*rpanfili@spectral.com; +1-781-273-4770; spectral.com Hyperspectral Modeling of Combustion Flare Emissions R. Panfili*, P. Vujkovic-Cvijin, X. Tan, R. Kennett, R. Taylor, H. Dothe, and L. Bernstein, Spectral Sciences, Inc., 4 Fourth Avenue, Burlington, MA 01803 ABSTRACT The petroleum refining process utilizes the open-flame combustion of hydrocarbons by industrial flares as a method to eliminate waste gases which cannot be recovered in a commercially viable manner. Efficient, but incomplete, combustion will result in the emission of trace amounts of ozone-forming highly-reactive volatile organic compounds (VOCs) and human carcinogens. There currently exists a need for continuous and autonomous monitoring and control technology to minimize emissions of VOCs and human carcinogens from flare emissions. Monitoring flare emission is challenging due to the low concentrations of the species of interest and varying environmental conditions. Hyperspectral imaging provides a method to achieve these goals without interfering in the operation of combustion flare facilities. We outline the development of a novel combination of modeling and laboratory measurements that utilize hyperspectral sensors to achieve these goals. Keywords: Industrial Pollution, Remote Monitoring 1. INTRODUCTION Techniques which can be used for the remote detection of pollutants arising from the incomplete combustion of hydrocarbons can be separated into two groups. In the first group, gases from the product are captured after combustion has completed and their constituents are non-destructively analyzed. Gas chromatography is an important example of this approach. In the second group, light emitted from the combustion products is measured and a retrieval method is used to interpret the measurements. This can be done by actively stimulating quantum transitions, as is the case with Differential Optic Absorption Spectrometry (UV DOAS). Alternatively, remote sensing can be performed passively. We focus on a passive remote sensing approach to monitoring emission products. The remote detection process involves first capturing the photons emitted by the combustion products and then resolving the molecular distribution that most likely leads to those emissions. This process can be performed using either simulated data or measurements with the results in either case processed by a retrieval algorithm. Simulated emission data is generated by utilizing several well established codes to provide the relevant physics. The ARCHES code [1] is used to describe the combustion process. The SAG code [2], augmented with available weather data, is used to describe the background environment. The SAMM code [3] uses the molecular concentrations from ARCHES and the atmospheric profiles from SAG to perform the radiation transport to the detector. The composite output of these three codes is then used as predictors for an inversion or retrieval algorithm. This inversion algorithm [4], which is conceptually similar to the method to be used for data analysis on the NASA Orbiting Carbon Observatory mission [5], has previously been discussed [6]. The complexity of the physical process being observed challenges the retrieval algorithm and an appropriate choice of detector coupled with its optimal usage will be necessary to obtain accurate results. This approach makes use of hyperspectral imagery, which partitions the radiant signal into a large number of spectral bands. Four our application, we intend to examine a broad swath of the infrared spectrum partitioned into thousands of individual bands. Not all spectral bands will contain relevant information. For example, some may be located at frequencies at which the molecules of interest do not radiate and incorporating those bands in a retrieval algorithm would be nonsensical. An efficient deciphering of the end product of combustion will therefore require sufficient knowledge of the spectroscopy of all molecules of interest. This field is relatively mature for combustion products which are present in the ambient environment and a great deal of existing academic research can be utilized. The spectroscopy of larger hydrocarbons and more exotic products of incomplete combustion, such as highly-reactive volatile organic compounds (HR-VOCs), are not as well developed and the applicability of readily available spectroscopy is a function of the molecules of interest. This paper is separated into this introduction and three additional sections. In Section 2 we provide a discussion of the spectroscopy of ambient radiators which are end products of combustion. In Section 3 we give details of a procedure to obtain spectroscopy for HR-VOCs in cases where they are not available from other sources. Finally, in Section 4 we give conclusions of this work to date and outline future directions for this project. 