GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L23703, doi:10.1029/2006GL028112, 2006 Click Here for Full Article dipole anomaly and its contribution to sea ice export from the in the 20th century Eiji Watanabe,1 Jia Wang,2 Akimasa Sumi,1 and Hiroyasu Hasumi1 Received 7 September 2006; accepted 26 October 2006; published 5 December 2006.

[1] The winter dipole anomaly (DA) in the Arctic dependence of the correlation on data period. Local atmo- atmosphere and its contribution to sea ice export are spheric forcing is also important for sea ice export. A sea- investigated by using a high-resolution coupled global level pressure gradient across the explains the general circulation model. The spatial distributions of the most variance in the sea ice export [Kwok and Rothrock, first two leading EOF modes of winter mean sea level 1999]. Based on previous studies that indicate a possible pressure (SLP) and geopotential height at 500 hPa north of important impact of the second-leading empirical orthogo- 70°N obtained by the long-term simulation (1900–2010) nal function (EOF) mode of Arctic sea level pressure (SLP) are highly similar to those derived from the National Center anomalies on sea ice [Wang et al., 1995; Skeie, 2000], Wu et for Environmental Prediction and the National Center for al. [2006] suggest the winter dipole anomaly (DA) structure the Atmospheric Research (NCEP/NCAR) reanalysis in the Arctic atmosphere. The DA also appears to have a datasets (1948–2004). The first-leading mode corresponds close relationship with sea ice motion based on the Inter- to the (AO). The DA is defined as the national Arctic Buoy Programme (IABP) dataset (1979– second-leading mode. The AO and DA account for 59% 1998) and the National Center for Environmental Prediction and 19% of the total variance, respectively. Composite and the National Center for the Atmospheric Research spatial patterns of SLP, sea ice thickness and velocity in the (NCEP/NCAR) reanalysis data (1960–2002 [Kalnay et extreme years when both the absolute values of principal al., 1996]). The DA is defined as the second EOF mode component (PC1 and PC2) exceed 1.0 standard deviation of monthly mean SLP north of 70°N during the winter indicate that the DA plays a great important role in sea ice (Oct.–Mar.). It is expected that the DA has a great export from the Arctic Ocean to the Sea due to its importance to sea ice export due to its strong meridionality, strong meridionality. Sea ice export is highly promoted although it accounts for only 13% of the total variance. (restricted) in the positive (negative) DA phase. The Available observational data are limited to only the recent dependence of sea ice export on the DA is comparable to few decades. It is difficult to estimate contribution of the or rather larger than that on the AO. Citation: Watanabe, E., DA to sea ice export quantitatively and to discuss signifi- J. Wang, A. Sumi, and H. Hasumi (2006), Arctic dipole anomaly cance of the correlation without inspection using long-term and its contribution to sea ice export from the Arctic Ocean in the coupled simulations. In this study, characteristics of 20th century, Geophys. Res. Lett., 33, L23703, doi:10.1029/ the DA and dependence of sea ice export on the DA are 2006GL028112. investigated using the long-term simulation results (1900– 2010) reproduced by a high-resolution coupled general circulation model (CGCM). 1. Introduction [2] Observational and modeling studies have indicated a 2. The 20th Century Climate Experiment recent rapid decrease of sea ice in the Arctic Ocean, and its close relationship with the Arctic Oscillation (AO [Thompson [3] The 20th century climate experiment was conducted and Wallace, 1998; Wang and Ikeda, 2000, 2001]) and the by the Model for Interdisciplinary Research on Climate North Atlantic Oscillation (NAO [Hurrell, 1995]) [Rigor (MIROC) version 3.2 [K-1 Model Developers, 2004], which et al., 2002; Zhang et al., 2003; Wang et al., 2005]. was developed by the Center for Climate System Research, However, the AO does not always explain sea ice varia- University of Tokyo (CCSR/UT), the National