1. NASA NPP-SDS VCST NPP VIIRS On-Orbit Calibration Using the Moon J. Fulbright a, Z. Wang a, and X. Xiong b a SSAI (formerly with Sigma Space), Greenbelt, MD, USA b Sciences and Exploration Directorate, NASA/GSFC, Greenbelt, MD, USA VIIRS Characterization Support Team EUMETSAT, Darmstadt, 1-4 December 2014

2. Summary VIIRS Instrument Description NPP Scheduled Lunar Observation Events Imaging method for radiometric results –Good scan detection –Results and comparison to SD gain data Comparison to GIRO Results Summary See: Fulbright et al. (2014; SPIE-San Diego), J. Sun et al. (2013; SPIE-San Diego) and previous VCST and MCST papers. 2

3. 3 Visible Infrared Imaging Radiometer Suite (VIIRS) Key instrument on S-NPP and future JPSS satellites –S-NPP launched on October 28, 2011 –JPSS-1 launch in 2017 Strong MODIS heritage –Spectral band selection –On-board calibrators –Operation and calibration Strategies for planning/scheduling Data analysis methodologies/tools Background

4. Suomi-NPP Key Parameters SNPP is on a polar SSO orbit, ~100 min period, ~14.7 rev/day VIIRS has three types of bands: –Imaging bands (“I-bands”) with 375-m nadir resolution. >32 Detectors per band >3 RSB I-bands, 2 TEB I-bands –Moderate-resolution bands (“M-bands”) with 750-m nadir resolution. >16 Detectors per band >11 RSB M-bands, 5 TEB M-bands –Day-Night Band with 750-m nadir resolution >Panchromatic (500-900 nm) with a radiance dynamic range from 2x10 -2 W/m 2 /sr We will only talk about the 14 RSB I- and M-bands here. 4

5. 5 Spectral Bands 1 DNB 14 RSB (0.4-2.3  m) 7 TEB Dual Gain Bands: M1-M5, M7, M13 16 Moderate (radiometric) bands, 5 Imaging bands, 1 DNB

6. VIIRS On-board Calibrators (MODIS Heritage) Rotating Telescope Aft Optics and HAM Blackbody Solar Diffuser Solar Diffuser Stability Monitor Extended SV Port (Lunar Observations) 6

7. VIIRS Scan Description 7 During each scan, the VIIRS rotating telescope assembly (RTA) passes over four data sectors: BB = Blackbody; used for TEB calibration. SD = Solar Diffuser view; used for reflected solar band (RSB) calibration. SV = Space View; used to find the dark view values for all bands (at -65.675° off nadir). Moon observed at this position. 2.5 ms required to scan SV EV = Earth View, the main data sector (±65° off nadir). SV SD EV BB 1.7793 s/scan

8. 8 VIIRS RSB Calibration via SD The primary calibration constant for the Reflected Solar Bands (RSB) for VIIRS is the F-factor: VIIRS uses sunlight scattered off the Solar Diffuser as the known source to determine the F-factor: where: H(B,t) is the degradation factor of the SD as a function of band, B, and time, t. SD observations are made once per orbit. ATBD, Eq. 68

9. 9 VIIRS RTA Mirrors received an unintended contamination of WO x. When exposed to solar UV, this causes a λ-dependent optics degradation. Some impact on OOB: ↓ Small impact on OOB: ↑ Modulated RSR and Calibration Impact

10. Lunar Observation Description 10

11. Lunar View Scheduling 11 VIIRS Roll Maneuver - Roll Angle: -14 o to 0 o - Phase Angle: -56 o to -55 o ; -51.5 o to -50.5 o J. Sun and X. Xiong, Solar and Lunar Observation Planning for Earth-Observing Sensor, Proc. SPIE 8176, no. 817637, 2011. The Moon is visible in the VIIRS Space View sector when: The Moon is crossing the yz-plane of VIIRS: x = velocity direction, z = nadir direction, y = orthogonal to both x and y The angle of the lunar vector from the y-axis is over 24.325°. SNPP can be rolled over so that the SV is pointed closer to the Earth, but not the other way. The lunar phase angle is -51° ± 1°. If the Moon is within 1° of the SV center, no roll is required. Lunar Observation Opportunities - Scheduled: Nine months each year; day side - Unscheduled: Nine months each year with two at day side and two at night side; Partial moon VIIRS Moon

