Global Biogeochemical Cycles

Interannual and seasonal variability in atmospheric N2O

Authors


Abstract

[1] The increase in atmospheric N2O observed over the last century reflects large-scale human perturbations to the global nitrogen cycle. High-precision measurements of atmospheric N2O over the last decade reveal subtle signals of interannual variability (IAV) superimposed upon the more prominent growth trend. Anthropogenic sources drive the underlying growth in N2O, but are probably too monotonic to explain most of the observed IAV. The causes of both seasonal and interannual variability in atmospheric N2O are explored on the basis of comparisons of a 1993–2004 atmospheric transport simulation to observations of N2O at five stations of the Advanced Global Atmospheric Gases Experiment (AGAGE). The complementary tracers chlorofluorocarbons (CFCs) 11 and 12 and SF6 also are examined. The model simulation does not include a stratospheric sink and thus isolates the effects of surface sources and tropospheric transport. Both model and observations yield correlations in seasonal and interannual variability among species, but only in a few cases are model and observed variability correlated to each other. The results suggest that tropospheric transport contributes substantially to observed variability, especially at Samoa station. However, some features of observed variability are not explained by the model simulation and appear more consistent with a stratospheric influence. At Mace Head, Ireland, N2O and CFC growth rate anomalies are weakly correlated to IAV in polar winter lower stratospheric temperature, a proxy for the strength of the mean meridional stratospheric circulation. Seasonal and interannual variability in the natural sources of N2O may also contribute to observed variability in atmospheric N2O.

1. Introduction

[2] Nitrous oxide (N2O) is a naturally occurring atmospheric greenhouse gas with a Global Warming Potential 300 times that of CO2 on a molecule per molecule basis [Prather et al., 2001]. The primary sink for N2O is photochemical destruction in the stratosphere, which releases reactive nitrogen that can catalyze ozone loss. Aside from N2, N2O is the only nitrogen gas that is sufficiently long-lived to become globally well mixed in the atmosphere. Since a small fraction of nitrogen involved in microbial N cycle transformations in both soils and oceans tends to leak off as N2O, the increase in atmospheric N2O from a preindustrial level of about 275 ppb to a 2005 value of 320 ppb (Figure 1a) signifies a large-scale perturbation to the global N cycle. When combined with the N2O atmospheric lifetime of ∼120 years and the total atmospheric burden of 1500 Tg N, the observed increase implies an anthropogenic source that now exceeds the natural microbial source by ∼50% and accounts for 1/3 of the total N2O source [Khalil et al., 2002; Hirsch et al., 2006].

Figure 1.

(a) Two-box model based on work by Sowers et al. [2002] assuming an anthropogenic N2O source equal to 2% of the anthropogenic N fluxes in Figure 1b. The model reproduces the general shape of the observed atmospheric increase, as given by ice core, firn, and atmospheric data [Machida et al., 1995; Battle et al., 1996; Thompson et al., 2004]. (b) Increase in anthropogenic N in synthetic fertilizer, livestock manure and fossil NOx from 1860 to the present [Holland et al., 2005b].

[3] The global N2O concentration was first accurately determined in the late 1970s, at which time its global atmospheric increase was discovered and quantified [Weiss, 1981]. The growth in atmospheric N2O is attributed mainly to an increase in the amount and rate of N cycled globally in association with human agriculture. Agricultural N cycle perturbations include both industrial fixation of N2 to make synthetic fertilizer [Vitousek et al., 1997; Galloway et al., 2004], and the acceleration of nitrogen cycling through cultivation, land-use change, and the expanding global population of livestock (Figure 1b). A smaller amount of nitrogen is anthropogenically fixed as NOx by fossil fuel combustion and subsequently deposited on the biosphere [Holland et al., 2005a]. The assumption that ∼2% of anthropogenic N leaks off as N2O has been shown to reproduce the general shape of the increase in atmospheric N2O over the last 150 years [Nevison et al., 1996; McElroy and Wang, 2005] (also see Figure 1a). However, only a few studies have examined the finer-scale seasonal variability in atmospheric N2O [Bouwman and Taylor, 1996; Levin et al., 2002; Liao et al., 2004; Nevison et al., 2004, 2005] and still fewer have discussed N2O’s interannual variability [Schauffler and Daniel, 1994; Wong et al., 1999; Ishijima et al., 2001].

[4] Previous studies of seasonal variability in atmospheric N2O have noted that the seasonal N2O minimum observed at Northern Hemisphere monitoring stations is out of phase with the predicted source minimum and may be caused in part by seasonal differences in the transmission of N2O-depleted air from the stratosphere [Bouwman and Taylor, 1996; Levin et al., 2002; Nevison et al., 2004; Liao et al., 2004]. The similarity in the minima of N2O and complementary halocarbon tracers (discussed below) provide support for this hypothesis. However, these studies could not rule out an additional or even dominant role for tropospheric transport variability for a variety of reasons. First, early modeling studies have found that all long-lived tracers tend to display “similar behavior” at sites remote from their sources and sinks [Plumb and McConalogue, 1988]. Second, trapping of polluted air in the thinner wintertime boundary layer is known to produce seasonality in Northern Hemisphere surface observations, regardless of changes in sources [Barrie and Hoff, 1984; Elkins et al., 1993]. Third, the seasonal amplitude ratios of N2O to other long-lived tracers are not quantitatively consistent with the inverse of their atmospheric lifetimes [Nevison et al., 2004], as one might predict for a purely stratospheric signal [Plumb and Ko, 1992].

