Vicarious Calibration of GOES Imager Visible Channel:
Use of Stars for Responsivity Trending
I-Lok Chang, Dejiang Han, Charles Dean,
Xiangqian Wu,Michael Weinreb
(Revised January 21, 2009)
There is no on-board device for
calibrating the visible channels of the GOES imagers and sounders.
However, the responsivity of the GOES imagers' visible channel
is being monitored with star observations. Approximately 4 different
stars are viewed twice per hour by the imagers to determine
the attitude and orbit of GOES. The reference
shows that these observations can also be used to trend the
responsivity of the visible detectors.
Until January 2004, the analysis of the star data was to fit the
observations taken from the processed star files in the Orbit and Attitude
Tracking System(OATS) directly with an exponential function. Since then,
we have begun to carry out the analysis with a better technique, one that
reduces the noise in the time series by screening out “bad” data. However,
to allow users to maintain continuity in their data, we present here the
results of both the original analysis (“Method 1”) and the better analysis
(“Method 2”).
Method 1
Analysis Technique: For each
star, data available are about one observation per star per day, for approximately 40 stars.
- Delete data +/-5 hours around S/C midnight to eliminate
degradation due to scan mirror distortion caused by heating from direct sunlight.
- Fit a time series of the remaining data to an exponential in time.
- Keep track of the fit coefficients for each star.
- Trend is the average of
all stars
Limitations:
- Technique is based on the irradiance of
a point source rather than the radiance of an extended source; therefore,
star trend may be slightly different from trend seen on Earth scenes.
- Although detectors used for Earth imaging
are the same as those used for star observation, the star observations
employ an additional stage of amplification in the electronics. We believe,
but cannot be certain, that the gain of the additional stage is constant
in time.
- The OATS analysis, aimed at determining the
time of star transit and not the signal amplitude, results in excessive
noise in the inferred amplitude data
Example of Raw Data:
- GOES-8
Star Beta - Cnc Imager Signal (Method 1)
- GOES-9
Star Beta - Cnc Imager Signal (Method 1)
- GOES-10
Star Beta - Cnc Imager Signal (Method 1)
- GOES-11
Star Beta - Cnc Imager Signal (Method 1)
- GOES-12
Star Beta - Cnc Imager Signal (Method 1)
Model: The fit models the responsivity decrease as follows:
R = e -A(t-to)
where
R = Relative responsivity (R = 1 at beginning of time series)
A = Degradation rate per day
t = Time in days since launch
t0 = Number of days between launch and beginning of time series
All coefficients are spacecraft specific and are given below:
| Spacecraft |
A |
Launch Date |
Date of Beginning
of
Time Series |
End of Availability
of Data |
| GOES-8 |
1.359 x 10-4 |
April 13, 1994 |
April 10,
1995 |
April 01,
2003 |
| GOES-9 |
1.481 x 10-4 |
May 23, 1995 |
August 7,
1995 |
May 16,
1998 |
| GOES-10 |
1.257 x 10-4 |
April 25,
1997 |
March 21,1998 |
|
| GOES-11 |
1.204 x 10-4 |
May 3, 2000 |
June 21,
2006 |
|
| GOES-12 |
1.182 x 10-4 |
July 23, 2001 |
April
01,
2003 |
|
Examples of GOES Imager fits:
- GOES-8
Imager Fits of Relative Responsivity Vs Time (Method 1)
- GOES-9
Imager Fits of Relative Responsivity Vs Time (Method 1)
- GOES-10
Imager Fits of Relative Responsivity Vs Time (Method 1)
- GOES-11
Imager Fits of Relative Responsivity Vs Time (Method 1)
- GOES-12
Imager Fits of Relative Responsivity Vs Time (Method 1)
Listed below are the average
GOES Imager visible channel responsivity decreases per
year.
| Spacecraft |
A (annual rate) |
Length of Time Series |
| GOES-8 |
4.96±
0.09%* |
Apr.
10, 1995 to Apr. 01, 2003 |
| GOES-9 |
5.41±
0.28%* |
Aug.
