Annual and seasonal variations of atmospheric ... - Wiley Online Library

GEOPHYSICAL RESEARCH LETTERS, VOL. 29, NO. 8, 1269, 10.1029/2001GL014082, 2002

Annual and seasonal variations of atmospheric CO2, N2O and CO concentrations retrieved from NOAA/TOVS satellite observations Alain Che´din,1 Anthony Hollingsworth,2 Noelle A. Scott,1 Soumia Serrar,1 Cyril Crevoisier,1 and Raymond Armante1 Received 17 September 2001; revised 27 November 2001; accepted 10 January 2002; published 30 April 2002.

[1] We show that atmospheric concentration variations of major greenhouse gases (CO2, N2O, CO) may be retrieved from observations of the National Oceanic and Atmospheric Administration (NOAA) polar meteorological satellite series. The method relies on the analysis of the differences between the observations and simulations from a radiative transfer model using collocated radiosonde data and fixed gas concentrations as the prime input. Over the time period considered (July 1987 – September 1991), the results are in good agreement with present knowledge of the atmospheric cycles (seasonal, annual) of the gases. A reanalysis of the more than 20 years archive of NOAA observations has considerable promise for an improved knowledge of the carbon cycle. INDEX TERMS: 1640 Global Change: Remote sensing; 1610 Global Change: Atmosphere (0315, 0325); 0330 Atmospheric Composition and Structure: Geochemical cycles

1. Introduction [2] Knowledge of present carbon sources and sinks, including their spatial distribution and their variability in time is essential for predicting future carbon dioxide atmospheric concentration levels. Because the atmosphere integrates over varying surface sources and sinks, the distribution of CO2 in the atmosphere and its time evolution can be used to quantify surface fluxes. However, this approach is currently limited by the sparse and uneven distribution of the global flask sampling programs. For example, regional carbon budgets are reconstructed from about 100 points. As a consequence, inferring surface sources and sinks from observed concentrations is still highly problematic [Rayner et al., 1999; Fan et al., 1998; Schimel et al., 2000; Braswell et al., 1997; Feely et al., 1999]. [3] Satellite measurements of the distribution of global atmospheric CO2 would in principle fill this gap in scale [Rayner and O’Brien, 2001]. Measurements that densely sample the atmosphere, in time and in space, would provide a crucial constraint, allowing uncertainty in transport versus other information (on source and sink characteristics) to be separated and reduced. In particular, two new high spectral resolution infrared atmospheric sounder instruments are currently being developed, the Atmospheric InfraRed Sounder (AIRS) to be launched on board of EOSAqua in 2002 and the Infrared Atmospheric Sounder Interferometer (IASI) to be launched on board of the Meteorological Operational (METOP) Satellite in 2005. In addition to their main mission of measuring atmospheric temperature and moisture global fields, both instruments are expected to provide, at least, vertically integrated atmospheric columns of several greenhouse gases amounts (CO2, N2O, CO, CH4, etc). Recent simulation studies 1 Laboratoire de Me´teórologie Dynamique, Institut Pierre-Simon Laplace, Ecole Polytechnique, Palaiseau, France. 2 ECMWF, Shinfield Park, Reading, UK.

Copyright 2002 by the American Geophysical Union. 0094-8276/02/2001GL014082

on IASI [Che´din et al., 2001a] and AIRS [Engelen et al., 2001] conclude to the feasibility of measuring CO2 within the accuracy recommended in [Rayner and O’Brien, 2001]. [4] Because the potential benefits are so large for global carbon cycle research, it is worthwhile to explore the possibility of retrieving useful information from the present generation of polar operational meteorological satellites. Even if the modest spectral resolution of these sounders is not a priori encouraging, the more than 20 years archive already accumulated merits consideration. [5] This was the purpose of the research reported here.

