Nighttime Cloud Microphysical Products With The VIIRS Day/Night Band

1. Andi Walther1, Steven Miller3, Andrew Heidinger2 1University of Wisconsin, Madison, WI 2NOAA/NESDIS/Center for Satellite Applications and Research, Madison 3CIRA, Fort Collins CO Nighttime Cloud Microphysical Products With The VIIRS Day/Night Band

2. You will hear something about: • The NOAA/NASA JPSS program and the VIIRS sensor with the Day/Night band channel • A lunar reflectance model • A new cloud retrieval: The Nighttime Lunar Cloud Optical and Microphysical Properties (NLCOMP)

3. Daytime retrieval (DCOMP) • Standard approach: simultaneous measurements in a visible and in an absorbing near-IR channel • Use of solar reflectance. Solar irradiance is known very exactly. • NIR channel can be in mixed solar/terrestrial region of spectrum around 3.8 micron • Forward model equation set for a given geometrical constellation neglecting atmospheric correction:

4. Daytime retrieval • Standard approach: simultaneous measurements in a visible and in an absorbing N-IR channel • Used in all satellite-based cloud data sets. • NIR channel may be in mixed solar/terrestrial region of spectrum around 3.8 micron • Forward model as bi-spectral reflectance image:

5. What about nighttime? • Current satellite-based cloud optical properties retrievals work only under daytime conditions with solar reflectance measurements. • Nighttime retrievals are difficult due to missing measurements in visible spectrum. (Split-window IR methods are sensitive only to thin cirrus; MW-based LWP retrievals have very coarse spatial resolution) • The new Day/Night band (DNB) of VIIRS sensor onboard Soumi-NPP is sensitive enough to measure also low-light radiance. • Converting upwelling radiance at the DNB sensor in lunar reflectance would enable us to retrieve cloud properties similar to daytime retrievals. • The new Nighttime Lunar Cloud Optical and Microphysical Properties retrieval (NLCOMP) is an extension of the Daytime COMP (DCOMP) retrieval to use moonlight reflectance for the first satellite-based quantitative cloud properties retrieval during night.

6. Some facts • Moon light is about 250 000 dimmer than sun (~10-5 W m-2 sr-1 μm-1 at full moon) • Current sensors (MODIS, AVHRR, etc..) in visible spectrum are only able to detect signals from around 100-102 W m-2 sr-1 μm-1 • DMSP-OLS offered low-light images, but the data were not calibrated, with low information depth (6-bit) and low spatial resolution. • DNB VIIRS onboard NPP-Soumi is the first channel which is both, highly sensitive to low-light in visible spectrum and providing a sufficient data depth (down to 10-5W m2 sr-1 as a band average with a 14-bit resolution) • DNB spatial resolution is uniformly 740m along and across the swath from nadir to the edge of the swath. • DNB has to be collocated with VIIRS M-band channels those pixels grow from nadir to the edge (up to 5 times larger pixels) for retrievals.

7. DNB radiance example image

8. From DNB radiance to moon light reflectance • The radiance to reflection retrieval was developed by Steven Miller and colleagues at CIRA • In contrast to solar irradiance the computation of down-welling lunar irradiance is a complex task due to many components which have to be considered: • Lunar phase • Lunar spectral surface albedo • Moon-Earth-Sun orbital geometry • Lunar zenith angle

9. From DNB radiance to moon light reflectance • The radiance to reflection retrieval was developed by Steven Miller and colleagues at CIRA • In contrast to solar irradiance the computation of down-welling lunar irradiance is a complex task due to many components which have to be considered: • Lunar phase • Lunar spectral surface albedo • Moon-Earth-Sun orbital geometry • Lunar zenith angle • .. From Miller and Turner 2009

10. From DNB radiance to moon light reflectance complex moon surface albedo • The radiance to reflection retrieval was developed by Steven Miller and colleagues at CIRA • In contrast to solar irradiance the computation of down-welling lunar irradiance is a complex task due to many components which have to be considered: • Lunar phase • Lunar spectral surface albedo • Moon-Earth-Sun orbital geometry • Lunar zenith angle • We expect an overall uncertainty in lunar reflection from 5% to 12%

11. Validation of lunar reflectance Lunar and solar reflection results for cloud-free scenes at the Salar de Uyuni salt flat (“Salzpfanne”) in Bolivia. Results show agreement which is consistent with assumed uncertainties of the lunar model.

12. Global coverage of calibrated lunar reflectance • Lunar cycle is about 29.5 days • Lunar reflectance requires filtering of solar zenith (19 below the horizon) • Sufficient global lunar reflectance coverage is ~60% of nighttime • Winter poles have coverage most of the time

13. The Nighttime Lunar Cloud Optical and Microphysical Properties (NLCOMP) retrieval • Is the nighttime adaption of the daytime equivalent DCOMP ( Daytime-COMP) • Retrieves Cloud Optical Thickness and Effective Radius, those can be used to derive cloud water path. • Input parameter: DNB visible lunar reflectance and M-12 brightness temperature • NLCOMP products has higher uncertainty than DCOMP due to higher uncertainty of lunar reflectance in contrast to solar reflectance. • Limitations: City lights, ships, diffuse lights, etc..