12. S-NPP Scheduled Lunar Observations Date (M/D/Y) UT Time Roll Angle Phase Angle Sector 01/04/201208:48:53-9.490-55.41SV 02/03/201204:21:32-5.445-56.19SV 02/03/201206:03:34-5.279-55.38SV 04/02/201223:05:11-3.989-51.24EV 05/02/201210:20:06-3.228-50.92EV 05/31/201214:47:14-0.081*-52.97EV 10/25/201206:58:15-4.048-51.02EV 11/23/201221:18:20-9.429-50.74EV 12/23/201215:00:50-7.767-50.90EV 01/22/201312:13:35-3.383-50.81EV 02/21/201309:31:25-1.712-50.71EV 03/23/201303:29:00-3.320-51.15EV 04/21/201319:47:54-3.882-50.82EV 05/21/201308:43:15-0.809*-50.67EV 10/14/201321:39:19-1.305-50.95EV 11/13/201306:57:41-7.981-50.66EV 12/12/201319:35:46-9.438-50.39EV 01/11/201409:59:45-6.727-51.30EV 02/10/201405:34:12-3.714-51.03EV 03/12/201401:11:43-3.944-51.05EV 04/10/201420:53:17-4.977-50.60EV 05/10/201413:13:00-4.177-50.91EV 06/09/201403:48:420.301*-51.05EV 10/04/201417:29:100.696*-50.81EV 11/03/201401:07:34-6.090-50.52EV 12/02/201408:41:44-10.841-50.83EV 26 orbits with a scheduled lunar observations. 23 events with the Earth View sector rotated electronically to be centered on the Space View. Earth View Sector data is co- registered, which allows us to get all bands in one viewing event. Earth View Sector data is aggregated for some bands, so there is a loss of spatial resolution. > 75 “free” (serendipitous) lunar observations not included here. 12 * = No Roll Maneuver Required

13. VIIRS Lunar Images 13 I1I4I2 M1M2M6 Scheduled Lunar Calibration: 20130221.0931 SV scan time  2.5 ms

14. Data Processing Methodology 14 The VIIRS irradiance is calculated by: N: the number of scans with complete lunar images c: pre-launch detector gain coefficients s: scan number dn: detector response after background correction and so-called detector-difference correction Ω: the IFOV of the detectors Good scans determined by a method that ensures the Moon is fully within the viewing area: The bright lunar edge is determined through the use of a Sobel filter. The edge pixels are fit to find a the lunar center and radius. A scan is only used if the whole Moon (plus a margin), illuminated or not, is within the data area. We find 6 to 13 good scans per event (maximum for December events).

15. Example of Lunar Center/Edge Detection 15 Black = Lunar dn contours Red = Lunar edge pixels (from Sobel filtering) Green = Lunar center (X), edge (bold), and edge + N pixels (dotted, N = 1-4) R Moon ~ 10 pixels (375-m I-band)

16. Lunar Images at Center Scans 16

17. Dark Correction Method 17 Dark Correction For each detector, we use the DN values for the samples on both sides of the lunar image (yellow; approximate position shown here). The mean value of the “dark” samples is then subtracted off the Moon region (blue). Sample Detector

18. Integration Method 18 Lunar Irradiance The lunar irradiance is the sum of the radiance values for all the dark-corrected pixel (dn) values in the Moon region (blue). The quadratic used in the radiance calculation is modified for “nearly dark” pixels (dn < 4). In that case, we set c 0 = 0. This helps prevent the deep space pixels from affecting the result while allowing scattered light inclusion. Sample Detector

19. Lunar F-factor Calculation The lunar F-factor is the ratio of the irradiance predicted by the USGS ROLO model and the VIIRS irradiance derived using pre- launch coefficients: The ROLO model result is generated by Tom Stone from USGS, with input photometric parameters and detector RSRs provided by VCST. Currently, two sets of irradiances are calculated using both pre-launch RSRs and RTA-degradation modulated RSRs. Only modulated RSRs results are used here. VCST uses lunar observation results as an independent confirmation of relative trending, not absolute calibration. 19

20. VIS/NIR Bands: SD vs. ROLO 20

21. VIS/NIR Bands: SD vs. ROLO 21

22. SWIR Bands: SD vs. ROLO 22

23. VIIRS Data and GIRO 23

24. Using VIIRS Data in GIRO For simplicity, each good VIIRS scan is put into an individual lunar observation file for GIRO. Contains data for all 14 RSB. A new SRF file is created for each lunar observation event to account for the modulated SRF. Homemade scripts automatically do file handling. The radiance imagette placed in the lunar observation file uses “prelaunch” values (the same as assuming F = 1). We do not need to use the oversampling factor based on our methodology. Successfully using GIRO v0.2.4. “Unscheduled” observations not considered yet, nor DNB. 24

25. ROLO Result Comparisons 25 The raw results (left) show disagreements of a few percent, but after normalization to the first event (right), the trending is very good).

26. F Moon_VCST /F moon_GIRO Comparison 26 The raw results (left) show disagreements of a few percent, but after normalization to the first event (right), the trending is very good). Band M1 (blue; 412 nm) shows trend of ~0.8% with time.

27. Summary and Conclusion 27 S-NPP/VIIRS has successfully observed the Moon for comparison with SD-derived gains for the last 3+ years. Seasonal variations to both SD gains and lunar gains and both are under investigation. We have successfully generated GIRO-compatible lunar observation and SRF files and have generated GIRO outputs. The GIRO results compare well (better to 1% after renormalization) to the ROLO results sent to VCST by Tom Stone directly. Future: Include DNB, but it includes data from three gains states. Create single-detector images (requires oversampling correction).

28. 28

29. Corrected VIIRS ROLO Model Comparison Fulbright (VCST)

30. ROLO Result Comparisons (OLD DRIVER!!) 30 The raw results (left) show disagreements of a few percent, but after normalization to the first event (right), the trending is very good).

31. ROLO: Old Driver vs. New Driver

33. New Driver ROLO Results (from Tom Stone) used this time! M11 2.2µm Using GIRO 4 th Release

34. New Driver ROLO Results (from Tom Stone) used this time! Using GIRO 4 th Release