2. Data Set Description and Analysis [6] The Television and InfraRed Operational Satellite-Next generation (TIROS-N) Operational Vertical Sounder (TOVS) has been flown aboard the National Oceanic and Atmospheric Administration (NOAA) polar meteorological satellites since 1978 [Smith et al., 1979]. This instrument consists of the High resolution Infrared Radiation Sounder (HIRS-2), the Microwave Sounding Unit (MSU) and the Stratospheric Sounding Unit (SSU). Scan widths are approximately 2200 km, providing global coverage every 12 h. HIRS-2 measures atmospheric and/or surface emission in seven channels located around 15.0 mm, five located around 4.3 mm, and a 11.0 mm window channel. Ozone emission is measured in a 9.6 mm channel. Other channels measure water vapor emission and surface emission at shorter wavelengths. The MSU measures atmospheric emission in four 55-Ghz O2 channels. In the 15 mm and 4.3 mm spectral bands, measured radiances mostly depend on the temperature of the atmosphere but also on the CO2 concentration and the concentrations of the other greenhouse gases (N2O, CO, O3, in particular). [7] Over the period of time considered here (July 1987 to September 1991) the concentration of CO2 has increased by about 1.8% [CMDL, 2000] and that of N2O by about 1.3% [IPCC, 1995]. Such an increase in CO2 concentration corresponds to a variation of the observed radiance (hereafter expressed in terms of equivalent blackbody temperature) of any channel sensitive to CO2. For example, the response of channel 5 (located at 14 mm) is about 0.17 K and that of channel 15 (4.46 mm) of about 0.09 K. Similarly, the increase in N2O concentration corresponds to a signal of 0.09 K in channel 13 (4.57 mm), of about 0.13 K in channel 14 (4.52 mm), and of about 0.11 in channel 15 (4.46 mm). CO increase during the same period of time, about 1 % per year, is not detectable from HIRS CO-sensitive channels 13 and 14 (0.02K over the period for channel 13). These numbers were computed from the Automatized Atmospheric Absorption Atlas (4A) line-byline radiative transfer model [Scott and Che´din, 1981] in its latest version-2000 with up-to-date spectroscopy from the ‘Gestion et Etudes des Informations Spectroscopiques Atmospheriques (GEISA): Management and Study of Atmospheric Spectroscopic Information’ [Jacquinet-Husson et al., 1999], using a set of 42 representative atmospheric situations [Garand et al., 2001]. [8] Seasonally, CO2 and CO concentrations display important variations, particularly in the Northern Hemisphere [GLOBALVIEW-CO2, 1999; Novelli et al., 1998]. However, in the midtroposphere, to which most channels considered are principally

110 - 1

110 - 2

CHEDIN ET AL.: ANNUAL AND SEASONAL VARIATIONS OF CO2, N2O AND CO

sensitive, these variations are smaller and delayed by 1 – 2 months [Nakazawa et al., 1991; Holloway et al., 2000]. Approximately at the Northern mid-latitudes, the peak-to-peak amplitude for CO2 is 6 ppmv and CO varies from 80 ppbv to 120 ppbv (gross mean). N2O seasonal variations are less well documented and apparently small. [9] These signals, annual and seasonal, can clearly be revealed by a careful analysis of the differences between TOVS measurements and radiation model-simulated measurements using spacetime collocated radiosonde temperature and humidity measurements as the prime input. [10] This information is directly available from the so-called ‘DSD5’ NOAA/National Environmental Satellite Data and Information Service (NESDIS) collocation archive [Uddstrom and McMillin, 1994]. Within this archive, each collocation is described by its latitude, longitude, time, satellite observing conditions and TOVS channels brightness temperatures, radiosonde in situ measurements (temperature and moisture profiles), estimates of the surface temperature (from HIRS surface emission sensitive channels). Because of the sensitivity of some channels to surface temperature, the NOAA/NESDIS collocation archive estimated value is improved by minimizing the difference remaining between observations and radiative model simulations for the ‘‘window’’ (transparent) channel 8 of HIRS, located near 11 mm. Also, the DSD5 ozone information is complemented by the ozone climatology of [Li and Shine, 1995]. The space-time window of the collocations is 300 km – 3 h. All collocations considered here are clouds free. For more details, see [Uddstrom and McMillin, 1994]. [11] The radiative transfer model assuming constant mixing ratios for CO2 (355 ppmv), N2O (308 ppbv), and CO (100 ppbv), the differences between its simulations and the observations should reflect the differences between the reference concentration values used in the model and the actual values. Simulations are carried out following a special procedure (including the air-mass classification of each collocation processed) in order to avoid introducing spurious seasonal variations by the radiative transfer model itself [Che´din et al., 2002].

