STAR VIIRS SDR Cal/Val Overview

1. STAR VIIRS SDR Cal/Val Overview Quanhua (Mark) Liu, Slawomir Blonski, and Changyong Cao with great contributions from Fuzhong Weng, Sirish Uprety, Scott Hu, Weizhong Chen, Dave Pogorzala, Sean Shao, and Yan Bai October 24, 2012

2. Progresses since VIIRS beta Maturity • VIIRS F factor Generation and Monitoring • VIIRS Calibration Data Analyze • Inter-comparisons between VIIRS, MODIS, AVHRR, and CrIS • Quantify Power Outage Using DNB • New Algorithm and First VIIRS Moon Temperature Image • Online VIIRS Global Quickview • Support EDR teams for their VIIRS Satellite Productions • VIIRS Observed Tropic Cyclones without Gap • VIIRS Gallery Enhancement

3. VIIRS SDR Tasks Assigned to NOAA/STAR Cao (management lead) and De Luccia (technical lead) coordinate and execute the 58 VIIRS SDR cal/val tasks defined in the OPSCON.

4. Interactions with Users, SDR and EDR Teams • Worked closely with EDR teams (e.g. SST, Ocean Color, fire and vegetation, etc EDR teams), NCEP, and JCSDA. • Established and maintained the SDR/EDR web blog on CasaNosa. • Got help from EDR teams and users, e.g., JCSDA (CRTM model), SST EDR team (VIIRS BT anomaly during BB WUCD), ocean color team (gain mismatch), fire detection team (M13 low gain calibration, BBR). • Supported the EDR teams and our users, for example, we helped JCSDA correct the error in using VIIRS RSR.

5. Moon’s Irradiance Observed from NPP VIIRS on May 31, 2012 and Comparison with Model (RAD 7) VIIRS M1 Channel + VIIRS M1-11 channels VIIRS M1 Channel

6. VIIRS and CRTM Modeling for M12 Striping Investigation (RAD 7) M1, M4, and M11 measured (R-Rm)/Rm *100 The STAR team applied the CRTM to simulate the VIIRS SDR data. It is found that the M12 striping reported by the SST EDR team is caused by the difference in VIIRS azimuth angles among detectors.

7. VIIRS – MODIS SNO Prediction Table of predicted SNOs for the next 14.0 days since TLE Epoch: 2/11/2012 • Based on information from the NOAA / STAR / NCC SNO (simultaneous nadir overpass) prediction website, https://cs.star.nesdis.noaa.gov/NCC/SNOPredictions, included all SNO datasets acquired by Suomi NPP VIIRS and by MODIS from both Aqua and Terra during six months since mid-February 2012 • The SNOs occur over snow-covered Antarctica (some at the Dome C site), providing bright surfaces in the VisNIR bands, as well as over northern Alaska, Canada, Greenland, Scandinavia, Siberia, and ocean (both dark and bright scenes) Suomi NPP and Terra SNO Example Suomi NPP and Aqua SNO Example VIIRS MODIS

8. SDR Comparison with AVHRR (RAD 8) VIIRS and AVHRR TEBs agree (~ 0.3 K). VIIRS and AVHRR RSB agree with the slope. Large bias in RSB needs to be further investigated.

9. RAD 9: Aqua vs. Terra in SNO Data • Compared TOA reflectance measured by VIIRS and MODIS during the SNO events between Suomi NPP and either Aqua or Terra • For the terrestrial ecology bands (used in NDVI calculations: VIIRS bands I1 and I2, MODIS bands 1 and 2): • There is no bias when comparing NPP VIIRS band I1 with Aqua MODIS band 1 • There is only small bias (~2%) between VIIRS band I2 and Aqua MODIS band 2 • When VIIRS is compared with Terra MODIS, the biases are larger (~4%): • There are clearly biases between MODIS bands 1 and 2 measurements from Aqua and Terra (similar biases occur for other bands as well) • Collection 5 MODIS Level 1B data were used for the Aqua vs. Terra comparison because Collection 6 products are not yet available for Terra MODIS

