1 INTRODUCTION

As an important part of the Asian summer monsoon，the SASM not only influences India and Southeast Asia but also China. For example，floods from strong monsoon events can extensively damage crops (Wang，2006) . The SASM is a typical type of tropical summer monsoon，namely an Indian trough，equatorial gyre with Somalia jet，and Mascarene High. The SASM is influenced by many factors. A number of scientists have studied the relationship between the SASM and oceanic variability，but they have almost always used sea surface temperature (SST) to represent oceanic signals. Li and Yanai (1996) declared that strong (weak) SASM years were associated with positive (negative) SST anomalies in the western tropical Pacific. Meehl and Arblaster (2003) showed that SASM interannual variability was primarily related to Pacific Ocean SST. The strongest Indian monsoon anomalies occur during the autumn of an El Niño development year. Goswami and Xavier (2005) claimed that the ENSO affects SASM rainfall mainly by controlling the length of the rainy season. Wang et al. (2003) pointed out that thermal conditions in the Pacific have a moderate impact on the strength of the Arabian Sea summer monsoon.

As is well known，there was an abrupt change in the climate system during 1976/1977 (Graham，1994; Miller et al.，1994) ，and in El Niño Southern Oscillation (ENSO) evolution (Wang，1995) . Convection over the tropical Pacific and Indian Ocean and associated circulation patterns have also changed since the 1976/1977 shift (Nitta and Yamada，1989;Wang，1995) . These changes can affect the SASM. Some researchers have found that the relationship between the ENSO and SASM changed during the 1976/1977 climate shift (Ajayamohan and Rao，2008; Sabeerali et al.，2012) . Ajayamohan and Rao indicated that extreme rainfall events over central India increased after the shift. Sabeerali et al. (2012) discovered that prior (posterior) to the 1976/1977 climate shift，most Indian summer monsoon withdrawal dates indicated a late (early) withdrawal.

Compared with SST，upper ocean heat content (HC) can much better represent oceanic variability because of its stability (Feng and Hu，2014) . Subsurface variability in the ocean not only memorizes the SST effect on the atmosphere but also conveys feedback of atmospheric signals to SST (Yan et al.，2010) . Many researchers have investigated the relationship between ocean heat content and climate change，finding that a varying content can be an important indicator of climate change (Chen and Hu，2003; Hu and Yu，2008; Feng et al.，2013) . Therefore，in this paper，we further investigate the relationship between HC over the Indo-Pacific Ocean and the SASM before and after the 1976/77 shift，in terms of HC rather than SST. The rest of the paper is organized as follows. Section 2 describes the datasets and methods. Section 3 introduces the definition of SASM intensity. The relationship between HC and that intensity and a possible HC physical mechanism influencing the intensity are discussed in Sections 4 and 5. Finally，Section 6 presents the conclusions.

2 DATA AND METHOD

The atmospheric data used are from the National Centers for Environmental Prediction/Department of Energy (NCEP/DOE) Reanalysis I and Ⅱ datasets. For a detailed description of these，refer to Kalnay et al. (1996) and Kanamitsu et al. (2002) . We used Reanalysis I monthly wind field and geopotential height with 17 pressure levels in the vertical，from January 1958 through December 1975. The horizontal resolution is 2.5°×2.5°，covering the globe. Monthly wind fields and velocity potential data for January 1978 through December 2013 were from Reanalysis Ⅱ. Ocean temperature data used to investigate the relationship between the SASM and HC in the Indo-Pacific Ocean were from European Centre for Medium-Range Weather Forecasts (ECMWF) Ocean Reanalysis System 4 (ORAS4) (Mogensen et al.，2012; Balmaseda et al.，2013) . These data span 1958 to 2013，with 1°×1° horizontal resolution and 42 vertical levels.

Correlation analysis was used to illustrate the potential relationship between the two variables. We also used composite analysis，which is commonly used to portray responses associated with a certain climate condition by averaging data over a given time period.

3 INTENSITYOFSOUTHASIANSUMMER MONSOON

There are several indexes for the SASM intensity. For example，the All Indian Summer Rainfall Index (AIRI) ，which is defined by seasonally averaged precipitation over all Indian subdivisions from June through September，has long been used as a measure of the Indian summer monsoon (Shukla and Paolino，1983) . To reflect variability of the broad-scale SASM，Webster and Yang (1992) used a circulation index defined by a time-mean zonal wind shear between 850 and 200 hPa，averaged over South Asia (40°-110°E，0°-20°N) (WYI hereafter) . Goswami et al. (1999) found that AIRI and WYI were poorly correlated. They defined a new precipitation index，termed Extended Indian Monsoon Rainfall (EIMR) ，averaged over a larger region (10°-30°N，70°-110°E) ，and a broad-scale circulation index，termed the monsoon Hadley index (MHI) ，defined by meridional wind shear between 850 and 200 hPa averaged over the same area as the EIMR. The EIMR index is strongly correlated with the MH index. Although the MHI exhibits a stronger correlation with AIRI than WYI，it mainly reflects the influence of the Hadley circulation on the SASM. For characterizing interannual variations of the large-scale SASM，Wang and Fan (1999) defined a convection index (CI) ，averaging negative Outgoing Longwave Radiation (OLR) anomalies over a given South Asian region. Li and Zeng (2002) defined a unified dynamical index for a monsoon，dynamical normalized seasonality (DNS) . This index is appropriate for various monsoon regions，and is useful for understanding climate variability of the monsoon system and exploring the relationship between monsoon variability and other major climate variations，such as the ENSO. SASM intensity is defined as a seasonal (June-July-August-September: JJAS) area-average DNS at 850 hPa within a South Asian domain (5°-22.5°N，35°-97.5°E) (Li and Zeng，2002，2003，2005) . Li and Zeng (2005) showed that the SASM index defined by DNS had a stronger correlation with AIRI than with WYI/MHI. Therefore，we use the DNS index of Li and Zeng (2002) .

