-
The research period in this study covers the late summers (from July to August) from 1979 to 2018. The data included the geopotential height and temperature data with spatiotemporal resolutions of 6 h and 1° × 1° from the European Centre for Medium-Range Weather Forecasts Re-Analysis Interim (ERA-Interim, Dee et al., 2011), the 500-hPa geopotential height with a resolution of 1° × 1° in the monthly reanalysis data came from ERA-Interim, the monthly SST data, and the monthly sea ice area fraction data with a horizontal resolution of 1° × 1° was provided by the Hadley Centre (Rayner et al., 2003).
-
In this study, the NCCV during mid-summer is objectively identified using the ERA-Interim geopotential height and temperature data from 1979 to 2018. See details of the identification method in Fang et al. (2021).
-
The NCCV index is defined as the absolute value of the sum of the geopotential height anomaly of the lowest-value grid point within the inner equipotential height line of the cold vortexes in all the NCCV occurrence days in late summer from 1979 to 2018. In this way, the adopted index calculation method was not only able to reflect the number of cold vortex days but also represented the intensities of cold vortex centers.
-
The objective classification method of the NCCV is given in the three steps below.
First, characteristic parameters were collected. According to the identification method of the NCCV, when we read the geopotential height and temperature data for each time into the script, the first time that meets the identification conditions is printed as the starting time of the cold vortex process, and the last time that meets the conditions is the ending time. The center position is defined as the grid point of the lowest geopotential height value within the inner equipotential height line of the cold vortex, while the generation and extinction positions are defined as the center position at the starting and ending time, respectively. Based on the objective identification results of the NCCV, the characteristic parameters for the NCCV processes in late summer are collected, including the latitudes and longitudes of the center positions from the starting times to the ending times.
Second, the trajectory description parameters were calculated. The generation position was collected to represent the source information of the cold vortex. Then, the meridional and zonal mean from the starting to ending times are calculated to describe the location of the cold vortex process. The meridional, zonal, and diagonal variances from starting times to ending times are calculated to describe the moving curvature and distance (e.g., a large diagonal variance corresponds to a straighter motion trajectory and a longer moving distance; see Zheng et al., 2013). The center positions of the cold vortex along its trajectory within the geographical scope of NEC are also collected to represent the relationship between its trajectory and the geographical location of NEC. The above parameters are preliminarily used to describe the trajectory of the NCCV.
Third, the cluster number is determined. The K-means clustering method is applied to the objective classification of the trajectories of cold vortexes. Before this, the value of the classification number, K, should be determined. Based on the dataset of trajectory describing parameters obtained in the previous step, two to seven initial classification numbers are preliminarily determined, and the shape coefficients corresponding to these classification numbers are calculated. The closer the shape coefficient is to 1, the more reasonable the classification number is, and thus an optimal classification number is determined. In this study, a classification number of three is selected. Thus, the NCCV is classified into three types in terms of trajectories.
-
The SST or SIC indices of the key areas were defined according to the regions where the correlation coefficient between the NCCV indexes in late summer and the SSTs or SICs in earlier months passed the 90% confidence level (see section 5). The specific definitions are given in Table 1.
Indexes Definitions North Indian Ocean Dipole Index (NIOD-I) The differences in the average SSTs of the Indian Ocean in April between the area bounded by 46°–76°E and 10°S–24°N and the area bounded by 86°–126°E and 50°–12°S. South Indian Ocean Dipole Index (SIOD-I) The differences in the average SSTs of the Indian Ocean in June between the bounded by of 70°–104°E and 28°S–6°N and the area bounded by 54°–90°E and 48°–30°S. SIC Index of Nansen Basin (NB-I) The average value of the SICs in North of the Nansen Basin in June in the area bounded by
4°W–120°E and 82°–88°N.SIC Index of Barents Sea (BS-I) The average value of the SICs in the Barents Sea in April in the area bounded by 30°–80°E and 70°–82°N. North Atlantic Triple Index (NAT-I) The differences between the SSTs of the North Atlantic Ocean in the average SST in the high latitude areas (60°W–0°, 50°–68°N) and lower latitude areas (80°–20° W, 10°–28° N) and middle-latitude areas in April (70°–36°W, 30°–48°N). Equatorial Pacific Index for Clus-1 (EP1-I) The average value of the SSTs in the equatorial Pacific in April in the area bounded by 150°E–150°W and 16°S–10°N. Northwest Pacific Index (NWP-I) The average value of the SSTs in Northwest Pacific in June in the area bounded by 122°–152°E and 18°–46°N. Equatorial Pacific Index for Clus-2 (EP2-I) The average value of the SSTs in the equatorial Pacific in June in the area bounded by 150°E–170°W and 10°S–26°N. Table 1. Definitions of the SST and SIC indexes.
