Vietnam. Abstract: singular value decomposition; ENSO; sea-surface temperature; precipitation; South Asian summer monsoon

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 27: (2007) Published online 22 May 2007 in Wiley InterScience ( Short Communication Relationship between the tropical Pacific and Indian Ocean sea-surface temperature and monthly precipitation over the central highlands, Vietnam Tinh Dang Nguyen, a,c * Cintia Uvo b and Dan Rosbjerg a a Institute of Environment & Resources, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark b Department of Water Resources Engineering, Lund University, SE Lund, Sweden c Faculty of Planning & Management of Water Resources Development Systems, Water Resources University, 175 Tayson, Dongda, Hanoi, Vietnam Abstract: The relationship between monthly sea-surface temperature (SST) in the tropical Pacific and the Indian Ocean and monthly precipitation over the Vietnamese central highlands (VCH) has been investigated by means of singular value decomposition. The seasonal variation of SST plays a critical role in the onset of the monsoon season and convective rain band movement in the inter-tropical convergence zone. The relationships between precipitation and SST in both oceans vary significantly through the rainy season. In April, ENSO is strongly correlated with precipitation over the VCH, while the Indian Ocean SST only shows a significant correlation with precipitation in the northern VCH. In May, there is no significant relationship between precipitation and SST in either of the oceans. In June, precipitation over the VCH is negatively correlated with the northern Indian Ocean and the southern Pacific SST. Through July to September, no significant relationships were found between the Indian Ocean SST and precipitation patterns. The equatorial central to the eastern Pacific SST, in turn, is positively correlated with precipitation in a small area from the north to the south of the VCH. In ober, precipitation over the VCH is strongly related to ENSO and positively correlated with the equatorial eastern Indian Ocean SST. For November, the north-western Pacific as well as the equatorial eastern Indian Ocean SST is positively and strongly correlated with precipitation over the VCH. Lag-time analyses demonstrate a potential link between the Pacific SST and monthly precipitation patterns through the rainy season from one to three months in advance, and between the Indian Ocean SST and monthly precipitation patterns in ober and November from one to two months in advance. The analysed results provide a strong basis for a predictive scheme, but further analysis of skill levels needs to be developed. Copyright 2007 Royal Meteorological Society KEY WORDS singular value decomposition; ENSO; sea-surface temperature; precipitation; South Asian summer monsoon Received 4 November 2005; Revised 31 March 2006; Accepted 30 November 2006 INTRODUCTION The Vietnamese central highlands (VCH) region is located in the southeast of the Indochina Peninsula, between approximately N and E, and comprises approximately square km of rugged mountain peaks, widespread forests, and flat plateaus of basaltic land. The region is located in the tropics, as is the whole country. The major rainy season, from May to ober provides about 70 to 80% of the annual rainfall. The Indochina Peninsula is embedded in the Asian monsoon system and is flanked by both the Indian monsoon and the Southeast Asian monsoon (Chen and Yoon, * Correspondence to: Tinh Dang Nguyen, Faculty of Planning & Management of Water Resources Development Systems, Water Resources University, 175 Tayson, Dongda, Hanoi, Vietnam. dangtinh@wru.edu.vn 2000). This unique region has complex wind patterns, where the monsoon activity reflects a transitional feature of two distinct monsoon subsystems, the south Asian summer monsoon (SASM) and the east Asian winter monsoon (EAWM), responsible for a wet and a dry season, respectively (Zhang et al., 2002). The wet season occurs from early May to mid-ober, whereas the dry season is from November to late April. The transition from the dry to the wet season is characterized by a sudden increase in rainfall in late April. The processes controlling the weather over the Indochina Peninsula are complex, as described by Zhang et al. (2002). The physical mechanism responsible for the SASM onset is the seasonal transition of the atmospheric circulation over mid and low latitudes due to the rapidly warming landmass of Asia (He et al., 1987; Yanai et al., 1992). The full SASM circulation normally begins in Copyright 2007 Royal Meteorological Society

