1. The apparent bluing of aerosols near clouds Alexander Marshak (GSFC) Guoyong Wen and Tam á s V á rnai (UMBC/GSFC) Lorraine Remer and Bob Cahalan (GSFC) Jim Coakley (OSU) and Norman Loeb (LRC)

2. Feb 5, 2008Alexander Marshak1 Aerosols and Clouds 1-km Terra MODIS Data 0.86-  m 0.55-  m Aerosol Optical Depth Fine Particle Fraction Fine Mixed Coarse Sun Glint Boundary Courtesy of Jim Coakley

3. Feb 5, 2008Alexander Marshak2 MODIS Aerosol Optical Depth and Cloud Cover Sun Glint Boundary Aerosols grow in humid environments near clouds. Aerosols grow through in-cloud processing. Aerosols remain when droplets evaporate New particle production in the vicinity of clouds. Illumination of cloud-free columns enhanced through scattering of sunlight by nearby clouds. Cloud contamination of the cloud-free pixels used to obtain aerosol properties. Aerosols are precursors to cloud formation. Courtesy of Jim Coakley

4. Feb 5, 2008Alexander Marshak3 What happens to aerosol in the vicinity of clouds? All observations show that aerosols seem to grow near clouds or (to be safer) “most satellite observations show a positive correlation between retrieved AOT and cloud cover”, e.g.: from Loeb and Manalo-Smith, 2005from Ignatov et al., 2005 Cloud Fraction (%) from Zhang et al., 2005

5. Feb 5, 2008Alexander Marshak4 What happens to aerosol in the vicinity of clouds? All observations show that aerosols seem to grow near clouds. However, it is not clear yet how much grows comes from “real” microphysics, e.g. increased hydroscopic aerosol particles, new particle production or other in-cloud processes. (“artificial”) the 3D cloud effects in the retrievals: cloud contamination, extra illumination from clouds

6. Feb 5, 2008Alexander Marshak5 How do clouds affect aerosol retrieval? clouds are complex and “satellite analysis may be affected by potential cloud artifacts” (Kaufman and Koren, 2006) ; Both cloud contamination (sub-pixel clouds) cloud adjacency effect (a clear pixel with in the vicinity of clouds) may significantly overestimate AOT. But they have different effects on the retrieved AOT: while cloud contamination increases “coarse” mode, cloud adjacency effect increases “fine” mode.

7. Feb 5, 2008Alexander Marshak6 The Ångström exponent and the cloud fraction vs. AOT from Kaufman et al., IEEE 2005 Atlantic ocean, June-Aug. 2002; each point is aver. on 50 daily values with similar AOT in 1 o res.; for AOT < 0.3, as AOT increases CF and the Ångström exponent also increase; the increase is due to transition from pure marine aerosol to smoke (or pollution); the increase in AOT cannot be explained by cloud contamination but rather aerosol growth.

8. Feb 5, 2008Alexander Marshak7 More clouds go with larger AOT and larger (not smaller!) Ångstr ö m exponent Courtesy of Norman Loeb (A-train presentation, Lille Oct. 2007) 25 1 o x1 o in each 5 o x5 o region over ocean (off the cost of Africa) are subdivided into two groups with  a  and  a  meteorology has been checked as similar for two groups

9. Feb 5, 2008Alexander Marshak8 More clouds go with larger AOT and larger (not smaller!) Ångstr ö m exponent Courtesy of Norman Loeb (A-train presentation, Lille Oct. 2007) 25 1 o x1 o in each 5 o x5 o region over ocean (over the entire globe) are subdivided into two groups with  a  and  a  meteorology has been checked as similar for two groups Difference in cloud fractionDifference in fine-mode fraction

10. Feb 5, 2008Alexander Marshak9 AOT and Ångstr ö m exponent vs. distance from the nearest cloud (AERONET data) from Koren et al., GRL, 2007 Time passed from the last cloud (min) AOT (0.47  m) Ångstr ö m exponent The Ångström exponent increases with distance to the nearest cloud while the AOT increases

11. Feb 5, 2008Alexander Marshak10 Enhancement of radiance near clouds Cumulus clouds over Atlantic from Koren et al., GRL, 2007 at 0.87  m 3D ? 20 1000x500 km scenes analyzed