2. AMBIENT RADIATORS The most significant combustion products as measured by quantity created are molecules that are also present in the ambient environment. This presents us with certain advantages as the radiative properties of the ambient environment have long been studied. Consequently, there exists a large body of academic work which can be directly applied to this problem. Most notably, databases of spectral line properties associated with molecules in the ambient environment are readily available. These lines correspond to specific quantum transitions and the strength of these transitions is a function, primarily, of the temperature of the ambient environment. As such, lines present in a database are directly applicable to higher temperatures encountered when monitoring the end products of a combustion process. Where databases which focus solely on the ambient environment fall short is in their dynamic range. Elevated temperatures increase the number of transitions which need to be accounted for. High-temperature spectroscopic databases account for these transitions, and we outline the specific high-temperature databases used for each molecule of interest in the following subsections. 2.1 Water Water is a primary product of combustion and is also present in varying quantities in the ambient atmosphere. Water concentrations are not directly significant when determining combustion efficiency through remote sensing. However, water is an exceptionally efficient infrared absorption species. Its presence has a significant effect on the pattern of infrared radiation which will be measured by a hyperspectral sensor. Examples of water line positions and strengths are shown in Figure 1. We see emission throughout nearly the entire infrared region and saturation at the heart of the strongest absorption bands. Through an analysis of the measured radiance, it is possible to determine both temperature and water vapor profiles. Figure 1. Background infrared radiance attributable to water at ambient temperatures near the ground. There are numerous databases of spectral line features available for the forward modeling of water radiance. We have made use of three of these databases in this project. The HITRAN database [7] can be used as long as path temperatures do not exceed 1000 K, the upper range of temperatures expected to be encountered in the thermosphere. This temperature limit may be sufficient for evaluating emissions from flares operating at a tiny fraction of design capacity, but will likely be insufficient for higher flow rates. For these instances, the HITEMP database [8] can be used for evaluation. This database is valid for temperatures up to 2000-3000 K, depending on the radiator, and its data format is identical to that of HITRAN. Substituting HTRAN spectroscopy with HITEMP is operationally trivial but comes at a cost of substantially increased computational time. The HITRAN database contains 69,201 water lines while the HITEMP database contains 114,241,164 water lines. The computational time involved in determining line-by-line absorption coefficients from spectroscopic databases scales linearly with the number of entries. Finally, we note that even more sophisticated databases are available [9]. We explored the use of one database with over 500,000,000 water lines and found the additional computational time associated with its usage was unnecessary for this application. Retrieval of water profiles or total water column amount has been attempted by a significant number of remote sensing missions. Examples include MODIS [10,11], which utilized the near-infrared region for its retrieval, as well as MIPAS [12, 13], TES [14] and AIRS [15], which utilized the mid-wave and long-wave infrared region for their retrievals. These sensors have determined column water vapor amounts from the transmittances based on pre-computed theoretical radiative transfer calculations consistent with each sensor's bandpass. Rapid analysis of measurements was performed using a look-up table procedure. An overview of the spectral bands and retrieval algorithm used by these missions is provided in Table 1. These measurements had a challenge unique to airborne retrievals and one which will not come into play when monitoring combustion flares. The retrieved water quantities were sensitive to the effects of spectrally dependent surface reflectance. It is necessary to remove this contribution in order to obtain the correct atmospheric water vapor transmittance. Since combustion flare emission is monitored with ground-based sensors looking up, this contribution will not appear. The retrievals are also very sensitive to the radiative signals of thin transparent clouds. Space-based measurements showing any signature of cloud contamination were excluded from the analysis. We will not have that luxury. Instead, the contribution of cirrus clouds (as well as other cloud types) will need to be accounted for in our retrieval. Table 1: Sampling of space-based sensors which monitor ambient water vapor. Sensor/Satellite Spectral bands Algorithm/data processing scheme Comments MODIS/ Terra and Aqua 0.905, 0.936, and 0.940 μm; atmospheric window channels at 0.865 and 1.24 μm Look-up table procedure The ratios