IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
1
Validating the AIRS Version 5 CO Retrieval With DACOM In Situ Measurements During INTEX-A and -B W. W. McMillan, Keith D. Evans, Christopher D. Barnet, Eric S. Maddy, Glen W. Sachse, and Glenn S. Diskin
Abstract—Herein we provide a description of the atmospheric infrared sounder (AIRS) version 5 (v5) carbon monoxide (CO) retrieval algorithm and its validation with the DACOM in situ measurements during the INTEX-A and -B campaigns. All standard and support products in the AIRS v5 CO retrieval algorithm are documented. Building on prior publications, we describe the convolution of in situ measurements with the AIRS v5 CO averaging kernel and first-guess CO profile as required for proper validation. Validation is accomplished through comparison of AIRS CO retrievals with convolved in situ CO profiles acquired during the NASA Intercontinental Chemical Transport Experiments (INTEX) in 2004 and 2006. From 143 profiles in the northern mid-latitudes during these two experiments, we find AIRS v5 CO retrievals are biased high by 6%–10% between 900 and 300 hPa with a root-mean-square error of 8%–12%. No significant differences were found between validation using spiral profiles coincident with AIRS overpasses and in-transit profiles under the satellite track but up to 13 h off in time. Similarly, no significant differences in validation results were found for ocean versus land, day versus night, or with respect to retrieved cloud top pressure or cloud fraction. Index Terms—Carbon monoxide (CO), infrared measurements, remote sensing, satellite validation.
I. I NTRODUCTION
T
HE ATMOSPHERIC Infrared Sounder (AIRS) was launched into a 705-km polar sun-synchronous orbit onboard NASA’s Aqua satellite on 4 May 2002 at the head of the A-Train satellite constellation. Since 31 August 2002, AIRS has acquired daily global views of the terrestrial troposphere [1]. AIRS primary design criterion was to retrieve day and night high-vertical-resolution temperature and moisture profiles in conjunction with the advanced microwave sounding unit
Manuscript received December 17, 2009; revised July 19, 2010; accepted December 5, 2010. W. W. McMillan, deceased, was with the Department of Physics, University of Maryland, Baltimore County, Baltimore, MD 21250-0001 USA. K. D. Evans is with the UMBC/JCET, University of Maryland, Baltimore County, Baltimore, MD 21250-0001 USA (e-mail:
[email protected]). C. D. Barnet is with the National Oceanic Atmospheric Administration (NOAA)/NESDIS/Center for Satellite Applications and Research (STAR), Camp Springs, MD 20746 USA. E. S. Maddy is with Dell, Inc., Fairfax, VA 22031 USA. G. W. Sachse is with the National Institute of Aerospace, Hampton, VA 23666 USA. G. S. Diskin is with the NASA Langley Research Center, Hampton, VA 23681 USA. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TGRS.2011.2106505
(AMSU) [2]. However, AIRS wide spectral coverage from 660 to 2655 cm−1 (15.38 to 3.75 um) and resolving power (λ/Δλ) of approximately 1200 includes emission features of several trace gases including carbon monoxide (CO) [1]–[3]. AIRS cross-track scanning swath width of 1650 km provides coverage of 70% of Earth every day [2]. Cloud clearing techniques combining the nine 1.1◦ AIRS fields-of-view within one 3.3◦ AMSU field-of-view yield successful AIRS+AMSU retrievals over approximately 50% of the planet for both day and night scenes up to 80% cloudy [4]–[6]. Hereafter, we term these AIRS retrievals as performed on the combined AIRS+AMSU field-of-regard (FOR), ∼45 km at nadir. CO is a direct product of incomplete combustion and a byproduct of the oxidation of methane and other volatile hydrocarbons (VOCs) [7], [8]. Roughly 50% of CO emissions are a result of biomass burning and naturally emitted VOCs [8], [9]. In concert with other trace gases including NOx , CO is a precursor to the formation of tropospheric ozone [8]. CO surface concentrations decreased in the 1990s largely as a result of improved industrial combustion efficiencies and vehicle emissions controls [10]–[12]. Natural and human-induced biomass burning, principally extensive wild and forest fires, comprise the largest sources of interannual variations in CO emissions [13]. Several studies have pointed to climate change as a possible driving force for increases in the size and severity of boreal forest fires resulting from changes in rainfall and drought [14], [15]. Monitoring changes in these natural and anthropogenic sources and subsequent long-range transport of their emissions are crucial to forecasting their impact on tropospheric chemistry and surface air quality. Nadir sounding instruments have been measuring CO from space for almost 30 years. For instance, CO was first measured from space by the Measurement of Air Pollution from Space (MAPS) instrument aboard the space shuttle in 1981 [16] and three subsequent flights [17], [18]. MAPS ushered in a new era in rigorous validation with an extensive correlative measurements campaign [19], [20] (and references therein). The launch of the Canadian Measurements Of Pollution in The Troposphere (MOPITT) instrument onboard NASA’s Terra satellite in 2000 provided the first routine global views of tropospheric CO on a weekly basis [21], [22] with subsequent extensive validation [23]–[25]. CO measurements also have been made by the Interferometric Monitor of Greenhouse gases (IMG) which flew onboard the Japanese ADEOS satellite [26], [27] and the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) onboard the European
0196-2892/$26.00 © 2011 IEEE
2
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
ENVISAT satellite [28], [29]. The Tropospheric Emission Sounder (TES) onboard NASA’s Aura satellite [30] has provided more detailed information on CO vertical profiles along nadir-only satellite tracks which prove difficult to validate [31]. Most recently, the Infrared Atmospheric Sounding Interferometer (IASI) onboard the European MetOp-A satellite now provides additional daily global views of CO from an operational weather platform [32]–[34]. Prior analyses of AIRS CO retrievals have demonstrated the utility of daily global views for tracking long-range transport of CO emissions from biomass burning [3], [35]–[39] and anthropogenic emissions [40]. The impact of this long-range CO transport on downwind air quality has been demonstrated by [41] and [42]. Previous validation and satellite intercomparison studies have found AIRS mid-tropospheric CO retrievals exhibit a high bias, particularly in the Southern Hemisphere where the smallest total column CO values are found. McMillan et al. [38] found out that AIRS version 4 (v4) CO retrievals were biased approximately 8% high at 500 hPa versus Aircraft in situ measurements during NASA’s Intercontinental Chemical Transport Experiment Phase A (INTEX-A) over North America. Yurganov et al. [13] and Warner et al. [43] found out that AIRS v4 CO retrievals were biased high versus MOPITT observations, less so over land in the Northern Hemisphere. Kopacz et al. [44] found out that AIRS v5 CO retrievals are biased high versus a global chemical transport model, MOPITT and SCIAMACHY, but low versus TES. Fisher et al. [45] found out that AIRS v5 CO retrievals are also biased high versus a chemical transport model and aircraft observations during the NASA Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS) experiment, but again demonstrated the ability of AIRS to track CO transport in the mid-troposphere. George et al. [33] found out that AIRS v5 CO retrievals are biased high versus IASI. In Section II of this paper, we discuss in detail the AIRS v5 CO retrieval algorithm and in Section III, we describe the optimization process employed on the AIRS v4 algorithm to produce the v5 algorithm. Validation of the AIRS v5 CO retrieval algorithm is presented in Section IV along with a discussion of the mechanics of validation using the AIRS v5 CO averaging kernels. Final conclusions are presented in Section V. II. AIRS V 5 CO R ETRIEVAL A LGORITHM Tropospheric CO abundances are retrieved from AIRS cloudcleared radiances in the R-branch side of the CO 1-0 fundamental vibration-rotation band between 2181.49 and 2221.12 cm−1 . Instrument design constraints resulted in AIRS discontinuous spectral coverage containing only this portion of the CO 1-0 band. AIRS resolving power yields an effective spectral resolution of approximately 1.8 cm−1 in the CO band, sufficient to separate the CO lines but insufficient to resolve the line shapes. Similar to the retrieval of all AIRS atmospheric products [4], [5], [46], a set of vertically overlapping trapezoidal perturbation functions are used to retrieve the CO profile. The following section details the differences between AIRS v5 CO retrievals and the v4 algorithm used in previous studies [3], [13], [38], [43]. These differences include the number of trapezoidal functions, the damping parameter, and the first-guess profile. In the present section, we summarize the formulation of the v5
AIRS CO retrieval algorithm and averaging kernels following from [5]. AIRS cloud clearing methodology is described in [4] and [47]. The overall AIRS v5 retrieval algorithm is described in [6]. A. Algorithm Description At each iteration of the AIRS CO retrieval, we solve for the ˆ [5, eq. (2)] coarse layer differences ΔA ˆ =U· ΔA
ϕ · U T S T M −1 [ΔΘ − δΘ] λ
(1)
where ϕ is the weighting matrix of the eigenvectors U given by [38, eq. (1)], [5, eq. (6)] λj ϕ = diag (2) λj + Δλj and where U [j × j] and λj are the eigenvectors and eigenvalues of the measurement sensitivity covariance [5, eq. (3)], respectively. S[n × j] is the derivative of the forward model for each channel index n with respect to each coarse layer j in ˆ and M [n × n] is the channel by channel noise covariance ΔA, matrix. ΔΘ[n × 1] is the observed minus computed brightness temperature for the n channels used in the CO retrieval. The 36 AIRS spectral channels employed in v5 of the CO retrieval range from 2181.49 to 2221.12 cm−1 [48]. δΘ[n × 1] is a background term [4]. Here, Δλj damps the least significant eigenvectors with values specified in [5, eq. (4)] √ λc · λj − λj , for λj ≤ λc (3) Δλ = 0 otherwise thus limiting the propagation of noise into the final solution. λc was determined empirically through comparisons of retrievals with in situ aircraft profiles as detailed in [49] and discussed in the next section. The coarse layers are described by overlapping trapezoids as depicted in Fig. 1(b) and discussed further in the following section. For comparison, the v4 trapezoids are shown in Fig. 1(a). Following from [5, eq. (1)] and [38, eq. (2)], we can write the CO averaging kernels for the AIRS retrieval algorithm as Φ = U · ϕ · UT .
