Journal of Geophysical Research: Atmospheres

POAM II retrieval algorithm and error analysis


  • J. D. Lumpe,

  • R. M. Bevilacqua,

  • K. W. Hoppel,

  • S. S. Krigman,

  • D. L. Kriebel,

  • D. J. Debrestian,

  • C. E. Randall,

  • D. W. Rusch,

  • C. Brogniez,

  • R. Ramananahérisoa,

  • E. P. Shettle,

  • J. J. Olivero,

  • J. Lenoble,

  • P. Pruvost


We present a detailed discussion of the version 5 algorithms used to analyze the Polar Ozone and Aerosol Measurement II (POAM II) solar occultation data. As this version is primarily an ozone and aerosol retrieval, the scope of the paper is limited to these constituents. Forward model algorithms for calculating atmospheric transmittance in the nine POAM II spectral channels are presented, as well as detailed instrument simulation models used to generate synthetic data. Inversion of POAM II data is accomplished in two sequential steps, using the method of optimal estimation. A nonlinear spectral inversion is performed first, which simultaneously inverts the measured slant optical depth profiles in all channels to obtain the separate gas and aerosol optical depth components. Inversion of the slant path line of sight integral then yields altitude profiles of constituent density and aerosol extinction. Examples of both real and simulated retrievals are presented, and comparisons are made between retrievals obtained with the version 5 operational algorithm and an independent algorithm developed at the Laboratoire d'Optique Atmosphérique, University of Lille, Lille, France. A detailed error analysis is also presented, providing estimates of the total random and systematic error budget for the ozone and aerosol retrievals. The results of this analysis indicate random errors of 5% or less in the ozone retrievals in the 10- to 50-km range. Random errors in the retrieved aerosol extinction are estimated to be 10 to 20% in the primary 1060-nm channel, increasing to 20 to 30% at 352 nm. Total systematic errors are estimated to be less than 5% for ozone and as much as 10 to 15% in the 1060-nm aerosol extinction and are primarily caused by systematic biases in the National Center for Environmental Prediction (NCEP) temperatures used in the analysis. Analysis of the retrieval averaging kernels shows that under present aerosol loading conditions, aerosol feedback into the ozone retrievals introduces biases of less than 2% at all altitudes.