Journal of Geophysical Research: Atmospheres

Northeast China summer temperature and North Atlantic SST

Authors


Abstract

[1] A previous study revealed a close relationship between interannual variations of northeast China (NEC) summer temperature and a tripole sea surface temperature (SST) anomaly pattern in the North Atlantic in preceding spring. The present study investigates the change in the above relationship and the plausible causes for the change. A tripole SST index is defined with its positive value corresponding to positive SST anomalies in the tropics and midlatitudes and negative SST anomalies in the subtropics. The tripole SST anomaly pattern has a weak correlation with NEC summer temperature during the 1950s through the mid-1970s, in sharp contrast to the 1980s and 1990s. This change is related to the difference in the persistence of the tripole SST pattern. Before the late 1970s, the tripole SST pattern weakened from spring to summer, and thus, the spring North Atlantic tripole SST pattern had a weak connection with NEC summer temperature. On the contrary, after the late 1970s, the tripole SST pattern displayed a tendency of persistence from spring to summer, contributing to circulation changes that affected NEC summer temperature. There are two factors for the persistence of the tripole SST pattern from spring to summer. One is the North Atlantic air-sea interaction, and the other is the persistence of SST anomalies in the eastern equatorial Pacific during the decay of El Niño–Southern Oscillation (ENSO). It is shown that the North Atlantic SST anomalies can have an impact on NEC summer temperature independent of ENSO.

1. Introduction

[2] This is a follow-up study of Wu et al. [2010] who showed that the relationship between El Niño–Southern Oscillation (ENSO) and northeast China (NEC) summer temperature has not been steady. Previous studies have indicated that cold summers in NEC tend to occur in the El Niño developing years, and the NEC summer temperature leans to being warmer than normal in La Niña developing years [Wang and Wu, 1997; Lian and An, 1998; Liu and Wang, 2001; Zhu et al., 2004; Sun and Wang, 2006; Zhu et al., 2007; Guan and Li, 2008]. One process linking ENSO to NEC summer temperature is through changes in the South Asian heating and a midlatitude Asian atmospheric circulation [Wu and Wang, 2002; Wu et al., 2003]. ENSO-induced anomalous heating over South Asia contributes to a zonal circulation pattern over midlatitudes of Asia, leading to anomalous circulation over East Asia and thus affecting climate in NEC. However, NEC summer temperature tended to be higher than normal in the El Niño developing years after the late 1970s [Zhu et al., 2007; Wu et al., 2010], which was in sharp contrast to the relationship before the late 1970s. Wu et al. [2010] linked this interdecadal change to a weakened connection from ENSO to the Indian summer monsoon [e.g., Krishna Kumar et al., 1999; Krishnamurthy and Goswami, 2000; Wang et al., 2001] and a weakened influence of the South Asian heating on midlatitude Asian circulation [Wu, 2002; Wu and Wang, 2002].

[3] Wu et al. [2010] revealed a close relationship between NEC summer temperature and a tripole sea surface temperature (SST) anomaly pattern in the North Atlantic Ocean during the preceding spring in the 1980s and 1990s when the relationship between ENSO and NEC summer temperature was weak. The connection from the North Atlantic tripole SST pattern to NEC summer temperature appears to be through an atmospheric wave pattern that extends from the North Atlantic through Eurasia to East Asia. The present analysis further examines the influences of North Atlantic SST anomalies on NEC summer temperature. One issue to be addressed in the present study is the interdecadal change in the relationship between NEC summer temperature and the tripole SST pattern of the North Atlantic.

[4] It has been indicated that ENSO can induce SST changes in the North Atlantic [e.g., Curtis and Hastenrath, 1995; Enfield and Mayor, 1997; Klein et al., 1999; Saravanan and Chang, 2000; Wu and Zhang, 2010]. The time lag between eastern equatorial Pacific (EEP) SST and tropical North Atlantic SST anomalies is about one to two seasons [e.g., Klein et al., 1999; Wu and Zhang, 2010]. This raises the possibility that ENSO may influence the following summer NEC temperature through the North Atlantic SST change. On the other hand, North Atlantic SST anomalies can be independent of ENSO [Enfield, 1996], being induced by atmospheric circulation changes over the North Atlantic [e.g., Watanabe et al., 1999; Marshall et al., 2001; Czaja and Frankignoul, 2002; Wang et al., 2004; Pan, 2005; Wu et al., 2009]. Thus, another issue that this study will address is whether there are influences of the North Atlantic SST on NEC summer temperature that are independent of ENSO.

[5] In section 2, we describe the data sets and analysis methods applied in this study. Section 3 addresses the interdecadal change in the relationship between NEC summer temperature and the North Atlantic SST. Section 4 investigates the evolution of North Atlantic SST anomalies in relation to ENSO and the North Atlantic air-sea interaction. In section 5, we discuss the impact of the North Atlantic SST anomalies on NEC summer temperature independent of ENSO. The summary and discussions are given in section 6.

2. Data and Methods

[6] This study uses the monthly mean surface air temperature of 160 China stations from 1951 to 2008, provided by the National Climate Center of the China Meteorological Administration. The present study also uses the gridded monthly mean terrestrial air temperature of the University of Delaware, which has a resolution of 0.5° × 0.5° and covers the period 1900–2008 [Matsuura and Willmott, 2009]. The SST used in this study is the NOAA Extended Reconstruction SST, version 3 [Smith et al., 2008], which has a resolution of 2.0° × 2.0° and is available from 1854 to present. This SST data set is provided by NOAA/OAR/ESRL Physical Science Division (PSD), Boulder, Colorado, USA, from its website at http://www.cdc.noaa.gov/.

[7] In addition, this study employs the geopotential heights at 200 hPa, 500 hPa, and 850 hPa and the winds at 850 hPa from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP-NCAR) [Kalnay et al., 1996]. These variables are available on 2.5° × 2.5° grids for the period from 1948 to present. A parallel analysis has been performed using geopotential heights and winds from the 40-year reanalysis of the European Center for Medium-Range Weather Forecasts (ERA40) [Uppala et al., 2005]. The ERA40 data are on 2.5° × 2.5° grids and cover the period from September 1957 to August 2002. The results based on ERA40 are very similar to those based on the NCEP-NCAR reanalysis and thus are not shown.

