Interannual Variation of the Asian-Pacific Atmospheric System in Association with the Northern Summer SST Changes

The purpose of this study is to analyze the dynamic and hydrological characteristics of the interannual variability of the northern summer (June­ August) ocean-atmosphere system in the Asian-Pacific region. In this ocean­ atmosphere system, there are two types of interannual variability modes. As indicated by the sea surface temperature (SST) variability, the first type is related to the variations of the mature phase of the El Nifio-Southern Oscillation (ENSO) events. Its temporal variability is characterized by al­ ternations between the maximum phases of the El Niii.o and La Nifia events. Its spatial structure is characterized by an elongated positive (negative) SST anomaly over the tropical eastern Pacific during the El Nifio (La Nifia) event. The second type is related to SST variability between the developing and decaying stages of the ENSO events. This mode is characterized by warm (cold) SST anomalies in the tropical central and eastern Pacific dur­ ing the developing stage of the El Nifio (La Nifia) event, and warm (cold) SST anomalies near the Peruvian coast during the decaying stage. In accordance with these two types of interannual SST variability, tropi­ cal convection and the upward branches of Walker circulation are found enhanced (suppressed) in association with the warm (cold) SST anomalies. The centers of tropical convection anomaly coincide well spatially with the centers of major vertical motion branches and SST anomalies in the tropi­ cal western Pacific. The centers are to the west of the centers of vertical motion branches and SST anomalies in the central and eastern Pacific. In the atmospheric system, the lower-tropospheric circulation anoma­ lies corresponding to the first interannual mode contain spatial structures largely opposite to the climatological mean circulation. These anomalies represent the weaker Asian low and Pacific subtropical high during the El Nifio event, which lead to weaker tropical monsoon westerlies and Pacific trade winds. The lower-tropospheric circulation anomalies corresponding to the second interannual mode are characterized by an anomalous low 1 Research and Development Center, Central Weather Bureau, Taipei, Taiwan *Corresponding author address: Dr. Jau-Ming Chen, Research and Development Center, Central, Weather Bureau, 64 Kung-Yuan Rd. Taipei, Taiwan, 100; E-mail: cjming@rdc.cwb.gov.tw 833 834 TAO, Vol. 11, No. 4, December 2000 centered in the western Pacific during the developing stage of the El Niiio event. This anomalous low later develops into an anomalous high during the decaying stage. For both types of interannual mode, water vapor con­ vergence toward the convection-enhanced region is observed. Such conver­ gence results in an increase in atmospheric water vapor and thus maintains the positive precipitation anomalies. Enhanced precipitation and tropical convection are found embedded in the lower-tropospheric anomalous lows and accompanied by intensified transient activity. Water vapor divergence, negative precipitation anomaly, and weaker transient activity are found for the convection-suppressed region. (


INTRODUCTION
In the Asian-Pacific region, the most significant interannual variability in the oceanic system is the EI Nifio-La Nifia oscillation. In the El Nifi o phase, the sea surface temperature (SST) exhibits strong warm anomalies along the equatorial eastern Pacific and minor cold anomalies over the tropical western Pacific (Rasmusson and Carpenter 1982). These SST anoma lies reverse their polarity in the La Nifia phase. The warm/cold SST anomalies are accompa nied with enhanced/suppressed tropical cumulus convection that further couples with the up ward/downward branch of the tropical Walker circulation anomaly (e.g., Deser and Wallace 1990; Wang 1992). The Walker circulation has been described as being composed of the planetary-scale divergent circulation (also called irrotational circulation) (Krishnamurti 1971).
Moreover, the divergent circulation is generally paired with the rotational circulation to depict the atmospheric circulation system (Sardeshmukh and Hoskins 1988; Chen and Chen 1990).
These dynamic relationships imply that variability of the Walker circulation is connected with variability of the planetary-scale atmospheric circulation. They also suggest that SST variabil ity can affect the atmospheric variability through the coupling of the tropical convection and Walker circulation.
The interannual SST anomalies associated with the El Niiio or La Niiia events generally reach their maximum intensity in the northern winter (December-February). Therefore, the relationship between interannual SST anomalies and atmospheric variability in the northern winter has been extensively studied and is better understood than in the other seasons (e.g., Horel and Wallace 1981;Blackmon et al. 1983; Graham et al. 1994). Recently, the interna tional meteorological community began to stress the importance of the dynamic relationship between the interannual SST variations and northern summer (June-August) atmospheric systems, in particular the Asian monsoon (CLIVAR 1995). Ju and Slingo (1995) and Li and Yanai (1996) pointed out that during the El Nifio phases, the Asian summer monsoon tended to have a delayed onset and lower monsoon rainfall. These fe atures were accompanied by a broader-scale reduction in the intensity of monsoon circulation and water vapor convergence toward the Asian monsoon region (Yang and Lau 1998). These results suggest that the El Nino-related Pacific SST anomalies may be systematically linked with the interannual vari ability of the northern summer atmospheric system in the Asian-Pacific region through the variability of monsoon precipitation and circulation.