2 1440 T. D. NGUYEN, C. UVO AND D. ROSBJERG mid-may (Matsumoto, 1997) and its withdrawal takes about 4 months, being replaced by the EAWM at latitudes around 18 N in September, from 10 N to18 N inober, and south of latitude 10 N in November (Wang and Wu, 1997). The strong temperature gradient between the Asian continent and the oceans maintains the monsoon system (Yanai et al., 1992; Li and Yanai, 1996). Early rainfall over the Indochina Peninsula is associated with the appearance of a strong convection that indicates the start of the SASM (Lau and Yang, 1997) and is linked to the South China Sea (SCS) monsoon (Ding et al., 1996; Lau and Yang, 1997). The onset of SCS monsoon induces a deep convection from the SCS to the north-western Pacific including the Indochina Peninsula (Lau and Yang, 1997; Xie et al., 1998). A large area of deep tropical convection, extending from the equatorial eastern Indian Ocean to the equatorial western Pacific Ocean, expands abruptly northward until July, retreats southward after July, and ends at the southern Indochina Peninsula in November (Qian and Lee, 2000). The inter-annual variation of the SASM rainfall over the Indochina Peninsula is mainly based on disturbances in the local weather caused by (1) westward-propagating weather disturbances (e.g. tropical cyclones, perturbations, and monsoon lows) over the SCS by short wave train anomalies emanating from the western tropical Pacific (Saha et al., 1981; Chen and Weng, 1999; Chen and Yoon, 2000) and (2) an east west inter-annual seesaw response to the tropical Pacific sea-surface temperature (SST) anomalies, the so-called Walker circulation for the Indo Pacific inter-annual interaction (Chen and Yoon, 2000). The SASM rainfall over the Indochina Peninsula can be interpreted as a result of northward seasonal rain belt migration, in an east west oriented precipitation belt called the inter-tropical convergence zone (ITCZ), over the equatorial Indian Ocean in the Boreal summer. Positive SST anomalies over the eastern Indian Ocean maintain the ITCZ (Goswami and Shukla, 1984), and the annual cycle of SST plays a very important role in the annual evolution of the SASM rainfall (Fennessy and Shukla, 1994a,b). The SASM and El Niño-Southern Oscillation (ENSO) are not independent phenomena but part of a coupled ocean atmosphere oscillation. The interaction between ENSO and the SASM is complicated (Yasunari, 1990; Webster and Yang, 1992; Torrence and Webster, 1999) and has been known since the pioneering work of Sir Gilbert Walker (Walker, 1924). The influence of ENSO on the SASM was explained by Sikka (1980) and Shukla (1987) as being caused by a change in the Walker circulation and impacts on the regional Hadley circulation associated with the SASM (Webster and Yang, 1992; Goswami, 1998). The relationship between ENSO and the monsoon has noticeably weakened after 1976 (Kumar et al., 1999; Kinter et al., 2002). Moreover, using an atmospheric general circulation model, the influence of ENSO on the SASM was investigated by Ju and Slingo (1995) and Soman and Slingo (1997). They concluded that the latitudinal position of the ITCZ is significantly influenced by ENSO, and as a consequence, the convective rainfall over the VCH is associated with the ITCZ. The influence of the SASM on ENSO through the atmospheric circulation over the central and eastern Pacific was identified by Yasunari (1990), Chung and Nigam (1999), and Kirtman and Shukla (2000). The relationship between the SASM rainfall and SST anomalies in the equatorial eastern Pacific region was shown by Goswami et al. (1999). More SASM rainfall occurs over the Indochina Peninsula during negative SST anomalies in the eastern tropical Pacific (Chen and Yoon, 2000) and vice versa. Warm SST anomalies in the central equatorial eastern Pacific Ocean can strongly suppress the convective activities over the equatorial western Pacific area and subsequently weaken the SASM variation through large-scale circulation (Wang et al., 2003). The circulation variations in the Indian Ocean can influence ENSO either through the atmosphere or the ocean. The proposed mechanism supportive of the atmospheric connections reveals the co-location of the rising zone of the transverse circulation components of the Indian monsoon and the ascending branch of the Walker circulation (Webster et al., 1992). Variations in the strength of the monsoon can influence trade winds over the Pacific, thus affecting the period and magnitude of ENSO (Barnett, 1984; Wainer and Webster, 1996). Since the warm phase of ENSO is associated with an eastward displacement of a pool of warm water from the equatorial western to the equatorial eastern Pacific, there is an oceanic connection. The Indonesian through-flow can modulate the warm pool in the western Pacific, thus influencing the period and magnitude of ENSO (Wyrtki, 1987). The atmosphere ocean system over the Indian Ocean