07, 1995 to May 16, 1998 |
| GOES-10 |
4.59±
0.08%* |
Mar
21, 1998 to Dec. 17, 2008 |
| GOES-11 |
4.39± 0.34%* |
Jun. 21, 2006 to Dec. 17,
2008 |
| GOES-12 |
4.31± 0.09%*,a |
Apr. 01, 2003 to Dec. 17,
2008 |
* The stated trend is the mean of the trends for approximately 40 stars, and
the stated error is the usual standard error of the mean.
a
As can be seen in the raw data plot (above) GOES-12 Star Beta-Cnc Imager Signal
(Method1), the GOES-12 data are unique in that thay have an extra band of points
running parallel to
the main sequence at approximately 13 units above it. Most of the points in this
upper band are star signals from detector 7. They occur above the main sequence
because of an error in
the value of the detector-7 normalization factor in the OATS database. Other
points in this band arise from the summing of star signals in multi-detector
crossings. The values of A
(annual degradation rates) for GOES-12 in the preceding table were derived, nevertheless,
with the data from all detectors, including detector 7. Removing the detector
7 data from the
analysis gives an annual degradation rate of 4.32± 0.09%/yr.
Method 2
Analysis Technique: For each star, data available are about one observation
per star per day, for approximately 60 stars.
- Delete data +/-5 hours around S/C midnight to eliminate
degradation due to scan mirror distortion caused by heating from direct sunlight.
- Use data from OATS
analysis, as in Method 1, but add the following steps to improve the data.
-
Delete all data from detectors 1 and 8 (partial crossing of these array-end
detectors give excessively low values).
-
Delete any star signal transit where signal is observed in more than
one detector (OATS computes excessively high signal).
-
Remove the OATS multiplication by detector-dependent constants. (OATS
normalizes the signal from each detector and divides by the expected noise.
This process is imperfect and results in serious noise.)
-
Fit a time series of the remaining data to an exponential in time.
-
Keep track of the fit coefficients for each star.
- Trend is the average of
all stars
Limitations:
- Technique is based on the irradiance of
a point source rather than the radiance of an extended source; therefore,
star trend may be slightly different from trend seen on Earth scenes.
- Although detectors used for Earth imaging
are the same as those used for star observation, the star observations
employ an additional stage of amplification in the electronics. We believe,
but cannot be certain, that the gain of the additional stage is constant
in time.
- Less data than in Method 1,
for two reasons. The first is the result of
the screening, obviously. The second reason is that much of the OATS data
before 1998 needed for screening were not accessible. The shorter time series
results in larger variabilities in the estimated trends.
Example of Raw Data:
- GOES-8
Star Beta - Cnc Imager Signal (Method 2)
- GOES-10
Star Beta - Cnc Imager Signal (Method 2)
- GOES-11
Star Beta - Cnc Imager Signal (Method 2)
- GOES-12
Star Beta - Cnc Imager Signal (Method 2)
Model: We model the responsivity decrease as follows:
R = e -A(t-to)
where
R = Relative responsivity (R = 1 at beginning of time series)
A = Degradation rate per day
t = Time in days since launch
t0 = Number of days between launch and beginning of time series
All coefficients are spacecraft specific and are given below:
| Spacecraft |
A |
Launch Date |
Date of Beginning
of
Time Series |
End of Availability
of Data |
| GOES-8 |
1.331 x 10-4 |
April 13, 1994 |
Oct 19,
1995 |
April 01,
2003 |
| GOES-9 |
Insufficient
Data |
May 23, 1995 |
|
|
| GOES-10 |
0.926 x 10-4 |
April 25,
1997 |
Jan. 04,2001 |
|
| GOES-11 |
1.216 x 10-4 |
May 3, 2000 |
Jun. 21, 2006 |
|
| GOES-12 |
1.216 x 10-4 |
July 23, 2001 |
Apr. 01,2003 |
|
Examples of GOES Imager fits:
- GOES-8
Imager Fits of Relative Responsivity Vs Time (Method 2)
- GOES-10
Imager Fits of Relative Responsivity Vs Time (Method 2)
- GOES-11
Imager Fits of Relative Responsivity Vs Time (Method 2)
- GOES-12
Imager Fits of Relative Responsivity Vs Time (Method
2)
Listed below are the average
GOES Imager visible channel responsivity decreases per
year.
| Spacecraft |
A (annual rate) |
Length of Time Series |
| GOES-8 |
4.86 ±
0.08%* |
Oct.
19, 1995 to Apr. 01, 2003 |
| GOES-9 |
Insufficient
Data |
|
| GOES-10 |
3.38±
0.06%* |
Jan.
04,
2001 to Dec. 17, 2008 |
| GOES-11 |
4.44±
0.28%* |
Jun.
21,
2006 to Dec. 17, 2008 |
| GOES-12 |
4.44± 0.05%* |
Apr.
01, 2003 to Dec. 17, 2008 |
* The stated trend is the mean of the trends for approximately 60 stars, and
the stated error is the usual standard error of the mean.
Home | Operations |
News | Daily Updates |
Info & History
GOES | POES |
DMSP | Contact Us |
Webmaster
Privacy Policy
|
DOC
NOAA
NESDIS