partially remove the effects of variation of surface reflectance with wavelengths MIPAS/Envisat (before failure of the interferometer slide in 2004) 4.15-14.6 μm in 5 spectral bands; spectral resolution of 0.035 cm-1 Constrained nonlinear least squares fitting. The retrievals are very sensitive to the radiative signals of thin transparent clouds MIPAS/Envisat (after failure of the interferometer slide in 2004) 4.15-14.6 μm in 5 spectral bands; spectral resolution of 0.065 cm-1 Constrained inverse modeling of limb radiances TES/Aura 3.3-15.4 μm band with 0.1 cm-1 spectral resolution. Used 7.5-9.1 μm for water Joint estimates of H2O, HDO, CH4, N2O Utilizes improved vertical resolution in the lower troposphere and boundary layer AIRS/Aqua 9-12 μm Physical iterative method using regression retrieval as a first guess. Retrieves temperature and water vapor profiles. 2.2 Carbon Dioxide Carbon dioxide is the primary carbon product expected from hydrocarbon combustion. It is present in the ambient atmosphere in a roughly uniform mixture with other atmospheric constituents. It is also, obviously, a significant product of hydrocarbon combustion. Any remote measure of carbon dioxide concentration will necessarily capture molecules arising from both the ambient environment and combustion product. Unlike water, carbon dioxide radiates in the infrared in a series of well-defined bands. Its fundamental bending mode radiates around 15 μm, which can be used for retrieval and a vibrational bending mode which we not use directly for retrievals. Its fundamental stretching modes appear at 7.2 and 4.3 μm, respectively. Examples of carbon dioxide line positions and strengths are shown in Figure 2. We see emission bands corresponding to the fundamental or overtone transitions except primarily in the 4.3 micron region where we see carbon dioxide primarily through absorption. Figure 2. Location and strength of infrared carbon dioxide lines. As was the case with water, there are numerous databases of spectral line features available for the forward modeling of carbon dioxide radiance. We once again make use of the HITRAN database for path temperatures which do not exceed 1000 K and the HITEMP database for higher temperatures. These databases span the range of spectral values we will be evaluating but are by no means the only available databases. Other examples include a database maintained by the Jet Propulsion Laboratory [16], CDSD-4000 [17] and HITELOR [18]. There are a number of remote sensing systems which monitor ambient carbon dioxide, some of which are listed in Table 2, owing to its importance in climate research. Examples include the TANSO/FTS instrument aboard GOSAT [19], the SCHIAMACHY instrument aboard Envisat [20, 21], and the Orbiting Carbon Observatory [22]. Unfortunately, and unlike other molecules, the bands used by these instruments for remote sensing of carbon dioxide cannot be used directly for this application. All three programs make use of the shortwave infrared bands centered at 1.6 and 2.0 μm. These have the advantage of being relatively clear of radiance contributions from other molecules, greatly simplifying the retrieval process. These bands have the disadvantage of occurring in a region where our proposed hyperspectral imager lacks sensitivity. That leaves us with the 15 micron and 4.3 micron bands utilized by AIRS [23] and perhaps the 2.7 micron band. All three regions pose challenges. The 15 micron band is a relatively weak signal, but is very temperature-sensitive. It is unfortunately once again in a region where our proposed instrument begins to lose sensitivity. The 4.3 micron band is an extremely efficient absorption region and we expect to encounter scenarios in which all photons along the path are absorbed except, perhaps in the wings of the band. Finally, the 2.7 micron band contains strong water features. It will be necessary to accurately separate the radiance response for these two radiators if we are to use this band. Table 2: Sampling of space-based sensors which monitor ambient carbon dioxide. Sensor/Satellite Spectral bands Algorithm/data processing scheme Comments TANSO-FTS/GOSAT 1.567-1.618 μm Scenes selected to be cloud-free. Retrieval quality examined to exclude low-quality and/or aerosol-contaminated results. SCIAMACHY/ Envisat Short-wave infrared upgrad (v2) of the WFM-DOAS retrieval algorithm The retrieved mole fractions are compared to global model simulations (CarbonTracker XCO2 and TM5 XCH4) optimised by surface measurements SCIAMACHY/ Envisat 1.0-1.7 μm (SCIAMACHY channel 6) Artificial Neural Network Technique Retrieval of CO2 Vertical Concentration Profiles AIRS/Aqua Long-wave infrared CO2 forecast models by ensemble Kalman filtering 6 hourly 3-D CO2 fields OCO/ Taurus XL launch vehicle, (launch failure) Two CO2 bands at 1.61 and 2.06 μm Channel Selection method using information content analysis Identified the 20 best channels from each CO2 spectral region to use in OCO retrievals. 4.3 and 15 μm absorption bands (AIRS or TES) are sensitive to temperature. 