Remotely operable compact instruments for measuring atmospheric CO 2 and CH 4 column densities at surface monitoring sites

Remotely operable compact instruments for measuring atmospheric CO2 and CH4 column densities at surface monitoring sites N. Kobayashi, G. Inoue, M. Kawasaki, H. Yoshioka, M. Minomura, I. Murata, T. Nagahama, Y. Matsumi, and T. Ibuki Research Institute for Humanity and Nature, Kyoto 603-8047, Japan Department of Molecular Engineering, Kyoto University, Kyoto 615-8510, Japan Department of Geophysics, Tohoku University, Sendai 980-8578, Japan Solar Terrestrial Environment Laboratory, Nagoya University, Nagoya 464-8601, Japan Received: 8 March 2010 – Accepted: 19 March 2010 – Published: 12 April 2010 Correspondence to: M. Kawasaki (kawasaki@moleng.kyoto-u.ac.jp) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
Carbon dioxide and methane are the most important anthropogenic greenhouse gases, with a contribution of 80% of radiative forcing to total anthropogenic greenhouse gases, leading to global warming (IPCC 2001).In order to determine Correspondence to: M. Kawasaki (kawasaki@moleng.kyoto-u.ac.jp) major source and sink regions on the earth, precise measurement of the global column density is an extremely pressing need.A greenhouse gas observing satellite (GOSAT: IBUKI) of Japan was launched on 23 January 2009.Data acquisition for the CO 2 and CH 4 column densities has progressed by an onboard Fourier transform spectrometer, FTS (Kuze et al., 2009).GOSAT observes an instantaneous field of view 10.5 km in diameter at every 160 km interval (Kuze et al., 2009;Yokota et al., 2009).For validation of the GOSAT data and also for covering the regions between the sparsely meshed observing points of GOSAT, it has been necessary to develop an apparatus with a high accuracy for measurement of the column density at surface monitoring sites.An FTS instrument with high resolution analyzes direct sunlight, offering enough uncertainty (Washenfelder et al., 2006;Ohyama et al., 2009;Wunch et al., 2010) for this purpose.Thus 16 ground-based FTIR spectrometers have been operated as Total Carbon Column Observing Network (TCCON, 2007).The data accumulated in this network were used for validation of SCIAMACHY (Dils et al., 2006) and GOSAT (Morino et al., 2010).However, FTS instruments have a high cost and are unsuitable for portable use.
In the present paper a desktop optical spectrum analyzer (OSA: Yokogawa Electric, AQ6370B-Custom) monitoring the near infrared (NIR) region was examined for applicability to measure CO 2 and CH 4 column densities in air.Spectral resolution of the OSA (0.16 cm −1 ) is higher than that of the FTS onboard GOSAT(∼0.26cm −1 ).Rotational lines in the regions of 1565-1585 and 1665-1685 nm were resolved for CO 2 and CH 4 measurements, respectively.The OSA instrument employed here costs less than one tenth of a FTS and is compact, portable, easy to use and rugged.
N. Kobayashi et  Ref.

OSA
Ref.As for the other method to measure atmospheric CO 2 column density, Wilson et al. (2007) have reported that a Fabry-Perot interferometer made of quartz glass has a high enough spectral resolution for resolving the rotational photoabsorption lines centred at 1575 nm (Wilson et al., 2007).A prototype instrument presented by them was composed of a solid etalon Fabry-Perot interferometer, off-axis parabolic mirrors, a beam splitter etc.The instrument is compact but seems to need precise optical alignment.In addition, the solid etalon has an appreciable heat capacity and hence temperature control of the solid etalon would be difficult within 1/100 deg accuracy for keeping the transmission wavelength through the etalon fixed (Wilson et al., 2007).FFPI, which has been developed for use in telecommunication industry, has the same optical features with the solid Fabry-Perot interferometer, and thus was tried to measure the CO 2 column density in the present work.The FFPI instrument assembled by simple fiber optics is essentially optical alignment-free, compact, easy to set up and strong against shock.Since the FFPI has a small heat capacity, transmission wavelength of the FFPI quickly responds to temperature change to reduce the time interval for measurement of the CO 2 column density.