can be influenced by ENSO events. The observational results reported by Yu et al. (2002) and Wu and Kirtman (2004a) show that the Indian Ocean plays a basic role in the ENSO cycle. Positive SST anomalies always appear in the Indian Ocean during warm ENSO events, and vice versa, through an atmospheric bridge mechanism (Lau and Nath, 1996; Klein et al., 1999). The dominant inter-annual variation of SST over the Indian Ocean is known as a basin wide warming (cooling), lagging a few months behind a mature warm (cold) phase of ENSO (Chambers et al., 1999; Yu and Rienecker, 1999; Lau and Nath, 2003; Wang et al., 2003). The delayed basin wide warming is remotely forced by SST anomalies in the equatorial eastern Pacific Ocean (Lau and Nath, 2003) and remotely forced winds over the Indian Ocean exciting the down welling Rossby waves and contributing to an off-equatorial warming in the western Indian Ocean (Chambers et al., 1999; Yu and Rienecker, 1999). Wind anomalies, therefore, oppose the climatologic circulation over the region and can lead to warming over the western and cooling over the eastern Indian Ocean (Yu et al., 2002). At the early onset stages of an ENSO warm phase, the eastern equatorial Indian Ocean is typically cold while the western begins to warm, and the warming peaks as the ENSO decays (Hendon, 2003).

3 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION 1441 To date, diagnostic studies concerning rainfall and other synoptic weather patterns over the central highlands of Vietnam or over Vietnam as a whole have been addressed in general. More detailed temporal and spatial analyses are potentially more useful when addressing the demand for water, especially in relation to agriculture and water resources management. This was the motivation for this study. Here, we utilize the monthly precipitation anomalies of the network stations over the VCH to investigate the relationships with monthly SST anomalies in both the tropical Pacific and the Indian Oceans. The section Data describes the data used, and the section Methodology shows the methodology. The results of the analyses are presented in the section Results, and the summary and conclusions are presented in the section Summary and Conclusions. DATA This study uses two main sets of data: the monthly SST anomalies over the Pacific and Indian Oceans and anomalies of the observed precipitation monthly totals. Both sets of data were standardized by dividing by the standard deviation. The available data span the period Considering the data limitation of 21 years, careful control of data quality, as well as over the statistical significance and possible physical interpretation of the results, was applied. The rainfall records were observed at 36 stations over the VCH. The data was quality controlled by the Vietnamese National Hydrometeorology Institute. We used the records from stations with no more than 10% missing data during the period from January 1980 to December 2000, thereby selecting 21 stations in the VCH (Figure 1). Among the selected records, the maximum proportion of missing data was 9.5% and the average was 2.8%. This small percentage allows the use of an algorithm to estimate the missing values without causing serious errors. For a particular station showing more than 3% missing data in a month, or a whole year of missing data, a multiple linear temporal regression was established on the basis of rainfall records from close by and highly correlated stations. After this step, only two stations with missing data points remained in the records. These remaining gaps in data were filled-in by inserting appropriate long-term means. The SST data set was obtained from the Comprehensive Ocean Atmosphere Data Set (COADS) (Reynolds et al., 2002). For the purpose of the present study, the tropical range of SST data in the Pacific and Indian Oceans was selected, i.e. 28 N 28 S; 120 E 85 W and 28 N 28 S; 30 E 105 E. A 4 latitude 4 longitude resolution was applied by using the averaged computation from the original 2 2 grid-point data set. The SST and precipitation data sets were standardized in order to use SST and precipitation anomalies at grid 15N LEGEND Political border Rainfall station Kon Tum North Wast region North east region HA NOI Red river delta region LAOS CHINA 14N Pleiku THAI LAND North central region Quan dao Hoang Sa Paracel Islands (Vietnam) Hue Da Nang South central EAST SEA CAMBODIA Central highlands region coast region Khanh Hoa Phu Quoc QULF OF THAI LAND North east south region Quan dao Truong Sa Ho Chi Minh City Spratley Islands Mekong Vung Tau (Vietnam) rive delta region Ca Mau Con dao 13N Buon Ma Thuot 12N 106E 107E 108E 109E Figure 1. The Distribution of rain gauge stations/districts over the Vietnamese central highlands.

