Temporal and spatial characteristics of positive and negative Indian Ocean dipole with and without ENSO

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113,, doi:10.1029/2007jd009151, 2008 Temporal and spatial characteristics of positive and negative Indian Ocean dipole with and without ENSO Chi-Cherng Hong, 1 Mong-Ming Lu, 2 and Masao Kanamitsu 3 Received 10 July 2007; revised 9 November 2007; accepted 26 December 2007; published 18 April 2008. [1] The differences in the temporal evolution and spatial characteristics of the Indian Ocean Dipole (IOD) between positive and negative events with and without ENSO have been investigated using observations for the period 1948 2002. To document such differences is particularly important for climate forecasts over far east Asia, since distinctly different monsoon activities over China, Korea, and Japan for different types of IOD are found in the composite maps of precipitation anomalies. The composite map of SST and wind during various stages of IOD and the ocean mixed layer heat budget showed that the IOD with and without ENSO has a large difference in its temporal evolutions and their triggering mechanisms. In both negative and positive IOD events without ENSO, the wind anomaly in the eastern Indian Ocean seems to be responsible for the formation of sea surface temperature anomalies, while the anomaly in the western Indian Ocean seems to be the oceanic dynamical response to the anomaly in the east. During the ENSO years, the temporal and spatial contrast of the asymmetry of the IOD evolution is smaller, and the SST anomaly is driven by the anomalies in incoming radiation due to changes in cloudiness caused by the ENSO associated anomalous atmospheric circulations and not by the local wind anomalies. Citation: Hong, C.-C., M.-M. Lu, and M. Kanamitsu (2008), Temporal and spatial characteristics of positive and negative Indian Ocean dipole with and without ENSO, J. Geophys. Res., 113,, doi:10.1029/2007jd009151. 1. Introduction [2] The Indian Ocean Dipole (IOD) is one of the dominant modes of the interannual variability of the Indian Ocean SST [Saji et al., 1999; Webster et al., 1999]. It has been argued that the IOD is an artificial mode [Dommenget and Latif, 2002] or that it is an ENSO forced mode [Baquero-Bernal et al., 2002; Yu and Lau, 2004]. But recent studies show IOD is a physical mode and can be identified without ENSO [Ashok et al., 2003; Saji and Yamagata, 2003a; Lau and Nath, 2004; Zhong et al., 2005; Cai et al., 2005; Behera et al., 2006]. [3] The abundant studies of IOD are not only due to its scientific interest but also to its practical implications. IOD influences the Indian and Asian monsoon precipitation [Vinayachandran et al., 1999; Zubair et al., 2003; Ashok et al., 2004; Kripalani and Kumar, 2004; Harou et al., 2006] and even areas outside the Asian monsoon region [Guan and Yamagata, 2003; Saji and Yamagata, 2003b]. [4] In many of the published papers, we found that positive and negative phases of IODs are not treated separately. We also noticed that the well-documented [Saji 1 Department of Science, Taipei Municipal University of Education, Taipei, Taiwan. 2 Central Weather Bureau, Taipei, Taiwan. 3 Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California, USA. Copyright 2008 by the American Geophysical Union. 0148-0227/08/2007JD009151 et al., 1999; Vinayachandran et al., 1999; Webster et al., 1999; Yu and Rienecker, 1999] strong IOD cases were all based on positive events, from which some of the studies of its mechanisms are based [Li et al., 2002, 2003; Zhong et al., 2005]. The study of negative events was less common than the positive events and the difference in the temporal evolution of positive and negative IODs has not been documented well. One of the few studies that examines the positive and negative phases of IOD separately is Saji and Yamagata [2003a]. Their conclusion is that from the statistical point of view, both negative and positive IOD events have the same characteristics, but they did not examine the details of the structural differences and their time evolution. [5] The influence of ENSO on IOD is another point of interest. It has been noted that ENSO remote forcing leads to an Indian Ocean basin scale warming in the following spring after the ENSO matures [Xie et al., 2002]. Nearly 40% of IODs occur simultaneously with ENSO; the positive IOD is associated with El Niño and the negative IOD generally with La Niña [Saji and Yamagata, 2003a; Ashok et al., 2004]. The warm and cold ENSO events do not evolve in the same way (with opposite sign) [An and Jin, 2004]. It was also recently demonstrated that the ENSO- Indian ocean coupling is different for warm and cold events [Kug et al., 2006]. In addition, long-term numerical simulations indicate that the large-scale atmospheric circulation which initiates IOD is different for the ENSO associated and non-enso associated IODs [Lau and Nath, 2004]. From these views of the positive and negative phase of the IOD 1of15