Institute for tions, because sea ice response lags behind atmospheric Environmental Studies (NIES), and the Frontier Research variations [Watanabe and Hasumi, 2005] and is affected by Center for Global Change (FRCGC) as subject No. 1 of some components other than the AO. While high correlation Kyousei project (K1). The horizontal resolution is T106 between sea ice export through the Fram Strait and the AO/ (1.1°) spectral resolution with 56 vertical levels in the NAO indices appears after the late 1970s [Kwok and atmospheric part, and is 1/4° zonally and 1/6° meridionally Rothrock, 1999], some studies show no significant correla- with the rotated to Greenland in sea ice and tion between them [Vinje, 2001]. With respect to this ocean parts. The ocean has 48 vertical levels. The sea ice contradiction, Hilmer and Jung [2000] investigate the model adopts 0-layer thermodynamics [Semtner, 1976] and elastic-plastic-viscous rheology [Hunke and Dukowicz, 1997]. The initial ocean temperature and salinity are given 1Center for Climate System Research, University of Tokyo, Chiba, by the climatology of Levitus and Boyer [1994] and Levitus Japan. et al. [1994]. The CGCM was spun up for 100 years with 2International Arctic Research Center, University of Alaska-Fairbanks, the climate forcing in 1900, then the model was integrated Fairbanks, Alaska, USA. from 1900 to 2100 with several external forcings (solar Copyright 2006 by the American Geophysical Union. irradiance, volcanic eruption, greenhouse gases, stratospher- 0094-8276/06/2006GL028112$05.00 ic and tropospheric ozone, carbonaceous and sulfate aero-

L23703 1of4 L23703 WATANABE ET AL.: THE 20TH CENTURY ARCTIC DA AND ICE EXPORT L23703 sols) based on the historical (1900–2000) and B1 (2001– 2100) scenarios of the Intergovernmental Panel on 4th Assessment Report (IPCC FAR). Note that neither restoring method nor flux adjustment is used throughout the spin up and simulation [Suzuki et al., 2005]. We analyze the simulation data for 110 years from 1900 to 2010 (K1, hereafter).

3. Simulating DA in the Arctic Atmosphere in the 20th Century 3.1. Winter Climatology [4] The winter climatology (Dec.–Feb. mean) of SLP and 500 hPa geopotential height is reproduced well by the long-term simulation. The spatial patterns are similar to that derived from the NCEP/NCAR reanalysis data, while the anti-cyclone over the Beaufort Sea seems to be weaker in the K1 than the reanalysis. Geopotential height is low over the entire Arctic Ocean, while the location of the low center is shifted to the Siberian side, compared to the reanalysis. Even by using the same period as the NCEP (i.e. 1948– Figure 1. Regressed winter mean SLP anomalies 2004 mean), the spatial patterns of SLP and geopotential (Dec.–Feb.) to (a) the first- and (b) the second-leading height are almost same as the 1900–2010 mean. The winter EOF modes derived from the K1 (1900–2010). (c) Same mean sea ice export through the Fram Strait is 0.071 ± 0.023 as Figure 1a except for the NCEP reanalysis (1948–2004). 6 3 À1 Sv (1 Sv = 10 m s ), which is smaller than the (d) Same as Figure 1b except for the NCEP reanalysis observational estimate [Vinje, 2001]. (1948–2004). Contour interval is 0.5 hPa. White solid and 3.2. EOF Analysis dash lines represent the 95% and 99% significance level of correlation coefficient between principal components and [5] Following Wu et al. [2006], the EOF analysis was winter mean SLP anomalies, respectively. applied to winter mean anomalies of SLP and 500 hPa geopotential height north of 70°N. The spatial distributions for the first two modes of SLP derived from the K1 and the when the PC1 of SLP is greater than 1.0 or smaller than NCEP/NCAR reanalysis data are shown in Figure 1. The À1.0 standard deviation, respectively. There are 18 (19) first-leading EOF mode has an annular structure, which years corresponding to the positive (negative) phase of the accounts for 59% (K1) and 63% (NCEP) of the total AO during the 110 simulation years (1900–2010). The variance, respectively. This annular structure corresponds composite AO fields are constructed by averaging the to the AO [Wang and Ikeda, 2000]. The second-leading SLP, geopotential height, sea ice thickness and velocity in mode, which accounts for 19% (K1) and 14% (NCEP) of these years (not shown). The SLP difference between in the the variance, displays an obvious seesaw pattern between positive and negative AO phases is statistically significant the Atlantic side (Greenland-Canadian Archipelago) and the based on the Student T-test, and is particularly, noticeable Siberian side of the Arctic region in both the K1 and the over the eastern Arctic. During the positive AO phase, SLP NCEP. There are two action centers with opposite signs - has a negative anomaly over the whole Arctic region, so that one center is located over the , the other the Beaufort High is weakened. The corresponding cyclonic exists over the . This structure is defined as the wind stress drives sea ice in the same manner and promotes DA [Wu et al., 2006]. Correlation coefficients between the sea ice export through the Fram Strait to the Greenland Sea principal components of the leading mode and the interann- (0.073 ± 0.024 Sv). In contrast, during the negative AO ually varying original anomaly are significant (more than phase, the dominant Beaufort High induces anti-cyclonic 99% confidence level) over each action center. The first- sea ice circulation and tends to restrict the sea ice export and second-leading modes of 500 hPa geopotential height, (0.063 ± 0.021 Sv). These results are consistent with which account for 51% and 19% (K1) and 46% and 22% previous studies [Rigor et al., 2002; Zhang et al., 2003]. (NCEP), also produce the same spatial patterns as that of [7] Next, the composite DA fields are constructed by SLP (not shown). The correlation coefficients of the prin- averaging the extreme years when the absolute value of cipal components (PC1 and PC2) between SLP and 500 hPa the PC2 of SLP exceeds 1.0 standard deviation. There are geopotential height are 0.86 (first mode) and 0.75 (second 23 (18) years corresponding to the positive (negative) phase mode) for the K1. Consequently, both the first-leading of the DA. The SLP difference between the positive and annular structure and the second-leading dipole structure negative DA phases displays the noticeable dipole structure, are regarded as quasi-barotropic, as shown by Wu et al. where one center lies over the Greenland Sea and the other [2006]. is located over the Laptev Sea. This pattern is highly similar to the regression map shown in Figure 1b. The positive DA 3.3. Composite Analysis structure produces an anomalous meridional wind stress, [6] Composite analyses were performed in order to which promotes the sea ice export (0.082 ± 0.021 Sv). In the illustrate typical fields in the positive and negative phases negative DA phase, the expands to the Pacific of the AO and the DA. First, we chose the extreme years

2of4 L23703 WATANABE ET AL.: THE 20TH CENTURY ARCTIC DA AND ICE EXPORT L23703

Figure 3. Same as Figure 2, but winter mean sea ice Figure 2. Composite fields of winter mean SLP (shaded: thickness (shaded: contour interval 20 cm) and velocity contour interval 2 hPa) and wind stress (unit vector 0.1 Pa) (unit vector 10 cm sÀ1). Values are sea ice export (Sv), in state (a) 1, (b) 2, (c) 3 and (d) 4. whose directions are shown by thick arrows. The net sea ice export (Sv) and the standard deviation (Sv) are shown in the side of the Arctic Ocean and anti-cyclonic sea ice bottom purple box. circulation appears. As a result, the sea ice export is significantly restricted (0.060 ± 0.015 Sv). The difference different depending on whether the DA phase is positive or of the export between the positive and negative DA negative (Figure 4). The lowest SLP anomaly is located phases (0.022 ± 0.036 Sv) is comparable to that of the over the Laptev Sea in state 1, while it is located over the AO (0.010 ± 0.045 Sv). Greenland Sea in state 2. On the other hand, the highest SLP [8] The composite fields described above indicate that anomaly is located over the Greenland Sea in state 3, and sea ice export from the Arctic Ocean is promoted (restricted) located over the Laptev Sea in state 