Direct aerosol radiative effects based on combined A-Train observations
Jens Redemann, Y. Shinozuka, J. Livingston, M. Vaughan, P. Russell, M.Kacenelenbogen, O. Torres, L. Remer
BAERI – NASA Langley - NASA Ames – SRI – NASA Goddard
http://geo.arc.nasa.gov/AATS-website/
email: [email protected]
Outline Goal: To devise a new, methodology to derive direct aerosol radiative
effects - Faerosol(z) based on CALIOP, OMI and MODIS
Motivation, data sets and role of field observations
Methodology for combining CALIOP, OMI and MODIS data
Checking consistency of input data
Proof of concept for 4-month data set – Jan., Apr., Jul., Oct. 2007
Impact of input data sets
Comparisons to AERONET and CERES flux products
Conclusions
MODIS
OMI
CALIOP
Goal: To use A-Train aerosol obs to constrain aerosol radiative properties to calculate Faerosol(z)
Myhre, Science, July 10, 2009: 1)Observation-based methods too large2)Models show great divergence in regional and vertical distribution of DARF.3)“remaining uncertainty (in DARF) is probably related to the vertical profiles of the aerosols and their location in relation to clouds”.
Target:Faerosol(z) + Faerosol(z)
Constraints/Input:- MODIS AOD (7/2 ) + AOD- OMI AAOD (388 nm) + AAOD- CALIPSO ext (532, 1064 nm) + ext- CALIPSO back (532 , 1064 nm) + back
Goal: To use A-Train aerosol obs to constrain aerosol radiative properties to calculate Faerosol(z)
Retrieval:ext (, z) + extssa (, z) + ssa
g (, z) + g
MODIS aerosol models: 7 fine and 3 coarse mode distribution models define size and refractive indices of bi-modal log-normal size distribution → 100 combinationsFree parameters: Nfine, Ncoarse
Issues to consider- Differences in data quality land/ocean- Impact of model assumptions- Spatial variability- Aerosols above & near clouds
Rtx code
Methodology: Role of field observationsUse suborbital observations to:
1) Guide choices in aerosol models
2)Test retrievals of aerosol radiative properties
3)Test calculated radiative fluxes
4)Study spatial variability = uncertainty involved in extrapolating to data-sparse regions (e.g., above clouds)
See poster A43A-0127 Kacenelenbogen et al.
Re 4): Shinozuka and Redemann, ACP, 2011
Solution space: expansion from over-ocean MODIS models
ARCTAS data are corrected after Virkkula [2010].
Role of suborbital observations: 1) Test realism of aerosol models
7 fine + 3 coarse modesSSA and EAE Lidar Ratio and EAE
ix̂ix : retrieved parameters
: observables
: uncertainties in obs.
: weighting factors
Current choices in retrieval method:1)Metric / error / cost function
2)4 Observablesxi = AOD 550nm (±0.03±5%)
AOD 1240 nm (±0.03±5%) - MODIS AAOD 388 nm ±(0.05+30%) -
OMI 532 ±(0.1Mm-1sr-1+30%), -
CALIOP
3)Minimize X and select the top 3% of solutions that meet for all i
2/12
ˆ
ˆ
i
ii
ii x
xxw
ix̂
iw
iii xxx ˆ|ˆ|
Example of successful retrieval from actual collocated MODIS, OMI, CALIOP (V3) data: Oct. 23, 2007
Consistency issues: AOD comparisons (CALIOP V3)
Eight months of data: January, April, July and October 2007 and 2009
Use CALIOP 5/40km-avg. (V3/V2) aerosol extinction profiles, and 5km
aerosol and cloud layer products
Find all instantaneously collocated, MODIS MYD04_L2 (10x10km)
aerosol retrievals traversed by 5km/40km CALIPSO track
Judicious use of data quality flags
Break down geographically → zonal mean AOD
See Redemann et al., ACPD for details
See Kacenelenbogen et al., ACP 2011
for potential explanations for CALIOP-MODIS
differences
Latitudinal distribution of AOD differences between MODIS and CALIOP V3
MODIS-CALIOP AOD
MODIS-CALIOP AOD
Latit
ude
Latit
ude
Redemann et al., ACPD
Main findings, ocean:1.bias differences of 0.03 - 0.04 (with CALIOP<MODIS for all months), 2.RMS of 0.09 – 0.12 3.r2 is ~ 0.4-0.5 …all after judicious use of quality flags
Shinozuka and Redemann, ACP, 2011
OMAERUV (Torres group) OMAERO (KNMI group)