Table 1. Years of Positive and Negative IOD Events a Type of IOD Year Positive 1961, 1963*, 1966, 1967, 1972*, 1976*, 1977, 1982*, 1983, 1987*, 1994, 1997* Negative 1954*, 1955*, 1956*, 1958, 1959, 1960, 1964*, 1971*, 1989*, 1992, 1996, 1998*, 2001 a The years with asterisks indicate an IOD year with ENSO and those without indicate a pure IOD year. and its response to ENSO, we feel it is essential to reexamine the evolution of spatial structure of IOD for positive and negative phases as well as with and without ENSO, individually. [6] The main purposes of this study are (1) to examine the relation between various IODs and the precipitation over Asia, (2) to describe the characteristics of time evolution and structure of the four types of IODs, (3) to examine the impact of ENSO on the IOD, and (4) to interpret the physical mechanisms of the evolution of each IOD event. In section 2, we discuss the observational data used. The relation between IODs and precipitation anomaly in far east Asia is discussed in section 3. Section 4 presents results of the composite analysis of the evolution of IODs, and section 5 analyzes the mixed layer heat budget. The conclusions are presented in section 6. 2. Data [7] The monthly NCEP reanalysis [Kalnay et al., 1996] winds, geopotential height, and NOAA Extended Reconstructed Sea Surface Temperature from 1948 to 2002 [Smith et al., 1996] are used as major data sources in this study. The CRU (Climatic Research Unit in the University of Norwich, United Kingdom; http://www.cru.uea.ac.uk/) data based on gauge observations is used for the diagnosis of precipitation. We also utilized COADS (Comprehensive Ocean-Atmosphere Data Set) and ERA40 (ECMWF 40 year reanalysis) [Uppala et al., 2005] to examine the robustness of the NCEP/NCAR Reanalysis based results. [8] For the ocean process, the Simple Ocean Data Assimilation (SODA) products of Carton et al. [2000] for the same period are used. The SODA data resolution is 1 degree in the zonal, and variable in the meridional direction, from 0.45 degrees at the equator to 1 degree in higher latitudes. There are 20 levels in the vertical, with a spacing of 15 m at the top and 737 m at the bottom. The usefulness of SODA for studying the Indian Ocean coupled dynamics has been demonstrated by Xie at al. [2002]. [9] For the anomaly calculations, the monthly climate average for 1971 2000 is first subtracted, and then a linear trend (of the basin scale warming of about 0.5 C/100 year) is removed. The period of climatology (30 years) is based on WMO specification. The results did not vary much when we utilized the 1948 2002 average as climatology. The removal of linear trend makes the IOD the first principal mode of SSTA in the Indian Ocean. [10] The IOD index as defined by Saji et al. [1999] is used, which applies the following two conditions: (1) The difference of normalized summer season (JJA) SSTA between the area 50 E 70 E,10 S 10 N (hereafter referred to as IOD-W) and 90 E 110 E,10 S 0 N (hereafter referred to as IOD-E) is larger than one standard deviation (s), and (2) JJA zonal wind anomaly averaged over the central Indian Ocean 70 E 90 E, 5 S 5 N, (hereafter referred to as IOD-U) is easterly (westerly) for the positive (negative) IOD event (see Figure 4 for the IOD-E, IOD-W, and IOD-U areas). [11] We further separated the IODs into pure-iod and ENSO-IOD depending on the coexistence of ENSO events. For the ENSO index, we used Trenberth s [1997] definition. Following the above definitions, 12 positive and 13 negative IOD events are identified (Table 1). In Table 1, an asterisk denotes the ENSO-IOD and no asterisk denotes the pure-iod. The pure negative case in 1992 follows a strong El Niño. Because the ENSO remote forcing is still strong (Niño 3.4 = 1.4) in the initiation stage, it was excluded for the spring season (MAM) composite. [12] We note that our definition of IOD is somewhat different from that used by Saji and Yamagata [2003a], but is closer to that of Saji et al. [1999]. Saji and Yamagata [2003a] filtered out interdecadal timescale, as well as the basin-wide anomaly and the detrend applied in both studies. Their SSTA criterion of IOD is that each pole is larger then 0.5 s and ours is that the difference is larger than 1.0 s. These differences resulted in the different characteristics between positive and negative phases, as discussed later in this section. [13] In order to make sure that our results are not dependent on reanalysis procedures, we repeated nearly all of the calculations performed by Saji and Yamagata [2003a] which are based on COADS data, using NCEP/ NCAR reanalysis, and confirmed that the reanalysis data can reproduce their results. We also repeated many of the calculations made in this study using the COADS data and ERA40 data whenever possible to confirm the robustness of our result. Note that ERA40 data starts from 1958 while NCEP/NCAR reanalysis starts from 1948 and some recomputation was necessary when comparing the two results. All the results agreed with each other except for an erroneously stronger decadal-scale meridional wind component over the western Indian Ocean and Arabian Sea in the NCEP/NCAR reanalysis. More details of this problem will be discussed in section 4.1. 3. Relation Between IOD and Precipitation Over Far East Asia [14] In this section, we will first examine the relation between the different types of IODs and the summer precipitation over far east Asia. The purpose of this section is to demonstrate the importance of studying each type of IOD for practical forecast implications, particularly over the far eastern Asian monsoon region. [15] A contrasting anomalous precipitation pattern during the positive and negative phase of IODs is shown in the composite rainfall anomaly maps in Figure 1. The composite for ENSO without IOD is to separate the sole effect of ENSO from the total effect of ENSO and IOD. 2of15

Figure 1. Composite 850-hPa stream line and precipitation anomaly for JJA during the positive phases of (a) pure-iod, (b) ENSO-IOD, (c) pure-enso, and the negative phases of (d) pure-iod, (e) ENSO- IOD, and (f) pure-enso years. Gray stipples indicate anomalous precipitation is greater than 95% confidence level. The CRU precipitation is presented. The IOD events are shown in Table 1. Positive phase of pure-enso events are 1957, 1965, 1969, 1986, 1991, and negative phase of pure-enso events are 1950, 1970, 1973, 1975, 1988, 1999, 2000. 3of15

Figure 2. Composite anomalous JJA local Hadley circulation along the longitudes of 100 E 120 E represented by the vertical wind velocity (unit: Pa s 1 ) for the negative phase of (a) pure-iod, (b) ENSO-IOD, and (c) pure-enso. The positive area represents the downward motion and the negative area represents the upward motion. Thick arrows indicate the vertical wind anomaly greater than 95% confidence level. Bonin High over the Yellow Sea, Korea, and Japan, which was identified as a major factor causing the dry and hot summer over far east Asia during the positive phase of the IOD [Guan and Yamagata, 2003], is evident in the IOD case without ENSO as shown in Figure 1a. [17] Interestingly, the rainfall anomaly patterns and signs over the far east Asia for both positive (Figure 1a) and negative (Figure 1d) pure-iod events are found to be very similar. Looking at the detail, the dry condition over northern China, Korea, and Japan is more intense during the positive pure IOD, while the wet condition over southern China and Taiwan is more intense during the negative pure IOD. The Bonin High strengthens in both phases, but its horizontal extent is narrower in the zonal direction and wider in the meridional direction during the negative phase. The most distinct contrast between the negative and positive phases of the pure-iod is found over northern Myanmar, Indonesia, and northeastern Indochina. [18] The impacts of negative IOD and La Niña on the rainfall over far east Asia appear to be complicated. The Bonin High is active during the negative pure IOD, leading to the dry condition over Korea and Japan (Figure 1d). Compared to the negative pure IOD cases, the anticyclonic circulation around Japan during the pure La Niña cases shifts to the north due to the enhancement of the convection over the maritime continent (Figure 1f). Figure 1f also shows an anomalous cyclonic circulation stretching from northern Indochina through southern China to the East China Sea when La Niña occurs without IOD. This anomalous cyclone provides a wet condition for Southeast Asia [Ropelewski and Halpert, 1987]. When La Niña appears concurrently with negative IOD (Figure 1e), the anomalous anticyclonic circulation over the South China Sea and the Philippine Sea is substantially intensified, which results in dry conditions over the areas of southern China and Taiwan and wet conditions over a wide region from Bangladesh, Assam, the west coast of Myanmar, and the Yangtze River, to Korea and Japan. An example of an extreme negative IOD and La Niña year is 1998. The massive flood along the Yangtze River in south central China killed 3656 people. [19] The anomalous anticyclonic circulation over the South China Sea and the Philippine Sea during the negative [16] The rainfall anomalies over India show opposite signs for the composites of the IOD with and without ENSO. For El Niño without IOD, the northern part of India is dry and the southern part is wet (Figure 1c). This is consistent with the findings of Ashok et al. [2004]. The rainfall anomalies over far east Asia also show opposite signs for the composites of the IOD with and without ENSO. The areas over central eastern China, Korea, and Japan are dry when the positive IOD occurs without ENSO but wet when El Niño occurs (Figures 1a and 1b). The Figure 3. Climatological annual cycle SST over IOD-E (closed circle line) and IOD-W (open circle line), and zonal component of wind over IOD-U (open square line). 4of15

