3.1 Characteristics of the SIOD
Empirical orthogonal function (EOF) decomposition of the seasonal averaged SSTA for boreal winter during 1979–2020 in the Southern Indian Ocean (0°-50°S, 30°-120°E) is performed (Liu and Zhang 2011; Wei 2007). The results demonstrate that the first mode of winter SSTA (variance contribution of 25.1%) exhibits a dipole-type distribution in a southwest-northeast direction (Fig. 2), with a positive southwest polar anomaly centered at 30°-40°S, 55°-80°E and a negative northeast polar anomaly centered at 18°-28°S, 80°-100°E. The two poles are the regions with the most intense SST variability and inverse phase variability. This dipole-type distribution of SSTA in the Southern Indian Ocean dipole mode is also generally consistent with other findings (Morioka et al. 2012; Suzuki et al. 2004; Reason 1999). In this paper, when the SSTA in the subtropical Southern Indian Ocean is positive in the southwest and negative in the northeast, we call it a positive dipole event, and the opposite is a negative dipole event (Suzuki et al. 2004).
3.1.1 Periodicity of the SIOD mode
Wavelet analysis is a common method for analyzing local power variations in a time series. It is feasible to determine the primary modes of variation and how they change over time by decomposing the time series into time-frequency space (Torrence and Compo 1998). A Morlet wavelet transform of the SIOD time series from December 1979 to February 2021 is shown in Fig. 3. Additionally, the 90% confidence interval for red noise is calculated to examine the significance of the wavelet power spectrum. From the wavelet local power spectrum (Fig. 3a), we can see that the lower part of the figure is the high-frequency region, and the upper part is the low-frequency region, corresponding to short- and long-periodic oscillations, respectively. The SIOD fluctuates more significantly on the time scale of 6-7a between 1990 and 2007, while it shows quasi-4-5a short-period oscillations around 1980–1985 and 2015–2019, and quasi-2a oscillations exist from 1994 to 2000.
Figure 3b shows the wavelet variance, which can reflect the fluctuations of the various scales (cycles) contained in the time series and their strength (energy magnitude) with scales (Qi and Chen 2010). Among them, the scale corresponds significantly to the energy serving as the primary cycle component in the provided time series. The dashed line is a reference line for the significance level, which shows that the wavelet variance of the SIOD peaks at 6-7a, and the value passes the \(\alpha =0.1\) significance level test, indicating that the SIOD oscillates and changes mainly at a 6-7a timescale during our research period.
3.1.2 Seasonal evolution of the SIOD mode
The dipole mode is evident on interannual and interdecadal scales (Yan et al. 2009). To investigate the spatial and temporal evolution of the SIOD pattern, we select the main positive and negative SIOD events for composite analysis based on the PC1 time series. The selection criterion for these events is as follows: positive (negative) SIOD events must have a standard deviation (SD) of PC1 larger than 0.6 (smaller than − 0.6) (Pang et al. 2021; Guan et al. 2014). Under this criterion, thirteen positive and eleven negative events are chosen. The specific years of positive (negative) events are 1980, 1981, 1985, 1992, 1993, 1996, 1998, 2000, 2005, 2007, 2010, 2016, 2020 (1979, 1982, 1987, 1990, 1997, 2002, 2009, 2011, 2012, 2015, 2019), noted as PSIOD (NSIOD) in the following. A PSIOD (NSIOD) composite is then defined as the average of these objectively selected positive (negative) events. We are not constrained to the DJF season in constructing the composite events. Instead, these composites span several seasons before and after DJF, so they can form a time series to describe the life cycle of composite events.
Figures 4a-d show the composite of the SSTA and surface wind anomalies in the Southern Indian Ocean during four consecutive seasons (i.e., SON, DJF, MAM, and JJA) for the PSIOD events, respectively. This shows that positive SST anomalies initiate in the southwestern part of the Southern Indian Ocean in boreal autumn (SON), slightly pronounced negative SST anomalies appear between 70°E and 90°E, and cold and warm poles begin to take shape. Positive and negative anomalies reach their maximum in boreal winter (DJF); at the same time, the biggest cold SSTA is associated with the strongest southeast wind anomalies prevailing along the western coast of Australia, which brings cool air from farther south. Cold anomalies also spread northwestward. The warm SSTA is associated with northwesterly wind anomalies prevailing between 30°S and 40°S (Fig. 4b). This mode is still maintained in late spring (MAM), but the center of the northeast region begins to recede significantly, the center of the significant value of the southwest pole anomaly expands, and both SST and sea surface wind anomalies are significantly reduced. In the following season (JJA), SSTA are no longer significant over large ocean areas, and the dipole mode disappears.