Li and Zeng (2002) suggested that the SASM sector was composed of two independent components，the Southwest Asian Summer Monsoon (SWASM) over Southwest Asia (2.5°-20°N，35°-70°E) and Southeast Asian Summer Monsoon (SEASM) over Southeast Asia (2.5°-20°N，70°-110°E) ，with very different relationships to monsoon rainfall over South Asia. The SWASM has a clear positive correlation with precipitation over most of north and central India and the Bay of Bengal. In contrast，the SEASM has significantly negative correlations with precipitation from the Arabian Sea across the tropical Indian Ocean to Indonesia and Malaysia. Correlation between the SWASM and SEASM is weak. Therefore，it is important to clarify the variability of SWASM and SEASM in predicting precipitation over South Asia. Therefore，we examine herein SEASM and SWASM relationships with Indo-Pacific HC.

4 RELATIONSHIP BETWEEN SWASM/ SEASM INTENSITY AND HC

Figure 1 shows correlation maps between HC and the SWASM before and after the 1976/77 climate shift. We see from Fig. 1a and b that before 1976/77，there were significant positive correlations (exceeding the 95% confidence level) in the southern tropical Indian Ocean and western tropical Pacific from spring to summer，with significant negative correlations in the eastern tropical Pacific. On the contrary，after 1977 in the tropical Pacific and tropical Indian Ocean，there was no clear relationship between HC and the SWASM (Fig. 1c and d) . The co-relationship between HC and the SWASM，before and after 1976/1977，is consistent with Sabeerali et al. (2012) .

Figure 2 shows correlation maps between SEASM intensity and HC in the Indo-Pacific Ocean before and after 1976/77. Figure 2a reveals that before 1976，there was no significant correlation between HC and the SEASM in the Indo-Pacific. By contrast，after that period，spring-summer HC showed a significant correlation with the SEASM in the central tropical Pacific (Fig. 2c and d) . This indicates a close relationship between the ENSO and SEASM after 1977.

5 POSSIBLE MECHANISMS OF THE EFFECTOFTROPICALINDO-PACIFICHC ON SWASM AND SEASM INTENSITY 5.1 HC and SWASM before 1976/77

According to the above correlation results，we know that SWASM intensity was strongly related to HC in the southern Indian Ocean and tropical Pacific Ocean during 1958-1975. When HC anomalies in the southern Indian Ocean were positive (negative) the previous spring，the SWASM was usually strong (weak) . When HC anomalies in the eastern Pacific were negative (positive) ，the SWASM was usually strong (weak) . Next，we use composite analysis to examine the possible mechanism.

According to Fig. 1a，we chose the southern Indian Ocean region (78.5°-90.5°E，19.5°-13.5°S) ，where HC is strongly related to SWASM intensity as in the southern tropical Indian Ocean. We calculated the standard HC averaged over this region. We defined a warm year as when this standard HC during spring was greater than 0.5 standard deviation. In contrast，a cold year was when that HC was less than -0.5. Over 1958-1975，warm years were 1958，1960，1961， 1964，1969，1970，1973 and 1975. Cold years were 1963，1965，1967，1968，1971 and 1974.

Figures 3 and 4 show composite warm-minus-cold year differences of 850-hPa geopotential height anomaly and wind anomaly fields. From spring to summer，a cyclonic circulation appeared as a response to ocean temperature warming in the southern Indian Ocean，and southerly wind over its western periphery strengthened the Mascarene High over South Africa (Figs. 3,4) . There was low pressure over the northern Indian Ocean (Fig. 3) . Pressure gradients between South Africa and northern Indian Ocean enhanced the Somali cross-equatorial flow (Fig. 4) . Anomalous westerlies in the Bay of Bengal were favorable for increasing SWASM intensity，so that monsoons became strong.

A similar composite analysis was done to examine remote linkage from the Pacific. Warm and cold years were defined based on the HC average in the eastern tropical Pacific (224.5°-244.5°E，2.5°S-2.5°N) . When the standard HC was <-0.5，the year was defined as cold (corresponding to a La Niña event) . For a standard HC in spring >0.5，the year was warm (El Niño event) . During 1958-1975，warm years were 1959，1963，1965，1966，1969，1972 and 1975. Cold years were 1964，1970，1971 and 1973. Figures 5-7 show the composite cold-minus-warm differences of 850 hPa wind，200 hPa wind，and 500 hPa vertical velocity (omega) during spring and summer.