Figure 4. Interannual variance of SST for JJA. The western, central, and eastern boxes represent the areas of IOD-W, IOD-U, and IOD-E respectively. ENSO-IOD years is a key factor to enhance the wet condition over the Yangtze River and to cause the opposite influence of IOD on the precipitation over far east Asia from southern China and the Philippines to Korea and Japan. The anomalous local Hadley circulation along the longitudes 100 E 120 E in Figure 2 suggests that both negative IOD and La Niña can enhance the upward motion over Indonesia and the downward motion over the South China Sea. However, for the areas to the north of 20 N, the influence of the IOD is evidently stronger than La Niña. The anomalous subsidence over southern China (25 N 30 N) appears in both pure-iod (Figure 2a) and ENSO-IOD (Figure 2b) plots, but not in the La Niña years without IOD (Figure 2c). When the negative IOD and La Niña occur together, the anomalous subsidence area extends to the south covering the latitudes of 15 N 25 N. The upward motion over 30 N 35 N is also substantially intensified at the upper troposphere when the negative IOD and La Niña occur together (Figure 2b). This enhanced upward motion is Figure 5. Time evolution of normalized IOD index (bar), SSTA over IOD-W (open circle lines), SSTA over IOD-E (closed circle lines) and zonal wind anomaly over IOD-U (open square lines). (a) Positive pure IOD, (b) negative pure IOD, (c) positive ENSO-IOD, and (d) negative ENSO-IOD. The normalized SSTA over Niño3.4 is also shown in thick gray lines. X-axis is month of the year, (0) being the year when the IOD occurs, ( 1) is the previous and (+1) is the following year. Y-axis is normalized variance. 5of15

Figure 6. Anomaly of 850-hPa winds (arrow) and SST anomaly (shading) during initiation period (MAM) for (a) positive pure-iod, (b) negative pure-iod, (c) positive ENSO-IOD, and (d) negative ENSO-IOD. Gray stipples indicate SSTA greater than 95% confidence level. Only the wind anomalies significant at 95% are plotted. consistent with the wet condition over the Yangtze River and Korea observed in Figure 1e. The surplus rains over Indonesia and central China, Korea, and Japan and the deficient rains over the Philippines, southern China, and Taiwan resemble the negative phase of Nitta s Pacific-Japan pattern [Nitta, 1987], which is explained as a Rossby wave response to the tropical heating over the Philippines by Kurihara and Tsuyuki [1987]. Thus, it is of profound importance to study the differences in the evolution of IOD associated with and without ENSO in order to understand and improve the east Asian monsoon rainfall prediction. 4. Characteristic Features of the IODs [20] In this section, we will examine the temporal and spatial characteristics of the positive and negative pure- IODs and ENSO-IODs in detail. Before discussing the time evolution and space characteristics of the IODs, it is important to understand the annual cycle of the Indian Ocean SSTs and equatorial zonal winds. The seasonal evolution of these variables over IOD-W, IOD-E, and IOD-U are presented in Figure 3. The dominant wind over the central equatorial Indian Ocean is westerly (square line) persisting from April to December. The SST over IOD-W is warmest in April and coldest in August due to the strong upwelling induced by the Somali Jet. Therefore, the amplitude of the annual cycle of SST over the IOD-W is much larger than that over IOD-E. The temperature difference between the areas IOD-W and IOD-E is largest in August and smallest in March and May. Except in April, the SST over IOD-E is warmer than that over IOD-W. Figure 4 shows the interannual variance of SSTs during the summer monsoon season (JJA). Figure 4 clearly displays that large variances are located near the eastern coast of the Indian Ocean, along Somalia and Saudi Arabia, and along the western coast of Sumatra. The large variances over IOD-E and IOD-W are due to the IOD. The other large variance (120 E, 20 S) located to the southwest of IOD-E is probably caused by the southward propagating oceanic Kelvin wave trapped in the Australian coastline [Meyers, 1996]. Another large variance off the coast of Somalia (50 E, 10 N) is probably attributed to the ENSO ocean wave from the plot of the D20 (depth of 20 C isotherm; not shown). 4.1. Pure-IOD [21] The evolution and spatial characteristics of pure-iod derived from the composite of the positive and negative events are shown in Figures 5, 6, and 7. For the positive IOD, we find that the easterly anomaly first appears over the equatorial Indian Ocean (IOD-U) in March and persists until 6of15