[5] Much has been learned about the carbon cycle by analyzing seasonal and interannual variability in atmospheric CO2 [Prentice et al., 2001; Baker et al., 2006]. This paper aims to promote a similar discussion of atmospheric N2O, which is often considered the nitrogen cycle counterpart to CO2. The analysis presented here is based on model results and observations at 5 long-term monitoring stations. We begin by examining the potential of interannual changes in the anthropogenic N2O source to explain the observed atmospheric interannual variability (IAV). We next examine how transport acts upon N2O to create variability in the troposphere. For this analysis, we present the results of an atmospheric transport model simulation with realistic surface sources of N2O and IAV in atmospheric transport. We include several complementary tracers, discussed below, in the simulation. While our primary motivation is to understand IAV, we revisit the analysis of seasonal variability presented by Nevison et al. [2004], since influences capable of creating seasonal cycles are likely to be relevant for IAV. Finally, we evaluate the influence of IAV in the cross tropopause exchange of stratospherically depleted air on tropospheric N2O and complementary tracers by comparing observed growth rates to IAV in lower stratospheric temperature, which is used as a proxy for variability in the mean stratospheric circulation.

[6] We note that this paper does not attempt to evaluate IAV in natural N2O sources (with the exception of a brief discussion of the relationship between ENSO and the N2O ocean source in the tropical Pacific). Such an evaluation is beyond the scope of the paper because there are still substantial uncertainties in the mean annual natural N2O budget [Khalil et al., 2002]. Atmospheric N2O data potentially can provide top-down constraints to help quantify natural microbial ocean and soil sources and discern the regional imprint of anthropogenic sources. However, variability in tropospheric N2O can only be exploited fully to constrain and identify sources with an improved understanding of the influence of atmospheric transport and the natural stratospheric photochemical sink [Nevison et al., 2005; Hirsch et al., 2006].

2. Methods

2.1. Complementary Tracers (CFC-11, CFC-12, and SF6)

[7] Chlorofluorocarbons (CFCs) 11 and 12 are used in this paper as complementary tracers to aid in the interpretation of variability in atmospheric N2O. Like N2O, the CFCs are long-lived, well-mixed species in the troposphere and are destroyed by photochemistry in the stratosphere. Thus CFCs can be used to examine both the influence of the stratospheric sink and of tropospheric transport variability. Unlike N2O, the CFCs are entirely man-made gases with relatively well-known surface sources that largely ceased in 1996. CFC-11, the shorter-lived of the two CFCs, has been declining in the troposphere since the early 1990s, while the longer-lived CFC-12 only began to level off in the early 2000s [Montzka et al., 1999; Prinn et al., 2000]. Despite these trends, substantial surface sources of CFCs may remain, either from illegal production, production in developing countries allowed under the Montreal Protocol, or the leakage of old refrigerators, air conditioners and closed-cell foams [Hurst et al., 2004].

[8] An additional complementary tracer considered in this study is sulfur hexafluoride (SF6). SF6 is a man-made compound that is used and released from electric power transmission equipment and other industrial applications. It has risen rapidly in the atmosphere from <0.5 ppt in 1978 to over 5 ppt by 2001 [Geller et al., 1997; Sidorov et al., 2002]. Unlike N2O and the CFCs, which have stratospheric photochemical sinks, SF6 is destroyed beginning only in the mesosphere and hence has a very long atmospheric lifetime of about 3200 years. SF6 also differs from N2O and the CFCs in its very rapid atmospheric growth rate and large interhemispheric gradient.

2.2. AGAGE Measurements

[9] Measurements from the Advanced Global Atmospheric Gases Experiment (AGAGE) provide the longest available continuous record of the very high precision data necessary to detect interannual variability signals in atmospheric N2O and CFCs [Prinn et al., 2000]. The AGAGE network consists of 5 monitoring stations at Mace Head, Ireland (53°N, 10°W), Trinidad Head, California (41°N, 123°W), Ragged Point, Barbados (18°N, 65°W), Cape Matatula, American Samoa (14°S, 171°W) and Cape Grim, Tasmania (41°S, 145°E). AGAGE measurements, which succeeded the earlier ALE and GAGE programs, began in 1993 at Cape Grim, in 1994 at Mace Head, 1995 at Trinidad Head, and 1996 at Barbados and Samoa. Samples are collected every 40 min using in situ gas chromatography. N2O and CFC-12 are measured on the same column and electron capture detector (ECD), while CFC-11 is measured on a different column and ECD. Monthly mean values are estimated on the basis of order 103 measurements with pollution events removed. The relative precision of the individual measurements is about 0.03% (0.1 ppb) for N2O and slightly less precise for the CFCs. SF6 measurements by ECD are available at only one AGAGE station, Cape Grim, beginning in 2001. Although every effort is made to minimize the introduction of systematic errors in the AGAGE calibration procedure [Prinn et al., 2000], it is important to recognize that the small changes interpreted here remain vulnerable to such errors, which are extremely difficult to quantify objectively.

[10] N2O, CFC and SF6 measurements also have been made by the NOAA HATS and CCGG groups at a wide variety of stations [Thompson et al., 2004; Hirsch et al., 2006]. The NOAA data are consulted but not analyzed in detail in the current study.

2.3. Surface Fluxes and the MATCH Tracer Transport Model

[11] Surface fluxes of N2O and the complementary tracers CFC-11, CFC-12, and SF6 were run from 1993 through 2004 in the Model of Atmospheric Transport and Chemistry (MATCH) [Rasch et al., 1997; Mahowald et al., 1997]. The model used T62 horizontal resolution (about 1.9° latitude by longitude) with 28 vertical levels and was run with a time step of 20 min using archived 6 hourly winds for the years 1993–2004 from the National Center for Environmental Prediction (NCEP) reanalyses. Stratospheric destruction was not included in the simulations, since past experience has suggested that models have difficulty capturing the processes by which stratospheric signals in N2O and the CFCs are transported to the troposphere [Nevison et al., 2004; Hirsch et al., 2006]. To avoid this uncertainty, the simulations were designed purely to isolate the influences of surface sources and transport on tropospheric variability in N2O, the CFCs and SF6.

