Summary of ADAPS 1KM Benchmark
Verification of calibration code
General Notes
All images are windowed to be 874 samples on each side of nadar
(1748 samples total, this is what is done in production).
Consequently, 150 samples on the extreme left and right are
dropped.
The actual data is not used for the image, but rather a computed
image wedge of data with values ranging from 0 - 1023 (range of
valid raw AVHRR data for all five bands). None-the-less, windows
with data ideal for an atmospheric correction test have been
selected, and this actual image data is saved, though not used
for now.
Criteria for testing
- Test with illumination correction applied and not applied. The
scenes to test this are taken at summer and winter solstice since
this is when the solar angles are extreme. The latitude selected
for testing is at approx. 40 degrees (mid-US).
- All test images are ascending, since ascending/descending images
should not be critical to these calibration tests.
- Test image with noise and dropped lines in it so that minor frame
data processing is verified.
- Test images from NOAA-11, NOAA-12, NOAA-14.
- Test using calibration tables as follows:
- NOAA-11
- PRELAUNCH tables
- TEILLET tables with data 2 years after launch (Sept 90)
- NOAA-12
- NOAA-14
- Run software tests varying the degradation technique, calibration
period, the number of calibration periods, and the
band sequence in AVHRRCAL.
- Do NOT vary the following quantities, but rather use values
consistent with 1KM production processing:
- output data type (use I*2)
- temperature range for thermal channels
- Correct for nonlinearity errors
- bands 4 and 5 of NOAA-N calculated with zero radiance of space
- Test all of the scaling options "GLOBAL", "US" , and "NONE".
- Generate the output in radiance or percent reflectance with illumination
correction. To emulate production, GLOBAL scaling should be tested with
radiance output values (without illumination correction appled);
US scaling should be tested with percent reflectance output values
(with illumination correction appled).
- Have one "bad" test image with values outside of the range "0 - 1023"
to verify that failure will occur.
Test case scenarios
- Test case scenario #1
-
- Original image name
- ah11062094213142 from archive tape 064038.
- Windowing of image
- 512 lines close to 40 degrees latitude were selected. So the
window is (2880,151,512,1748) for all 5 bands.
- Rational of test scenario
- This image was chosen because it is approx. at summer solstice
(extreme solar angles), is an ascending pass over the US at
approx. 40 deg latitude.
Optical calibration degradation file being used is "PRELAUNCH".
The test is run with illumination correction applied (reflectance
output values) and without illumination correction applied
(radiance output values).
- Test case scenario #2
-
- Original image name
- ah11092990203303 from archive tape 053509.
- Windowing of image
- 512 lines close to 40 degrees latitude. The window is
(2200,151,512,1748) for all 5 bands. This window includes the Great
Salt Lake and surrounding area.
- Rational of test scenario
- This image was chosen because it is approx. two years past the launch
of the NOAA-11 satellite. Consequently, this is a good time frame to
verify the TEILLET calibration method. It is also during the time
of the fall equinox. It is an ascending pass over the US at
approx. 40 deg latitude.
Optical calibration degradation file being used is "TEILLET".
Illumination correction is applied.
- Also run a SOFTWARE test
- The calibration period and number of calibration periods is varied
from the default. Also, the order of bands is altered from the
default to make certain the correct AVHRR channels are still
processed. NOTE: The variation of CALPER and NCALS will cause
differences in accuracy in the output, especially if illumination
correction is being applied. The biggest differences can be seen
in AVHRR band 3 right where the input signal becomes large enough to
force the computed thermal value to its minimum value (160 degrees).
- Test case scenario #3
-
- Original image name
- ah14012295185257 from archive tape 075414.
- Windowing of image
- 512 lines at the northern-most part of the scene (not close to 40
degrees latitude). The window is (1,151,512,1748) for all 5 bands.
This window was chosen because there are multiple dropped lines in
the image window, and the available data is fairly cloud free.
- Rational of test scenario
- This image was chosen because it is a NOAA-14 scene with drop
lines surrounded by fairly clear data. Note that this data is
not near 40 deg lat but in Canada.
Optical calibration degradation file being used is "PRELAUNCH".
Illumination correction is being applied.
- Test case scenario #4
-
- Original image name
- ah12122192131236 from archive tape 027117.
- Windowing of image
- 512 lines at the center part of the scene (close to 40 degrees
latitude). The window is (2400,151,512,1748) for all 5 bands.
- Rational of test scenario
- This image was chosen because it is a NOAA-12, it is a '92 scene,
it is a winter solstice scene at approx. 40 degrees latitude.
Optical calibration degradation file being used is "PRELAUNCH".
Illumination correction is and is not being applied (two tests).