(4)
Each row of the matrix Φ[j × j] corresponds to the averaging kernel for the respective trapezoid depicted in Fig. 1(b). As illustrated in Fig. 1(d), in the thermal infrared and with AIRS limited spectral resolution (Δv/v ∼ 1200), the typical AIRS CO averaging kernels are peaked broadly in the midtroposphere, exhibiting significant correlation between the different trapezoids or coarse layers. Thus, rarely is there more than one independent piece of information present in the AIRS CO retrievals. However, the location and vertical distribution of this information varies from scene to scene depending on the actual amount and vertical distribution of CO, the temperature structure of the atmosphere, and the temperature contrast between the surface and air. The degrees of freedom (DOF) in a particular retrieval is given by [5, eq. (7)] DOF = Trace(Φ)
(5)
MCMILLAN et al.: AIRS V5 CO RETRIEVAL WITH DACOM IN SITU MEASUREMENTS
3
Fig. 1. (a) At top left, the four nominal trapezoidal perturbation functions for the AIRS v4 CO retrieval. (b) At top right, the nine nominal trapezoidal perturbation functions for the AIRS v5 CO retrieval. (c) At bottom left, AIRS v4 CO averaging kernels for a relatively clear scene over the central U.S. on August 30, 2006. (d) At bottom right, AIRS v5 CO averaging kernels for the same scene as in (c).
and it represents the fractional number of significant eigenvectors used in the retrieval. This is not the same DOF for signal defined by [50] but similarly represents the information content of the AIRS CO retrieval. The CO averaging kernels are critical to use when one wants to compare another set of CO profiles to AIRS retrievals. The process of convolving in situ aircraft CO profiles with the AIRS CO averaging kernels is discussed in detail in the subsequent validation section. The AIRS CO verticality diagnostic function V , first defined in [38, eq. (3)], provides a quick assessment of where in the vertical a given AIRS CO retrieval is most sensitive. Here, we define the coarse layer verticality as the sum of the columns of the averaging kernel matrix V=I·Φ
(6)
where I [1 × 9] is a vector of ones. This is similar to the column operator of [51] but for the AIRS retrieval methodology. B. V5 CO Products Two distinct sets of files are produced during the AIRS v5 operational processing, standard and support product files. The standard product files contain subsets of the full retrieval
TABLE I AIRS v5 CO TRAPEZOID LAYERS (hPa/10)
products designed to meet a large and diverse community of users. As such, efforts have been made to minimize the total number and size of the variables in the standard product files. The support files are much more complete. They include all quantities necessary to compute spectra in a manner identical to the retrieval algorithm and contain a large number of diagnostic variables. The standard product files contain CO volume mixing ratios computed for the coarse layers defined by the faces of the nine trapezoidal perturbation functions along with the verticality function. The support files contain CO column densities for each of the 100 radiative transfer layers and the full averaging kernel matrix. CO_trapezoid_layers for v5 are listed in Table I. Its values define the faces of the nine trapezoids with an assumed tenth element nominally below the 100th layer of the AIRS fast radiative transfer model [52], [53]. The CO trapezoid layers are fixed vector for all retrievals in a granule of AIRS data. In practice, it is a fixed vector for all of the v5 retrievals. Each AIRS retrieval scene, field-of-regard (FOR), sees a different part of the Earth. As the topography varies, the actual surface
4
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
ing restrictions, in addition to the v5 Constituent Good: Qual_CO = 0 for DOFs ≥ 0.75 & skin temperature > 250 K, Qual_CO = 1 for 0.75 > DOFs ≥0.5 and skin temperature >250 K, and Qual_CO = 2 for DOFs < 0.5 or skin temperature ≤250 K. Although many of the CO_dof < 0.4 appear over cold surfaces or in very cloudy scenes, the additional surface skin temperature constraint rejects spurious AIRS CO retrievals year round over the Antarctic plateau. The same is true over central Greenland and the Arctic in wintertime, when the atmosphere is more isothermal and the thermal signals are inherently weak. In the validation analysis presented in the following results section, we further restricted our good matchups to retrievals where the AIRS retrieved cloud fraction was ≤ 80%. III. V5 O PTIMIZATION
Fig. 2. Estimated maximum root-mean-square (RMS) errors in retrieved CO for AIRS v5 retrievals due to errors in retrieved surface temperature, temperature profile, and water vapor profile.
pressure also varies. A given retrieval may not use all nine of the trapezoids depending on the surface pressure. The trapezoid closest to the surface for each retrieval is truncated at the surface pressure before averaging. In the v5 retrieval, the estimated retrieved errors in surface skin temperature, temperature profile, and water profile [54], [55] are smoothed with fits to the sensitivity test results to compute a CO retrieval error for each of the 100 layers. The smoothed fits are plotted in Fig. 2. In the V5 algorithm error estimation, we have neglected errors due to instrument noise, spectral calibration, and spectroscopy; however, the estimation is the square root of the sum of the squares of these errors. Thus, we believe this to be an upper limit to the error in AIRS retrieved CO. The most important CO variable in the support product files is Qual_CO, the quality control flag for AIRS CO retrievals. In AIRS v5 retrievals, Qual_CO has three possible values 0, Best Qual_CO =
1, 2,
Questionable Bad.