4 1442 T. D. NGUYEN, C. UVO AND D. ROSBJERG points and stations, respectively, as the input to the single value decomposition (SVD) analysis. METHODOLOGY The technique employed for data analysis is based on a multivariate analysis procedure known as SVD, used for simultaneous statistical analysis of two data sets varying both spatially and temporally. The choice of this method was based on its simplicity, and previous examples of its successful use, e.g. Bretherton et al. (1992) and Uvo et al. (1998). The technique is applied to the cross-covariance matrix between two field data sets (Bretherton et al., 1992). The use of SVD allows isolating sets of orthogonal pairs of spatial patterns with maximum squared temporal covariance between two physical variables. Bretherton et al. (1992) introduced a conceptual framework for comparison of four multivariate techniques. Among them, SVD directly produces explicit measures of relatedness between patterns, while at the same time being simple to perform and interpret, with no userspecified parameters required. The first use of SVD in a climatologic context was made by Prohaska (1976). Wallace et al. (1992) show the application of SVD for a geographic problem and demonstrate that SVD is the best among the techniques compared. SVD has also been applied in other climatologic studies, e.g. Lanzante (1984), Hsu (1994), and Lau and Nath (1994). In Uvo et al. (1998), the SVD technique was used to illustrate the relationship between monthly SST anomalies in both the Atlantic and Pacific oceans and monthly precipitation anomalies over the northeast of Brazil, considering both simultaneous and lag-time analyses. SVD of the cross-covariance matrix C YZ of two fields (Y and Z) generates two singular vector matrices and one diagonal matrix of singular values. The singular vector matrices describe spatial patterns for each field that have an overall covariance given by the corresponding singular value. Wallace et al. (1993) define a normalized squared covariance (NSC) that is associated with each pair of spatial patterns. The NSC ranges from 0 to 1, where NSC = 0ifthetwofieldsareinnowayrelated,and NSC = 1 if the variations at each grid-point in the first field are perfectly correlated with variations in all grid points in the second field. In this study, NSC values are used to analyse the relationships between SST variations in different oceanic basins with the rainfall variability over the VCH. Once the two singular vectors of the cross-covariance matrix have been obtained from SVD, it is possible to produce heterogeneous correlation maps by projecting the data onto the appropriate singular vector (Bretherton et al., 1992). The heterogeneous correlation map is the vector of correlations between the grid-point values of one field and the k th mode of singular vector in the other field. In the present study, the patterns shown by the heterogeneous correlation map indicate how well the precipitation (SST) anomalies pattern relates to the k th expansion coefficient of the SST (precipitation) anomalies pattern. A test of the null hypothesis based on the Student s t distribution performed in Bendat and Piersol (1986) was used to determine whether the correlation coefficients of the heterogeneous correlation maps differed significantly from what can be expected by chance. RESULTS SVD was employed to investigate the relationship between precipitation over the VCH and SST in the tropical Pacific and Indian Oceans considering both simultaneous and lag-time analysis. The analyses allowed us to determine not only the association between SST and precipitation over the VCH but also precipitation patterns over the VCH related to the Pacific and Indian Ocean SST during the study period. The results obtained from SVD showed that the first mode is the most significant for these analyses, corresponding to a much higher NSC and larger explained variance values compared to the other modes. For all the SVD analyses performed, the NSC reduced sharply from the first to the second mode and beyond (Figure 2). Therefore, the results presented here are obtained by analysing the first mode of the SVD analysis, the socalled leading mode. Simultaneous analyses Heterogeneous correlation maps between monthly Pacific and Indian Ocean SST and precipitation anomalies from the early rainy season to the late rainy season (April to November) are presented in Figure 3. The statistical significance of the correlation coefficients shown in the heterogeneous correlation maps was tested by taking into account the autocorrelation of each time series. The significance test indicated that coefficient correlations above 0.4 and below 0.4 are statistically significant at above 90%. NSC(%) Mode PAC IND Figure 2. Explanation of Normalized Squared Covariance (NSC). 3