[12] The CFC-11 and CFC-12 fluxes used to drive MATCH were obtained from the Global Emissions Inventory Activity (GEIA) database www.geiacenter.org), which is based on a 1986 spatial distribution [McCulloch et al., 1994]. Emissions for the years 1993–2000 were estimated as recommended on the GEIA website by scaling the 1986 emissions by the global emissions in the appropriate year [McCulloch et al., 2001, 2002]. This approach assumes that, while global emissions change relatively rapidly, distribution is affected only by relative economic activity and population dynamics, which have slower rates of change with time. Nevertheless, the application of the 1986 distribution to years beyond 1990 involves significantly increasing uncertainty. Furthermore, GEIA provides global emissions totals only through 2000, while our simulation ran through 2004. For lack of better information, we applied the 1986 distributions and 2000 absolute values to the years 2001–2004. We also lacked seasonal information on the CFC fluxes and thus assumed that the fluxes decreased linearly over the calendar year according to the decreasing global emission totals. SF6 sources were prescribed in a similar manner, using the 1995 spatial distribution of the EDGAR-95 database [Olivier and Berdowski, 2001] and scaling the emissions each year to match the global total growth rate of SF6 in the atmosphere inferred from atmospheric measurements [Peters et al., 2004].

[13] The N2O surface flux was estimated as the total of soil, ocean and anthropogenic components. Monthly mean soil emissions totaling 6.2 Tg N2O-N/yr were obtained from the CASA terrestrial biogeochemistry model [Potter et al., 1996]. Monthly mean ocean fluxes totaling 3 Tg N2O-N/yr were obtained from the ocean biogeochemistry model of Jin and Gruber [2003]. Anthropogenic emissions were estimated by assuming a 5 Tg N2O-N/yr source distributed uniformly over the year according to the spatial pattern of nitrogen fertilizer consumption. In contrast to the CFC and SF6 fluxes, the N2O surface fluxes were assumed to be cyclostationary; that is, the same monthly mean cycle of seasonal fluxes was repeated each year over the 1993–2004 simulation. The cyclostationary approach was adopted owing to the lack of detailed information about recent temporal changes in N2O sources. Such changes have been relatively small over the 1990s compared to the CFCs, whose sources decreased by 60% or more over the decade.

2.4. Numerical Analysis Methods

[14] Seasonal and interannual variability was calculated by fitting each species in the AGAGE or MATCH monthly mean time series to a polynomial + harmonic function.

equation image

The optimal fit was determined by recursive least squares regression, in which F(t) in the case of AGAGE data was weighted by the inverse of the σ2 associated with each monthly mean. The harmonic component (last 4 terms) of equation (1) was assumed to represent the average seasonal cycle of each time series. To estimate IAV, all terms of equation (1) were subtracted to obtain detrended, deseasonalized concentration anomalies. The latter were smoothed with a 12-month running mean, followed by 3 successive 3-month wide triangular smoothings to remove subannual variability. Growth rate anomalies were estimated as the central difference slope of the smoothed curve. For some applications, the seasonal cycles and growth anomalies were normalized by the global mean tropospheric mixing ratios of N2O, CFC-11, CFC-12 and SF6 to facilitate comparison of the 4 species.

[15] For both seasonal and interannual variability, the correlations between N2O, CFC-11, CFC-12 and SF6 were quantified on the basis of shape and phasing, using the R correlation statistic, and on amplitude of variability, using the ratio of standard deviations for each time series. Similar statistics were calculated for correlations between the AGAGE (observed) and MATCH (modeled) time series for each individual species. These comparisons are shown in Figures 45678 and the statistics are compiled in Tables 123.

Table 1. Correlations in Mean Seasonal Cycle Among Species for MATCH and AGAGEa
 F11 Versus F12
MATCH SeasonalAGAGE Seasonal
Rσf11:σf12Rσf11:σf12
  • a

    R coefficients <0.5 are not reported.

cgo0.901.110.631.80
smo0.980.820.881.04
rpb0.980.890.691.58
thd0.970.800.741.47
mhd0.981.240.791.44
 
 N2O Versus F12
MATCH SeasonalAGAGE Seasonal
Rσn2o:σf12Rσn2o:σf12
cgo-0.390.700.82
smo0.890.500.890.65
rpb0.650.15-0.59
thd-0.15-0.55
mhd0.740.130.640.71
 
 N2O Versus F11
MATCH SeasonalAGAGE Seasonal
Rσn2o:σf11Rσn2o:σf11
cgo-0.360.630.46
smo0.920.610.900.63
rpb0.700.170.630.37
thd-10.19-0.37
mhd0.770.110.830.49
 
 SF6 Versus F12
MATCH SeasonalAGAGE Seasonal
Rσsf6:σf12Rσsf6:σf12
cgo0.842.51-2.93
smo0.952.88--
rpb0.943.25--
thd0.865.19--
mhd0.931.34--
Table 2. Correlations in Mean Seasonal Cycle and Interannual Variability Between MATCH and AGAGE for Individual Speciesa
 Mean Seasonal CycleInterannual Variability
RσMATCH:σAGAGERσMATCH:σAGAGE
  • a

    R coefficients <0.5 are not reported.

