nasa viirs characterization support team (vcst ......aqua use definitive ephemeris data 27 hour...
TRANSCRIPT
-
(SNPP + J1/N20 + J2) VIIRS Geolocation Status
NASA VIIRS Characterization Support Team (VCST)
Geometric Calibration Group
Guoqing (Gary) Lin, SSAI/GSFC code 619
Robert E. Wolfe, NASA/GSFC Code 619
James C. Tilton, NASA/GSFC Code 606
Ping Zhang, Bin Tan, SSAI/GSFC Code 619
John Dellomo, GST/GSFC Code 619
NASA MODIS/VIIRS STM Calibration Workshop
Silver Spring, Maryland
18 October 2018
-
Acknowledgements
• Thanks to Carol Davidson & her Land SIPS (formerly
Land PEATE) Team for processing control point
residuals and testing Geo LUTs updates
• Thanks to Fred Patt of NASA Ocean Group for helping
us understanding and resolving issues related to
ephemeris and attitude data
• Thanks to the NOAA STAR VIIRS SDR Team and many
other branches of the JPSS Program
Lin et al., 18 Oct 2018 VCST/GEO 2
-
Outline
• SNPP VIIRS Geolocation Performance and Trends
• J1/N20 VIIRS Geolocation Performance and Trends
• Some expectations of J2 VIIRS
• Improvements and future work
• Conclusions
Lin et al., 18 Oct 2018 VCST/GEO 3
-
VIIRS Geolocation Performance
Residuals SNPP
VIIRS C. 1.0
J1/N-20
VIIRS Prel.
Aqua
MODIS C6.1
Terra
MODIS C6.1
Track mean 13 m 7 m 1 m 0 m
Scan mean 4 m -24 m 0 m 0 m
Track RMSE 58 m 62 m 46 m 43 m
Scan RMSE 52 m 59 m 53 m 44 m
Data-days 2447 (6.7 yrs) 282 (0.8 yrs) 5910 (16.2 yrs) 6725 (18.4 yrs)
Missing days 1 6 10 59
Daily matched
GCPs w/ I1/B1
204 183 223 258
• Nadir equivalent accuracy (RMSE – Root Mean Square Error) . (MODIS for reference)
– Meet Spec: 125 m (1s); within 20% I1 HSI (375 m) = 75 m @ nadir for VIIRS
– Band-to-band mis-registration to other bands adds bias to RMSE to :
• MODIS – VIIRS differences
Aqua use definitive ephemeris data 27 hour latency
SNPP attitude data is not as good, see Slide #13
DEM resolutions: older 1 km for VIIRS vs newer 0.5 km for MODIS C6/C6.1
22 s RMSE
Lin et al., 18 Oct 2018 VCST/GEO 4
-
-112.5
-75
-37.5
0
37.5
75
112.5
12 13 14 15 16 17 18 19
Tra
ck (
ad
j.)
res.
(m)
.
Years since Jan. 1, 2000
Daily 16-day Global 16-day S. Hemisphere 16-day N. Hemisphere
-112.5
-75
-37.5
0
37.5
75
112.5
12 13 14 15 16 17 18 19
Scan
(ad
j.)
res.
(m)
Years since Jan. 1, 2000
Lin et al., 18 Oct 2018
SNPP C1 geolocation errors
Started 1/19/2012
Cryo door open
VCST/GEO 5C1.1 RMSE Track: 58 m Scan: 52 m, nadir equivalent
±20% I-band pixel size
-
-112.5
-75
-37.5
0
37.5
75
112.5
-60 -40 -20 0 20 40 60
Tra
ck r
esid
uals
(m
)
Scan angle (deg)
-112.5
-75
-37.5
0
37.5
75
112.5
-60 -40 -20 0 20 40 60
Scan r
esid
uals
(m
)
Scan angle (deg)
SNPP Scan Angle Residuals
Lin et al., 18 Oct 2018
Track
Scan
VCST/GEO 6
±20% I-band pixel size
Strange pattern is being corrected,
see Slides 10 - 12
-
-112.5
-75
-37.5
0
37.5
75
112.5
17.75 18 18.25 18.5 18.75 19
Tra
ck (
ad
j.)
res.
(m)
.
Years since Jan. 1, 2000
Daily 16-day Global 16-day S Hemi 16-day N Hem
-112.5
-75
-37.5
0
37.5
75
112.5
17.75 18 18.25 18.5 18.75 19
Scan
(ad
j.)
res.
(m)
.