(7)
In detail, retrievals with Qual_CO = 0 (Best) are those where the internal Constituent Good flag is true and DOFs > 0.5. The Constituent Good flag is an internal indicator set to true if the total water vapor error estimate is less that 35% of the total retrieved water vapor [6], [56]. This requires that a final infrared retrieval was successful and implies that the total AIRS retrieved cloud fraction for the FOR is ≤ 90%. Qual_CO = 1 (Questionable) occurs for retrievals where Constituent Good = true and 0.5 > DOFs ≥ 0.4. Qual_CO = 2 (Bad) happens for retrievals where Constituent Good = false or DOFs ≤ 0.4. The additional constraints on Qual_CO based on DOFs were determined empirically based on examination of global data. DOFs > 0.5 means more than 50% of the retrieval is being determined by the measured AIRS radiances. Conversly, DOFs < 0.5 implies most of the information in the retrieval is coming from the first guess. Ongoing validation work suggests that Qual_CO should be tightened for the CO retrieval algorithm v6 with the follow-
The AIRS v4 CO retrieval algorithm [38] was optimized in 2006 [49]. These optimized changes to v4 were incorporated into v5 of the AIRS CO retrieval algorithm along with other changes to the overall AIRS science team retrieval algorithm. In this section, we summarize the differences between v4 and v5 of the AIRS CO retrieval algorithm as they pertain to the validation of v5. In the following section, we provide independent validation of the v5 AIRS CO retrieval algorithm using aircraft in situ profiles from NASA’s INTEX-A and INTEX-B field experiments. A. Trapezoids and Averaging Kernels As noted in [38], a major shortcoming of the v4 AIRS CO retrieval algorithm lies in the use of only four trapezoidal perturbation functions with a single thick trapezoid covering the lower half of the troposphere from 500 hPa to the surface. Thus, only the 400–500 hPa region of the v4 profiles were validated [38]. The four rather “blocky” averaging kernels are depicted in Fig. 1(c) and poor specificity of where the retrieval was finding information in the vertical, particularly, for the thick trapezoid closest to the surface. Fig. 1(b) presents the optimized set of nine trapezoidal perturbation functions utilized by AIRS v5 CO retrievals. There is no real increase in the information content between v4 and v5; however, the additional trapezoids better define where the retrieval algorithm finds information in the radiances as demonstrated by the smoother averaging kernels in Fig. 1(d). Limiting the trapezoids at the top and bottom to values of 0.5 forces the averaging kernels to roll off more steeply in these regions where there is less sensitivity. To facilitate intercomparison of CO retrievals between AIRS and the MOPITT instrument, the effective pressures of the AIRS v5 optimized trapezoids (see Fig. 1) were chosen to closely match the MOPITT v3 reporting levels [49]. The averaging kernels represent the retrieval’s response to a perturbation of the corresponding trapezoidal function. The gross similarity of the shapes of the nine averaging kernels for each retrieval indicates the low information content for the CO in the AIRS spectra. Typically, the DOF for AIRS v5 CO retrievals is near 0.8. The shapes of the v5 averaging kernels illustrate that sensitivity peaks in the mid-troposphere, near 500 hPa, and the retrievals are sensitive to a weighting of the total tropospheric CO column. However, from retrieval to retrieval, the shape of the averaging kernels can change depending on
MCMILLAN et al.: AIRS V5 CO RETRIEVAL WITH DACOM IN SITU MEASUREMENTS
5
Fig. 4. DC-8 ground-track locations for all INTEX-A (circles) and -B (squares) profiles that were used in this validation analysis. Open symbols denote in-transit profiles and filled symbols mark spiral profiles.
Fig. 3.
AIRS v4 and v5 first-guess profiles as described in the text.
the true distribution of CO in the atmosphere [42] and changes in the thermal structure. Sharp vertical gradients of both CO and temperature can enhance sensitivity to layers where such gradients occur. B. First-Guess Profile The AIRS v4 CO retrieval algorithm utilized a single global first-guess profile given by the AFGL 1976 standard atmosphere [58] for ease of use with the AIRS forward model. The optimization studies in [49] found that the MOPITT a priori CO profile [22] was a better choice for a single global first guess and further facilitated intercomparison with MOPITT v3 CO retrievals. For comparison, both the AIRS v4 and v5 firstguess CO profiles are shown in Fig. 3. In addition, we hoped the reduction in total column of the first guess would reduce the high bias (AIRS larger than in situ) found in the validation of v4 [38]. As discussed in more detail in the subsequent v5 validation section, v5 AIRS CO retrievals are still biased high. To resolve a computational issue with the AIRS forward model caused by very low CO mixing ratios at the top of the MOPITT v3 a priori profile, for pressures < 10.25 hPa, the MOPITT v3 a priori relaxes to the AFGL CO profile in the AIRS v5 first guess. The impact of this on the retrievals is insignificant as AIRS CO retrievals are insensitive to the small number of CO molecules in the upper atmosphere. C. Damping and Removal of Model Error Term From V4 The changes to the number of CO perturbation trapezoids and the first-guess CO profile necessitated a re-evaluation of the damping parameter as part of the optimization process [49]. The reduction of the damping factor weakened the stiffness of the CO retrieval algorithm, provided more flexibility to depart from the first-guess profile, and increased the dynamic range of the v5 CO retrievals. Previous studies have found that the v4 CO retrievals exhibit a smaller dynamic range than MOPITT [13], [43]. An additional reduction in the overall damping on the v5 AIRS Science Team retrieval algorithm was achieved by removing an ad hoc physics error term from the channel noise covariance matrix. However, the AIRS v5 CO retrieval algorithm still exhibits a smaller dynamic range in retrieved
column CO when compared to other satellite instruments [33]. This restricted dynamic range is due in part to AIRS lower spectral resolution and smaller DOF. IV. V5 VALIDATION For validation of AIRS v5 CO retrievals in this paper, we utilized in situ CO profiles acquired by the Differential Absorption CO Measurement (DACOM) instrument [59] flown onboard the NASA DC-8 during NASA’s INTEX-A in 2004 [60] and INTEX-B in 2006 [61]. We used final quality-assured 1-min averaged DACOM data from the NASA Langley Distributed Active Archive Center and INTEX archive. AIRS v5.0.14.0 standard and support product files were downloaded from the NASA Goddard Earth Sciences Data and Information Services Center. V5.0.14.0 retrievals are virtually identical to the current v.5.2.0 retrievals. Previous validation of the AIRS CO retrievals used DC-8 spiral profiles from INTEX-A [38]. During the spiral profiles, the DC-8 ascended or descended about a fixed point timed to coincide with a satellite overpass. For INTEX-A, separate profiles were flown for MOPITT and AIRS. During INTEX-B, most of the spiral profiles were flown for validation of instruments onboard NASA’s Aura satellite. Those timed for nadir observations of the TES are close in time to AIRS overpasses due to the approximately 15-min temporal spacing of the Aura and Aqua satellites in the A-train orbit at that time. Owing to the atmospheric chemistry focus of INTEX-A and -B, a larger number of in-transit aircraft profiles were performed in addition to the spiral profiles. As described below, we find no statistically significant differences in our CO validation results using intransit versus spiral profiles. A. Mechanics of Validation For each in situ profile, all good AIRS CO retrievals within a 200-km radius of the profile midpoint were selected for comparison. Here, a good CO retrieval is one where v5 Qual_CO = 0, the total retrieved cloud fraction was ≤ 80%, and Tskin > 250 K. For the INTEX data, there were no retrievals where Tskin < 260 K. For spiral profiles, the midpoint is located in the center of the spiral. For in-transit profiles, the midpoint is the latitude/longitude of the vertical midpoint of the profile. A 200-km radius was selected for consistency with MOPITT validation studies [23] because the ground-track
6
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
TABLE II INTEX SUMMARY VALIDATION STATISTICS FOR AIRS v5 DAYTIME RETRIEVALS VERSUS FIRST GUESS (FG), IN SITU DACOM (DA), AND CONVOLVED IN SITU DACOM (CV). THE EFFECTIVE PRESSURES FOR EACH OF THE VALIDATED TRAPEZOIDS ARE NOTED ALONG WITH THE NUMBER OF MATCHUPSFOR EACH TRAPEZOID (#)
profiles typically fall within this area, and to provide enough retrievals for a statistical measure of the variance. The number of retrievals (FORs) per in situ profile ranged from 3 to 56 with an average of 29.1. Each in situ profile with co-located AIRS retrievals is termed a matchup, and we present hereafter the mean results for the matchups. For each matchup, we convolve the one in situ profile with each of the individual retrieval averaging kernels within the 200-km radius and compute the average and standard deviation for the ensemble of AIRS retrievals for the matchup. To prepare an in situ profile for convolution, we first integrate the in situ profile to obtain CO column densities for each of the 100 AIRS layers. In our analysis, layer column densities for AIRS layers not completely covered by in situ measurements are filled with “Not a Number” valued to effectively remove them from the comparison. Here, we present results for trapezoids completely filled by in situ measurements. Convolution then proceeds via the following equation from [5]: X = X0 + FΦF (X − X0 )
(8)
where X [100 × 1] is the log of the in situ 100 layer CO column densities, X [100 × 1] is the log of the convolved in situ layer CO column densities, and X0 [100 × 1] is the log of the firstguess 100 layer CO column densities. Because of the exponential form of the absorption through Beer–Lambert’s Law, a linear change in radiance is related to a logarithmic change in trace gas column abundance in a layer. The AIRS CO retrieval algorithm approximates logarithmic derivatives using finite difference perturbations of the CO column density in each layer. Thus, the averaging kernel convolution must be performed with the logarithm of the layer column densities [5]. F [100 × 9] is the matrix of trapezoid functions as depicted in Fig. 1(b), and F [9 × 100] is the pseudo-inverse of F computed via F = [FT F]−1 FT
(9)
where T and −1 represent the matrix transpose and the inverse operations, respectively. Φ is the previously described averaging kernel matrix defined in (4). The following results depicting comparisons between convolved in situ CO profiles and AIRS retrieved CO profiles have been converted from layer column densities to mixing ratios. For the trapezoid layer mixing ratios, we follow the same prescription used in the AIRS standard product files and compute the mean CO mixing ratio across the face of each trapezoid. The effective pressure for each of these trapezoids is determined by (9). These coarse layer averages better illustrate the vertical sampling and smoothing of the v5 CO retrieval. For some matchups, examination of the 100 layer comparisons can be instructive as discussed below and in [38].