12. Feb 5, 2008Alexander Marshak11 Airborne aerosol observations in the vicinity of clouds Courtesy of Jens Redemann From airborne extinction rather than scattering observations 3D effects decrease AOT rather than increase it

13. ARM Shortwave Spectrometer transition between cloudy and clear skies clear cloudy two Cu clouds during the first and last 5 to 8 sec. clear sky is evident about 15 sec. away from these periods; the measurements in the intervening period (5 to 12 s and 75 to 82 s) are difficult to classify; depending on the remote sensing criterion used for cloud detection, would be called either cloudy or clear. measurements of P. Pilewskie (Chiu et al., 2008, in preparation)

14. Feb 5, 2008Alexander Marshak13 A simple 3D RT experiment cloudy 0.3 km 5 km 5 cc 5 Distance (km) clear 0

15. Feb 5, 2008Alexander Marshak14 A simple 3D RT experiment black surface vegetated surface 0 Courtesy of Toby Zinner

16. Aerosol-cloud radiative interaction (a case study) Collocated MODIS and ASTER image of Cu cloud field in biomass-burning region in Brazil at 53 o W on the equator, acquired on Jan 25, 2003 Wen et al., 2006

17. Feb 5, 2008Alexander Marshak16 ASTER image and MODIS AOT Thick clouds Thin clouds from Wen et al., JGR, 2007 ASTER image MODIS AOT

18. Feb 5, 2008Alexander Marshak17 A striking example: const AOT Modeled (with const AOT but MODIS 3D cloud structure) vs Observed Reflectance. Cor. coef. = 0.77

19. Feb 5, 2008Alexander Marshak18 Cloud effect at 90-m resolution Thin clouds, =7 Thick clouds, =14 AOT 0.66 =0.1  ~ 0.0046  ~0.05 ≈50%  ~ 0.014  ~0.14 ≈140% enhancement:  3D -  1D

20. Feb 5, 2008Alexander Marshak19 Effect of distance to a cloudy pixel 3D enhancement (red) Cumulative distr. (blue) Nearest cloud dist. (km) 0.66  m

21. Feb 5, 2008Alexander Marshak20 aerosol or molecule molecule + aerosol MODIS sensor Conceptual model to account for the cloud-induced enhancement surface

22. Feb 5, 2008Alexander Marshak21 Contributors to cloud enhancement Rayleigh scattering Aerosols Surface reflectance AOT=0.1 AOT=0.5 AOT=1.0 For dark surfaces and aerosols below the cloud tops,the enhancement only weakly depends on the AOT enhancement:  3D -  1D 3D cloud enhancement Surface albedo

23. Feb 5, 2008Alexander Marshak22 aerosol or molecule from Wen et al., JGR 2008: molecule (82%) + aerosol (15%) MODIS sensor Conceptual model to account for the cloud enhancement (at 0.47  m) surface (3%)

24. Feb 5, 2008Alexander Marshak23 Assumption for a simple model Molecular scattering is the main source for the enhancement in the vicinity of clouds thus we retrieve larger AOT and fine mode fraction

25. Feb 5, 2008Alexander Marshak24 How to account for the 3D cloud effect on aerosols? The enhancement is defined as the difference between the two radiances: one is reflected from a broken cloud field with the scattering Rayleigh layer above it and one is reflected from the same broken cloud field but with the Rayleigh layer having extinction but no scattering Broken cloud layer Rayleigh layer from Marshak et al., JGR, 2008

26. Feb 5, 2008Alexander Marshak25 Stochastic model of a broken cloud field AR = 2 A c = 0.3 Clouds follow the Poisson distr. and are defined by average optical depth, cloud fraction, A c aspect ratio, AR = hor./vert. AR = 1

27. Feb 5, 2008Alexander Marshak26 Stochastic model of a broken cloud field AR = 2 A c = 0.3 Clouds follow the Poisson distr. and are defined by average optical depth, cloud fraction, A c aspect ratio, AR = hor./vert. AR = 1

28. Feb 5, 2008Alexander Marshak27 Cloud-induced enhancement at 0.47  m LUT: The enhancement vs for AR = 1. A c =1 corresponds to the pp approximation.

29. Feb 5, 2008Alexander Marshak28 Cloud-induced enhancement: our simple model and 3D RT calculations The enhancement vs for A c = 0.6 and 3 cloud AR = 0.5, 1 and 2. Different dots are from Wen et al. (2007) MC calculations for the thin and thick clouds.

