The Accuracy of Retrieved Cloud Properties Impacted by Systematic Error
by Leandra Merola
Mentor: Odele Coddington
Laboratory for Atmospheric and Space Physics, University of Colorado – Boulder
questions? – email me at lm681@bard.edu
Overview
For this project I looked at the accuracy of cloud optical
property retrievals, focusing on the biases induced by an
overlying aerosol layer above the cloud.
In order to measure clouds we use an inverse
problem solving method that compares modeled
data with measured to find the optical depth and
effective radii pair that matches the irradiance that
we have measured.
Measurements • Cloud retrievals are made by comparing measured and
modeled albedo (upwelling irradiance/downwelling
irradiance.)
• Solar Spectral Flux Radiometer (SSFR) is an instrument
aboard an airplane that measures.
Tools for Characterizing Cloud Retrievals
GENRA
(Generalized Nonlinear Retrieval Analysis)
GENRA is a statistical program that lets us study cloud retrievals
from many cloud types with and without systematic error in an
efficient way.
Retrieval wavelengths avoid
gas absorption bands.
Modeling • The forward model takes into account the physics of liquid
water absorption to create a look up table of cloud albedo
defined by a cloud optical thickness and effective radii pair.
•The forward model does not typically consider sources of
systematic error, such as aerosols, which can result in a biased
retrieval.
7
Max Info
6
5
1
1
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0
0
515 nm
pdf 0.5
745 nm
1015 nm
1240 nm
1625 nm
40
0.2
1015 nm
15
0.3
0
0
1500
2000
7
1240 nm
6
1625 nm
5
0.2
0.1
1000
Shannon Information Content for
Effective Radii
745 nm
0.4
0.1
500
Wavelength (nm)
515 nm
pdf 0.5
Scaled Albedo
2
True(tau,re) = (40,15 micron)
0.9
Albedo
3
Output with No Systematic Error
1
0.3
The measured data that I worked with came from the SSFR
instrument which is aircraft bound and measures upwelling and
downwelling irradiance. We use the albedo for our cloud retrievals.
Shannon Information Content for
Optical Thickness
4
0.4
“Knowledge of cloud properties, including their spatial and
temporal variability, is needed for understanding and
quantifying the role of clouds in climate variability and for
modeling clouds and their effects in climate and weather
models.” (Vukicevic, et al. 2010)
•A formal mathematical theory to
quantify the information gained
by making a measurement.
•It’s a scalar.
•Maximum information content is
dependent on resolution of the
look up table.
• Makes use of the pre-existing
look up table.
•Defines pdfs of measured and
modeled albedo.
•Aerosol impact is treated as a
systematic error (shift) in the
model pdf.
•Solution pdf is the expected
behavior in retrieved cloud
properties.
Statistical tool • Using a statistical program known as GENRA (Generalized
Nonlinear Retrieval Analysis), it was possible to characterize
the bias induced by an overlying absorbing aerosol layer
above the cloud for different cloud types in an efficient way.
•Future research includes using GENRA to study other biases
such as 3D cloud effects.
Why Study Clouds?
Shannon Information Content
PDFs
Max Info
4
3
0
20
Bias – The Ultimate Enemy
40
60
30
Scaled Albedo
1
Effective Radius
0
0
500
1000
1500
2000
Wavelength (nm)
0.9
0.8
0.8
0.7
0.7
0.6
0.6
515 nm
745 nm
1015 nm
515 nm
pdf 0.5
745 nm
1015 nm
0.4
1240 nm
0.3
37
0.2
1625 nm
1625 nm
0.4
0.1
0
0
20
40
515 nm
15
0.3
0.1
0
1240 nm
pdf 0.5
745 nm
1015 nm
0.2
60
80
1240 nm
1625 nm
100
0
10
Optical Depth
20
30
Effective Radius
2 Wavelength Retrieval
(e.g. satellite retrievals)
1
Results
By looking at the Shannon
Information Content for the
output from the albedo and
comparing it to the output from
the scaled albedo we see that no
new information is gained.
0.9
0.8
0.8
0.7
0.7
0.6
0.6
pdf 0.5
pdf 0.5
0.4
15
0.4
0.3
0.3
The red stars
are points
taken from
actual SSFR
data when
aerosols were
present.
(Coddington,
et al. 2010)
0.2
35
0.2
0.1
0.1
0
0
0
0
20
40
60
80
10
20
30
Effective Radius
100
Optical Depth
24 Wavelength Retrieval
(e.g. future retrievals!)
1
1
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
pdf 0.5
pdf 0.5
37
0.4
Increased Optical Thickness
Solid line
20
True(tau,re) = (40,15 micron)
0.9
Increased Effective Radii
Dashed line
10
1
1
This graph shows a
snap shot of a 2
wavelength retrieval
for many cloud types.
As the lines become
orthogonal they
become independent
so that one property
can have bias when
the other does not.
0
Output with Systematic Error
0.9
Here you can see that
an aerosol changes the
albedo, which would
change the optical
properties retrieved,
giving us a biased view
of the cloud.
100
Optical Depth
1
• Overlying absorbing aerosols reduce cloud albedo.
• These aerosols bias the cloud retrieval giving us
inaccurate information about the cloud, which could
be confused with the indirect aerosol effect.
• Aerosol 1st Indirect Effect – when aerosols
physically change cloud microphysical properties
and therefore change its albedo.
80
Albedo
2
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
0
20
40
15
60
Optical Depth
80
100
0
5
10
15
20
Effective Radius
25
30
We were able to characterize the bias
induced by aerosol over all cloud
types and found that retrievals of
thicker clouds have a large, low bias of
optical depth and thin clouds have a
large low bias of effective radii.