Likun Wang Ph.D. Candidate

1. The Homogeneity of Midlatitude Cirrus Cloud Structural Properties Analyzed from the Extended FARS Dataset Likun Wang Ph.D. Candidate

2. Content • Motivation • FARS high cloud dataset • Proposed Method • Proposed future research

3. Why are cirrus clouds important? • Influence on the radiation balance of the climate system (Liou, 1986) • Macrophysical properties • Cloud top, base, thickness, cover, overlap • Microphysical properties • Ice water content (IWC) and ice crystal size distribution • Ice crystal habit

4. Why are cirrus clouds important? (con’t) • Important in the chemistry of the upper troposphere • Contribute to upper troposphere ozone depletion (Borrman et al. 1996; Kley et al. 1996) • Perturb chlorine chemistry (Solomon et al. 1997 )

5. Reality v.s. GCM • Using Plane Parallel Homogeneous (PPH) approximation

6. Reality v.s. GCM (con’t) • No horizontal inhomogeneities • e.g. the distribution characteristics of cloudy and clear sky regions • e.g. the horizontal variability of microphysical properties within a layer

7. Reality v.s. GCM (con’t) • Limited vertical inhomogeneities • e.g. How clouds overlap? • maximum overlap for adjacent levels & random overlap for non adjacent levels is assumed • e.g. the vertical variability of microphysical properties within a layer

8. Why PPH can’t represent reality ? PPH without homogeneities ICA With homogeneities

9. PPH v.s. ICA • Independent column approximation (ICA) • Sliced grid box into different column • Radiative transfer calculations of a cloud field are done in for every column • then an average value is determined

10. PPH v.s. ICA ------Albedo Bias Carlin et al. personal communication; Cahalan et al. 1994; Barker,1996 αPPH> αICA Overestimate αPPH Bias αICA Albedo τm τ2 τ1 Optical Thickness

11. OLRPPH< OLRICA Underestimate Bias PPH v.s. ICA ------ OLR Bias • OLR(ICA)-OLR(PPA) ~ 14 W/m- 2 (Fu et al. 2000)

13. Author Inhomogeneous structure Length Scale (KM) Instruments Comments Heymsfield (1975) Uncinus top generating cell 1-2 Radar, aircraft observation Minnesota, Illinois, Colorado, Wyoming. Auria and Campistron (1987) cirrus generating cell 1.3 and 0.7 Radar PEP* project, in Spain, 1987. Sassen et al. (1989) Mesoscale Unicinus Complexes (MUC) cirrus uncinus cell ~15- ~100 ~1 Lidar, radar and aircraft observation FIRE data, Colorado,(1983), Utah(1985), Wisconsin(1986). Starr and Wylie (1990) MUC Small scale cellular structure 20-500 Rawinsonde and satellite observation FIRE data, Wisconsin, 1986 Sassen et al. (1990) MUC cirrus uncinus cell ~120 ~1 Lidar and aircraft observation FIRE data, Wisconsin, 1986 Grund and Eloranta (1990) MUC 4-12 Lidar FIRE data, Wisconsin, 1986 Smith et al. (1990) Convective cell 4-10 Aircraft observation FIRE data, Wisconsin, 1986 Gultepe and Starr (1995) Gravity waves Quasi-two-dimensional waves Larger two-dimensional esoscale wave 2-9 10-20 100 Aircraft observation FIRE data, Wisconsin, 1986 Gultepe et al. (1995) Coherent Structure 0.2-10 Radar and Aircraft observation FIRE II data, Kansas, 1991 Smith and Jonas (1996) Convective cell Gravity waves Turbulence 2 2 0.05-0.6 Aircraft observation EUCREX**, England, Scotland, Iceland, 1993 Demoz et al. (1998) Convective cell Gravity waves 1.2 2-40 Aircraft observation SUCCESS***, Oklahoma, 1996 Inhomogeneous structure observed from cases study

14. How about cirrus? • the complexity of internal structure exists • scale: 10-2 ~ 105 m • Include: • Turbulence • Kelvin-Helmholtz waves • Small scale cellular structure, convective cell • Gravity waves • Mesoscale Unicinus Complexes (MUC)

15. How about cirrus? (con’t) • Starr and Cox (1985) • embedded cellular structures develop in the simulation of cirrostratus cloud layer • horizontal scales : ~1 km or less • Dobbie and Jonas (2001) • radiation could have an important effect on cirrus clouds inhomogeneity