CGO
 F11-0.44-0.32
 F12-0.77-0.48
 N2O0.670.38-0.19
 SF60.630.81--
SMO
 F110.980.82-0.37
 F120.980.890.590.71
 N2O0.970.960.620.40
RPB
 F11-1.20-0.36
 F12-3.41-0.50
 N2O-0.87-0.13
THD
 F110.971.23-0.57
 F120.902.71-0.70
 N2O-1.02-0.26
MHD
 F11-1.04-3.72
 F12-1.02-4.98
 N2O0.730.61-0.68
Table 3. Correlations in Interannual Variability Among Species for MATCH and AGAGEa
 F11 Versus F12
MATCH IAVAGAGE IAV
Rσf11:σf12Rσf11:σf12
  • a

    R coefficients <0.5 are not reported.

cgo0.861.51-2.34
smo0.980.730.781.31
rpb0.950.920.901.30
thd0.920.930.861.15
mhd0.971.280.681.71
 
 N2O Versus F12
MATCH IAVAGAGE IAV
Rσn2o:σf12Rσn2o:σf12
cgo-0.250.720.66
smo0.830.470.860.79
rpb-0.16-0.60
thd0.690.210.880.58
mhd0.780.120.880.91
 
 N2O Versus F11
MATCH IAVAGAGE IAV
Rσn2o:σf11Rσn2o:σf11
cgo-0.170.720.28
smo0.840.650.790.60
rpb0.600.170.720.46
thd0.630.230.830.50
mhd0.850.100.700.53
 
 SF6 Versus F12
MATCH IAVAGAGE IAV
Rσsf6:σf12Rσsf6:σf12
cgo0.721.95--
smo0.862.54--
rpb0.563.25--
thd0.798.62--
mhd0.951.59--

3. Results and Discussion

3.1. Atmospheric Growth Rate and the Anthropogenic Source

[16] A time series for atmospheric N2O is plotted using monitoring network data spanning from the late 1970s to the present (Figure 2a). As discussed above, the driving force behind the 0.7ppb/yr (∼3.4 Tg N/yr) increase observed over this period is known in general but not in detail. The time derivative of the N2O curve (Figure 2b) reveals IAV in the atmospheric growth rate, some of which may be an artifact of the low signal-to-noise ratio in N2O data and the difficulty of the early instruments in capturing subtle seasonal and interannual signals. However, the AGAGE data shown in Figure 2 have achieved a precision of 0.03% and provide a credible record of IAV in atmospheric N2O over the last decade. The juxtaposition of the anthropogenic N2O source (estimated as 2% of fertilizer + manure) against the observed atmospheric growth rate (Figure 2b) indicates that interannual changes in the anthropogenic source are unlikely to explain the IAV in the atmospheric data, unless the assumed 2% anthropogenic N2O emission coefficient varies by 30% or more year to year. Such a large interannual variation cannot be ruled out, given the uncertainty in the emission coefficient. However, it is not supported by available studies using biogeochemical models of the anthropogenic and/or natural soil emissions of N2O. These models, which are driven by observed temperature and precipitation, yield IAV of only ±0.2 to 0.4 Tg N2O-N/yr [Potter and Klooster, 1998; Prinn et al., 1999], substantially less than the observed N2O growth rate variability of ±1.0 to 1.5 Tg N/yr (Figure 2b). The analysis presented above suggests some parallels between N2O and carbon dioxide, in which the primary anthropogenic source of CO2, fossil fuel combustion, is relatively steady and monotonic with time and cannot explain the large IAV observed in the atmospheric CO2 growth rate [Prentice et al., 2001]. In the next sections, we explore alternative explanations for the variability observed in atmospheric N2O, including transport and stratospheric influences.

Figure 2.

(a) Increase in atmospheric N2O from ALE/GAGE/AGAGE measurements at Cape Grim, Tasmania. (b) Time derivative of atmospheric N2O increase juxtaposed against the assumed anthropogenic N2O source of 2% of agricultural nitrogen, assuming a conversion factor of 4.8 Tg N per ppb N2O [Kroeze et al., 1999].

3.2. Comparison of MATCH Simulation and AGAGE Data

3.2.1. Latitudinal Gradients

[17] The north-south latitudinal gradients in N2O, CFCs and SF6, when combined with transport changes, are potential drivers of seasonal and interannual variability, especially in the Southern Hemisphere. As an important first evaluation of the MATCH transport model simulation, the modeled and observed gradients between Trinidad Head (THD) and Cape Grim (CGO) are compared in Figure 3. The comparison shows that MATCH captures the N2O gradient fairly accurately but tends to overestimate the THD-CGO gradient for the CFCs by an average of ∼50% between 1996 and 2004. Similar plots (not shown) of the gradient between Mace Head (MHD) or Barbados (RPB) and CGO show an even larger overestimate of the MHD-CGO CFC gradient (factor of 2), but a relatively small (∼10 to 20%) overestimate for RPB-CGO. The reasons for the overestimate of the CFC gradients may include the uncertainty in the global emission totals used in the MATCH simulation. MATCH is also known to produce a relatively strong “seasonal rectifier effect” in CO2 simulations compared to other tracer transport models [Gurney et al., 2003], which suggests that the model may overestimate boundary layer effects. In addition, hemispheric differences in the return of N2O and CFC-depleted air from the stratosphere, i.e., greater exchange in the north, may affect the gradient, but are not included in the MATCH simulation [Holton et al., 1995]. Finally, AGAGE data are routinely screened for pollution events, but there is no easy way to do a comparable filtering of MATCH results. Pollution events have a particularly strong influence at Mace Head, which sits near large industrial sources of CFCs and large agricultural sources of N2O in Europe [Biraud et al., 2000].

Figure 3.

Gradient between Trinidad Head (41°N) and Cape Grim (41°S) calculated using monthly mean time series from 1995–2005. The gradient is expressed as a percentage of the average tropospheric mixing ratio of the two stations. Black line is MATCH results, and gray line is AGAGE data. (a) N2O, (b) CFC-11, and (c) CFC-12.