The above reviews reveal that past studies dealing with the interannual variability of the northern summer ocean-atmosphere system paid particular attention to the relationships be tween the Pacific SSTs, monsoon precipitation, and atmospheric circulation. In addition to these aspects, the interannual characteristics of other atmospheric processes require further investigation. For example, the precipitation anomaly was found important in the connection between Asian monsoon circulation and Pacific SSTs. Nevertheless, how the interannual vari ability of the precipitation was maintained is still not clear. To answer this question, we need to examine the interannual variability of the hydrological processes. In another aspect, the synoptic-scale transient activity was found to vary coherently with the planetary-scale atmo spheric circulation anomalies in the northern winter (Lau 1988; Matthews and Kiladis 1999).
However, the dynamic relationships between the interannual variations of transient activity and atmospheric circulation in the northern summer still remain unclear. Further studies are surely needed to understand the interannual variability of the northern summer ocean-atmo sphere system.
The purpose of this study is to analyze the dynamic and hydrological characteristics of the interannual variability of the northern summer ocean-atmospheric system. In this study, we analyze the interannual characteristics of the following oceanic and atmospheric fields: SST, tropical cumulus convection inferred from Outgoing Longwave Radiation (OLR), Walker circulation, atmospheric circulation, transient activity, and hydrological processes. Each of these fields is selected for a particular reason. SST anomalies are employed as the reference to define the phases of interannual variability. The tropical cumulus convection and Walker cir culation are used to portray the response of the atmosphere to the underlying SST forcing.
Atmospheric circulation and transient activity are employed to illustrate the response of the atmospheric flow system to the summer SST changes. Hydrological processes are analyzed in order to determine how moisture fields maintain the interannual precipitation anomaly. By incorporating the above analyses, we hope to gain a more comprehensive understanding of the dynamic characteristics of the interannual variability of the northern summer ocean-atmo sphere system.
The outline of this study is as follows. Data used to portray the northern summer interannual modes are described in Section 2. The climatology and interannual variability of SSTs are presented in Section 3. Anomalies of the tropical cumulus convection and Walker circulation associated with the summer SST changes are discussed in Section 4. In Sections 5,6, and 7, climatologies and interannual variations of atmospheric circulation, transient activity, and hydrological processes, respectively, are analyzed. Concluding remarks are in Section 8.

DATA
The following monthly-mean data for the period 1979-95 are analyzed in this study: (a) SST and OLR compiled by the Climate Prediction Center of the National Centers for Environ mental Prediction (NCEP); (b) moisture and upper air wind fields of the NCEP/NCAR (National Center for Atmospheric Research) reanalysis data (Kalney et al. 1996); and (c) precipitation estimates compiled from rain gauge measurements and satellite data (Xie and Arkin 1996). SST data are in a 2° x 2° grid and the other data are in a 2.5° x 2.5° grid. In all analyses, the northern summer (June-August) mean fields are used. In previous studies, it was found that an interdecadal trend has existed in the ocean-atmosphere system during the past four decades (e. g., Deser and Blackmon 1993;Jones 1994). In order to examine the interannual variability phe nomenon specifically, the interdecadal trend embedded in the 1979-95 northern summer mean fields is removed from the analyses using a least-square-fit method.

SUMMER SSTS
In order to make a better interpretation of the dynamic characteristics of the interannual SST anomaly, the spatial pattern of the interannual SST anomaly is compared with the clima tological SST pattern. The SST climatology of the 1979-95 northern summer over the Asian Pacific region is shown in Fig. la. This SST climatology displays a major warm center (e.g., above 28°C) over the tropics and northern subtropics of the western Pacific. Warm SSTs extend westward to the oceans surrounding the Indian subcontinent and eastward to the central Pacific. Over the eastern Pacific, a belt of warm SSTs spreads just north of the equator in parallel with the overlying Intertropical Convergence Zone (ITCZ). To the south of this warm SST belt, tropical SSTs decrease signific antly from the date line toward the east end of the Pacific. This SST distribution forms a cold tongue over the tropical eastern Pacific. In the tropics, the northern summer SST climatology is warmer in the western than in the eastern Pacific.
In the analyses of the interannual characteristics of the northern summer SSTs, we first evaluate the intensity of the interannual variab ility which is inferred from the root-me an-square (RMS) value of SST anomalies (L).SST). The L). denotes a deviation from the 1979-95 sum mer mean. The RMS distribution of the 1979-95 northern summer SST anomalies over the Asian-Pacific region is displayed in Fig. lb. Its salient feature is a maximum center over the eastern-Pacific Nifio-3 region (i.e., 150°W-90°W, 5°S-5°N). Since the SST value averaged over the Nifio-3 region is used as an index to depict the evolution of the El Nino-Southern Oscillation (ENSO) phenomenon, the existence of the SST RMS maximum center over the Nifi o-3 region implies the dominance of ENSO-related SST variability in the interannual varia tion of the northern summer SSTs. In contrast to the noticeable RMS pattern over the eastern Pacific, the RMS pattern over the western Pacific and Indian Oceans is much weaker and not organized. Apparently, the major center of action for the interannual variability of the northern summer SSTs resides over the tropical eastern Pacific.
