Spring Arctic sea ice as an indicator of North American summer rainfall

Authors

  • Mirong Song,

    1. State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029, China
    Search for more papers by this author
  • Jiping Liu,

    Corresponding author
    1. State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029, China
    • LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, 100029 China.
    Search for more papers by this author
  • Chunyi Wang

    1. Chinese Academy of Meteorological Sciences, Beijing, 100081, China
    Search for more papers by this author

Abstract

Associations between the spring Arctic sea ice concentration (SIC) and North American summer rainfall were discussed using the singular value decomposition analysis. Results show that a reduced SIC in the North Atlantic and Pacific sectors of the Arctic, and an enhanced SIC in much of the central Arctic Basin and Beaufort Sea, are accompanied by dry conditions over the western United States, the northern Great Plains, the Midwest, westernmost Canada, and central and eastern Greenland, and wet conditions over the southern United States, Alaska, northern Canada, and western Greenland. Atmospheric circulation anomalies associated with the SIC variability show two wave-train structures, which are persistent from spring to summer, leading to the identified relationship between the spring Arctic SIC and North American summer rainfall. This relationship indicates a potential long-term outlook for the North American summer rainfall as the decline of the spring sea ice in the north Atlantic and Pacific sectors of the Arctic is expected to continue as climate warms. Copyright © 2011 Royal Meteorological Society

1. Introduction

Sea ice is a key player in global climate system (e.g. IPCC, 2007). High albedo of sea ice substantially reduces solar energy absorbed by surface. Insulating effect of sea ice drastically reduces exchanges of heat, moisture, momentum, and CO2 between the atmosphere and the ocean. Sea ice can potentially influence global thermohaline circulation through impacts on dense water formation and fresh water budget. Sea ice is also a key indicator of climate change. Many studies indicate that the Arctic sea ice has remarkably decreased (extent and area) and thinned for the period of satellite observations (e.g. Rothrock et al., 1999; Comiso et al., 2008). Because the albedo, insulating, and water density effects are so strong, large interannual variability and rapid decline of the Arctic sea ice is expected to have substantial impacts on global climate.

Some modelling studies have suggested that changes of the Arctic sea ice strongly impact the overlying atmosphere (e.g. Alexander et al., 2004), causing feedback to atmospheric circulation well beyond the Arctic boundary. Herman and Johnson (1978) noted significant far-field atmospheric anomalies in response to the ideal ice edge difference in the Arctic. Murray and Simmonds (1995) found that the reduced Arctic ice cover could result in a weakening of mid-latitude westerlies, affecting the speeds and intensities of storms. Honda et al. (1999) identified significant downstream anomalies over the Bering Sea, Alaska, and North America in the form of a stationary-wave train in response to heavy and light ice extent in the Sea of Okhotsk. Deser et al. (2004) and Singarayer et al. (2006) indicated that anomalies in sea ice extent might be more efficient than anomalies in sea surface temperature at exciting atmospheric response. All these studies suggest substantial influences of the Arctic sea ice variability on extrapolar climate.

The purpose of this study is to discuss relationships between the spring Arctic sea ice and North American summer rainfall, which has received relatively little attention, and possible mechanisms responsible for their associations.

2. Data and method

Datasets used in this study include: (1) Arctic sea ice concentrations (SIC) retrieved from the Scanning Multichannel Microwave Radiometer and the Special Sensor Microwave/Imager based on a bootstrap algorithm (http://nsidc.org/data/docs/daac/nsidc0079_bootstrap_seaice.gd.html for details), which provide the longest quality-controlled record for studying interannual and decadal sea ice variability; (2) precipitation from the Global Precipitation Climatology Project based on estimates from gauge measurements and satellite retrievals (Adler et al., 2003, for details); (3) 200-mb and 700-mb winds, and 500-mb geopotential height from the National Centers for Environmental Prediction (NCEP) reanalysis II (Kanamitsu et al., 2002).

The common time period for all the datasets is from 1979 to 2007. For each year, we generated spring (March–April–May) and summer (June–July–August) averages. Anomalies were calculated by subtracting the climatology. The singular value decomposition (SVD) analysis (Bretherton et al., 1992) was used to show coupled patterns between the spring Arctic SIC and North American summer precipitation. To identify pairs of coupled patterns between two fields, the SVD analysis is performed on the covariance matrix with weights proportional to the area of the grid between two space- and time- dependent data fields. The result is a spatially orthogonal set of singular vectors (analogous to the eigenvectors in the empirical orthogonal function (EOF) analysis), and a set of singular values associated with each pair of vectors (analogous to the EOF eigenvalues) that explain some quantifiable fraction of the squared covariance between two fields. As in the EOF analysis, the time series of expansion coefficients of the SVD analysis is determined by projecting the original field onto the singular vectors.