2.3 Carbon Monoxide Carbon monoxide is the primary carbon product expected from incomplete hydrocarbon combustion. It is also present in the ambient atmosphere in a roughly uniform mixture with other atmospheric constituents. As with carbon dioxide, any remote measure of carbon monoxide concentration will necessarily capture molecules arising from both the ambient environment and combustion product. Most carbon monoxide infrared emission is present in either the 2.3 or 4.7 micron band. Examples of carbon monoxide line positions and strengths are shown in Figure 3. As with carbon dioxide, we utilize the HITRAN and HITEMP spectral databases for carbon monoxide lines. Figure 3. Location and strength of infrared carbon monoxide lines. Remote sensing measurements of carbon monoxide have made use of both of these bands. Examples of atmospheric carbon monoxide monitoring instruments, some of which are summarized in Table 3, include MOPITT [24], AIRS [25], SCIAMACHY [26], TES [26], IASI [27,28], and IMG [29] sensors. A variety of algorithms including digital gas correlation, nonlinear least squares, global fit adjustment, maximum likelihood and neural networks are being used. These sensor and algorithms are mature, and the agreement between different sensors and with in situ measurements is good. The relatively small remaining differences can be largely explained by different averaging kernels and the use of different sources of a priori information. Table 3: Sampling of space-based sensors which monitor ambient carbon monoxide. Sensor/Satellite Spectral bands Algorithm/data processing scheme Comment MOPITT/Terra 4.7 um band retrieval algorithm based on the maximum likelihood method AIRS/Aqua 4.54-4.63um alternative retrieval algorithm for AIRS using the Optimal Estimation (OE) technique AIRS OE results are much more realistic than AIRS V5 operational results, especially in the lower troposphere. SCIAMACHY/ Envisat 2.3 um GEOSChem Using GEOSChem for intercomparison platform shows global consistency between the different satellite datasets. TES/Aura 4.7 um GEOSChem for, AIRS, and SCIAMACHY. TES: a posteriori Using GEOSChem as a common intercomparison platform shows global consistency between the different satellite datasets and with the in situ data. IASI/MetOp-A 4.7 μm University of Leicester IASI Retrieval Scheme (ULIRS) IMG 4.7 um Digital gas correlation, nonlinear least squares, global fit adjustment, and neural networks CO column amounts agreed within 4% with MOPITT, TES and IASI. Larger discrepancies obtained for the IMG instrument. The assumed vertical temperature profile shown to be a critical parameter for accurate CO retrieval. 2.4 Methane Methane is another significant product of incomplete combustion which is also present in the ambient atmosphere. Both ambient and combustion-generated methane concentrations must be accounted for in order to obtain an accurate retrieval. Methane has a large number of degrees of freedom owing to the number of atoms in the molecule as well as significant degeneracies in its vibrational modes owing to the symmetry of the molecule. The four most significant regions for infrared retrieval are the triply-degenerate bending mode around 7.6 μm, the doubly-degenerate bending mode at 6.5 μm and two stretching modes around 3.3 μm. An example of methane line positions and strengths is shown in Figure 4. We utilize the HITRAN and HITEMP spectral databases for methane lines but recognize significant deficiencies exist owing to the complexity of methane spectroscopy. Figure 4. Location and strength of infrared methane lines. Atmospheric retrieval of methane has experienced a recent period of intense academic interest. This is due in no small part to the Cassini-Huygens probe and its observations of Saturn's moon Titan. Methane spectroscopy is critically important to interpreting these observations as it is a significant component of Titan's atmosphere [30]. This has led to several new spectroscopic databases of methane lines, including one specifically constructed to better understand Titan's radiance signature [31] and a second of more terrestrial origins [32]. Atmospheric methane is monitored by numerous satellites, including TANSO/FTS [33], SCIAMACHY [20], MOPITT [24] and AIRS [19,34] instruments in the near-infrared band around 1.65 μm, mid-wave infrared band around 3.4 μm and in the long-wave infrared band around 7.6 μm. A summary of these instruments is provided in Table 4. The long-wave infrared observations are sensitive to carbon dioxide and methane in the middle to upper troposphere, whereas short-wave infrared observations can address gas abundances near the surface. Because the major sources and sinks of carbon dioxide and methane exist near the surface, short wave infrared observations are typically used more than long-wave infrared observations for satellite-based remote sensing. The calculus of a ground-based measurement differs and both the mid-wave and long-wave band can be used for our purposes. Table 3: Sampling of space-based