Report

North Atlantic Oscillation Dynamics Recorded in Greenland Ice Cores

See allHide authors and affiliations

Science  16 Oct 1998:
Vol. 282, Issue 5388, pp. 446-449
DOI: 10.1126/science.282.5388.446

Abstract

Carefully selected ice core data from Greenland can be used to reconstruct an annual proxy North Atlantic oscillation (NAO) index. This index for the past 350 years indicates that the NAO is an intermittent climate oscillation with temporally active (coherent) and passive (incoherent) phases. No indication for a single, persistent, multiannual NAO frequency is found. In active phases, most of the energy is located in the frequency band with periods less than about 15 years. In addition, variability on time scales of 80 to 90 years has been observed since the mid-19th century.

The North Atlantic oscillation (NAO) is one of the Northern Hemisphere's major multiannual climate fluctuations and typically is described with an index based on the pressure difference between Iceland and the Azores (1). On multiannual time scales, variations in the NAO have a strong impact on North Atlantic and European climate (2) and also on the recent surface temperature warming trend in the Northern Hemisphere (3). In recent decades the winter index remained predominantly in a positive state, and there is evidence that during this period the variability might have increased (4). Analysis of various NAO indices (5) showed maximum amplitudes in the frequency bands of about 2, 7 to 8, 20, and 70 years, but none of these peaks is strongly statistically significant. A longer NAO time series can provide more reliable information about the nature of NAO variability and possible dominant time scales associated with this climate oscillation. This information is necessary for both testing theoretical and numerical models and quantifying natural and anthropogenic changes in NAO behavior.

A number of proxy data such as historical records (6), tree ring data (7), and ice core data (8) potentially can be used to estimate long-term NAO variability. Here we reconstruct an annually averaged proxy index from Greenland ice accumulation rates. The correlation between ice accumulation and NAO was shown (9) to be strongly negative in western Greenland, whereas it was weak in central Greenland. The data used for the reconstruction were measured on the NASA-U core (10) located in western Greenland at 73.84°N and 49.49°W and 2370 m above sea level. The annual mean temperature is about −27°C and summer melt layers are formed rarely. A quality check of the record against two shorter neighboring cores shows (11) that the multiannual variability recorded in the core can be taken as representative for that region after the high-frequency part is removed.

The correlation between net snow accumulation at the NASA-U drill site and the mean sea level pressure distribution over the entire North Atlantic region is illustrated in Fig. 1. Both data sets are based on 15 years of monthly mean ECMWF (European Centre for Medium-Range Weather Forecasts) reanalysis data (9,12). The pattern of explained pressure variation by 1 SD in snow accumulation (scaled with −1.0) shows a clear dipole over the Atlantic that strongly resembles the NAO pattern, with negative values over Iceland and positive values at lower latitudes. Thus, it can be expected that the variability of the measured NASA-U ice accumulation rates also reflects the variability of the NAO index.

Figure 1

Teleconnection map between snow accumulation at the NASA-U drill site (*) and mean pressure at sea level. Contours indicate pressure variation associated with 1 SD in snow accumulation. Contour interval is 0.5 hPa and for clarity values are scaled with −1.0. Long dashed line is the zero line. Regions with statistically significant correlation (above 99% confidence level according to Student's t test) are shaded. Both data sets are monthly mean ECMWF reanalysis (9, 12) data for 1979–1993.

In Fig. 2 the proxy NAO index derived from the normalized annual mean ice accumulation rates over the entire ∼350 years (shaded) is compared with the measured annual mean NAO index for the past 130 years (1, 13) (solid line). In both data sets the linear trend and the high-frequency part are removed (11). The correlation coefficient between the two indices is 0.57; hence, about one-third (0.572) of the total variability is explained by a linear relationship. Correlation coefficients with other commonly used NAO indices were also determined (Table 1), showing that the proxy index also captures a good part of the winter index variability. The lowest correlation is found with reconstructed winter NAO variability from tree-ring widths (14). Our proxy index shows particularly low values around 1880, whereas the highest values are around 1695. Periods with a persistently low index (Fig. 2, thin line) occurred in the second half of the 19th century and also around 1950 to 1975 and 1675 to 1690 (Maunder minimum). Persistently positive anomalies are found in the early 18th and early 20th centuries. From 1975 to 1990 the proxy index appears to underestimate the increase in the NAO index; however, it should be noted that these periods correspond to the top meters of the ice core, where the equivalent accumulation analysis is difficult because of uncertainties in density measurements.