[18] The SF6 gradient is not shown in Figure 3 owing to lack of AGAGE data in the Northern Hemisphere. Between 1995 and 2004, MATCH estimates a THD-CGO SF6 gradient of ∼0.35–0.4 ppt, a value comparable to that presented by Denning et al. [1999]. Thus the MATCH SF6 gradient supports an accurate calculation of interhemispheric transport in the model, despite the overestimate of the CFC gradients described above. Expressed as a percentage (7–8%) of the absolute tropospheric mixing ratio (∼5 ppt), the SF6 gradient is at least 4 times larger than the CFC gradients and ∼16 times larger than the N2O gradient.

3.2.2. Seasonal Cycles

[19] MATCH model seasonal cycles are strongly correlated among the 3 man-made tracers (CFC-11, CFC-12 and SF6) at the 5 stations presented in Figure 4 and Table 1. The amplitude of the cycles declines between the northern midlatitude stations and Cape Grim. The amplitude of the CFC-11 cycle (normalized by mean tropospheric mixing ratio) is comparable or slightly smaller than that of CFC-12 at all stations except Mace Head, while the amplitude of the SF6 cycle is consistently at least 2.5 times larger than that of CFC-12. The seasonal cycle of MATCH N2O is strongly correlated to the man-made tracers at Samoa (SMO), more weakly correlated at Mace Head and Barbados and uncorrelated at Trinidad and Cape Grim.

Figure 4.

Mean seasonal cycles of N2O (solid black line), CFC-11 (dotted black line), and CFC-12 (dash-dotted gray line) normalized by mean tropospheric mixing ratio. (left) MATCH results. (right) AGAGE observations. (a, f) Cape Grim, (b, g) Samoa, (c, h) Barbados, (d, i) Trinidad Head, and (e, j) Mace Head.

[20] The observed AGAGE seasonal cycles tend to be correlated between the CFCs, although more weakly than in the MATCH simulation (Figure 4 and Table 1). In contrast to MATCH, the normalized AGAGE CFC-11 amplitude is consistently larger than the normalized CFC-12 amplitude, except at Samoa, where the two are about the same. The AGAGE N2O seasonal cycle is uncorrelated to the CFC cycles at Barbados and Trinidad Head, strongly correlated at Samoa and more weakly correlated at Mace Head and Cape Grim. The decline in amplitude of the N2O and CFCs cycles between the northern midlatitudes and Cape Grim is weaker than in the MATCH simulation.

[21] AGAGE SF6 data (measured only at Cape Grim) show that the minimum in the observed SF6 seasonal cycle leads the observed N2O and CFC minima by 2–3 months (Figure 5a), whereas the MATCH simulation predicts coincident January minima for SF6 and the CFCs (Figure 5b). The AGAGE results are consistent with NOAA N2O and SF6 data at Cape Grim and other mid- to high-latitude Southern Hemisphere stations (http://www.cmdl.noaa.gov/ccgg/iadv/). The MATCH and AGAGE SF6 cycles are similar in amplitude and phasing, although the MATCH minimum leads the AGAGE minimum by 1 month (Figure 5c). In contrast, the MATCH and AGAGE cycles are farther out of phase for the CFCs (Figure 6 and Table 2).

Figure 5.

(a) Mean seasonal cycles of N2O (solid black line), CFC-11 (dotted black line), CFC-12 (gray dash-dotted line), and SF6 (dashed line) normalized by mean tropospheric mixing ratio for AGAGE observations at Cape Grim. (b) Same as Figure 5a but for MATCH results. (c) Mean SF6 seasonal cycle at Cape Grim for MATCH results (black line) and AGAGE data (gray dash-dotted line).

Figure 6.

Mean seasonal cycles for MATCH results (black line) and AGAGE data (gray dash-dotted line). (top) N2O, (middle) CFC-11, and (bottom) CFC-12. (a, f, k) Cape Grim. (b, g, l) Samoa. (c, h, m) Barbados, (d, i, n) Trinidad Head. (e, j, o) Mace Head.

[22] The comparison of SF6 and CFC results at Cape Grim suggests that two different influences may be acting in the extratropical Southern Hemisphere: (1) A transport influence, driven by the north-south tracer gradients [Denning et al., 1999], which tends to yield a January or February minimum. This influence is captured by the MATCH model for the SF6 and CFC cycles and, given the relatively good correlation between MATCH and AGAGE SF6, may be the primary driver of the observed SF6 cycle. (2) A stratospheric influence that produces an April–May minimum in the CFCs and creates a larger normalized amplitude in CFC-11 than in CFC-12. The hypothesized stratospheric influence can explain the larger seasonal amplitude observed for CFC-11 relative to CFC-12, since CFC-11 is destroyed more than twice as fast and at lower stratospheric altitudes than CFC-12 [Wong et al., 1999; Nevison et al., 2004]. Since SF6 is destroyed beginning only in the mesosphere, it might be affected only weakly by the stratospheric influence.

[23] Comparisons of AGAGE and MATCH seasonal cycles for other species and at other stations reveal only a limited number of significant correlations (Figure 6 and Table 2). All species (N2O and both CFCs) are strongly correlated at Samoa with roughly comparable amplitude ratios between AGAGE and MATCH. These cycles are most likely caused by seasonal variations in north-south transport, which are known to produce seasonal cycles in halocarbons observed at Samoa, with the normalized amplitude of the cycles declining with decreasing interhemispheric gradient [Prinn et al., 2000].