3. The Method [12] For the time period considered and for NOAA-10 observations, we have applied an inverse model to these differences with the aim of interpreting them in terms of annual and seasonal variations of the mixing ratio of CO2, N2O and CO (relative to the above reference mixing ratio values). [13] Our method is to express the difference between simulated and observed TOVS brightness temperature, TB, of a given channel, as a sum over the contributing gases of the products of the differences, q, between the radiative transfer model reference mixing ratio profile and the unknown true profile for a given gas, by the channel-gas Jacobian J, @TB/@q, computed from the 4A model. [14] For a collocation i and channel n, discretizing the atmospheric column within layers, the precise formula is: TB ¼ TBcalc ðv ; iÞ TBobs ðv; iÞ i X Xh true qref x J ðv ; layer; gas; iÞ: ¼ gas qgas ðiÞ gas layer

layer

ð1Þ

Jacobians come from the Thermodynamic Initial Guess Retrieval (TIGR) Climatological data set [Che´din et al., 1985; Chevallier et al., 1998]. This library of atmospheres consists of about 2300 situations selected by statistical methods out of 150,000 radiosonde reports. Clear sky transmittances, brightness temperatures and associated Jacobians for all TOVS channels are computed for each situation in TIGR by the 4A model for all satellite observing conditions and the results are also stored within the TIGR data set. For each collocation, the radiosonde profiles (temperature,

moisture) are compared to all the situations in TIGR and the closest is retained. In the above formula, the Jacobians are those of this closest TIGR situation. They approximate the true Jacobians to better than 5 – 10%. [15] For each collocation and each channel, equation (1) may be more simply written as: TB ¼

XX

q J :

ð2Þ

gas layer

[16] Assuming that both the reference and the true mixing ratio profiles are constant throughout the atmospheric column reduces the above expression to the sum over the contributing gases of the product q by the vertical integral of J: TB ¼

X gas

q

X

J:

ð3Þ

layer

It must however be pointed out that if this assumption holds for CO2 and N2O, it is more questionable for CO. This has to be kept in mind. [17] Channels are selected for their sensitivity to the gases considered and for their relative insensitivity to water vapor or to upper stratospheric temperature or to surface temperature and emissivity, rather poorly (or not) measured by radiosondes. Eight HIRS channels satisfy this criterion. Channels 2 to 5, around 15 mm, as well as channel 15, around 4.5 mm, are sensitive to CO2. Channels 2 and 3, the least sensitive, peak in the stratosphere. Channels 4, 5 and 15, the most sensitive, peak in the mid to high troposphere. Channels 13 and 14 are sensitive to CO and N2O and peak in the mid troposphere, as channel 15, also sensitive to N2O but not to CO. The influence of surface emissivity on the channels selected is small and may be neglected. This may be shown by examining the difference between the simulated and the observed brightness temperature for the transparent channel at 4.0 mm (channel 18): the 3-month running mean difference displays values between 0.1K and +0.1K. Signatures of no more than one third of these values may thus be expected for channel 13 (at 4.56 mm), the most sensitive of the channels selected to surface emissivity. More details on the sensitivities and shapes of the gas-Jacobians are given in [Che´din et al., 2002]. The ozone channel 9, centred at 9.6 mm, is also considered because of the non-negligible sensitivity of channel 5 to this absorber. Thus, for a set of N collocations (summed over 3 months for seasonal variations and over 12 months for annual trends), a total of Nx8 equations with 4 unknowns are solved to minimize the TB0s and calculate a ‘‘least-squares’’ estimate of the mixing ratios of CO2, N2O, CO and O3.