Verification of normalized difference calculation
General Notes
1KM production conditions are held constant:
- Output data type is always "byte"
- The offset is always 1.0
- A scale factor of 100.0 is always specified
- The normalized limit value is set to 200
- Always round instead of truncating
Test case scenarios
- Test case scenario #1
-
Test with an image of mask values specifying NONE and US scaling
(results should be the same since neither does anything special with
"mask" values).
- Test case scenario #2
-
Run test #1 only specifying GLOBAL scaling (GLOBAL scaling does use
mask values). Compare the results with predetermined answers.
The mask combinations used in this test are equal.
- Test case scenario #3
-
Repeat test #2 only in this test specify an
invalid combination of mask values. Verify that a fatal error results.
- Test case scenario #4
-
Repeat test #3 except specifying NONE and US scaling. Verify that a
fatal error does NOT occur since special "mask" values do not exist for
NONE and US scaling algorithms. Verify that the same answers are obtained
for both scaling algorithms (US and NONE). Compare results to
pre-determined answers.
- Test case scenario #5
-
Run a test using GLOBAL scaling where both bands are equal. Compare the
results to pre-determined answers.
- Test case scenario #6
-
Run a test using NONE and US scaling where both bands are equal. Compare the
results to pre-determined answers.
- Test case scenario #7
-
Run a test where one input band has values which increment from 10 to 1023
vertically, the other input band has values which increment from 10 to 1023
in the horizontal direction. Use GLOBAL scaling. This scenario ensures
the testing of all possible combinations of values except for mask values
(illegal combinations of mask values would force this test to fail). Compare
the results to pre-determined answers.
- Test case scenario #8
-
Run a test where one input band has values which increment from 0 to 1023
vertically, the other input band has values which increment from 0 to 1023
in the horizontal direction. Use NONE and US scaling. This scenario ensures
the testing of all possible combinations of values (there are no mask values
for NONE and US scaling). Compare the results to pre-determined answers.
Verification of 1KM production flow results
General Notes
Process AVHRR scenes much like that done in production Global
1KM processing and verify the output with established results.
Four test scenes have been selected:
-
"ah11062692204528" is a NOAA-11 scene acquired at summer solstice
over study area 3.
-
"al14122395204111" is a NOAA-14 scene acquired at winter solstice
over study area 3.
-
"ah15033099130104" is a NOAA-15 scene acquired at spring equinox
over study area 1.
-
"ah14062095223227" is a NOAA-14 scene acquired at summer solstice
over study area 1.
Processing flow
The following processing flow will be performed on each test
scene:
Begin by running line detection on the raw AVHRR archive image. Verify
the results with pre-determined answers.
Call GACOSINFO to get the information for this study area. This information
will be used throughout the remainder of the test.
Call AVHRRIN-LATRANGE to convert the AVHRR archive image to a LAS image.
Only convert the data which falls within the proper latitude, longitude, and
quadrant number. Only convert 874 pixels on either side of nadar (truncate
extreme 150 pixels on both edges of the image).
Calibrate the data next. Call AVHRRCAL for the line detected LAS image.
Specify VERMOTE tables and no illumination correction to be applied.
Calibrate all 5 AVHRR bands with radiance output values, "I*2" output data
type, and with GLOBAL scaling applied. Verify the results with pre-determined
answers.
Call AUTONAV-MULTI to test the navigation of the scene. Several attempts
will be made to navigate the image using DCW and WVS data. If the scene
successfully navigates verify the results by comparing the ;addr file (only
thing that changes because of navigation). If the scene does not successfully
navigate, continue on.
Next reproject the image. Call PROJPRM, SWINDOW, RECTIFY, and GEOM to generate
reprojected bands of AVHRR data ("I*2" data using GLOBAL scaling). Use nearest
neighbor resampling. Call ANGINTERP to generate bands of sun and satellite
angles in the same projection space as the AVHRR data (byte data using GLOBAL
scaling). Verify the results by comparing them with pre-determined answers.
The angle bands will be needed for atmospheric correction.
Call NORMD to generate normalized difference results. The results are byte
data using GLOBAL scaling. A scale factor of 100 and a normalized limit of
185 is specified.
Map the solar angle band to the NDVI band as well as bands 1 and 2 of the
AVHRR image. Do not use solar angles greater than 79 degrees. Verify the
results by comparing them with pre-determined answers.
Prepare to atmospherically correct bands 1 and 2 of the AVHRR image.
Get the appropriate elevation image, ozone image, and water mask.
Apply the water mask and interrupt area mask to the AVHRR image.
Call ATMOCOR to atmospherically correct the visible AVHRR bands.
The output is radiance values using GLOBAL scaling.
Verify the results by comparing them with pre-determined answers.
After atmospherically correcting AVHRR bands 1 and 2, recalculate the
NDVI of bands 1 and 2 using the same parameters as the first time.
Verify the results by comparing them with pre-determined answers.