Variations in Atmospheric N2O Concentration During Abrupt Climatic Changes

See allHide authors and affiliations

Science  09 Jul 1999:
Vol. 285, Issue 5425, pp. 227-230
DOI: 10.1126/science.285.5425.227


Nitrous oxide (N2O) is an important greenhouse gas that is presently increasing at a rate of 0.25 percent per year. Records measured along two ice cores from Summit in Central Greenland provide information about variations in atmospheric N2O concentration in the past. The record covering the past millennium reduces the uncertainty regarding the preindustrial concentration. Records covering the last glacial-interglacial transition and a fast climatic change during the last ice age show that the N2O concentration changed in parallel with fast temperature variations in the Northern Hemisphere. This provides important information about the response of the environment to global climatic changes.

Nitrous oxide (N2O) is an atmospheric trace gas with a relatively long lifetime of about 120 years (1). The main sources of N2O in preindustrial times have been tropical soils, the ocean in upwelling regions, and soils in temperate regions (estimated contributions are 45, 30, and 25%, respectively, but with high uncertainties) (1). The main sink is photodissociation in the stratosphere (1). The atmospheric concentration of this greenhouse gas [which, apart from water vapor, is third in importance after carbon dioxide (CO2) and methane (CH4)] reached 314 parts per billion by volume (ppbv) in 1998 (2) and increases by about 0.25% per year (1). Knowledge about the atmospheric N2O concentration before 1976 comes mainly from ice core analyses (3–7). Our record covering the past millennium allows us to narrow the range of the preindustrial concentration.

Records of CH4 variations parallel to climatic oscillations have provided a wealth of information about the global significance of climatic oscillations and about the response of the environment to such variations (8–10). Like CH4, N2O has an important source in the tropics [wet forest soils and dry savannahs for N2O; wetlands for CH4 (1)], but in contrast to CH4 it has an important oceanic source too. Therefore, we expect that our measurements, which cover time periods of drastic climate changes, will provide new information about environmental responses to those changes. We present examples from the last glacial-interglacial transition [16.5 to 10.5 thousand years before the present (kyr BP)] and from a fast climate oscillation during the last ice age (37.0 to 32.5 kyr BP).

The air trapped in polar ice samples was extracted with the melt-refreezing method used for CH4 analyses (10, 11). The N2O concentration of the extracted air was measured with a gas chromatograph equipped with an electron capture detector. Tests with bubble-free ice and standard gas (the working standard was 304 ± 4 ppbv) confirmed that N2O can be extracted with the melt-refreezing method despite the high solubility of N2O in water (12), if the freezing speed is kept low. The small sample size (about 40 g) allowed us to measure several samples per annual layer for samples from this millennium as well as one sample per annual layer for the transition from the last glacial epoch to the Holocene. Data presented in this study are not corrected for gravitational fractionation (13).

Samples for the first measurement series, covering the past 1000 years (Fig. 1), are from cores drilled at Summit (central Greenland, 72°34′N, 37°38′W) as part of the EUROCORE project and the Greenland Ice Core Project (GRIP). The preindustrial N2O concentration was relatively stable, with an average of 270 ± 5 ppbv for the time period between 1400 and 1750 A.D. Previously published data (3–6) are highly scattered and allow only a rough estimate of the preindustrial concentration of between 260 and 285 ppbv (Fig. 1). An exception are the precise results by Machidaet al. (7) covering the past 300 years. The low scatter of our measurements and the good agreement with the measurements by Machida et al., which were performed with a dry extraction technique, as well as the good agreement with the direct measurements on atmospheric air (14) and firn air measurements from Antarctica (15), indicate that our measurements for this time period are reliable and allow us to investigate changes in the atmospheric N2O concentration that occurred together with climatic changes in past climate epochs.

Figure 1

Comparison of our N2O measurements with previously published values for the past 1000 years. Shown are EUROCORE (•) and GRIP mean values (⧫) from this study and the calculated 1σ band of EUROCORE and GRIP mean values (shaded area). Previously published data are from Lawdome (+) (3), Byrd (▿), Crête (⋄), Camp Century (□) (4), D57 (○) (5), Dye 3 (▵) (6), H15 (▴) (7), the South Pole (⋆) (15), and direct atmospheric measurements from Cape Grim (×) (14). Each point of our record is either a single measurement or the mean value of two to six measurements on samples from the same core section. The error bars represent the analytical reproducibility (1σ). For the age scale, the difference between the age of the air and of the surrounding ice has been taken into account (20). The 1σ band through our data was obtained by a Monte Carlo simulation (25) and marks the range of a likely evolution of the N2O concentration for this period.