4. Results [18] Results presented here concern the latitude band 20N – 60N over land, which corresponds, by far, to the largest number of items in the statistics. Daytime collocations are rejected due to possible contamination of the observations by solar radiation. [19] The number, N, of collocations entering the minimization process, so far, a simple least squares estimate, is shown on Figure 1. The dashed line is for the 3-month (running) total and the solid line for the 12-month total. The latter is roughly constant around 1300, when the former displays a seasonal variation around a mean of about 300, illustrating the fact that fewer collocations are, inherently, available within the 20N – 60N latitude zone at night in summer than in winter. These numbers lead to a highly over-determined (and well-conditioned) system in both the seasonal and annual modes. The system of equations is solved month by month and a (number of items-weighted)

110 - 3

1600

160

1400

140

1200

CO concentration (ppbv)

Number of collocations


1000 800 600 400

100 80 60 40

200 0 8707

120

8807

8907

9007

20 8707

9107

8807

month

Figure 1. Number of collocations (almost simultaneous observations by the satellite NOAA-10 and by radiosondes) for the latitude band 20N – 60N, at night, over land, entering the gas mixing ratio retrieval process. Dashed line: 3-month monthly running mean; solid line: 12-month running mean. 3-month or 12-month monthly running filter is applied to these monthly results. [20] Figure 2 illustrates the results for the retrieval of the seasonal and annual variations of the CO2 mixing ratio. The annual variations (solid line) agree quite well with what is expected and even the evolution in time of the rate of increase, slower, faster, slower, seems in phase with the data from CMDL (see, for example, [Fung, 2000]). The overall trend is in good agreement with the 20N – 60N global trend from [GLOBALVIEW-CO2, 1999] displayed by the open circle line. The seasonal variations (dashed line) display a peak-to-peak amplitude in fairly good agreement with the expectation [GLOBALVIEW-CO2, 1999]; the spring peak appears principally in May with a 1 – 2 month delay with respect to surface data as measured, for example, at Cape Meares (45N, 124W) land site (open-squares line). Two ‘‘bumps’’, after July 1989 and July 1990 render the interpretation of the autumn peaks

8907 month

9007

9107

Figure 3. Seasonal (dashed) and annual (solid line) variations of the CO mixing ratio retrieved from NOAA-10, for the latitude band 20N – 60N, at night, over land. The reference mixing ratio value is 100 ppbv. more difficult. They might be due to the geographical locations of the collocations, but other factors may interfere: the number of items, for example. [21] Figure 3 shows the results for CO. The overall trend (small increase followed by a small decrease) is reasonable. The seasonal cycle is in remarkable agreement with the expectations, with peaks delayed by about one month, as for CO2, and a sharp decrease from maxima to minima and a slow increase from minima to maxima [Novelli et al., 1998]. The anomalous value of June 1989 is due to too small a number of items. [22] Figure 4 shows the results for N2O. The annual trend is almost constant throughout the period and larger than what is currently reported: about 1.4 ppbv/year instead of 1.0. After a detailed analysis, the explanation cannot be found in the sensitivity of channels 13 and 14 to the surface temperature. The geographical distribution of the collocations is another possible explanation, however quite speculative. The seasonal variations are very irregular and no clear cycle can be identified. This is, however, in agreement with our present knowledge of the N2O mixing ratio variations.

370

5. Discussion [23] On the whole, these results retrieved from an infrared sounder like HIRS, with its limited spectral resolution (more than

360

355

320

350

315

345

340 8707

8807

8907 month

9007

9107

Figure 2. Seasonal (dashed) and annual (solid line) variations of the CO2 mixing ratio retrieved from NOAA-10, for the latitude band 20N – 60N, at night, over land. These variations are relative to the mean value used by the radiative transfer model, here 355 ppmv. The open-circles line shows the annual variations of CO2 from GLOBALVIEW-CO2. The open-squares line shows the seasonal variations of CO2 as measured at Cape Meares land site (45.48N, 123.97W). As expected, HIRS retrieved cycle is delayed by 1 – 2 months with respect to surface measurements.

N2O concentration (ppbv)

CO2 concentration (ppmv)

365

310

305

300

295

290 8707

8807

8907 month

9007

9107

Figure 4. Same as Figure 3, for N2O. The reference mixing ratio is 308 ppbv.