sensors which monitor ambient methane. Sensor/Satellite Spectral bands Algorithm/data processing scheme Comment TANSO/FTS GOSAT 1.65 um Full physics retrieval algorithm by NASA Orbiting Carbon Observatory (OCO) Short-wave infrared providing sensitivity to the near-surface absorbers, while thermal infrared providing mid-tropospheric sensitivity. SCIAMACHY/ Envisat Short-wave infrared upgraded version (v2) of the retrieval algorithm WFM-DOAS CH4 results show that atmospheric methane has started to rise again in recent years, which is consistent with surface measurements. MOPITT/Terra 2.26-4.62 um Maximum likelihood method Instrument description, algorithm description AIRS/Aqua near 7.6 μm Comparison with in situ aircraft CH4 profiles implied a 2% increase in methane absorption coefficients for strong absorption channels 3. OTHER HYDROCARBONS We also require HR-VOC spectroscopy in to predict their concentrations in combustion flare exhaust plumes. Unfortunately, the spectroscopy of HR-VOCs is notably less mature. Spectral databases for many radiators in the EPA CERCLA list can be found scattered throughout a collection of spectral databases, but the HR-VOCs of current concern (benzene, toluene, ethylbenzene and xylene) are not readily available. The approach we have taken is to turn to analytic methods as a means of obtaining the necessary spectral data. We have achieved varying degrees of success, dependent strongly upon the size and configuration of the molecule. Here we outline the approach taken to generate the necessary spectroscopy for HR-VOCs. We then provide a few comparisons of the cross-section obtained from this analytical approach when compared to experimentally measured HR-VOC cross-sections. 3.1 Procedure For cases in which necessary molecular spectra are not available, we will rely on analytical methods to produce the needed spectroscopy. The procedure to generate spectra utilizes well-established physical chemistry models which exploit the relationship between the structure of a molecule and its optical signature. The frequency and intensity of peaks in any molecule's spectrum are directly related to its vibrational and rotational energy levels and the temperature of the background environment. The process of analytically constructing a molecule's spectroscopy then reduces to determining the energy spacing and strength of allowed quantum transitions. For small molecules, this has often been performed by finding the set of normal modes and the coupling between those modes. The same approach can be taken for the smallest HR-VOCs, such as formaldehyde where normal modes can be determined experimentally. This approach becomes intractable for molecules containing more than three or four atoms. Another tool will be needed to construct the spectroscopy of larger molecules. In those cases, we utilize quantum mechanical and statistical mechanical methods to match individual features to specific transitions. The process of generating HR-VOC spectroscopy makes extensive use of existing algorithms developed by the academic community. These algorithms perform first principles quantum mechanical calculations to predict the normal mode spectrum of relatively large HR-VOCs. As one example, these algorithms have been recently used to calculate harmonic and anharmonic properties of organic molecules from quantum chemistry calculations models such as Gaussian03 [35], GAMESS [36, 37] and NWCHEM [38]. Combining the output of these simulations with statistical mechanical notions [39], we can predict the spectroscopic signatures of HR-VOCs. The end result is a series of molecule-specific spectroscopic databases functionally equivalent to those available for ambient radiators. This database of line strengths can be used directly to calculate absorption cross-sections for these molecules at arbitrary temperatures and pressures in order to interpret hyperspectral data [40]. 3.2 Examples The accuracy of these calculations can be surmised through a comparison with experimental measurements. Perhaps the best current collection of HR-VOC experimental cross-sections has been assembled by the Pacific Northwest National Laboratory (PNNL). These measurements provide a reasonably high spectral resolution at between one and three fixed temperature and pressure combinations, depending on the particular molecule of interest. We make use of the cross-sections contained in the PNNL infrared spectral library (IRSL) for validation and to identify the any shortcomings in our approach. We will begin by providing an example of a small and intermediately-sized hydrocarbon. The small hydrocarbon example is formaldehyde, a carcinogen present on the Environmental Protection Agency Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) list. The spectroscopy of this molecule is reasonably well-understood and spectral line databases are available in the open literature [7], a useful third information source for the purposes of validating the procedure. A comparison of the spectrum obtained through analytically