30. Feb 5, 2008Alexander Marshak29 Ångstr ö m exponent Ångstr ö m exponent vs for A c = 0.5 and AR = 2. Three cases: clean, polluted and very polluted. 0.47  m vs. 0.65  m 0.65  m vs. 0.84  m clean polluted very polluted The cloud adjacency effect increases the Ångstr ö m exponent

31. MODIS observations (Várnai and Marshak, 2008, in preparation) Collection 5 MOD02, MOD06, MOD35 products September 14-29 in 2000-2006 (2 weeks in 7 years) North-East Atlantic (45°-50°N, 5°-25°W), south-west from UK Viewing zenith angle < 10° Pixels included in plots: Ocean surface with no glint or sea ice MOD35 says “confident clear”, all 250 m subpixels clear Highest cloud top pressure nearby > 700 hPa (near low clouds) Nearby pixels are considered cloudy if MOD35 says definitely cloud.

32. Example of the region: Sep 22, 2005

33. Average reflectance vs. dist. to clouds for 0.45, 0.65, 0.87, 2.1 and 11  m mean and std

34. Reflect. Dist to Clouds vs Cloud Optical Depth 0.47  m # of pixelslog of # of pixels Dist. to cloud (km)

35. Reflect. Diff from Values at 10 km vs Cloud Optical Depth 2.10  m

36. Reflect. Diff from Values at 10 km vs Cloud Optical Depth 0.87  m

37. Reflect. Diff from Values at 10 km vs Cloud Optical Depth 0.65  m

38. Reflect. Diff from Values at 10 km vs Cloud Optical Depth 0.47  m

39. Cloud enhancement vs. dist. to clouds for 0.45, 0.65, 0.87, and 2.1  m

40. Latency effect for 2.1 um

41. Point spread function effect for 0.53  m (preliminary results) with Jack Xiong

42. Cloud contamination in 0.47  m (preliminary results) with Jack Xiong latency effect removed (red curve); assumed that at 2.1  m the increase is due to undetected subpixel clouds (blue curve) assumed that 2.1  m has the same point spread function as 0.53  m (green curve)

43. Work in progress select a few MODIS subscenes with broken low Cu; retrieved AOT; over ocean with no glint, etc; analyze AOT, CF, average COD over many 10 x 10 km areas; use a simple stochastic model and RT to estimate upward flux; use CERES fluxes to convert BB to spectral fluxes; use ADM to determine spectral fluxes from MODIS radiances; estimate cloud enhancement and compare the results; use a simple linearization model. with Norman Loeb and Lorraine Remer

44. Work in progress The NASA ARCTAS aircraft mission will take place in spring/summer 2008. Cloud-scattered Light Experiment To characterize the cloud-scattered light field to interpret satellite radiance observ. in the “halo” region surrounding clouds. Using SSFR to measure up- and down-welling spectral radiative flux as a function of - distance to (water) cloud - COD, cloud base, cloud height, AR - vertical distribution of aerosols - surface spectral albedo - etc. Below cloud and well-above cloud with multiple horizontal distances from cloud with Ralph Kahn, Jens Redemann and Phil Russell

45. Feb 5, 2008Alexander Marshak44 Conclusions No clear understanding from satellites alone of what happens to aerosols at the vicinity of clouds. (The twilight zone?) Accounting for the 3D cloud-induced enhancement helps. For certain conditions, 3D cloud enhancement  3D  1D only weakly depends on AOT and molecular scatt. is the key source for the enhancement. The enhancement increases the “apparent” fraction of fine aerosol mode (“bluing of the aerosols”). MODIS observations confirm that the cloud induced enhancement increases with cloud optical depth. Retrieved AOT can be corrected for the 3D radiative effects.

46. Point spread function effect for 0.53  m (preliminary results) with Jack Xiong point spread function and latency effect removed (blue curve); assumed that at 2.1  m the increase is due to undetected subpixel clouds (green curve) assumed that 2.1  m has the same point spread function as 0.53  m (black curve)

47. Feb 5, 2008Alexander Marshak46 Cloud enhancement vs. cloud reflectance The upwelling flux is needed to estimate the cloud- induced enhancement of cloud-free pixels