Figures 4e-h illustrate the composite of the Southern Indian Ocean SSTA during the consecutive SON, DJF, MAM, and JJA seasons in the NSIOD years. The evolution is similar to that of the PSIOD events, with significant warming in the northeastern Southern Indian Ocean in autumn, followed by continued dipole enhancement, a peak in winter, and the formation of an obvious center of positive and negative dipoles. In MAM, the patterns of the SST and surface wind anomalies persist while their magnitudes are reduced, and then the dipole mode alters and breaks down quite quickly in JJA, with similar changes in the strength of sea surface wind anomalies from boreal autumn to the following summer. In conclusion, although the modes and evolution characteristics of the SSTA in the Southern Indian Ocean are different in PSIOD and NSIOD years, they all exhibit obvious signs of seasonal locking; that is, they occur and develop in boreal autumn, reach their peak in boreal winter, weaken the following spring, and die out the next summer.
3.2 Relationship between the SIOD mode and ENSO
To quantify the relationships between the SIOD mode (i.e., PC1) and ENSO for DJF, the PC1 time series corresponding to EOF1 is shown in Fig. 5 (red curves), together with the Niño3.4 index (SSTAs averaged within 5°S-5°N, 170°W-120°W, blue curves) (Zhang et al. 2014; Zhang et al. 2012). From the PC1 time series, we can see that the SIOD mode oscillates on a 4–6-year time scale, similar to the wavelet analysis results in section 3.1.1. Additionally, the fluctuations of the PC1 time series are consistent with the corresponding Niño3.4 index. The level of reverse phase changes between PC1 and Niño3.4 can be quantified by the correlation coefficient, -0.63, which passes the 95% significance test. Such a substantial correlation implies that the SIOD and ENSO are closely related.
Considering the high correlation between the SIOD and ENSO, we remove ENSO years from the PC1 time series in the subsequent composite analysis to separate ENSO signals and then explore the influence of simple SIOD signals on climate in China. El Niño and La Niña years are selected according to the Oceanic Niño Index (ONI, https://ggweather.com/enso/oni.htm), which is a running 3-month mean SSTA for the Niño3.4 region (i.e., 5°N-5°S, 120°-170°W); weak El Niño and La Niña events are ignored. Among the PSIOD events, four years (1998, 2007, 2010, 2020) are PSIOD and La Niña co-occurrence years, which are denoted as PSIOD + La_Niña. We call the remaining positive SIOD years P_PSIOD, which represents pure positive SIOD years. Similarly, pure negative SIOD years are called as P_NSIOD, and NSIOD + El_Niño represents years when NSIOD events and El Niño events occur simultaneously. The specific years chosen for the following composite analysis are listed in Table 1. The composite using P_PSIOD and P_NSIOD is considered only for the effect of the pure SIOD signal with the ENSO signal removed in the following discussion.
Table 1
| Classification | Year |
| P_PSIOD | 1980, 1981, 1985, 1992, 1993, 1996, 2000, 2005, 2016 |
| PSIOD + La_Niña | 1998, 2007, 2010, 2020 |
| P_NSIOD | 1979, 1990, 2012, 2019 |
| NSIOD + El_Niño | 1982, 1987, 1997, 2002, 2009, 2015 |
3.3 Influence of the pure SIOD on the ITCZ position
It has been shown (Qian and Guan 2007; Wang et al. 2006; Yang et al. 2007) that the SIOD affects precipitation in China by influencing the intensity of the Maskelyne high, circum-Pacific waves, meridional circulation anomalies, and mid- and low-latitude SST anomalies. While the ITCZ is the region of strong convection and maximum rainfall (Philander et al. 1996; Liu et al. 2020), its changes are likely to affect precipitation in the mid-latitudes (Wen et al. 2005); Ge et al. (2000) showed that the location of the ITCZ axis is closely related to precipitation in China. To analyze the effect of the SIOD on the ITCZ position, we use bilinear interpolation to transform the original resolution of the 2.5° × 2.5° precipitation data to 0.5° × 0.5°, and then we select the following Eq. (1) (Adam et al. 2016b) to calculate the latitude of the ITCZ:
$${\phi _{\hbox{max} }}=\frac{{\int_{{20^\circ S}}^{{20^\circ N}} {\phi {{[\cos (\phi )P]}^N}d\phi } }}{{\int_{{20^\circ S}}^{{20^\circ N}} {{{[\cos (\phi )P]}^N}d\phi } }}$$
1
where the independent variable \(\phi\) is the latitude and is the rainfall over the period in the area to be studied; \(N{\text{=}}10\) is considered to reliably identify the precipitation maximum and smooth the grid discretization noise. Figure 6 shows the composite of the precipitation, 850 hPa wind vector field, and ITCZ location under three scenarios: seasonal mean, P_PSIOD, and P_NSIOD in boreal winter and the following spring.
