1. Reduction of stripe noise in ACSPO clear-sky radiances and SST Marouan Bouali and Alexander Ignatov NOAA/NESDIS/STAR and CSU/CIRA 1 SPIE Security, Defense and Sensing 2013, Baltimore, MD

2. Motivation Native resolution SST imagery (≤1km) derived from whiskbroom scanners (MODIS, VIIRS) is affected with stripe noise 2 Sea Surface Temperatures (°C) Terra MODIS Aqua MODIS NPP VIIRS 05/01/2013 T2012294110000.L2_LAC_SSTA2012294031500.L2_LAC_SSTACSPO_V2.12_NPP_VIIRS_2012-10-20_1510-1519 SPIE Security, Defense and Sensing, Baltimore, MD

3. Accuracy of SST retrieval 3 NPP VIIRS (0.75 km) Stripe noise in level 1B or SDRs BTs can lead to SST errors of up to ± 0.3K SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2012-10-20_1510-1519

4. SST Fronts 4 NPP VIIRS (0.75 km) Sobel filter Striping introduces artificial structures and affects the analysis of thermal fronts (orientation, intensity and location) SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2012-10-20_1840-1849

5. 5 Prior to assimilating into UHR L4 SST, MODIS and VIIRS L2 SSTs should be destriped SPIE Security, Defense and Sensing, Baltimore, MD UHR L4 SST will be increasingly used as input in NWP models due to better weather prediction (LaCasse et al. 2008) ESA MEDSPIRATION NASA JPL MUR NASA JPL G1SST Ultra-High Resolution Level 4 SSTs 05/01/2013

6. Mitigation of striping 6  Currently, blackbody (BB) and space view (SV) are used for absolute BT calibration on a scan-by-scan basis  This practice ensures that sensor uniformity performance is within pre-launch specification BUT it does not guarantee full mitigation of striping effect  Post-processing of L1B (SDR) is required to generate improved L1B, from which SST products with improved image quality can be produced SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013

7. Destriping Literature 7  ~ 40 years of research (since Landsat MSS, 1972)  Techniques with potential for operational use are intended for highly pronounced stripe noise whereas,  Stripe noise in 11 & 12 µm bands (MODIS, VIIRS) is below 0.05 K..  No work has been done so far for stripe noise reduction in SST imagery other than…  Spatial smoothing of the atmospheric term in SST algorithm SPIE Security, Defense and Sensing, Baltimore, MD SST = a 0. BT 11 + a 1. BT 12 + a 2.SST ref.[(BT 11 -BT 12 ) * H](sec( Ɵ ) -1) Low signal-to-stripe noise ratio (SSNR) Low-pass filter 05/01/2013

8. Towards Operational Destriping 8 Requirements Automatic: Minimize supervision and algorithm tuning Robust: Maximize improvement of image quality, i.e., minimize:  Residual stripes  Distortion  Processing artifacts Fast: Suitable for real-time processing of satellite data. SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013

9. Adaptive Destriping 9 Scene-based denoising algorithm based on  Directional Hierarchical Decomposition (DHD) using on a unidirectional quadratic variational model  Nonlocal filtering “Adaptive Reduction of Striping for Improved SST Imagery from S-NPP VIIRS”, JTech, 2013 (in review) SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013

10. Adaptive Destriping 10 Directional Hierarchical Decomposition SPIE Security, Defense and Sensing, Baltimore, MD Observed image (f) Horizontal gradient Vertical gradient 1 st guess “True Image” (u 0 ) 1 st guess noise (v 0 ) 05/01/2013 = +

11. Adaptive Destriping 11 SPIE Security, Defense and Sensing, Baltimore, MD u0u0u0u0 f-v 1 v0v0v0v0 v1v1v1v1 v2v2v2v2 v3v3v3v3 v4v4v4v4 v5v5v5v5 v6v6v6v6 f, Noisy image f-v 2 f-v 3 f-v 4 f-v 5 f-v 6 05/01/2013 Iteration #1 #2#3#4#5#6

