Transcript Global transport and radiative forcing of biomass burning

Quantifying aerosol direct
radiative effect with MISR
observations
Yang Chen, Qinbin Li, Ralph Kahn
Jet Propulsion Laboratory
California Institute of Technology, Pasadena, CA
California Institute of Technology, June 05, 2007
Overview

Aerosol direct radiative effect (Scattering, Absorption)
• Aerosol forcing (anthropogenic)
• Aerosol radiative effect (natural + anthropogenic)

Method to estimate global aerosol direct radiative effect
• Model-based, radiative transfer calculation
• Satellite-based
• -0.1~-0.9 W/m2 for annual mean global aerosol direct forcing
(IPCC,2007).

Previous satellite based studies
• CERES (Clouds and the Earth’s Radiant Energy System)  TOA fluxes
! Coarse resolution (~10x10 km2)
• MODIS (Moderate Resolution Imaging Spectroradiometer)  Aerosols
! Has difficulty in retrieving aerosol properties over bright land

This study
• Use MISR (Multi-angle Imaging SpectroRadiometer) to quantify global
aerosol shortwave direct radiative effect (SWDRE), on both land and
oceans.
Introduction to MISR (Multi-angle
Imaging SpectroRadiometer)

On board satellite TERRA

9 view angles at earth
surface:
•

Four spectral bands at each
angle:
•
•
•
•

446
558
672
866
nm
nm
nm
nm
(Blue)
(Green)
(Red)
(NIR)
Global Mode:
•
•
•
•

±70.5º. ±60.0º, ±45.6º, ±26.1º,
nadir
275 m sampling resolution for
nadir camera and red band of
other cameras
1.1 km for the other channels
400-km swath
Global coverage: 9 days at
equator, 2 days at poles
Continuous data retrieval
since Feb 2000.
MISR products and images
MISR products used
TOA albedo
MISR images
Nadir view
Cloud mask
(2.2x2.2 km2)
AOD
(17.6x17.6 km2)
AOD
TOA albedo
Cloud mask
(1.1x1.1 km2)
BHRPAR
(1.1x1.1 km2)
1°x 1° grid
Method
Fclearsky  ITOA  with_ aerosol  no _ aerosol 
Fallsky  Fclearsky  1 Cloud _ fraction
Global distribution of MISR AOD,
albedo, and BHRPAR (July, 2002)
26 BHRPAR bins:
0~0.1: each 0.01 interval
0.1~0.4: each 0.02 interval
Above 0.4: 1 level
Global TOA albedo~AOD correlation
over ocean
List of selected regions
Regions over land
Regions over ocean
Each region has 10°x5° area
TOA albedo~AOD correlation over
ocean regions
The slopes indicate the ability of aerosols to affect TOA radiative flux.
TOA albedo~AOD correlation
over remote ocean regions
Global regression over
ocean for each SZA
Alternative method:
do global regression
for each solar zenith
angle.
Global TOA albedo~AOD
correlation over land
List of selected regions
Regions over land
Regions over ocean
Each region has 10°x5° area
Albedo~AOD correlation over land
A
East US
East US
Albedo~AOD correlation over land
A
East US
Central Africa
Albedo~AOD correlation over land
A
East US
Sahara desert
Aerosol direct radiative effect
(a) Clear-sky and (b) all-sky aerosol direct radiative effect (W/m2) for
July 2002.
Aerosol direct radiative effect
From this study (July, 2002):
Aerosol DRE (Clear sky)
(W/m2)
Aerosol DRE (All sky)
(W/m2)
Global
-4.70
-1.49
Over ocean
-4.54
-1.95
Over land
-4.88
-1.18
From previous satellite-based studies:
Source
Aerosol DRE (W/m2)
Spatial coverage
Temporal coverage
Satellite data source
Zhang and
Christopher, 2005
-6.4 ± 2.6
Cloud-free
oceans
09/2000-08/2001
CERES, MODIS
Christopher and
Zhang, 2002
-6
Cloud-free
oceans
09/2000
CERES, MODIS
Loeb and Kato,
2002
-4.6 ± 1
Cloud-free
tropical oceans
01/1998-08/1998,
03/2000
CERES, TRMM VIRS
Loeb and ManaloSmith, 2005
-5.5, -3.8
Cloud-free
oceans
03/2000-12/2003
CERES, MODIS
-2.0, -1.6
All-sky oceans
Correlation between aerosol
SWDRE and 0.56 mm AOD
Correlation between aerosol SWDRE and AOD (a) Over ocean (b) Over
land
Uncertainties

Satellite retrieval of aerosol, TOA albedo and
surface properties.

Cloud contamination.

Diurnal variability.

TOA albedo narrow-to-broadband conversion.

Surface heterogeneity.
Use SVM classifier to calculate
smoke aerosol effect - method

SVM: Support Vector Machine

SVM classifiers
Clouds
• Clouds
• Aerosols



Smoke
Dust
Other
• Ice/Snow
• Water
• Land
Land
Water
Fsmoke _ aerosols  ITOA  smoke   aerosol _ free 
Smoke
Dust
Use SVM classifier to calculate
smoke aerosol effect - result
(a) Clear-sky aerosol SWDRE
from smoke aerosols (W/m2)
(b) MODIS wild fire occurrence
from Fire Information for Resource
Management System (NASA/U of
Maryland, 2002)
Use SVM classifier to calculate
smoke aerosol effect - uncertainty



Threshold for differentiating ‘Aerosols’ and ‘Surface’
pixels is arbitrary, which may cause the underestimation
of total number of ‘smoky’ pixels.
Since many ‘Surface’ pixels actually have some aerosol
loading, the albedo for ‘Surface’ pixels is highly
overestimated.
The SVM scene classification is still in provisional quality,
and the aerosol sub-classification validation has not yet
been completed.
Due to the above reasons, we chose not to use SVM scene
classifiers in the global aerosol SWDRE estimation.
Conclusions and future work

Conclusions
• By using MISR datasets, first satellite-based attempt to
estimate global aerosol direct radiative effect over both
ocean and land has been made.
• Aerosols and TOA albedo show different correlations in
absorptive aerosol dominated region and non-absorptive
aerosol dominated region.
• Over land, the slope of AOD ~ TOA albedo decreases as
BHRPAR increases, indicating the aerosol scattering and
absorbing effect on TOA albedo is smaller over brighter
surfaces.

Future work
• Extend the approach to include seasonal and interannual variability.
Acknowledgment

MISR data were obtained from the NASA Langley Atmospheric
Sciences Data Center (http://eosweb.larc.nasa.gov/).