In recent years, there is more information available to the public regarding atmospheric teleconnection activity including a recently discovered phenomenon called the Asian–Pacific Oscillation (APO). Published studies describing this teleconnection from different research organisations are on the increase. There are 15 papers alone regarding this topic collected from the proposer as a first or corresponding author (see the reference list). Zhao et al. (2007, hereafter referred to as Z07) in their earliest paper on the subject proposed a seesaw pattern for ‘eddy temperature’ variations in the upper troposphere between Asia and the North Pacific. They called this pattern the APO, and they found that the index based on this teleconnection varies coherently with monsoon activity, subtropical jet movement and location, as well as other important circulation systems (Liu et al., 2011a, 2011b; Wang et al., 2012). They described the APO as an appropriate index for describing the evolution of the circulation over Asia and the North Pacific area. However, there are some issues with the current understanding of APO theory. For example, does the APO index as defined play a role in representing the general circulation locally or more broadly?
The APO index was constructed by using the difference between the mean summer season zonal temperature anomalies in the upper troposphere (500–200 hPa) within the region defined by 15–50°N, 60–120°E and the same variable in the region 15–50°N, 180–120°W. The result was a dipole pattern (Fig. 1). However, the definition of this index may be extended from previously published work. The spatial pattern in the first component of the set of empirical orthogonal functions (EOF1, see Fig. 2) does not support an exclusively Pacific Region Basin seesaw as shown in Fig. 1. The two figures show different information although the correlation between the APO index and the time series of EOF1 was strong (0.93). Many smaller-scale centres distributed throughout Fig. 2 roughly coincide with the characteristic topography of the North Hemisphere. This implies that consideration of the APO as a dipole pattern germane to the Pacific Ocean Basin alone (Fig. 1) is limiting since previous authors used Pacific Region centres of EOF1 (Fig. 2) to confirm the existence of APO. In their analysis, Z07 neglected a large part of the Northern Hemisphere (NH) (0–60°N, 120°W −30°E in Fig. 1), but which was included here (Fig. 2). The area beyond the Pacific should be considered in order to explore whether an additional action centre exists. Its existence could yield new and useful information about the general circulation. This paper aims to reexamine the APO phenomenon as thoroughly as possible. The investigation begins with a reassessment of the current understanding of APO theory. Note that instead of a complete review of the collected papers themselves, we would rather focus on a detailed discussion of a number of key points regarding the APO phenomenon including a proposal for an improved and expanded index.
In Section 2, the data used here is provided. A statistical analysis regarding the role of the Atlantic Region action centre is carried out in Section 3. A discussion of EOF1 behaviour in different sectors is carried out, and we revisit the use of geopotential height for the APO index in Section 4. A detailed discussion about the limitations of the APO definition is addressed in Section 5, and these issues arise from the analysis from earlier sections. Finally, Section 6 provides a discussion that highlights the main results of this work. We also arranged an appendix table (Table A1) to summarise indices and the domains corresponding to the different acronyms.
We used the European Center for Medium-Range Weather Forecast (ECMWF) reanalysis (ERA) monthly mean data set (ERA-40, see Gibson et al., 1997; Uppala et al., 2005), which was the same data set used by other APO researchers. The primary variables accessed here are the temperature (K) and the geopotential height at the 200–500 hPa levels over the entire NH during the summer period (JJA) from 1958 to 2001. The National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) global atmospheric reanalysis dataset (Kalnay et al., 1996) was used for comparison to the ERA derived results. The monthly mean Arctic oscillation (AO) index was used in this study as well, and the values were provided by Climate Prediction Center (CPC). The North Atlantic Oscillation (NAO) index was derived from the time series of the leading EOF of sea level pressure (SLP) anomalies over the Atlantic sector (Hurrell, 1995).