-
The CAM5.3, as part of the Community Earth System Model version 1.2.2 (CESM1.2.2) modeling framework by the National Center for Atmospheric Research (NCAR), is used in this study to further verify the physical linkages between the previous SST (or SIC) and the late summer 500-hPa geopotential height field. This model has been used in previous studies of climate over East Asia (Huang et al., 2019; Zhang et al., 2020). The finite-volume dynamical core configured with a horizontal resolution of 1.9° latitude × 2.5° longitude (f19_f19) and 30 vertical hybrid levels is selected. More details about this model can be found in Neale et al. (2012).
Nine sets of numerical simulation experiments are conducted to examine the correlations of the SST and SIC in the preceding period with the 500-hPa geopotential height field in late summer. In the control experiment, the monthly climatological data is used as the initial field, while the model integration starts on 1 January of the first year. In the anomaly experiment, the output from 1 January in the sixth year in the control experiment is used as the initial field, with the SST (SIC) anomaly added to the SST (SIC) key areas in the corresponding months for each year. For SST, the sums of the climatic state and twice the regression coefficients calculated by the observed SST regressing onto the SST indexes in key areas were used as the anomalies. For the SIC, the regional averages of sea ice area fraction in the key areas from 1979 to 2018 are calculated. Then, the min and the max regionally-averaged values during these 40 years were selected, and the ice area fractions replaced by the maximum or minimum in the key areas for each grid were used as anomalies. The specific designs are shown in Table 2, and the corresponding SST (SIC) key areas are shown in Table 1. All sets of experiments are continuously integrated into the fifteenth year. During the integration, the SST (SIC) anomaly remains unchanged. The model results from the sixth year to the fifteenth year are selected for significance tests of differences.
Experiments Description SSTc Forced by climatological SSTs. SST1 The North Indian Ocean dipole (NIOD) SSTs anomaly forcing in the Indian Ocean was added in April. SST2 The South Indian Ocean dipole (SIOD) SSTs anomaly forcing in the Indian Ocean was added in June. SICMin1 The ice area fractions in the Nansen Basin (NB) were replaced by the minimum of the regional average in June from 1979 to 2018 (the minimum appeared in 2014). SICMax1 Same as SICMin1 but the ice area fractions were replaced by the maximum (the maximum appeared in 2005). SICMin2 The ice area fractions in the Barents Sea (BS) were replaced by the minimum of the regional average in April from 1979 to 2018 (the minimum appeared in 1995). SICMax2 Same as SICMin2 but the ice area fractions were replaced by the maximum (the maximum appeared in 1979). SSTSIC1 The NIOD SSTs anomaly forcing in the Indian Ocean were added in April, while the ice area fractions in the NB were replaced by the maximum of the regional average in June. SSTSIC2 The SIOD SSTs anomaly forcing in the Indian Ocean were added in June, while the ice area fractions in the BS were replaced by the maximum of the regional average in April. Table 2. The forcing areas and months of SST or SIC anomaly for nine experiments using the CAM5.3 atmospheric model.
Table 2 summarizes the details of all experiments. The difference between the result from the anomaly experiment adopting SST1 (SST2) and that from the control experiment adopting SSTc was regarded as the influence of SST anomaly in the key area on atmospheric circulations at 500 hPa in late summer. The difference between the experimental result adopting SICMax1 (SICMax2) and SICMin1 (SICMin2) was regarded as the influence of SIC anomaly in the key area of atmospheric circulations in the late summer. Moreover, the difference between the results of experiments adopting SSTSIC1 (SSTSIC2) and SICMin1 (SICMin2) was regarded as the combined effect of SST and SIC in the key areas of atmospheric circulations in late summer.
