Spectral Estimates of the Troposhere Using Least Squares and Maximum Likelihood

Given the significant cost and effort required to field infrared sensors capable of cutting-edge performance, each step of the development process, from experiment concept to calibrated sensor, must be optimized to insure maximum quality in the final processed data. This is particularly important if the final data is to be used as a basis for decisions about the properties of future sensors or for claims about the characteristics of the Troposphere. The spectra estimation process depends critically on how well the technique anticipates and models the operational properties of the system, how well the optical and electrical characterizes of the system are characterized, how closely the temporal properties of the system approximate a linear, time-invariant system, and how well system noise characterization are factored into the process that ends with quality spectra. To this end an alternate spectrum estimation algorithm, Expectation Maximum inversion (EM), is investigated and compared against the standard Fast Fourier Transform (FFT) operating on data collected during the flight of the spectrometer developed for the Far Infrared Spectroscopy of the Troposphere (FIRST) program. A discussion of the characteristics of the FFT and EM transform is given along with some preliminary results