[24] The AGAGE and MATCH N2O seasonal cycles are also weakly correlated at Cape Grim, possibly reflecting the MATCH model’s capture of the oceanic seasonal cycle in N2O [Nevison et al., 2005]. AGAGE and MATCH are also correlated at Mace Head for N2O, although we tend to distrust the MATCH results there, owing to erratic cycles in the CFCs and the strong influence of unfiltered pollution events. The only other significant correlations between AGAGE and MATCH occur for both CFC-11 and 12 at Trinidad Head. The MATCH cycles must be caused by tropospheric transport/boundary layer influences, since the prescribed CFC sources lack seasonality. Thus MATCH shows that transport alone can create a summertime minimum in the CFCs at Trinidad Head and may be at least partly responsible for the seasonal cycles observed there. However, some aspects of the observed cycles are not well reproduced by MATCH. Namely, MATCH overestimates the amplitude of the CFC-12 cycle by nearly a factor of 3, possibly reflecting exaggerated boundary layer effects, and provides no mechanism by which the normalized CFC-11 amplitude should be larger than the CFC-12 amplitude, as observed by AGAGE. In contrast, such an observation could be explained by a stratospheric influence, as discussed above for Cape Grim.

[25] If a stratospheric influence is acting at the Trinidad Head and Mace Head stations, the above results suggest that, unlike at Cape Grim, it may be in phase with the tropospheric transport influence, making the two signals difficult to separate. NOAA SF6 data tend to support this idea by showing a primary August minimum in SF6 at Mace Head, coincident with the CFC minima. Similarly, Peters et al. [2004] find that the typical SF6 seasonal cycle at northern midlatitude NOAA stations has a summer minimum, which they attribute to greater summertime vertical mixing over continents that transports SF6 to the free troposphere instead of trapping it in the planetary boundary layer.

3.2.3. Interannual Variability

[26] When MATCH results and AGAGE data are considered separately, both model and observations tend to show correlated IAV among CFC-11, CFC-12, N2O and SF6 (Figure 7 and Table 3). (Note that IAV in AGAGE SF6 is not analyzed here, since data are available only at Cape Grim and only from 2001 onward.) The only exceptions for MATCH are N2O versus both CFCs at Cape Grim and N2O versus CFC-12 at Barbados. The only exceptions for AGAGE are N2O versus CFC-12 at Barbados and CFC-11 versus CFC-12 at Cape Grim. At Mace Head, the amplitude of IAV for MATCH CFCs and SF6 is 5 times larger than that at any other station and is probably strongly influenced by unfiltered pollution events (Figure 7e).

Figure 7.

Atmospheric growth rate anomalies in N2O (solid black line), CFC-11 (dotted black line), and CFC-12 (gray line) normalized by mean tropospheric mixing ratio. (left) MATCH results. (right) AGAGE observations. (a, f) Cape Grim. (b, g) Samoa. (c, h) Barbados. (d, i) Trinidad Head. (e, j) Mace Head. Note different y axis scale for MATCH results at Mace Head (Figure 7e).

[27] Some notable differences between AGAGE and MATCH are first that the normalized amplitude of CFC-11 variability is consistently larger than that of CFC-12 in the AGAGE data, whereas it is smaller for MATCH results at 3 of the 5 stations. Second, the amplitude of IAV in N2O relative to the CFCs is considerably smaller in MATCH (σN2O:σCFC-11 = 0.1 to 0.2; σN2O:σCFC-12 = 0.1 to 0.25) than in AGAGE (σN2O:σCFC-11 = 0.3 to 0.5; σN2O:σCFC-12 = 0.6 to 0.9). (This statement excludes Samoa station, which is discussed separately below.) Third, IAV among the three northern stations and, to a lesser extent, between the two southern stations appears correlated in AGAGE but not in MATCH, suggesting coherent hemispheric influences in the observations but not in the model. Finally, there is a factor of 2 decrease in MATCH IAV moving from Trinidad Head south to Cape Grim, whereas the amplitude of AGAGE IAV is more or less comparable across both hemispheres.

[28] Comparison of MATCH and AGAGE results for individual species (Figure 8 and Table 2) reveals that MATCH IAV is generally smaller than AGAGE IAV, with the exception noted above at Mace Head. Samoa is the only station with a correlation coefficient of R > 0.5 between model and observations, and even there the correlations are fairly weak. IAV in interhemispheric transport affecting Samoa has been strongly linked to tropical circulation changes associated with the El Niño–Southern Oscillation (ENSO). For species with positive north-south gradients, warm ENSO (i.e., El Niño) conditions result in a slowdown of the atmospheric growth rate observed at Samoa due to shifts in the low-level convergence pattern, resulting in a lessened influence of winds from the Northern Hemisphere and a heightened influence of southeasterly winds [Prinn et al., 1992; Elkins et al., 1993; Cunnold et al., 1994; Hartley and Black, 1995]. The MATCH-AGAGE correlations at Samoa suggest that the model is at least partially able to capture this process.

Figure 8.

Atmospheric growth rate anomalies for MATCH model results (black line; note that at Mace Head, MATCH CFC-11 and CFC-12 (dash-dotted lines) have been divided by 5 to keep them on scale) and AGAGE data (gray line). (top) N2O, (middle) CFC-11, and (bottom) CFC-12. (a, f, k) Cape Grim. (b, g, l) Samoa. (c, h, m) Barbados. (d, i, n) Trinidad Head. (e, j, o) Mace Head.

[29] Other factors, such as biogeochemical source variability, may also contribute to the IAV in N2O observed at Samoa. Some of the highest oceanic N2O concentrations on record have been measured in the subsurface waters of the eastern tropical Pacific, upwind of Samoa, where suboxic conditions lead to large microbial N2O yields [Cohen and Gordon, 1978]. An El Niño-induced reduction in upwelling, which ventilates these waters to the atmosphere, would therefore be expected to reduce the ocean N2O source. Indeed, we find that the AGAGE N2O atmospheric growth rate at Samoa is weakly correlated (R ∼ 0.6) with a negative slope to the Niño4 temperature index in the central tropical Pacific (Figure 9). On the basis of the negative slope alone, the cause of the correlation at Samoa cannot be easily distinguished between reduced upwelling versus a slowdown in interhemispheric transport, since both responses would reduce the N2O growth rate. However, consideration of AGAGE CFC data tends to implicate changes in atmospheric circulation and interhemispheric transport rather than biogeochemistry. Like N2O, the CFC growth anomalies are also negatively correlated to the Nino4 SST indices (R ∼ 0.6). If reduced ocean upwelling were primarily responsible for the negative correlation for N2O, one would predict a positive correlation for the CFCs, since they, in contrast to N2O, are undersaturated in deep ocean water [Dutay et al., 2002].

