GRASP Algorithm: Retrieval of the detailed properties of atmospheric aerosol from PARASOL and other sensors

Transcription

1 GRASP Algorithm: Retrieval of the detailed properties of atmospheric aerosol from PARASOL and other sensors Oleg Dubovik (University of Lille-1, CNRS, France) GRASP team: P. Litvinov 1,T. Lapyonok 1, F. Ducos 1, D. Fuertes 1, X. Huang 1, A. Lopatin 1, B.Torres 1, Y. Derimian 1, L. Li 1, C. Chen 1 M. Aspetsberger 2, C. Federspiel 2, etc. 1 - University of Lille-1, CNRS, France 2- Catalysts GmbH, High Performance Computing, Linz, Austria GRASP 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

2 GRASP: Generalized Retrieval of Aerosol and Surface Properties MERIS AERONET POLDER lidar AERONET SentineI - 4 laboratory single scattering GRASP open source MISR shape Surface reflectance BRDF BPDF 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

3 Smoke Desert Dust AERONET retrievals are driven by 31 variables : dv/lnr - size distribution (22 values); n(l) and k(l) - ref. index (4 +4 values) C spher (%) - spherical fraction (1 value) Particle Size Distribution: 0.05 mm R (22 bins) 15 mm Complex Refractive Index at l = 0.44; 0.67; 0.87; 1.02 mm Maritime

4 Single - Pixel Retrieval: RT calculation on fly!!! f j* - PARASOL data: Angular measurements (~15 angles) of - Intensity (l = 0.49; 0.67; 0.87; 1.02 mm) - Polarization (l = 0.49; 0.67; 0.87 mm) a j - Parameters to be retrieved: -Aerosol propetries: - size distribution; - real refractive index - imaginary refractive index; - particle shape, - height -Surface properties (over land): - BRF parameters; - BPRF parameters ì ï í ï î f j * 0 j * = æ ç ç è F j S j ö a j + ø D j m D j a PARASOL A Priori Constraints limiting derivatives (e.g. Dubovik 2004) of - for aerosols (e.g. in AERONET, Dubovik and King 2000) : - aerosol size distribution variability over size range; - spectral variability of complex refractive index; - for surface (e.g. in AERONET/satellite retrievals, Sinuyk et al. 2007) : - spectral variability of BRF/ PBRF parameters. Multi-term LSM statistically optimized Solution (Dubovik and King 2000, Dubovik 2004) :,where W = S T j j S j ; W = 1 i e C ; g = e 2 f 2 f j 2 f e a

5 Time-Variability Constraints The concept of multi-pixel retrieval ( t 3 ; x ; y ) t ( t 2 ; x ; y ) z X ( t 1 ; x ; y ) x X X-Variability Constraints 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

6 Multi - Pixel Retrieval: ì ï í ï ï î æ * f ö 1 ç * è 0 1 ø æ * f ö 2 ç * è 0 2 ø æ * f ö 3 ç * è 0 3 ø... * 0 t * 0 x * 0 y æ ç çç ç çç æ ç è ö 0 0 ø æf ö 2 0 ç 0 ès 2 ø æf ç ö è ø F 1 S 1 a è 3 ø = S 3 + ç çç ç çç è S t,1 S t,2 S t,2 S x,1 S x,2 S x,3 S y,1 S y,2 S y,3 ö ø æ ç çç ç a 1 a 2 ö Single-Pixel Data (PARASOL measurements and physical a priori constraints) are used by the same way as in Single-Pixel retrieval. Multi-Pixel a priori constraints (e.g.dubovik et al. 2008): - limited spatial variability of each aerosol /surface parameter - limited temporal variability of each aerosol /surface parameter NOTE: degree of variability constraints (smoothnes) can be different and adequately chosen for each parameter Multi-term LSM Multi-Pixel Solution:,where W = S T x x S x ; W = S T y y S y ; W = S T t t S t ; g x = e 2 f e ; g = e f 2 y x e ; g = e f 2 t y 2 2 e t 2

7 PARASOL: the space borne instrument most suitable for enhanced aerosol/surface characterization PARASOL daily coverage image, March 3, 2013 life time: dec INTENSITY for aerosol (0.44, 0.49, 0.56, 0.67, 0.865, 1.02 mm) for gas absorption: (0.763, 0.765, mm) POLARIZATION (Q, U): (0.49, 0.67, mm) Swath: about 1600 km cross-track Global coverage: every 2 days 1 pixel spatial resolution: 5.3km 6.2km 0 _ 0