Figure 7. Same as Figure 6, but during development phase (JJA). December (shown as a square line in Figure 5a), while the SST anomaly characterized by negative SSTA over IOD-E and positive SSTA over IOD-W does not appear until May (shown as open and closed circle lines in Figure 5a). As the JJA climatology of the zonal wind over IOD-U is westerly (Figure 3), the easterly anomaly over IOD-U implies weaker westerly winds near the equator. The anomalous easterlies over IOD-U and its associated negative anomalous SST intensify sharply during summer after the Indian summer monsoon matures. The rapid development found in IOD-E does not occur in IOD-W. [22] The evolution of the negative phase of the pure-iod is presented in Figure 5b. Compared with the positive IOD shown in Figure 5a, the equatorial westerly zonal wind anomaly (square line) appears earlier (December of the previous year) and with much smaller amplitude. The development of the SSTA over IOD-E (closed circle line) starts well before March (Figure 5b), which is much earlier than the positive IOD case (Figure 5a). [23] The spatial-temporal patterns of positive and negative IOD events are illustrated using the composite maps of SST and 850-hPa wind anomalies during MAM (initiation phase) and JJA (development phase) for the positive (Figures 6a and 7a) and negative (Figures 6b and 7b) events, respectively. [24] For the positive events, the figures suggest that the negative SSTA over the IOD-E in the development phase is formed in response to the surface wind easterly anomalies during the leading season, MAM (Figure 6a). This anomalous offshore flow can potentially cause colder SSTA due to anomalous upwelling along the western coast of Sumatra where thermocline is shallow, which will be further examined from the ocean heat budget in section 5. Over the Western Indian Ocean, we do not see apparent anomalies in winds, but there is a slight warm anomaly in SST in the preceding season. [25] For the negative IOD event, we observe a weak westerly anomaly over the eastern Indian Ocean in the leading season (Figure 6b). It is also noted that a significant warm SSTA over IOD-E has already formed in the leading season (Figure 6b), which is followed by stronger than normal convective activity over the Indonesian maritime continent (Figure 8b). The negative pure IOD event exhibits a gradual evolution in phase with the seasonal cycle. The most notable feature over IOD-W is that the southerly wind anomaly along the coast of Africa appears only in negative IODs for both the preceding and development seasons. Although this wind anomaly in the preceding season is consistent with the cooling over the African coast in the following summer (Figure 7b) due to the possible enhancement of upwelling, the wind anomaly over IOD-W during MAM is found to be strongly dependent on the observational data used and may not be a reliable feature. A comparison of the NCEP/NCAR reanalysis composite wind 7of15

Figure 8. Hovmüller diagrams of precipitation (shaded) and SSTA averaged over 10 S 5 N. (left) Positive pure IOD and (right) negative pure IOD. Contour interval is 0.3 C. anomaly to that of ERA40 shows (Figure 9) that the southerly anomaly along the African coast in MAM does not appear in ERA40, although it shows up in both analyses during JJA. Note that since the ERA40 period started from 1958, but our analysis started from 1948, the composite for the NCEP/NCAR reanalysis was also computed from the same year for this comparison. We also compared the COADS wind with the NCEP/NCAR and ERA40 reanalyses over the Arabian Sea and found that the NCEP/NCAR reanalysis overestimated the meridional wind during MAM 1958 1963, while ERA40 was reasonable (Figure 10). The average overestimation of meridional wind anomaly during 1958 1963 is 1.25 ms 1, almost one and a half of the temporal standard deviation (s v =0.9ms 1 ). Because a half of the negative pure-iods (Table 1) appear in this period, we concluded that the large southerly anomaly over IOD-W in the NCEP/NCAR reanalysis during MAM was erroneous. The NCEP/NCAR Reanalysis fits the COADS well after 1963, the average deviation of meridional wind anomaly decreases to 0.2 ms 1. [26] The relation between the time evolution of the SSTA associated with pure IOD and the seasonal march of SST is summarized below. The SSTA over IOD-E is larger than that over IOD-W in the mature phase of both positive and negative IOD events. For the positive events, the seasonal variation of SSTA over IOD-E and the seasonal cycle are in phase but the SSTA has larger amplitude than the seasonal change (compare Figure 3 and Figure 5). Figure 5a shows that the positive IOD-E anomaly appears in April, 1 month ahead of the climatologically warmest month of the year (Figure 3), while the largest negative anomaly appears in October, 1 month after the climatologically coldest month of the year. The positive SSTA over IOD-W during the period April to October (Figure 5a) leads to weaker than climatological seasonal SST variation over IOD-W. In other words, the positive IOD event appears as a stronger than normal seasonal change of SST over IOD-E but weaker than normal variation over IOD-W. By the same token, the negative IOD event appears as a stronger than normal seasonal change of SST over IOD-W but weaker than normal seasonal change over IOD-E. 4.2. ENSO-IOD [27] We plot the temporal evolution of the ENSO-IOD indices (Figures 5c and 5d) along with their composite maps of SST and 850-hPa wind anomalies (Figures 6c and 6d and Figures 7c and 7d). In general, the positive IOD events occur during El Niño, while the negative IOD events occur during La Niña as seen from the SST anomaly over Niño 3.4 (gray lines in Figures 5c and 5d). [28] For the positive ENSO-IOD, we find that El Niño can enhance the positive SSTA over IOD-W (compare Figures 5a and 5c). In addition, the anomalous easterly is observed over most of the equatorial Indian Ocean for ENSO associated IOD, but it is much more restricted to the eastern Indian Ocean for positive pure IOD. The anomalous easterlies over the Eastern Indian Ocean are distributed in a narrow belt over the entire Indian Ocean, from north Sumatra to Sri Lanka and the southern tip of the Indian subcontinent extending to the eastern equatorial 8of15

Figure 9. Comparison of 850-hPa wind anomaly for negative pure IOD during MAM and JJA using two different data sets, (left) NCEP/NCAR and (right) ERA40 reanalyses. Note the complete disappearance of the southerly anomaly over the African coast in ERA40 during MAM. Figure 10. Comparison of monthly averaged meridional component of wind averaged over the domain (45 E 55 E, 5 S 10 N) between NCEP/NCAR reanalysis, ERA40 and COADS. The correlation between NCEP/NCAR and COADS and ERA40 and COADS area also shown. 9of15