5 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION (a) Apr (b) May Figure 3. Heterogeneous correlation map for the first mode in the SVD expansion for monthly Pacific SST and precipitation (upper panel), and Indian Ocean SST and precipitation (lower panel). NSC value is presented on the top right of the left panel. Solid lines represent positive correlation coefficients and dashed lines the negative ones. Light and dark shading shows correlation coefficients above 0.4 and 0.6 (below 0.4 and 0.6), respectively. Analysing the heterogeneous correlation maps between the Pacific and Indian Ocean SST and precipitation anomalies, it is possible to notice that, in general, precipitation over the VCH shows better correlations with SST over the Pacific Ocean than over the Indian Ocean through the rainy season. The highest correlations (>0.6), corresponding to highest NSC values (>0.25), were found in the rainy season in early April, ober, and November in both the Pacific Ocean SST and precipitation patterns. The highest correlations between the Indian Ocean SST and precipitation appear during the last two months of the rainy season (NSC > 0.22). Figure 3(a) shows the correlations between precipitation over the VCH and the Pacific and Indian Ocean SST anomalies in April, reflecting the transition from the dry to the wet season. The SST anomaly in the central equatorial Pacific, representative of ENSO, is negatively correlated with precipitation over the VCH, indicating that an El Niño (La Niña) event would imply a decrease (increase) in precipitation over most of the VCH. On the other hand, SST in the equatorial eastern and southwestern Indian Ocean is positively correlated with precipitation in the northern part of the VCH in that month. However, results show that SST in the Pacific plays a more important role for rainfall variations over the VCH

6 1444 T. D. NGUYEN, C. UVO AND D. ROSBJERG (c) Jun (d) Jul Figure 3. (Continued). in April than SST in the Indian Ocean, as expressed by NSC values of 0.25 and 0.17, respectively. ENSO events affect rainfall over the VCH through variation of the convective activity over the north-western Pacific (Yoo et al., 2004). Thus, what the heterogeneous correlation maps for April show is an increase in rainfall over the VCH associated with positive (negative) SST anomalies in the north-western (central equatorial) Pacific; this is in agreement with Xie et al. (1998) and Zhang et al. (2004), who stated that a warm (cold) SST in the central to equatorial eastern Pacific, representative of the El Niño (La Niña), suppresses (enhances) convection corresponding to a cold (warm) SST in the north-western Pacific and the equatorial eastern Indian Ocean. The convection then affects the SASM circulation, subsequently influencing rainfall over the VCH. Heterogeneous correlation maps for May are shown in Figure 3(b). The relationships between SST and precipitation are not very strong. This finding was corroborated by means of simple correlation between precipitation and SST (not shown). Precipitation in the central VCH is weakly negatively correlated with SST in the equatorial eastern Pacific (NSC = 0.16) and in the tropical off-equatorial Indian Ocean (NSC = 0.20). This lack of significant relationship could be explained by the fact that during May the SASM is not yet fully set up and that the precipitation regime over the VCH is complex and controlled by winds of different origins (Zhang et al., 2002). In June (Figure 3(c)), precipitation over the VCH is negatively correlated with SST in the southern Pacific and the northern Indian Ocean. In this case, NSC values of