Years since Jan. 1, 2000
Lin et al., 18 Oct 2018
J1/N20 preliminary geolocation errors
Started 12/14/2017
Nadir door open
VCST/GEO 7Preliminary RMSE Track: 62 m Scan: 59 m, nadir equivalent
-25
-20
-15
-10
-5
0
5
10
15
20
25
17.92 18.00 18.08 18.17 18.25 18.33 18.42 18.50 18.58 18.67 18.75
Mo
un
tin
g r
ota
tio
n (
arcs
ec)
Time (years from c. 2000)
J1 R/P/Y correction needed & in-work
Roll Pitch Yaw
J1 temporal pointing variation is large (>
SNPP). Yaw and pitch pointing variations
cause geolocation error ~ 250 m @ edge
of scan.
MODIS-like correction is being worked
for both SNPP and J1 pointing variations.
±20% I-band pixel size
VCST/GEO 7
-
J1 DNB geolocation
Lin et al., 18 Oct 2018
Courtesy of Wenhui Wang and Slawomir Blonski of NOAA STAR
-750
-600
-450
-300
-150
0
150
300
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
96
296
496
696
896
109
6
129
6
149
6
169
6
189
6
209
6
229
6
249
6
269
6
289
6
309
6
329
6
349
6
369
6
Mea
sure
d s
can
off
set
(m)
FPA
X-o
ffse
t (m
m)
Sample# at mid-points of 114 scan chips
J1/N20 DNB scan geo offset (LSIPS LUT r2)
FPA equivalent Local ground offset-750
-600
-450
-300
-150
0
150
300
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
96
296
496
696
896
109
6
129
6
149
6
169
6
189
6
209
6
229
6
249
6
269
6
289
6
309
6
329
6
349
6
369
6
Mea
sure
d s
can
off
sdet
(m)
FPA
X-o
ffse
t (m
m)
Sample # at mid-points of 167 scan chips
J1/N20 DNB scan geo offset (LSIPS LUT r1)
FPA equivalent Local ground offset
nadirnadir
~5 km bias ~200 m
~200 m ~20 m
On-orbit Cal, further
VCST/GEO 8
-
Expectations for J2 VIIRS
Lin et al., 18 Oct 2018
• Band-to-band co-registration -- good
• Pointing – good
• Geolocation – should be good with on-orbit calibration
• Line spread functions (LSFs) / modulation transfer functions
(MTFs) – mostly good, except DNB at low light
VCST/GEO 9
0
1
-1 -0.5 0 0.5 1 1.5 2 2.5 3
Spat
ial r
esp
on
se
Distance from sample center (HSI)
Illustration of Simulated J2 DNB LSF at low light
nominal LSFSidelobe -1Sidelobe-2System LSF
Scan direction
Incorrect J2 voltage setting causes the charge in
the current sample to remain behind in the
transfer gate and be deferred into the next
sample in the scan direction
0
5
10
15
20
25
30
35
40
45
50
10 100 1000 10000
% s
ide
lob
e
Peak dn
%sidelobe vs peak_dn (HGA & HGB)
"Optimal light level"
"Bonus light levels"
Envelope
dn at Lmin =
3 nW/cm^2/sr
Agg32, Agg1
Envelope
-
Answer:
Lsight = Ltel - 1/M ( Ltel – Lhamvector)
where M = -4 (not +4 as we are currently
using), which affects line of sight due to
the parts of RTA/HAM motion non-
linearity (non-synchronization), which
are large enough to skew the errors
outside of ± 20% I-band pixel size
-20
-15
-10
-5
0
5
10
15
20
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300
LOS d
evia
tion (
arcs
ec)
Encoder Sequence, 3rd scan 20130410T1325z -- S-NPP3rd scan 20180104T1812z -- J1/N20
VIIRS HAM encoder non-linearity
S-NPP J1/N20
-112.5
-75
-37.5
0
37.5
75
112.5
-60 -40 -20 0 20 40 60
Scan r
esid
uals
(m
)
Scan angle (deg)
SNPP VIIRS geolocation errors in scan direction
VIGMU:VIIRS instrument geometric model update
• Puzzle: ground geolocation
SW is supposed to correct
RTA/HAM motion non-linearity
• Long term trend from SNPP
VIIRS still shows the pattern,
but in the opposite direction
Lin et al., 18 Oct 2018 VCST/GEO 10
-
Image flipping
before HAM
Image flipping
Image flipping after
HAM
Lsight = Ltel - 1/M ( Ltel – Lhamvector)
where M = -4 (not +4 as we are
currently using), which affect non-
synchronization of RTA/HAM only
Lin et al., 18 Oct 2018 VCST/GEO 11