Fig. 5. (a) At top, map of INTEX-A DC-8 flight track (black and white dashed line) on 1 July 2004, noting the location of a spiral validation profile (large open black circles off the coast of California) in the center of co-located AIRS good CO retrievals (color filled circles). (b) At bottom, the CO profile comparison for the matchup profile in (a) for the nine trapezoid layer mixing ratios. All DACOM CO measurements are accurate to 2%.
B. Results The primary goals of NASA’s field experiments were to study the intercontinental transport of chemical products into and out of the U.S. INTEX-A focused on transport within the U.S. and toward Europe [60]. INTEX-B studied transport into the U.S. from the west (Asia) and south (Mexico) [61]. A secondary goal of both experiments was to provide crucial in situ validation measurements for several NASA satellite instruments including AIRS. Fig. 4 depicts the spatial coverage of the 143 INTEX-A and -B profiles utilized in this analysis. Due to differing vertical extents, every profile does not contribute to comparisons for every trapezoid. The 505- and 706-hPa effective pressure
MCMILLAN et al.: AIRS V5 CO RETRIEVAL WITH DACOM IN SITU MEASUREMENTS
7
FIg. 6. (a) At top, map of INTEX-B DC-8 flight track (black and white dashed line) on 22-23 March 2006, noting the location of an in-transit descent profile (large open black circles) in the center of co-located AIRS good CO retrievals (color filled circles). (b) At bottom, the CO profile comparison for the matchup profile in (a) for the nine trapezoid layer mixing ratios. All DACOM CO measurements are accurate to 2%.
Fig. 7. Summary of validation statistics on AIRS CO trapezoidal layers for all daytime INTEX matchups in terms of (a) bias, top, and (b) RMS, bottom. Symbols and lines for both plots are defined in the legend in (b).
trapezoids have the largest number of comparisons with 141 each, but not necessarily the same 141 profiles, and there were 128 and 129 profiles for the daytime, respectively. INTEX-A provided 15 spiral and 71 in-transit profiles. INTEX-B provided 18 spiral and 39 in-transit profiles. The number of matchups contributing to the daytime comparisons for each trapezoid is listed in Table II. 1) Individual Profiles: Fig. 5(a) shows the location of one INTEX-A spiral profile from 1 July 2004 off the coast of northern California and the AIRS v5 CO retrievals at 505 hPa. This case was presented for validation of AIRS v4 CO [38]. The black circles off the coast of northern California denote the good AIRS profiles used in the validation comparison for this matchup. Fig. 5(b) presents the comparison of this matchup of the mean CO over the trapezoid layers computed from the 100-layer support products, including the measured and convolved in situ profiles, the AIRS first-guess profile, and one standard deviation of the AIRS retrieved profile. These trapezoid layer CO values are essentially the same as those found in the corresponding AIRS standard product files and thus represent validation of the AIRS v5 CO Standard Product.
In Fig. 5(b), the divergence of the AIRS mean retrieval from the first guess between 900 and 100 hPa indicates the tropospheric region where AIRS radiances for these scenes provide information on CO abundances. For pressures > 900 hPa, the correspondence of the AIRS retrieval and the first guess does not necessarily indicate that the retrieval contains no information there. To ascertain the available information, it is necessary to examine the CO averaging kernels, or as demonstrated in Fig. 5(b), to examine the similarity of the convolved in situ to the first guess. In this case, for pressures between 900 and 950 hPa, the 100 layer convolved in situ coincides with the first guess; thus, the retrieval algorithm indicates AIRS radiances contributed no information in this tropospheric region. The convolved profiles only appear over the altitude ranges where in situ data was acquired in the AIRS 100 layers or filled a trapezoid. Between 900 and 300 hPa, the change in shape from the measured in situ to the convolved in situ reflects the broad vertical smoothing of the AIRS averaging kernels and the impact of the amount of first guess remaining in the final retrieval.
8
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
Fig. 8. Validation results for all (day and night) INTEX matchups for the four v5 validatable trapezoids: 351, 505, 706, 853 hPa. INTEX-A matchups appear as red diamonds, with INTEX-B matchups as blue squares. Open symbols denote in-transit profiles and filled symbols mark spirals. The 1 : 1 line is plotted in solid black and a 10% high AIRS bias is plotted as a dashed black line.
The trapezoid layer mixing ratio comparisons in Fig. 5(b) clearly demonstrate the trapezoids with sufficient aircraft measurements for validation. The four trapezoids in the middle, Pef f = 351, 505, 706, 853 hPa, show very similar CO mixing ratios to the 100 layer convolved mixing ratios (not shown). In Fig. 5(b), we note the convolved in situ lies on the low side of the mean AIRS retrieval and is within the one sigma variance of the retrievals for this matchup. All DACOM CO measurements are accurate to 2%. This matchup is typical of the majority of the 143 spiral and in-transit matchups used in this analysis as 80% of the profiles had either an upper or lower tropospheric enhanced CO layer. About 24% of the profiles show only an enhanced CO layer in the upper troposphere, 29% show only an enhanced CO layer in the lower troposphere, 27% show enhanced layers on both the upper and lower troposphere, and 20% of the profiles do not show any enhanced layers. No statistical difference is seen in the matchup differences due to these layers probably due to the 200-km averaging of the AIRS FORs. Fig. 6(a) notes the location of the in-transit profile acquired as the DC-8 descended on return to California following the first phase of INTEX-B. With a local time of 0354 h, this is one of only three INTEX nighttime DC-8 profiles. However, as discussed subsequently, a total of 13 in situ matchup profiles have nighttime AIRS data as the closest temporal comparator. Fig. 6(b) again demonstrates AIRS’ lack of sensitivity be-
Fig. 9. Validation results for 505-hPa trapezoid for the 13 cases where nighttime AIRS retrievals were the closest temporal matchup to in situ profiles. Symbol definitions are identical to Fig. 8.
low approximately 900 hPa. With in situ measurements from 200 hPa, this deeper aircraft profile actually provides validation for five trapezoids, Pef f = 253, 351, 505, 706, and 853 hPa. The bottom trapezoid, here with a mean effective pressure of 915 hPa, is nearly validated but lacks CO measurements all the way to the surface. Too few in situ profiles extend high enough
MCMILLAN et al.: AIRS V5 CO RETRIEVAL WITH DACOM IN SITU MEASUREMENTS
9
Fig. 10. Correlation plots of the matchup error for the 505 hPa with (a) the time difference between AIRS overpass and the in situ profile, (b) the mean DOF for each matchup, and (c) the mean cloud fraction for each matchup. (d) represents the correlation of the mean matchup DOF with the mean retrieved cloud fraction. Symbols are defined in Fig. 8.