We can see that the ITCZ is a deep convective cloud area near the equator in the climatic state. The convergence of the surface northeast and southeast winds results in the largest amount of precipitation (Nicholson 2018). Compared to the winter seasonally averaged ITCZ axis, the P_PSIOD annual ITCZ position varies more in the land region and less in the Indian Ocean during the Northern Hemisphere winter (Fig. 6a). The largest latitudinal differences are found in the 30°-50°E and 60°-75°E ranges; For the P_NSIOD years, the ITCZ position also changes significantly on land along the southeast coast of Africa, but only slightly at sea. In spring (Fig. 6b), the ITCZ positions in the three cases vary more than in the preceding winter. Within the range of 30°-45°E and 55°-65°E, it can be seen that when pure PSIOD events occur, the ITCZ moves northward, and when pure NSIOD events occur, the ITCZ moves southward. The ITCZ axes are located south of the equator in winter and spring, in good agreement with previous findings (Broccoli et al. 2006; Berry and Reeder 2014).
To further understand the specific magnitude of the impact of SIOD events on the ITCZ in the Indian Ocean, we calculate the average of the ITCZ position over 35°-75°E for the three cases from December to May (Table 2). This shows that the ITCZ is located in the Southern Hemisphere for all three cases during December-February and moves significantly northward in May, whether in SIOD dipole years or in the seasonal climate state. The difference is that the ITCZ of P_PSIOD years fluctuates little from January to March and reaches its southernmost position in March, while the ITCZ of P_NSIOD years goes to its southernmost location in February and then moves significantly northward. From the monthly and seasonal average ITCZ positions, it can be seen that compared with the mean ITCZ position for the 42 years, the ITCZ position moves south in winter for P_PSIOD years and north for P_NSIOD years, and vice versa in spring, i.e., the ITCZ position shifts northward in P_PSIOD years and southward in P_NSIOD years, with the movement being larger in spring.
Table 2
Latitudes of ITCZ axes averaged over 35°-75°E (unit: °N/°S)
| Month | P_PSIOD | P_NSIOD | Mean |
| 12 | -9.31 | -7.12 | -8.11 |
| 1 | -11.31 | -10.63 | -10.91 |
| 2 | -11.31 | -12.69 | -12.11 |
| Ave(Winter) | -10.65 | -10.15 | -10.38 |
| 3 | -11.53 | -11.88 | -11.58 |
| 4 | -6.02 | -9.13 | -7.42 |
| 5 | 1.01 | -0.23 | -0.81 |
| Ave(Spring) | -5.51 | -7.08 | -6.60 |
3.4 Correlation between ITCZ location and precipitation in China
To discuss the relationship between the ITCZ position and precipitation in China, the correlation coefficient is calculated (Fig. 7). It can be seen that the correlation between the location of the Indian Ocean ITCZ and precipitation in China in SIOD years is stronger than in the 42-year seasonal climatology, in both winter and spring (Fig. 7a, d). In the winter of P_PSIOD years (Fig. 7b), the ITCZ position is strongly and positively correlated with precipitation in most areas of China, except for a weak negative correlation in Guangdong Province and central and western Tibet. Furthermore, most of the positive correlation areas pass the 95% confidence test. In the P_NSIOD years (Fig. 7c), the regions with negative correlations are concentrated in Xinjiang Province and a small portion of northern Heilongjiang Province. There are obvious positive correlations in the middle and lower reaches of the Yangtze River, with correlation coefficients up to 0.8 or higher, and most regions also pass the 95% confidence test.
In the spring following P_PSIOD years (Fig. 7e), the correlation changes significantly, and the positive correlation is concentrated in a small portion of the northeast, Guangxi, and neighboring provinces. There is a significant negative correlation between the middle and lower reaches of the Yangtze River. In the P_NSIOD years (Fig. 7f), positive correlations persist in Qinghai, Sichuan, Chongqing, and the southern region, but there are still negative correlations in the middle and lower reaches of the Yangtze River.
3.5 Precipitation anomalies in China in pure SIOD years
Figure 8 shows the geographic distribution of precipitation anomalies in China during the winter and ensuing spring corresponding to P_PSIOD and P_NSIOD years. From Fig. 8a, we can see that the precipitation decreases in most parts of China, including the middle and lower reaches of the Yangtze River, particularly Guangdong, Jiangxi, Fujian, Zhejiang, and Taiwan (passing the 90% significant test), during the winter of P_PSIOD years. The situation in the winter of P_NSIOD years is different (Fig. 8b); precipitation decreases in Guangxi, Guangdong, and Fujian but increases roughly to the north of these regions, especially in the middle and lower reaches of the Yangtze River. In the following spring in P_PSIOD years (Fig. 8c), precipitation is concentrated in a small portion of Guangxi, Yunnan, and Tibet and decreases in most areas, especially in the Yangtze River basin. In P_NSIOD years (Fig. 8d), precipitation increases in East China and the middle and lower reaches of the Yangtze River. In combination with the areas where the SIOD has a great impact on precipitation in China in winter and spring, we will focus on the middle and lower reaches of the Yangtze River for analysis in the following paragraphs.