[8] Monthly mean surface turbulent heat fluxes used in the present study are from an objectively analyzed air-sea flux data set (OAFlux) [Yu and Weller, 2007]. The OAFlux is based on a synthesis of satellite retrievals, ship reports, and atmospheric model reanalysis products using a variational objective analysis. These fluxes are available on 1° × 1° grids. We use the OAFlux surface heat fluxes for the period since 1958. The notation for surface heat fluxes is positive for upward fluxes.

[9] As in the study by Wu et al. [2010], the analysis in this study is focused on interannual variations. For this purpose, a harmonic analysis is applied to time series, and only variations with periods shorter than 8 years have been retained to reconstruct the time series. This excludes the possible contamination of interannual relationship by interdecadal changes and avoids the impact of plausible unrealistic interdecadal variability in the NCEP-NCAR reanalysis [Yang et al., 2002; Inoue and Matsumoto, 2004; Wu et al., 2005; Greatbatch and Rong, 2006]. The reason for the unrealistic interdecadal change in the NCEP-NCAR reanalysis is unclear.

[10] The area-mean June-July-August (JJA) NEC temperature is obtained by averaging the temperature at stations to the east of 120°E and to the north of 42.5°N, following Wu et al. [2010]. The NINO3.4 (5°S–5°N, 170°–120°W) SST in December-January-February (DJF) is used as an index for ENSO. A tripole SST index is constructed as follows: (S* + N*)/2 − M*, where S*, N*, and M* refer to normalized March-April-May (MAM) SST anomalies averaged over (a) 7.5°–17.5°N and 50°–20°W, (b) 37.5°–47.5°N and 60°–30°W, and (c) 20°–30°N and 80°–50°W, respectively. A positive index denotes positive SST anomalies in the tropics (a) and the midlatitudes (b) and negative SST anomalies in the subtropics (c). The selection of these regions is based on the correlation between JJA NEC temperature and MAM SST by Wu et al. [2010, Figure 7d]. The MAM is chosen because it is the season when the tripole pattern in the SST correlation distribution is most pronounced. Note that the locations of these regions are not the same as obtained by regressing SST anomalies onto the North Atlantic Oscillation (NAO) in previous studies [e.g., Marshall et al., 2001]. In this study, our focus is on the SST anomaly pattern that has a strong relationship with the NEC summer temperature anomalies. We follow the same notation for seasons as Wu et al. [2010]. For example, JJA(0) and DJF(1) refer to summer and the following winter, respectively.

3. Interdecadal Change in the Relationship Between NEC Summer Temperature and North Atlantic Spring SST

[11] Figure 1a shows the time series of normalized MAM(0) North Atlantic tripole SST index and JJA(0) NEC temperature interannual anomalies. The NEC temperature and the North Atlantic tripole SST index display both same-sign and opposite-sign anomalies during the period 1951–2008. We use a 0.5 standard deviation to select anomalous years, and present the statistics of the relationship in Table 1. Use of a different criterion leads to changes in the number of years selected, but does not alter the statistical relationship discussed below. During 1951–1977, there are four same-sign years and three opposite-sign years, respectively. During 1978–2008, there are 6 same-sign years and 11 opposite-sign years, respectively.

Figure 1.

(a) Time series of normalized MAM(0) North Atlantic tripole SST index (solid line) and JJA(0) NEC temperature (dashed line); (b) sliding correlation between MAM(0) North Atlantic tripole SST index and JJA(0) NEC temperature displayed at the central year of 21-year window. The plots are based on interannual anomalies.

Table 1. Years When JJA(0) NEC Temperature Anomalies and MAM(0) North Atlantic Tripole SST Index Have the Same Signs and Opposite Signsa
 1951–19771978–2008
  • a

    The bold and bold-italic numbers denote ENSO-related and ENSO-independent opposite-sign years, respectively. Refer to the text for explanation of ENSO-related and ENSO-independent years.

Same-sign years1955, 1957, 1958, 19641988, 1989, 1999, 2000, 2003, 2004
Opposite-sign years1951, 1961, 19761979, 1980, 1981, 1982, 1983, 1992, 1993, 1994, 1998, 2007, 2008

[12] Figure 1b shows the sliding correlation between JJA(0) NEC temperature and MAM(0) tripole SST index with a window of 21 years. Apparently, the correlation is significant and negative during 1973–1992. Weak positive correlation is seen during 1962–1965. The sliding correlation calculated using unfiltered time series displays a similar change but with weaker negative correlation during 1973–1992. The increase in negative correlation during 1969–1973 is mainly due to the opposite-sign anomalies of NEC temperature and tripole SST index during 1979–1983. The decrease in negative correlation around 1993 is due to the replacement of large opposite-sign anomalies during 1982–1983 by large same-sign anomalies during 2003–2004. Following Wu et al. [2010], we choose two epochs, 1955–1975 and 1978–1998, in the contrasting analysis throughout this study. The correlation coefficient between MAM(0) tripole SST index and JJA(0) NEC temperature is 0.25 and −0.72, respectively, for the above two epochs. The change in the correlation appears to concur with an increase in the amplitude of the tripole SST index since the late 1970s (Figure 1a).

[13] The above contrasting feature is further demonstrated in Figure 2, which displays the correlation of JJA(0) China station temperature with MAM(0) tripole SST index for 1955–1975 and 1978–1998, respectively. The patterns of correlation are very different between the two periods. During 1978–1998, significant negative correlation is seen in NEC and north China, and significant positive correlation is observed in the western part of south China (Figure 2b). During 1955–1975, the correlation is positive in NEC, negative in central China, and weak in southern China (Figure 2a). The above contrast of correlation is confirmed by the correlation based on the gridded air temperature from the University of Delaware (figures not shown). During 1978–1998, the negative correlation in NEC extends westward to Mongolia and southward to the Korean Peninsular. During 1955–1975, moderate negative correlation is seen over eastern Russia.