Instrumental designs
A block diagram for data acquisition is shown in Fig. 1: two small telescopes for the FFPI and OSA are installed on a sun tracker with a GPS, the tracking resolution of which is stated to be 0.00225 • (Prede, ASTX-2, 20 kg).Geophysical data from the GPS and metrological data are accumulated in a laptop computer PC1 that controls also the temperature of the FFPI element.Photoabsorption intensity by the FFPI and solar intensity reference signal are detected by NIR detectors with a wide power range from +20 to −90 dBm (Yokogawa, AQ2200-211).OSA spectra and reference solar intensity signals are recorded by another laptop computer PC2.The  grating-operated OSA is commercially available and has an automatic self-alignment function for optics as well as wavelength calibration.

Collimation of sunlight
A small telescope was designed to collimate sunlight onto an optical fiber: The sunlight was prefiltered by a long-pass filter (HOYA, R100, λ ≥1000 nm) that was attached to the telescope body by sealing with an O-ring for preventing water leakage and then focused by a lens (50 mm in diameter and f =100 mm) on an optical fiber.The sunlight through the optical fiber was transmitted into the OSA or FFPI analyzer.
The small telescope was onboard the portable sun tracker.

Optical spectrum analyzer (OSA)
The optical spectrum analyzer disperses radiation in the wavelength regions of 600-1800 nm and is sensitive down to −90 dBm or 1 pW.The incidence slit width is designed to be equal to a core-diameter of an optical fiber.A multimode optical fiber (MMF) with 62.5 µm in core-diameter was found to transform a practically analyzable photon flux and to give an FWHM=0.16± 0.01 cm −1 at 1572 nm measured by using a tunable laser (ANDO, AQ4321D, line width of 0.2 MHz).Typical photoabsorption lines of the atmospheric CO 2 measured by the OSA are shown in Fig. 2a.The spectral scanning time was 34 s.The intensity of the sunlight in the scanning interval sustained fluctuation of air or shielding by thin cloud.For compensating the fluctuation of the sunlight intensity, 1000-1700 nm radiation was monitored throughout the measurement with an InGaAs detector as shown by a gray line in Fig. 2a.Optical bundle fibers were adopted for the monitoring the sunlight intensity.The light intensity through the central optical fiber was utilized for the spectral measurements while the intensity through the outer optical fibers was recorded as reference signal for normalization purpose.

Fiber Fabry-Perot interferometer (FFPI)
The instrument described here employs a single mode optical fiber (SMF) FFPI to measure the CO 2 column density through photoabsorption of sunlight.In this design, the sunlight passes through the atmosphere where it undergoes some photoabsorption by atmospheric CO 2 .The NIR radiation of 1570-1575 nm where strong CO 2 photoabsorption lines lie was isolated via a narrow bandpass filter, which was made of a quartz glass (Optical Coatings Japan) and installed at the incidence of the optical fiber.The isolated radiation was fed into a 2×2 optical coupler of SMF (Tatsuta Electric Wire&Cable) to split into two channels with transmitting intensity ratio of 10:1 where the first stronger channel is for the spectral analysis by the FFPI (Nippon Electric Glass, single mode optical fiber of 13 mm long×1.25 mm in diameter) (Sakamoto and Nishii, 2005) and the second weaker one is for reference signal.Output intensity of the FFPI reached several nW and was sufficient for measurements.The FFPI and the optical coupler were set in a small incubator kept at 25 ± 1 • C .
The FFPI was directly jacketed with low-expansion glass ceramics, and gave a free spectral range (FSR) shown in Fig. 3.The FFPI transmittance fringes were aligned with spacing of the CO 2 photoabsorption lines (Devi et al., 2007) so that photoabsorption due to CO 2 is primarily detected and gives the I 0 and I values in the Beer-Lambert law as will be discussed below.The sunlight intensity via the second channel strongly depended on changes of the NIR solar flux and was recorded as reference signal with 200 ms interval.This fast time response was required to follow the rapid solar flux intensity change that is probably caused by temporal airfluctuation and shielding by clouds.The signal ratios of the two channels were used to infer the atmospheric CO 2 abundance.
The FFPI has a temperature coefficient toward the transmittance wavelength as shown in Fig. 3b and thus the solar light wavelength passing through the FFPI is able to be on-and off-aligned with CO 2 photoabsorption lines by controlling the temperature with a Peltier device (Cell System, TDU-5000R, <150 VA).When the FFPI spectral feature overlaps the CO 2 rotational lines, the NIR detector gives the minimum intensity that effectively corresponds to the I value in the Beer-Lambert law of Eq. ( 1).At the minimum spectral overlap condition, the maximum intensity is recorded, which corresponds to the I 0 value: where σ t is the effective total photoabsorption cross section of CO 2 in the measuring wavelength region, N the concentration of CO 2 in a unit volume, and L the length from the earth's surface to the space.The product N×L is the CO 2 column density and determinable when the σ t is available for the optical system employed.

