Development of GEMS Cloud Data Processing Algorithm Yong-Sang Choi 1, Bo-Ram Kim 1, Heeje Cho 2,...

download Development of GEMS Cloud Data Processing Algorithm Yong-Sang Choi 1, Bo-Ram Kim 1, Heeje Cho 2, Myong-Hwan Ahn 1 (Former COMS PI), and Jhoon Kim 3 (GEMS.

If you can't read please download the document

Transcript of Development of GEMS Cloud Data Processing Algorithm Yong-Sang Choi 1, Bo-Ram Kim 1, Heeje Cho 2,...

  • Slide 1
  • Development of GEMS Cloud Data Processing Algorithm Yong-Sang Choi 1, Bo-Ram Kim 1, Heeje Cho 2, Myong-Hwan Ahn 1 (Former COMS PI), and Jhoon Kim 3 (GEMS PI) 1 Ewha Womans University, Seoul 2 Seoul National University, Seoul 3 Yonsei University, Seoul
  • Slide 2
  • Clouds significantly affect gas/aerosol retrievals! Cloud contamination causes errors in air mass factor, and errors in gas/aerosol retrievals. 2
  • Slide 3
  • Comparison of UV cloud products InstrumentWavelength regionProducts GOME-2300 800 nmO 3, NO 2, BrO, SO 2, H 2 O, HCHO, and OClO OMI280 500 nmO 3, aerosols, surface UV irradiance, NO 2, BrO, SO 2, HCHO, and OClO SCIAMACHY240 2380 nmO 3, SO 2, NO 2, BrO, HCHO, OClO, H 2 O, CO, CH 4, CO 2, and clouds TROPOMI270 500 nm, 675 775 nm, 2305 2385 nmO 3, SO 2, NO 2, HCHO, H 2 O, CO and CH 4, and clouds and aerosols 3
  • Slide 4
  • 223 253 283 876543210876543210 460 a 465 a 470 a 475 a 480 485 a 490 O 2 -O 2 absorption band (Acarreta, Haan, and Stammes 2004 JGR) Absorption cross-section of the O 2 -O 2 collision complex near 477 nm, based on measurements by Newnham and Ballard [1998] at 283 K (red curve) and 223 K (blue curve). The curve for 253 K (green) was obtained by interpolation. (adopted from OMI ATBD) 4
  • Slide 5
  • GEMS cloud algorithm will provide CH and CF by using O 2 -O 2 absorption. Main cloud products: - Cloud height ( z c ) - Effective cloud fraction ( c f ) Main bands: -O 2 -O 2 absorption band (460490 nm) 5
  • Slide 6
  • GEMS cloud algorithm will provide CH and CF by using O 2 -O 2 absorption. Main cloud products: - Cloud height ( z c ) - Effective cloud fraction ( c f ) Main bands: -O 2 -O 2 absorption band (460490 nm) 6 ISSUE: How to build DOAS and LUT
  • Slide 7
  • DOAS-calculated O 2 -O 2 absorption factors (Rc and Ns) are compared with LUT, to get cloud products. LUT variables (7D or 8D) Radiance spectra Solar zenith angle Viewing zenith angle Relative azimuth angle Surface altitude Surface albedo Absorption cross section RcRc NsNs 7 Calculation of Rc and Ns Temperature profile?
  • Slide 8
  • DOAS-calculated O 2 -O 2 absorption factors (Rc and Ns) are compared with LUT, to get cloud products. Radiance spectra Solar zenith angle Viewing zenith angle Relative azimuth angle Surface altitude Surface albedo Absorption cross section RcRc NsNs 8 LUT variables (7D or 8D) Temperature profile? Calculation of Rc and Ns ISSUE: How to effectively extract cloud fraction and cloud height from LUT fitting?
  • Slide 9
  • C Sequence of cloud height ( z c ) and cloud fraction ( c f ) retrievals zczc OBS DB 9 Cloud height for C f (Albedo = 0.8) Cloud fraction
  • Slide 10
  • C Sequence of cloud height ( z c ) and cloud fraction ( c f ) retrievals zczc OBS DB 10 Cloud height for C f (Albedo = 0.8) Cloud fraction zczc
  • Slide 11
  • Generation of GEMS synthetic cloud- radiation data TOA UV Radiance (Reflectance) Forward RT simulation Cloud Information 3D NWP model Retrieval Algorithm 11
  • Slide 12
  • Data and models Cloud-to-radiance conversion SCIATRAN (ver 3.1) Cloud properties WRF model simulation* of Case: Typhoon Muifa ( 03UTC, August 6, 2011 ) *by Prof. S.-Y. Hongs team in Yonsei Univ. WRF output Geometry Liquid & ice water contents Centroid cloud height 12
  • Slide 13
  • Assumptions Standard atmosphere McLinden climatology Surface albedo based on WRFs land category Cloud water droplet size = 10 m Cloud ice particle size = 50 m of fractal shape 13
  • Slide 14
  • A test for c f retrieval R clear clear sky simulation R cloud overcast simulation assuming Lambertian reflector ( A g = 0.8) R meas Synthetic data 14
  • Slide 15
  • A test for c f retrieval R clear clear sky simulation R cloud overcast simulation assuming Lambertian reflector ( A g = 0.8) R meas Synthetic data 15
  • Slide 16
  • Relation between effective cloud fraction and cloud optical thickness in synthetic data 16 1% of cloud pixels exceed c f value of 1. These clouds are optically thick ( c 25), having albedo over 0.8.
  • Slide 17
  • Error analysis with synthetic data Errors in cloud height Errors in effective cloud fraction 1%0.06% 10%0.1% 50%0.4% Artificial errors were given to cloud height, and then the sensitivity to errors in effective cloud fraction were tested with our synthetic data. Results show that the sensitivity is fairly small, meaning that cloud height limitedly affects the retrieval accuracy of effective cloud fraction. 17
  • Slide 18
  • C Sequence of cloud height ( z c ) and cloud fraction ( c f ) retrievals zczc OBS DB 18 Cloud height for C f (Albedo = 0.8) Cloud fraction zczc Probably Fine How about this?
  • Slide 19
  • Error analysis with synthetic data Errors in cloud height Errors in effective cloud fraction 1%0.06% 10%0.1% 50%0.4% Artificial errors were given to cloud height, and then the sensitivity to errors in effective cloud fraction were tested with our synthetic data. Results show that the sensitivity is fairly small, meaning that cloud height limitedly affects the retrieval accuracy of effective cloud fraction. 19
  • Slide 20
  • Future study topics and call for In-depth discussion issues in Cloud/aerosol breakout session in GEMS International Science Workshop (October 2013, Korea) Cloud/aerosol effects on various gas retrievals Accuracy of cloud/aerosol products Validation plans for cloud/aerosol products Synthetic cloud-radiation simulators Comparison of algorithms using different bands: O 2 -A, O 2 -B, O 2 -O 2, Raman scattering Etc. 20 Yong-Sang Choi ([email protected])