The primary temporal characteristics of the northern summer SST changes may be illus trate d by the time series of L). SST anomalies averaged from the maximum RMS center, i.e., the Nifio-3 region. This L). SST time series (Fig. le) has a standard deviation of 0.804°C. Using   in the periods 84/9-85/6 and 88/5-89/6. The above phase relationships demonstrate clearly that there is an ENSO signature on the interannual variability of the northern summer SSTs.
In order to examine the spatial and temporal characteristics of the northern summer SST changes simultaneously, SST anomalies are subjected to the Empirical Orthogonal Function (EOF) analysis. The first two EOF eigenmodes of the interannual SST anomalies over the tropical Asian-Pacific region are displayed in Fig. 2   In the first SST eigenmode, the first eigenvector (El; Fig. 2c) shows a spatial structure resembling the typical ENSO pattern. This is a major positive center stretching from the east ern-Pacific Nifio-3 region westward to the equatorial central Pacific, and it is surrounded by horseshoe-shaped negative anomalies with weaker magnitudes over the tropical western Pa cific and the subtropical eastern Pacific. With such a typical ENSO pattern, the El(SST) pat tern resembles the SST RMS pattern in Fig. 1 b. The El(SST) also shows that, during its posi tive phase, the SST anomaly is warmer in the eastern Pacific and colder in the western Pacific. Such a longitudinal distribution in the tropical Pacific is out of phase with SST climatology (Fig. la). This result indicates that the first SST mode tends to weaken (enhance) the east-west contrast of the northern summer tropical Pacific SSTs during its positive (negative) phase.
The temporal variation of the first SST eigenmode, as illustrated by the first eigencoefficient (Cl; Fig. 2a), seems to vary coherently with the Nifio-3 LiSST time series in Fig. le. Cl(SST) contains maximum phases in summers of 82,83,87 and 91 and minimum phases in those of 84, 85 and 88. These maximum (minimum) phases correspond to the relatively warm (cold) phases of the Nifio-3 Li SST time series. The temporal coherence between Cl(SST) and the Nifi o-3 Li SST time series is statistically substantiated by a 0.92 correlation coefficient. According to these temporal features, we conclude that the dominant interannual variability in Cl(SST) is associated with the oscillations between the major El Nifi o and La Nifia events (or the mature phases of ENSO events). Based upon the spatial and temporal features depicted by the El (SST) and Cl(SST), we interpret the first SST eigenmode as an ENSO-related mode which reflects the interannual SST variability caused by the alternation between the major El Nifio and La Nifia events. This SST mode also reveals that the intensity of tropical Pacific SSTs is weakened (enhanced) during the mature phase of the El Nifio (La Nifia) event.
The second SST eigenmode has only about one third the intensity of the first, as inferred from the percentage of variance. The second eigenvector (E2; Fig. 2d) features a major nega tive center over the tropical central Pacific, and two major positive centers over the South China Sea-western Pacific region and the eastern Pacific adjacent to the Peruvian coast. The amplitudes of these three centers are comparable. E2(SST) thus shows a three-center pattern which differs from the single-dominant-center pattern of El(SST).
The temporal variation of the second eigencoefficient (C2; Fig. 2b) exhibits large nega tive phases in the summers of 82 and 86, and large positive phases in the summers of 83 and 88. These summers happen to be in the periods around the major El Nifio events. The summers of 82 and 83 are in the developing and decaying stages, respectively, of the 82/4-8317 El Nifio event. Similarly, the summers of 86 and 88 are in the prior and posterior periods, respectively, of the 86/8-88/2 El Nifio event. These temporal relationships suggest that the major negative (positive) C2(SST) phase corresponds to the developing (decaying) stage of the major El Nifi o event. This type of phase relationship also applies to another major El Nifio event during 9113-9217. C2(SST) switches from a negative phase in the summer of91 into a positive phase in that of 92, but with a much weaker change in amplitude. On the other hand, a reversed phase relationship is found for the La Nifi a events. During the major La Nifia events in the periods of 84/9-85/6 and 88/5-89/6, C2(SST) fluctuates from positive phases in the developing stages during the summers of 84 and 88 into relatively negative phases in the decaying stages during those of 85 and 89. According to the spatial and temporal structures of the second SST mode, one can infer that warm SST anomalies are located in the central Pacific during the developing phase of the El Nifio event (negative E2, negative C2). Warm SST anomalies later move to the eastern Pacific during the decaying phase (positive E2, positive C2). Based upon the above analyses, we conclude that the second SST mode represents the interannual SST variability induced by changes in the summer SST anomalies between the developing and decaying stages of ENSO events.