Optical spectrum analyzer (OSA)
Figure 4 shows parts of photoabsorption spectra of CO 2 observed by the OSA (solid curve) and the FTS onboard GOSAT (line with dots).The optical resolution of the former is better than that of the latter.One of the campaigns for the validation of the GOSAT measurements was carried out as a joint program of the National Institute for Environmental Studies of Japan (NIES) and the Japan Aerospace Exploration Agency (JAXA) on 26 August 2009 at the Moshiri observatory of Nagoya University in Hokkaido Japan (Latitude 44.3660; Longitude 142.2605; 290 m a.s.l.).The following instruments and an aircraft participated in the campaign: a ground-based Bruker IFS 120HR FTS (Nagoya University), an aircraft to collect flask samples for CO 2 , and CH 4 in the atmosphere (NIES), balloons to measure pressure, temperature, relative humidity and atmospheric CO 2 concentration (Nagoya University), and the OSA (Kyoto University).The spectral resolution of the Bruker IFS 120HR FTS was set at 0.02 cm −1 , that  is corresponding the TCCON observational protocol.The balloon measurements were performed with use of a small NIDR instrument that was on-board calibrated with standard gases (Nihon Sanso, 0.01 ppm accuracy).Photoabsorption spectra of CO 2 and CH 4 were measured by the FTS and the OSA under the same weather conditions.The spectra of the FTS were retrieved independently by two groups, Kyoto University and NIES.In Kyoto, the spectra of the FTS and OSA were retrieved by adopting a nonlinear least squares spectral fitting algorithm developed by us for the present work, details of which are given in the Appendix.In NIES, the FTS spectra were retrieved using a nonlinear least-squares spectral fitting algorithm (GFIT, Toon et al., 1992), which scales an a priori profile to produce a synthetic spectrum that achieves the minim residual between the synthetic and measured spectrum.GFIT is commonly used in TCCON and the retrieval conditions with GFIT are followed as those in TCCON.A flask sampling instrument was installed in the Beechcraft Kingair 200T aircraft and air over Moshiri was sampled.