10. RAD 9: MODIS Collection 5 vs. Collection 6 • Aqua MODIS radiometric calibration has been recently improved in production of Collection 6 datasets • The largest change has occurred for bands 8 and 9 that are comparable with VIIRS bands M1 and M2, respectively • When, instead of Collection 5 data, Aqua MODIS Collection 6 data are used in SNO comparisons with VIIRS : • M1 bias is reduced from +4% to -1% • M2 bias is reduced from 1% to near zero • Observed temporal variation of the M1 bias may be due to VIIRS polarization sensitivity (will investigate) • 6Sv radiative transfer modeling conducted for VIIRS band M1 (including out-of-band response) and for MODIS band 8 (using a snow surface reflectance and a range of atmospheric conditions) agrees better with the Collection 6 data

11. RAD 9: SNO Comparisons for Other Bands • VIIRS band M3 measurements agree very well with Aqua MODIS band 10 Collection 6 data: no bias is observed • VIIRS band M6 and MODIS band 15 data are often saturated in SNO observations: • A few data points collected during the last six months show no bias (with large uncertainty) • VIIRS band M7 shows 2% bias versus Aqua MODIS band 2 (similarly to band I2 which has almost the same spectral response) • Bands M7 and I2 are the most affected by the VIIRS telescope throughput degradation, but the bias remains stable (within uncertainty of the SNO measurements) thanks to weekly updates of the radiometric calibration coefficients

12. RAD 9: VIIRS band M5 vs. MODIS band 1 • In SNO observations, there is a large bias (~9%) between VIIRS band M5 and MODIS band 1 • Spectral responses of MODIS band 1 and VIIRS band M5 are quite different (unlike for band I1 which shows no bias from MODIS band 1) • Although the average bias is ~9%, the bias changes with time between 5-6% and ~18% • The temporal dependence is correlated with the solar zenith angle (SZA) changes • 6Sv radiative transfer modeling (using a snow surface reflectance and a range of atmospheric conditions) predicts a very similar SZA dependence for the SNO observations

13. RAD 9: Summary • SNO comparisons with MODIS have shown that: • The implemented regular updates of the radiometric calibration coefficients have stabilized the VIIRS radiometric calibration • Biases between the Suomi NPP VIIRS and Aqua MODIS measurements in reflective solar bands are small for bands with similar spectral responses:

14. PTT1: Operability, Noise, SNR Verification • Objective is to verify detector operability, noise, and SNR of all detectors in all Reflective Solar Bands and gain states, and to verify the absence of artifacts in calibration view data • Task is repeated through the life of the sensor to monitor performance • Mean and noise are calculated for raw counts (expressed as the DN) for all calibration views (OBC BB, SD, SV) • Mean and noise for offset corrected counts for illuminated SD views are also calculated • SNR is calculated for SD data (when SD is illuminated by the Sun) Orbit 5000 Oct 14, 2012 SD View SV, BB View

15. PTT1: Mean DN/Gain Time Series Vis Band: SD degradation (with unexplained change since July) NIR Band: RTA degradation SWIR Band: RTA degradation with Safe Mode aftermath (and unmarked yaw maneuver) Asymmetric changes during the Suomi NPP first year on orbit

16. PTT1: Predictions of RTA Degradation • Band M7 • Detector 8 • High gain • HAM A Error estimates with weekly LUT updates Good agreement between the predicted rate of calibration coefficient changes and the one from the F-factor look-up tables used in the operational processing, despite recent, minor differences • Multi-exponential non-linear optimization: • Best fit with 3 exponential terms (versus 2-exp and 4-exp: 4th term negligible) • Degradation time constants: 1, 13, and 139 days with the weights of 7%, 23%, and 70%, respectively • Final gain: 66% of initial value

17. Desert Ocean • Bias trends at Desert and Ocean are consistent for matching VIIRS and MODIS channels. • M5 shows the largest bias (~ 8%) which can be explained due to RSR differences between MODIS and VIIRS channels • Most of the VIIRS channels agree with MODIS within 2% after accounting for spectral differences except channel M2 which needs more investigation.. • Observed Bias = (V/M - 1) ×100%

18. VIIRS and MODIS Comparison (Feb. 25)(PTT-3) Bias = MODIS - VIIRS

19. PTT-4: DNB Image DNB image quality was investigated and a high quality DNB image was sent to JPSS program office per request.

