Cloud Filtering Algorithm For Nadir Viewing FTIR Spectrometers Over Oceanic Regions

1
Cloud Filtering Algorithm For Nadir Viewing FT-IR
Spectrometers Over Oceanic Regions
C.J. Ferguson, J.S. Ribberink, W.F.J. Evans and
A.R.M. Rutledge
2
Introduction
A significant issue facing all satellite-based
nadir viewing FT-IR spectrometers is how to
determine whether the instrument is viewing a
clear or clouded sky condition. Presented here is
the description of an algorithm used to discern
whether a particular spectral radiance
measurement is of a clear or clouded sky
condition over oceanic regions.
3
The development, initial testing and validation
of this algorithm was performed using the
Interferometric Monitor of Greenhouse Gases
(IMG) data set and could easily be altered to
cloud filter similar data from other nadir
viewing FT-IRs.
4
Algorithm
5
General Description
Basically, the algorithm extracts surface
temperature information from many spectral
radiance measurements within a specified
geographic region and time period. It is then
assumed that the average of the four highest
surface temperature values represents the oceans
surface temperature for that region and time
period, if a number of conditions are satisfied.
Since cloud temperatures are typically colder
then ocean temperatures the algorithm is able to
determine if a particular measurement is clouded
by the difference in surface temperatures.
6
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature
threshold value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption threshold value
Clear Sky Measurements
7
The IMG instrument took brightness temperature
measurements in cycles where each cycle took 6
nadir measurements, 1 deep space and an internal
black body. Each cycle lasted 110 seconds.
8
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature
threshold value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption threshold value
Clear Sky Measurements
9
For April 1997 there were 38,500 measurements
which were sorted into 2482 bins leaving on
average 15.5 measurements per bin.
10
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature
threshold value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption threshold value
Clear Sky Measurements
11
Extraction of spectrums characteristics, which
are then used to determine the bins threshold
parameters
12
Here we see the file produced for each bin, which
displays the characteristics of each measurement
and the corresponding threshold values for that
bin.
13
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature
threshold value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption threshold value
Clear Sky Measurements
14
The file is flagged as manual if it fails the
consistency checks.
15
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature
threshold value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption threshold value
Clear Sky Measurements
16
The cirrus cloud check parameter is predetermined
to be 2 K and any measurement that produces a
value or - 2 K is determined to be cirrus
cloud contaminated. (Ackerman et al, GRL, 1998)
17
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature average
value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption average value
Clear Sky Measurements
18
The DT of 4 K was chosen to compensate for areas
of high temperature variation. If a smaller grid
system was used the DT could also decrease
making the filter more accurate.
19
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature average
value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption average value
Clear Sky Measurements
20
The CO2 absorption parameters are extracted from
each spectral radiance measurement by taking the
following absorption free points and absorption
points and dividing them to produce an average
CO2 transmission value.
CO2 was chosen because its concentration is
consistent throughout the atmosphere and the
absorption bands have minimal interference from
other absorption bands.
21
Each measurement is checked to see if its CO2
transmission value Is within 0.005 of the
established average value.
22
Here we see the relationship between CO2
absorption and surface temperature. The blue
data points represent Ocean only, where the
black represents land data. The red line shows
the linear regression of the data. The bottom
figure shows the threshold of 0.005 above the
linear regression.
23
IMG Brightness Temperature Measurements
Measurements are sorted into 5? Latitude by 5?
Longitude bins and then by month
The surface temperature, cirrus cloud and CO2
absorption information is extracted to determine
the threshold parameters for each bin.
If the Bin does not pass a consistency check then
the threshold parameters are taken from the
neighbouring bins.
If neighbouring bins fail to provide threshold
parameters then the bin is flagged for manual
cloud clearing
Each measurement within the bin is first checked
for cirrus cloud contamination
The measurements are then checked to see if their
surface temperature falls within 4 K of
previously determined surface temperature
threshold value
Clouded Measurements
The measurements are then checked to ensure that
the CO2 absorption is within 0.5 of the CO2
absorption threshold value
Clear Sky Measurements
24
The algorithm then reads the filter results from
the file and moves the measurements into their
appropriate folders.
