Figure 1 shows the composite V200 and U200 anomalies for the positive and negative JMDI cases (hereafter referred to as "JMDI+" and "JMDI-" cases), respectively. The V200 anomalies in the JMDI+ cases (Fig. 1a) are characterized by a well-organized wave-like pattern along the Asian jet, from West Asia to East Asia. The amplitudes of these alternate anomalies are roughly the same for each cell, which are all above 3 m s-1. The wave-like pattern along the Asian jet agrees with the SRP-related anomalies, which are represented by the red contours, and coincide well with the anomalies related to the SRP in previous studies (e.g., Lu et al., 2002; Yasui and Watanabe, 2010). By contrast, there is no clear wave-like pattern along the Asian jet for the JMDI- cases (Fig. 1b). The V200 anomalies tend to be scattered to the north and south of the Asian jet, but are weak along the jet. These anomalies differ remarkably from the SRP-related anomalies (also shown as red contours in Fig. 1b), even showing opposite signs over the Caspian Sea and Central Asia. Therefore, the relationship between the JMD and SRP is close for positive JMDI years but vague for negative years.
On the other hand, the composite U200 anomalies for the JMDI+ and JMDI- cases also show evident distinctions. For the JMDI+ cases (Fig. 1c), the most prominent feature of the U200 anomalies is a north-south seesaw pattern, with westerly and easterly anomalies on the two sides of the climatological mean jet axis, respectively. Consistent with previous studies (Du et al., 2016; Hong and Lu, 2016), this seesaw pattern appears much stronger over West Asia and East Asia, with the amplitudes of the centers around these regions being above 7 m s-1. Meanwhile, the anomalies over Central Asia are quite weak, as indicated by the strength being only about 2 m s-1. (Hong and Lu, 2016) suggested that the JMD can be further promoted by the SRP, manifested by intensified wind anomalies over West Asia and East Asia but weakened anomalies over Central Asia. Therefore, the inhomogeneous U200 anomalies in the zonal direction for the JMDI+ cases (Fig. 1c) can also be viewed as a manifestation of a close JMD-SRP relationship. The U200 anomalies in the JMDI- cases (Fig. 1d) also exhibit a seesaw pattern, similar to that for the JMDI+ cases. However, these anomalies are much weaker and tend to shift westward.
To check that the abovementioned asymmetry in the JMD-SRP relationship is not merely the result of a particular index used, we adopt Upc1 as another index for the JMD. This index is similar to the JMDI, with the correlation coefficient between them being as high as 0.96, but the two indexes differ appreciably, as explained later. Figure 2 shows the composite V200 and U200 anomalies for the positive and negative Upc1 cases (hereafter referred to as "Upc1+" and "Upc1-" cases), respectively. It is clear that for the Upc1+ cases (Figs. 2a and c), there is a well-organized wave-like pattern in V200 anomalies along the Asian jet, which nearly overlaps with the anomalies related to the SRP (red contours in Fig. 2a), and the north-south seesaw pattern of U200 anomalies is stronger over West Asia and East Asia. These anomalies show great resemblance to those for the JMDI+ cases (Figs. 1a and c), verifying the robustness of the JMD-SRP relationship in northward JMD years. By comparison, the anomalies for the Upc1- cases (Figs. 2b and d) show much weaker intensity compared to those for the Upc1+ cases. Hardly any significant V200 anomalies exist along the Asian jet over the western Eurasian continent, deviating greatly from those related to the SRP (red contours in Fig. 2b), and the intensities of the U200 anomalies are roughly only around half those for the Upc1+ cases.
Figures 1 and 2 indicate that when different indexes are used for the JMD, the related anomalies are similar for the northward JMD but quite different for the southward JMD cases. The reason for this similarity and difference is that the cases selected by the two indexes are similar for the northward JMD cases but quite different for the southward JMD cases, as shown in Fig. 3 and Table 1. Almost all cases (13 of 14) are the same for the northward JMD cases, but 10 cases in total are distinct for the southward JMD cases. Specifically, 1960, 1968, 1983, 1988 and 2003 are included in the JMDI- cases but are absent in the Upc1- cases; while 1965, 1974, 1991, 2002 and 2012 are selected for the Upc1- cases but are not chosen for the JMDI- cases.
In addition, the anomalies tend to be more spatially coherent for the northward JMD cases than the southward cases. The pattern correlation coefficients in the U200 anomalies within the domain (20°-60°N, 0°-150°E) between individual cases and the corresponding composites (e.g., positive cases with the positive composite and negative cases with the negative composite) are generally greater for the positive cases than the negative cases (Table 2).
Scatterplots of standardized JMDI and Upc1 with the SRPI, as shown in Fig. 4, better illustrate the asymmetry in the JMD-SRP relationship. In general, there is a positive relationship between the JMD and SRP: the SRPI is positive (negative) in most positive (negative) JMDI/Upc1 years. This positive relationship is confirmed by the correlation coefficient between the JMDI (Upc1) and SRPI, which is 0.40 (0.54) and significant at the 0.01 level. However, Fig. 4 also indicates an interesting detail of this positive relationship: with increment of the JMDI, the SRPI increases obviously when the JMDI is positive but tends to decrease when the JMDI is negative (Fig. 4a). This can be verified by the correlation coefficients between the two indexes, which drop dramatically from 0.60 for the positive JMDI years to -0.16 for the negative JMDI years. The correlation coefficient is negative, although weak, for the negative JMDI years. The scatterplot of Upc1 and SRPI (Fig. 4b) shows quite similar features to Fig. 4a, with contrasting correlation coefficients of 0.62 and -0.04 for positive and negative Upc1 years, respectively.
All the above results on the asymmetry in the JMD-SRP relationship are obtained from the viewpoint of JMD indexes. If this relationship is viewed from the perspective of positive and negative SRPI years, the asymmetry can also be found. The correlation coefficient between JMDI (Upc1) and SRPI is 0.55 (0.59) and significant at the 0.01 level in positive SRPI years, but is only 0.04/0.20 in negative SRPI years. For brevity, we only show the results based mainly on the JMD indexes in this paper.