-
VIGMU tests – J1 & NPP
Lin et al., 18 Oct 2018
-100
-80
-60
-40
-20
0
20
40
60
80
100
-60 -40 -20 0 20 40 60
Scan r
esid
uals
(m
)
Scan angle (deg)
J1, 4/7/ - 5/8/2018
-100
-80
-60
-40
-20
0
20
40
60
80
100
-60 -40 -20 0 20 40 60
Scan r
esid
uals
(m
)
Scan angle (deg)
SNPP, 4/7/ - 5/8/2018
VCST/GEO 12
-
10
100
1,000
10,000
183 366 548 731 914 1,096 1,279 1,461 1,644 1,827 2,009 2,192
Pe
ak
RSS
(RP
Y)
(arc
sec)
Time (Days since 10/28/2011)
3s=108, 99.73% 3s=36, 99.73%
SNPP SC attitude performance- - Spec outage and trend
Lin et al., 18 Oct 2018
8/23/2012
using ~90% data
Large circles for control spec outage; Small dots hint knowledge spec outage
Star tracker cooling improved SNPP attitude performance
We are seeking for further improvements3
We are developing SW with Kalman filter to refine the attitude for NASA SIPSs
J1 is performing better but we are monitoring
Roll: 5,666”
Trackers TEC 0oC-10oC
9/22/2016 Day 1791
Data up to
8/7/2017
Day 2110
1st year
3. My eRooms > S-NPP Flight Operations and Support > FARB > All Discussion Topics--Artifacts and Minutes > DR 6348--SNPP STAR TRACKER
DEGRADATIONS OVER MISSION LIFE: ATTITUDE EXCURSIONS AND LUNAR INTRUSIONS > SNPP ADCS and Geolocation Report
2nd year 3rd year 4th year 5th year 6th year
VCST/GEO 13
https://jpss-erooms.ndc.nasa.gov/eRoomhttps://jpss-erooms.ndc.nasa.gov/eRoom/JPSSFlightOperationsandSupport/SNPPFLIGHTOPERATIONSANDSUPPORThttps://jpss-erooms.ndc.nasa.gov/eRoom/JPSSFlightOperationsandSupport/SNPPFLIGHTOPERATIONSANDSUPPORT/0_4cbdhttps://jpss-erooms.ndc.nasa.gov/eRoom/JPSSFlightOperationsandSupport/SNPPFLIGHTOPERATIONSANDSUPPORT/0_4e86https://jpss-erooms.ndc.nasa.gov/eRoom/JPSSFlightOperationsandSupport/SNPPFLIGHTOPERATIONSANDSUPPORT/0_17c66
-
SNPP Events
data exclusion for now
Lin et al., 18 Oct 2018
-1125
-750
-375
0
375
750
1125
0 6 12 18 24 30 36 42 48 54 60 66 72
viir
s G
eo E
rr (
nad
ir e
qu
iv m
)
Time (hours from 2015-08-19T00:00z)
snpp scc 1 sec time err track_shift scan_shift
1 second time error
onboard of SNPP
14:24:40.6z
21:16:30.6z
Leap second insertions: 2012-07-01, 2015-07-01, 2017-01-01
Inclination adjustment maneuvers: poor pointing ~ 2 hours after
#1-4: 2014/ 7/30+9/10+10/15+11/12;
#5: 2015 9/23; #6: 2016 9/29;
#7: 2017 9/28; #8: 2018 9/27.
Drag make-up maneuvers: 28 DMUs up to 2018-07-17.
VCST/GEO 14
We will “hide” the
granules where
geolocation accuracy is
poor.
-
0
100
200
300
400
500
-90 -75 -60 -45 -30 -15 0 15 30 45 60 75 90
Dis
tan
ce (m
)
Latitude (deg)
nadir distance_to_next_scanline
J1 VIIRS Scan-to-scan underlaps
J1/N20 VIGMU test data
• Widest underlaps occur at nadir near 15N at ~ 70
m in this case. They narrow down as J1 goes
north or south due to increasing altitude. They
also close in off nadir angles (@ ~10 deg) due to
bowtie effects
• High terrain widens the underlaps.
• SNPP VIIRS has less of this issue because of its
shorter focal length and scan speed (~0.4%)
Underlap
overlap
-90
-60
-30
0
30
60
90
-180 -150 -120 -90 -60 -30 0 30 60
350
370
390
410
430
450
470
-30 -15 0 15 30 45 60
Dis
tan
ce (m
)
Latitude (deg)
nadir distance_to_next_scanline
Underlap
Within scan
Inter-scan
Inter-HAM-sides
where F = effective focal length = Mag x aft optic focal length, p = detector “pitch” interval in the track direction, n = #
detectors, h = range from satellite to earth terrain surface altitude, T = scan period, i=inclination angle (in ECI) < 90 deg for J1,
VECI = spacecraft ground speed in the inertial frame, Vearth0 = speed of earth rotation at equator, Overlap < 0 indicates underlap.