Figure 2

Normalized proxy NAO index based on western Greenland ice accumulation rates (shaded) and normalized instrumental NAO index (1, 13) (thick line). Data are annual means averaged from spring to spring (10). Linear trends and high-frequency parts (11) are removed. Also shown is a 15-year running median of the proxy index (thin line).

Table 1

Correlation coefficients between proxy NAO and other common NAO indices. In all data the linear trend and the high-frequency variability are removed (11). Annual means are averaged from spring to spring (13). Hurrell annual and winter indices (1, 13), based on instrumental data, are for 1865–1994. Jones early instrumental NAO index (6, 13) is for 1825–1994 and 1865–1994 (in parentheses); Cook tree-ring reconstructed winter mean index (7) is for 1701–1980 and 1865–1980 (in parentheses).

View this table:

To explore the dynamics of the NAO index, we used a Morlet wavelet analysis (15, 16) instead of a classic Fourier analysis. The statistical significance of the local wavelet power spectrum was tested by a Monte Carlo method. Our null hypothesis states that the NAO index is an autoregressive (AR-1) noise time series with autocorrelation coefficient α estimated from the observed data. Both time series, the proxy index as well as the instrumental index, are nearly white with α = −0.1 and +0.1, respectively.

The local wavelet spectrum of the proxy NAO index (Fig. 3A) shows a number of time sequences with spectral power above the 90 and 95% confidence levels. The most pronounced sequences occur during 1685 to 1720, 1730 to 1775, 1870 to 1900, and 1900 to 1930; a weaker phase has prevailed from 1960 onward. The spectrum does not show one or more dominant and persistent NAO frequencies throughout the past 350 years but is characterized by a highly nonstationary behavior. One striking feature is that most of the area with significant coherent oscillation is located in the frequency band of less than 15 years. Maximum power is around 5 to 7, 9 to 11, and 12 to 14 years. In addition, we note that from 1850 onward the proxy NAO index also shows significant (95%) power at 80 to 90 years. Roughly comparable multiannual (16, 17) and century scale (17, 18) variability was found in a number of independent temperature proxy analyses. This is consistent with the profound effect of NAO-like variability on the northern hemisphere and European surface temperature (2, 3). Note that in the first 200 years of the proxy NAO record no indication of such a century-scale variability is found and that the onset of this oscillation is comparable to that expected from greenhouse gas forcing (19).

Figure 3

Local wavelet power spectrum for the proxy NAO index (A) and for the instrumental NAO index (B) as in Fig. 2, based on a Morlet wavelet with a characteristic frequency of six calculated as described in (15). Amplitudes are scaled with the variance of the respective index; hence, the expected power of white noise is 1. Logarithmic vertical axes indicate equivalent periods, and horizontal axes indicate time. The 90 and 95% confidence limits are shown as thin lines, and the cone of influence marks regions where edge effects might underestimate the amplitudes. Hatched region indicates periods for which the proxy index should not be interpreted.

The local wavelet power spectrum of the measured annual NAO index for the instrumental period (Fig. 3B) shows a number of active phases with maximum amplitude in frequency bands similar to those in the spectrum of the proxy index (Fig. 3A). Although there are some differences, the two local wavelet power spectra are similar and indicate that our proxy NAO index correctly reproduces the multiannual variability exhibited by the instrumental annual NAO index.

A number of mechanisms have been proposed to explain NAO-like variability, including uncoupled and coupled atmosphere-ocean interactions (20). For a truly internally or externally forced climate oscillation, one would expect to see statistically significant wavelet power throughout most of the 350 years indicated in Fig. 3A, although the dominant frequency might change in time (21). In contrast, the observed power spectrum of the proxy index suggests that the NAO is an intermittent climate oscillation characterized by temporally active (coherent) and passive (incoherent) phases. Atmosphere-ocean interaction on the typical time scales of 5 to 15 years might occur during active phases but would be absent during passive phases, although spatially coherent patterns still may exist. Note that we cannot exclude the possibility that the proxy NAO index represents simply stochastic variability. Monte Carlo simulations based on a large number of autoregressive noise time series with the same autocorrelation coefficient as the proxy index showed (22) that randomly distributed active and passive sequences with wavelet spectra above the 95% significance level can also occur by chance.

Our study shows that it is possible to reconstruct a proxy NAO index from carefully selected Greenland ice core data and that intermittency is an important feature in the NAO for the past 350 years. If this intermittency is not simply caused by a stochastic process, it may have implications for climate prediction in the Atlantic region and Europe (23), because the predictability should be increased during coherent active NAO phases.

  • * To whom correspondence should be addressed. E-mail: christof.appenzeller{at}climate.unibe.ch

REFERENCES AND NOTES

View Abstract

Stay Connected to Science

Navigate This Article