• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
IGARSS_LIU_XU_2011.ppt
 

IGARSS_LIU_XU_2011.ppt

on

  • 934 views

 

Statistics

Views

Total Views
934
Views on SlideShare
917
Embed Views
17

Actions

Likes
0
Downloads
1
Comments
0

1 Embed 17

http://www.grss-ieee.org 17

Accessibility

Categories

Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    IGARSS_LIU_XU_2011.ppt IGARSS_LIU_XU_2011.ppt Presentation Transcript

    • A PRINCIPAL COMPONENT-BASED RADIATIVE TRANSFER MODEL AND ITS APPLICATION TO HYPERSPECTRAL REMOTE SENSING Xu Liu NASA Langley Research Center, Hampton VA 23681 W. Wu, Hui Li, D. K. Zhou, A. M. Larar, and P. Yang
    • Presentation outline
      • Why radiative transfer model is a key component
      • How to deal with large amount of hyperspectral data
      • Examples of PCRTM applications
      • Summary and Conclusions
    • Why radiative transfer model is important for hypersectral remote sensing
      • A radiative transfer model is needed to quantify the relationship between satellite data and the properties listed above
      • It is a key component in retrieval algorithms
        • For geophysical and climate parameter retrievals
        • For satellite data assimilations
      • It is needed to perform end-to-end sensor performance simulations
        • A Key component in climate OSSE
        • Help to refine instrument specifications
      • Hyper or ultra spectral data contains thousands of spectral channels
        • T, H 2 O, O 3 , CO 2 , CO, CH 4 , and N 2 O vertical profiles or column densities
        • Cloud height, particle size, optical depth, and phase
        • Surface skin temperature and emissivity spectra
        • Outgoing Longwave Radiation (OLR), TOA radiative flux, cooling rate …
    • How to deal with large amount of hyperspectral data?
      • Line-by-line (LBL) forward model is too slow to handle huge amounts of satellite data
        • Need to perform too many radiative transfer calculations at ~1 million of wavelengths
      • Traditional channel-based rapid radiative transfer models are still too slow
        • 1000-9000 spectral channel radiance spectrum needed
        • NWP models only assimilate 1-2 hundred channels due to the speed limitation
      • Principal-Component-based Radiative Transfer Model (PCRTM) is ideal
        • Channel-to-channel spectral correlations are captured by eigenvectors
        • Reduce dimensionality of original spectrum by a factor of 10-90
        • Spectral correlations are used to reduce number of radiative transfer calculations
        • Very accurate relative to line-by-line (LBL) RT model
        • 3-4 orders of magnitude faster than LBL RT models
        • A factor of 2-100 times faster than channel-based RT models
      • PCRTM model has been w ell tested using real satellite and airborne remote sensing data
        • AIRS, IASI, CrIS, CLARREO,and NAST-I PCRTM models have been generated
      • References on the PCRTM model and retrieval algorithms
        • Liu et Applied Optics 2006
        • Saunders et al., J. Geophys. Res., 2007
        • Liu et al. Q. J. R. Meterol. Soc. 2007, Liu et al. Atmospheric Chemistry and Physics 2009
    • PCRTM is very accurate and has been applied to IASI, nast-I, AIRS, and CLARREO
    • Apply PCRTM to Orbiting Carbon Observatory (OCO) O 2 A-band
      • Model reflectance of O 2 A-band
      • OCO spectral resolution (0.045 nm)
      • 6 EOF, 7 Mono needed for R-branch of O 2 A-band
      • Maximum RMS error < 2.32 x 10 -5 for 7500 sample
        • Various clouds
        • Aerosols
        • Ocean and various land surface types
        • Various atmospheric profiles
      • Bias error close to zero
    • Apply PCRTM to SCIAMACHY O 2 A-band
      • Model reflectance of O 2 A-band
      • SCIAMACHY spectral resolution (~0.2 nm)
      • 5 EOF, 7 Mono needed for R-branch of O 2 A-band
      • RMS error < 3 x 10 -5
      • Bias error close to zeros
      • Will extend the method to CLARREO spectral resolution (~4 nm spacing)
        • Need even less point
        • Will enable much faster OSSE and end-to-end simulations
    • Why PCRTM is very fast ?
      • PCRTM needs far less radiative transfer calculations and needs small number of predictors to calculate channel radiances
        • 1-2 orders of magnitude smaller than channel-based RT models
      • PCRTM provides derivatives of radiance with respect to atmospheric parameters for each forward model
        • Saves 10-100 forward model runs compared to finite difference method
      NAST-I Spectral Band Number of Channels No. of RT Calc. for All NAST Channels Predictors per Channel PCRTM 8632 310-900 0.04-0.1 PFAST 8632 8632 ~40 OSS 8632 22316 2.59
    • PCRTM computes cloud radiative contribution efficiently
      • PCRTM can handle as many as 40 layers of clouds
        • Parameterizes cloud emissivity and transmissity
        • Compares well with DISORT
        • Orders of magnitude faster than DISORT
        • Only slightly slower than clear sky radiative transfer calculation
    • PCRTM retrieval compares well with radiosondes
      • Temperature, moisture, and ozone cross-sections
      • Plots are deviation from the mean
      • Fine water vapor structures captured by the retrieval system
      • A very cloudy sky condition
    • PCRTM retrieved surface temperature and emssivity and comparison with field validation data
      • Comparison of PCRTM retrieved surface skin temperature with ARIES measured Tskin
      • Comparison of retrieved ocean emissivity with ARIES aircraft measurements
    • Example of retrieved cloud properties Cloud Optical Depth (truth) Cloud Optical Depth Cloud Effective size (  m) Effective Size (truth,  m)
    • PCRTM retrieved monthly mean climate related properties (T, water)
      • Atmospheric temperature at 9 km for July 2009
      • Surface skin temperature for July 2009
      • Surface emissivity for July 2009
      • Atmospheric carbon monoxide mixing ratio for July 2009
    • Applying PCRTM to calculate the OLR and comparison with CERES observations
      • Work done by Fred Rose and Seiji Kato at NASA Langley
      • PCRTM used to calculate cloudy radiance from 50 cm-1 to 2800 cm-1 using MODIS/CERES cloud fields and model atmospheres
      • PCRTM OLRs are compared with CERES observations
      • Orders of magnitude faster than Modtran
      • Good agreement for 6 years of record
    • Summary and Conclusions
      • Forward model is a key component in analysing hyperspectral data
        • End-to-end sensor trade studies
        • Realistic global long term data simulations and OSSE experiment
        • Key component in satellite data analysis and data assimilations
      • PCRTM is a useful tool specific for hyperspectral data with thousands of channels
        • PCRTM compresses thousands of spectral channels into 100-200 EOFs
        • 3-4 orders of magnitude faster than Line-by-line models
        • 2-100 times faster than traditional forward model
        • Very accurate relative LBL models
        • Multiple scattering cloud calculations included
        • Model has been developed for AIRS, NAST,IASI, CLARREO
        • Work started to extend the method to UV-VIS spectral region (OCO, SCIAMACHY)
      • Retrieval algorithms based on PCRTM has been successfully used for IASI, NAST and other hyperspectral sensors
        • Atmospheric temperature, water, trace gases, cloud properties, surface skin temperature and surface emissivities are retrieved simultaneously
        • Retrievals have been validated using radionsondes and field data