computed spectroscopy and the measured spectrum in the PNNL IRSL database is shown in the left-hand plot of Figure 5. The second example shows an isomer of butane, a hydrocarbon which may or may not be present in a fuel mixture. The resulting comparison is shown in the right-hand plot of Figure 5. We emphasize that the PNNL IRSL database was not used in any part of the analytic process. No corrections have been applied to the calculated spectrum in order to improve the fit. Figure 5. Comparison between model and measurement for the cross-sections of formaldehyde (left) and 2-butene (right). The final examples show cross-sections generated for much larger HR-VOCs. Figure 6 shows a comparison between the PNNL measurement and computed values for three molecules. The example molecules are ethyl-benzene (left), toluene (center) and ortho-xylene (right). We see the center position of the measurement and computation overlap indicating the computation is accurately determining the center value of the energy spacing. One notable exception is one of the peaks of toluene around 700 wavenumbers is missing. Less successful is the modeling of the cross section falloff along the center point value. While the total frequency-integrated cross-sections are approximately equal, indicating the total band transition strength has been accurately computed, the contour of the computed band profile does not match what has been experimentally measured. As the molecules being modeled become larger, determining normal modes and coupling constants becomes more complex. The illustrated fits can be improved in principle through further increasing the level of theory applied to the computations at an increased computational cost. Figure 6. Comparison between model and measurement for the cross-sections of three hydrocarbons. 4. CONCLUSIONS AND FUTURE DIRECTION The remote detection of HR-VOCs in the exhaust plume of a combustion flare is a challenging task. A detailed analysis of hyperspectral imagery provides an opportunity to quantifying HR-VOC concentrations from remote detection observations [41]. This process of retrieving path quantities from observed radiance measurements requires detailed molecular spectroscopy. We have outlined the procedure currently used, which is to utilize existing molecular spectroscopic databases when available and to perform quantum chemistry computations to construct an equivalent database when not available. The examples provided illustrate the construction of cross-sections from libraries of publicly available spectroscopic databases as well as cross-sections for molecules which spectroscopy have been analytically obtained. ACKNOWLEDGEMENTS Funding for this research comes from the Department of Energy through Contract No. DE-SC003373. Substantial support has been provided to this project by Dr. Jeffrey A. Mercier of Sandia National Laboratory. We also acknowledge numerous substantive discussions of benefit to this effort with Mr. Russ Nettles of the Texas Commission on Environment Quality. REFERENCES [1] Shroll, R. M., S. Adler-Golden, J. W. Duff and J. H. Brown, "Users' Manual for SAG-2, SHARC/SAMM Atmosphere Generator," AFRL-TR-03-1530 (2003). [2] Spinti, J.P., Thornock, J.N., Eddings, E.G., Smith, P.J., and Sarofim, A.F., "Heat Transfer to Objects in Pool Fires," in Transport Phenomena in Fires [3] H. Dothe, H., Duff, J.W., Panfili, R., Gruninger, J.H., Kennett, R.G., Acharya, P.K., Berk, A., and Bernstein, L.S., "Users' Manual for SAMM® Version 3," SSI-TR-599 (2010). , WIT Press, Southampton, U.K. (2008). [4] Connor, B.J., Boesch, H., Toon, G., Sen, B., Miller, C., and Crisp, D., "Orbiting Carbon Observatory: Inverse method and prospective error analysis," Journal of Geophysical Research 113, D05305 (2008). [5] Crisp, D., Atlas, R.M., Breon, F.-M., Brown, L.R., Burrows, J.P., Ciais, P., Connor, B.J., Doney, S.C., Fung, I.Y., Jacob, D.J., Miller, C.E., O'Brien, D., Pawson, S., Randerson, J.T., Rayner, P., Salawitch, R.J., Sander, S.P., Sen, B., Stephens, G.L., Tans, P.P., Toon, G.C., Wennberg, P.O., Wofsy, S.C., Yung, Y.L., Kuang, Z., Chudasama, B., Sprague, G., Weiss, B., Pollock, R., Kenyon, D. and Schroll, S., "The Orbiting Carbon Observatory (OCO) mission," Advances in Space Research 34, 700-709 (2004). [6] R. Panfili, H. Dothe, P. Vujkovic-Cvijin, X. Tan, R. Kennett, R. Taylor, L.S. Bernstein, P.J. Smith, J.N. Thornock, K. Gross, and J. Seebold, "High-fidelity modeling of flare combustion and emission detection," American Flame Research Committee Industrial Flare Colloquium, Houston, Texas (2011). [7] L.S. Rothman, I.E. Gordon, A. Barbe, D. ChrisBenner, P.F. Bernath, M. Birk, V. Boudon, L.R. Brown, A. Campargue, J.-P. Champion, K. Chance, L.H. Coudert, V. Dana, V. M.Devi, S. Fally, J.-M. Flaud, R.R. Gamache, A. Goldman, D. Jacquemart, I. Kleiner, N. Lacome, W.J. Lafferty, J.