16. Big difficulties: • Case analysis is not enough to disclose the characteristics of cirrus clouds inhomogeneities • Need a high resolution and long-term datasets • Different scale processes often happen together and coexist in the same cloud system and not easy to locate • Need an efficient analysis tool

17. Content • Motivation • FARS high cloud dataset • Proposed Method • Proposed future research

18. FARS Site • Located 40 49’00’’N, 111 49’38”W • Instruments • Passive Remote Sensors • Active Remote Sensors • Polarization Cloud Lidar (PCL) ---Ruby lidar • Two-color Polarization Diversity Lidar (PDL) • 95 GHz Polarimetric Doppler Radar

19. Ruby lidar • Two channels • Vertical polarization transmitted • Manually "tiltable" ± 5° from zenith • 0 .1 Hz PRF, 7.5 m maximum range resolution • Maximum 2K per channel data record length • 1-3 mrad receiver beamwidths • 25 cm diameter telescope • 0.694 µm wavelength, 1.5J maximum output

20. FARS high cloud dataset • October,1987 --- Now • Typical 3-hour data (10 sec resolution) • Using the average wind speed: 25 m/s • Spatial scale : 250 m ~ 270 km • Mainly focus on higher, colder and thinner cirrus cloud independent with low clouds (lidar limit)

21. FARS Data (Oct. 1987 - Dec. 2001) Total: 3216 hours

22. FARS Data per month Max: 404 hours(OCT) Min: 177 hours (JUN)

23. Content • Motivation • FRAS high cloud dataset • Proposed Method • Proposed future research

24. Signal from lidar • P0 is the power output (J) , • c speed of the light (m s-1), • t the pulse length (m), • Ar the receiver collecting area (m2), •  the volume backscatter coefficient (m sr)-1, •  the volume extinction coefficient area (m-1), •  the multiple forward-scattering correction factor. • m and c denote contributions from molecules and cloud.

25. Signal from lidar • Calibrate the scattering and extinction due to air molecules under the pure molecular scattering assumption (Sassen 1994) • Assume a relationship (Klett 1984): • It is possible to gather the information on inhomogeneous properties by analyzing P(R)•R2

26. From Time series to spatial series data • Assume that the internal cloud properties vary much more with space than with typical observation periods • Also assume cirrus moves faster horizontally than vertically • Using radiosonde data, we can transfer time series data to spatial series data

28. Why wavelet?

29. Spectrum of two process(Fourier transform)

30. But using wavelet

31. Continuous Wavelet Transform (CWT) • the element transform wavelet function can be defined : • Where • τ is translation parameters • s is scale parameters

32. ψ can be many forms including morlet, Mexican hat …

33. Continuous Wavelet Transform (CWT) • CWT is defined as follows : Where • x(t) is the signal • Ψ*(t) is the wavelet function • τ and s , the translation and scale parameters, respectively

34. Content • Motivation • FRAS high cloud dataset • Proposed Method • Proposed future research

35. Proposed future work • Examining structural inhomogeneity of broken cirrus cloud cases • Determining the statistics of broken cirrus fractional cloud amounts • Determining cloud layer overlap for multiple layer cirrus clouds without low water clouds • Creating the relationship between the cloud top temperature and the length scales of cloud distribution

36. Proposed future work • Examining inhomogeneous properties in ‘homogeneous’ cirrus • Check all the cirrostratus cases • Locate inner inhomogeneous dynamics process such as gravity waves, Kelvin-Helmholtz waves and convective cell • Evaluate statistics characteristics of these process

37. Proposed future work • Furthering the knowledge of cirrus cloud structures and the dynamics to the major cloud generating mechanisms • Classified into four kinds type • Check every type’s inner structures • Try to find the relationship between inner structures and dynamics

38. Proposed future work • Calculating the bias of radiative quantities due to the neglect of cirrus cloud inhomogeneities • Use Fu and Liao’s radiation transfer model • Structural characteristics • Quantify the bias of albedo and OLR between ICA and PPH

39. Purpose of research Final Purpose is: Characterize the vertical and horiziontal inhomogeneities of midlatitude cirrus cloud FARS lidar data radiosonde data spatial series data cloud detection method wavelet method cloud fraction cloud overlap length scale of cloud distribution

40. Purpose of research (con’t) Final Purpose is: Quantify the radiative bias due to the neglect of midlatitude cirrus cloud inhomogeneities using radiation transfer models Characteristics from data analysis Radiation Transfer Model LW Radiation Bias Albedo Bias

41. Thank you! Need hard work!