3. Results

Figure 1 shows the spatial patterns and temporal coefficients of the first SVD mode between the spring Arctic SIC and North American summer rainfall, which account for 20% of the total covariance. The spatial distribution of the spring SIC (hereafter referred to as SVD1-SIC, Figure 1(a)) shows negative values in the north Atlantic (Baffin Bay/Davis Strait, the Labrador Sea, and the majority of the Barents Sea and the Greenland Sea) and Pacific (the Sea of Okhotsk, the Bering Sea, and the southern Chukchi Sea) sectors of the Arctic, and positive values in much of the central Arctic Basin and Beaufort Sea. The spatial distribution of the North American summer rainfall (hereafter referred to as SVD1-NAR, Figure 1(b)) exhibits broad negative values extending from western to central North America, and positive values in northern and southeastern North America. As shown in Figure 1(c), the corresponding temporal coefficients of both SVD1-SIC and SVD1-NAR show coherent variations. Besides strong year-to-year variability, there is a statistically significant upward trend (0.89 per decade for SVD1-SIC and 0.74 per decade for SVD1-NAR). The correlation between the temporal coefficients of SVD1-SIC and SVD1-NAR is 0.95, which still maintains high value with the trend removed (0.93), indicating strong covariability between SVD1-SIC and SVD1-NAR on interannual time scale. For the period of 1979–2007, the standardized time series of both SVD1-SIC and SVD1-NAR underwent a change from one of negative phase to one of positive phase. That is, there is a tendency toward less spring SIC in the North Atlantic and Pacific sectors of the Arctic, and more spring SIC in much of the central Arctic Basin and Beaufort Sea from 1979 to 2007. Meanwhile, summer rainfall over North America featured dry conditions over the western United States, the northern Great Plains, the Midwest, westernmost Canada, and central and eastern Greenland, and wet conditions over the southern United States, Alaska, northern Canada, and western Greenland from 1979 to 2007.

Figure 1.

First SVD mode of the spring Arctic SIC (SVD1-SIC) and North American summer rainfall (SVD1-NAR) for the period 1979–2007. The spatial pattern of (a) SVD1-SIC and (b) SVD1-NAR (the units are arbitrary), and (c) the standardized temporal coefficients of SVD1-SIC (red line) and SVD1-NAR (blue line). The solid and dash lines are original and detrended temporal coefficients, respectively

To insure the reliability of the first SVD modes, we performed the EOF analysis on the spring Arctic SIC and North American summer rainfall, respectively, for the period 1979–2007. The EOF modes shown in Figure 2 are statistically significant and distinct from the other EOFs based on the method proposed by North et al. (1982). It appears that the first EOF mode of the spring Arctic SIC shows similar spatial pattern to that of SVD1-SIC, although the sign in the Labrador Sea and northeastern Bering Sea is somewhat reversed (Figure 2(a)). Its corresponding principle component has significant correlation with the temporal coefficients of SVD1-SIC (0.50). Similarly, the spatial pattern and temporal evolution of the second EOF mode of the North American summer rainfall is in good agreement with that of SVD1-NAR (Figure 2(b), the correlation between the time series is 0.96). Thus, sea ice (rainfall) variability in SVD1-SIC (SVD1-NAR) somehow reflects the dominant spatiotemporal variability of the spring Arctic SIC (North American summer rainfall). Additionally, the spatial pattern of SVD1-SIC (SVD1-NAR) is similar to the linear trend of the spring Arctic SIC (North American summer rainfall) for the period of 1979–2007 (Figure 3), although some differences exist. All the above analyses suggest that the relationships between the spring Arctic SIC and North American summer rainfall arise from a combination of forcings by fluctuations in SIC as well as by the decline of sea ice due to global-scale warming.

Figure 2.

EOF mode of the (a) spring Arctic SIC and (b) North American summer rainfall for the period 1979–2007 (the units are arbitrary)

Figure 3.

Linear regression trends of the spring Arctic SIC (% per decade) and North American summer rainfall (mm/day per decade). Contours give the trends that are significant at the 90% confidence level

The question is, how is sea ice variability in SVD1-SIC linked to rainfall variability in SVD1-NAR? We examined changes of large-scale atmospheric circulation associated with SVD1-SIC. Specifically, we examined the regression patterns of the NCEP 200-mb zonal wind, 500-mb geopotential height, 700-mb zonal and meridional winds in summer (Note the summer fields being analysed are for summer following spring used for the SVD-SIC/NAR analysis.) with respect to the standardized time coefficients of SVD1-SIC.