8 Test with synthetic measurements Aerosol Optical Thickness PARASOL over Banizoumbou in January, February 2008 I + Q, U I Q, U

9 Test with synthetic measurements Single Scattering Albedo PARASOL over Banizoumbou in January, February 2008 I + Q, U I Q, U

10 EXAMPLES of PARASOL/GRASP retrievals NO ASSUMPTIONS on aerosol and surface All calculation on the fly AOT(0.56) SSA(0.56) AOT(0.56) - loading SSA(0.56) - absorption Processed at ICARE

11 EXAMPLES of PARASOL/GRASP retrievals NO ASSUMPTIONS on aerosol and surface All calculation on the fly Surface Albedo (0.56) NDVI Albego(0.56) - surface NDVI Processed at ICARE

12 1 year of PARASOL data compared with AERONET over Africa at 6 sites: AOD (440) AOD (870) 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

13 1 year of PARASOL data compared with AERONET over Africa at 6 sites: Angstrom Exponent Aerosol SSA 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

14 1 year of PARASOL data compared with AERONET Ilorin complex mixture of dust and biomass burning AOD Angstrom Exponent SSA 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

15 Comparison with other aerosol products PARASOL /GRASP PARASOL /fine mode operational MODIS /Dark Target MODIS /Deep Blue

16 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

17 Retrieved seasonal variability of aerosol AOD AOD(565 nm) Dust sources Biomass burning sources 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

18 Aerosol transport tracking with GRASP Stable dust aerosol sources Daily orbits Biomass burning plume evolution 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

19 Aerosol types identification with GRASP (AOD, SSA, AE) SSA (670 nm) SSA (865 nm) SSA (1020 nm) AOD (443 nm) AOD (565 nm) AOD (670 nm) September, 8, 2008.

20 Is GRASP is fast enough for operational processing? Accurate, 3 sec per pixel, accelerated is ~ 0.3 or less per pixel GRASP accelerated vs. GRASP accurate

21 AOD (440 nm): June August, 2008 PARASOL/GRASP No assumptions!!!

22 AOD (560 nm): June August, 2008 PARASOL/GRASP No assumptions!!!

23 AOD-fine (440 nm): June August, 2008 PARASOL/GRASP No assumptions!!!

24 AOD-coarse (1020nm): June August, 2008 PARASOL/GRASP No assumptions!!!

25 Angstrom Exponent: June August, 2008 PARASOL/GRASP No assumptions!!!

26 SSA (560 nm): June August, 2008 PARASOL/GRASP No assumptions!!!

27 AAOD (560 nm): June August, 2008 Absorption PARASOL/GRASP No assumptions!!!

28 Residual : June August, 2008 Characterizes error PARASOL/GRASP No assumptions!!!

29 AOD (440 nm): June August, 2008 PARASOL/GRASP No assumptions!!!???

30 July 11, 2008 Kilauea volcano ash plume (Hawaii, Halemaumau Crater, 2008)

31 July, 2008 Kilauea volcano ash plume (Hawaii, Halemaumau Crater, 2008)

32 GRASP retrieval: Kilauea volcano (Hawaii, Halemaumau Crater, June-August, 2008) AOD, 443 nm Fine mode AOD, 443 nm Halemaumau Crater ash plume SSA, 443 nm

33 Aerosol Mean Height: June August, 2008 PARASOL/GRASP No assumptions!!!

34 Sensitity tests for retrieving aerosol MEAN HEIGHT For different AOD For aerosol types GRASP retrieval of the altitude, m K= a= b= RMSE= Dust Urban Smoke Maritime Assumed aerosol altitude, m 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

35 Conclusion of sensitivity tests: PARASOL data have solid sensitivity to aerosol height; Sensitivity is higher to fine mode aerosol and less to large non-spherical dust; There is dependence on assumption about atmospheric aerosol vertical profile. Which profile to use?

36 DHR(870): June August, 2008 PARASOL/GRASP No assumptions!!!

37 BPDF

38 Advanced surface retrieval with GRASP The same NDVI and DHR but different Polarized reflectance R p (J v,j 0,j) = a f (J v,j 0,j)F p (m,g ) Polarized reflectance provides new information about surface type!

