-2
0
2
-4.11-4.11
τ=0.202
-2.09
∆=2.02
(-49%)
-2.09
∆=2.02
(-49%)
τ=0.103
TOA SW
SGP(b)
RL RL-RM
Aero
sol
DR
E [
Wm
-2]
-10
-8
-6
-4
-2
0
2
-7.46-7.46
τ=0.223
-5.29
∆=2.17
(-29%)
-5.29
∆=2.17
(-29%)
τ=0.161
Surface SW
TWP(c)
RL RL-RM
-10
-8
-6
-4
-2
0
2
-6.67-6.67
τ=0.202
-3.36
∆=3.31
(-50%)
-3.36
∆=3.31
(-50%)
τ=0.103
Surface SW
SGP(d)
CALIPSO’s lack of sensitivity causes a significant reduction of 30–50% in the magnitudeof the aerosol DRE
CALIPSO aerosol DRE bias estimates (9/11)
Introduction Method Lidar ratio Sensitivity
Effect of detection sensitivity
• Method to force RL aerosol occurrenceprofile to match CALIPSO’s byremoving aerosol in each collocatedoverpass.
• “RL-RM”: RL degraded to CALIPSO’ssensitivity
RL RL-RM
Aero
sol
DR
E [
Wm
-2]
-10
-8
-6
-4
-2
0
2
-6.87-6.87
τ=0.223
-4.87
∆=2.00
(-29%)
-4.87
∆=2.00
(-29%)
τ=0.161
TOA SW
TWP(a)
RL RL-RM
-10
-8
-6
-4
-2
0
2
-4.11-4.11
τ=0.202
-2.09
∆=2.02
(-49%)
-2.09
∆=2.02
(-49%)
τ=0.103
TOA SW
SGP(b)
RL RL-RM
Aero
sol
DR
E [
Wm
-2]
-10
-8
-6
-4
-2
0
2
-7.46-7.46
τ=0.223
-5.29
∆=2.17
(-29%)
-5.29
∆=2.17
(-29%)
τ=0.161
Surface SW
TWP(c)
RL RL-RM
-10
-8
-6
-4
-2
0
2
-6.67-6.67
τ=0.202
-3.36
∆=3.31
(-50%)
-3.36
∆=3.31
(-50%)
τ=0.103
Surface SW
SGP(d)
CALIPSO’s lack of sensitivity causes a significant reduction of 30–50% in the magnitudeof the aerosol DRE
CALIPSO aerosol DRE bias estimates (9/11)
Introduction Method Lidar ratio Sensitivity
Effect of detection sensitivity
• Method to force RL aerosol occurrenceprofile to match CALIPSO’s byremoving aerosol in each collocatedoverpass.
• “RL-RM”: RL degraded to CALIPSO’ssensitivity
RL RL-RM
Aero
sol
DR
E [
Wm
-2]
-10
-8
-6
-4
-2
0
2
-6.87-6.87
τ=0.223
-4.87
∆=2.00
(-29%)
-4.87
∆=2.00
(-29%)
τ=0.161
TOA SW
TWP(a)
RL RL-RM
-10
-8
-6
-4
-2
0
2
-4.11-4.11
τ=0.202
-2.09
∆=2.02
(-49%)
-2.09
∆=2.02
(-49%)
τ=0.103
TOA SW
SGP(b)
RL RL-RM
Aero
sol
DR
E [
Wm
-2]
-10
-8
-6
-4
-2
0
2
-7.46-7.46
τ=0.223
-5.29
∆=2.17
(-29%)
-5.29
∆=2.17
(-29%)
τ=0.161
Surface SW
TWP(c)
RL RL-RM
-10
-8
-6
-4
-2
0
2
-6.67-6.67
τ=0.202
-3.36
∆=3.31
(-50%)
-3.36
∆=3.31
(-50%)
τ=0.103
Surface SW
SGP(d)
CALIPSO’s lack of sensitivity causes a significant reduction of 30–50% in the magnitudeof the aerosol DRE
CALIPSO aerosol DRE bias estimates (9/11)
Introduction Method Lidar ratio Sensitivity
Global implications
• Aerosol that goes undetected is consistent with random noise considerations• CALIPSO’s SNR is too low to detect all aerosol during both day and night.
• Even for large aerosol optical depths,the bias remains significant
• The global mean ocean AOD asmeasured by CALIPSO is 0.09(Winker et al., 2013)
• AOD=0.09 → -35% to -50% aerosolDRE bias at the two ARM sites
Clear-sky ocean
Passive sensor-based −5.0 Wm−2
(Yu et al. ACP 2006)
CALIPSO-based −3.21 Wm−2 (-36%)
(Oikawa et al. JGR 2013)
CALIPSO-based −2.6 Wm−2 (-48%)
(Matus et al. JCLIM 2015)
CALIPSO aerosol DRE bias estimates (10/11)
Introduction Method Lidar ratio Sensitivity
Global implications
• Aerosol that goes undetected is consistent with random noise considerations• CALIPSO’s SNR is too low to detect all aerosol during both day and night.