-Y. Mandin, S.T. Massie, S.N. Mikhailenko, C.E. Miller, N. Moazzen-Ahmadi, O.V. Naumenko, A.V. Nikitin, J. Orphal, V.I. Perevalov, A. Perrin, A. Predoi-Cross, C.P. Rinsland, M. Rotger, M.Simeckova, M.A.H. Smith, K. Sung, S.A. Tashkun, J. Tennyson, R.A. Toth, A.C. Vandaele, and J. VanderAuwera, "The HITRAN 2008 molecular spectroscopic database," Journal of Quanitative Spectrosscopy and Radiative.Transfer 110, 533-572 (2009). [8] L.S. Rothman, I.E. Gordon, R.J. Barber, H. Dothe, R.R. Gamache, A. Goldman, V.I. Perevalov, S.A. Tashkun, and J. Tennyson, "HITEMP, the high-temperature molecular spectroscopic database," Journal of Quanitative Spectrosscopy and Radiative.Transfer 111, 2139-2150 (2010). [9] Barber R.J., Tennyson J., Harris G.J., Tolchenov R.N., "A high accuracy computed water line list," Monthly Notices of the Royal Astronomical Society 368, 1087 (2006). [10] Gao, B.C. and Kaufman, Y.J., "Water vapor retrievals using Moderate Resolution Imaging Spectroradiometer (MODIS) near-infrared channels," Journal of Geophysical Research 108, 4389 (2003). [11] Gao, B.C. and Kaufman, Y.J., "The MODIS Near-IR Water Vapor Algorithm, Product ID: MOD05," Algorithm Technical Background Document, http://modis-atmos.gsfc.nasa.gov/_docs/atbd_mod03.pdf. [12] M. Milz, T. von Clarmann, H. Fischer, N. Glatthor, U. Grabowski, M. Hopfner, S. Kellmann, M. Kiefer, A. Linden, G. Mengistu Tsidu, T. Steck, G. P. Stiller, B. Funke and M. Lopez-Puertas, and M.E. Koukouli, "Water vapor distributions measured with the Michelson Interferometer for Passive Atmospheric Sounding on board Envisat (MIPAS/Envisat)," Journal of Geophysical Research 110, D24307 (2005). [13] Stiller, GP, Kiefer, M., Eckert, E., von Clarmann, T., Kellmann, S., Garcia-Comas, M., Funke, B., Leblanc, T., Fetzer, E., Froidevaux, L., et al., "Validation of MIPAS IMK/IAA temperature, water vapor, and ozone profiles with MOHAVE-2009 campaign measurements," Atmospheric Measurement Techniques 5, 289-320 (2012). [14] Worden, J., Kulawik, S., Frankenberg, C., Payne, V., Bowman, K., Cady-Peirara, K., Wecht, K., Lee, J.E., and Noone, D., "Profiles of CH4, HDO, H2O, and N2O with improved lower tropospheric vertical resolution from Aura TES radiances," Atmospheric Measurement Techniques 5, 397-411 (2012). [15] Kwon, E.H., Li, J., Li, J., Sohn, BJ, and Weisz, E., "Use of total precipitable water classification of a priori error and quality control in atmospheric temperature and water vapor sounding retrieval," Advances in Atmospheric Sciences 29, 263-273 (2012). [16] Toth, R.A., Brown, LR, Miller, CE, Malathy Devi, V., and Benner, D.C., "Spectroscopic database of CO2 line parameters: 4300-7000 cm-1," Journal of Quantitative Spectroscopy and Radiative Transfer 109, 906-921 (2008). [17] Tashkun SA, Perevalov VI., "CDSD-4000: High-resolution, high-temperature carbon dioxide spectroscopic databank," Journal of Quantitative Spectroscopy and Radiative Transfer 112, 1403-1410 (2011). [18] Scutaru D, Rosenmann L, Taine J., "Approximate intensities of CO2 hot bands at 2.7, 4.3 and 12 μm for high temperature and medium resolution applications," Journal of Quantitative Spectroscopy and Radiative Transfer 52, 765-781 (1994). [19] Yoshida, Y., Ota, Y., Eguchi, N., Kikuchi, N., Nobuta, K., Tran, H., Morino, I., and Yokota, T., "Retrieval algorithm for CO2 and CH4 column abundances from short-wavelength infrared spectral observations by the Greenhouse Gases Observing Satellite," Atmospheric Measurement Techniques 4, 717-734 (2011). [20] Schneising, O., Buchwitz, M., Reuter, M., Heymann, J., Bovensmann, H., and Burrows, JP, "Long-term analysis of carbon dioxide and methane column-averaged mole fractions retrieved from SCIAMACHY," Atmospheric Chemistry and Physics 11, 2863-2880 (2011). [21] Carvalho, A.R., Ramos, F.M., and Carvalho, J.C., "Retrieval of carbon dioxide vertical concentration profiles from satellite data using artificial neural networks," TEMA-Tendencias em Matematica Aplicada e Computacional 11, 205--216 (2010). [22] Kuai, L., Natraj, V., Shia, R.L., Miller, C., and Yung, Y.L., "Channel selection using information content analysis: A case study of CO2 retrieval from near infrared measurements," Journal of Quantitative Spectroscopy and Radiative Transfer 111, 1296-1304 (2010). [23] J. Liu, I. Fung, E. Kalnay, J.-S. Kang, E.T. Olson and L. Chen, "Simultaneous assimilation of AIRS Xco2 and meteorological observations in a carbon climate model with an ensemble Kalman filter," Journal of Geophysical Research 117, D05309 (2012). [24] Wang, J., J. C. Gille, P. L. Bailey, L. Pan, D. Edwards, and J. R. Drummond, "Retrieval of tropospheric carbon monoxide profiles from high-resolution interferometer observations: A new digital gas correlation (DGC) method and applications," Journal of Atmospheric Science 56, 219-232 (1999). [25] J. X. Warner, Z. Wei, L. L. Strow, C. D. Barnet, L. C. Sparling, G. Diskin, and G. Sachse, "Improved agreement of AIRS tropospheric carbon monoxide products with other EOS sensors using optimal estimation retrievals," Atmospheric Chemistry and Physics Discussions 10, 11851-11883 (2010). [26] M. Kopacz, D. J. Jacob, J. A. Fisher, J. A. Logan, L. Zhang, I. A. Megretskaia, R. M. Yantosca, K. Singh, D. K. Henze, J. P. Burrows, M. Buchwitz, I. Khlystova, W. W. McMillan, J. C. Gille, D. P. Edwards, A. Eldering, V. Thouret, and