Our EOF analyses reveal the existence of significant ENSO signatures in the interannual variability of the northern summer SSTs. The dominant interannual variability of the first SST mode is related to variability of the tropical Pacific SS Ts between major El Nifio and La Nifia events. In fact, this kind of SST variability also dominates the interannual variability of the northern winter SSTs and has already attracted research attention to study its impact on the northern winter atmospheric system (e.g., Lau and Nath 1994; Lau et al. 1998). On the other hand, the second SST mode is often ignored because of its weak intensity. However, our analysis suggests that this SST mode is important because it represents the interannual SST variability resulting from the variability of SSTs between the developing and decaying stages of ENSO events. Therefore, it is of interest to study the impacts of the first two SST modes on the northern-summer atmosphere.

INTERANNUAL VARIABILITY OF WALKER CIRCULATION AND OLR
In this section, we analyze anomalies of the Walker circulation and tropical cumulus con vection in order to examine the impacts of the interannual SST mode on the vertical motion of the atmosphere. The Walker circulation anomalies associated with the first and second SST modes are shown in Figs. 3a and 3b, respectively. To construct these anomalies, we first com pute two types of composites based upon the negative phases or positive phases of each SST eigencoefficient. The difference between these two composites (positive phase -negative phase) is considered as the typical response of the Walker circulation to the positive phase of the interan rual SST mode. The Walker circulation is expressed by a mass flux function 'I' m = P i P,uddp (Newell et al. 1974), where p1 = 100 mb, p0 = 1000 mb, uct is the zonal divergent wind, and pi s the pressure. A positive (negative) '¥ m cell indicates a two-dimen sional counterclockwise (clockwise) motion. 'I'm anomalies associated with the first/second SST mode are extracted from the longitude-height section at the equator/10°S in order to portray vertical motions above the one-center/three-center pattern in the first/second SST eigenvector. As shown in Fig. 3a, '¥ anomalies related to the first SST mode contain one

841
east. In association with these two 'Pm cells, the updraft sections are located at around 120°W and 90°E and match the tropical warm E2(SST) anomalies (see Fig. 2d) over the eastern Pa cific and Indian Ocean, respectively. Meanwhile, the downdraft section is located at around 160°E and corresponds to an underlying cold E2(SST) anomaly in the tropical central Pacific.
The above spatial relationship describes that the major updraft (downdraft) branches of the Walker circulation anomaly match the anomalous warm (cold) SSTs.
The interannual variability of tropical cumulus convection associated with the interannual SST modes is inferred from OLR anomalies constructed in the same manner as (m anomalies.
OLR anomalies corresponding to the fi rst SST mode ( We also compare the spatial patterns of OLR and 'Pm anomalies and find some interest ing results. In both modes, the centers of positive OLR anomalies over the tropical western Pacific are located within the major downdraft sections of '¥ m anomalies, such as the sections at around 130°E in Fig. 3a and 160°E in Fig. 3b. However, the centers of negative OLR anoma lies over the tropical central Pacific in the vicinity of the date line are located westward about 30°-50° longitude from the major updraft sections of 'Pm anomalies, such as the sections at around 150°W in Fig. 3a and 120°W in Fig. 3b. These spatial relationships indicate that the centers of OLR anomalies and tropical vertical motion are more consistent in geographical distribution in the western Pacific than in the central and eastern Pacific. The above results may be attributed to the longitudinal dependence of the air-sea interaction processes (Wang 1992). In the tropical western Pacific, the atmospheric stability is weak. Vertical motion and convection are both sensitive to changes in external boundary condition such as the underlying SSTs. This characteristic leads to a spatial consistency among the anomalous centers of these fields. On the other hand, the tropical atmosphere over the central and eastern Pacific becomes more stable because it is underneath the subsidence of the climatological Walker circulation.
In this region, the pressure gradient force associated with the SST gradient induces the bound ary-layer convergence that determines atmospheric heating. The OLR center tends to match the locations of atmospheric heating and the zonal SST gradient, rather than the center of SST anomalies. Therefore, the anomalous OLR centers over the central and eastern Pacific are located westward from the anomalous SST center and the major branch of the Walker circula tion anomaly. that the low-latitude X E (850 mb) pattern is sensitive to the thermal structure of the underlying sea surface condition. Thus, the X E (850 rob) field is often used to illustrate the distribution of the atmospheric heat source (e.g., convergence sector) and heat sink (e.g., divergence sector) in accordance with the external heat conditions.