110 - 4


one order of magnitude smaller than that of AIRS or IASI), and small number of channels (8 are used here), appear amazingly good. We may conclude that a dedicated reanalysis of the more than 20 years of the NOAA/TOVS archive offers considerable promise for an improved knowledge of the carbon and other biogeochemical cycles. [24] The main limit of the method developed here is its dependency upon collocations between satellite observations and radiosonde measurements because of their inability to measure all necessary information (ozone profiles; surface temperatures, particularly over land; upper stratospheric temperatures; etc.) and the uneven distribution of the radiosonding stations. Moreover, the requirement of a cloud free situation over each station at the time of the satellite overpass greatly reduces the number of acceptable items. An alternative and promising approach is now being developed based on the coupling between HIRS infrared channels, sensitive to both temperature and gas concentrations, and MSU microwave channels only sensitive to temperature. The method does not require collocations between satellite and radiosonde observations, and will use all clear fields of view available for each orbit, bringing a large number of items per month for the areas presently considered for the reconstruction of the atmospheric carbon dioxide distribution [Rayner and O’Brien, 2001]. For example, an area of 500 km 500 km, observed twice a day during one month, at a spatial resolution of 100 km 100 km (a compromise between HIRS and MSU spatial resolutions), leaves, on the mean, at least 500 clear items. [25] A comparison between the satellite retrieval annual variations curve for the CO2 mixing ratio and the equivalent curve from [GLOBALVIEW-CO2, 1999] shows negative differences between the beginning of the period and April – May 1989, and positive differences afterward (Figure 2). The mean difference is close to zero. It is difficult to say which is best between the retrieved values (averages over the collocations) and GLOBALVIEW-CO2 values (averages over the whole 20N – 60N latitude band). The mean absolute value of the differences is about 2 ppmv and may be taken as a rough estimate of the mean expected error when about 1300 items are processed (see Figure 1). Extrapolated to the satelliteonly approach, with about 500 items available (over an area of 500 km 500 km for one month), this would give a mean expected error of 2square root (1300/500) ffi 3 ppmv, close to 1% of the mean mixing ratio. This rough estimate includes significant (if not dominant) errors brought by the collocations. [26] These results also strengthen our hope to greatly improve our knowledge of the global distribution of a variety of radiatively active gases with the coming second generation vertical sounders like AIRS or IASI, both characterized by a much higher spectral resolution allowing a more sophisticated retrieval scheme than the one developed here for the present TOVS channels. [27] Acknowledgments. We are happy to thank the two anonymous referees for their fruitful comments and suggestions. We have also benefited from the large facilities of IDRIS, the computer center of CNRS.

References Braswell, B. H., D. S. Schimel, E. Linder, and B. Moore III, The response of global terrestrial ecosystems to interannual climate variability, Science, 278, 870 – 872, 1997. Che´din, A., N. A. Scott, C. Wahiche, and P. Moulinier, The improved initialisation inversion method: A high resolution physical method for temperature retrievals from satellites of the TIROS-N series, J. Climate Appl. Meteor., 24, 128 – 143, 1985. Che´din, A., A. Hollingsworth, N. A. Scott, R. Saunders, M. Matricardi,