3.6 Pure SIOD influences on precipitation in the middle and lower reaches of the Yangtze River in China through the Indian Ocean ITCZ and possible reasons
Figure 6 and Table 2 show that the P_PSIOD events cause a southward shift in the Indian Ocean ITCZ during winter, and the Indian Ocean ITCZ position is positively correlated with precipitation in the middle and lower reaches of the Yangtze River; the precipitation in this region then decreases. In contrast, P_PSIOD events the next spring cause the Indian Ocean ITCZ to move northward, and the location of the Indian Ocean ITCZ is negatively correlated with the precipitation in the middle and lower reaches of the Yangtze River. Hence, the rainfall in the region decreases. However, in a pure negative SIOD event, the movement directions of the ITCZ and precipitation are just opposite to those in a pure positive SIOD event. That is, a P_NSIOD event shifts the ITCZ position northward in winter when precipitation increases in the middle and lower reaches of the Yangtze River, which is consistent with the positive correlation between the ITCZ position and rain in the region. Later in spring, the ITCZ shifts southward, and the ITCZ position is negatively correlated with the precipitation in the middle and lower reaches of the Yangtze River, where rainfall increases.
The ITCZ is a zone of intense convection generated by the convergence of trade winds (Qiu et al. 1999; Mischell and Lee 2022). Figure 9 shows the time series (PC) regression distribution of the SIOD to the 850 hPa wind field in winter and spring. Significant and continuous correlation can be seen between the SIOD and the 850 hPa wind, and most areas pass the 90% confidence test. In winter (Fig. 9a), the wind field in the southwestern and northeastern regions of the Southern Indian Ocean exhibits an anti-spatial pattern. A stronger anticyclone in the subtropical Southern Indian Ocean corresponds to the southwest pole of the SIOD mode. A cyclonic circulation corresponding to the northeast pole of the SIOD mode crosses the equator and turns into a westerly flow, which meets the northeast flow over the northwest Pacific in the middle and lower reaches of the Yangtze River and southern China. This may be the reason for the precipitation change in China, particularly in the middle and lower reaches of the Yangtze River. In the ensuing spring (Fig. 9b), the anticyclone and cyclone corresponding to the southwest and northeast poles of the SIOD mode still exist, but unlike in winter, both the cyclone and anticyclone centers move northward, and the cross-equatorial airflow and the northeast airflow over the northwest Pacific still meet in the middle and lower reaches of the Yangtze River and southern China. This also reflects the close connection between the SIOD and precipitation in the middle and lower reaches of the Yangtze River and southern China during winter and the next spring.
To further investigate the specific movement of the ITCZ position under different events in winter and spring and the possible reasons for its impact on precipitation, Fig. 10 shows the wind field and geopotential height anomalies at 500 hPa.
In winter, when only P_PSIOD events occur (Fig. 10a), a high-pressure anomaly in eastern China is enhanced, which may cause the ITCZ to move southward. In addition, cold air from the northern continental region moves southward. Southeastern China is controlled by this airflow, which is not conducive to local water vapor convergence and rise, so precipitation in the middle and lower reaches of the Yangtze River is abnormally low during this period. In the following spring, when only P_PSIOD events take place (Fig. 10c), an anticyclone exists at 0°-20°S, 30°-80°E. The northwest Pacific subtropical high (green curve) withdraws eastward, which makes the ITCZ move northward and be unfavorable to the sustained transport of warm and humid airflow behind the high pressure to the continent; hence the precipitation in the Yangtze River basin is relatively low. When only P_NSIOD events occur in winter (Fig. 10b), two anticyclones exist at 0°-20°S, 40°-50°E, and 30°-60°S, 40°-50°E, corresponding to the increase in two high pressure anomalies. In addition, the southeasterly winds south of the equator strengthen. Together these may cause the ITCZ to move northward. At the same time, the area controlled by the subtropical high in the 500 hPa height field is larger, and the westward extension ridge point is relatively westward. The warm and humid airflow from the Indian Ocean is transported across the Indian Peninsula and Bay of Bengal near 70°E to southern China, where it easily forms precipitation. In the following spring (Fig. 10d), the subtropical high in the western Pacific Ocean strengthens and extends westward, causing the ITCZ to move southward (Guan et al. 2022). The South China Sea is under a subtropical high-pressure system. The warm and humid airflow behind the high pressure is continuously transported to southern China, where it merges with moisture from the Arabian Sea, causing increased rainfall in eastern and southern China, especially in the middle and lower reaches of the Yangtze River and its southern regions.