Figure 2.

Correlation of JJA(0) China station temperature with MAM(0) North Atlantic tripole SST index in (a) 1955–1975 and (b) 1978–1998. Contour values are ±0.1, ±0.3, ±0.5, and ±0.7. Shading denotes regions where the absolute values of correlation are larger than 0.433, which is significant at the 95% confidence level. The correlation is calculated based on interannual anomalies.

[14] To understand the connection from the MAM North Atlantic SST to the JJA NEC temperature, we show in Figure 3 the evolution of SST anomalies from DJF(0) to JJA(0) and in Figure 4 the JJA(0) geopotential height anomalies obtained by regression with respect to the MAM(0) tripole SST index for 1955–1975 and 1978–1998, respectively. Also included in Figure 3 are 850 hPa wind anomalies over the North Atlantic to understand the regional atmosphere-ocean relationship. The following descriptions correspond to positive tripole SST index years, but also apply to negative tripole SST index years except for a switch in the sign of anomalies.

Figure 3.

Interannual anomalies of (a, b) DJF(0), (c, d) MAM(0), and (e, f) JJA(0) SST (°C; contours) and 850 hPa winds (vectors, scale at top) in (left) 1955–1975 and (right) 1978–1998 obtained by regression on the MAM(0) North Atlantic tripole SST index. Contour values are ±0.1°C, ±0.2°C, ±0.4°C, ±0.6°C, and ±0.8°C. Shading denotes regions where the absolute values of the corresponding correlation are larger than 0.433, which is significant at the 95% confidence level. Wind vectors are plotted only where the corresponding correlation vectors are longer than 0.433 over the North Atlantic.

Figure 4.

Interannual anomalies of JJA(0) geopotential height (m) at (a, b) 200 hPa, 500 (c, d) hPa, and (e, f) 850 hPa in (left) 1955–1975 and (right) 1978–1998 obtained by regression on the MAM(0) North Atlantic tripole SST index. In Figures 4a and 4b contour values are ±2 m, ±4 m, ±8 m, ±12 m, ±16 m, and ±20 m. In Figures 4c and 4d contour values are ±1 m, ±3 m, ±6 m, ±9 m, ±12 m, and ±15 m. In Figures 4e and 4f contour values are ±1 m, ±2 m, ±4 m, ±6 m, ±8 m, and ±10 m. Shading denotes regions where the absolute values of the corresponding correlation are larger than 0.433, which is significant at the 95% confidence level.

[15] As expected, the North Atlantic tripole SST anomaly pattern is seen in both 1955–1975 and 1978–1998 (Figure 3). However, the temporal evolution of SST anomalies shows pronounced differences between the two periods. During 1955–1975, the tripole SST anomaly pattern appears in DJF(0) and MAM(0) (Figures 3a and 3c), and it is weak in JJA(0) except for the positive anomalies in the tropics (Figure 3e). During 1978–1998, SST anomalies are weak in DJF(0) (Figure 3b) and the tripole SST anomaly pattern is maintained from MAM(0) to JJA(0) (Figures 3d and 3f).

[16] Remarkable differences are also seen in tropical Indo-Pacific SST anomalies. During 1955–1975, SST anomalies are weak in the tropical Indian and Pacific Oceans (Figures 3a, 3c, and 3e). During 1978–1998, the temporal evolution of SST anomalies features a delayed decay of warm ENSO events (Figures 3b, 3d, and 3f). Large positive SST anomalies are seen in EEP. Negative SST anomalies are in the western North and South Pacific. Positive SST anomalies appear in the tropical Indian Ocean and the South China Sea. From DJF(0) to JJA(0), positive SST anomalies decrease in EEP and increase in the tropical Indian Ocean and the South China Sea.

[17] Notable differences are further seen in the North Pacific. During 1955–1975, the SST anomalies in the North Pacific are of relatively small-scale features (Figures 3a, 3c, and 3e). During 1978–1998, negative SST anomalies are seen in the central North Pacific and positive SST anomalies are along the west coast of North America (Figures 3b, 3d, and 3f). In JJA(0), negative SST anomalies extend westward from the central North Pacific to Japan (Figure 3f).

[18] During 1955–1975, weak positive height anomalies are observed over NEC (Figures 4a, 4c, and 4e), consistent with the weak correlation between the tripole SST index and JJA NEC temperature. During 1978–1998, there are large negative height anomalies over NEC (Figures 4b, 4d, and 4f). The anomalous northwesterly winds in the western part of the anomalous cyclone bring relatively cold air from the high latitudes and thus decrease temperature over NEC, which explains the high negative correlation between the tripole SST index and JJA NEC temperature.

[19] The negative height anomalies over NEC during 1978–1998 are connected with two wave patterns. One is a meridional wave pattern along the East Asian coast. This wave pattern is more obvious at lower and middle troposphere (Figures 4d and 4f). It appears to be emanated by anomalous cooling over the western North Pacific induced by the negative SST anomalies in the tropical western North Pacific (Figure 3f) [Nitta, 1987; Huang and Sun, 1992]. This is confirmed by anomalous descent at 500 hPa over the tropical western North Pacific (figure not shown). The other is a wave pattern that extends from the North Atlantic through Eurasia to East Asia (Figures 4b, 4d, and 4f). This wave pattern may include the effects of SST anomalies in both the tropics and the midlatitudes. Notice that over Central America and the Caribbean Sea are negative height anomalies at lower levels (Figure 4f) and positive height anomalies at upper levels (Figure 4b), featuring a Gill-type response to anomalous heating associated with the warm SST anomalies in the tropical North Atlantic (Figure 3f). This is confirmed by the anomalous ascent at 500 hPa over the tropical western North Atlantic and the Caribbean Sea (figure not shown). Further evidence for the contribution of tropical North Atlantic anomalous heating to this wave pattern will be presented later in this section. The midlatitude SST anomalies may contribute to the wave pattern through thermodynamic changes, as evidenced in the overlapping of negative SST and height anomalies (Figures 3f and 4f).