Figure 11. Same as Figure 6, but during mature phase (SON). Indian Ocean during the development phase (Figure 7c). The months of initiation, peak and termination of the anomaly over IOD-E are about the same for the positive ENSO- and pure-iod (Figures 5a and 5c). Over IOD-W, the time period is shifted forward by several months, due to the effect of ENSO, which normally peaks in winter. [29] For the negative ENSO-IOD, the westerly anomaly over the IOD-E is stronger in MAM (Figure 6d) when compared with its pure-iod counterpart. But only a very limited anomaly is seen off the coast of Sumatra. However, the difference becomes rather small during the mature phase in SON (Figure 11d). Note that a large area of anomalous easterlies appears over Indochina in JJA (Figure 7d), which reflects the remote influence of ENSO. The wind anomaly over the western Indian Ocean for this phase lacks the strong southerly anomaly found in the pure IOD case. Thus, the main difference between the negative phases of pure- IOD and ENSO-IOD during the development phase is located over both the IOD-W area and over Indochina and southeastern Sumatra (Figures 7b and 7d). Compared to the positive IOD with El Niño, the influence of La Niña on negative IOD tends to be weaker due to the weaker La Niña forcing (Figures 5c and 5d). [30] The influence of ENSO on IOD is much greater during the decaying stage of IOD. As the El Niño causes a basin scale warming over the Indian Ocean after its mature phase [Xie et al., 2002; Chowdary and Gnanaseelan, 2007], it leads to the warming of IOD-W making the west pole persist longer. The basin scale warming makes the SSTA over IOD-E turn sharply from negative to positive (Figure 5c) compared to the weak negative SSTA which persisted until the following spring in the pure-iod case. Thus, characteristic features of the ENSO-IOD during the decaying stage seem to be explained mainly by the superposition of the remote ENSO impact and pure IOD features. [31] In summary, the temporal evolutions of pure-iod and ENSO-IOD are very different. Under the influence of ENSO, the time evolution of the IOD index is more gradual than pure IOD in a positive event and more rapid than pure IOD in a negative event. This is in accord with the mature phase of El Niño occurring in winter when positive ENSO- IOD develops and La Niña occurring in summer when negative ENSO-IOD develops. 5. Ocean Mixed Layer Heat Budget [32] In this section, we will present the role of the ocean advection and surface heat fluxes for the formation of SSTA over IOD-E and IOD-W by analyzing the mixing layer heat budget. The budget of the deviation from climatological mean of vertically integrated heat in the ocean mixed layer is expressed by the following equation: @hti 0 @t ¼ hv rti 0 þ 1 rc P H hq SW þ Q LW þ Q LH þ Q SH i 0 þ R 0 ð1þ 10 of 15

Figure 12. Time series of computation (closed square) and observed (open square) mixing layer temperature tendency of east pole for the total IOD events. Closed circle lines denote the positive IOD phase and open circle lines are negative IOD phase. (a) IOD-E and (b) IOD-W. The shading denotes the 1s envelop of the standard deviation of heat flux. where the brackets are defined as vertical average from the ocean surface to the bottom of the mixed layer and V =(u, v, w) is three-dimensional ocean current. There are four levels in the mixed layer. The term (V rt ) will be called ocean advection in this paper, and prime denotes deviation from monthly climatology. Thus, the term (V rt) 0 can be decomposed to V rt 0, V 0 rt, and V 0 rt 0 and includes contribution from seasonal variation to the heat anomaly. We note that the advection term is threedimensional and therefore it includes vertical temperature advection from the bottom of the mixed layer. Q SW, Q LW, Q LH, Q SH represent the net downward shortwave radiation flux at the ocean surface, net upward long-wave radiation flux, surface latent heat flux, and sensible heat flux, and the sum of the fluxes will be called heat flux. R is the residual term representing either the model errors or a process other than heat fluxes and temperature advection, H is the mixed layer depth, which varies from 30 m to 50 m by season [Du et al., 2005]. Here we take the mean (40 m) for the budget analysis. Before analyzing the heat budget, we first compare the estimate of the mixing layer temperature (hereafter MLT) total tendency as a sum of ocean advection and heat flux, with the observed tendency. This displays that the tendency from the budget analysis reproduced the features of the observed tendency (Figure 12), both evolutions and peaks, fairly well. For the positive phase, the computation seems to overestimate the tendency. The possible reason is that the ocean analysis (SODA only supports ocean data) and the surface heat fluxes are from different sources and the NCEP Reanalysis tends to overestimate surface heat flux in the Indian Ocean [Yu et al., 2007]. However, the estimated tendency is within the error range of one standard deviation of the surface heat flux (Figure 12, shading). Thus, we felt confident that this result will provide at least qualitative nature of the ocean advection and heat flux. Similar to the description of the time evolution and spatial patterns, we first show the results of pure-iod and then discuss the ENSO-IOD. [33] The computed MLT total tendencies averaged separately over IOD-E and IOD-W for pure-iod are shown in Figure 13. For the IOD-E, both ocean advection and latent heat contributes to the SSTA for positive and negative IOD (Figure 13a). Figure 13 also illustrates that the heat flux (peaks in May June) leads the ocean advections (peak in August September) by 2 3 months, which means the latent heat flux is important in the initiation of IOD. We found that the tendency due to advection and latent heat flux for positive pure-iod is larger than for negative pure-iod. This difference in forcing is the major cause of the asymmetric behavior between the positive and negative phase, resulting in the SSTA magnitude of IOD-E for the positive phase being much larger than that of the negative phase. The cause of the asymmetry in evaporation (latent heat) can be explained by the difference in total wind speed, i.e., the anomalous wind over IOD-E is in direction with mean flow for positive IOD but against for negative IOD (Figures 14a and 14b), which leads to a stronger surface wind for positive IOD and weaker wind for negative IOD. Note that the JJA SSTA amplitude is almost the same between the positive and the negative IODs, implying that the asymmetry of wind speed anomaly is not attributable to SSTA. We note that the asymmetry of wind anomaly is highly correlated with the low level anomalous anticyclonic circulation located to the south of Australia (not shown), previously suggested by Lau and Nath [2004]. [34] In contrast to IOD-E, the amplitude of advection and heat flux over IOD-W are much smaller (Figure 13c), which could explain why the SSTA amplitude of IOD-W is smaller than IOD-E. Figure 13c also shows that the MLTover IOD-W is primarily driven by ocean advection and the maximum forcing of advection (October November) lags that of IOD-E (August September) by 1 2 months. This is consistent with the model diagnosis performed by Li et al. [2002] and may be explained by the time lag of wind-driven ocean current in responding to the surface wind stress. The lag relationship between the ocean advection and heat flux observed for IOD-E is not observed for IOD-W. Instead, a second SSTA peak of pure IOD-W driven by ocean advection is noted in April May for positive IOD (Figure 13c). This suggests that the wind anomalies over the equatorial eastern Indian Ocean during MAM (Figure 6a) may play an important role in triggering the SSTA over IOD-W. 11 of 15