Figure 9.

Atmospheric growth rate of N2O observed at Samoa plotted together with monthly mean Niño4 (160°E–150°W; 5°N–5°S) sea surface temperature time series in the central tropical Pacific (www.cpc.noaa.gov/data/indices/sstoi.indices).

[30] The presentation of MATCH results above suggests that tropospheric transport influences indeed can lead to correlated IAV among long-lived tracers. However, they cannot explain the magnitude of IAV, particularly for N2O, observed at most stations or the observed correlations in IAV between N2O and CFCs at Cape Grim. Furthermore, the MATCH model fails to reproduce the phasing of observed IAV in all species at any station other than Samoa. Below, we explore the possibility that the model simulation, which contained no stratospheric sink, was missing an important factor contributing to IAV in the surface variability of N2O and the CFCs. We note that the model tendency to strongly underestimate the amplitude of IAV for N2O may also suggest a role for surface source variability, but the exploration of this hypothesis is beyond the scope of the current study.

3.3. Stratospheric Circulation and the N2O Sink

[31] In section 3.1, it was suggested that IAV in the anthropogenic N2O source is unlikely to explain the observed ±1.0 to 1.5 Tg N/yr IAV in the N2O growth rate (Figure 2b). In contrast, a simple back-of-the-envelope calculation suggests that IAV in the stratosphere sink term of the N2O budget can reasonably account for the observed growth rate variability. The IAV in the Northern Hemisphere net downward extratropical stratosphere to troposphere mass flux has been estimated as ±20 × 108 kg/s annually [James et al., 2003; Schoeberl, 2004]. Combining this with the observed gradient of N2O between the lowermost stratosphere and the troposphere of ∼20 ppb [Hintsa et al., 1998], and converting by the molar mass ratios of N2O-N and bulk air, yields IAV in the stratospheric backflux of ∼±1.2 Tg N/yr. Year to year changes in the N2O gradient between the lowermost stratosphere and the troposphere could further increase IAV.

[32] The growth rates of N2O, as well as of the CFCs, thus are likely to be affected discernibly by the dilution of the troposphere with CFC and N2O-depleted air from the stratosphere. The large-scale mean meridional (Brewer-Dobson) stratospheric circulation brings air poleward and downward from the tropical middle and upper stratosphere into the lowermost stratosphere, where it exchanges with the troposphere. The downwelling branch of this circulation pattern occurs primarily in the winter hemisphere. The mean meridional circulation is driven by the interaction of differential radiative heating and cooling between the tropics and poles and forcing by planetary and gravity waves. Strong wave activity strengthens the mean meridional circulation and weakens the zonal-mean wind circulation [Newman and Nash, 2000]. Enhanced radiative heating in the tropical stratosphere, for example, due to injection of volcanic aerosol, can also strengthen the mean meridional circulation by increasing the tropical upwelling branch [Schauffler and Daniel, 1994]. IAV in the mean meridional circulation manifests in year to year changes in the strength and isolation of the Arctic and Antarctic polar vortexes and in the poleward transport/mixing and descent of stratospheric trace species [Waugh et al., 1999; Huck et al., 2005].

[33] Stratosphere-troposphere exchange (STE) of chemically depleted air occurs mainly at extratropical latitudes through synoptic-scale mixing events such as tropopause folds. The latter are not necessarily coordinated instantaneously with the mean meridional circulation but should balance over a sufficiently long timescale, for example, greater than seasonal. Since the high-energy solar radiation capable of photolyzing N2O and the CFCs has been absorbed by the time it reaches the lowermost stratosphere, no additional photochemistry is expected in this region. Therefore IAV in the mean meridional circulation, regardless of the details of STE, should provide a reasonable measure of variability in the amount of N2O and CFC-depleted air that enters the troposphere over an annual cycle [Holton et al., 1995].

[34] Polar temperature in the lower stratosphere can serve as a proxy for the strength of the mean meridional circulation, with cold years reflecting a strong polar vortex and a weak meridional circulation and warm years reflecting a weak vortex and a strong meridional circulation [Newman et al., 2001]. In the year 2000, an unusually cold year in the Arctic stratosphere [Manney and Sabutis, 2000], the atmospheric growth rate anomalies of N2O and the CFCs are strongly positive at Mace Head, Trinidad Head and Barbados. This observation appears consistent with a hypothesized reduction in the amount of N2O and CFC-depleted air brought down from the upper stratosphere to exchange with the troposphere.

[35] A more systematic examination at Mace Head, which has the longest time series of the 3 northern AGAGE stations, reveals a modest correlation between Arctic winter lower stratospheric temperature and the atmospheric growth rate anomalies of N2O and CFC-12 and a weaker correlation for CFC-11 (Figures 10a–10c). In this analysis, winter lower stratospheric temperature was estimated from satellite data at 100 hPa, averaged from January through March, 60°N–90°N (Paul Newman, personal communication, 2005). The growth rate anomaly versus stratospheric temperature plots were based on a small set of single annual data points. Uncertainty in these slopes was quantified by systematically removing each individual year of the station data sets from the plot and by varying the range of months used to calculate the average growth rate for a given year. The standard range was assumed to be the 10-month span from October through July for Mace Head data, in an effort to encompass stratospheric downwelling, polar vortex breakup and subsequent midlatitude stratosphere-troposphere exchange. The average growth rate over the 10-month span from June through March was used as the standard for a similar analysis (see below) in the Southern Hemisphere.