CO 2 column density
Spectra shown in Fig. 2 were observed by the OSA and the FTS, and enlarged views of the spectra after the retrievals using our peak fitting algorithm are shown in Fig. 5.The atmospheric CO 2 column densities obtained by the OSA and the FTS measurements are given in Table 1 for dry air.Relative humidity, temperature and pressure in the air were measured over the Moshiri observatory up to 26 km and took account of in calculating the column densities.
A CO 2 balloon was launched at 13:30 until 14:26 (Nakayama, T. and Takekawa, S., personal communication, 2009).The CO 2 concentration measured was 374.4 ± 3.6 ppm on the surface level and 388.8 ± 2.3 ppm at 10 km.The column density deduced from the concentrations is given in Table 1 where the atmospheric CO 2 concentra-tion above 10 km was treated after the method in evaluation of CO 2 column average volume mixing ratio (VMR) over Tsukuba (Araki et al., 2010).The CO 2 concentration in the stratosphere above 20 km is considered to be five years older than that of the global mean concentration of CO 2 in the troposphere, which is 385.2 ppm in 2008 with a growth rate of 1.93 ppm/yr (WMO 2008).Thus the CO 2 concentrations in the troposphere and the stratosphere in 2009 are estimated to be 387.1 and 377.6 ppm, respectively.Between 10 km and 20 km in altitude the concentration is assumed to be linear (Araki et al., 2010).Atmospheric CO 2 column densities measured by GOSAT (V00.20) have been officially announced and are quoted in Table 1 for 6 August and 2 September 2009.The column densities of GOSAT were measured over Nayoro City in Hokkaido Japan located 20 km east to the Moshiri observatory.
Figure 6 shows time-series of the CO 2 column densities measured by the OSA, the FTS, the balloon and GOSAT in Table 1.These four independent measurements are consistent.A little large distribution of the column densities in the FTS (small open circles) probably arises from thin clouds since it was cloudy in the morning at the Moshiri distinct on 26 August 2009.The weather improved in the afternoon but was not cloud-free.The column density of CO 2 measured at the Park Falls Tall Tower site in Wisconsin USA has shown similar distribution of the CO 2 columns on a partly cloudy day (Washenfelder et al., 2006).Reference signal correction for the OSA spectra (large open circles) seems to be effective for a better distribution.
In the present retrievals the CO 2 concentration in the air is a fitting parameter and assumed to be constant (see the Appendix A).The values thus deduced correspond to the column average VMR of CO 2 in dry air, which is defined as the ratio of the deduced column density of CO 2 to total column of dry air (Washenfelder et al., 2006): The column average VMR for the balloon measurement is calculated to be 378.02± 5.16 ppm and consistents with those of the OSA and the FTS.The total column of dry air in Eq. ( 2) is replaced by the following relation (Washenfelder et al., 2006): where the dry-air mole fraction of O 2 is 0.2095 and highly constant in dry air.The Eq. (2) abbreviated to xCO 2 is then given by The FTS spectra from 09:50 to 16:01 were retrieved by NIES using GFIT (denoted by FTS-GFIT hereafter) and the xCO 2  values were deduced by Eq. ( 4).xCO 2 retrieved with GFIT is airmass corrected and aircraft calibrated using a factor of 0.990.The arithmetic mean of the xCO 2 for 275 retrievals in the FTS-GFIT was 377.05 ± 2.47 ppm as given in Table 1.Flask samples of the atmospheric CO 2 were collected by the aircraft from 530 to 7300 m over the Moshiri observatory and the CO 2 concentration was determined by NIES against the NIES 95 CO 2 scale (Machida et al., 2009).We obtain the column average VMR or xCO 2 to be 376.56 ± 0.02 ppm for the flask sampling if we assume that the CO 2 concentration of 375.85 ± 0.01 ppm at 7300 m continues up to 10 km.Above 10 km the same assumption as in the balloon discussion was applied.(Yokota et al., 2009).The reason to give the low xCO 2 is now under review.
Figure 7 shows the xCO 2 time-series of the OSA (large open circles) and the FTS (small open circles) retrievals by use of the present fitting algorithm as well as the FTS-GFIT (small solid circles).Most of the data concentrate at 377 ppm but some points of the FTS distribute a little wide probably due to shielding effect by thin cloud.The xCO 2 distribution of the OSA is good because of the reference signal correction.The column densities for the FTS-GFIT and the flask samplings in Table 1 were calculated by multiplying the column average VMR of CO 2 and total column of dry air (see Eq. 2) deduced in the analysis of the balloon data, where direct sampling by aircraft has been considered to be high-reliable.The column densities of the OSA, the balloon, the FTS-GFIT and the aircraft agree within ±0.2 % difference.The FTS analysis gives a 1.3% less value while the column average VMR is entirely consistent with that of the aircraft.The VMRs for the others are also in excellent agreement.These observations imply that the OSA has the performance enough to measure the CO 2 column density as well as FTS used for a ground-based standard.