P. Nedelec, "Global estimates of CO sources with high resolution by adjoint inversion of multiple satellite datasets (MOPITT, AIRS, SCIAMACHY, TES)," Atmospheric Chemistry and Physics 10, 855-876 (2010). [27] Illingworth, SM, Remedios, JJ, Boesch, H., Moore, DP, Sembhi, H., Dudhia, A., and Walker, JC, "ULIRS, an optimal estimation retrieval scheme for carbon monoxide using IASI spectral radiances: sensitivity analysis, error budget and simulations Atmospheric Measurement Techniques 4, 269-288 (2011). [28] Illingworth, SM, Remedios, JJ, Boesch, H., Moore, DP, Sembhi, H., Dudhia, A., and Walker, J., "A new optimal estimation retrieval scheme for carbon monoxide using IASI spectral radiances--Part 1: Sensitivity analysis, error budget and simulations," Atmospheric Measurement Techniques Discussions 3, 3747-3802 (2010). [29] Cathy Clerbaux, Juliette Hadji-Lazaro, Sébastien Payan, Claude Camy-Peyret, Jinxue Wang, David P. Edwards, and Ming Lu, "Retrieval of CO from nadir remote-sensing measurements in the infrared by use of four different inversion algorithms," Applied Optics 41, 7068-7078 (2002). [30] Panfili, R. and Dothe, H. and Gruninger, J. and Duff, J.W. and Brown, J.H., "Radiance computations for Planetary Atmospheres under Non-Equilibrium Conditions," Advanced Maui Optical and Optical and Space Surveillance (AMOS) Technologies Conference Proceedings, (2007) [31] Bailey, J., Ahlsved, L., and Meadows, VS, "The near-IR spectrum of Titan modeled with an improved methane line list," Icarus 212, 218-232 (2011). [32] Nikitin, AV, Lyulin, OM, Mikhailenko, SN, Perevalov, VI, Filippov, NN, Grigoriev, IM, Morino, I., Yokota, T., Kumazawa, R., and Watanabe, T., "GOSAT-2009 methane spectral line list in the 5550--6236cm-1 range," Journal of Quantitative Spectroscopy and Radiative Transfer 111, 2211-2224 (2010). [33] Robert Parker, Hartmut Boesch, Austin Cogan, Annemarie Fraser, Liang Feng, Paul I. Palmer, Janina Messerschmidt, Nicholas Deutscher, David W. T. Griffith, Justus Notholt, Paul O. Wennberg, and Debra Wunch, "Methane observations from the Greenhouse Gases Observing SATellite: Comparison to ground-based TCCON data and model calculations," Geophysical Research Letters 38, L15807 (2011). [34] Xiaozhen Xiong, Chris Barnet, Eric Maddy, Colm Sweeney, Xingpin Liu, Lihang Zhou, and Mitch Goldberg, "Characterization and validation of methane products from the Atmospheric Infrared Sounder (AIRS)," Journal of Geophysical Research 113, G00A01 (2008). [35] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, Montgomery, Jr., J. A., T. Vreven, K. N. Kudin, J. C. Burant, J. M. Millam, S. S. Iyengar, J. Tomasi, V. Barone, B. Mennucci, M. Cossi, G. Scalmani, N. Rega, G. A. Petersson, H. Nakatsuji, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, M. Klene, X. Li, J. E. Knox, H. P. Hratchian, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, P. Y. Ayala, K.Morokuma,G. A. Voth, P. Salvador, J. J. Dannenberg, V. G. Zakrzewski, S. Dapprich, A. D. Daniels, M. C. Strain, O. Farkas, D. K. Malick, A. D. Rabuck, K. Raghavachari, J. B. Foresman, J. V. Ortiz, Q. Cui, A. G. Baboul, S. Clifford, J. Cioslowski, B. B. Stefanov, G. Liu, A. Liashenko, P. Piskorz, I. Komaromi, R. L. Martin, D. J. Fox, T. Keith, M. A. Al- Laham, C. Y. Peng, A. Nanayakkara, M. Challacombe, P. M. W. Gill, B. Johnson, W. Chen, M. W. Wong, C. Gonzalez, and J. A. Pople, "Gaussian 03 manual," (2004). [36] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.H. Jensen, S. Koseki, N. Matsunaga, K.A. Nguyen, S.J. Su, T.L.Windus,M. Dupuis, and J.A. Montgomery, "General atomic and molecular electronic structure system," Journal of Computational Chemistry 14, 1347-1363 (1993). [37] M.S. Gordon and M.W. Schmidt, "Advances in electronic structure theory: GAMESS a decade later," Theory and Applications of Computational Chemistry, the first forty years 14, 1167-1189 (2005). [38] M. Valiev, E.J. Bylaska, N. Govind, K. Kowalski, T.P. Straatsma, H.J.J van Dam, D. Wang, J. Nieplocha, E. Apra, T.L. Windus, and W.A. de Jong, "NWCHEM: a comprehensive and scalable open-souce solution for large schale molecular simulations," Computer Physics Communications 101, 1477 (2010). [39] M. Page, C. Doubleday Jr., and J.W. McIver Jr., "Following steepest descent reaction paths the use of higher energy derivatives with ab initio electronic-structure methods," Journal of Chemical Physics 93, 5634-5642 (1990). [40] R. Panfili and R. Taylor, "Utilizing Hyperspectral Imagery to Identify Highly-Reactive Volatile Organic Compounds in Combustion Flares," IEEE Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, Shanghai, China (2012). [41] R. Panfili, P. Vujkovic-Cvijin X. Tan, R. Kennett, R. Taylor, H. Dothe, L. Bernstein, P. Smith, J. Thornock, K. Gross and J. Seebold, "Using models to interpret remote sensing measurements of industrial flare emissions," 34th Annual Review of Atmospheric Transmission Models Meeting, Albuquerque, New Mexico (2012).