Figure 3 shows the same basic information as portrayed in Fig. 1, except including the complementary area included in Fig. 2. Note that there is a warm and cool zonal temperature anomaly over Asia and North Pacific, respectively. There is an additional cool anomaly (30°N, 52.5°W), however, centred over the North Atlantic Ocean, roughly coinciding with the location of the cool region in Fig. 2. This temperature anomaly over the North Atlantic has the same order of magnitude as the Pacific Region centres. In fact, the Atlantic Region anomaly appears to be stronger than the other two when the NCEP/NCAR global reanalysis data were used to produce Fig. 3 (not shown). Note that in this analysis, the warm and cool temperature anomalies are congruent with high and low pressure regions in the NH general circulation at all levels since, on the time scale used, the upper troposphere is largely equivalent barotropic (Hurrell et al., 1998).
The analysis above implies that the Atlantic Region centre identified above is equally important to the other two centres in order to explain the summer season circulation regime using a hemisphere-wide teleconnection pattern (Figs. 2 and 3). Also, this indicates teleconnectivity between the Atlantic and Pacific Ocean basins especially during the summer season, which cannot be demonstrated using a regional index such as the Pacific North American (PNA) Index. Teleconnection activity during the summer between these two ocean basins has been hinted at in earlier studies of the general circulation over North America (Namias, 1982, 1983). These papers found that the temperature anomalies over the Pacific and Atlantic were often of similar sign during the summer season. Additionally, this extended APO temperature pattern matches what would be expected since the mean upper tropospheric, summer season, NH temperature climatology is more zonal in structure (Hurrell et al., 1998; Fig. 1.5) when compared to the winter pattern. The winter season pattern is opposite to the APO centres, showing a trough over East Asia and ridges over the oceanic region. It is likely that the NH land–ocean distribution may be partially responsible for the APO, and this would argue for an extended APO.
In order to investigate whether the Atlantic Region cool anomaly or low centre is closely associated with the other two action centres, we propose a new index as described here. Based on the information displayed in Fig. 3, we chose the region 15–50°N, 0–60°W to define the boundaries for the additional action centre representing the Atlantic Region anomaly. Thus, an Asian–Pacific–Atlantic (APA) index is presented as follows:
where T′ is the zonal temperature anomaly averaged over the 500–200 hPa layer. Here, we call the first and second terms of the right side of eq. (1) the Asian Component (AC) and the Pacific–Atlantic Component (P–AtC), respectively. The latter has two subcomponents, that is, the Pacific Component (PC) and the Atlantic Component (AtC) (terms 2a and 2b, respectively). Note that the AC was assigned as twice the weight of the other two subcomponents in order to make their weighted sum equal to zero following the approach of Wallace and Gutzler (1981). Table 1 shows the relationships among them. Note that all of the correlation coefficients exceeded the 99.9 % confidence level except for that between the PC and AtC. Here we adopt 0.479 as a critical value beyond which a correlation is regarded as highly significant (significance at the 99.9 % confidence level1 ). The correlation coefficient between the AC and P–AtC was −0.877, which is higher than that between AC and PC (−0.78) separately, and the latter are the APO patterns proposed by Z07. The out-of-phase relationship between the AC and AtC was statistically significant as well (r=−0.61). The correlation coefficient between the PC and AtC was 0.28, and this approaches statistical significance at the 95 % confidence level (0.297). Table 2 indicates that the NH EOF1 was highly correlated with the APO, APA, and their components. In particular, the correlation coefficient between the NH EOF1 time series and the APA index was 0.905, which is similar to that between NHs EOF1 and APO index (0.93, Table 2). Additionally, the correlation between Asian and Pacific–Atlantic Region teleconnection centres is more significant than those used to construct the APO. The APA index is defined as a tripole pattern, and should be more informative in a statistical and physical sense for circulations covering the entire NH than previously published APO formulations. This does not preclude the use of the APO for Pacific Region only general circulation studies.