In Fig. 2, the N2O record for the transition period from the last glacial epoch to the Holocene measured along the GRIP ice core is compared with the GRIP records of oxygen isotope δ18O (16) (used as a proxy for temperature) and CH4 (8,17). The N2O concentration increased during the transition from around 200 ppbv in the late glacial epoch to about 270 ppbv in the early Holocene. This increase was interrupted by a decrease during the Younger Dryas (YD) period (12.7 to 11.6 kyr BP).

Figure 2

GRIP N2O (solid and open circles), δ18O (16) (upper trace), and CH4 (17) (lower trace) records for the last glacial-interglacial transition [the GRIP time scale is in years before 1989 (20, 26)]. For N2O, the results from all individual samples are plotted together with their analytical uncertainty (1σ). We assume that high N2O values (open circles) are caused by artefacts. Solid circles indicate reliable results. Error bars indicate estimated 1σ uncertainty (4 ppbv). Measurements that are out of range are marked by arrows at the corresponding time. The light shading denotes the time period of the B/A; the dark shading denotes the YD event. The uncertainty in the difference between ice and gas ages (Δage) and therefore in the comparison of N2O and CH4 to δ18O is about 100 years (17).

The N2O record shows a few surprisingly high values, which deviate from their neighboring samples (partly from the same core section) by much more than our mean analytical reproducibility (3.7 ppbv). Values that we assume to be too high are marked as open circles in Fig. 2. The highest values measured in this section (11.4 kyr BP, 404 ppbv; 16.3 kyr BP, 297 ppbv) are out of range and are marked by arrows at the corresponding time. The long lifetime of N2O excludes variations of 20 ppbv or more in only a few years. Therefore, the high values cannot represent the atmospheric N2O concentration and are probably due to an artefact (18).

The question arises whether the synchronous evolution of the N2O record with the records of δ18O and CH4 could be an artefact caused by variations of the impurities in the ice with climate changes. Calcium (Ca++) is a proxy for impurities in the ice and shows variations that are parallel to climatic changes. The GRIP Ca++ concentrations are elevated (19) and the N2O values are depleted during the YD event. However, there is a clear depth shift for the beginning as well as for the termination of the two signals in the ice core. Such a shift is expected if the measured N2O signal corresponds to the atmospheric concentration, because of the age difference between ice and enclosed air (20). If reactions with chemical impurities caused the N2O signal, no such shift would be expected. We conclude that it is very unlikely that the whole N2O record during fast climate changes is an artefact caused by chemical impurities in the ice.

In the GRIP record covering a fast climate oscillation during the last ice age—Dansgaard-Oeschger (D-O) event 8 (36.5 to 33.5 kyr BP)—the N2O concentration increased from below 210 ppbv to over 250 ppbv and decreased again to less than 210 ppbv within 3.5 kyr (Fig. 3). The highest value measured in this time period (302 ppbv at 36.0 kyr BP) is marked by an arrow at the corresponding time in Fig. 3. For the same time period, 13 samples of the Byrd ice core from Antarctica (80°S, 120°W) were analyzed as well (Fig. 3). Given today's small interhemispheric gradient of approximately 1 ppbv (1), we expect that the interhemispheric difference was also small during D-O event 8 and that therefore the measurements from the two cores should be almost the same within the analytical uncertainty. The Byrd measurements indicate an increase in the N2O concentration from approximately 210 ppbv to a maximum value of 247 ppbv around 35 kyr BP. The increase at the beginning of D-O event 8 and the absolute values during the mild phase are in good agreement with our GRIP results. This agreement between two records from ice cores that differ significantly in their chemical impurities strongly supports our assumption that the general N2O record measured in Greenland ice represents the atmospheric concentration.

Figure 3

GRIP N2O (solid circles, left axis), δ18O (16) (upper trace), and CH4 (17) (lower trace) records for D-O event 8. We assume that high N2O values (open circles and arrow at the corresponding time) are caused by an artefact. For this time period, the uncertainty in Δage is better than 300 years (17). N2O measurements from the Byrd core for the same time period are shown as solid diamonds (right axis). Error bars indicate estimated 1σ uncertainty (4 ppbv). The time scale of the Byrd gas record was synchronized to that of the GRIP gas record with an uncertainty of 200 years (17).

