December 18 2012 Calvin Liang Stephanie Weiss NGAS AampDP Path Forward to Investigate NCC EDR Issues Proposed steps to investigate NCC issue Collect SVDNB GDNBO and IVOBC granules from August 1617 ID: 298468
Download Presentation The PPT/PDF document "NCC Status" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
NCC Status
December 18, 2012
Calvin Liang
Stephanie Weiss
NGAS A&DPSlide2
Path Forward to Investigate NCC EDR Issues
Proposed steps to investigate NCC issue:
Collect SVDNB, GDNBO, and IVOBC granules from August 16-17
th
, 2012 – Around new moonConstruct stray light correction lookup table (LUT) from new moon dataApply stray light correction LUT to new moon granules using NGAS stray light correction algorithmDetermine best functional fit to corrected radiances and account for night glow. (In work right now)Produce a new GVVSSE/GVVSLE LUT based on the new functional fits.Remove stray light in problematic granulesRun the latest version of the NCC EDR algorithm in ADL on stray light corrected granules (results are shown using ADL 3.1)Inspect output to determine whether some of the problems identified in the granules go away with stray light correctionIf it is found that fixing stray light does not completely solve the problems, then we will delve into the NCC EDR algorithm logic (Wrote offline NCC code for testing) ADL code has been modified along with appropriate LUTs to address fill granules.
2
Completed
In Work
PlannedSlide3
Radiance
vs Solar Zenith After
Stray Light Correction (Updated)
3
Noise floor 7e-11Current SZA cutoff = 105Straylight removedSome overcorrection but very little
New Lunar table that excludes night glow.
C. Liang, NGAS
New solar table to account for night glow. Slide4
The NCC Algorithm
4
Pseudo-
albedo
α is the NCC output and is defined as:where LDNB is the DNB radiances from the SVDNB product. L
sol, and
L
lun
are solar reference and lunar radiances computed for a reference earth that is
Lambertian
to first order.
arf
is the asymmetric reflectance factor which takes into account a second order BRDF effect on the reference earth
E
and
G
are the irradiance and GVVSSE/GVVSLE gain valuesSlide5
VNCCO-npp-d20120612-t0010428-e0012136-b03229
5
It is recommended to extend the solar/lunar zenith angle to span [0,180].
Lunar
Solar
Currently the Ops code cuts off at 105 for both solar and lunar zenith anglesThis limit creates problems since both
L
sol
and
L
rad
are necessary to balance each other.
C. Liang, NGASSlide6
6
The discontinuity arises because the solar angle limit is 105. Need solar and lunar values to transition smoothly.
Lunar
Solar
Lunar
Solar
C. Liang, NGAS
VNCCO-npp-d20120612-t0010428-e0012136-b03229Slide7
7
With recommended changes, ADL now produces expected smooth transition between solar and lunar conditions
Lunar
Solar
Lunar
Solar
C. Liang, NGAS
VNCCO-npp-d20120612-t0010428-e0012136-b03229 (new)Slide8
Aurora Imagery (SDR, 12 granules), d20120612-t0000485-e0017522-b03229
8
C. Liang, NGASSlide9
Aurora Imagery (NCC, 12 granules), d20120612-t0000485-e0017522-b03229
9
C. Liang, NGAS
Large value is due to 1 pixel having a light source that was magnified by gain.Slide10
Proposed GVVSSE/GVVSLE LUT
10
GVVSLE just needs to be larger than the GVVSSE in the night glow region.
Night glow will be handled by GVVSSE table
C. Liang, NGASSlide11
Proposed GVVSSE/GVVSLE LUT
11
GVVSLE just needs to be larger than the GVVSSE in the night glow region.
Night glow will be handled by GVVSSE table
C. Liang, NGASSlide12
Summary
New updates to the ADL LUT and code have removed the fill value granule issues previously observed.Previous out of bounds issues primarily were due two issues:
The 105 solar angle cut-off
The Ops table did not account for night glow
It is recommended to extend the solar and lunar zenith angles out to 180GVVSSE will be modified to account for night glow/ GVVSLE will NOT account for night glow because the solar table will do soIt is also recommended to remove the [0,5] albedo limit. User of the NCC product can use whatever limit they choose to suit their needs.A pure LUT change to avoid code changes is not possible:Although the solar/lunar angle thresholds (sza_threshold/lza_threshold), minimum NCC radiance (min_ncc_radiance), and albedo limits can be adjusted by simple LUT changes, the size of the table cannot be adjusted through a LUT change.Keeping the 0.1 degree resolution of the GVVSSE/GVVSLE table (10 bins per degree) would require a change to the parameter ZENITH_ANGLE_GAIN_BINS from the value 1051 to 1801; this parameter is in “ViirsNccStructs.h” This will require a recompile of the codeAnother route is to change the angular resolution of the GVVSSE/GVVSLE table such that number of solar/lunar angle and gain values remain 1051; this corresponds to an angular resolution of ~0.1714 deg. However, this will also require a code change as the gradation of the table is NOT calculated from the LUT solar/lunar angles but hard coded into the file “NccPtrStruct.f” in the parameter “GVVSE_BIN_WIDTH” which is set to 10 in the ops code. This value would need to be changed to 5.83333333.12Slide13
Path Forward (Overall)
The change from 0.1 to 0.1714 deg resolution in solar/zenith angles did not produce any visual artifacts. Thus, in order to avoid a change in LUT size, which would involve recompilation of modules upstream from the NCC EDR code, it is recommended to update the GVVSSE_BIN_WIDTH parameter to 5.8333333.Albedo
limits should be expanded to a range that allows for negative values and positive
albedos
much larger than 5. Allowing negative albedos will require a change to “min_ncc_radiance” to a value < -1.5-e9 (the minimum allowed DNB radiance).Set the high albedo limit to be 1000. This may be overkill but will allow for extraneous cases (bright light source at large lunar angles for near new to new moon conditions)SZA and LZA thresholds should be set to [0,180]; these threshold are in the same LUT as “min_ncc_radiance”13Slide14
Path Forward (2 Step Process)
Make LUT/PCT changes:Albedo limits in:
ProCmnProductDictionary_CFG.xml
Set “
min_ncc_radiance” to value < -1.5e-9 in: *VIIRS-NCC-EDR-AC*xml_InternalModify GVVSSE to account for night glowModify GVVSLE table so that it increases exponentially beyond the GVVSSE night pointBoth GVVSSE/GVVSLE will continue to have the 105 solar/lunar zenith angle limitMake code changes along with LUT changes:Set sza_threhold/lza_threshold to 180 in : *VIIRS-NCC-EDR-AC*xml_InternalChange GVVSE_BIN_WIDTH to 5.833333 in: NccPtrStruct.fLeave the number of LUT elements constant at 1051 for angle and gainUpdate GVVSSE and GVVSLE table to extend out to 180 with a resolution changed from 0.1 to 0.1714 deg. Because there is a code change involved, the small code change will not make it into build Mx7.0; will not likely get in until fall of 2013. The LUT changes in step 1 can be implemented relatively quickly and will lead to improvement in the images.14Slide15