In addition, statistical methods, including linear correlation analysis, partial correlation analysis, regression analysis, and composite analysis methods, are also used. By establishing a regression equation of the NCCV intensity index in late summer with the SST (SIC) index in the key area, the variance ratio between the regression result and the observed NCCV intensity was calculated as the contribution rate of the SST (SIC) in the preceding period to the NCCV intensity in late summer (Shi, 2002).
-
Figure 4 shows the spatial distributions for the correlations of the Clus-1 index with the SSTs in late summer and the preceding period, respectively. It can be seen that there are persistent correlations between the Indian Ocean SST and the Clus-1 index from March through late summer, where the signals appear the strongest in April before gradually weakening. Specifically, significant negative correlations appear over the southeastern Indian Ocean, while significant positive correlations are found over the Arabian Sea and the areas to its south. Statistically, the persistent correlations of this distribution pattern reflect the continuous influence of the SST anomaly of the North Indian Ocean dipole (NIOD) in the preceding spring on Clus-1; that is, the positive (negative) phase of NIOD in April of the preceding period corresponds to stronger (weaker) Clus-1. Therefore, it can be preliminarily inferred that the anomalous NIOD in the preceding April is an SST factor related to the anomalous Clus-1 in its intensity.
Figure 4. Correlation coefficients of the Clus-1 indexes with the SSTs in (a) March, (b) April, (c) May, (d) June, and (e) late summer (the doted areas have passed the significance test at the 90% confidence level). The regions shown in red boxes from left to right are NIOD, EP1, and NAT in (b).
Figure 5 shows the spatial distributions for the correlations of the Clus-2 index with the SSTs from April to the late summer, respectively. It can be seen that the correlations exhibit a dipole pattern in the north-south direction over the Indian Ocean in the preceding period and are at their strongest in June. Specifically, significant negative correlations appear over the equatorial Indian Ocean, and significant positive correlations appear south to 30°S. Statistically, this correlation distribution pattern reflects the influence of the SST anomaly of the South Indian Ocean dipole (SIOD) in the early summer on Clus-2, that is, the positive (negative) phase of SIOD in the preceding June corresponds to a stronger (weaker) Clus-2.
Figure 5. Correlation coefficients of the indexes of Clus-2 with the SSTs in (a) April, (b) May, (c) June, and (d) late summer (the dotted areas have passed the significance test at the 90% confidence level). The regions shown in red boxes from left to right are SIOD, NWP, and EP2 in (c).
To verify whether the key circulation patterns of the NCCV in late summer are related to the SST anomaly in the Indian Ocean during the preceding period through the air-sea interaction, the NIOD index (NIOD-I) in April and the SIOD index (SIOD-I) in June of the preceding period are defined, respectively. The results of the 500-hPa geopotential height field regressed onto the NIOD-I in April and SIOD-I in June are shown in Figs. 6a and 6b. It can be seen that the regression of 500-hPa geopotential height onto the NIOD-I shows significant positive anomalies over Lake Baikal and eastern Siberia and significant negative anomalies to the east of NEC. It also presents an EAP teleconnection wave train along the coasts of East Asia. This distribution pattern is basically consistent with the distribution of the significant correlations between the Clus-1 index and the 500-hPa geopotential height in late summer. It can be preliminarily inferred that the EAP pattern along the coasts of East Asia is related to the NIOD SST anomaly in the preceding April through the air-sea interactions, thus causing the Clus-1 intensity anomaly.
Figure 6. The 500-hPa height field in late summer regressed onto the (a) NIOD-I in the preceding April and (b) SIOD-I in the preceding June. Dotted areas indicate where the regression values have passed the significance test at the 90% confidence level.
