1404 Volume 55, Number 10, 2001 APPLIED SPECTROSCOPY 0003-7028 / 01 / 5510-1404$2.00 / 0 q 2001 Society for Applied Spectroscopy Comparison of Spectral and Interferogram Processing Methods Using Simulated Passive Fourier Transform Infrared Remote Sensing Data RONALD E. SHAFFER * and ROGER J. COMBS Naval Research Laboratory, Chemistry Division, Code 6116, Washington, D.C. 20375 (R.E.S.); and U.S. Army, SBC Command, Aberdeen Proving Ground, Maryland 21010 (R.J.C.) Computer-generated synthetic single-beam spectra and interfero- grams provide a means of comparing signal processing strategies that are employed with passive Fourier transform infrared (FT-IR) sensors. With the use of appropriate radiance models and spectrom- eter characteristics, synthetic data are generated for one-, two-, and four-component mixtures of organic vapors (ethanol, methanol, ac- etone, and methyl ethyl ketone) in two passive FT-IR remote sensing scenarios. The single-beam spectra are processed by using Savitsky– Golay smoothing and rst-derivative and second-derivative lters. Interferogram data are processed by Fourier ltering using Gauss- ian-shaped bandpass digital lters. Pattern recognition is performed with soft independent modeling of class analogy (SIMCA). Quan- titative models for the target gas integrated concentration-path- length product are built by using either partial least-squares (PLS) regression or locally weighted regression (LWR). Pattern recogni- tion and calibration models of the ltered spectra or interferograms produced comparable results. Discrimination of target analytes in complex mixtures requires a sufciently large temperature differ- ential between the infrared background source and analyte cloud. Quantitative analysis is found to be possible only when the temper- ature of the analyte cloud is stable or known and differs signi- cantly from the background temperature. Net analyte signal (NAS) methods demonstrate that interferogram and spectral processing methods supply identical information for multivariate pattern rec- ognition and calibration. Index Headings: Chemometrics; Multivariate calibration; Pattern recognition; FT-IR; Interferogram; Single-beam spectra; Radio- metric model; Remote sensing; Net analyte signal. INTRODUCTION In Fourier transform infrared (FT-IR) remote sensing, an interferometer-based optical system is used to monitor the atmosphere between the spectrometer and an infrared source. 1–3 FT-IR remote sensing measurements that rely on an instrumentally controlled infrared source are termed active measurements, while those that view the uncontrolled background infrared radiation present in the operating environment are categorized as passive. Be- tween the two congurations, the passive conguration is much more operationally versatile because no con- trolled background infrared source is required. This ver- satility makes it well suited for both ground-based and airborne operation scenarios. Passive FT-IR sensors offer a potential for a wide variety of applications ranging from regulating stack emission to hazardous waste site reme- diation monitoring. Airborne measurements using passive Received 15 February 2001; accepted 13 June 2001. * Author to whom correspondence should be sent. Current address: General Electric, Corporate Research and Development, P.O. Box 8, Schenectady, NY 12301. FT-IR spectrometry are an important approach to iden- tifying and assessing the downwind environmental im- pact of hazardous pollutants. Despite the obvious instrumental advantages, passive FT-IR has primarily remained a research tool, and prac- tical use of the technology for chemical analysis is not widespread outside of the military. 4–7 This situation is in part due to the various difculties that are often encoun- tered in the analysis of passive FT-IR remote sensing data. The primary requirement for a detection capability is the need for a signicant radiometric temperature dif- ferential between the analyte cloud and the infrared back- ground. This temperature difference, as well as the an- alyte concentration, determines the relative strength of the spectral signature. Szczepanski and Fountain have an- alyzed a xed concentration of carbon dioxide with a 4 cm 21 resolution laboratory FT-IR spectrometer and multi- path heated gas cell for investigating the effects on quan- tication of analyte temperature impacting both signal in- tensity and band shape. 8 Mattu et al. have studied the effects of analyte (sulfur dioxide) concentration and tem- perature on quantitative analysis with an 8 cm 21 resolu- tion passive FT-IR conguration viewing low-angle back- grounds through a short-pathlength heated gas cell. 9 Low- angle sky backgrounds contain signicant interferent contributions due to the presence of minor atmospheric constituents viewed through a long atmospheric path. Ex- traction of the analyte spectral signature from the back- ground and interferent features as well as correction for the internal emission prole of the spectrometer places stringent requirements on the interferometer thermal sta- bility 10 and the signal processing capabilities. An addi- tional requirement on a ruggedized sensor is the need for accurate interferogram sampling in often environmentally harsh conditions. Several solutions to these challenges have been pro- posed. Jaakkola et al. 11 have shown that in active FT-IR remote sensing, low-resolution (.4 cm 21 ) spectra are suf- cient for automated quantitative and qualitative analysis. Acquiring low-resolution spectra enhances the reliability of the interferometer by reducing the moving mirror ac- tuation distance needed. Some processing and measure- ment schemes have been devised for cases where the spectrometer is stationary. For example, a differential de- tection algorithm based on a dual-beam interferometer design has been developed for the passive detection of atmospheric vapors. 12,13 This approach is based on the assumption that the dual-beam spectrometer design per- forms an optical subtraction of adjacent elds of view,