[20] To further understand the relationship between NEC summer temperature and the North Atlantic SST, in Figure 5 we show composite SST anomalies in DJF(0), MAM(0), and JJA(0) and in Figure 6 we show composite height anomalies in JJA(0) at 200 hPa, 500 hPa, and 850 hPa for the cases when JJA(0) NEC temperature and MAM(0) North Atlantic tripole SST index anomalies are opposite (Table 1). Included in Figure 5 are 850 hPa wind anomalies over the North Atlantic. The composite fields are constructed by including all the opposite-sign cases, but with the sign of anomalies reversed when NEC temperature anomalies are negative. This increases the number of cases in the composite, but neglects the difference between the cases of positive- and negative-sign anomalies. Shading in Figures 5 and 6 denotes the magnitude of composite anomalies larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies in the same type of years. The following discussions correspond to positive NEC summer temperature anomalies, but are also applicable to negative NEC temperature anomalies except for a reversal in the sign of anomalies. Note that the composite JJA(0) NEC temperature anomalies are 0.64°C for all the opposite-sign cases.

Figure 5.

Composite interannual anomalies of (a) DJF(0), (b) MAM(0), and (c) JJA(0) SST (°C; contours) and 850 hPa winds (vectors, scale at top) for opposite-sign cases. Contour values are ±0.1°C, ±0.2°C, ±0.4°C, ±0.6°C, and ±0.8°C. Shading denotes regions where absolute anomalies are larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies. Wind vectors are plotted only where their length is longer than 0.7 standard deviations over the North Atlantic.

Figure 6.

Composite interannual anomalies of JJA(0) geopotential height (m) at (a) 200 hPa, (b) 500 hPa, and (c) 850 hPa for opposite-sign cases. In Figure 6a contour values are ±2 m, ±4 m, ±8 m, ±12 m, ±16 m, ±20 m, ±24 m, ±28 m, and ±32 m. In Figure 6b contour values are ±1 m, ±3 m, ±6 m, ±9 m, ±12 m, ±15 m, ±18 m, ±21 m, and ±24 m. In Figure 6c contour values are ±1 m, ±2 m, ±4 m, ±6 m, ±8 m, ±10 m, ±12 m, ±14 m, and ±16 m. Shading denotes regions where absolute anomalies are larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies.

[21] The North Atlantic SST anomalies are small in DJF(0) (Figure 5a). The tripole SST anomaly pattern is obvious in MAM(0) (Figure 5b) and it is maintained from spring to summer, although the magnitude of SST anomalies becomes weaker and the location of these anomalies is displaced (Figure 5c). The evolution of negative SST anomalies in EEP features a decay of La Niña events. This feature implies that the tripole SST pattern tends to follow the La Niña events. The EEP SST anomalies, however, are not particularly significant in DJF(0) (Figure 5a), suggesting that not every tripole SST event in the North Atlantic is preceded by ENSO. In the tropical Indian Ocean, negative SST anomalies develop from MAM(0) to JJA(0), but they are not significant. In the North Pacific, positive SST anomalies extend from the Japan Sea to the western extratropical North Pacific in JJA(0) (Figure 5c). Negative SST anomalies are seen along the west coast of North America in MAM(0) and JJA(0) (Figures 5b and 5c).

[22] There are positive height anomalies over NEC during the opposite-sign cases (Figure 6). These height anomalies connect with a wave train over the North Atlantic and Eurasia, which is probably related to the influence of both tropical and extratropical North Atlantic SST anomalies. Over the western North Pacific and East Asia, there is a meridional wave pattern at the lower and middle troposphere (Figures 6b and 6c). It appears to be related to anomalous upward motion over the tropical western North Pacific (figure not shown) that induces anomalous heating and excites a wave pattern along East Asia [Nitta, 1987; Huang and Sun, 1992]. This wave train also contributes to the height anomalies over NEC.

[23] The connection of atmospheric circulation over the North Atlantic with climate over East Asia has been indicated in previous studies. Through diagnosis of observations and a linear barotropic model, Watanabe [2004] has shown that the NAO signal can extend downstream in February through a wave train along the Asian jet stream when the NAO accompanies the Mediterranean convergence anomaly. As a result, the interannual variability of the NAO is tied to the East Asian climate variability. Wu et al. [2009] have presented evidence that anomalous NAO in spring (April-May) can induce a tripole SST pattern in the North Atlantic that persists into the ensuing summer and excites downstream development of subpolar teleconnection across the northern Eurasia extending to East Asia. Their numerical experiments with a simple general circulation model provide supports for the influence of the North Atlantic SST anomalies on the Eurasian atmospheric circulation and the East Asian summer climate. The results of the present study agree with these previous studies, although the locations of the tripole SST anomaly pattern and the associated atmospheric circulation differ somewhat from previous studies.

[24] To support the connection of the East Asian circulation with the tropical North Atlantic anomalous heating, we display in Figure 7a the wave activity fluxes [Plumb, 1985; Karoly et al., 1989; Yang and Gutowski, 1994] associated with the composite 500 hPa geopotential height anomalies. The wave activity fluxes signify the propagation of stationary wave activity. The divergence region of wave activity fluxes indicates a source for stationary wave activity. Also shown in Figure 7a are composite 500 hPa vertical p-velocity anomalies in the domain of 5°N–35°N and 120°W–0°W. Anomalous descent is observed over tropical North Atlantic and the Caribbean Sea, which is consistent with local negative SST anomalies (Figure 5c). The subtropical North Atlantic appears to be a source region of wave activity fluxes that extend northeastward to West Europe and then southeastward to East Asia. The wave activity flux source region extends to the Caribbean Sea at 200 hPa (not shown). This suggests that the contribution of the tropical North Atlantic SST anomaly induced anomalous heating to the East Asian circulation change.

Figure 7.

