1. All-Sky Microwave Radiative Transfer Modeling for DA: Advancing the CRTM to Microphysics-Consistent Cloud Optical Properties JSDSA Satellite Data Assimilation Summer Colloquium 5 August 2015 Scott Sieron (email: sbs5130@psu.edu) Advisor: Fuqing Zhang (Penn State) Major Collaborators: Eugene Clothiaux (Penn State), Lu Yinghui

2. Self Introduction 2009 – 2013 B.S. + M.S. Meteorology, The Pennsylvania State University 2013 – present PhD Dept. of Meteorology, The Pennsylvania State University Ongoing project began summer 2014

3. Quick Aside: M.S. Project, CloudSat Hurricane Overpasses Cloud top height of eyewall and near-storm convection to diagnose cyclone intensity (Wong and Emanuel 2007) Publication did not include CloudSat: thesis concluded that the vertical cross-section was insufficient sampling Sieron, S. B., F. Zhang, and K. A. Emanuel (2013), Feasibility of tropical cyclone intensity estimation using satellite-borne radiometer measurements: An observing system simulation experiment, Geophys. Res. Lett., 40, 5332–5336, doi:10.1002/grl.50973.10.1002/grl.50973 CloudSat overpass through eye and eyewall of Typhoon Dolphin. Image courtesy nasa.gov

4. Background: Microwave Imaging channels: small clear-air opacity, signal dominated by surface and hydrometeors At higher frequencies (>~50 GHz), Water surface has high emissivity (high T B in clear air) Hydrometeor impact dominated by snow/graupel/hail scattering At lower frequencies (<~50 GHz), Water surface has low emissivity (low T B in clear air) in H-polarization Hydrometeor impact dominated by rain absorption/emission, may be augmented by ice scattering Primary O 2 Absorption Bands Primary H 2 O Absorption Band Temperature Sounding Channels Moisture Sounding Channels Imaging Channels Imaging Channel Clear-Air Atmospheric Opacity vs. MW Freq. (AMSU channels demarked)

5. COLD WARM 270250 230 210 190 Hurricane Karl 09/17/10 0113Z (SSMI/S image courtesy NRL) COLD WARM 260240220 200 180 160 Hurricane Karl 09/17/10 0113Z (SSMI/S image courtesy NRL) High-mid frequency (91 GHz) Low-mid frequency (37 GHz)

6. Background: Data Assimilation of Microwave Radiances Global DA uses sounding channels: informative of vertical profile of temperature and moisture in clear (and cloudy) sky There is potential for value in regional-scale (hurricane) DA of precipitation information from imaging channels Can our observation operator (Community Radiative Transfer Model, CRTM) represent the radiance impacts of the hydrometeors with sufficient accuracy for DA? Want to avoid (a high magnitude of) bias correction If not, then could the process be beneficial to the CRTM and the forecast model?

7. About Clouds and Precipitation in CRTM CRTM clouds are specified by hydrometeor type (cloud water, cloud ice, rain, snow, graupel, hail) Radiative properties are calculated for spheres; snow and graupel are represented as “soft spheres” with densities < 917 kg/m 3 amount of hydrometeor (vertically-integrated mass per volume) size of hydrometeors (effective radius) Radiative properties are contained in lookup tables Have dimensions of cloud effective radius and (for liquid) layer temperature

8. Microwave and Precipitation Hydrometeor size is very important in microwave: When [particle radius] < ~1/6 wavelength, scattering increases by ~[particle mass] 2 Rayleigh scattering of a homogenous sphere Considering spherical solid particle of ever-increasing size: scattering per mass growth slows, oscillates, then declines Mie scattering of a homogenous sphere These MW wavelengths are only several millimeters Largest precipitation particles exceed ~1 mm radius and are removed from well- behaved scattering regime Mass extinction (thick solid), scattering (dashed) and absorption (thin solid) coefficients (m 2 g -1 ) of solid ice spheres as a function of radius for three imaging channels. Wavelength and 1/6-wavelength demarked.

