Impact of the southern annular mode on extreme changes in indian rainfall during the early 1990s

The variability in rainfall amounts in India draws much attention because it strongly influences the country’s ecological and social systems. Indian rainfall is associated with climate factors, including El Niño/Southern Oscillation and the Indian Ocean Dipole. Here we identified the Southern Annular Mode (SAM), the primary pattern of climate variability in the Southern Hemisphere, as the ultimate forcing leading to decadal changes in Indian rainfall. Through statistical analyses using observational data covering the period from 1979 to 2015, we show an increase in the decadal rainfall amount in the early 1990s over the Indian region. Examining atmospheric environmental conditions, we demonstrate that conditions have become more favorable over the past few decades. Specifically, during the positive SAM phase since the early 1990s, changes in the atmospheric fields have evoked anomalous vertical motion over the continent and the Indian Ocean, enhancing the southerly cross-equatorial flow by increased land–sea thermal contrast, thereby increasing decadal rainfall in the region.


Decadal variability in Indian rainfall
shows the annual average rainfall and low-level winds at 850 hPa in boreal summer (June-September) in the Asian monsoon region from 1979 to 2015. The prevailing horizontal wind velocity clearly changes direction to southwesterly in the Northern Hemisphere from southeasterly in the Southern Hemisphere around the tropical Indian Ocean. The rainfall pattern of the Indian region tends to be inhomogeneous and is generally heavier on land than in the ocean.  Fig. 1a, according to Li et al. 16 ). The 37-year time series reveals apparent extreme circumstances alternating between rainy and arid periods. More specifically, the number of extreme rain events has increased considerably over the past few decades, which agrees with previous findings that these events are happening with increasing frequency in the Indian region during monsoon season 17 . Hence, Indian rainfall involves not only interannual variability but also decadal variability.
To better describe the long-term patterns in Indian rainfall, we applied an 11-year running mean to a time series of extreme rain events. Figure 1c shows a conspicuous increase during this period, in particular from late 1980 to early 2010, revealing the decadal variability in Indian rainfall. Deka et al. 18 suggested that the increase in rainfall variability in recent decades is capable of increasing the uncertainty of flood and drought events in the Indian region. Unlike previous studies that focused only on summer rainfall, the analyses of year-round precipitation data in the current study reveal the occurrence of extreme rainfall in not only boreal summer but also the ensuing months, in particular since 1993, as shown in Table S1. Several studies pointed out that Asian monsoon has undergone climate shifts in recent decades, following 1976/77 PDO regime shift 19 and since the mid-to-late 1990s 20,21 . Nevertheless, the year 1993 coincides with the recent phase change in the SAM index, which implies its potential impact on extreme Indian rainfall. Nevertheless, the dominant climate factor leading to decadal variability in Indian rainfall requires closer examination.
The relative importance of climate factors, including ENSO, the IOD, the PDO, and the SAM, can be quantified through multiple regression analyses. Because the SAM is characterized by both interannual variability 22 and decadal variability 23 , the regression employed in the SAM was performed separately on interannual and decadal time scales. The regression coefficient revealed that decadal variability in the SAM has an advantage over the other indices, with statistical significance above the 99% confidence level. Although both ENSO and the IOD are statistically significant, their main signal is characterized by interannual variability rather than decadal variability. Figure 2a shows a monthly time series of normalized Indian rainfall anomalies (black curve) along with normalized decadal SAM indices (red curve) during the period from 1979 to 2015; a 121-month running mean was applied to both time series. The two time series show close agreement, with a correlation coefficient of 0.91, which is above the 99% significance level. The high correlation suggests the substantial impact of the decadal SAM on Indian rainfall. The SAM and Indian rainfall have similar variability during the study period, with a negative phase before 1993 that changes to a positive phase and increases considerably thereafter. Both likely have a long-term upward trend over the past three decades, with the exception of the time period around 2005. Figure 2b-e show the atmospheric conditions affected by the SAM, including the dynamic conditions of mean sea level pressure and potential vorticity, as well as the thermodynamic condition of specific humidity, all of which are projected by the SAM (single regression analyses). Terrestrial land adjacent to the northern Indian Ocean has a slightly anomalous low, whereas an anomalous high extends from 60 to 85°E in longitude and from 5 to 50°S in latitude in the central basin (Fig. 2b). The meridional pressure gradient pattern, therefore, can create convergence from the central Indian Ocean to the study area at the low level, also advantageously impacting the organization of the convective system. Figure 2c shows potential vorticity at mid-level (500 hPa), with positive anomalies in the study area but negative ones over the rest of the Indian Ocean. In general, anomalous positive potential vorticities at mid-level are conducive to the formation of convective instability, whereas anomalous negative values are unfavorable for convection.
In addition to the enhanced meridional pressure gradient and increased potential vorticity of the dynamic conditions, the thermodynamic conditions are also performed in this study. Figure 2d-e show regression results for specific humidity at 850 hPa and 500 hPa, respectively; spatially coherent anomalous wet conditions are shown in the study area, whereas the Indian Ocean, in particular south of the equator, displays anomalous dry conditions. This implies a large-scale vertical environment filled entirely with moisture over the study area. The conditions are supportive of stronger convective activity, in particular in the study area. These results also imply that the SAM alters large-scale dynamics and thermodynamics to produce favorable conditions for rainfall over the study area on a decadal time scale.