20. STAR PRT Temperature Monitoring (PTT-5, PTT-7)

21. PTT6: RADIOMETRIC PERFORMANCE MONITORING

22. PTT-7: Telemetry Trending and Monitoring Objective: To monitor telemetry parameters. Methods and Tools: OBC-IP reader, VIIRS LUTs, long-term monitoring system Results and Recommendations: STAR monitoring system provided visualized figures for VIIRS telemetry parameters. Black solid and dashed lines are for measured values at HAM A and B sides. Lines in red are predicted based on single operational BB temperature (see green triangle).

23. VIIRS GLOBAL QUICKVIEW • A VIIRS RGB composite image is generated from M3, M4, and M5 bands. This VIIRS global quick-view image serves as a fast check of the VIIRS data quality. Users can also use the image to select their interested scenario: a granule for ocean, land, clear sky, and clouds. VIIRS M15 NEdT for HAM A and B sides. Significance: STAR is closely monitoring the VIIRS performance and serves users.

24. New method to verify VIIRS geolocation at SNOx - STAR • STAR team used the SNO prediction to investigate the geolocation consistency between the VIIRS and MODIS. The differencing animation image on the upper right not only shows cloud movement, but also geolocation discrepancy for land features. • The VIIRS geolocation discrepancy was reported. • After the VIIRS new geolocation LUT was implemented, the “land movement” issue was resolved (see lower-right image). • The VIIRS geolocation error is generally less than 100 meter (Wolfe et. al., Apr. 17, 2012). Differencing image shows geolocation discrepancies for land features (20:55 Dec. 20, 2011) Animation image at 18:05 Feb. 25, 2012

25. Scientific AdvancementsQuantify power outage using DNB • Despite the straylight effect, the Day/Night Band has been used to detect a major power outages in the Washington, DC on the night of June 29, 2012. • An analysis of the data after the storm showed that most areas had power restored within 3 days. Blue: mostly clear sky Red: Cloudy ROI: (101 by 71 pixels) ~ 64 km × 64 km 07/06 06/29 07/02 06/23 06/30 VIIRS DNB of the Washington/Baltimore area on June 26th (top)and June 30th. The suburbs west of DC and Baltimore, in particular show dark areas. VIIRS DNB radiance time series before and after the power outage (6/29) shows that most of the power was restored in three days.

26. VIIRS Image about Moon Temperature VIIRS can occasionally observe moon through the space view window. Since gain is stable, VIIRS radiance for moon can be represented by (1) Although Cold_space_count is not available, we can use radiance for BB (2) subtracting Eq.(1) from Eq.(2) to derive the radiance for moon (3)

27. VIIRS observed tropic cyclones without gap • Using the wide swath and the high spatial resolution of the VIIRS, STAR team found • the landing of the typhoons Saola and Damrey, and the development of the typhoon Haikui. Damrey Haikui Saola Significance: VIIRS SDRs has wide a swath and high spatial resolution, uniquely for monitoring global tropic cyclones without gap.

28. Summary • STAR ICVS online monitors VIIRS including F factor generation and monitoring. • VIIRS, MODIS and AVHRR agree well over the SNO scenes and desert. • VIIRS and CrIS cross check can be performed at any time and any location. Both sensors for TEB agree, but a scene-dependent bias for M15 is still under investigation. • Progress made for DNB and DNB can be used to quantify city light and power outage. • Good Quality of the First VIIRS Moon Temperature Image. • Moon irradiance model works well for VIIRS moon observations. • VIIRS M12 striping at daytime is mainly caused by the difference in VIIRS azimuth angles among detectors. CRTM can simulate the striping effect. • More high-quality VIIRS images in Gallery.