Corresponding to one positive standard deviation change in the time coefficients of SVD1-SIC, at upper levels, there is a weakening of the westerly wind (∼30°N), and a strengthening of the zonal wind to the north (∼50°N) (Figure 4a). Meanwhile, the westerly flow (∼60°N) is weakened, although it is not statistically significant (Figure 4(a)). As a result of the changes of the westerly wind, the north-south tripole is evident in the 500-mb geopotential height anomalies forming two wave train structures (Figure 4(c)). One is over the northern Pacific and western North America, with anomalous centers over the Bering Sea, the Gulf of Alaska, and the western United States (Figure 4(c)). The other is over the Atlantic coast of North America, with anomalous centers over the western Atlantic, the northern Atlantic, and Greenland to the Barents Sea (Figure 4(c)). At low levels, an anomalous cyclonic circulation occupies southeastern North America. The anomalous easterlies in the northern flank of the anomalous cyclone impinge on central North America, and are merged with the southern flank of the anomalous anticyclone extending from the western United States to the subtropical eastern Pacific. This anomalous cyclone is also coupled to an anomalous anticyclone over the western Atlantic (Figure 4(e)). It appears that the atmospheric responses become more statistically significant from upper to lower levels.

Figure 4.

Regression coefficients of the 200-b zonal wind (m/s, a and b), 500-mb geopotential height (m, c and d), and 700-mb winds (m/s, e and f) in summer (left panel) and spring (right panel) on the standardized time series of SVD1-SIC. The light shading gives anomalies that are significant at the 90% confidence level

In fact, the two wave-train structures start to emerge in spring. Moreover, the aforementioned atmospheric anomalies in the 200-mb zonal wind, 500-mb geopotential height, and 700-mb winds are more pronounced (with more statistically significant grid points) in spring (Figure 4(b), (d) and (f)). For the wave train over the Pacific coast of North America, the locations of the anomalous centers do not change much (Figure 4(e) and (f)), with slight changes to the center W2 (westward) and center W1 (eastward) from spring to summer. By contrast, for the wave train over the Atlantic coast of North America, the locations of the anomalous centers show obvious eastward propagation from spring to summer (E1 and E2 in Figure 4(e) and (f)).

Furthermore, we removed the trend of the time series of SVD1-SIC, and repeated the above regression analysis. The atmospheric anomalies accompanied with the detrended and original SVD1-SIC are similar, although some differences exist (not shown). This suggests that the aforementioned atmospheric anomalies are driven primarily by interannual variability of SVD1-SIC.

The drying of western North America is coincident with the anomalous easterly/northeasterly winds, which opposes the climatological westerly/southwesterly winds associated with the subtropical high, thereby limiting moist air from the eastern Pacific into western North America, and decreases rainfall there. The anomalous northerly winds along the Great Plains (∼100°W) and northwesterly winds over the Midwest tend to block the climatological winds, thereby inhibiting warm and moist air from the Gulf Coast reaching upper Great Plains and the Midwest, and suppresses rainfall there. The wet conditions along the Atlantic coast stem from the combined effects of the anomalous cyclonic circulation over southeast North America and the anomalous anticyclonic circulation over the western Atlantic, which tend to shift the climatological southerly/southeasterly winds eastward and encourage warm and moist air from the Gulf Coast and subtropical western Atlantic to the eastern United States. The anomalous northerly winds over southern Alaska reduce warm and moist air from the northeastern Pacific, resulting in below-normal precipitation in Alaska. The above-normal precipitation over the western and central Greenland is due to the anomalous advection and ascent of warm and moist air from the northern Atlantic. Additionally, concomitant with the aforementioned shift of the high-latitude westerly is the poleward shift of storm tracks, which is helpful for a general increase in precipitation for northern North America.

4. Discussion and conclusion

In this study, we found that the decrease (increase) of the spring SIC in the north Atlantic and Pacific sectors of the Arctic (in much of the central Arctic Basin and Beaufort Sea) is associated with two wave patterns in the atmosphere, which is persistent from spring to summer. This contributes to the summer rainfall deficit (surplus) in the western to central (northern and southeastern) North America. Using the atmospheric model, Zhao et al., (2004) examined atmospheric response to heavy and light sea ice conditions the Sea of Okhotsk and Bering Sea in spring. The models show significant responses propagating eastward from the Bering Sea to North America in the form of a stationary wave train. Also, the numerical experiments of Alexander et al. (2004) and Magnusdottir et al. (2004) show that sea ice anomalies in the North Atlantic can result in an out-of-phase relationship in geopotential height between the polar cap and farther south in the North Atlantic. Thus, the wave-train structures identified here are in good agreement with their model results.