Compared with the traditional definition of the IOD (Saji et al., 1999), the positive anomaly area of the west pole, in the definition of the NIOD in this paper, is larger and farther north, while the negative anomaly area of the east pole is farther south. Positive anomalies appear over most of the tropical Indian Ocean and the areas to its north. Previous studies have shown that the SST anomaly in the tropical Indian Ocean is often associated with anomalous anticyclones over the Northwest Pacific and the climate anomalies in East Asia (Chowdary et al., 2011; Wang et al., 2013a, b; Yang et al., 2015b). Since the end of the 1980s, the Hadley and Ferrell cells over the Northern Hemisphere, related to the SST anomaly in the tropical Indian Ocean, have been significantly enhanced, with their activity scopes extending further northwards to middle latitudes. Therefore, the influence of the Hadley and Ferrell cells on the atmospheric circulations over the middle latitudes of Eurasia increases (Han et al., 2018). Zhao et al. (2019) found that the positive SST anomalies of the Indian Ocean in the preceding March and April can affect the western Pacific subtropical high and the precipitation over the NEC by adjusting the atmospheric circulations (such as the vertical motions and Hadley cell). It is also indicated that the SST anomaly in the tropical Indian Ocean has an important forcing effect on the anticyclone in the Western Pacific. The warmer Indian Ocean SST can stimulate a Kelvin wave propagating into the Northwest Pacific, which leads to an anomalous anticyclone near the Philippine Sea and excites the Pacific-Japan (PJ) teleconnection type along the coast of East Asia (Xie et al., 2009; Wu et al., 2010). Based on the conclusions in previous studies, it can be inferred that the NIOD anomaly could adjust the EAP pattern through the eastward-moving Kelvin wave, which further affects the Clus-1 intensity.
Figure 6b shows that the regression of 500-hPa geopotential height onto the SIOD-I exhibits a wave train of a “+ − +” pattern from west to east over the middle and high latitudes of Eurasia. Significant negative anomalies appear over Lake Baikal, and significant positive anomalies appear over the Okhotsk Sea. Nevertheless, the positive anomalies near the Ural Mountains fail the 10% significance test. This distribution pattern is similar to the “Ural-Baikal-Okhotsk” coordination pattern between the Clus-2 intensity index and the 500-hPa geopotential height field in late summer. It can be preliminarily inferred that the “Ural-Baikal-Okhotsk” coordination pattern in late summer is related to the anomalous SIOD SST in the preceding June, which further enhances the anomalous Clus-2 intensity. Previous studies have suggested that the SIOD is closely related to regional air-sea interaction. Not only can it change the distribution of heat flux in the atmosphere over South Asia and the tropical Pacific, but it can also affect the extratropical atmospheric circulation (Behera and Yamagata, 2001). Suzuki et al. (2004) and Yang et al. (2007) pointed out that the changes in the SST of the extratropical southern Indian Ocean in the preceding spring have an important impact on the summer atmospheric circulation and precipitation in NEC. The meridional circulation anomaly caused by the SIOD mode will strengthen the subtropical high in the Northeast Pacific, which is conducive to the maintenance of the anomalous anticyclonic circulation in the South China Sea and the Philippines. This teleconnection is also one of the ways in which SIOD affects the precipitation in eastern China (Xu et al., 2013). The correlation coefficient between the SIOD-I and the SIOD indexes, as defined by Yang and Ding (2007), reaches as high as 0.93. Therefore, it is suggested that the 500-hPa circulation field over East Asia is affected by the SIOD in the preceding June by affecting the meridional circulation and the western Pacific subtropical high, which further causes the intensity anomaly of Clus-2.
It should be noted that, except for the Indian Ocean, the SSTs of equatorial Pacific (EP) and NAT in the preceding April are also significantly correlated with Clus-1, and the SSTs of Northwest Pacific (NWP) and the EP in the preceding June are significantly negatively correlated with Clus-2. To determine the SST factors affecting the NCCV, the partial correlation coefficients between the two types of NCCV indexes and the SST index of each region in the preceding months of April and June and the contribution rates of SST indexes to the NCCV indexes were calculated (Table 3). For each type of NCCV, a partial correlation indicates the correlation between the NCCV and the SST, excluding the influence of the other two SST key areas. For example, the partial correlation between the NIOD-I and NCCV index for clus-1 indicates the correlation excluding the influence of the EP and NAT. It can be seen that the partial correlation coefficients of the NIOD-I in the preceding April and the SIOD-I in the preceding June with the NCCV intensity index are 0.44 and 0.40, respectively, which have passed the significance test at the 90% confidence level, while the partial correlation coefficients between other SST indexes and NCCV intensity indexes fail the 10% significance test. At the same time, the contribution rates of NIOD-I and SIOD-I to the NCCV intensity index are 29.5% and 31.2%, respectively, which are higher than those of other SST indexes. Therefore, the NIOD-I and SIOD-I are selected as the major factors in the preceding period that contribute to the first two types of the NCCV.