In the rotational circulation, the salient feature of the climatological 'P E (850 mb) field ( Fig. 4b) is two east-west pairs oflows and highs. In the northern hemisphere (NH), '¥ 8 (850 843 mb) is dominated by a monsoon low over the Asian continent and a subtropical high over the North Pacific. This low-high pair is accompanied by another low-high pair in the southern hemisphere (S H), but with opposite polarity. The above 'l'E (850 mb) patterns contain a unique phase relationship with the X E (850 mb) patterns in Fig. 4a. In each hemisphere, the east-west pair of low and high reside over two sides of the maximum X E (850 mb) convergence center around the Philippines. This result indicates that the maximum phase of X E (850 mb) (i.e., convergence center) more or less matches the node phase of 'l' E (850 mb), where the zero contour line is located. Their phase difference, about one quarter wavelength or 90 degrees, reveals the existence of a spatially quadrature relationship between the X E (850 mb) and 'l' E (850 mb) fields. Since a quadrature relationship was found between the atmospheric waves and the idealized tropical heat source that excites these waves (Matsuno 1966;Gill 1980), the quadrature relationship is usually interpreted as existing between the heat source field (like X E field) and its responding atmospheric wave fields (like 'l'E field). The interannual variability of the planetary-scale atmospheric circulation is illustrated by the first two eigenmodes of X l850 mb) and 'l'E (850 mb ). As displayed in Figs Fig. 2d). Over the Indian Ocean and South Ch ina Sea, the El[X E (850 mb)] exhibits a conver gent (positive) center which corresponds to a warm E2(SST) anomaly. It is clear that the convergent (divergent) center of the X E (850 mb) field is connected with the warm (cold) SST anomaly. We also compare the spatial patterns of El[Xi850 mb)] and El[ 'P E (850 mb)J (Fig.  Sd). In the NH, the El[X E (850 mb)] divergent center around 150°W is surrounded by a pair of El['P E (850 mb)] anomalies with an anomalous high to the west and an anomalous low to the east. In the SH, this divergent center is sandwiched by another east-west pair of 'P E anomalies with opposite polarity to the NH 'P E anomalies. It is also clear that there is a spatially quadra ture relationship between El[ X E (850 mb)] and El[ 'P E (850 mb)J. The above spatial relation ships between E2(SST), El [ X E (850 mb)], and El[ 'P E (850 mb)] suggest that the interannual SST variability can exert impacts on the interannual variability of the atmospheric waves (inferred from El[ 'P E (850 mb)]) by affecting the heat distribution of the summer atmosphere  'P E (850 mb)(c-d).
In the spatial structure, the E2[ X 6 (850 mb) J pattern (Fig. 6b) shows an anomalous con vergence region (positive anomaly) over the eastern Pacific with a center at around 120°W, wh ich corresponds to an underlying tropical warm El(SST) anomaly (see Fig. 2c). Over the western and central Pacific, an anomalous divergence region (negative anomaly) is accompa nied by the relatively cold El(SST) anomaly. It appears that the convergence (divergence) region in E2[ X6(850 mb)J tends to match the underlying warm (cold) El(SST) anomaly. Such a spatial relationship is consistent with that found previously between E2(SST) and El[ 'l'E (850 mb)]. We also examine the spatial relationship between E2[X 6 (850 anomalous low over the North Pacific. Such an east-west distribution is largely opposite to 'l' E (850 mb) climatology (Fig. 4b), which exhibits a low in Asia and a subtropical high in the North Pacific. In the SH, the east-west distribution of the circulation in the Pacific region also tends to be opposite between E2[ 'PE (850 mb)] and 'l' E (850 mb) climatology. Note that the second 'l' E (850 mb) mode is connected to the first SST mode. During the El Nino phase, this SST mode exhibits warm anomalies over the central-eastern Pacific and cold anomalies over the western Pacific. Such an east-west SST distribution is also opposite to the summer SST climatology. Since the tropical SST patterns are reversed between the El(SST) and SST climatology, it is reasonable to find a reversal in the spatial pattern between the E2['¥ E (850 mb)] and '¥ E (850 mb) climatology. These results suggest that during the positive (negative) phase of the second 'l' E (850 mb) mode which corresponds to the El Nino (La Nina) phase, the East-Asian summer monsoon circulation becomes weaker (stronger). Wang (2000) conducted composite analyses with the observational data and found that during the El Nino (La Nina) phase, the tropical central Pacific warming (cooling) and the weak (strong) East-Asian mon soon is connected by an anomalous high (low) over the western Pacific, which is clearly shown in the E2[ 'l' E (850 mb)]. Thus, Wang's finding regarding the dynamic relationship between the East-Asian monsoon and ENSO is supported by our diagnostic results for the second 'PE (850 mb) mode and first SST mode.  (Fig. 7) shows a maximum center stretching from the subtropical western Pacific off the coasts of the Asian continent southeastward to the equatorial central Pacific. The subtropical western Pacific is the location of the climatological X E (850 mb) con vergence center (see Fig. 4a) where tropical cyclone activity is significant (McBride, 1995). Thus, the RMS( ' ¥ ; ) center over the subtropical western Pacific is likely to reflect the active transient variability associated with tropical cyclone activity. The other maximum RMS ( ' ¥ ; ) center over the equatorial central Pacific resides along the equatorial rim of the Pacific sub tropical high (see Fig. 4b) where tropical disturbances, such as easterly waves and tropical cyclones, move westward following the prevailing tropical trade winds (e.g., Wallace and Chang 1969). This result suggests that the RMS('¥ ; ) center over the tropical central Pacific is likely to be related to the frequent occurrence of the equatorial easterly waves and tropical cyclones.