to make for a meaningful validation of the 253 hPa trapezoid with this data set. For this case, the AIRS retrievals are sensing some of the enhanced upper tropospheric CO between 200 and 400 hPa, as evidenced by the mean retrieval and convolved DACOM trapezoid values in Fig. 6(b). However, the vertical smoothing inherent in AIRS’ retrievals due to the limited spectral resolution prevents AIRS from identifying the “C” shape of the in situ profile. 2) Overall Statistics: Fig. 7(a) and (b) summarize the validation results for all 143 INTEX matchups on the AIRS CO trapezoidal layers in terms of bias and root-mean-square (RMS) differences. Here, we define the first guess bias = (first guess trapezoid layers)–(in situ DACOM trapezoid layers), the in situ DACOM bias = (AIRS retrieved trapezoid layers)–(in situ DACOM trapezoid layers), and the convolved in situ DACOM bias = (AIRS retrieved trapezoid layers)–(convolved in situ DACOM trapezoid layers). Both the bias and RMS are greatly reduced from the first guess by the retrieval. Although it appears the unconvolved in situ yields a lower bias between 300 and 800 hPa, outside this region, the unconvolved bias is much larger, and the RMS is larger everywhere. The convolution appropriately converts the in situ to a profile as AIRS would have observed it by smoothing the profile and including the amount of first guess as dictated by the retrieval averaging kernels. The numerical values for the results in these plots are found in Table II along with the
number of matchups contributing to each trapezoid. Recall, due to differing vertical extents, every in situ profile does not contribute to comparisons for every trapezoid. Fig. 8(a)–(d) present the validation comparisons for each of the four v5 validatable trapezoids, 351, 505, 706, and 853 hPa, for every INTEX matchup. From these plots, one discerns no significant differences in the validation results between spiral and in-transit profiles or between INTEX-A and INTEX-B. The results for all four trapezoids cluster between 10% high AIRS dashed line and the 1 : 1 solid black line with the 853 hPa trapezoid showing the lowest overall bias. Fig. 9 excerpts the 505-hPa trapezoid results for the 13 matchups where AIRS nighttime spectra are the closest temporal matchup to the in situ profiles. Only three of these profiles were acquired at night. The other three trapezoids show similar results. The spread is slightly smaller for the nighttime cases, but with 1/10 as many matchups (13 night versus 128 day), there is little significance to this difference. The mean bias is nearly identical to the daytime cases. The plots in Fig. 10 examine the correlation, if any, between the matchup error (AIRS–convolved in situ) and several parameters of the matchups using the same symbols as in Fig. 8. With similar results for all four trapezoids, we present only the results for the 505-hPa trapezoid. The lack of correlation of the matchup biases with the time difference between AIRS overpass and the in situ profile further supports the use of
10
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
in-transit in situ profiles for validation of AIRS CO retrievals, as shown in Fig. 10(a). Fig. 10(b) reveals essentially no correlation of the matchup bias with the mean DOF. This is true because the DOF from AIRS is generally < 1 and the convolution procedure removes any first guess bias from the matchup. Fig. 10(c) exhibits a slight negative correlation between the matchup bias and the mean retrieved cloud fraction for each matchup. The offset of this trend at low cloud fraction represents the overall bias in AIRS v5 CO retrievals. The slope of the trend relates to the decrease in information content of the AIRS v5 CO retrieval with increasing cloudiness as demonstrated in Fig. 10(d). As a scene becomes more cloudy, the AIRS v5 CO retrieval sees less of the true total CO column and thus the DOF decrease. Nevertheless, the anti-correlation between cloudiness and both, DOFs and mean matchup bias is very small, 0.26. No correlations were seen between the matchup bias and the retrieved surface skin temperature, land versus ocean scenes, matchup latitude, mean satellite zenith angle, or retrieved cloud top pressure. V. C ONCLUSION This analysis provides a full description of the AIRS v5 CO retrieval algorithm and all associated CO standard and support products. We describe the optimization process used to produce the AIRS v5 CO retrieval algorithm from the previous v4 algorithm. Utilizing previously published AIRS v5 algorithm development, we completely describe the proper method for convolving in situ measurement profiles with the AIRS CO averaging kernel and first guess CO profile. Such convolved profiles are thus transformed into the AIRS v5 retrieval space for comparison to AIRS v5 CO retrievals for validation. The presented validation analysis utilizes a set of 143 in situ CO profiles acquired by the DACOM instrument flown onboard the NASA DC-8 during NASA’s INTEX-A and INTEX-B field experiments in 2004 and 2006, respectively. Overall, we find AIRS v5 CO retrievals are biased high by 6%–10% between 900 and 300 hPa with an RMS error of 8%–12%. No statistically significant differences were found between validation using spiral profiles flown to be temporally coincident with AIRS overpass and in-transit profiles flown to meet other aircraft research objectives up to 13 hours distant in time. Combined, INTEX-A and -B yielded only 33 spiral profiles flown for validation. The additional 110 in-transit profiles significantly improves the robustness of our results. Although representing only 10% of the case, the 13 nighttime matchups exhibit no statistically significant difference versus the daytime matchups. Similarly, no significant differences in validation results were found for ocean versus land scenes, with respect to latitude, mean satellite viewing angle, or retrieved surface skin temperature, cloud top pressure, or cloud fraction. Improvement of the bias of AIRS CO retrievals remains as a challenge for the next version, v6, of the AIRS retrieval algorithm. At present, consideration is being given to a more conventional optimal estimation CO retrieval algorithm for v6 [34]. ACKNOWLEDGMENT The authors would like to thank the AIRS Project Office at JPL and the entire AIRS Science Team for their stalwart support. Additional support for this analysis has come from the