Index Month Partial correlation coefficient Variance ratios Clus1 NIOD-I April 0.44 29.5% Clus1 NAT-I April 0.19 11.8% Clus1 EP1-I April 0.05 8.7% Clus2 SIOD-I June 0.40 31.2% Clus2 NWP-I June 0.21 21.8% Clus2 EP2-I June 0.20 19.0% Table 3. The partial correlation coefficients between NCCV indexes for Clus-1 (Clus-2) and SST indexes, excluding the influence of the other two SST indexes for Clus-1 (Clus-2), and the variance ratios between NCCV indexes and SST indexes.
Figure 7 shows the sliding correlation coefficients between each SST index and the NCCV intensity index. It can be seen that the sliding correlation coefficients of the NIOD-I in the preceding April and the SIOD-I in the preceding June with the NCCV intensity index have passed the significance test at the 90% confidence level during the entire period, while the coefficients between other SST indexes and the NCCV intensity index are relatively small; further proving that the Indian Ocean SST in the preceding period is the key factor affecting the anomalous NCCV in late summer.
Figure 7. Panels (a, c, e) show the 21-year sliding correlation coefficients between the Clus-1 index and SST indexes in the preceding April, and panels (b, d, f) show the 21-year sliding correlation coefficients between the Clus-2 index and SST indexes in the preceding June.
It is known that the Arctic SIC plays an important role in regulating the climate change due to its high albedo and its tendency to block the exchanges of heat, water, and momentum between the atmosphere and the ocean (Honda et al., 1999; Rigor et al., 2002; Alexander et al., 2004; Wu et al., 2004). Changes in the SIC will induce changes in various factors, such as surface heat flux, near-surface temperature, and sea level pressure (Alexander et al., 2004). Previous studies show that the Arctic SIC provides a complementary precursor for Chinese summer rainfall variability (Wu et al., 2009). Wu et al. (2013) found that the winter-spring SIC and North Atlantic SST anomalies can cause atmospheric circulation anomalies over northern Eurasia in summer by affecting the atmospheric circulation over southern Newfoundland in spring. He et al. (2018) also indicated that the June SIC variability in the Barents Sea exerts the most significant impact on the East Asian rainfall pattern in August. As an active climatic factor in middle and high latitudes in late summer, a question arises; is the NCCV also affected by the SIC in the preceding period? To resolve this, the correlation coefficients of Clus-1 and Clus-2 with the Arctic SIC in the preceding period are calculated (Fig. 8). It is found that the Clus-1 index is well correlated with the SIC in the preceding June, where there are significant positive correlations in the NB north of the Severnaya Zemlya. The Clus-2 index has a robust correlation with the SIC in the preceding April, where there are significant positive correlations in the Barents Sea, the Kara Sea, and the regions to their north. The results indicate that the NCCV anomaly in late summer may be affected by the Arctic SIC anomaly in the preceding period.
Figure 8. (a) The correlations between the Clus-1 index and the SIC in the preceding June and (b) correlations between the Clus-2 index and the SIC in the preceding April. The region in (a) is the Nansen Basin and (b) is the Barents Sea. The lined areas indicate where the correlations have passed the significance test at 90% confidence level.