The first two eigenmodes of RMS( '¥ ; ) are displayed in Fig. 8   This result reveals that the interannual variation of transient activity is more significant in the Pacific than in the Indian monsoon region.

CLIMATOLOGY AND INTERANNUAL VARIABILITY OF HYDROLOGICAL PROCESSES
The hydrological characteristics of the interannual variability of the atmospheric system are analyzed in this section. Hydrological processes of the atmosphere are often depicted by a water-budget equation: where W, Q, E, and P are the total atmospheric precipitable water, vertically-integrated water vapor flux, evaporation, and precipitation, respectively. Deser and Wallace (1990) pointed out that in the El Nifio phases, the precipitation anomaly over the tropical central Pacific is main tained primarily by the convergence of water vapor flux, rather than in-situ evaporation. Simmonds et al. ( 1999) also found that over China, the dominant terms in the interannual variability of the summer hydrological processes are atmospheric moisture convergence and precipitation. In view of theses findings, we thus focus our analyses of the hydrological pro cesses on precipitation, divergence of water vapor flux ( V • Q ), and precipitable water.
The northern summer climatologies of P, V • Q, and W are displayed in Fig. 9. The P pattern (Fig. 9a) can be divided into several heavy precipitation sectors. One sector covers the Asian monsoon region, with maximum centers over the Arabian Sea, Bay of Bengal, and South China Sea. Heavy precipitation spreads eastward along the tropical western Pacific east of the Philippines and extends farther to form two rain belts over the equatorial central Pacific: the ITCZ and the South Pacific Convergence Zone (SPCZ). To supply moisture for the afore mentioned heavy precipitation, V • Q (Fig. 9b) exhibits a convergence of water vapor flux (negative pattern) in the Bay of Bengal, South China Sea, tropical western Pacific, and SPCZ. Water vapor also converges into the central-Pacific ITCZ, except that its intensity is quite weak. Following the water vapor convergence, large amounts of atmospheric moisture, as interpreted from the precipitable water field in Fig. 9c, accumulate in heavy precipitation re gions and provide the moisture source for precipitation. This relationship leads to a resem blance in spatial structure between the P and W fields. The above hydrological processes reveal that the convergence of water vapor flux increases atmospheric moisture content to maintain heavy precipitation. The interannual variability of hydrological processes is portrayed by the first two eigenmodes of P, V • Q, and W fields. The first eigenmodes of these fields are shown in Fig.  10. For the temporal variation, time series of C 1 (P) (Fig. lOa), C 1 ( V • Q) (Fig. lOc ), and C 1 (W) (Fig. lOe) contain common features such as large positive phases in the summers of 82, 87 and 91 and large negative phases in those of 84, 85 and 88. These temporal features are consis tent with that of C 1 (SST) (see Fig. 2a), except for a discrepancy in the phase of the summer of 83 in which Cl(SST) contained a positive maximum. Nevertheless, a large positive phase is found in Cl(P), but there is a minor positive phase in Cl(W) and even a minor negative phase in C 1 ( V • Q ). The above differences suggest that in the El Nino summer of 83 variations in SST and moisture fields, such as W and V • Q, have a relationship somehow different from those happening in other major El Nino summers. In spite of the phase discrepancy in the summer of 83, the correlation coefficients between Cl(SST) and Cl(P), Cl(V • Q), and Cl (W) are 0.86, 0.78, and 0.84, respectively. These high correlation coefficients suggest that the first eigenmodes of P, V • Q, and W are in connection with the first SST eigenmode, which represents the interannual SST variability induced by alternations between the El Nino and La Nina events.
In the spatial structure, the E 1 (P) pattern ( Fig. 1 Ob) is noticeable by two major negative P anomalies over the southeast-Asian maritime continents and the northwestern Pacific east of the Philippines and Taiwan. Between these two negative anomalies, a major positive P anomaly is located over the equatorial central Pacific with a center west of the date line. The negative P anomaly over the southeast-Asian maritime continents is accompanied by a divergence of water vapor flux as indicated by a positive center in El ( V • Q) (Fig. lOd). It is also linked to a negative center in El(W) (Fig. lOf) which indicates a decrease in the atmospheric water vapor content. A similar situation occurs over the northwestern Pacific east of Taiwan. A divergent (positive) El( V • Q) anomaly and a decreased (negative) El(W) anomaly accom pany the negative P anomaly there. However, the intensity of the El( V • Q) anomaly and El (W) anomaly over the northwestern Pacific is weaker than that over the southeast-Asian mari time continents. For the maj or positive P anomaly over the tropical central Pacific , it is located in an environment with extra water vapor content (positive El(W) center in Fig. lOf)  decrease) in atmospheric water vapor following the convergence (divergence) of water vapor flux. These dynamic relationships among the interannual hydrological modes are consistent with those found from the climatological modes.