(a) Wave activity fluxes (vectors, scale at the top-right) associated with composite interannual anomalies of JJA(0) 500 hPa geopotential height and composite interannual anomalies of JJA(0) 500 hPa vertical p-velocity (contours, interval of 0.2 × 10−2 Pa s−1 with zero contours suppressed; shading for magnitude of anomalies exceeding 0.7 standard deviations; shown only for the domain of 5°N–35°N, 120°W–0°W) for opposite-sign cases. (b) Barotropic model height perturbation difference (contour interval of 50 m) averaged over days 31–40 and imposed idealized convergence and divergence anomaly difference (shaded; contour interval 3 × 10−6 s−1) for JJA climatological mean divergence.

[25] The role of tropical North Atlantic heating in East Asian circulation change is further demonstrated by results of experiments with a barotropic model [Sardeshmukh and Hoskins, 1988] in response to idealized convergence and divergence anomalies. The barotropic model used is spectral with the truncation at rhomboidal wave number 40. For our purpose, divergence and convergence anomalies are prescribed over the tropical North Atlantic and the Caribbean Sea with a maximum intensity of 7 × 10−6 s−1 at 15°N and 60°W. The location for those anomalies is based on vertical motion anomalies in Figure 7a. We have performed two experiments: one with climatological summer mean divergence plus convergence anomalies and the other with climatological summer mean divergence plus divergence anomalies. The model is integrated for 40 days. Figure 7b shows the difference of response between the two experiments (i.e., imposed convergence minus imposed divergence) averaged over model days 31–40 with thick contours in the shaded region denoting the difference of imposed divergence and convergence anomalies. A wave pattern of height anomalies appears over the North Atlantic and Eurasia, which bears similarity to the observational composite (Figure 6a). In particular, positive height anomalies are seen over NEC in both the model and observations (Figures 7b and 6a). However, the location of anomalies over Eurasia displays differences from observations. Albeit this discrepancy, the model results support the impact of tropical North Atlantic heating on the East Asian circulation. The smaller height anomalies over NEC in the model compared with the observational composite indicate an important contribution of other factors, such as anomalous heating over the tropical western North Pacific and thermodynamic effects of midlatitude North Atlantic SST anomalies.

[26] From above analyses, the apparent link between spring North Atlantic tripole SST pattern and summer NEC temperature during 1980s and 1990s is due to the persistence of the SST anomaly pattern in the North Atlantic and the influence of summer North Atlantic SST anomalies on atmospheric circulation. During the 1950s and 1960s, the lack of persistent North Atlantic SST anomalies leads to a weak link between spring North Atlantic SST and summer NEC temperature.

4. Evolution of the North Atlantic SST Anomalies

[27] The development of the tripole SST anomaly pattern may be related to changes in atmospheric circulation over the North Atlantic, as suggested by previous studies [e.g., Watanabe et al., 1999; Marshall et al., 2001; Czaja and Frankignoul, 2002; Wang et al., 2004; Pan, 2005; Wu et al., 2009]. The atmospheric changes can affect SST through various processes including surface heat flux and oceanic advection [e.g., Cayan, 1992; Visbeck et al., 2003], as supported by the relationship between lower-level wind and SST anomalies shown in Figure 3. During 1955–1975, lower-level wind anomalies feature an anticyclone over the midlatitudes and a cyclone over the subtropics in DJF(0) and MAM(0) (Figures 3a and 3c). The anomalous easterlies over the midlatitudes may contribute to warm SST anomalies through reducing surface evaporation and inducing northward Ekman transport [Marshall et al., 2001; Visbeck et al., 2003]. Similar wind anomalies are observed during 1978–1998 (Figures 3b and 3d). The tripole SST pattern appears earlier in 1955–1975 than in 1978–1998, likely because of the influence of wind anomalies in the preceding fall. During 1955–1975, the midlatitude North Atlantic wind anomalies change from anticyclonic in AM(0) to cyclonic in JJA(0) (Figures 3c and 3e). During 1978–1998, the anomalous midlatitude anticyclone shifts from west in MAM(0) to east in JJA(0) (Figures 3d and 3f).

[28] A similar wind-SST relationship over the North Atlantic is seen on Figure 5. Negative SST anomalies develop in the tropics following enhanced easterlies, and those in the midlatitudes are preceded by anomalous westerlies. Warm SST anomalies appear in the subtropics under anomalous anticyclonic winds. The persistence of the tripole SST anomaly pattern from MAM(0) to JJA(0) appears to be consistent with the maintaining (though with reduced magnitude) of anomalous westerlies over the midlatitudes and anomalous easterlies over the tropics.

[29] As noted above, the evolution of the tripole SST pattern is different between 1955–1975 and 1978–1998. In the former period, the tripole SST pattern decays in JJA(0), whereas in the latter period the tripole SST pattern is maintained to JJA(0). Then, what are the causes for this difference? One plausible factor is the tropical Pacific SST anomaly. Previous studies have shown that ENSO exerts an impact on the tropical North Atlantic SST [e.g., Curtis and Hastenrath, 1995; Enfield and Mayor, 1997; Klein et al., 1999; Saravanan and Chang, 2000; Wu and Zhang, 2010]. The tropical North Atlantic SST warms up in spring following the mature phase of El Niño events due to a change in the ENSO-induced surface heat fluxes. In turn, these SST anomalies may be conveyed to the midlatitudes through atmospheric circulation and consequent surface heat flux and oceanic changes. The ENSO effects, however, experienced a change around the late 1970s.

[30] Figure 8 compares the lag-lead correlation of monthly NINO3.4 SST with MAM(0) tripole SST index for 1955–1975 and 1978–1998, respectively. During 1978–1998, the correlation is positive and significant before May and weakens and switches to negative after June, suggesting that the tripole SST pattern develops during the decay of warm ENSO events. During 1955–1975, the correlation is positive but weak during the entire 2-year span, suggesting a weak link between ENSO and the tripole SST pattern. This change seems to be related to the change in the ENSO decay time [Lee et al., 2008; Wu and Zhang, 2010]. Before the late 1970s, ENSO decays quickly after the mature phase. As such, its influence on the tropical North Atlantic SST is weak. After the late 1970s, ENSO shows larger persistence, leading to tropical North Atlantic warming (or cooling) in spring. Wu et al. [2009] attributed the persistence of the tripole SST pattern from spring to summer to oceanic memory. The lack of persistence of the tripole SST pattern from spring to summer during 1955–1975 indicates the role of factors other than the oceanic memory, such as the lack of remote ENSO forcing.