25
Validation
26
Validation of the algorithm using data from the
Ocean Temperature and Colour Scanner (OCTS)
Instrument
Image source www.kuroshio.eorc.jaxa.jp/ADEOS/
27
The OCTS instrument measured several channels of
radiation from visible to infrared, as outlined
in the figure below.
Using the data from the infrared channels the
instrument was able to Create global sea surface
temperature images.
Image source www.kuroshio.eorc.jaxa.jp/ADEOS/
28
In order to validate the sea surface temperatures
derived by the Cloud filter algorithm, a
comparison was made with the sea surface data
from the OCTS instrument for April 1997.
Image source www.kuroshio.eorc.jaxa.jp/ADEOS/
OCTS
IMG
29
Band 1 2 3
From the visible channels, the OCTS
instrument also provided day time images of the
field of view of the IMG instrument and the
surrounding area. These images are available
for most day time IMG measurements
Band 1 2 3
Image Source www.eorc.jaxa.jp/AtmChem/IMG/
30
IMG measurements were then checked for cloud
contamination for the month of April and
recorded as either clear, clouded or partial
accompanied by the appropriate percentage. OCTS
cloud contamination results were then compared
with the results from the cloud filter
algorithm, in order to determine the validity of
the cloud filter.
Image Source www.eorc.jaxa.jp/AtmChem/IMG/
31
This comparison was first carried out on areas of
low temperature variation and are outlined below.
Pacific Cloud 0/4 Partial 51 99
0/7 Partial 20 - 50 7/12 Partial lt10
8/10 Clear 14/15
Atlantic Cloud 1/15 Partial 51 99 0/5
Partial 20 50 6/16 Partial lt10
10/11 Clear 14/15
Indian Cloud 0/10 Partial 51 99
0/2 Partial 20 50 1/10 Partial lt10
0/1 Clear 4/6
32
This comparison was then carried out on areas of
high temperature variation and are outlined below.
Pacific Cloud 3/13 Partial 51 99
2/2 Partial 20 - 50 1/1 Partial lt10
3/3 Clear 3/4
Atlantic Cloud 0/22 Partial 51 99 0/3
Partial 20 50 3/14 Partial lt10
6/6 Clear 6/7
Indian Cloud 0/18 Partial 51 99
1/5 Partial 20 50 1/9 Partial lt10
1/5 Clear 4/4
33
High Temperature Variability Cloud 3/53
(6) Partial 51 99 3/10 (30) Partial 20 -
50 5/24 (21) Partial lt10 10/14
(71) Clear 13/15 (87)
Low Temperature Variability Cloud 1/29
(3) Partial 51 99 0/14 (0) Partial 20 -
50 14/38 (37) Partial lt10 18/22
(82) Clear 32/36 (89)
Overall Total Cloud 4/82 (5) Partial 51 99
3/24 (13) Partial 20 - 50 19/62
(31) Partial lt10 28/36 (78) Clear 45/51
(88) all cirrus cloud contaminated
34
Conclusion
35
1. The cloud filtering method uses the infrared
portion of the electromagnetic spectrum and
therefore works during the day and night.
2. This method of cloud filtering is based
entirely on the satellite FT-IR measurements and
does not require any input from external data
sets.
3. The algorithm requires roughly 15 measurements
per bin and therefore if twice as many
measurements were available in a one month
period the area covered by each bin could be
reduced to half. increasing the cloud filtering
accuracy and spatial resolution. Therefore with
the introduction of imaging FT-IRs the ability
and accuracy of this method will be greatly
increased.
4. The algorithm is designed to work with any
nadir viewing FT-IR data at various spatial
resolutions.
36
Acknowledgements
The authors gratefully acknowledge Enbridge Inc
and ORDCF for providing the financial support for
this work. Support was also provided by the ACE
project. (U. Waterloo P. Bernath)