Figure 13. Time series of computation mixing layer temperature tendency for (a and b) IOD-E and (c and d) IOD-W. Open square line represents surface heat fluxes, closed square line denotes ocean advections, and the cross line is the observed temperature tendency. Figures 13a and 13b are pure IOD and Figures 13c and 13d denote ENSO-IOD. Dark lines represent the positive IOD phase and gray lines are negative IOD phase. [35] For ENSO-IOD, the heat budget displays that El Niño enhances the ocean advection (Figure 13b, dark closed square line), but its impact on surface heat flux is weak (Figure 13b, dark closed circle line), especially over IOD-E. Here, we need to emphasize that the peak forcing by advection lags latent heat by 2 3 months, occurring in the mature phase when the downward shortwave radiation (negative feedback to SSTA) dominates the heat flux (Figures 13a and 13b, dark closed circle line), and counteracts the effect of advection. Thus, a larger advection does not necessarily imply larger total forcing. In fact, the SSTA amplitude over IOD-E is primarily determined by the latent heat, which is controlled by the local wind speed. Figure 14 shows positive pure-iod has a larger latent heat flux than ENSO-IOD in JJA, which dominates in determining SSTA for the positive pure-iod. In contrast to El Niño La Niña enhances the evaporation (Figures 13b and 13d, gray closed circle lines), but their influence on ocean advection is not significant (Figures 13b and 13d, gray closed square lines). There is a difference in the SSTA associated with IOD between La Niña and El Niño (or negative and positive ENSO-IOD) as examined from the ocean heat content anomaly averaged in the layer 0 125 m (Figure 15). The SSTA penetrates deeper for positive IOD while it is limited to a much shallower layer for negative IOD. Such characteristics do not change when ENSO and IOD appear concurrently. In other words, the influence of La Niña on Indian ocean SST is limited to the near-surface but the El Niño can extend to the subsurface. 6. Conclusions [36] The characteristic features of the positive and negative phases of Indian Ocean Dipole events are examined by classifying them into those associated with ENSO (ENSO- IOD) and without (pure-iod). Our interest is to document the difference in four types of IOD, examine their possible impact on far eastern Asian monsoon precipitation and to speculate on the differences in the triggering and developing mechanisms of the IODs. The monthly NCEP reanalysis winds, geopotential height, NOAA Extended Reconstructed SST from 1948 to 2002, COADS, ERA40, the SODA products of Carton et al. [2000], and the CRU precipitation observation are used for the diagnostics. [37] The connection between IOD and the dry/wet precipitation anomaly over far east Asia is one of the interesting findings in this study. The north-south orientated area of rain anomaly over east Asia is known to be influenced by the interdecadal variability of the Pacific High [Chang et al., 2000a, 2000b] and Tibetan Plateau topographic forcing 12 of 15

Figure 14. Same as Figure 7, but for the latent heat flux (contour) and SSTA (shading). Wind vector denotes the 10 m anomalous horizontal wind. Contour interval is 5 W/m 2. [Hsu and Liu, 2003]. We recreated Figure 1e by separating the negative ENSO-IODs into two periods (1948 1977 and 1978 2002) and found that no significant difference exist between the two composites. We further examined the zonally averaged vertical circulation over 25 N 35 N to investigate whether or not the subsidence over southern China is related with the topographic forcing (not shown) and found that the vertical circulation anomaly in the eastern flank of Tibetan Plateau is insignificant suggesting that the rain band anomalies shown in Figure 1e is primary attributed to the SSTA related with IOD and ENSO and less to the interdecadal variability and to the topographic forcing. [38] The north-south orientated area of east Asian monsoon rainfall anomaly due to IOD is also described by Kripalani et al. [2005], who focused on the delayed impacts (three to four seasons) of positive IOD on east Asian monsoon. Since some of the negative IODs are followed by positive IOD events, this lagged impact is consistent with our results for the IOD years when the Indian Ocean SSTA is dominated by TBO (tropical biennial oscillation). However, the simultaneous impacts of negative IODs on the rainfall over China and Korea-Japan become important when TBO is weak, especially during the negative IOD years (1954, 1955, 1956, 1971). Particularly, the negative IOD combined with La Niña (negative ENSO- IOD) significantly influences east Asian summer rainfall by enhancing the local anomalous Hadley circulation over the northwestern Pacific. This particularly strong impact of IOD is only apparent for the negative phase of ENSO-IOD, which is only identified by classifying the IODs into four different types. The evolution of positive SSTA over IOD-E Figure 15. Scatter diagram of normalized SSTA and H125 (heat content anomaly averaged from 0 125 m) for IOD-E during the mature phase (SON). 13 of 15