The action centres in Fig. 2, however, were located at higher latitudes than those in Figs. 1 or 3. This implies that regions which best match with EOF1 centres may not be the same regions indexed by the APO or APA (Fig. 2). In addition, the correlations among the values derived from the respective action centres in Fig. 3 were much lower than those among the area averaged values of the action centres derived at higher latitudes. This suggests that the APO or APA may not be as strong as the teleconnection patterns found in Wallace and Gutzler (1981). In order to find an index that more closely matched with EOF1, we performed many tests by changing the positions or size of the boxed areas on the locations of the three action centres in Fig. 3. Since there are many plausible choices for such a purpose, here we only show one example, that is, a new APA index as follows:
This new APA index (nAPA) was constructed by covering more area, especially by extending the boxes into higher latitudes than did the APO or APA index. Similarly, the first, second, and the third terms on the right side of the eq. (2) can be defined as a new Asian Component (nAC), new Pacific Component (nPC) and new Atlantic Component (nAtC), respectively. We can give the second and third terms a combined name as in eq. (1), or the new Pacific–Atlantic Component (nP–AtC).
The correlation coefficient between the new APA index and the time series of EOF1 in the NH is now 0.944, which is higher than the 0.93 found for the APO index (Z07). The relationship among each term in eq. (2) as shown in Table 3 indicated that the constituent terms for the nAPA index appear to be more closely interdependent than those of the APA or APO. The out-of-phase relationship between nAC and nAtC or between nAC and nP–AtC became much more significant than the corresponding components for the APO or APA indexes, highlighting the role and importance of the Atlantic Region action centre. The relationship between nPC and nAtC is also stronger since the correlation reached 0.642 (significant at greater than 99.9 %). Again, this suggests that upper level temperature over the two oceans tends to vary in phase. If we follow the explanation proposed by Z07, a new Asian–Atlantic oscillation composed of nAC and nAtC (r=−0.86) or a new Asian–Pacific–Atlantic oscillation composed of nAC and nP–AtC (r=−0.90) would be more acceptable than the APO (r=−0.78).
Here, we show the results of other tests in this study of the APO phenomenon, and two prominent calculations are introduced in detail below.
Since previous studies emphasised the importance of EOF1, it is necessary to investigate how different the EOF1 calculations are when the size of the domain changes. Wang et al. (2010) pointed out that changing the domain in a geopotential field could generate quite different responses in EOF analysis especially for a regional teleconnection pattern such as the Okhotsk–Japan teleconnection. The question to be answered here is: how appropriate is the use of the hemisphere-wide EOF1 time series in order to represent the regional-scale APO phenomenon in the temperature field?
Figure 4a–c shows the spatial patterns of EOF1 using the domain as in Fig. 1, including the Pacific–Atlantic Region, and an extended Pacific–Atlantic Region, respectively. Comparing Fig. 4 with Fig. 2 demonstrates that the information displayed in Fig. 4 was almost unchanged in each panel except for some slight drift in the position of the action centres. This implies that the EOF1 distribution basically reflected a NH pattern since the response was not sensitive to the domain change. The correlation coefficients in Table T0004 and 5 T0005 provide more evidence supporting this point. The correlations among the time series of EOF1 in their respective regions far exceed the 99.9 % confidence level (Table 4). In other words, any EOF1 analysis in one these domains would show the same result as using one across the entire NH according to previous studies of the APO theory.
Interestingly, the time series correlation between e-P–At (extended area in the Pacific and Atlantic) and P–At regions was 0.957, whereas that between NH and e-P–At regions was slightly higher at 0.962. This suggests that including a larger domain (approaching that of the entire NH) would be key in strengthening the regional calculations. The values in Table 5 show that the EOF1 in every domain were highly correlated with every index we produced here. In order to investigate the correlation between the regional EOF1 and NH EOF1 in detail, we produced correlations in bands of 30 degrees longitude over the domain of 60° latitude ×180° longitude as shown in Fig. 5. All of correlation coefficients exceeded 0.85, far above the value needed for confidence at the 99.9 % level. Three peaks in the correlation coefficient greater than 0.95 were found in the bands bordered by 60°E–90°W, 180–30°E and 60°W–150°E, respectively. The highest correlation was 0.981 in the 60°W–150°E band. Thus, Fig. 5 and Table 5 suggest that an index for the Atlantic–Asian region may best be reflected by the NH APA, although the patterns in the Asian–Pacific and Pacific–Atlantic regions were also equally robust.