9. Microphysics Scheme Details, Example

10. Testing the CRTM, All-sky Microwave DA – WRF Simulations Hurricane Karl, initialized at 21Z 16 Sept. from EnKF analysis after assimilating airborne Doppler radar radial velocities Same as Masashi’s experiments WRF version 3.6.1 (Skamarock et al. 2008) PSU WRF-EnKF: Zhang et al. (2009); Weng and Zhang (2012) Ensemble size: 60 WSM6 microphysics (5 species, 1 moment) 3 hour forecast

11. 260240220200180160 270250230210190 37 GHz 89 GHz Observations (SSMI/S)

12. 260240220200180160 270250230210190 36.5 GHz89 GHz WSM6, Particle Size Distribution Means as CRTM Cloud Effective Radii

13. WSM6 Scheme Note: CRTM assumes cloud ice is sufficiently small so as to not scatter, which is an invalid assumption for the sizes seen here Mean particle radius (microns)

14. WSM6, Specified and Uniform CRTM Radii 260240220200180160 270250230210190 36.5 GHz89 GHz Cloud: 15 μm Rain: 500 μm Ice: 50 μm Snow: 1000 μm Graupel: 1000 μm

15. Testing the CRTM, All-sky Microwave DA – First Attempts Pre-specified radii: unacceptable Relatively ad-hoc Simply not representing enough physics to be comfortable for DA Mean radius: too warm, too little scattering Mean particle radius of a cloud < effective scattering particle radius of a cloud because scattering is dominated by the large particles Mean of appropriately transformed distribution could produce better results, but… It often exceeds the CRTM lookup table effective radius dimension (1500 μm) At these wavelengths, the D 6 scattering relationship often breaks down for large particles, so using this transform will lead to over-estimated scattering

16. Testing the CRTM, All-sky Microwave DA – Next Efforts Create new cloud optical property lookup tables Model properties of single particles as specified by MP scheme Maxwell-Garnett mixing formula for ice dielectric constants (Turner et al., in prep) Product of the Henyey-Greenstein and Rayleigh scattering phase functions, and Legendre coefficients thereof, as specified by Liu and Weng (2006) Calculate per-mass optical properties of clouds constructed with particle size distribution as specified by MP scheme (We allow for scattering by cloud ice) MP scheme will be perfectly and natively interfaced with CRTM (Though both the MP schemes and CRTM remain a source of error/bias)

17. Testing the CRTM, All-sky Microwave DA – Next Efforts Build lookup tables for multiple MP schemes: WSM6 (Dudhia et al. 2008) Goddard (Lang et al. 2007) Morrison (Bryan and Morrison 2012) Modify CRTM source codes accordingly Redo WRF simulation with these MP schemes, compare: hydrometeor concentration and particle sizes resulting forward CRTM simulations Using 16+2 streams at all locations (removing effective radius broke the Mie parameter stream determination method)

18. 260240220200180160 270250230210190 36.5 GHz89 GHz WSM6, Particle Size Distribution Means as CRTM Cloud Effective Radii

19. 260240220200180160 270250230210190 36.5 GHz89 GHz WSM6, Particle Size Distribution Means as CRTM Cloud Effective Radii

20. WSM6, New Look-up Tables 260240220200180160 270250230210190 89 GHz36.5 GHz

21. Goddard, New Look-up Tables 260240220200180160 270250230210190 89 GHz36.5 GHz

22. Morrison, New Look-up Tables 260240220200180160 270250230210190 89 GHz36.5 GHz

23. Results and Discussion Scheme-specified cloud optical properties: too cold, too much scattering Consistent with many studies involving radar, and passive microwave using the simpler Goddard-SDSU radiative transfer solver [Zupanski et al. 2011; Zhang et al. 2013; Han et al. 2013; Chambon et al. 2014] Conclusion: too much or too big of snow and/or graupel in upper troposphere Using fewer than 16+2 streams in CRTM leads to not-as-cold brightness temperatures Simulations with only rain + cloud water (emitters) are very similar Goddard has most snow and graupel, also has substantial cloud ice scattering Morrison is heavier on snow, lighter on graupel WSM6 is lighter on snow, heavier on graupel Graupel stays near convective cells, creates very cold splotches Snow spreads out

24. Future Certain: I’m continuing PhD work on this project Uncertain: What work to be done and when Comparing to Goddard-SDSU Refining these modifications, working in CRTM repository Stream number estimation Revamp data structures, scheme selection interface Tangent linear, adjoint, K-matrix ? Waiting for better microphysics scheme (Goddard 2- moment) Working toward improved microphysics scheme ? Ensemble parameter estimation Bias correction OSSE* *though as long as this bias is present, such experiments will yield results of substantially limited value