NASA EOS, Global Carbon Cycle, Tropospheric Chemistry, and Atmospheric Chemistry Modeling and Analysis Programs as well as the NOAA Global Carbon Cycle Program. The authors would also like to thank all the scientists and the DC-8 crews involved in the INTEX field experiments. WWM and KDE thank the ARF group at UMBC for feedback. WWM thanks R. Force for her patient editing. This paper (and the research involved) was completed and submitted before Wallace passed away on 10 March 2010. R EFERENCES [1] M. T. Chahine, T. S. Pagano, H. H. Aumann, R. Atlas, C. Barnet, L. Chen, E. J. Fetzer, M. Goldberg, W. F. Irion, B. H. Lambrightsen, S. Y. Lee, J. Lemarshall, W. W. McMillan, E. T. Olse, L. L. Strow, and J. Susskind, “The Atmospheric Infrared Sounder (AIRS): Providing new insights into weather and climate for the 21st century,” Bull. Amer. Meteorol. Soc., vol. 87, pp. 911–926, 2006. [2] H. H. Aumann, M. T. Chahine, C. Gautier, M. D. Goldberg, E. Kalnay, L. M. McMillin, H. Revercomb, P. W. Rosenkranz, W. L. Smith, D. H. Staelin, L. L. Strow, and J. Susskind, “AIRS/AMSU/HSB on the aqua mission: Design, science objectives, data products, and processing systems,” IEEE Trans. Geosci. Remote Sens., vol. 41, no. 2, pp. 253–264, Feb. 2003. [3] W. W. McMillan, C. Barnet, L. Strow, M. Chahine, M. McCourt, P. Novelli, S. Korontzi, E. Maddy, and S. Datta, “Daily global maps of carbon monoxide: First views from NASA’s Atmospheric Infrared Sounder,” Geophys. Res. Lett., vol. 32, no. L11 801, 2005, DOI:10.1029/2004GL012821. [4] J. Susskind, C. D. Barnet, and J. M. Blaisdell, “Retrieval of atmospheric and surface parameters from AIRS/AMSU/HSB data in the presence of clouds,” IEEE Trans. Geosci. Remote Sens., vol. 41, no. 2, pp. 390–409, Feb. 2003. [5] E. S. Maddy and C. D. Barnet, “Vertical resolution estimates in version 5 of AIRS operational retrievals,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 8, pp. 2375–2384, Aug. 2008. [6] J. Susskind, J. M. Blaisdell, L. Iredell, and F. Keita, “Improved temperature sounding and quality control methodology using AIRS/AMSU data: The AIRS science team version 5 retrieval algorithm,” IEEE Trans. Geosci. Remote Sens., 2011, to be published. [7] P. J. Crutzen, L. E. Heidt, J. P. Krasnec, and W. H. Pollock, “Biomass burning as a source of atmospheric gases CO, H2 , N2 O, NO, CH3 Cl and COS,” Nature, vol. 282, pp. 253–256, Nov. 1979. [8] J. A. Logan, M. J. Prather, S. C. Wofsy, and M. B. McElroy, “Tropospheric chemistry: A global perspective,” J. Geophys. Res., vol. 86, pp. 7210– 7254, Aug. 1981. [9] A. M. Thompson, K. E. Pickering, R. R. Dickerson, W. G. Ellis, D. J. Jacob, J. R. Scala, W. Tao, D. P. McNamara, and J. Simpson, “Convective transport over the central United States and its role in regional CO and ozone budgets,” J. Geophys. Res., vol. 99, no. D9, pp. 18 703– 18 711, 1994. [10] P. C. Novelli, K. A. M. Asarie, P. M. Lang, B. A. Hall, and R. C. Myers, “Reanalysis of tropospheric CO trends: Effects of the 1997/1998 wildfires,” J. Geophys. Res., vol. 108, no. D15, pp. AC H14.1– AC H14.14, 2003, DOI:10.1029/2002JD003031. [11] D. Parrish, “Critical evaluation of US on-road vehicle emission inventories,” Atmos. Environ., vol. 40, no. 13, pp. 2288–2300, Apr. 2006. [12] R. C. Hudman, L. T. Murray, D. J. Jacob, S. Turquety, S. Wu, D. B. Millet, M. Avery, A. H. Goldstein, and J. Holloway, “North American influence on tropospheric ozone and the effects of recent emission reductions: Constraints from ICARTT observations,” J. Geophys. Res., vol. 114, no. D7, p. D07 302, 2009, DOI:10.1029/2008JD010126. [13] L. N. Yurganov, W. W. McMillan, A. Dzhola, E. Grechko, N. Jones, and G. Van der Werf, “Global AIRS and MOPITT CO measurements: Validation, comparison, and links to biomass burning variations and carbon cycle,” J. Geophys. Res., vol. 113, no. D9, p. D09 301, 2008, DOI:10.1029/2007JD009229. [14] L. Yurganov, T. Blumenstock, E. Grechko, F. Hase, E. Hyer, E. Kasischke, M. Koike, Y. Kondo, I. Kramer, F.-Y. Leung, E. Mahieu, J. Mellqvist, J. Notholt, P. Novelli, C. Rinsland, H. Scheel, A. Schultz, A. Strandberg, R. Sussmann, H. Tanimoto, V. Velazco, R. Zander, and Y. Zhao, “A quantitative assessment of the 1998 carbon monoxide emission anomaly in the northern hemisphere based on total column and surface concentration measurements,” J. Geophys. Res., vol. 109, no. D15, p. D15 305, 2004, DOI:10.1029/2004JD004559.
MCMILLAN et al.: AIRS V5 CO RETRIEVAL WITH DACOM IN SITU MEASUREMENTS
[15] E. S. Kasischke and M. R. Turetsky, “Recent changes in the fire regime across the north american boreal region—Spatial and temporal patterns of burning across Canada and Alaska,” Geophys. Res. Lett., vol. 33, p. L09 703, 2006, DOI:10.1029/2006GL025677. [16] H. G. Reichle, S. M. Beck, R. E. Haynes, W. D. Hesketh, J. A. Holland, W. D. Hypes, H. D. Orr, R. T. Sherrill, H. A. Wallio, J. C. Casas, M. S. Saylor, and B. B. Gormsen, “Carbon monoxide measurements in the troposphere,” Science, vol. 218, no. 4576, pp. 1024–1026, Dec. 1982. [17] H. G. Reichle, V. S. Connors, J. A. Holland, R. T. Sherrill, H. A. Wallio, J. C. Casas, E. P. Condon, B. B. Gormsen, and W. Seiler, “The distribution of middle tropospheric carbon monoxide during early October 1984,” J. Geophys. Res., vol. 95, no. D7, pp. 9845–9856, 1990. [18] V. S. Connors, B. B. Gormsen, S. Nolf, and H. G. Reichle, Jr., “Spaceborne observations of the global distribution of carbon monoxide in the middle troposphere during April and October 1994,” J. Geophys. Res., vol. 104, no. D17, pp. 21 455–21 470, 1999. [19] H. G. Reichle, V. S. Connors, and A. Thompson, “Preface to MAPS special section,” J. Geophys. Res., vol. 103, no. D15, pp. 19 283–19 284, 1998. [20] P. C. Novelli, K. A. Masarie, and P. M. Lang, “Distributions and recent trends of carbon monoxide in the lower troposphere,” J. Geophys. Res., vol. 103, pp. 19 015–19 033, 1998. [21] J. R. Drummond and G. S. Mand, “The Measurements of Pollution in the Troposphere (MOPITT) instrument: Overall performance and calibration requirements,” J. Atmos. Ocean. Technol., vol. 13, pp. 314–320, 1996. [22] M. N. Deeter, L. K. Emmons, G. L. Francis, D. P. Edwards, J. C. Gille, J. X. Warner, B. Khattatov, D. Ziskin, J. F. Lamarque, S. P. Ho, V. Yudin, J. L. Attie, D. Packman, J. Chen, D. Mao, and J. R. Drummond, “Operational carbon monoxide retrieval algorithm and selected results for the MOPITT instrument,” J. Geophys. Res., vol. 108, no. 4399, pp. 1–11, 2003, DOI:10.1029/2002JD003186. [23] L. K. Emmons, M. N. Deeter, J. C. Gille, D. P. Edwards, J. L. Attie, J. Warner, D. Ziskin, G. Francis, B. Khattatov, Y. Yudin, J. F. Lamarque, S. P. Ho, D. Mao, J. S. Chen, J. Drummond, P. Novelli, G. Sachse, M. T. Coffey, J. W. Hannigan, C. Gerbig, S. Kawakami, Y. Kondo, N. Takegawa, H. Schlager, J. Baehr, and H. Ziereis, “Validation of Measurements of Pollution in the Troposphere (MOPITT) CO retrievals with aircraft in situ profiles,” J. Geophys. Res., vol. 109, no. D3, p. D03 309, 2004, DOI:10.1029/2003JD004101. [24] L. Emmons, G. Pfister, D. Edwards, J. Gille, G. Sachse, D. Blake, S. Wofsy, C. Gerbig, D. Matross, and P. Nedelec, “Measurements of Pollution in the Troposphere (MOPITT) validation exercises during summer 2004 field campaigns over North America,” J. Geophys. Res., vol. 112, p. D12 S02, 2007, DOI:10.1029/2006JD007833. [25] L. K. Emmons, D. P. Edwards, M. N. Deeter, J. C. Gille, T. Campos, P. Nedelec, P. Novelli, and