C. Clerbaux, and J. Etcheto, IASI capability to monitor CO2, submitted to J. Geophys. Res., 2001a. Che´din, A., S. Serrar, R. Armante, N. A. Scott, and A. Hollingsworth, Signatures of annual and seasonal variations of CO2 and other greenhouse gases from NOAA/TOVS observations and model simulations, J. Climate, 15, 95 – 116, 2002. Chevallier, F., F. Cheruy, N. A. Scott, and A. Che´din, A neural network approach for a fast and accurate computation of a longwave radiative budget, J. Appl. Meteor., 37, 1385 – 1397, 1998. CMDL, See the NOAA Climate Modelling and Diagnostics Laboratory (CMDL) cooperative sampling network sites: www.cmdl.noaa.gov, 2000. Engelen, R. J., A. S. Denning, K. R. Gurney, and G. L. Stephens, Global observations of the carbon budget: I. Expected satellite capabilities for emission spectroscopy in the EOS and NPOESS eras, submitted to J. Geophys. Res., 2001. Fan, S., M. Gloor, J. Mahlman, S. Pacala, J. Sarmiento, T. Takahashi, and P. Tans, A large terrestrial carbon sink in North America implied by atmospheric and oceanic carbon dioxide data and models, Science, 282, 442 – 446, 1998. Feely, R. A., R. Wanninkhof, T. Takahashi, and P. Tans, Influence of El Nino on the equatorial Pacific contribution for atmospheric CO2 accumulation, Nature, 398, 597 – 601, 1999. Fung, I., Variable carbon sinks, Science, 290, 1313 – 1313, 2000. Garand, L., D. S. Turner, M. Larocque, J. Bates, S. Boukabara, P. Brunel, F. Chevallier, G. Deblonde, R. Engelen, M. Hollingshead, D. Jackson, G. Jedlovec, J. Joiner, T. Kleespies, D. S. McKague, L. McMillin, J.-L. Moncet, J. R. Pardo, P. J. Rayer, E. Salathe, R. Saunders, N. A. Scott, P. Van Delst, and H. Woolf, Radiance and Jacobian intercomparison of radiative transfer models applied to HIRS and AMSU channels, accepted in J. Geophys. Res., 2001. GLOBALVIEW-CO2, Cooperative Atmospheric Data Integration Project Carbon Dioxide. CD-ROM, NOAA/CMDL, Boulder, Colorado. [Also available on Internet via anonymous FTP to ftp.cmdl.noaa.gov, Path: ccg/ co2/GLOBALVIEW], 1999. Holloway, T., H. Levy II, and P. Kasibhatla, Global distribution of carbon monoxide, J. Geophys. Res., 105, 12,123 – 12,147, 2000. IPCC, Climate Change 1995, Contribution of working group I to the Second Assessment Report of the Intergovernmental Panel on Climate Change, edited by Cambridge University Press, 572 pp., 1996. Jacquinet-Husson, et al., The spectroscopic GEISA data bank, J. Quant. Spectrosc. Radiat. Transfer, 62, 205 – 254, 1999. Li, D., and K. P. Shine, A 4-Dimensional Ozone Climatology for UGAMP Models, UGAMP Internal Report No. 35, 1995. Nakazawa, T., K. Miyashita, S. Aoki, and M. Tanaka, Temporal and spatial variations of upper tropospheric and lower stratospheric carbon dioxide, Tellus, B43, 106 – 117, 1991. Novelli, P. C., K. A. Masarie, and P. M. Lang, Distributions and recent changes of carbon monoxide in the lower troposphere, J. Geophys. Res., 103, 19,015 – 19,033, 1998. Rayner, P. J., I. G. Enting, R. J. Francey, and R. Langenfels, Reconstructing the recent carbon cycle from atmospheric CO2, d13C and O2/N2 observations, Tellus, B51, 213 – 232, 1999. Rayner, P. J., and D. M. O’Brien, The utility of remotely sensed CO2 concentration data in surface source inversions, Geophys. Res. Lett., 28, 175 – 178, 2001. Smith, W. L., H. M. Woolf, C. M. Hayden, D. Q. Wark, and L. M. McMillin, The TIROS-N operational vertical sounder, Bull. Amer. Meteor. Soc., 60, 1177 – 1187, 1979. Schimel, D. S., J. Mellilo, H. Tian, and A. D. McGuire, Contribution of increasing CO2 and climate to carbon storage by ecosystems in the United States, Science, 287, 2004 – 2006, 2000. Scott, N. A., and A. Che´din, A fast line-by-line method for atmospheric absorption computations: the ‘‘4A’’ Automatized Atmospheric Absorption Atlas, J. Appl. Meteor., 20, 801 – 812, 1981. Uddstrom, M. J., and L. M. McMillin, System Noise in the NESDIS TOVS forward model, Part I: Specification, J. Appl. Meteor., 33, 919 – 938, 1994.

A. Che´din, N. A. Scott, S. Serrar, C. Crevoisier, and R. Armante, Laboratoire de Me´teórologie Dynamique, Ecole Polytechnique, 91128 Palaiseau, France. ([email protected]) A. Hollingsworth, ECMWF, Shinfield Park, Reading RG29AX, UK. ([email protected])