39 MERIS: - radiances at seven wavelengths: (413, 443, 490, 510, 560, 665, and 870 nm); - single view AEROSOL: - size distribution (5 or more bins); - spectral index of refraction (7 l); - sphericity; SRFACE: - BRDF (3 spectrally dependent parameters); 41 = (5 (SD) +14 (ref. ind.) + 1 (nonsp.) + 21 (BRDF) ) 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

40 The concept MERIS/GRASP retrieval works good with synthetic measurements AOD SSA Smoke Dust 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

41 GRASP: Generalized Retrieval of Aerosol and Surface Properties Dubovik et al. 2011, 2014 Retrieves both: surface (over land) and detailed aerosol properties First results: surface GRASP/MERIS 3 months average DHR (January - March 2008) MERIS 3 months PARASOL 3 months Excellent agreement! 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

42 GRASP: First results: aerosol MERIS Bright surfaces AOD560 (January-March 2008) PARASOL Banizoumbou, Niger Ilorin, Niger AERONET validation

43 GRASP/ MERIS land/water (January - Маrch 2008, 10 km resolution) АОТ(670 nm) Dakar AERONET validation ESA CAWA project

44 GRASP/ MERIS retrieval land/water (average January - Маrch 2008, 10 km resolution) Surface DHR(670 nm) Dakar ESA CAWA project

45 GRASP/ MERIS (January - Маrch 2008, 10 km resolution) No assumptions!!! AOT(440)

46 GRASP/ MERIS (January - Маrch 2008, 10 km resolution) No assumptions!!! DHR(670)

47 GRASP/ MERIS (January Маrch, 2008, 10 km resolution) AOT(440) Real or artifacts???

48 March, 22, 2008 MERIS/GRASP MODIS/Terra

49 March, 23, 2008 MERIS/GRASP MODIS/Terra

50 Conclusions: GRASP concept: retrieval of aerosol and surface simultaneously under minimum assumptions GRASP/PARASOL retrieval are promising for diverse products: - Aerosol AOD (even over bright surfaces) and size distributions; - Aerosol absorption, refractive index etc., - Aerosol height; non-sphericity (???); - BRDR + BPRF First results are very promising for several satellte sensors : - PARASOL, MERIS, GOCI (geostationary) ; Promising for synergy retrievals etc. ; - etc. 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

51 Sphericity fraction : June August, 2008 PARASOL/GRASP No assumptions!!!

52 Retrieval of Non-Sphericity and Aerosol Layer Height from Synthetic Measurements AOD (440) AOD (670) AOD (870) Sphericity fraction Aerosol height Low aerosol mass loading! 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

53 The concept of multi-pixel retrieval Dubovik et al th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

54 Aerosol sources identification with GRASP (AOD, SSA, AE) PARASOL/GRASP AOD (565 nm) Dust sources Cloud free for PARASOL and cloudy pixels and MODIS? Plume of biomass burning MODIS/AQUA AOD (550 nm) PARASOL/GRASP SSA (565 nm) PARASOL image. September, 15,

55 GRASP/MERIS: AOD(560 nm), January March 2008 GRASP is somewhat slow compared to conventional retrievals, but easily mitigated using advanced IT technology such as GPGPU First results: Aerosol, 35 km resolution No assumptions!!! ESA CAWA project 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

56 GRASP/MERIS: AOD(560 nm), January - March 2008 Retrieves both: surface (over land) and detailed aerosol properties No assumptions!!! First results: Surface reflectance, 35 km resolution ESA CAWA project 7th ICAP Meeting, Barcelona, Spain, June 16-19, 2015

57 European Geosciences Union General Assembly 2015, Vienna Austria April 2015 AOD retrieval in Ocean/Land zones West coast of Africa (23N,-15W) June 11, 2008 August 14, 2008

58 What is done Reduced number of Gaussian quadrature in forward RT and Jacobean matrix calculations. Used initially: 10(7) To speed up: 7(3) Reduced number of Fourier expansion term. Used initially: 10 To speed up: 5 Reduced accuracy of RT-calculation: Initial absolute accur.: To speed up: Reduced number of bins in Size distribution. Used initially: 9 To speed up: 5 Reduced number of iterations in inversion. Used initially: 12 To speed up: 6