Figure 8.

Lag-lead correlation of monthly NINO3.4 SST with MAM(0) North Atlantic tripole SST index for (solid line) 1955–1975 and (dashed line) 1978–1998.

[31] Another plausible factor for the different persistence of the tripole SST pattern is the change in atmospheric circulation over the North Atlantic. During 1955–1975, wind anomalies over the midlatitude North Atlantic in JJA(0) are opposite to those in MAM(0) (Figures 3c and 3e). Anomalous winds switch from easterly in MAM(0) to westerly in JJA(0), which may lead to a switch in anomalous surface evaporation and oceanic Ekman transport. This feature is consistent with the quick decrease of positive SST anomalies. During 1978–1998, anomalous easterly winds are maintained in JJA(0) (Figure 3f), and so do anomalous surface evaporation and oceanic Ekman transport, favoring the persistence of positive SST anomalies. The switch of wind anomalies from MAM(0) to JJA(0) in 1955–1975 and the maintenance of wind anomalies in 1978–1998 may be attributed to the switch or maintenance of SST anomalies in the midlatitudes, which leads to an argument for the role of the North Atlantic air-sea interaction.

[32] To confirm the role of surface evaporation in the North Atlantic SST change, we show in Figure 9 composite anomalies of surface latent heat flux in February-March (FM(0)), May-June (MJ(0)), and July-August (JA(0)) for the cases when JJA(0) NEC temperature and MAM(0) North Atlantic tripole SST index anomalies are opposite (Table 1). The latent heat flux anomaly pattern over the North Atlantic in FM(0) resembles closely the tripole SST anomaly pattern in MAM(0) (Figure 9a versus Figure 5b). Positive latent heat flux anomalies are observed over the subtropics and midlatitudes where anomalous winds are in the same direction as mean winds (Figures 5a and 5b). Negative latent heat flux anomalies are seen over the Gulf of Mexico and subtropical western North Atlantic, likely being related to the advection of warm air from the south that reduces the sea-air temperature and humidity differences. The results support the contribution of anomalous North Atlantic circulation to the development of the tripole SST anomaly pattern through wind-evaporation effects, consistent with previous studies [Klein et al., 1999; Marshall et al., 2001; Visbeck et al., 2003; Wu and Zhang, 2010]. The latent heat flux distribution shows a pronounced change from FM(0) to MJ(0). In MJ(0), positive latent heat flux anomalies are seen over western subtropics (Figure 9b) where SST anomalies are positive in MAM(0) (Figure 5b). This signifies the SST forcing of surface latent heat flux. On the other hand, these anomalous fluxes contribute to the weakening of positive SST anomalies in the subtropics (Figure 5c). Over most of the tropics and midlatitudes, negative latent heat flux anomalies appear over regions of negative SST anomalies. These latent heat flux anomalies are relatively small, which favors the maintenance of SST anomalies from spring to summer (Figures 5b and 5c). In JA(0), negative latent heat flux anomalies are seen over the tropical North Atlantic (Figure 9c) with negative SST anomalies (Figure 5c), which in turn may weaken the SST anomalies.

Figure 9.

Composite interannual anomalies of (a) FM(0), (b) MJ(0), and (c) JA(0) surface latent heat flux anomalies (W m−2) for opposite-sign cases. Contours values are ±2 W m−2, ±4 W m−2, ±8 W m−2, ±12 W m−2, ±16 W m−2, ±20 W m−2. Shading denotes regions where absolute anomalies are larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies.

[33] From the above analyses, atmospheric wind changes induce the tripole SST anomaly pattern that matures in spring and early summer. The summer SST anomalies in turn lead to latent heat flux anomalies likely through modulating sea-air humidity difference and force atmospheric circulation changes. These heat flux anomalies act to feed negatively back on summer SST. Thus, the observed summer SST anomalies also include the impact of atmospheric feedback. As such, the relationship of summer SST with summer atmospheric circulation has a mixed feature of both SST forcing of atmosphere and the atmospheric feedback on SST and thus it appears to be weaker than that of spring tripole SST pattern.

5. Impact of the North Atlantic SST Independent of ENSO

[34] Analyses in the preceding sections indicate that the North Atlantic tripole SST pattern is linked to NEC summer temperature anomalies through atmospheric teleconnection. On the other hand, ENSO can contribute to the North Atlantic SST change [Klein et al., 1999; Wu and Zhang, 2010] as well as the East Asian summer climate [Wu et al., 2003]. Thus, it may be hypothesized that the relationship between NEC summer temperature and spring North Atlantic tripole SST anomaly pattern arises because both of them are induced by ENSO. For this hypothesis to be valid, there should be a connection between the preceding winter EEP SST and the spring North Atlantic tripole SST pattern as well as a connection between the preceding winter EEP SST and the NEC summer temperature.

[35] Indeed, MAM(0) tripole SST index displays a significant positive correlation with the preceding winter and spring NINO3.4 SST in the 1980s and 1990s, but not in the 1950s and 1960s (Figure 8). This correlation is also seen in the sliding correlation with a 21-year window between DJF(0) NINO3.4 SST and MAM(0) tripole SST index in Figure 10. However, the correlation between DJF(0) NINO3.4 SST and JJA(0) NEC temperature is well below the 95% confidence level during most of the period, except in the 1990s when a moderate negative correlation is observed (Figure 10). This correlation invalidates the hypothesis that the tripole SST pattern-NEC temperature connection is due to the influence of the preceding ENSO on both of them. This correlation, however, suggests the possibility that the preceding ENSO may affect NEC summer temperature through the North Atlantic SST change. Thus, a question arises as to whether there is an independent influence of the North Atlantic SST on NEC summer temperature.

Figure 10.