G. Sachse, “Measurements of Pollution in the Troposphere (MOPITT) validation through 2006,” Atmos. Chem. Phys., vol. 9, no. 5, pp. 1795–1803, 2009. [26] J. Wang, J. C. Gille, P. L. Bailey, L. Pan, D. Edwards, and J. R. Drummond, “Retrieval of tropospheric carbon monoxide profile from high resolution interferometer observations: A new Digital Gas Correlation (DGC) method and applications,” J. Atmos. Sci., vol. 56, no. 2, pp. 219–232, Jan. 1999. [27] B. Barret, S. Turquety, D. Hurtmans, C. Clerbaux, J. Hadji-Lazaro, I. Bey, M. Auvray, and P. F. Coheur, “Global carbon monoxide vertical distributions from spaceborne high-resolution FTIR nadir measurements,” Atmos. Chem. Phys., vol. 5, no. 4, pp. 2901–2914, 2005. [28] J. P. Burrows, E. Hlze, A. P. H. Goede, H. Visser, and W. Fricke, “SCIAMACHY—Scanning imaging absorption spectrometer for atmospheric chartography,” Acta Astronaut., vol. 35, no. 7, pp. 445–451, 1995. [29] M. Buchwitz, I. Khlystova, H. Bovensmann, and J. P. Burrows, “Three years of global carbon monoxide from SCIAMACHY: Comparison with MOPITT and first results related to the detection of enhanced CO over cities,” Atmos. Chem. Phys., vol. 7, no. 9, pp. 2399–2411, 2007. [30] R. Beer, “TES on the AURA mission: Scientific objectives, measurements, and analysis overview,” IEEE Trans. Geosci. Remote Sens., vol. 44, no. 5, pp. 1102–1105, May 2006. [31] M. Luo, C. Rinsland, B. Fisher, G. Sachse, J. Logan, H. Worden, S. Kulawik, G. Osterman, A. Eldering, R. Herman, and M. Shepherd, “TES carbon monoxide validation with DACOM aircraft measurements during INTEX-B 2006,” J. Geophys. Res., vol. 112, p. D24 S48, 2007, DOI:10.1029/2007JD008803. [32] S. Turquety, J. Hadji-Lazaro, C. Clerbaux, A. Hauglustaine, S.A. Clough, V. Cass, P. Shlssel, and G. Mgie, “Operational trace gas retrieval algorithm for the Infrared Atmospheric Sounding Interferometer,” J. Geophys. Res., vol. 109, p. D21 301, 2004, DOI:10.1029/2004JD004821. [33] M. George, C. Clerbaux, D. Hurtmans, S. Turquety, P. F. Coheur, M. Pommier, J. Hadji-Lazaro, D. Edwards, H. Worden, M. Luo,
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46] [47]
11
C. Rinsland, and W. McMillan, “Carbon monoxide distributions from the IASI/METOP mission: Evaluation with other spaceborne remote sensors,” Atmos. Chem. Phys., vol. 9, no. 2, pp. 9793–9822, 2009. E. S. Maddy, C. D. Barnet, and A. Gambacorta, “A computationally efficient retrieval algorithm for hyperspectral sounders incorporating a priori information,” IEEE Geosci. Remote Sens. Lett., vol. 6, no. 4, pp. 802–806, Oct. 2009. A. Stohl, T. Berg, A. Fjaeraa, C. Forster, A. Herber, C. Lunder, W. McMillan, S. Oltmans, S. Solberg, K. Stebel, J. Strom, K. Torseth, and K. Yttri, “Arctic smoke record air pollution levels in the European Arctic during a period of abnormal warmth, due to agricultural fires in Eastern Europe,” Atmos. Chem. Phys., vol. 7, no. 2, pp. 511–534, 2007. A. Stohl, C. Forster, H. Huntrieser, W. McMillan, A. Petzold, H. Schlager, and B. Weinzierl, “Aircraft measurements over Europe of an air pollution plume from Southeast Asia aerosol and chemical characterization,” Atmos. Chem. Phys., vol. 7, no. 3, pp. 913–937, 2007. A. Thompson, J. B. Stone, J. C. Witte, S. K. Miller, S. J. Oltmans, T. L. Kuscera, J. T. Merrill, G. Forbes, D. W. Tarasick, E. Joseph, F. J. Schmidlin, W. W. McMillan, J. Warner, E. J. Hintsa, and J. Johnson, “Intercontinental Chemical Transport Experiment Ozonesonde Network Study (IONS) 2004: 2. Tropospheric ozone budgets and variability over northeastern North America,” J. Geophys. Res., vol. 112, p. D12 S13, 2007, DOI:10.1029/2006JD007670. W. W. McMillan, J. X. Warner, M. M. Comer, E. Maddy, A. Chu, L. Sparling, E. W. Eloranta, R. M. Hoff, G. Sachse, C. Barnet, I. A. Razenkov, and W. Wolf, “AIRS views of transport from 12-22 July 2004 Alaskan/Canadian fires: Correlation of AIRS CO and MODIS AOD with forward trajectories and comparison of AIRS CO retrievals with DC8 in situ measurements during INTEX-A/ICARTT,” J. Geophys. Res., vol. 113, p. D20 301, 2008, DOI:10.1029/2007JD009711. H. Tanimoto, K. Sato, T. Butler, M. G. Lawrence, J. A. Fisher, M. Kopacz, R. M. Yantosca, Y. Kanaya, S. Kato, T. Okuda, S. Tanaka, and J. Zeng, “Exploring CO pollution episodes observed at Rishiri Island by chemical weather simulations and AIRS satellite measurements: Long-range transport of burning plumes and implications for emission inventories,” Tellus B, Chem. Phys. Meteorol., vol. 61, no. 2, pp. 394–407, 2009. L. Zhang, D. J. Jacob, K. F. Boersma, D. A. Jaffe, J. R. Olson, K. W. Bowman, J. R. Worden, A. M. Thompson, M. A. Avery, R. C. Cohne, J. E. Dibb, F. M. Flock, H. E. Fuelberg, L. G. Huey, W. W. McMillan, H. B. Singh, and A. J. Weinheimer, “Transpacific transport of ozone pollution and the effect of recent Asian emission increases on air quality in North America: An integrated analysis using satellite, aircraft, ozonesonde, and surface observations,” Atmos. Chem. Phys., vol. 8, pp. 6117–6136, 2008. G. Morris, S. Hersey, A. Thompson, S. Pawson, E. Nielsen, P. Colarco, W. McMillan, A. Stohl, S. Turquety, J. Warner, B. Johnson, T. Kuscera, D. Larko, S. Oltmans, and J. Witte, “Alaskan and Canadian forest fires exacerbate ozone pollution over Houston, Texas on 19 and 20 July 2004,” J. Geophys. Res., vol. 111, p. D24 S03, 2006, DOI:10.1029/2006JD007090. W. W. McMillan, R. Pierce, L. C. Sparling, G. Osterman, K. McCann, M. L. Fischer, B. Rappengluck, R. Newsom, D. Turner, C. Kittaka, K. Evans, S. Biraud, B. Lefer, and A. Andrews, S. Oltmans, “An observational and modeling strategy to investigate the impact of remote sources on local air quality: A Houston, Texas, case study from the Second Texas Air Quality Study (TexAQS II),” J. Geophys. Res., vol. 115, p. D01301, 2010, DOI:10.1029/2009JD011973. J. Warner, M. M. Comer, C. Barnet, W. W. McMillan, W. Wolf, E. Maddy, and G. Sachse, “A comparison of satellite tropospheric carbon monoxide measurements from AIRS and MOPITT during INTEX-A,” J. Geophys. Res., vol. 112, p. D12 S17, 2007, DOI:10.1029/2006JD007925. 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),” Atmos. Chem. Phys. Discuss., vol. 9, no. 5, pp. 19 967–20 018, 2009. J. A. Fisher, D. J. Jacob, M. T. Purdy, M. Kopacz, P. Le Sager, C. Carougel, C. D. Holmes, R. M. Yantosca, R. L. Batchelor, K. Strong, G. S. Diskin, H. E. Fuelberg, J. S. Holloway, E. J. Hyer, W. W. McMillan, J. Warner, D. G. Streets, Q. Zhang, Y. Wang, and S. Wu, “Source attribution and interannual variability of Arctic pollution in spring constrained by aircraft (ARCTAS, ARCPAC) and satellite (AIRS) observations of carbon monoxide,” Atmos. Chem. Phys. Discuss., vol. 9, pp. 19 035–19 080, 2009. E. T. Olsen, AIRS/AMSU/HSB Version 5 Level 2 Product Pressure Levels, Layers, and Trapezoids. Pasadena, CA: JPL, 2007, Ver. 1.0. M. T. Chahine, “Remote sensing of cloud parameters,” J. Atmos. Sci., vol. 39, pp. 159–170, 1982.