CH 4 column density
Photoabsorption spectra of CH 4 measured by the OSA were independently retrieved by two groups, NIES and Kyoto University as in the CO 2 analysis described above.Two peaks at 1674.447 and 1677.601nm in Fig. 8 were used for the present retrieval algorithm in Kyoto.The xCH 4 in NIES using GFIT is a fitting parameter in the retrieval, being not aircraft calibrated.The column densities and column average VMRs deduced are shown in Fig. 9 and are stable all day long.The numerical values are given in Table 2 along with the GOSAT observation.
The xCH 4 for the aircraft was calculated by use of the estimation as in the balloon measurement, that is, the concentration of 1.850 ppm sampled at 7300 m was assumed to continue up to 10 km.The global mean concentration of CH 4 in 2008 is 1.797 ppm with a growth rate of 2.5 ppb (WMO 2008), giving 1.787 ppm for the stratospheric concentration above 20 km in 2009.The CH 4 concentration between 10 and 20 km linearly decreases from 1.850 to 1.787 ppm.Table 2 shows the data of the FTS-GFIT retrieval, aircraft and GOSAT(V00.30).
The column density and the xCH 4 of the OSA are less than those of the aircraft by 2.5% while the values for the FTS-GFIT and GOSAT are about 6% smaller.

Fiber Fabry-Perot Interferometer (FFPI)
A spectral profile of the sunlight passing through the narrow bandpass filter is shown in Fig. 10, where the sharp dips superimposed on the peak are assigned to the Rbranch lines in the 30012←00001 band of CO 2 (Devi et al., 2007).The spacing of the CO 2 photoabsorption lines between R8e and R22e is 0.324 ± 0.016 nm, which is close to the FRS = 0.317 ± 0.002 nm of the FFPI in Fig. 3.The peak width of the transmitted radiation through the FFPI is λ=0.072 ± 0.002 nm at FWHM, which is wider than the  pressure broadening of the R00e line of CO 2 , i.e., 0.050 nm at 1013 hPa (Nakamichi et al., 2006).The temperature coefficient of the FFPI was found to be 13.58 ± 0.16 pm/deg by monitoring wavelength shift of the transmitted light in Fig. 3.By modulating the temperature of the FFPI between 303 and 318 K (dotted curve in Fig. 11), the wavelength transmitted through the FFPI is shifted and hence the signal intensity via the FFPI periodically changes as shown by the solid curve in Fig. 11.In a 67 s cycle of the temperature modulation, the highest value gives the I 0 while the lowest one corresponds to the I in Eq. ( 1).Small dip in the peak-top results from overshooting the temperature to give a minimum overlap between the FSR and the spacing of CO 2 photoabsorption lines.The hump in the bottom, which is opposite, results from over- shooting the temperature to give a maximum overlap.The dip and hump ensure the maximum and the minimum signal intensities, respectively.
To obtain the effective total photoabsorption cross section, σ t , a calibration curve in Fig. 12 was measured by changing CO 2 pressure in a photoabsorption cell (l=174.1 cm) in the laboratory.The slope gives a proportional constant for the present optics, (2.696 ± 0.051) ×10 −23 cm 2 /molecule, which corresponds to the effective total photoabsorption cross section.By adopting this value as σ t , the CO 2 slant column is calculated by Eq. (1).Airmass correction approximated by the following formula should be performed in obtaining the CO 2 column density (Kasten and Young, 1989): where Z is the solar zenith angle.
The CO 2 column density thus obtained, however, is only apparent since the effective total photoabsorption cross section σ t was determined under artificial conditions in the laboratory.For accessing a true value the temporal column density should be normalized by CO 2 column density measured by a standardized instrument.We measured the atmospheric CO 2 column densities by use of the FFPI and the OSA under the same weather conditions at the Katsura campus of Kyoto University in Kyoto Japan (Longitude 34.983, Latitude 135.677, 160 m a.s.l.).The coefficients for the normalization in the present work were determined by averaging six days measurements.An example of the normalized FFPI is shown in Fig. 13.The OSA measurements shown by the solid circles give the column density of (7.93 ± 0.06)×10 21 molecules/cm 2 and the that revised by normalizing parameter for the FFPI shown by the open circles is (7.94 ± 0.01)×10 21 molecules/cm 2 on 31 October   2009.It is clear that the statistics of the distribution for the FFPI is much better than that for the OSA.Standardizing parameter in the revising procedure of the FFPI should be determined one time for the individual FFPI instrument because the effective σ t cross section depends strongly on the bandwidth and the transmission efficiency of the sunlight through the narrow pass filter employed.
The instrument composed of an FFPI optical device for obtaining the CO 2 column density is much less expensive and easier to operate than the OSA, though it needs to be calibrated and normalized one time by means of a standardized instrument.The fiber devices employed in the present work are essentially optical alignment-free, therefore rugged, and will perform their function at surface sites which may be unsuitable for FTS or OSA instruments.