What is the mechanism that helps the atmosphere maintain the memory (two wave patterns) from spring to summer? Blanchard-Wrigglesworth et al. (2011) analyse the lagged correlation of the Arctic sea ice in observations and a GCM ensemble. They showed a persistence of total sea ice area anomalies for the 1–5-month time scale, longer in winter and summer months and shorter in spring and fall months. Figure 5(a) shows the regression pattern of the summer Arctic SIC on the standardized time series of SVD1-SIC. It shows that the reduction of the spring SIC in the North Pacific and Atlantic sectors of the Arctic also occurs in summer, although sea ice retreats in the Arctic Ocean. This also closely resembles the linear trend of the summer Arctic SIC for the period of 1979–2007 (Figure 5(b)). We speculate that the spring SIC anomalies are persistent in the following summer. As a result, the accumulated impacts of the anomalous SIC persistency from spring and summer influence the large-scale atmospheric circulation (e.g. jet streams and storm tracks) through changing surface albedo and exchanges of surface fluxes between the atmosphere and the ocean, which further modulates distribution of stationary waves in the atmosphere.

Figure 5.

Regression coefficients of the summer Arctic SIC (%) corresponding to one standard deviation change in the time series of SVD1-SIC, and linear regression trend of the summer Arctic SIC (% per decade). Contours give anomalies and trends that are significant at the 90% confidence level

We also examined the possible linkages between sea ice variability in SVD1-SIC and some of the most prominent large-scale modes of climate variability in the Northern Hemisphere (http://www.cdc.noaa.gov/ClimateIndices/Analysis). As shown in Table I, the North Pacific Index (NPI), Pacific Decadal Oscillation (PDO), and NAO have little correlation with SVD1-SIC. As compared to other climate indices, the Pacific-North America pattern (PNA) shows moderate correlation with SVD1-SIC, but it is not statistically significant and only accounts for 12% of the shared variance. Also, the spatial pattern of the 500-mb geopotential height anomalies in spring associated with the PNA cannot fully capture the wave structure over the Pacific coast of North America identified in this study, i.e. the two anomalous centres over North America associated with the PNA (not shown) are too far east relative to that in Figure 4(d). This indicates that SVD1-SIC is relatively independent physical process in influencing the extrapolar atmospheric circulation.

Table I. Correlations between SVD1-SIC and climate indices in spring (detrended correlations are in parenthesis). Correlations are not statistically significant at 95% confidence level
 PNANPIPDONAO
SVD1-SIC− 0.35 (−0.33)0.25 (0.27)− 0.23 (−0.02)− 0.13 (−0.18)

Recognition of the associations between the spring Arctic SIC and North American summer rainfall might provide us a useful long-term outlook of water resource availability for summer North America. As the decline of the spring sea ice in the North Atlantic and Pacific sectors of the Arctic have begun (Figure 3(a)), which is expected to continue associated with increased greenhouse gases in the atmosphere (e.g. Arzel et al., 2006; Stroeve et al., 2007; Serreze et al., 2007), our results suggest that the available water resource for summer might be reduced in the western United States, the northern Great Plains, Midwest, and westernmost Canada in the 21st century, whereas, the opposite is case in the southern United States, Alaska, and northern Canada.

Additionally, Greenland precipitation variability has important consequences for mass balance of the Greenland ice sheet and associated freshwater discharge into the North Atlantic. The decreased precipitation over central and eastern Greenland associated with SVD1-SIC might enhance the loss of mass there through reducing snow accumulation. Meanwhile, the increased precipitation over western Greenland in conjunction with SVD1-SIC might contribute to an increase rate of summer ice sheet melt there, since summer precipitation is mainly in a form of rain.

Further model simulations forced with the spatial pattern found in SVD1-SIC will be conducted to shed light on the actual magnitude of the extrapolar atmospheric responses to sea ice variability in SVD1-SIC, and detailed dynamic and thermodynamic processes by which sea ice variability in SVD1-SIC influences the extrapolar atmospheric circulation.

Acknowledgements

We thank for the reviewers for their helpful comments. This research was supported by the CAS Hundred Talent Program, National Basic Research Program of China (2011CB309704), NSFC (40876099, 40930848), National Key Technology R&D Program (2008AA121-704), and China Meteorological Administration (GYHY200806006).

Ancillary

Advertisement