Figure 10.

(left) Atmospheric growth rate anomaly calculated from AGAGE data at Mace Head plotted versus mean Arctic (60°N–90°N) winter (January–March) temperature at 100 hPa, an indicator of the strength of the stratospheric mean meridional circulation. (right) Same as left column except at Cape Grim, Tasmania and using mean Antarctic (60°S–90°S) winter/spring (September–October) temperature at 100 hPa. (a, d) N2O. (b, e) CFC-11. (c, f) CFC-12.

[36] Slower growth in N2O and the CFCs occurs in warmer years and faster growth occurs in colder years at Mace Head, consistent with the stratospheric influence hypothesized above (Figures 10a–10c). This correlation is sensitive to individual years. In particular, the 1997 and 2000 data points weaken and strengthen the correlation, respectively, for all 3 species. The MATCH results presented earlier show that tropospheric transport can produce IAV in N2O at Mace Head of comparable amplitude to observations (although the MATCH results may be biased by unfiltered pollution events). It is possible that these transport effects may counteract the stratospheric signal in some years, for example, 1997, but reinforce it in other years, for example, 2000 (Figure 8e).

[37] In contrast to Mace Head, analysis of Cape Grim data shows no correlation between the observed atmospheric growth rate anomalies of N2O or the CFCs versus Antarctic winter lower stratospheric temperature, averaged from September through November from 60°S to 90°S (Figures 10d–10f). The poor correlations at Cape Grim may reflect the fact that the stratospheric signal is more robust in the Northern Hemisphere, where stronger planetary wave generation occurs owing to rougher surface topography and greater synoptic disturbance. Interestingly, the growth rate anomalies for all three species are unremarkable in 2002, despite the unprecedented early stratospheric warming that occurred at the end of September of that year. Unusually low concentrations of N2O and other tropospheric source gases were observed in the Antarctic lower stratosphere at that time, indicating strong poleward transport/mixing and descent [Manney et al., 2005].

4. Conclusions

[38] Like its carbon cycle counterpart CO2, atmospheric N2O is increasing in the atmosphere as a result of human perturbation to global biogeochemistry. However, the more subtle interannual variability in atmospheric N2O may be governed largely by the influences of atmospheric dynamics and the backflux of photochemically depleted air from the stratosphere. Thus IAV in N2O is probably not informative about anthropogenic fluctuations in the nitrogen cycle (e.g., changes in synthetic fertilizer use), except perhaps on longer timescales than the ∼12 year span of this study.

[39] An atmospheric transport model simulation with N2O, CFC and SF6 sources prescribed for 1993–2004, no stratospheric sink, and IAV in meteorological drivers yields generally good correlations in IAV among the four species. This results raises the possibility that “similar behavior” among species due simply to transport effects in the troposphere may be responsible for the correlations in IAV that are generally observed among N2O and CFCs in AGAGE data. However, the amplitude of model IAV tends to be underestimated for all species, especially N2O, and declines from the Northern to the Southern Hemisphere in a manner inconsistent with observations. Furthermore, the model and observed IAV is generally uncorrelated except at Samoa station in the southern tropics, where changes in north-south interhemispheric transport associated with ENSO cycles influence the growth rate anomalies of N2O and the CFCs.

[40] The transport model also yields correlated seasonal cycles between N2O and the CFCs, with exceptions at Trinidad Head and Cape Grim, where the local ocean N2O source may outweigh transport effects. The model predicts a January minimum in the CFCs and SF6 at Cape Grim, more or less in phase with observed SF6 but leading the observed April–May minimum in the CFCs by 2–3 months. Thus the model (with no stratospheric sink) provides no mechanism by which the SF6 minimum should be out of phase with the CFC and N2O minima. At Trinidad Head, the model August seasonal minima in the CFCs are in phase with AGAGE data, suggesting that tropospheric transport variability may contribute substantially to the observed summer minima in the Northern Hemisphere. However, the model provides no explanation for the larger amplitude of the CFC-11 seasonal cycle relative to the CFC-12 cycle that is observed by AGAGE and might be expected from a stratospheric influence.

[41] At the Northern Hemisphere Mace Head, Ireland, AGAGE station, IAV in the growth rate anomalies of N2O and the CFCs is weakly correlated to polar winter lower stratospheric temperature. The latter is a proxy for the strength of the winter downwelling branch of the Brewer-Dobson circulation, which brings N2O- and CFC-depleted air from the middle and upper stratosphere down to the lower stratosphere where it exchanges with the troposphere. Longer high-precision time series are needed to confirm that these correlations with stratospheric circulation are real rather than driven by a few years with cold or warm extremes. At the Southern Hemisphere Cape Grim station, correlations with stratospheric circulation are not evident in the simple analysis presented here based on the lower stratospheric temperature proxy.

Acknowledgments

[42] C. D. N. thanks the Climate and Global Dynamics and Atmospheric Chemistry Divisions of the National Center for Atmospheric Research (NCAR) for their support of her work. NCAR is operated by the University Corporation for Atmospheric Research under the sponsorship of the National Science Foundation. R. G. P. acknowledges support from NASA grant NAG5-12669. R. F. W. acknowledges support from NASA grants NAG5-12806 and NAG5-12806. We thank Adam Hirsch for helpful discussions, Paul Newman for stratospheric temperature data, Nicolas Gruber, Xin Jin, Chris Potter, and Wouter Peters for surface source estimates used in the MATCH simulation, Laurie Porter for his special effort to implement SF6 measurements at Cape Grim, and the entire AGAGE team for making their data available for this study.

Ancillary