4. Consequently, the sign in the positive (negative) phases for both the AO and the DA. of dominant SLP anomaly may be represented by the AO, Next, the cross composite analyses by combining the AO and the location of the anomaly is characterized by the DA. and the DA were performed in order to compare the These climate states seem to be quasi-barotropic, since the dependence of the export on the DA with that on the AO. composite fields of 500 hPa geopotential height agree well In these analyses, we chose the extreme years when both with the SLP. The total sea ice export from the Arctic absolute values of the PC1 and the PC2 exceed 0.5 standard Ocean reaches maximum in state 1, and minimum export in deviation in order to construct following four climate states: state 4. The export in state 3 is comparable to that in state 2. positive AO and positive DA (state 1), positive AO and Note that even when we use 1.0 standard deviation as a negative DA (state 2), negative AO and positive DA (state 3), and negative AO and negative DA (state 4). There are 14 to 15 extreme years in each state. In state 1, a low pressure over the Greenland Sea and the is strengthened and extends eastward along the Siberian coast and a relatively high pressure lies over the Canadian Archipelago (Figure 2a). Corresponding wind stress causes cyclonic sea ice circulation and advects more sea ice toward the Green- land Sea (0.082 ± 0.020 Sv; Figure 3a). In state 2, a low pressure over the Barents Sea is also deepened, and the Siberian High is expanded to the Pacific side of the Arctic Ocean (Figure 2b). As a result, sea ice is accumulated along the Canadian coast. Even during the positive AO phase, the sea ice export is largely restricted in the negative DA phase (0.066 ± 0.019 Sv; Figure 3b). On the other hand, in state 3, the sea ice export is promoted in spite of the negative AO phase (0.075 ± 0.018 Sv; Figure 3c), while sea ice is favorably retained inside the Arctic Ocean in state 4. The export is significantly larger in state 3 than in state 4, where it would be least favorable for sea ice to export through the Fram Strait (0.056 ± 0.014 Sv; Figure 3d). The composite Figure 4. Same as Figure 2, but winter mean SLP (shaded: SLP anomalies from the climatology in each climate state contour interval 1 hPa) and wind stress (unit vector 0.05 Pa) suggest that the location of the dominant anomaly center is anomalies from the climatology (1900–2010 winter mean).

3of4 L23703 WATANABE ET AL.: THE 20TH CENTURY ARCTIC DA AND ICE EXPORT L23703 criterion for construction of composite fields, the results are K-1 Model Developers (2004), K-1 coupled model (MIROC) description, almost same. Hence, we can conclude that the dependence edited by H. Hasumi and S. Emori, K-1 Tech. Rep., 1, 34 pp. Cent. for Clim. Syst. Res., Univ. of Tokyo, Chiba, Japan. of the sea ice export from the Arctic Ocean on the DA is Kalnay, E. C., et al. (1996), The NCEP/NCAR 40-year reanalysis project, comparable to, or rather larger than, that on the AO. Bull. Am. Meteorol. Soc., 77, 437–471. Kwok, R., and D. A. Rothrock (1999), Variability of Fram Strait ice flux and North Atlantic Oscillation, J. Geophys. Res., 104, 5177–5189. 4. Conclusions and Discussion Levitus, S., R. Burgett, and T. S. Boyer (1994), World Ocean Atlas 1994, vol. 3, Salinity, NOAA Atlas NESDIS 3, 111 pp., NOAA, Silver Spring, [9] The winter DA in the Arctic atmosphere and its Md. contribution to sea ice export were investigated using a Levitus, S., and T. S. Boyer (1994), World Ocean Atlas 1994, vol. 4, Temperature, NOAA Atlas NESDIS 4, 129 pp., NOAA, Silver Spring, climate GCM. The simulated spatial distributions of the first Md. two leading EOF modes of winter mean SLP and 500 hPa Rigor, I. G., J. M. Wallace, and R. L. Colony (2002), Response of sea ice to geopotential height anomalies north of 70°N are highly the Arctic Oscillation, J. Clim., 15, 2648–2663. Semtner, A. J. (1976), A model for the thermodynamic growth of sea ice in similar to that derived from the NCEP/NCAR reanalysis numerical investigations of climate, J. Phys. Oceanogr., 6, 379–389. datasets. The composite