There have been only brief comparisons between temperature and other important meteorological variables in the previously published APO literature. These comparisons are necessary in order to determine if the APO corresponds to teleconnectivity as derived by Wallace and Gutzler (1981). Wallace and Gutzler (1981) used the 500 hPa geopotential height in their teleconnection indexes. Also Z07 suggested that geopotential height could be used to represent the APO, but they only looked at the Pacific Region and did not construct the EOF1 time series based on this data.
The geopotential height is used here instead of temperature as shown in Fig. 6a and b. There were large differences between Fig. 3 and 6a, although a weak high and two weak low centres were located in Asia, Pacific and Atlantic oceans, respectively. However, other strong centres appeared (e.g. North America – high; Fig. 6a). An APO or APA-like structure was nearly non-existent when compared with the pattern appearing in Fig. 3. Additionally, no APO or APA-like signal can be found in Fig. 6b, which is very different from that shown in Fig. 2. It is especially evident that there was a dipole-like pattern occurring in the Atlantic Region in Fig. 6a, which might imply a different dynamic mechanism from the eddy-temperature-based APA.
The correlation coefficient between the eddy height in 60–120°E, 15–50°N and in 180–120°W was 0.378, which does exceed the 95 % confidence level. The correlation between the NH EOF1 time series and the eddy height anomaly for the two regions (H-APO), however, was not statistically significant (0.127) at all. The NH EOF1 height fields demonstrate a relatively good relationship with the eddy temperature APO index and the NH EOF1 temperature (Table 6). However, the AO was significantly correlated negatively with APO (r=−0.318) although the AO has no significant correlation with EOF1 (Table 7).
Zhou and Wang (2015) showed a significant correlation between the NAO index and an APO index they constructed. However, there was no significant correlation between NAO index and the APO indexes as shown in Z07 and here. The strength of the relationship between APO and indexes such as NAO or AO could be adjusted subjectively as needed. However, researchers have argued that while the NAO is limited to the North Atlantic, it is closely linked to the broader hemispheric AO pattern. This does not imply that a study based on the NAO alone necessarily leads to less valuable or useful results than one using the AO alone. However, the relationship between AO and NAO is very different from that of the APO and APA. Both of the former indexes can be expressed in terms of the SLP or geopotential height (Thompson and Wallace, 2000), whereas both the latter indexes can be explained in the upper tropospheric and hemispheric-wide eddy temperature EOF1. However, note that NAO does not belong to an EOF 1 in the NH scale, which is different from AO and APO-like indices.
Thus, it is difficult to sufficiently compare the two pairs of indexes. Using eddy temperature might have limitations in expressing teleconnection patterns when compared with geopotential height as discussed here. Note also that, unlike in the temperature field, higher correlations between the values obtained by using a large enough domain for the NH EOF1 time series does not occur in the geopotential field. A weaker result here might be expected since the geopotential height is a function of both temperature and pressure although the atmospheric state appeared to be quasi-barotropic (see Section 2). Lastly, note that the difference between the Zhou and Wang (2015) study and this study here is that we included a discussion using the AO index for comparison. Also, in this study we investigated the relationship between the AO and APO indexes in detail. This analysis indicates that the AO potentially has a stronger connection to the APO since both phenomena can be explained via the first principle component of the respective meteorological elements in the NH scale. More studies regarding this issue are desirable in future.
In this analysis, an anomalous low centre over the North Atlantic Region with same order of magnitude was shown to be as important as the other two centres (Fig. 3). This low centre coincides with a coherent low over the North Atlantic in the spatial pattern of EOF1 (Fig. 2) based on which Z07 documented the APO (Fig. 1). However, the APO index as proposed formerly is a more regional index or phenomenon (Fig. 1), whereas our analysis includes the Atlantic Ocean Basin (Fig. 2). Previous research did not consider the linkage to the Atlantic Region, but Figs. 2 and 3 show that this linkage is an important part of the teleconnection phenomenon. As discussed in Section 4, our analysis proposes an extension of the previously published methodologies for expressing the APO.