25. References Chambon, P., S. Q. Zhang, A. Y. Hou, M. Zupanski, and S. Cheung, 2014: Assessing the impact of pre-GPM microwave precipitation observations in the Goddard WRF ensemble data assimilation system. Quart. Jour. Roy. Meteor. Soc., 140, 1219–1235. Han, M., S. A. Braun, T. Matsui, and C. R. Williams, 2013: Evaluation of cloud microphysics schemes in simulations of a winter storm using radar and radiometer measurements. J. Geophys. Res. Atmos., 118, 1401–1419. Liu, Q., and F. Weng, 2006: Advanced doubling-adding method for radiative transfer in planetary atmospheres. J. Atmos. Sci., 63, 3459‒3465. Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, M. G. Duda, X.-Y. Huang, W. Wang, and J. G. Powers, 2008: A description of the Advanced Research WRF version 3. NCAR Technical Note 475, http://www.mmm.ucar.edu/wrf/users/docs/arw_v3.pdf. Weng, Y., and F. Zhang, 2012: Assimilating Airborne Doppler Radar Observations with an Ensemble Kalman Filter for Convection-permitting Hurricane Initialization and Prediction: Katrina (2005). Mon. Wea. Rev., 140, 841-859. Wong, V., and K. A. Emanuel, 2007: Use of cloud radars and radiometers for tropical cyclone intensity estimation, Geophys. Res. Lett., 34, L12811, doi:10.1029/2007GL029960. Zhang, S. Q., M. Zupanski, A. Y. Hou, X. Lin, and S. H. Cheung, 2013: Assimilation of Precipitation-Affected Radiances in a Cloud-Resolving WRF Ensemble Data Assimilation System. Mon. Wea. Rev.,141, 754–772. Zhang, F., Y. Weng, J. A. Sippel, Z. Meng, and C. H. Bishop, 2009: Cloud-resolving Hurricane Initialization and Prediction through Assimilation of Doppler Radar Observations with an Ensemble Kalman Filter. Mon. Wea. Rev., 137, 2105-2125. Zupanski, D., S. Q. Zhang, M. Zupanski, A. Y. Hou, and S. H. Cheung, 2011: A Prototype WRF-Based Ensemble Data Assimilation System for Dynamically Downscaling Satellite Precipitation Observations. J. Hydrometeor., 12, 118–134.

26. Extra Slides

27. Cloud Ice Goddard Morrison

29. WSM6 10.65-H18.7-H23.8-V 36.5-H89.0-H 165.5-H

30. WSM6, New Look-up Tables Coarsened to 15x15 km 260240220200180160 270250230210190 89 GHz36.5 GHz

31. WSM6, New Look-up Tables 260240220200180160 270250230210190 89 GHz36.5 GHz

32. CRTM Results, prescribed and uniform hydrometeor radii 37 GHz cloud field should be even warmer -> there is too much scattering too many instances of convective cells producing an anomaly of low temperatures 89 GHz cloud field should not be that cold -> there is too much scattering many instances of much too low brightness temperatures (~50 K)

33. CRTM Results, Effective Radii == Scheme Particle Size Distribution (PSD) mean 37 GHz cloud field is too warm -> insufficient scattering everywhere 89 GHz sufficient/excessive scattering in areas of strongest convection most of cloud field is too warm, insufficient scattering

34. Why didn’t these simulations work well? Pre-specified radii: unacceptable Simply not representing enough physics to be used comfortably in DA or evaluating microphysics scheme Mean radius: too warm, too little scattering Mean particle radius of a cloud < effective scattering particle radius of a cloud b/c scattering is dominated by the large particles Mean of transformed distribution would produce better results, but… It often exceeds the CRTM lookup table effective radius dimension (1500 μm) At these high frequencies, D 6 relationship breaks down for large particles

35. What else have we tried? Analytically determining effective particle radii from microphysics (MP) scheme particle size distributions (PSD) for the scattering-dominated frozen hydrometeors Assume all particles do Rayleigh scattering (problematic) Result: multiply previous radius estimates by 4.9 This would put most graupel, snow > 1000 μm; CRTM results would have too much scattering