In an eastern Pacific cold year，i.e.，La Niña conditions，the western tropical Pacific usually shows a warm anomaly. At lower levels，there is a pair of anomalous anticyclones over the Pacific from spring to summer，resulting in dominant anomalous easterlies along the equatorial west-central Pacific; in the tropical Indian Ocean，there are anomalous westerlies，which strengthens the SWASM (Fig. 5) . Consistent with 850-hPa atmospheric fields，in the upper troposphere (200 hPa) ，there is a pair of anomalous cyclones over the Pacific，along with anomalous easterly flows over the tropical Indian Ocean (Fig. 6) . Vertical motion features at 500 hPa (Fig. 7) show anomalous ascent over the tropical western Pacific，and descent over the tropical central eastern Pacific and southern Indian Ocean.These anomalous atmospheric circulations show enhancement of the Walker circulation. Therefore，the thermal state of the Pacific influences the SWASM through the Walker circulation.

We conclude that the southern Indian Ocean and tropical Pacific Ocean can influence the SWASM via Somali cross-equatorial flow and the Walker circulation，respectively，with both local and remote effects. It is interesting that some eastern tropical Pacific cold (warm) years are classified as southern Indian Ocean warm (cold) years. However，when both the Pacific and southern Indian are in warm (cold) years，their contribution to SWASM variability is complex. In the future，their relative contribution should be addressed using coupled models.

5.2 HC and SEASM after 1976/77

We concluded in the previous section that SEASM intensity is strongly related to HC in the tropical central Pacific. According to the standard HC average over the region 165.5°-190.5°E，4.5°S-3.5°N，we selected 11 warm years (1980，1982，1986，1989，1991，1997，2004，2006，2008，2011 and 2012) and 7 cold years (1983，1984，1988，1992，1998，1999 and 2010) from 1978-2013. We performed composite analysis of 850-hPa wind，200-hPa wind，and velocity potential in summer to investigate atmosphere response to ocean temperature.

During summer when the equatorial central Pacific warms，to its west at lower levels (Fig. 8a) are anomalous westerlies; to its east are anomalous easterlies. A pair of anomalous surface cyclones form in the northern and southern hemisphere over the eastern Indian Ocean/western Pacific and Maritime Continent (Fig. 8a) . As a result，there is westerly flow from the Bay of Bengal to western tropical Pacific via the South China Sea. There is enhanced cross-equatorial flow over the eastern tropical Indian Ocean. The wind field in the upper troposphere is consistent with lower levels (Fig. 8b) . Consistent with the wind fields，Fig. 9a shows a dipole pattern with the center of convergence in the equatorial central Pacific and center of divergence over the Indo-western Pacific in the lower troposphere. The counterpart in the upper troposphere also exhibits a dipole structure，but with divergence in the central Pacific and convergence over the Indian Ocean -Maritime Continent. All these features are favorable for strengthening the SEASM.

6 SUMMARY AND DISCUSSION

We differentiated the SASM into the SWASM (2.5°-20°N，35°-70°E) and SWESM (2.5°-20°N， 70°-110°E) ，and investigated their relationship with Indo-Pacific HC before and after the 1976/1977 climate shift. The principal results are as follows.

(1) The SWASM was strongly linked to HC in the southern Indian Ocean and tropical Pacific. Before the 1976/1977 climate shift，when the southern Indian Ocean warmed (cooled) ，the SWASM was usually strong (weak) . When the eastern tropical Pacific cooled (warmed) ，the SWASM was again usually strong (weak) . The southern Indian Ocean and tropical Pacific influence the SWASM by modulating the Somali cross-equatorial flow and Walker circulation related to the remote forcing of ENSO，respectively. However，after the 1976/1977 climate shift，there was no significant correlation between the SWASM and HC in the tropical Pacific or tropical Indian Ocean.

(2) There was substantial correlation between the SEASM and central tropical Pacific HC after the 1976/77 climate shift. When central tropical Pacific HC was positive (negative) ，the SEASM was usually strong (weak) . The mechanism of the central tropical Pacific influence on the SEASM is similar to that of the tropical Pacific influence on the SWASM，as discussed above. However，before the 1976/1977 climate shift，their relationship was weak.

As mentioned above，Li and Yanai (1996) ，Meehl and Arblaster (2003) ，and Wang et al. (2003) suggested the influence of Pacific SST on SASM intensity. Their studies used SST data，which could not fully represent upper ocean variability. In contrast，in the present paper，we have addressed SASM intensity as a function of HC.

From the analysis，we conclude that HC in the Indo-Pacific Ocean had different effects on SWASM and SEASM intensities before and after the climate shift. According to Kumar et al. (1999) ，this difference may have been caused by a southeastward shift of Walker circulation anomalies and an enhanced land-sea pressure gradient. The results reported herein were based on statistical analysis; physical processes require further study using numerical experiments.