Since the first P mode is related to the first SST mode, it is of interest to compare the El (P) pattern with the spatial patterns of the other interannual modes associated with the first SST mode, such as the OLR anomaly in Fig . 3c, E2[RMS( 'f'; )J in Fig. 8d, and ['f'E(850 mb)] in Fig. 6d. Over the northwestern Pacific, the negative El(P) anomalies east of the Philippines are in accordance with the suppression of the tropical convection (positive OLR anomaly) and transient activity (negative E2[RMS( '¥; )] anomaly). All of th ese anomalies are located in an 850 mb anomalous high background (positive E2[ '¥ E (850 mb)] anomaly). A similar spatial relationship between the aforementioned fields occurs over the southeast-Asian maritime con tinents where another major negative E 1 (P) anomaly exists . As for the positive E 1 (P) anomaly, the one over the tropical central Pacific west of the date line matches well with the enhanced tropical convection (negative OLR anomaly). However, the enhanced transient activity (positive E2[RMS( '¥; )] anomaly) and an anomalous low (negative E2[ 'l' E (850 mb)] anomaly) are located somewhat to the north of this positive El(P) anomaly. The above spatial relationships indicate that the hydrological fields exhibit a clear dynamic connection with the large-scale convection, atmospheric circulation, and transient activity in the interannual time scale.
In the second set of hydrological modes, the temporal fluctuations of C2(P) (Fig. 1 la), C2 ( V • Q) (Fig. l lc), and C2(W) (Fig. 1 le) are not as coherent with one another as those found from the first modes. However, these time series still contain some common features. They all undergo evident phase changes between the developing and decaying periods of the maj or El Nifio events. For the 82/4-83/7 and 86/8-88/2 El Nifio events, these eigencoefficient time se ries vary from negative phases in the summers of 82 and 86 into positive phases in those of 83 and 88. For the third major El Nino event during the period 91/3 -92/7, changes in eigencoefficient phase from a negative state in the summer of 91 into a positive state in that of 92 occur in C2(P) and C2( V • Q ), but not in C2(W). Note that the phase change around the maj or El Nifi o events is the salient feature of the second SST mode. We thus compute correla tion coefficients between the second SST mode and the second hydrological modes to assess the degree of temporal coherence for these modes. The correlation coefficients between C2 (SST) and C2(P), C2( V • Q), and C2(W) are 0.70, 0.69, and 0.77, respectively. Since the correlation coefficients are _so high (about 70% ), it is legitimate to consider that the second set of hydrological modes correspond to the interannual SST variability resulting from contrasts between the developing and decaying stages of the major ENSO events, as interpreted from the second SST mode.
In the spatial structure, the salient pattern in E2(P) (Fig. 1 lb) is a major negative anomaly over the western Pacific with a center east of the Philippines. To the southwest of this negative P anomaly, there is a positive P anomaly centered at the southeast-Asian maritime continent.
In the moisture fields, a north-south stratified pattern stands out clearly in E2( V • Q) (Fig. l ld) and E2(W) (Fig. 1 lt). This stratified pattern suggests that water vapor is divergent (positive pattern) out of the North Pacific in the 10°N-25°N zone in connection with the decrease (negative anomaly) in precipitable water and precipitation. Water vapor is then transported southward to converge (negative pattern) over the equatorial zones in the vicinity of the southeast-Asian maritime continent where positive anomalies of precipitable water and precipitation are located. The above spatial relationships indicate that water vapor is convergent (divergent) to increase (reduce) the atmospheric water vapor content and leads to a favorable condition for the en hancement (suppression) of precipitation. There is no doubt that the mutual dynamic relation ship between the second modes of hydrological fields is consistent with that found from the first modes.
A further comparison is made between anomalies of precipitation, OLR, and atmospheric fl ows. Let us take the major negative E2(P) center east of the Philippines as the example. This·  Fig. 3d). This result suggests a systematic linkage between the second set of hydrological modes and the interannual modes of the atmospheric flows and large-scale convection activity.