Regression analysis was carried out to verify whether the Arctic SIC anomaly in the preceding period caused the anomalous circulation pattern corresponding to the NCCV in late summer. Figures 9a and 9b show the results of the NB index (NB-I) in the preceding June and the BS index (BS-I) in the preceding April, regressed against the 500-hPa geopotential height in late summer, respectively. It can be seen that the regression field of NB-I presents a significant positive anomaly appearing over Lake Baikal, which is consistent with the key circulation pattern during the same period corresponding to Clus-1 (Fig. 3a). However, the meridional wave along East Asia is not as significant as that from Fig. 3a, indicating that the anomalous SICs in the NB in the preceding June may affect the Clus-1 intensity by means of the Lake Baikal high pressure through a wave train in mid-high latitudes. The regression field of BS-I presents significantly positive anomalies over the Ural Mountains and to the south of the Okhotsk Sea, while the negative anomalies over Lake Baikal are not significant. This circulation pattern is basically consistent with the key circulation pattern during the same period corresponding to Clus-2 (Fig. 3b), indicating that the anomalously greater SICs in the preceding April in the Barents Sea are favorable for the formation of “Ural-Baikal-Okhotsk” coordination pattern in late summer, which further affects the Clus-2 intensity. The increase in SIC will result in a decrease in the heat flux between the ocean and the atmosphere, a decrease in the SST, and an increase in the SLP (Alexander et al., 2004). The integrated effects of persistent SIC anomalies from spring to early summer may affect the sub-polar westerlies in summer by changing the albedo and heat flux between the atmosphere and the ocean, thus, further affecting the re-distribution of Rossby waves in Eurasia (Wu et al., 2009). So the anomalously greater SICs in the NB may cause the anomalous Lake Baikal high pressure on its south side along the same longitude. Moreover, the anomalously greater SICs in the BS may cause the anomaly of the Ural high pressure on its south side. The anomalously greater SICs from the spring to early summer are favorable for forming the anomalous high pressure to the south. This pattern could induce the anomalous atmospheric circulations downstream, in middle and high latitudes of East Asia, by affecting the Rossby wave train in the westerlies, further leading to anomalous NCCV intensities.
Figure 9. The 500-hPa geopotential height fields in late summer regressed onto the (a) NB-1 and (b) BS-1 series. Dotted areas indicate where the regressions have passed the significance test at the 90% confidence level.
Figure 10 shows the time series of NCCV indexes for Clus-1 and Clus-2, as well as the corresponding SST and SIC indexes. The NCCV index for Clus-1 shows a consistent interannual variation with NIOD-I and NB-I, featuring correlation coefficients above 0.4, all significant at the 90% confidence level. The NCCV index for Clus-2 is consistent with SIOD-I and BS-I, with correlation coefficients above 0.5, significant at the 90% confidence level. The seven years with high NCCV index values above 0.5 for Clus-1 (1981, 1986, 1995, 2003, 2007, 2010, 2017) all corresponded with positive anomalies of NIOD-I and NB-I, except for 1995. For Clus-2, the eight years with high index values above 0.8 (1979, 1981, 1982, 1987, 1988, 1993, 2000, and 2010) all corresponded with positive anomalies of SIOD-I and BS-I, except for 1988 and 2000. At the same time, only two years (1979, 1990) fail to correspond with the negative anomalies of both NIOD-I and NB-I among the fourteen years (1979, 1983, 1984, 1988, 1989, 1990, 1992, 1997, 2008, 2011, 2012, 2014, 2015, 2018) with low values below –0.75 for Clus-1. The nine years with low NCCV index values below –0.8 for Clus-2 (1991, 1995, 1996, 2001, 2002, 2005, 2007, 2009, 2016) all correspond with the negative anomalies of SIOD-I and BS-I, except for 2001 and 2005. The results suggest a combined effect of SST and SIC on the NCCV in mid-summer.
-
The above observational analyses have revealed the possible effects of the Indian Ocean SST and Arctic SIC in forcing late summer air circulation. In this section, the results obtained by the statistical methods in section 5 will be verified based on the sensitivity tests of CAM5.3. The experimental design has been introduced in detail in section 2.2.4.