Sliding correlation between (solid line) DJF(0) NINO3.4 SST and JJA(0) NEC temperature, (dashed line) between DJF(0) NINO3.4 SST and MAM(0) North Atlantic tripole SST index, and (dotted line) between MAM(0) North Atlantic tripole SST index and JJA(0) NEC temperature displayed at the central year of the 21-year window. The correlation is calculated based on interannual anomalies.

[36] To address this question, we have examined year-to-year variations of DJF(0) NINO3.4 SST anomalies for the opposite-sign cases (Table 1). For the 14 cases during the analysis period, there are 5 cases (1983, 1992, 1994, 1998, and 2008) for which positive (negative) MAM(0) tripole SST index is preceded by positive (negative) DJF(0) NINO3.4 SST anomaly. For these cases, the development and persistence of the tripole SST pattern may be contributed by the impact of ENSO on tropical North Atlantic circulation and surface heat fluxes [Curtis and Hastenrath, 1995; Enfield and Mayor, 1997; Klein et al., 1999; Wu and Zhang, 2010]. These cases are denoted as ENSO-related cases. The composite JJA(0) NEC temperature anomalies for these cases are 0.77°C. There are five cases (1961, 1979, 1980, 1982, and 1993) for which the DJF(0) NINO3.4 SST anomalies are small. For these cases, the development and persistence of the tripole SST pattern is independent of ENSO. These cases are denoted as ENSO-independent cases. The composite JJA(0) NEC temperature anomalies for these cases are 0.56°C. In the remaining four cases (1951, 1976, 1981, and 2007), positive (negative) tripole SST index is preceded by negative (positive) DJF(0) NINO3.4 SST anomaly.

[37] Here, we compare the ENSO-related and ENSO-independent cases. Figures 1113 show composite anomalies of SST, surface latent heat flux, and height for the two types of cases, respectively. In constructing the composite patterns, we have reversed the sign of anomalies for the years when NEC summer temperature anomalies are negative. The shading in Figures 1113 denotes composite anomalies larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies in the same type of years.

Figure 11.

Composite interannual anomalies of (a, b) DJF(0), (c, d) MAM(0), and (e, f) JJA(0) SST (°C; contours) and 850 hPa winds (vectors, scale at top) for (left) ENSO-independent and (right) ENSO-related opposite-sign cases. Contour values are ±0.1°C, ±0.2°C, ±0.4°C, ±0.6°C, ±0.8°C, ±1.0°C, ±1.2°C, and ±1.4°C. Shading denotes regions where absolute anomalies are larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies. Wind vectors are plotted only where their length is longer than 0.7 standard deviations over the North Atlantic.

Figure 12.

Composite interannual anomalies of (a, b) FM(0), (c, d) MJ(0), and (e, f) JA(0) surface latent heat flux anomalies (W m−2) for (left) ENSO-independent and (right) ENSO-related opposite-sign cases. Contours values are ±2 W m−2, ±4 W m−2, ±8 W m−2, ±12 W m−2, ±16 W m−2, ±20 W m−2. Shading denotes regions where absolute anomalies are larger than 0.7 standard deviations of individual anomalies with respect to composite anomalies.

Figure 13.

Composite interannual anomalies of JJA(0) geopotential height (m) at (a, b) 200 hPa, (c, d) 500 hPa, and (e, f) 850 hPa for (left) ENSO-independent and (right) ENSO-related opposite-sign cases. In Figure 13a contour values are ±2 m, ±4 m, ±8 m, ±12 m, ±16 m, ±20 m, ±24 m, ±28 m, and ±32 m, In Figure 13b contour values are ±1 m, ±3 m, ±6 m, ±9 m, ±12 m, ±15 m, ±18 m, ±21 m, and ±24 m. In Figure 13c contour values are ±1 m, ±2 m, ±4 m, ±6 m, ±8 m, ±10 m, ±12 m, ±14 m, and ±16 m. Shading denotes regions where absolute anomalies are larger than 0.7.

[38] In both cases, SST anomalies in the North Atlantic are weak in DJF(0) (Figures 11a and 11b). A tripole SST anomaly pattern develops in MAM(0) (Figures 11c and 11d) and is maintained in JJA(0) (Figures 11e and 11f). The development of the tripole SST anomaly pattern is related to wind anomalies. In DJF(0), there are cyclonic and anticyclonic wind anomalies over the high- and midlatitudes, respectively. Westerly anomalies over the midlatitudes induce negative SST anomalies through enhancing upward latent heat fluxes (Figures 12a and 12b) and cold oceanic advection. Easterly anomalies over the tropics enhance surface evaporation (Figures 12a and 12b), contributing to negative SST anomalies (Figures 11c and 11d). Southeasterly anomalies over the Gulf of Mexico and western subtropical North Atlantic bring warmer air from the lower latitudes, reducing the sea-air temperature and humidity differences. This process suppresses surface latent heat fluxes (Figures 12a and 12b), contributing to warm SST anomalies in the subtropics (Figures 11c and 11d). The surface latent heat flux anomaly pattern displays a pronounced change from FM(0) to MJ(0) (Figures 12a12d). Surface latent heat flux anomalies in JA(0) (Figures 12e and 12f) appear to follow SST anomalies in many regions, signifying the SST forcing of surface latent heat flux. This in turn leads to the weakening of SST anomalies in summer in these regions.

[39] In the ENSO-related cases, large SST anomalies in the tropical Pacific (Figures 11b, 11d, and 11f) can force atmospheric circulation changes in the tropical North Atlantic and induce surface heat flux anomalies [Curtis and Hastenrath, 1995; Enfield and Mayor, 1997; Klein et al., 1999]. Thus, the development and persistence of the tripole SST anomaly pattern may be largely contributed by the impact of ENSO. In the ENSO-independent cases, tropical Pacific SST anomalies are weak (Figures 11a, 11c, and 11e), confirming that ENSO is not a factor for the development of the tripole SST anomaly pattern.