CONCLUDING RE MARKS
In this study, we have analyzed the dynamic and hydrological characteristics of the interannual variability of the northern summer (June-August) ocean-atmosphere system in the Asian-Pacific region for the period 197 9-95. As indicated by the temporal and spatial features of the interannual SST anomalies, there are two types of interannual modes in the northern summer ocean-atmosphere system. The first type is characterized by the interannual variabil ity associated with variations of the mature phase of the ENSO events. Its maj or temporal fluctuation is an alternation between the El Nifi o and La Nifia events. Its salient spatial pattern is an elongated warm (cold) SST anomaly over the eastern-Pacific Nifio-3 region during the El Nifio (La Nifia) event. The second type is characterized by the interannual variability induced by contrasts between the developing and decaying stages of the ENSO events. Its salient fea ture is warm ( Based upon analysis results in this study, we construct a schematic diagram (Fig. 12) to illustrate the framework of the interannual variability of the northern summer ocean-atmo sphere system in the Asian-Pacific region. In the ocean sector, SST changes related to ENSO phenomena constitute the major variability in the first two dominant interannual modes. These interannual SST modes then exert systematic impacts on the atmospheric system due to the ocean-atmosphere coupling process, which is inferred from vertical motions associated with the tropical convection and Walker circulation. Following the coupling process, the dynamic and hydrological processes of the atmospheric system exhibit systematic linkages with the interannual SST variability. In the dynamic processes, the planetary-scale atmospheric  environments. Thus, it is likely that the dynamic processes and hydrological processes of the atmospheric system can connect together through the tropical transient activity and its accom panying precipitation. We further construct a table to summarize our analysis results. All atmospheric modes related to the first and second SST modes are separated into two groups and listed in Table 1. This table shows an interesting relationship: the atmospheric modes related to the first (second) SST mode are the most (second most) dominant modes in the hydrological fields (such as P, \7 • Q, and W), but the second most (most) dominant modes in the dynamic fields (such as XE (850 mb), 'PE (850 mb), and RMS( '¥;). In other words, the most significant interannual modes in the atmospheric flow system are related to the second SST mode, even if the intensity of the second SST mode is only one third of the first SST mode. Recall that the first SST mode contains significant anomalies in.the tropical eastern Pacifi c and no clear pattern in the west ern Pacific [see El(SST) in Fig. 2c] . On the other hand, the second SST mode has organized patterns in the South China Sea-western Pacific region [see E2(SST) in Fig. 2d]. The above discrepancy in SST spatial structure implies that the interannual variability of the northern summer atmospheric flow system in the tropical Asian-Pacific region is more sensitive to local SST variations near the western Pacific (as inferred from the second SST mode), than it is to remote SST variations in the eastern Pacific (as inferred from the first SST mode). Also listed in Table 1 are the correlation coefficients between any two fields selected from the same group. Correlation coefficients between the first SST mode and its associated atmospheric modes reach a level of about 0.8. For the second SST mode, correlation coeffi cients with the atmospheric modes are above 0.8 in the dynamic fields, but these are -lowed to around 0.7 in the hydrological fields. These correlation coefficients are indeed high enough and demand a reasonable explanation. Recall that both SST modes represent the tropical SST variability associated with the ENSO phenomena. The tropical SST variability adjusts the tropical vertical motions and thus affects the moisture transport between the atmosphere arid ocean. Variability in the moisture transport then leads to variability in the atmospheric water vapor content as well as the atmospheric hydrological processes. Therefore, through the mois ture exchange processes between the ocean and atmosphere, the interannual SST anomalies and the atmospheric hydrological fields are likely to be closely linked, which explains the high correlation coefficients between these two fields . Meanwhile, the moisture transport is a key factor in deciding the tropical precipitation. Since the latent heat release of tropical precipita tion forms the primary heat source for atmospheric motions, the tropical SST variability can change the atmospheric heating distribution by affecting moisture transport and its associated precipitation anomaly. Following the redistribution in atmospheric heating, the atmospheric flow fields (such as atmospheric circulation and transient activity) can change in accordance with the SST anomaly. These processes explain the high correlations between the interannual SST modes and atmospheric dynamic fields.
One interesting way to utilize the result of this study is to compare it with the interannual variability of the northern winter ocean-atmosphere system. Let us take the spatial pattern as an example for comparison. During the El Nino period, the tropical SST anomalies exhibit a similar pattern in winter artd summer: a strong warm anomaly along the eastern Pacific and a weak cold anomaly in the western Pacific. As mentioned before, this longitudinal pattern is largely opposite to the northern summer SST climatology. Following the phase reversal in tropical SST anomalies, the summer T E (850 mb) anomaly exhibits a spatial pattern primarily opposite to the summer T E (850 mb) climatology, as shownby the second T E (850 mb) mode in this study. In winter, the maximum center of the northern winter SST climatology is located in the central Pacific which forms a warm pool in the vicinity of the date line (e.g. , Philander 1 990), while the cold tongue still remains in the eastern Pacific. There thus exists a phase shift between the maximum centers of the northern winter SST climatology (i.e., central Pacific) and the El Nifio SST anomaly (i.e., eastern Pacific). Such a phase shift is not large enough to be out of phase as in the case of the northern summer. Following the phase shift in tropical SST patterns, a phase shift occurs between the climatology and interannual anomaly of the northern winter atmospheric circulation. Rogers and Raphael (1992) and others have shown that the northern winter 500 mb geopotential high anomaly during the El Nifi o period contains anomalous centers surrounding the climatological ridges and troughs, instead of overlapping them. The above comparisons reveal that the spatial patterns of the El Nino-related anomalies and the climatology are more or less opposite in the northern summer season, but only in certain phase shifts in the northern winter season.
This study analyzes only some selected processes of the interannual variability of the northern summer ocean-atmosphere system. Many more processes of this interannual vari ability system, including radiation processes and ocean dynamics, still need further analysis.
This study is regarded as a pilot study with which we hope to attract more research attention to study the northern summer ocean-atmosphere system.