The general pattern of N2O variations is similar to that for CH4, with lower concentrations during cold climate stages and higher concentrations during warm ones. Like CH4, N2O seems to be coupled with Greenland temperature and therefore to the climate changes in the Northern Hemisphere, whereas the temperature reconstructed for Antarctica shows a different pattern (17). The preindustrial emission rate of the natural sources is estimated at 11 × 1012g of nitrogen per year (11 Tg of N year−1) (21). We assume for the following estimates that the lifetime of N2O was constant and therefore that the equilibrium concentration was proportional to the annual emission rates. Lower concentrations in the past suggest that emission rates assuming equilibrium states were reduced by 10% (1.1 Tg of N year−1) during the YD event and by about 26% (2.9 Tg of N year−1) in the late glacial relative to preindustrial emission rates. During D-O event 8, the concentration changed by about 50 ppbv, corresponding to an emission rate change of 2 Tg of N year−1. These are significant changes in any single natural source. Additional information about environmental changes in response to climatic changes is expected from measurements of changes in the interhemispheric gradient of N2O and of time lags between changes in the concentrations of CH4 and N2O. The interhemispheric gradient cannot be determined with the present precision of measurements. Time lags are especially pronounced during the two transitions from mild events with a relatively high N2O concentration to cold events with a low N2O concentration. The decreases of N2O at the end of the warmer epochs started at the same time or even before the decreases of CH4 and the final decrease of δ18O after a last peak value, but occurred more slowly. Only part of the delayed decrease of N2O as compared to CH4 after a climate cooling can be attributed to the longer atmospheric lifetime of the former.

At the end of D-O event 8, the CH4 concentration showed its highest value at about 34.5 kyr BP (Fig. 3) and decreased over about 500 years to a value typical for the cold phases. Assuming a linear decrease of the N2O source from 10.4 to 8.4 Tg of N year−1 in the same time period as the CH4sources, we would expect a N2O concentration below 210 ppbv at about 33.8 kyr BP (22). Therefore, we conclude that the reduction of N2O sources was slower than that of methane sources at the end of D-O event 8.

The difference is even more pronounced at the transition from the Bølling/Allerød (B/A) (14.5 to 12.7 kyr BP) to the YD (12.7 to 11.6 kyr BP) (Fig. 3). The CH4 concentration decreased very quickly, indicating a reduction of methane sources to late glacial levels within 250 years or less. If N2O sources had been reduced in the same time span to late glacial levels, we would expect a concentration of about 200 ppbv in the middle of the YD period despite the longer lifetime of N2O. The N2O sources were reduced more slowly than the CH4 sources and, in contrast to CH4 sources, did not reach a usual cold phase level (about 8 Tg of N year−1) during the YD event.

Time lags during increases of the N2O concentration at the beginning of mild phases or of the Holocene are less obvious. The increase at the beginning of D-O event 8 is difficult to determine because there is a gap of results between 36.2 and 36.7 kyr BP. An increase in the N2O concentration of 50 ppbv occurred either in 1250 years or in a little less than 1000 years. However, the increase in the source was a gradual one. The increases at the beginning of the B/A and the Holocene are difficult to determine because of results that we consider to be affected by artefacts. N2O started to increase slowly at about 16 kyr BP to reach a level of about 220 ppbv in 14.8 kyr BP and was followed by a fast increase of about 45 ppbv in 300 years. At the YD-to-Holocene transition, the N2O concentration increased by 30 ppbv in about 200 years. Considering the present uncertainty, the increases at the beginning of the B/A and the Holocene can either be explained by an immediate increase of the sources or a gradual increase within 300 and 200 years, respectively, as shown by very simple model calculations.

The main candidates for changing sources of N2O are the oceans and terrestrial soils. The N2O production in soils depends mainly on the input of organic matter, fertility, moisture status, temperature, and oxygen status (23). The oceanic N2O source depends not only on N2O production but also on the transport of N2O from depth (24) to the surface water. Climate changes influence both sources, but their individual contribution and even the sign of their response to climate variations are difficult to estimate. Because both oceanic and terrestrial sources can change very quickly, even the rate of a N2O increase (for example, from the last glacial to the B/A) is not sufficient to constrain the sources. On the basis of our data, we cannot yet determine the sources responsible for the N2O concentration changes discussed here.

An identification of sources will only become possible when the isotopic signature of N2O trapped in ice can be measured accurately. However, it is not a priori clear whether N2O trapped in ice keeps its isotopic signature over centuries and millennia. Comparison of additional high-resolution N2O records with CH4 (mainly terrestrial source) and CO2 records (largely influenced by the oceanic source) and identification of leads and lags of concentration changes of these three greenhouse gases during climate changes will provide more detailed information about the response of the environment to climate changes. In addition, model simulations of oceanic and terrestrial sources should be performed to further constrain the N2O source history on the basis of measurements of past N2O changes.

  • * Present address: Princeton University, Princeton, NJ 08544, USA.

  • To whom correspondence should be addressed. E-mail: stauffer{at}


View Abstract

Stay Connected to Science

Navigate This Article