14. Daytime retrieval • Standard approach: simultaneous measurements in a visible and in an absorbing N-IR channel • Use of solar reflectance because solar irradiance is known very exactly • NIR channel can be in mixed solar/terrestrial region of spectrum around 3.8 micron • Forward model equation set for a given geometrical constellation:

15. Nighttime retrieval • Standard approach: simultaneous measurements in a visible and in an absorbing N-IR channel • Use of lunarreflectancebut lunar irradiance is not known very exactly • NIR channel can be in mixed solar/terrestrial region of spectrum around 3.8 micron, but no reflection value. • Forward model equation set for a given geometrical constellation:

16. Nighttime retrieval As Surface albedo from MODIS climatology Tsfc Surface temperature from NWP εsfc Surface emissivity from NWP Tc Cloud temperature from PATMOS-x ACHA cloud algorithm ( Heidinger et al. 2009) • Standard approach: simultaneous measurements in a visible and in an absorbing N-IR channel • Use of lunar reflectance but lunar irradiance is not known very exactly • NIR channel can be in mixed solar/terrestrial region of spectrum around 3.8 micron, but no reflection value. • Forward model equation set for a given geometrical constellation:

17. NLCOMP information content M-12 (3.8μm) channel M12 M12 Cloud transmission and emissivity have the tendency to counterbalance in respect to optical thickness.

18. NLCOMP information content: Thin clouds Thin clouds: Surface dominates the signal (thin cloud, warm surface, cold cloud) : The bigger the REF the smaller the radiance at sensor.

19. NLCOMP information content: Thick clouds Thick clouds: Cloud dominates the signal (thick clouds, warm clouds, cold surface): The bigger REF the bigger the radiance at sensor.

20. “blind spot” for REF retrieval In the transition zone between both regimes builds a “blind spot” where NLCOMP has no skill to retrieve REF. This position of the zone is shifted to thinner or thicker clouds depending on cloud and surface temperature The ‘blind spot’ COD can be determined during processing. Optimal estimation provides high uncertainty value.

21. NLCOMP results:COD on April 26 2013 IR-based DCOMP 06:30PM NLCOMP 01:30AM DCOMP 09:30AM LARC Shortwave Infrared Infrared Split Window Technique (SIST) algorithm (Minnis et al., 1998).

22. NLCOMP result example 17 January 2013 ( Full moon)

23. Cloud Water Path 11 μm BT Lunar Reflectance Solar Reflectance

24. Global results 27 Jan 2013

25. Global COD distribution

26. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night.

27. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night. • A lunar down-welling irradiance predictor was developed which enables us to use DNB lunar reflectance as an input of a nighttime cloud property algorithm NLCOMP.

28. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night. • A lunar down-welling irradiance predictor was developed which enables us to use DNB lunar reflectance as an input of a nighttime cloud property algorithm NLCOMP. • Comparisons to daytime results demonstrates consistency between DCOMP and NLCOMP for COD.

29. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night. • A lunar down-welling irradiance predictor was developed which enables us to use DNB lunar reflectance as an input of a nighttime cloud property algorithm NLCOMP. • Comparisons to daytime results demonstrates consistency between DCOMP and NLCOMP for COD. • Information content studies indicate that certain constellations of cloud and surface temperature posses a ‘blind spot’ region of COD, where a retrieval of REF is not possible or very limited.

30. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night. • A lunar down-welling irradiance predictor was developed which enables us to use DNB lunar reflectance as an input of a nighttime cloud property algorithm NLCOMP. • Comparisons to daytime results demonstrates consistency between DCOMP and NLCOMP for COD. • Information content studies indicate that certain constellations of cloud and surface temperature posses a ‘blind spot’ region of COD, where a retrieval of REF is not possible or very limited. • The current version of NLCOMP is limited to non-urban regions.

31. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night. • A lunar down-welling irradiance predictor was developed which enables us to use DNB lunar reflectance as an input of a nighttime cloud property algorithm NLCOMP. • Comparisons to daytime results demonstrates consistency between DCOMP and NLCOMP for COD. • Information content studies indicate that certain constellations of cloud and surface temperature posses a ‘blind spot’ region of COD, where a retrieval of REF is not possible or very limited. • The current version of NLCOMP is limited to non-urban regions. • NLCOMP is implemented in PATMOS-x cloud retrieval scheme

32. Summary • The new DNB channel of VIIRS offers observations of low-light signals in a visible channel during night. • A lunar down-welling irradiance predictor was developed which enables us to use DNB lunar reflectance as an input of a nighttime cloud property algorithm NLCOMP. • Comparisons to daytime results demonstrates consistency between DCOMP and NLCOMP for COD. • Information content studies indicate that certain constellations of cloud and surface temperature posses a ‘blind spot’ region of COD, where a retrieval of REF is not possible or very limited. • The current version of NLCOMP is limited to non-urban regions. • NLCOMP is implemented in PATMOS-x cloud retrieval scheme • NLCOMP will help to close the nighttime observation gap of cloud optical properties. This will be especially valuable in high latitudes winter where cloud observations are missed for longer periods (example: Alaska)

33. Next steps • Generation of static background signal maps to mask out cities. • Global maps of the “blind spot” where REF retrieval is limited. • Use of lunar reflectance to correct IR-based nighttime cloud mask for missed low-level clouds over land.

34. Acknowledgments AMS/EUMETSAT WIEN 2013 • NLCOMP work funded by NOAA JPSS Risk reduction project. • PATMOS-x originally funded by NOAA/NESDIS/STAR and the NOAA PSDI Program. • VIIRS work funded by the JPSS Cal/Val Program.

35. Thank you