• Even for large aerosol optical depths,the bias remains significant
• The global mean ocean AOD asmeasured by CALIPSO is 0.09(Winker et al., 2013)
• AOD=0.09 → -35% to -50% aerosolDRE bias at the two ARM sites
Clear-sky ocean
Passive sensor-based −5.0 Wm−2
(Yu et al. ACP 2006)
CALIPSO-based −3.21 Wm−2 (-36%)
(Oikawa et al. JGR 2013)
CALIPSO-based −2.6 Wm−2 (-48%)
(Matus et al. JCLIM 2015)
CALIPSO aerosol DRE bias estimates (10/11)
Introduction Method Lidar ratio Sensitivity
Global implications
• Aerosol that goes undetected is consistent with random noise considerations• CALIPSO’s SNR is too low to detect all aerosol during both day and night.
• Even for large aerosol optical depths,the bias remains significant
• The global mean ocean AOD asmeasured by CALIPSO is 0.09(Winker et al., 2013)
• AOD=0.09 → -35% to -50% aerosolDRE bias at the two ARM sites
Clear-sky ocean
Passive sensor-based −5.0 Wm−2
(Yu et al. ACP 2006)
CALIPSO-based −3.21 Wm−2 (-36%)
(Oikawa et al. JGR 2013)
CALIPSO-based −2.6 Wm−2 (-48%)
(Matus et al. JCLIM 2015)
CALIPSO aerosol DRE bias estimates (10/11)
Introduction Method Lidar ratio Sensitivity
Conclusions
• The results presented here strongly suggest that newer estimates of the globalaerosol DRE that rely solely on CALIPSO aerosol observations (Oikawa et al. JGR2013); Matus et al. JCLIM 2015) are biased weak (i.e. too small in magnitude).
• This study demonstrates that our knowledge of the global aerosol DRE remainsincomplete.
• While CALIPSO allows for more consistent global estimates of the aerosol DRE in allscene types, its detection sensitivity is likely not sufficient for detecting allradiatively-significant aerosol.
• Passive sensors outperform CALIPSO in observing thin AOD since CALIPSO issensitive to the backscatter in a relatively small volume while passive sensorsmeasure the vertically-integrated scattering.
• However, the limitation of accurate passive retrievals to cloud-free ocean as well aspotential biases from cloud contamination makes fully and accurately assessingglobal aerosol DRE difficult.
We don’t know the global aerosol DRE
CALIPSO-inferred aerosol direct radiative effects: Bias estimates using ground-basedRaman lidars; TJ Thorsen, Q Fu; Journal of Geophysical Research, 2015.CALIPSO aerosol DRE bias estimates (11/11)
Introduction Method Lidar ratio Sensitivity
Effect of assumed lidar ratios
• CALIPSO’s processing:Detect → cloud/aerosol → 6 aerosol subtypes → lidar ratio → extinction → flux
• The wavelength difference betweenCALIPSO (532 nm) and RL (355 nm)precludes a direct assessment ofCALIPSO’s lidar ratios. Instead theaerosol DRE is computed with¶ Directly retrieved RL extinction
· Lidar ratio fixed (climatology±bias)
• If the selection of lidar ratio byCALIPSO can reproduce theclimatological value at a particularlocation, then the aerosol DRE can beaccurately calculated. Climo lidar ratio bias [%]
-80 -60 -40 -20 0 20 40 60 80
Aero
sol
DR
E b
ias
[%]
-60
-50
-40
-30
-20
-10
0
10
20
30TWP TOA
TWP Surface
SGP TOA
SGP Surface
• Rogers et al. AMT (2014) found approximately a +20% bias in CALIPSO’s lidarratio which would correspond to about +10% bias in the aerosol DRE.
CALIPSO aerosol DRE bias estimates (12/11)
Introduction Method Lidar ratio Sensitivity
0
1
2
3
4
5
6 (a)TWPDay
Night
Solid: RL
Dashed: RL-RM
PD
F [
km
]
0
1
2
3
4
5
6 (b)SGPDay
Night
Solid: RL
Dashed: RL-RM
Extinction coefficient [1/km]
10-3
10-2
10-1
100
0
2
4
6
8
10 (c)UndetectedDay
Night Dotted-dashed: TWP
Dotted: SGP
CALIPSO aerosol DRE bias estimates (13/11)
Introduction Method Lidar ratio Sensitivity
CALIPSO aerosol layer classifications
Counts (thousands)0 1 2 3 4
Marine
Dust
Polluted continental
Clean continental
Polluted dust
Smoke
TWP
(a)
0 3 6 9 12 15 18
SGP
(b)
CALIPSO aerosol DRE bias estimates (14/11)
Introduction Method Lidar ratio Sensitivity
0 30 60 90 120 1500.0
0.5
1.0
1.5
2.0(a) Aerosol
Solid: TWP
Dashed: SGP
0 4 8 12 16 200
3
6
9
12
15(b) Rain
0 10 20 30 40 50 60
Fre
quen
cy [
%]
0
2
4
6
8
10
12(c) Liquid
0 10 20 30 40 50 600
2
4
6
8
10(d) Ice
Lidar ratio [sr]
0 4 8 12 16 200
3
6
9
12
15(e) HOI
CALIPSO aerosol DRE bias estimates (15/11)
Introduction Method Lidar ratio Sensitivity
N = 2303
slope = 0.97
r = 0.91
RMS = 26.2%
bias = -4.3%
(a) TWP: all, 10min
10-2
10-1
100
RL
-FE
X a
ero
sol
op
tica
l d
epth
N = 19403
slope = 1.01
r = 0.89
RMS = 35.8%
bias = -0.3%
(b) SGP: all, 10min
Sun photometer aerosol optical depth
10-2
10-1
100
10-2
10-1
100
Fre
qu
ency
[%
]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
CALIPSO aerosol DRE bias estimates (16/11)