Observations of 3D Radiative Effects in MODIS Cloud Optical Thickness Retrievals

1
Observations of 3D Radiative Effects in MODIS
Cloud Optical Thickness Retrievals

• Alexander Marshak and Tamás Várnai
• (UMBC/JCET and NASA/GSFC)

2
Outline
Introduction Importance of 3D effects Pre-MODIS
study of 3D effects LANDSAT, AVHRR,
POLDER Theoretical simulations of MODIS
images Uncertainties based on stochastic
models MODIS observations and products Statistica
l asymmetry Effect of view angles Effect of
sub-pixel variability Wavenumber
spectra Conclusions
3
What are 3D Radiative Effects?
Effects of neighboring pixels (e.g., shadowing)
Sub-pixel variability
1km MODIS pixel
1km MODIS pixel
4
Importance of understanding 3D effects
Based on a sample set of RT calculations carried
out at 250-m resolution for SZA60o
5
Some earlier (pre-MODIS) studies at high
resolution, diffusion smoothes images
60 km by 60 km Landsat scene
Smoothing scale (mfpcloud thickness)1/2 from
Marshak et al. (1995) and Davis et al. (1997)
from Cahalan and Snider (1989)
6
Earlier studies For oblique sun, clouds appear
too thick forward reflection is too low
Based on AVHRR data
Based on Polder data
from Loeb and Coakley (1998)
from Buriez et al. (2001)
7
How are 3D effects studied?
Theoretical simulations of horizontally
inhomogeneous clouds and application of 3D
radiative transfer
Statistical analysis of satellite data and
products
Combination of theoretical simulations with a
statistical analysis of satellite data.
8
Theoretical simulations
Test datasets are realizations of realistic
stochastic cloud models
9
Theoretical simulations (cont.)
MAS 35x35 km2 field of marine Sc measured at 50-m
resolution
Estimated error bounds for optical thickness
retrieved at 1-km resolution
10
MODIS data

• Examine MODIS data how frequently 3D effects
  influence the measured radiances and the
  retrieved cloud properties
• Optical thickness
• Particle effective radius
• Cloud water path

11
Statistical asymmetry in clouds
Cold Warm
Toward sun Away from sun Tfront gt Tbehind
Tfront lt Tbehind
12
Procedure for MODIS images
March Sept. 2001 NH SH, Latitude 45 120
images (107 pixels) 2000 km x 450 km Solar
zenith angle 40-60 1 km radiance, cloud
products All clouds (e.g., ice liquid)
13
Number of pixels
of illuminated and shadowed pixels in 50x50
km2 areas are statistically equal
14
Symmetry at 11 mm
So is IR brightness temperature
15
Asymmetry at 0.86 and 2.1 mm
Each dot corresponds to a 50x50 km2 area.
Averaged reflectancies over illuminated pixels
are plotted vs. shadowed ones. The ill. slopes
are much brighter than the shad. ones!
16
Effects on MODIS productst and reff
Comparison of mean optical depth, t, and mean
effective radius, reff, at the illuminated and
shadowed portion of 50 by 50 km areas 3D effects
have a strong influence!
17
Influence on cloud water path
Strong asymmetry in cloud water path
18
Histogram of absolute and relative asymmetries of
t
19
Viewing angles
pdf of t for each column
Some optical thicknesses, t
20
Viewing angles (cont.)
FWD 115o BWD 50o
21
Four bidirectional reflectances 3D simulations
qo60o jo0o
qv60o jv0o
qv60o jv90o
qv0o jv0o
qv60o jv180o
22
Influence of sub-pixel variability250 m vs. 1 km
mean1 km 11.7 mean250 m 11.8 std1 km
8.3 std250 m 8.8 std250 m-gt1 km 8.5
Cu clouds, cloud fraction 0.9, SZA33o
23
Estimate level and sign of uncertainties for
each pixel

• Asymmetry
• 1 km MODIS data
• Biases caused by neighboring pixels
• Combine with theoretical simulations
• Biases due to sub-pixel variability
• 250 m MODIS images

24
50 by 50 km example
()
51x51 km2 cloud fields over the North Atlantic
(1305 UTC 1 Nov 2000) for SZA75o. Cloud
fraction 0.78 and mean tau39.
25
Overall uncertainties for 50 by 50 km areas
26
Wavenumber spectrum
VIS Band 2
NIR Band 7
IR Band 11
27
Wavenumber spectrum (cont.)
MODIS, 250 m data 1024 x 1024 pxls
MISR, 275 m data Nadir view 1024 x 1024 pxls
28
Observations of 3D Radiative Effects in MODIS
Cloud Optical Thickness Retrievals

• Alexander Marshak and Tamás Várnai
• (UMBC/JCET and NASA/GSFC)