Table 2. Characteristics of Positive and Negative Phases of Pure and ENSO IODs a Phase Evolution Positive Fast Pure IOD Negative Slow Positive Slower Than Pure IOD Positive ENSO-IOD Negative Faster Than Pure IOD Negative IOD-E Anomaly cold warm cold warm Initiation May December-1 April May April Peak October July October October Termination December Jan+1 December November Peak Amplitude 1.8 s 0.9 s 1.2 s 0.7 s Maintenance ocean adv. latent heat ocean adv. latent heat ocean adv. latent heat ocean adv. latent heat IOD-W Anomaly warm cold warm cold Initiation May Apr April-May April Peak Oct Oct December October Maintenance ocean adv. ocean adv. ocean adv. ocean adv. latent heat Termination Jan+1 December Summer+1 Feb+1 Peak Amplitude 0.5 s 0.7 s 1.3 s 1.0 s IOD-U Anomaly easterly westerly easterly westerly Initiation April December -1 March April Peak September October October October Termination December December May+1 Jan+1 Peak Amplitude 1.5 s 0.8 s 1.4 s 0.5 s a Season+1 indicates season of the following year, Season-1 indicates season of the previous year, and s is the standard deviation. during a La Niña year will be important in predicting the east Asian summer rainfall. [39] We found that the characteristics of the time evolution and spatial structure of IODs are very different between pure- IOD and ENSO-IOD. The positive and negative pure-iods also have different evolutions and structures. The characteristics of the IODs are summarized in Table 2 to facilitate the reader s understanding of the results described below. [40] For the pure IOD, the development of the positive phase is much more abrupt and quicker than the negative phase and acquires larger amplitude. Negative IOD tends to start earlier, but both terminate around the same time. Regarding the spatial structure, positive IOD has a much larger anomaly over IOD-E, while anomalies in both IOD-E and IOD-W are of about the same magnitude for negative IOD. The atmospheric anomaly pattern during the mature phase shows a typical Gill-type response to tropical forcing [Gill, 1980], agreeing with the modeling works by Baquero- Bernal et al. [2002] and Lau and Nath [2004]. [41] In contrast, the ENSO-IOD evolves in a quite different manner. The positive ENSO-IOD, which is always associated with El Niño, develops slowly compared to positive pure IOD, while the negative ENSO-IOD, which is associated with La Niña events, develops more quickly than the negative pure IOD. The peak amplitude of the anomaly is again weaker for negative IOD than for positive IOD as in the pure-iod cases. [42] Both ENSO and IOD together affect the SSTA over Indian Ocean when the two occur at the same time and it is hard to estimate the individual influence. The comparison of SSTA evolution between pure-iod and ENSO-IOD (Figure 5) shows that the SSTA over IOD-E is primary dominated by IOD for the development stage (JJA), whereas the ENSO remote forcing play a much important role in contributing to the SSTA over IOD-W and the basin scale warming in the following spring of IOD events. [43] The ocean mixed layer heat budget shows that the maintenance mechanisms of IOD-E and IOD-W are different. Both ocean advection and heat flux contribute to the SSTA over IOD-E, but the SSTA over IOD-W is primary driven by ocean advection. This result is consistent with the diagnostic study of an oceanic general circulation model by Li et al. [2002], although they examined only positive IOD events. We further point out the amplitude of total forcing (heat flux plus ocean advection) over IOD-E is larger for positive than for negative IOD. For IOD-E, the latent heat flux is more important in the initiation of IOD as suggested by the observation that the heat flux leads ocean advections by 2 3 months. It is also noted that El Niño is well correlated with the ocean subsurface temperature anomaly over IOD-E, but the impact of La Niña on heat budget is limited to the near-surface, which enhances only the latent heat. Whether this asymmetric influence is related with the asymmetry of ENSO is now under investigation. Finally, the findings in this paper are summarized in Table 3. It clearly shows what aspect of the IOD can be prolonged and Table 3. Temporal and Spatial Characteristics of Positive and Negative Phases of the IOD and Its Associated Processes of Mixed Layer Ocean Heat Budget Enhanced by El Niño Mixed Layer Heat Budget Months Indian Ocean Anomalies Positive IOD MAM IOD-E easterly AMJ IOD-W warm SSTA ocean advection MJJ IOD-E cold SSTA heat flux ASO yes IOD-E cold SSTA ocean advection SON yes IOD-W warm SSTA ocean advection Negative IOD MAM IOD-E westerly MJJ yes IOD-E warm SSTA heat flux JAS yes IOD-E warm SSTA ocean advection ASO yes IOD-W cold SSTA ocean advection SON yes IOD-W cold SSTA heat flux 14 of 15

