The Ill-posed Nature of the Satellite Temperature Retrieval Problem and the Limits of Retrievability

Owen E. Thompson Department of Meteorology, University of Maryland, College Park, MD 20742

Search for other papers by Owen E. Thompson in
Current site
Google Scholar
PubMed
Close
,
Donald D. Dazlich Department of Meteorology, University of Maryland, College Park, MD 20742

Search for other papers by Donald D. Dazlich in
Current site
Google Scholar
PubMed
Close
, and
Yu-Tai Hou Department of Meteorology, University of Maryland, College Park, MD 20742

Search for other papers by Yu-Tai Hou in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The inverse problem of satellite temperature profile retrieval is well known to be ill-posed. Ibis means that not only is a vertical temperature profile solution not unique, but that two solutions can be very different from each other. A set of atmosphere-like, and true atmospheric examples of significantly dissimilar inverse solutions, were sought and found, using an 11-channel simulated HIRS sounding radiometer. Using the Riemann-Lebesgue Lemma for guidance, it is shown that simultaneous, numerical solutions of an atmospheric character may differ by as much as 10 K between 10–1000 mb. However, an empirical search for dissimilar solutions in the natural atmosphere reveals an extremely low probability of finding two significantly different RAOBs which produce radiance measurements whose differences cannot be resolved by the satellite radiometer. The empirical results are used to derive a first estimate of the limits of retrievability, analogous to the limits of predictability derivable from the ill-posed nature of the numerical weather prediction problem.

Abstract

The inverse problem of satellite temperature profile retrieval is well known to be ill-posed. Ibis means that not only is a vertical temperature profile solution not unique, but that two solutions can be very different from each other. A set of atmosphere-like, and true atmospheric examples of significantly dissimilar inverse solutions, were sought and found, using an 11-channel simulated HIRS sounding radiometer. Using the Riemann-Lebesgue Lemma for guidance, it is shown that simultaneous, numerical solutions of an atmospheric character may differ by as much as 10 K between 10–1000 mb. However, an empirical search for dissimilar solutions in the natural atmosphere reveals an extremely low probability of finding two significantly different RAOBs which produce radiance measurements whose differences cannot be resolved by the satellite radiometer. The empirical results are used to derive a first estimate of the limits of retrievability, analogous to the limits of predictability derivable from the ill-posed nature of the numerical weather prediction problem.

Save