[40] In the ENSO-related cases, there are negative SST anomalies in the tropical Indian Ocean and the South China Sea (Figures 11b, 11d, and 11f). Positive SST anomalies develop in the central North Pacific in DJF(0) and MAM(0). Negative SST anomalies are seen along the west coast of North America in MAM(0) and JJA(0). In both the ENSO-related and ENSO-independent cases, a band of positive SST anomalies extends from the Japan Sea to the western North Pacific in JJA(0) (Figures 11e and 11f), which could be related to the eastward extension of warmer surface air temperature from NEC that favors ocean surface warming [Wu et al., 2010].

[41] In the ENSO-independent cases, a wave pattern of height anomalies extends from the North Atlantic through Eurasia to East Asia (Figures 13a, 13c, and 13e). Anomalous descent at 500 hPa is present over the tropical North Atlantic (figure not shown), suggesting that the emanation of the above wave pattern may be linked to anomalous cooling over the tropical North Atlantic in association with local negative SST anomalies (Figure 11e). In the ENSO-related cases, negative height anomalies span the entire tropics at 200 hPa and 500 hPa (Figures 13d and 13f), featuring a response to negative SST anomalies in EEP [Pan and Oort, 1983; Newell and Wu, 1992; Yulaeva and Wallace, 1994; Sobel et al., 2002; Chiang and Sobel, 2002]. A meridional wave pattern is observed over the western North Pacific and East Asia (Figures 13b, 13d, and 13f). This wave pattern appears to be emanated by anomalous heating over the Philippine Sea, as indicated by anomalous ascent over the western North Pacific (figure not shown). A meridional wave pattern is seen over the North Atlantic, which is associated with anomalous descent over the tropical North Atlantic (figure not shown) in relation to negative SST anomalies. The wave pattern, however, does not extend to East Asia, which differs from the ENSO-independent cases.

6. Summary and Discussions

[42] The present study has documented the interdecadal change in the relationship between NEC summer temperature and the North Atlantic SST around the late 1970s. It is found that the connection of the North Atlantic tripole SST anomaly pattern to NEC summer temperature is weak before the late 1970s, which is in sharp contrast to the feature in the 1980s and 1990s when the NEC summer temperature is closely related to the tripole SST pattern. Obvious difference is also seen in the evolution of the tripole SST pattern. Before the late 1970s, the persistence of the tripole SST pattern from spring to summer is weak. As such, the connection of spring tripole SST pattern with summer atmospheric circulation and NEC temperature is weak. After the late 1970s the tripole SST pattern is maintained from spring to summer. Because of the SST persistence and the impact of summer SST anomalies on atmospheric circulation, the spring tripole SST pattern displays a close relation with the summer atmospheric circulation and NEC temperature, and thus it appears to be a precursory signal for summer NEC temperature anomalies.

[43] Two factors may have contributed to the difference in the persistence of the tripole SST pattern. One is the North Atlantic air-sea interaction, as evidenced by a coherent change from spring to summer in both SST and wind anomalies. The other is the SST anomalies in EEP. Before the late 1970s, ENSO usually decays quickly, and thus ENSO-induced tropical North Atlantic warming is small. After the late 1970s, ENSO decays slower and the SST anomalies in EEP are maintained through spring. This may contribute to the tropical North Atlantic warming and thus favor the maintenance of the tripole SST pattern.

[44] While the development and persistence of the tripole SST pattern are contributed by the atmospheric circulation anomalies, there is evidence signifying the forcing of summer SST anomalies on the atmospheric circulation. Thus, air-sea interactions are involved in the evolution of the North Atlantic SST and wind anomalies, and the relative importance of atmospheric forcing of SST and SST forcing of atmosphere appear to be season-dependent in the North Atlantic. The observed summer North Atlantic SST anomalies may have also included negative feedback from the atmosphere (e.g., surface latent heat flux). As such, the spring North Atlantic tripole SST pattern appears to have a stronger correlation with the summer NEC temperature than the summer SST pattern. This time lag has an implication for seasonal forecasting of summer NEC temperature that is an important factor for the crop yield in NEC [Ding, 1980; Sun et al., 1983].

[45] Summarizing the results of the present analysis and Wu et al. [2010], before the late 1970s the concurrent impact of developing ENSO on NEC summer temperature is more robust, whereas after the late 1970s the impact of the North Atlantic SST is more important. The development of the tripole SST anomaly pattern can be related to the decaying ENSO in some cases. The tripole SST anomaly pattern can also develop by air-sea interaction processes over the North Atlantic, and thus provide an impact on NEC summer temperature that is independent of ENSO.

[46] The present study focuses on the impact of ENSO and the North Atlantic SST anomalies on NEC summer temperature. It is possible that SST anomalies in other regions may play a role for NEC summer temperature variability. There are notable SST anomalies in the North Pacific (Figures 3, 5, and 11) that may be a source for the NEC summer temperature variability. For example, Yoon and Yeh [2010] indicated that the North Pacific SST anomalies associated with the Pacific Decadal Oscillation could modulate the relationship between El Niño and the East Asian summer monsoon through changes in extratropical atmospheric circulation, such as the Eurasian pattern. The EEP SST anomalies of the preceding winter may also influence NEC summer temperature through other paths, for example, through their connection with the East Asian winter monsoon [Wang et al., 2007, 2008] and the link between the East Asian winter and summer monsoons [Chen et al., 2000]. This speculation needs further investigation for validation.

Acknowledgments

[47] Comments of three anonymous reviewers have led to a significant improvement of this paper. The authors acknowledge the support of the National High Technology Research and Development Program of China (2006AA12Z207) and HKSAR RGC Project 447807. This work was partly accomplished in COLA and RW acknowledges the support by grants from the NSF (ATM-0830068), NOAA (NA09OAR4310058 and NA09OAR4310186), and NASA (NNX09AN50G). S.L., L.S., Y.L., and Z.G. acknowledge the support by the Governor's Foundation of Jilin Province, China. Z.G. acknowledges the support by R&D Special Fund for Public Welfare Industry (Meteorology) (GYHY201106015). Y. L. acknowledges the support by R&D Special Fund for Public Welfare Industry (Meteorology) (GYHY201106016) and National Natural Science Foundation of China for Youth Science Foundation (40705036).