𝑂𝑣𝑒𝑟𝑙𝑎𝑝 = 𝑛𝑝
𝐹ℎ − [𝑉𝐸𝐶𝐼 − 𝑉𝑒𝑎𝑟𝑡ℎ0 cos 𝑖]𝑇, if < 0 underlap
Lin et al., 18 Oct 2018 VCST/GEO 15
-
On-orbit focal length measurements
Lin et al., 18 Oct 2018
-1
-0.5
0
0.5
1D
NB
DN
B
DN
B
DN
B
M2
M4
M3 I1 I2 I2 I2 I2
M7
M7
M7
M7
M5 I3 I3 I3 I3
M10
M11 I5 I5 I5 I5
M15EF
L d
evia
tio
n (%
No
min
al)
Band Name
EFL on-orbit measurements for SNPP VIIRS
-1
-0.5
0
0.5
1
DN
B
DN
B
DN
B
M2
M4
M3 I1 I2 I2 I2
M7
M7
M7
M5 I3 I3 I3
M10
M11 I5 I5 I5
M16
M15EF
L d
evia
tio
n (%
No
min
al)
Band Name
EFL on-orbit measurements for J1/N20 VIIRS
VisNIR FPA LWIRS/MWIR
VisNIR FPA LWIRS/MWIR
VCST/GEO 16
• A +0.1% EFL change means +12 m change in scan-to-scan underlap where it exits.
See poster for details, including BBR
-
Future work1) Routine monitor and LUTs update as needed
2) VIGMU (VIIRS instrument geometric model update):
test production
3) Corrections for temporal pointing variations:
test production
4) Refresh ground control points with Landsat-8 images
5) Replace attitude in SC diary with Kalman filtered data
6) Replace ephemeris in SC diary with GPS data
7) Update DEM from 1 km to 500m resolution
8) Automate GEO LUT updates
9) Create Level-1 geolocation web (needed?)
Anything else?
Any change in priority order above?Lin et al., 18 Oct 2018 VCST/GEO 17
-
Conclusions
• SNPP VIIRS geolocation performance is good
– mean errors for I- & M-bands are ~ 10 m and uncertainties are ~ 60 m
at nadir, statistically
– DNB geolocation accuracy is only spot-checked and it is ok
– The attitude system underperforms geolocation error up to 7 km
occurred occasionally. Attitude HW needs maintenance, and their
output will be tweaked on the ground (re-working ephemeris and
attitude data)
• J1/N20 VIIRS geolocation performance is ok– Early on-orbit calibration reduced biases from O(1km) to O(10m), including DNB
– Correction for large pointing variation was worked, and will be operationalized soon
• J2 (N21) VIIRS geolocation is expected to perform fine
Lin et al., 18 Oct 2018 VCST/GEO 18
-
Questions?
Q: Do random variables exist?
A: Not really.
Lin et al., 18 Oct 2018 VCST/GEO 19
Q: What do we do?
A: Careful trending and de-trending.
-
Backup Slides
Lin et al., 18 Oct 2018 VCST/GEO 20
-
y = 2.691x - 28.462R² = 0.3732
-75
-37.5
0
37.5
75
12 13 14 15 16 17 18 19
Tra
ck (
ad
j.)
res.
(m)
.
Years since Jan. 1, 2000
16-day Global16-day Global Linear (16-day Global)
y = 0.382x - 1.8738R² = 0.0165
-75
-37.5
0
37.5
75
12 13 14 15 16 17 18 19
Scan
(ad
j.)
res.
(m)
.
Years since Jan. 1, 2000
SNPP Long-Term Trending and Correction
Lin et al., 18 Oct 2018
De-trending was worked and will be operationalized soon
Linear trend and annual cycle
Linear trend and weak annual cycle
VCST/GEO 21
-
SNPP sun angle dependence
Lin et al., 18 Oct 2018
y = 0.0012x2 - 0.3587x + 19.644R² = 0.4179
-112.5
-75
-37.5
0
37.5
75
112.5
0 30 60 90 120 150 180
Shift
(mete
rs)
Sun angle (degrees)
Track residuals (adjusted & demeaned) vs. Sun angle
y = 0.0016x2 - 0.3713x + 16.848R² = 0.4163
-112.5
-75
-37.5
0
37.5
75
112.5
0 30 60 90 120 150 180
Shift
(mete
rs)
Sun angle (degrees)
Scan residuals (adjusted & demeaned) vs. Sun angle
Correction is in-workVCST/GEO 22
-
SNPP orbit position errors (wrt GPS)
SNPP
2016-09-
21+22
Lin et al., 18 Oct 2018 VCST/GEO 23
Spec : 75 m, 3-sigma
Plot courtesy: Fred Patt
Spec : 75 m, 3-sigma
Correction is in-work
-
J01
2018-06-
10+11
Along-track pos err wrt gps
Cross-track pos err wrt gps
Spec : 75 m, 3-sigma
J1/N20 orbit position errors (wrt GPS)
Plot courtesy: Fred PattCorrection is in-workLin et al., 18 Oct 2018 VCST/GEO 24