fields of SLP, sea ice thickness and Skeie, P. (2000), Meridional flow variability over the Nordic seas in the velocity in the extreme years when both the absolute values Arctic Oscillation framework, Geophys. Res. Lett., 27(16), 2569–2572. of PC1 and PC2 exceed 0.5 standard deviation indicate that Suzuki, T., T. Takashi, T. Sakamoto, T. Nishimura, N. Okada, S. Emori, A. Oka, and H. Hasumi (2005), Seasonal cycle of the Mindanao Dome in the DA plays an important role in sea ice export through the the CCSR/NIES/FRCGC atmosphere-ocean coupled model, Geophys. Fram Strait due to its meridional component in wind Res. Lett., 32, L17604, doi:10.1029/2005GL023666. anomaly. The difference of total sea ice export from the Thompson, D. W., and J. M. Wallace (1998), The Arctic Oscillation sig- nature in the wintertime geopotential height and temperature fields, Geo- Arctic Ocean between the positive and negative DA phases phys. Res. Lett., 25, 1297–1300. is comparable to that between the positive and negative AO Vinje, T. (2001), Fram Strait ice fluxes and : 1950– phases. The cross composite fields show that the depen- 2000, J. Clim., 14, 3508–3517. Wang, J., A. van der Baaren, and L. A. Mysak, (1995), A principal compo- dence of sea ice export on the DA is larger than that on the nent analysis of gridded sea-level pressure, surface air temperature, and AO. Hence, the variations characterized by the DA should sea-ice concentration of the arctic region, 1953–1993, C2GCR Rep. 95–4, be also taken into account for analysis of sea ice export. The 18 pp., McGill Univ., Montreal, Que., Canada. (Available from Jia Wang, composite SLP fields show that the location of the dominant [email protected]) Wang, J., and M. Ikeda (2000), Arctic Oscillation and Arctic Sea-Ice SLP anomaly seems to be characterized by the DA, while Oscillation, Geophys. Res. Lett., 27, 1287–1290. the sign of the anomaly is represented by the AO. However, Wang, J., and M. Ikeda (2001), Arctic sea-ice oscillation: Regional and whether the DA is physically independent of the AO or not seasonal perspectives, Ann. Glaciol., 33, 481–492. Wang, J., M. Ikeda, S. Zhang, and R. Gerdes (2005), Linking the Northern remains unknown. Possible mechanisms for the develop- Hemisphere sea ice reduction trend and the quasi-decadal Arctic sea ice ment and maintenance of the DA should be addressed. oscillation, Clim. Dyn., 24, 115–130. Watanabe, E., and H. Hasumi (2005), Arctic sea ice response to wind stress variations, J. Geophys. Res., 110, C11007, doi:10.1029/2004JC002678. [10] Acknowledgments. We are supported by the Frontier Research Wu, B., J. Wang, and J. E. Walsh (2006), Dipole anomaly in the winter Center for Global Change and International Arctic Research Center, Arctic atmosphere and its association with sea ice motion, J. Clim., 19, through JAMSTEC, Japan. The was run on the Earth 210–225. Simulator of JAMSTEC, Yokohama, Japan. We also thank Akasofu for Zhang, X., M. Ikeda, and J. E. Walsh (2003), Arctic sea ice and freshwater IARC/CCSR Exchange Students Funding. J. Wang also acknowledges the changes driven by the atmospheric leading mode in a coupled sea ice- support from Coastal Marine Institute (CMI) of University of Alaska/ ocean model, J. Clim., 16, 2159–2177. Minerals Management Service (MMS).

References ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ Hilmer, M., and T. Jung (2000), Evidence for a recent change in the link H. Hasumi, A. Sumi, and E. Watanabe, Center for Climate System between the North Atlantic Oscillation and Arctic sea ice export, Geo- Research, University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa-shi, Chiba phys. Res. Lett., 27, 989–992. 277-8568, Japan. ([email protected]) Hurrell, J. W. (1995), Decadal trends in the North Atlantic Oscillation: J. Wang, International Arctic Research Center, University of Alaska- Regional temperature and precipitation, Science, 269, 676–679. Fairbanks, 930 Koyukuk Drive, Fairbanks, AK99775-7335, USA. Hunke, E. C., and J. K. Dukowicz (1997), An elastic-viscous-plastic model ([email protected]) for sea ice dynamics, J. Phys. Oceanogr., 27, 1849–1867.

4of4