Mechanism
The trans-hemisphere influence of the SAM on decadal variability in Indian rainfall can be explained by crossequatorial circulation. During the positive SAM phase since the early 1990s, the SST anomaly over the Southern Ocean has shown anomalous warm (cool) conditions in the subtropics (extratropics), as shown in Fig. 3a. The anomalous ascending and descending motions are commonly accompanied by anomalous SSTs over the subtropical and extratropical regions, respectively. The change in the vertical motion weakens Ferrel and Hadley cells in the Southern Hemisphere by offsetting the general circulation. Simultaneously, adjacent meridional circulations of the Hadley cell in the Northern Hemisphere adjust and weaken 24 . Associated with changes in vertical motion, the strong positive anomalous low-level cloud cover (LCC) extends to the eastern coast of South Asia and the East Asian continent, whereas the Indian Ocean (0-50°S and 55-95°E) shows a negative anomalous LCC (Fig. 3b). LCC has a large impact on the surface net heat flux (NHF). A higher LCC generally traps more radiation, which is released from the surface. Figure 3c also shows that a positive anomalous NHF (in which the Earth's surface gains heat) prevails in the Asian continent, whereas an obviously negative anomalous NHF (in which the Earth's surface releases heat) is commonly found over the Indian Ocean. Accordingly, the surface air temperature (SAT; Fig. 3d) exhibits a substantially strong positive anomaly across the Asian continent, including South Asia and East Asia, whereas only a slightly positive anomalous SAT or even a negative anomalous SAT appears over almost the entire Indian Ocean.
The SAT pattern indicates a remarkable land-sea thermal contrast between the Indian Ocean and Asian continent that induces a southerly cross-equatorial flow over the Indian Ocean. Consequently, the pronounced www.nature.com/scientificreports/ intensification of the southerly cross-equatorial flow in the Indian Ocean leads to increased rainfall, as shown in Fig. 4a in the regression pattern of meridional wind speed at 850 hPa. For the sake of explicitness, we further quantified meridional wind speed anomalies averaged over the equatorial Indian Ocean (5°S-5°N and 55-85°E). Figure 4b shows a monthly time series of the anomalies at 850 hPa with a 121-month running mean during the period from 1979 to 2015. The time series reveals a prominent rise beginning in the early 1990s, when the SAM shifted to its positive phase; this indicates that the anomalous southerly wind was gradually enhanced. During the positive SAM phase, the meridional wind speed changed slightly beginning around 1993 (blue shading) accompanied by a slight change in rainfall, whereas it became more pronounced after 2004 (red shading), leading to increased rainfall, which bears out the impact of the decadal SAM on the variability in Indian rainfall.
Although the regression patterns are sufficient to explain the linear pattern effected by the SAM, the issue arises as to whether the manifested impact is associated with natural variability in the climate. To address this issue, we used differences between the positive and negative SAM phases for all parameters of this study (both dynamic and thermodynamic) over the period from 1979 to 2015 (Figures S1 and S2). The composite results are consistent with the results of the linear regression, indicating that the SAM is the dominant climate factor over the Indian region on the decadal time scale.

Summary
The current study demonstrates trans-hemisphere influences on variability in Indian rainfall from the middle and high latitudes of the Southern Hemisphere. Our results indicate that an increase in the SAM is the ultimate forcing leading to significantly enhanced Indian rainfall over the past few decades. In the early 1990s, a positive SAM phase resulted in anomalous warming of the SST in the subtropics and anomalous cooling in the extratropics, offsetting the general circulation and weakening Ferrel and Hadley cells in the Southern Hemisphere. A   β n X n + β 0 + ε A = α 1 S + α 0 + ε