Additionally, here we obtained higher correlations among the terms of an APA index constructed by changing the calculation to include three action centres of greater area (Fig. 3) than those using a more regional APO index.
The larger geographical areas included here for the index components produced even higher correlations between the indexes and the EOF1 time series. Since EOF1 accounted for only 0.21 of the contribution to the total NH variance, this result suggests that the APO cannot act as a teleconnection pattern significantly independent of any other pattern constructed using random selection and EOF1 alone. Thus, it is demonstrated in previous sections that an approach used which includes more centres of action is more physically meaningful.
Although Z07 did not originally include the Atlantic Region low, it was a surprise that this low centre was included in a later publication without any reasonable explanation (see Fig. 3a in Zhao et al., 2010). Additionally, their study only used data within the 300–200 hPa layer, which is inconsistent with the original formulation of the APO (500–200 hPa layer). Instead, APO studies that followed (e.g. Zhao et al., 2009, 2010, 2011a, 2012a) did not assign great importance to, or use, the Atlantic Region lows. Rather, we consider this feature as an important accessory that needs further research.
It is likely that the importance of the Atlantic Region centre is the result of Rossby Wave propagation. Wallace and Gutzler (1981) pointed out that connections between the flow regime over the Pacific and North America have been quantified for a long time in the literature going back a least (now) 75 years when Allen et al. (1940) noted a negative correlation between SLP in the Aleutian region and the western USA. Later, Lorenz (1951) and Namias (1951) noted that there were strong correlations between the North Pacific and Eastern USA flow regimes. Additionally, Jiang and Lau (2008) demonstrated a summer season Pacific Region wave train showing centres of action between the Bering Sea region and North America. Rossby Wave propagation may explain the higher correlations shown here when the calculation domains were extended poleward (Section 3). In fact, there was strong wave activity flux crossing the North American continent, which was linked to strong positive/negative phases of the APA (not shown).
Zhao et al. (2010) then postulated that the APO index could be formulated using data obtained from 300 to 200 hPa layer only, which was then demonstrated subsequently in Zhao et al. (2011a). This does provide for a more consistent definition of the APO index. This change in definition may not be convincing as it now does not include a large portion of the upper and middle troposphere. Here, it is recommended that examining the 500–200 hPa level for all three centres produces a better statistical result and it includes a significant portion of the upper airflow. Also, the 300–200 hPa layer could be located largely within the stratosphere in regions of low pressure during the winter season (Hurrell et al., 1998). The revised APO index using the proposed new equation should be renamed, however, since the action centres now include an Atlantic centre of action within NH (Fig. 2). The APA name has been proposed here. Additionally, using the definition proposed in this work, one can more readily identify an anomalous hemisphere-wide general circulation change that might be related to teleconnections between the Pacific and Atlantic Ocean basins.
The new index that was constructed is better correlated with EOF1 as well and provides for a better physical explanation in describing hemispheric-scale patterns than previous APO formulations. Small differences in highly significant correlations, however, may not be a sufficient reason for generating a new index. Additionally, we have pointed out that the eddy temperature has weaknesses in identifying teleconnection patterns compared with using geopotential height or pressure.
Some uncertainty and inconsistency in published APO theory is summarised here:
Therefore, the conclusion that two subregions belong to one pattern by using correlation would be unreliable especially in the temperature field. It is apparent that two patterns with qualitative differences (see Figs. 1 and 2) might represent the same atmospheric processes physically as long as there is a high correlation between them. Additionally, there was no reasonable explanation in previously published work justifying the use of the eddy temperature as the variable for the analysis of the APO. This technique was used typically to express wave train propagation in the atmosphere (Wang et al., 2010). However, it is difficult to find APO-related temperature wave propagation in previous studies and our results. In fact, the temperature anomalies alone from the time mean field can also produce a similar picture to Fig. 1. The eddy technique should not be necessary.