Concluding remarks
Two instruments presented here demonstrate the capability of compact, easy-to-operate, portable, and remotely operable measurement of the atmospheric column density: an OSA with high enough resolution to resolve rotational lines of CO 2 and CH 4 , in comparison with a FTS onboard GOSAT.Using a least squares spectral fitting algorithm developed by us for the present work, the OSA gives atmospheric CO 2 and CH 4 column densities which agree with those obtained by a ground-based FTS with high resolution of 0.02 cm −1 and balloon measurements.The column average concentrations of CO 2 and CH 4 were deduced as a parameter in the present retrievals and they are in agreement with those determined

Fig. 8 .
Fig. 8. CH 4 photoabsorption spectrum for OSA retrieval.Black curve denotes the observed spectrum and red one calculated.The arrows are the photoabsorption lines of CH 4 . Fig.10

Fig. 10 .
Fig. 10.Spectral profile of the sunlight through a narrow bandpass filter.Assignments for the R-branch in the CO 2 (30012 ← 00001) transition are shown.

Fig. 11 .
Fig. 11.Signal intensity of the CO 2 photoabsorption (solid curve) measured by modulating the FFPI temperature in a 67 s cycle (dotted curve).The highest value gives the I 0 while the lowest one corresponds to the I value in Eq. (1).Small dip at the top and hump at the bottom result from overshooting the etalon temperature. Fig.12

Fig. 12 .
Fig. 12. Absorbance of CO 2 for determination of the effective total photoabsorption cross section measured with neat CO 2 at room temperature in the laboratory. Fig.13

Fig. 13 .
Fig. 13.CO 2 column density measured by FFPI on 31 October 2009 at the Katsura Campus of Kyoto University, Kyoto Japan.Open circles are the CO 2 column densities after normalization and the solid circles are those measured by OSA at the same time.
al.: Remotely operable compact instruments

Table 1 .
Column densities and column average concentrations of CO 2 at the Moshiri observatory measured by different instruments.
a Column average volume mixing ratio.b 26 August 2009 except GOSAT.c Optical spectrum analyzer (OSA), Yokogawa AQ6370B-Custom.d Fourier transform spectrometer, Bruker IFS 120HR.Retrieved by the algorithm in the present work (see the Appendix).e Retrieved using GFIT.Retrieved values are airmass corrected with a factor of −0.0075 corrected and aircraft calibrated with a factor of 0.990.f Calculated by multiplying the concentration and total column of dry air.g Flask sampling and analyzed by NIES.h Measured over Nayoro City in Hokkaido, Japan, located 20 km east to the Moshiri observatory.
The xCO 2 values in Table 1 agree well to each others.The xCO 2 for GOSAT is clearly low.It has been announced that the xCO 2 of GOSAT lies lower than expectation of an offline global atmospheric transport model developed by NIES by 10-15 ppm

Table 2 .
Column densities and column average concentrations of CH 4 at the Moshiri observatory.Retrieved values are not airmass corrected and not aircraft calibrated e Calculated with multiplying the concentration and total column of dry air.g Flask sampling and analyzed by NIES.h Measured over Nayoro City in Hokkaido, Japan, located 20 km east to the Moshiri observatory c Optical spectrum analyzer.d Time series of column density and column average VMRs of CH 4 .OSA: large open circles; FTS-GFIT: small solid circles; GOSAT: gray solid circles.On 26 August 2009.