12. Preliminary Results: Data 12  Data: 3 days of level 1B/SDR TOA calibrated BTs -Terra MODIS -Aqua MODIS -NPP VIIRS  Destriping algorithm applied to SST bands @ 3.7, 11 and 12µm  Destriped BTs used as input in ACSPO prior to cloud masking and SST production  Cloud mask and SST image quality compared with/without destriping SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013

13. SST image quality (0.75 km) 13 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-27_1910-1919 (from original SDRs)

14. SST image quality (0.75 km) 14 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-27_1910-1919 (from destriped SDRs)

15. SST Fronts (0.75 km) 15 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-27_1910-1919 (from original SDRs) 05/01/2013

16. SST Fronts (0.75 km) 16 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-27_1910-1919 (from destriped SDRs) 05/01/2013

17. SST image quality (0.75 km) 17 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-24_0300-0309 (from original SDRs) 05/01/2013

18. SST image quality (0.75 km) 18 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-24_0300-0309 (from destriped SDRs)

19. SST Fronts (0.75 km) 19 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-24_0300-0309 (from original SDRs) 05/01/2013

20. SST Fronts (0.75 km) 20 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-24_0300-0309 (from destriped SDRs)

21. SST image quality (0.75 km) 21 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-21_0740-0749 (from original SDRs) 05/01/2013

22. SST image quality (0.75 km) 22 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-21_0740-0749 (from destriped SDRs)

23. SST image quality (4 km) 23 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-21_0740-0749 (from original SDRs) 05/01/2013

24. SST image quality (4 km) 24 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-21_0740-0749 (from destriped SDRs)

25. SST Fronts (4 km) 25 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-21_0740-0749 (from original SDRs) 05/01/2013

26. SST Fronts (4 km) 26 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-21_0740-0749 (from destriped SDRs)

27. Impact on cloud mask 27 SPIE Security, Defense and Sensing, Baltimore, MD ACSPO_V2.12_NPP_VIIRS_2013-01-31_0600-0609 (from original SDRs) 05/01/2013

28. Impact on cloud mask 28 SPIE Security, Defense and Sensing, Baltimore, MD 05/01/2013 ACSPO_V2.12_NPP_VIIRS_2013-01-31_0600-0609 (from destriped SDRs)

29. Destriping performance 29 The Normalized Improvement Factor (NIF) quantifies the improvement in image quality SPIE Security, Defense and Sensing, Baltimore, MD Monitoring of the NIF index over a 3 day period indicates stable destriping performance (5% ≤ NIF ≤ 30%) 05/01/2013

30. Conclusion 30 SPIE Security, Defense and Sensing, Baltimore, MD  Stripe noise is clearly visible in MODIS and VIIRS level 2 SST and cloud mask  It introduces relative errors of up to 0.3 K at the pixel level .. and compromises downstream SST applications - classification/clustering, cloud mask, thermal fronts detection…  Downsampling SST from 1 km (or 0.75 km for VIIRS) to 2/4 km resolution does not really mitigate the problem…  On-orbit calibration/characterization can reduce striping but cannot remove it fully, due to numerous factors contributing to stripe noise  Scene-based post-processing to reduce stripe noise is the only practical approach for improved SST imagery 05/01/2013

31. Current status and future work 31 Currently Rotational buffer of destriped VIIRS SST SDRs (M12, M15, M16) Rotational buffer of ACSPO VIIRS SST with destriped BTs Future work Estimate SST coefficients from destriped BTs Generate ACSPO SST with new SST coefficients Evaluate quantitative impact on SST retrieval. Destriping VIIRS/MODIS SST 05/01/2013

32. Thank you! Questions? 3205/01/2013

33. MODIS and VIIRS SST bands 33 Terra /Aqua MODIS Resolution at Nadir1 km FPA -> 10 Detectors S/MWIR λ (µm) NE∆T (K) Band 20 3.7 0.05 Band 223.9 0.07 Band 234 0.07 LWIR λ (µm) NE∆T (K) Band 31 11 0.05K Band 32 120.05K Suomi NPP VIIRS Resolution at Nadir0.75 km FPA -> 16 Detectors S/MWIR λ (µm) NE∆T (K) M12 3.7 0.40K LWIR λ (µm) NE∆T (K) M15 110.07K M16 120.07K Destriping VIIRS/MODIS SST 05/01/2013