cloud property retrieval from hyperspectral ir measurements jun li, peng zhang, chian-yi liu, xuebao...

Download Cloud property retrieval from hyperspectral IR measurements Jun Li, Peng Zhang, Chian-Yi Liu, Xuebao Wu and CIMSS colleagues Cooperative Institute for

If you can't read please download the document

Upload: sylvia-reeves

Post on 18-Jan-2018

215 views

Category:

Documents


0 download

DESCRIPTION

AIRS granule 143, March 03, 2004 MODIS natural color Wet Air Dry Air Desert Dust AIRS window BT (K)

TRANSCRIPT

Cloud property retrieval from hyperspectral IR measurements Jun Li, Peng Zhang, Chian-Yi Liu, Xuebao Wu and CIMSS colleagues Cooperative Institute for Meteorological Satellite Studies (CIMSS) University of Wisconsin - Madison Madison, WI 53706, U.S.A. 5th Annual Workshop on Hyperspectral Science of UWMadison MURI, Airborne, LEO, and GEO Activities June 7 9, 2005 Pyle Center, UW-Madison, Madison, Wisconsin What can we see about the cloud information/properties from hyperspectral IR ? How to extract the cloud information? Validation of results AIRS granule 143, March 03, 2004 MODIS natural color Wet Air Dry Air Desert Dust AIRS window BT (K) Less H2O absorption More H2O absorption Clear Dry air over ocean Clear Wet air over ocean AIRS clear brightness temperature spectra: Dry air mass versus wet air mass over ocean Desert Vs Dust aerosol Dust MODIS (BT11 - BT12) Desert AIRS window BT MODIS Natural color MODIS dust optical thickness MODIS dust effective radius Preliminary result for dust property retrieval from MODIS on March 03, 2004 Dust aerosol Desert Low surface 8.9 um over desert Negative slope between 11 and 12 um High ice cloudy spectrum with positive slope. MODIS BT11 - BT12 AIRS window BT (K) Cloud information in Hyperspectral IR data Cloud/aerosol top height Cloud/aerosol discrimination Cloud/aerosol particle size Cloud/aerosol optical depth How to extract the cloud information? AIRS sub-pixel cloud detection and characterization using MODIS data (Li et al. 2004a) Cloud property retrieval from AIRS radiances (Li et al. 2004b; 2005) with the help of MODIS AIRS sub-pixel cloud detection and characterization using MODIS data (Li et al. 2004a). 1km MODIS cloud mask tells if an AIRS footprint is clear or cloudy 1km MODIS cloud phase mask tells if an AIRS footprint is ice clouds, water clouds or mixed phase clouds 1km MODIS classification mask tells if an AIRS footprint is single-layer clouds or multi-layer clouds Aqua MODIS RGB Natural Color 17 September 2003 MODIS cloud mask with 1 km spatial resolution ClearCloudy AIRS cloud coverage from MODIS Unknown Mixed Phase Ice Clouds Water Clouds Clear AIRS phase mask from MODIS MODIS cloud phase mask with 1 km spatial resolution Water Land L. Cld Mid Cld L. Cld H. Cld H. Cld Mid Cld Mid. Cld F3 F2 F1 MODIS 1km classification mask superimposed to the AIRS footprints of the study area. MODIS classification mask gives the cloud layer information with each AIRS footprint. Mid level High level Very high AIRS window BT (K) Cloud property retrieval from AIRS radiances (Li et al. 2004b; 2005, JAM) Cloud property retrieval from AIRS radiances only using Minimum Residual (MR) algorithm, with help from MODIS cloud mask and cloud phase mask. Fast Cloudy Radiative Transfer Model: coupled Observed AIRS Radiance Measurements CTP: Cloud-Top Pressure; ECA: Effective Cloud Amount at 10 wavenumbers CPS: Cloud Particle Size in diameter; COT: Cloud Optical Thickness at 0.55m CTP and ECA: 670 790 cm -1 COT and CPS: 790 1130 cm -1 Minimum Residual (MR) algorithm for cloud retrieval (Li et al. 2004b; JAM) Cost function clear cloudy COT, CPS (window region) CTP (CO2 region) COT=10.0 COT=5.0 COT=1.0 COT=0.05 CPS=10 CPS=20 CPS=30 CPS=50 Ice Cloud Particle Size=30 m (cloud top pressure = 300 mbar) L.CLD H.Cld CTP=345 hPa,COT=2.7 CTP=238 hPa,COT=3.2 CTP=187 hPa,COT=4.3 AIRS cm -1 F1: CTP=251 hPa, COT=0.34 F2: CTP=248 hPa, COT=1.18 1km MODIS classification mask F1 F2 1km MODIS 11m BT (K) F1: CTP=251, CPS=47.56, COT=0.34 F2: CTP=248, CPS=67.50, COT=1.18 Wavenumber (cm**-1) AIRS cm -1 1km MODIS classification mask superimposed to AIRS footprints F3: Thick ice clouds F3: CTP=258, CPS=33.90, COT=1.62 AIRS window BT image AIRS cloud-top pressure retrieval AIRS cloud optical thickness retrieval Validation Compare with MODIS natural color image Compare with MODIS classification results Ground measurements Lidar observation AIRS CTP (hPa) MR algorithm AIRS CTH (m) Barrow Box A MPACE: Mixed Phase Artic Cloud Experiment Granule 223, 17 October 2004 CTP comparison with Lidar AIRS time is 22:17:32 MODIS CTH=5.5 km; AIRS CTH=7.6 km Vertical Layers AIRS Effective Radius=38.6 um AIRS time = 22:17:32 UTC AIRS OD=1.44 Advantage of AIRS cloud properties? 1.Better Cloud-top pressure information than MODIS for low clouds (Li et al. 2004b) 2.Provide cloud microphysics during both day and night COT AIRS COT MODIS COT averaged to AIRS FOV MODIS provides no retrieval even during daytime due to large solar zenith angle ! Limitations on hyperspectral IR data on cloud properties? When clouds are thick, cloud microphysical information in AIRS radiances is saturated, errors of cloud microphysical property retrievals might be large Thin and low clouds over ocean are difficult to retrieve due to large footprint size Lowcld MODIS natural color MODIS classification mask Lowcld Midcld ClrSfc MODIS 1.38 um band shows no cirrus clouds MODIS 0.64 um band shows bright clouds AIRS CTP - Wrong MODIS CTP - Right averaged to AIRS FOV Granule 223, 17 October 2004 AIRS ECA - Right MODIS ECA - Wrong averaged to AIRS FOV What is the solution? MODIS/AIRS combination 1.Hyperspectral IR data provide fluent cloud/aerosol top and microphysical property information 2.With a fast and accurate cloudy/aerosol radiative transfer model, those cloud properties can be retrieved quantitatively 3.Lidar data is crucial for validation of the hyperspectral IR cloud property retrieval 4.Combination of MODIS/AIRS is very important to achieve the reliable retrievals with good accuracy, further investigation is needed on hyperspectral versus high spatial information and the combination Summary