12
[48] E. T. Olsen, AIRS Version 5 Retrieval Channel Sets. Pasadena, CA: JPL, 2007, Ver. 1.0. [49] M. M. Comer, “Retrieving carbon monoxide abundances from the Atmospheric IntraRed Sounder (AIRS),” Ph.D. dissertation, Univ. Maryland, Baltimore County, MD, 2006. [50] C. D. Rodgers, “Characterization and error analysis of profiles retrieved from remote sounding measurements,” J. Geophys. Res., vol. 95, no. D5, pp. 5587–5595, 1990. [51] C. D. Rodgers and B. Connor, “Intercomparison of remote sounding instruments,” J. Geophys. Res., vol. 108, no. D3, p. 4116, 2003, DOI:10.1029/2002JD002299. [52] L. L. Strow, S. Hannon, S. De Sousa-Machado, and H. Motteler, “Validation of the AIRS radiative transfer algorithm,” in Proc. Opt. Remote Sens. Atmos., Tech. Dig. Ser., 2003, pp. 12–15. [53] E. T. Olsen, AIRS/AMSU/HSB Version 5 Level 2 Pressure Levels. Pasadena, CA: JPL, 2007, Ver. 1.0. [54] D. Tobin, H. E. Revercomb, R. O. Knuteson, B. M. Lesht, L. L. Strow, S. E. Hannon, W. F. Feltz, L. A. Moy, E. J. Fetzer, and T. S. Cress, “ARM site atmospheric state best estimates for AIRS temperature and water vapor retrieval validation,” J. Geophys. Res., vol. 111, p. D09 S14, 2006, DOI:10.1029/2005JD006103. [55] M. G. Divakarla, C. D. Barnet, M. D. Goldberg, L. M. McMillin, E. Maddy, W. Wolf, L. Zhou, and X. Liu, “Validation of AIRS temperature and water vapor retrievals with matched radiosonde measurements and forecasts,” J. Geophys. Res., vol. 111, p. D09 S15, 2006, DOI:10.1029/2005JD006116. [56] E. T. Olsen, AIRS Version 5 Release Level 2 Standard Product Quickstart. Pasadena, CA: JPL, 2007, Ver. 1.0. [57] E. T. Olsen, AIRS/AMSU/HSB Version 5 Level 2 Quality Control and Error Estimation. Pasadena, CA: JPL, 2007, Ver. 1.0. [58] G. Anderson, S. Clough, F. Kniezys, J. Chetwynd, and E. Shettle, “AFGL atmospheric constituent profiles (0–120 km),” AFGL (OPI), Hanscom AFB, Bedford, MA, Tech. Rep. AFGL-TR-86-0110, 1986. [59] G. W. Sachse, G. F. Hill, L. O. Wade, and M. G. Perry, “Fast-response, highprecision carbon monoxide sensor using a tunable diode laser absorption technique,” J. Geophys. Res., vol. 92, no. D2, pp. 2071–2081, 1987. [60] H. B. Singh, W. H. Brune, J. H. Crawford, and D. J. Jacob, “Overview of the summer 2004 Intercontinental Chemical Transport EXperimentNorth America INTEX-A,” J. Geophys. Res., vol. 111, p. D24 S01, 2006, DOI:10.1029/2006JD007905. [61] H. B. Singh, W. H. Brune, J. H. Crawford, F. Flocke, and D. J. Jacob, “Chemistry and transport of pollution over the Gulf of Mexico and the Pacific: Spring 2006 INTEX-B campaign overview and first results,” Atmos. Chem. Phys. Discuss., vol. 9, no. 7, pp. 2301–2318, 2009.
W. W. McMillan received the degree in physics Phi Beta Kappa, cum laude from Rhodes College, Memphis, TN, in 1985. He earned the M.A. and Ph.D. degrees in planetary atmospheres from The Johns Hopkins University, Baltimore, MD, in 1990 and 1992, respectively. After a National Research Council Post-Doctorate at NASA’s Goddard Space Flight Center, Greenbelt, MD, from 1992 to 1994, he moved to the Department of Physics at the University of Maryland, Baltimore County (UMBC), Baltimore, where he was an Associate Professor and Fellow in the Joint Center for Earth Systems Technology. He had worked with the AIRS project since coming to UMBC and was the lead scientist on the CO product from 2003–2010. He passed away on March 10, 2010.
Keith D. Evans received the B.S. degree in nuclear science with a physics option, from Virginia Polytechnic Institute and State University in Blacksburg, VA, in 1979. He earned the M.S. degree in physics from American University, Washington, DC, in 1984 and the M.S. degree in meteorology from the University of Maryland, College Park, in 1997. After working on solar energy and Department of Defense projects in the 1980s, he has spent the last 20 years studying trace gases in the atmosphere using lidar and satellite data. He is currently a Research Analyst in the Joint Center for Earth Systems Technology at the University of Maryland, Baltimore County, Baltimore.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
Christopher D. Barnet received the B.S. degree in electronics technology in 1976 and the M.S. degree in solid state physics in 1978 from Northern Illinois University, DeKalb, and the Ph.D. degree in remote sensing of planetary atmospheres using visible and infrared instruments aboard the Voyager spacecraft from New Mexico State University, Las Cruces, in 1990. His postdoctoral research at NASA and the Canadian Institute for Space and Terrestrial Science focused on ultraviolet, visible, and near-infrared observations of the planets using the Hubble Space Telescope. Since 1995, he has worked on advanced algorithms for terrestrial infrared and microwave remote sensing and has actively supported NASA’s Advanced Infrared Sounder (AIRS) science team. In June 2003, he was with the Office of Satellite Applications and Research (STAR) of NOAA/NESDIS, Camp Springs, MD, where he is exploiting operational sounder missions to provide measurements of ozone, carbon monoxide, carbon dioxide, and methane in the free troposphere. He is a member of the Infrared Atmospheric Sounding Interferometer (IASI) sounding team and the NPOESS Sounder Operational Algorithm Team (SOAT) and currently leads the calibration and validation team for the Cross-track Infrared Sounder and Advanced Technology Microwave Sounder (ATMS) Environmental Data Records (EDRs).
Eric S. Maddy received the B.S. degrees in physics and mathematics from Frostburg State University, Frostburg, MD, in 2001, the M.S. and Ph.D. degrees in atmospheric physics from the University of Maryland, Baltimore County (UMBC), Baltimore, on 2003 and 2007, respectively. He has been with QSS Group, Inc./Perot Systems Government Services (PSGS), Inc./Dell, Inc., Fairfax, VA, since 2004, focusing on the development and analysis of algorithms for deriving temperature, moisture and carbon trace gases from operational hyperspectral sounders.
Glen W. Sachse received the B.S. degree in physics from Virginia Polytechnic Institute and State University, Blacksburg, in 1970 and the M.S. degree in physics from the College of William and Mary, Williamsburg, VA, in 1977. He began developing airborne diode-laser-based gas sensors in the mid-1970s and subsequently participated in ∼30 major airborne science missions. He developed/investigated passive and active remote sensors and is the holder of related patents. He participated in nonintrusive measurements of the atmospheres within the sealed enclosures of the Charters of Freedom. He currently consults with NASA through the National Institute of Aerospace, Hampton, VA.
Glenn S. Diskin received the B.M.E. degree in mechanical and aerospace engineering from the Cooper Union, New York, NY, in 1984, the M.S. degree in mechanical and aerospace engineering from The George Washington University, Washington, DC, in 1986, and the Ph.D. degree in mechanical and aerospace engineering from Princeton University, Princeton, NJ, in 1997. He has developed and applied nonintrusive ground-based and airborne laser-based instrumentation for a variety of applications in the aerospace and atmospheric science fields. Since 1986, he has been with the NASA Langley Research Center, Hampton, VA.