enhanced by ENSO. Particularly note that the IOD-E SSTA warm anomaly can be intensified by La Niña during the East Asian summer monsoon season from May to September. The warm SST can intensify the subtropical anticyclone over the Philippine Sea and South China Sea and result in different signs of the precipitation anomaly over the east coast of South China and Taiwan for the negative IOD with and without ENSO. The concurrent negative IOD and La Niña can be favorable for the stationary Mei-Yu front over the Yangtze River that frequently generates disastrous floods. [44] The results presented in this paper are based on limited samples from observations, and some of the conclusions are still of a hypothetical nature, thus requiring further clarification with numerical models and analysis of ensemble long integrations. [45] Acknowledgments. This study was partially supported by NSC- 96-2745-M-002-010, NSC-96-2625-Z-052-008, and NOAA NA17RJ1231. The views expressed herein are those of the authors and do not necessarily reflect the views of NOAA. We thank Diane Boomer for the proofreading. Valuable suggestions by three reviewers are highly appreciated. References An, S.-I., and F.-F. Jin (2004), Nonlinearity and asymmetry of ENSO, J. Clim., 17, 2399 2412. Ashok, K., Z. Guan, and T. Yamagata (2003), A look at the relationship between the ENSO and the Indian Ocean dipole, J. Meteorol. Soc. Jpn., 81, 41 56. Ashok, K., N. H. Saji, and T. Yamagata (2004), Individual and combined influences of ENSO and Indian Ocean dipole on the Indian summer monsoon, J. Clim., 17, 3141 3155. Baquero-Bernal, A., M. Latif, and S. Legutke (2002), On dipolelike variability of sea surface temperature in the tropical Indian Ocean, J. Clim., 15, 1358 1368. Behera, S. K., J.-J. Luo, S. Masson, S. A. Rao, H. Sakuma, and T. Yamagata (2006), A CGCM study on the interaction between IOD and ENSO, J. Clim., 19, 1688 1705. Cai, W., H. H. Hendon, and G. Meyers (2005), Indian Ocean dipolelike variability in the CSIRO Marks 3 coupled model, J. Clim., 18, 1449 1668. Carton, J., G. Chepurin, X. Cao, and B. Giese (2000), A simple ocean data assimilation analysis of the global upper ocean 1950 95. Part I: Methodology, J. Phys. Oceanogr., 30, 294 309. Chang, C.-P., Y. Zhang, and T. Li (2000a), Interannual and interdecadal variations of the East Asian summer monsoon and tropical Pacific SSTs. Parts I: Roles of the subtropical ridge, J. Clim., 13, 4310 4325. Chang, C.-P., Y. Zhang, and T. Li (2000b), Interannual and interdecadal variations of the East Asian summer monsoon and tropical Pacific SSTs. Parts I: The meridional structure of the monsoon, J. Clim., 13, 4326 4340. Chowdary, J. S., and C. Gnanaseelan (2007), Basin-wide warming of the Indian Ocean during El Niño and Indian Ocean dipole years, Int. J. Climatol., 27, 1421 1438. Dommenget, D., and M. Latif (2002), A cautionary note on the interpretation of EOFs, J. Clim., 15, 216 225. Du, Y., T. Qu, G. Meyers, Y. Masumoto, and H. Sasaki (2005), Seasonal heat budget in mixed layer of the southeastern tropical Indian Ocean in a high-resolution general circulation model, J. Geophys. Res., 110, C04012, doi:10.1029/2004jc002845. Gill, A. E. (1980), Some simple solutions for heat-induced tropical circulation, Q. J. R. Meteorol. Soc., 106, 447 462. Guan, Z., and T. Yamagata (2003), The unusual summer of 1994 in East Asia: IOD teleconnections, Geophys. Res. Lett., 30(10), 1544, doi:10.1029/2002gl016831. Harou, A. P., R. F. Lajoie, D. R. Kniveton, and M. R. Frogley (2006), The influence of the Indian Ocean Dipole mode on precipitation over Seychelles, Int. J. Climatol., 26, 45 54. Hsu, H.-H., and X. Liu (2003), Relationship between the Tibetan Plateau and East Asian summer monsoon rainfall, Geophys. Res. Lett., 30(22), 2066, doi:10.1029/2003gl017909. Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull. Am. Meteorol. Soc., 77, 437 472. Kripalani, R. H., and P. Kumar (2004), Northeast monsoon rainfall variability over south Peninsular India vis-á-vis the Indian Ocean dipole mode, Int. J. Climatol., 24, 1267 1282. Kripalani, R. H., J.-H. Oh, J.-H. Kang, S. S. Sabade, and A. Kulkarni (2005), Extreme monsoons over East Asia: Possible role of Indian Ocean Zonal Mode, Theor. Appl. Climatol., 82, 81 94, doi:10.1007/s00704-004-0114-z. Kug, J.-S., T. Li, I.-S. Kang, J.-J. Luo, S. Masson, and T. Yamgata (2006), Role of ENSO-Indian coupling on ENSO variability in a coupled GSM, Geophys. Res. Lett., 33, L09710, doi:10.1029/2005gl024916. Kurihara, K., and T. Tsuyuki (1987), Development of the barotropic high around Japan and its association with Rossby wave-like propagations over the North Pacific: Analysis of August 1984, J. Meteorol. Soc. Jpn., 65, 237 246. Lau, N.-C., and M. J. Nath (2004), Coupled GCM simulation of atmosphere ocean variability associated with zonally asymmetric SST changes in the Tropical Indian Ocean, J. Clim., 17, 245 265. Li, T., Y.-S. Zhang, C.-P. Chang, E. Lu, and D. Wang (2002), Relative role of dynamic and thermodynamic processes in the of the Indian Ocean dipole, Geophys. Res. Lett., 29(23), 2110, doi:10.1029/2002gl015789. Li, T., B. Wang, C.-P. Chang, and Y. Zang (2003), A theory for the Indian Ocean dipole-zonal mode, J. Atmos. Sci., 60, 2119 2135. Meyers, G. (1996), Variation of Indonesian throughflow and the El Niño- Southern Oscillation, J. Geophys. Res., 101, 12,255 12,263. Nitta, T. (1987), Convective activities in the Tropical Western Pacific and their impact on the Northern Hemisphere summer circulation, J. Meteorol. Soc. Jpn., 65, 373 390. Ropelewski, C. F., and M. S. Halpert (1987), Global and regional scale precipitation patterns associated with the El Nino/Southern Oscillation, Mon. Weather Rev., 115, 1606 1626. Saji, N. H., and T. Yamagata (2003a), Structure of SST and surface wind variability during Indian Ocean dipole mode events: COADS observations, J. Clim., 16, 2735 2751. Saji, N. H., and T. Yamagata (2003b), Possible impacts of Indian Ocean dipole mode events on global climate, Clim. Res., 25, 151 169. Saji,N.H.,B.N.Goswami,P.N.Vinayachandran,andT.Yamagata (1999), A dipole mode in the tropical Indian Ocean, Nature, 401, 360 363. Smith, T. M., R. W. Reynolds, R. E. Livezey, and D. C. Stokes (1996), Reconstruction of historical sea surface temperature using empirical orthogonal functions, J. Clim., 9, 1403 1420. Trenberth, K. E. (1997), The definition of El Niño, Bull. Am. Meteorol. Soc., 78, 2771 2777. Uppala, S., P. Kallberg, A. Simmons, M. Fiorino, A. Hernandez, X. Li, K. Onogi, S. Saarinen, and N. Sokka (2005), The ERA-40 re-analysis, Q. J. R. Meteorol. Soc., 131, 2961 3012. Vinayachandran, P. N., N. H. Saji, and T. Yamagata (1999), Response of the equatorial Indian Ocean to an unusual wind event during 1994, Geophys. Res. Lett., 26, 1613 1616. Webster, P. J., A. M. Moore, J. P. Loschnigg, and R. R. Leben (1999), Coupled ocean-atmosphere dynamics in the Indian Ocean during 1997 1998, Nature, 401, 356 360. Xie, S. P., H. Annamalai, F. A. Schott, and J. P. McCresry (2002), Structure and mechanisms of South Indian Ocean climate variability, J. Clim., 15, 864 878. Yu, J.-Y., and K.-M. Lau (2004), Contrasting Indian Ocean SST variability with and without ENSO influence: A coupled atmosphere-ocean study, Meteorol. Atmos. Phys., 90, doi:10.1007/s00703-004-0094-7. Yu, L., and M. M. Rienecker (1999), Mechanisms for the Indian Ocean warming during the 1997 98 El Niño, Geophys. Res. Lett., 26, 735 738. Yu, L., X. Jin, and R. A. Weller (2007), Annual, seasonal, and interannual variability of air-sea heat fluxes in the Indian Ocean, J. Clim., 20, 3190 3209. Zhong, A., H. H. Hendon, and O. Alves (2005), Indian Ocean variability and its association with ENSO in a global coupled model, J. Clim., 18, 3634 3649. Zubair, L., A. R. Suryachandra, and T. Yamagata (2003), Modulation of Sri Lankan Maha rainfall by the Indian Ocean Dipole, Geophys. Res. Lett., 30(2), 1063, doi:10.1029/2002gl015639. C.-C. Hong, Department of Science, TMUE, 1 Ai Kuo West Road, Taipei, Taiwan 10001. (cchong@tmue.edu.tw) M. Kanamitsu, Scripps Institution of Oceanography, University of California, San Diego, 8605 La Jolla Shores Drive, La Jolla, CA 92093-0224, USA. M.-M. Lu, Central Weather Bureau, No. 64, Gongyuan Road, Taipei, Taiwan 10048. 15 of 15