Figure 11 shows the composite analysis of the numerical model results. The analysis uses the July-August average of the 500 hPa geopotential height, averaged over the assessment period (years 6 to 15), which is then differenced, as described in the caption. Numerical simulations reveal that the model results separately forced by the Indian Ocean SST or the Arctic SIC are quite different from the observations in Fig. 3 (Figs. 11a–d), while the overall distribution pattern in model results jointly forced by the SST and SIC anomalies is similar to the observations (Figs. 11e-f). In Fig. 10, the regional anomaly forcing for Fig. 11e is that for Fig. 11a, plus that for Fig. 11c, which is similar to Fig. 11f. In the model results (Fig. 11e) jointly forced by the NIOD SST in the preceding April and the NB SIC in the preceding June, there are meridional anomalies from north to south along the coasts of East Asia. Specifically, negative geopotential height anomalies appear over the Okhotsk Sea, accompanied by positive anomalies to its south. This circulation pattern is similar to the EAP pattern in Fig. 3a, except that the high pressure in the Russian Far East is not well simulated. In addition, the positive correlation over Lake Baikal in Fig. 3a is also revealed in the model results. Figures 11a and 11c show a poor performance compared with Fig. 11e, as Fig. 11a fails to present the positive anomaly over Lake Baikal. In model results jointly forced by the SIOD SST in the preceding June and the BS SIC in the preceding April (Fig. 11f), positive anomalies appear over the area near the Ural Mountains and east to the Okhotsk Sea. Negative anomalies appear from Lake Baikal to NEC, which is basically consistent with the “Ural-Baikal-Okhotsk” pattern in Fig. 3b. Figures 11b and 11d show a poor simulated result from Lake Baikal to NEC. Thus, the numerical simulation results in Figs. 11e−f reproduce the observations well, which indicates that the combined effects of SST and SIC in the preceding period can create substantial atmospheric anomalies persisting into late summer and cause anomalous circulations for the first two types of the NCCV in late summer.
Figure 11. Composite analysis for the simulated 500-hPa geopotential height field in late summer. The difference between (a) SST1 and SSTc, (b) SST2 and SSTc, the difference between (c) SICMax1 and SICMin1, (d) SICMax2 and SICMin2, the difference between (e) SSTSIC1 and SICMin1, (f) SSTSIC2 and SICMin2. Dotted areas indicate where the values have passed the significance test at the 90% confidence level.
Notably, the observed results of separated SST and SIC are basically linear relationships, while the numerical results suggest a highly nonlinear interaction. We selected years with positive values above 0.6 and negative values below 0.6 for both the SST and SIC indexes (Table 4). We then calculated the difference between the high-value and low-value years using the July-August averaged 500 hPa geopotential height. The composite results are shown in Fig. 12 and present an EAP-like distribution in East Asia with positive anomalies over Lake Baikal for Clus-1 and a “Ural-Baikal-Okhotsk” pattern in the mid-high latitudes for Clus-2, which suggests that the model results are consistent with the statistical analysis results, that is, the mechanisms for the influence of the SST and SIC on atmospheric circulations of NCCV are nonlinear and complicated.
High-value years Low-value years Clus-1 1982, 2003, 2010 1983, 1984, 2008, 2014 Clus-2 1979, 1981, 1982, 1985 1995, 2007, 2008, 2015, 2016 Table 4. The years with positive values above 0.6 and negative values below 0.6 for both the SST and SIC indexes for Clus-1 and Clus-2.
Indexes | Definitions |
North Indian Ocean Dipole Index (NIOD-I) | The differences in the average SSTs of the Indian Ocean in April between the area bounded by 46°–76°E and 10°S–24°N and the area bounded by 86°–126°E and 50°–12°S. |
South Indian Ocean Dipole Index (SIOD-I) | The differences in the average SSTs of the Indian Ocean in June between the bounded by of 70°–104°E and 28°S–6°N and the area bounded by 54°–90°E and 48°–30°S. |
SIC Index of Nansen Basin (NB-I) | The average value of the SICs in North of the Nansen Basin in June in the area bounded by 4°W–120°E and 82°–88°N. |
SIC Index of Barents Sea (BS-I) | The average value of the SICs in the Barents Sea in April in the area bounded by 30°–80°E and 70°–82°N. |
North Atlantic Triple Index (NAT-I) | The differences between the SSTs of the North Atlantic Ocean in the average SST in the high latitude areas (60°W–0°, 50°–68°N) and lower latitude areas (80°–20° W, 10°–28° N) and middle-latitude areas in April (70°–36°W, 30°–48°N). |
Equatorial Pacific Index for Clus-1 (EP1-I) | The average value of the SSTs in the equatorial Pacific in April in the area bounded by 150°E–150°W and 16°S–10°N. |
Northwest Pacific Index (NWP-I) | The average value of the SSTs in Northwest Pacific in June in the area bounded by 122°–152°E and 18°–46°N. |
Equatorial Pacific Index for Clus-2 (EP2-I) | The average value of the SSTs in the equatorial Pacific in June in the area bounded by 150°E–170°W and 10°S–26°N. |