Remote sensing: survival strategies in the jungle of averaging kernels and covariance matrices

Outer space, stars, exoplanets, planets in the solar system, and even the Earth's middle and upper atmosphere have in common that it is inconvenient, expensive, and often technically unfeasible to make in situ measurements there. Remote sensing, e.g., by means of radiance measurements, is a relatively cheap and convenient alternative. The conversion of the measured radiances to the quantities of interest, e.g., temperature and composition distributions of atmospheres, typically involves ill-posed and often even under-determined inverse radiative transfer calculations. These can be solved by adding some sort of prior information on the atmospheric state to the measurement. Observations containing prior information, however, are prone to misinterpretation, because it is not obvious which fraction of the results is based on real measurement information and which is just prior information the scientist has included. Diagnostic tools like averaging kernels and smoothing error estimates help to correctly interpret the data but are not without pitfalls. Related problems will be discussed using measurements of the Earth's atmosphere by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) as an example.

24/06/2014 - 14:30
Dr. Thomas von Clarmann
KIT / Germany