AOD 380nm AOD 380nm
ssa 380nm ssa 380nm
Consistency issues: Choice of OMI data
OMAERO data collocated with MODIS and CALIOP is a reasonable representation of global OMAERO data
OMAERUV data collocated with MODIS and CALIOP is a poor representation of global OMAERUV
OMAERO data collocated with MODIS and CALIOP is a reasonable representation of global OMAERO data
OMAERUV data collocated with MODIS and CALIOP is a poor representation of global OMAERUV
Target:Faerosol(z) + Faerosol(z)
Constraints/Input:- MODIS AOD (7/2 ) + AOD- OMI AAOD (388 nm) + AAOD- CALIPSO ext (532, 1064 nm) + ext- CALIPSO back (532 , 1064 nm) + back
Retrieval of aerosol radiative properties from A-Train observations
Retrieval:ext (, z) + extssa (, z) + ssa
g (, z) + g
MODIS aerosol models: 7 fine and 3 coarse mode distributions define standard deviation and refractive indices of bi-modal log-normal size distribution → 100 combinationsFree parameters: Nfine, Ncoarse
Rtx code
10-2
10-1
100
101
102
0
0.5
1
1.5
2
2.5
Diameter (m)
dN /
dlo
gD
Number
10-2
10-1
100
101
102
0
5
10
15
20
Diameter (m)
dV /
dlo
gD
Volume
123456789
123456789
Target:Faerosol(z) + Faerosol(z)
Constraints/Input:- MODIS AOD (7/2 ) + AOD- OMI AAOD (388 nm) + AAOD- CALIPSO ext (532, 1064 nm) + ext- CALIPSO back (532 , 1064 nm) + back
Retrieval of aerosol radiative properties from A-Train observations
Retrieval:ext (, z) + extssa (, z) + ssa
g (, z) + g
MODIS aerosol models: 7 fine and 3 coarse mode distributions define standard deviation and refractive indices of bi-modal log-normal size distribution → 100 combinationsFree parameters: Nfine, Ncoarse
Rtx code
10-2
10-1
100
101
102
0
0.5
1
1.5
2
2.5
Diameter (m)
dN /
dlo
gD
Number
10-2
10-1
100
101
102
0
5
10
15
20
Diameter (m)
dV /
dlo
gD
Volume
123456789
123456789
Comparison:CERES Fclear
Airborne Fclear
Comparison:AERONET AOD, ssa, gAirborne test bed data
AERONET Inversion collocation with MODIS-OMI(OMAERO)-CALIPSO
AERONET V2.0 (circles): •~68,200 collocated MODIS-OMI-CALIPSO obs in 4 months of data•93-96% of those have valid MOC retrieval•576 of those have collocated AERONET AOD measurements (±100km, ±1h)•Only 45 of those have collocated AERONET SSA (V2/L2) retrieval
The top 3% X for all 4 months.
Xavg=0.1640 Xstd= 0.0968Xavg+Xstd=0.2608
AOD and SSA retrievals from MODIS – OMI - CALIPSO
AOD~540nm
SSA~540nm
24h avg direct radiative forcing from MODIS – OMI - CALIPSO
Surface
TOA
ConclusionsA. MODIS-CALIOP AOD comparisons: bias differences of 0.03 - 0.04 (CAL<MOD),
RMS differences of 0.09 – 0.12 after judicious use of quality flags.
B. A methodology for the retrieval of aerosol radiative properties from MODIS AOD, OMI AAOD and CALIOP 532 has been devised. Proof of concept study complete for January+April+July+October 2007:– Results sensitive to choice of OMI data (OMAERUV vs. OMAERO)– OMAERUV possibly more accurate, but data collocated with
MODIS+CALIOP not representative of global OMAERUV data set– Tests with collocated AERONET observations sparse– Tests with CERES irradiance measurements difficult to interpret
C. Next questions to answer:1) What is the trade-off between uncertainty and spatial sampling in input satellite data
sets for calculating aerosol-induced changes in TOA or surface fluxes, i.e., how much uncertainty is involved in spatial extrapolation to data-sparse regions?
2) How to test the “consistency” between various satellite input data sets?
3) How do uncertainties in satellite data propagate to retrievals and flux estimates?
D. Next steps to take:– Continue to investigate spatial variability in suborbital data to extend the MOC
retrievals to aerosol above cloud (AAC) based on reduced data set
– Extend study to use more MODIS channels and more OMI retrievals
– Constrain OMI AAOD retrievals with CALIOP height input
– Compare vertical distribution of direct forcing to models