We have found some fundamental differences between the previously published APO work and our analysis here based on physical properties. For example, Fig. 7b in Zhao et al. (2012b) could be interpreted to show that there was westward-propagating wave activity over Eurasian continent. However, the westward wave propagation does not occur in a location where a westerly jet is climatically located over the continent, according to the wave theory proposed by Hoskins and Karoly (1981). The simulation and prediction of the APO pattern using models (Zhao et al., 2010; Chen et al., 2013; Zhou and Zhao, 2013; Liu et al., 2015) could be difficult because there may be so many problems in the APO definition, the APO meaning, and the impacts of the APO as shown in our analysis here. Similarly, some composite analyses presented in previous work cannot be explained readily through a physical mechanism.
Also, Zhao et al. (2010) indicated that their Fig. 12b shows a structure associated with zonal wave number two. However, their work only shows the Pacific Ocean Basin, and one might find five or more irregular centres when examining the NH flow regime. This does not necessarily indicate a wave train structure. Additionally, previous studies have provided no explanation of the mechanism for driving the APO. T0008
Thus, we believe that a complete reexamination of the current theory regarding the physical mechanism of APO-caused climate anomalies would be necessary because of the mentioned problems above. The new index that included the Atlantic Region cool anomaly definitely expresses a more hemispheric-scale teleconnection than the old index. Although the use of the eddy temperature may largely limit the practical application of all the indexes, the first step for the further study of the APO-like phenomenon should be to construct an index with a rigorous or standard definition. An ideal reexamination would include information based on (1) including the Atlantic low, and (2) overcoming the limitation of using eddy temperature. A large-scale dynamics-based analysis that supports our claim would be beyond the scope of this study.
Using the results presented here, the out-of-phase between the AC and PC or AtC, might be a by-product of the large-scale scope of thermodynamic equilibrium among continents and oceans in the NH. However, any APO index does not describe accurately the land–sea equilibrium because the temperature centres in Fig. 2 did not precisely coincide with the NH land–sea distribution. Since there are so many possible problems in defining the APO, we would like to suggest new avenues for APO study:
We must acknowledge that the previous studies proposed a thought-provoking question, that is, how can we explain a teleconnection in the temperature field for so many indexes that were highly correlated with each other. In order to improve our understanding of this phenomenon, further research must develop a more rigorous research plan in overcoming the defects mentioned here. If these are overcome, this will be beneficial to climate research.
81It is the test for correlation coefficient. In this case, the number of samples was 44 and the degrees of freedom were 42. Thus, if the coefficient is larger than 0.479, we can say that the result is significant at the 99.9 % confidence level.
This study was supported by National Natural Science Foundation of China (No. 41575055 and 41375091) and the project of the Ministry of Science and Technology of China (No. 2012CB417202). The authors also thank the anonymous reviewer whose comments were helpful in making this manuscript a stronger contribution. The values of the AO and the NAO data using in this study can be found at www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_ao_index/ao.shtml and www.climatedataguide.ucar.edu/climate-data/hurrell-north-atlantic-oscillation-nao-index-pc-based respectively.
GibsonJ. K., KållbergP. W., UppalaS., HernandezA., NomuraA., co-authors. ERA Description. 1997. ECMWF re-analysis project report series 1, pp. 72. Online at: www.ecmwf.int/publications.
HurrellJ. W., van LoonH., SheaD. J. KarolyD. J., VincentD. G. Meteorology of the Southern Hemisphere. Meteorological Monographs, 49. 1998. Boston, MA: American Meteorological Society. NCAR/CAS/95-08, 1–46 pp.
WangY., JianZ., ZhaoP. Extratropical modulation on Asian summer monsoon at precessional bands. Geophys. Res. Lett. 2012; 39(14): doi: http://dx.doi.org/10.1029/2012GL052553.
ZhaoP., YangS., WangH., ZhangQ. Interdecadal relationships between the Asian–Pacific Oscillation and summer climate anomalies over Asia, North Pacific, and North America during a recent 100 years. J. Clim. 2011a; 24: 4793–4799.
ZhouB., WangZ. On the significance of the interannual relationship between the Asian-Pacific Oscillation and the North Atlantic Oscillation, J. Geophys. Res. Atmos. 2015; 120: 6489–6499. doi: http://dx.doi.org/10.1002/2015JD023328.