7 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION (e) Aug (f) Sept Figure 3. (Continued) and 0.19 for the Indian and Pacific Ocean analyses, respectively, show that a more significant role in rainfall variation over the VCH is played by the Indian Ocean. In June, the full SASM is set up and rainfall over the VCH is associated with convective cloudiness and SASM variation (Zhang et al., 2002). A possible physical mechanism associated with Figure 3(c) is that warm (cold) SST in the southern Pacific can weaken (enhance) the monsoon circulation and moisture advection from the equatorial western Pacific to the Indochina Peninsula, and subsequently decrease (increase) rainfall over the VCH. On the other hand, a warm (cold) SST in the northern Indian Ocean would suppress (enhance) the moisture advection from the Indian Ocean to the Indochina Peninsula, and finally decrease (increase) rainfall over the VCH. In July (Figure 3(d)), SST anomalies in the central tropical Pacific Ocean are positively correlated with precipitation over the central and northern part of the VCH (NSC = 0.16), whereas the SST anomaly in the southwestern Indian Ocean is strongly negatively correlated with precipitation in a small part of the southern VCH (NSC = 0.17). Analyses of the Indian Ocean mainly reveal the same patterns as found in June, although the precipitation area is confined to the south of the VCH, and the SST is confined to the south-western Indian Ocean. During this month, precipitation over the VCH is associated with cyclonic activities over the SCS (Chen and Yoon, 2000). The strong SASM can enhance the surface easterlies in the central equatorial Pacific, induce an eastward propagating up-welling Kelvin wave, and give rise

8 1446 T. D. NGUYEN, C. UVO AND D. ROSBJERG (g) (h) Nov Figure 3. (Continued). to negative (positive) SST anomalies in the eastern (western equatorial) Pacific through air sea interactions (Wu and Kirtman, 2004b). In August (Figure 3(e)), precipitation over the central VCH is positively correlated with SST in the eastern equatorial Pacific (NSC = 0.17) and negatively correlated with SST in the central southern Indian Ocean (NSC = 0.17). During September (Figure 3(f)), precipitation over the southern part of the VCH is positively (negatively) correlated with the tropical eastern (south-western) Pacific SST anomalies (NSC = 0.2). SST in the northern and western Indian Ocean is positively correlated with precipitation in a small part of the southern VCH (NSC = 0.16). Although this is the month when the monsoon normally starts to withdraw northward, rainfall variation over the VCH is still characterized by disturbances of the SCS and the SASM variation (Wang and Wu, 1997). A possible physical mechanism related to Figure 3(f) is that positive SST anomalies in the western and northern Indian Ocean may cool the SST over the equatorial eastern Indian Ocean and suppress the SASM variation over peninsular Indochina, and subsequently reduce rainfall over the VCH and enhance rainfall variation over the eastern African continent and the maritime Indian subcontinent. In ober (Figure 3(g)), the central equatorial Pacific SST is highly negatively correlated with precipitation over the VCH (NSC = 0.3). The equatorial eastern Indian Ocean SST is positively correlated with precipitation over the VCH (NSC = 0.23). In this month, precipitation over

9 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION 1447 the VCH is associated with the SASM withdrawal (Wang and Wu, 1997) and deep convective activities over the eastern Indian Ocean (Goswami and Shukla, 1984; Qian and Lee, 2000). A physical mechanism associated with that shown in Figure 3(g) is that a warm (cold) SST over the central to eastern equatorial Pacific highly enhances a cold (warm) SST over the eastern Indian Ocean through air sea interactions, and then the warm SST over the equatorial eastern Indian Ocean enhances the convective rain band to the VCH. Figure 3(h) shows the analyses from November, late in the rainy season. Precipitation over the VCH is highly positively correlated with the north-western Pacific SST (NSC = 0.28) and the equatorial eastern Indian Ocean SST (NSC = 0.22). During this month, the southeasterly wind is already replaced by a north-easterly wind, and precipitation over the VCH is still governed by convective activities (Goswami and Shukla, 1984; Qian and Lee, 2000). The heterogeneous correlation maps for November show that a warm SST over the north-western Pacific and eastern Indian Oceans is associated with an increased rainfall over the VCH; this is in agreement with Qian and Lee (2000). They revealed that warm SST over the north-western Pacific and eastern Indian Oceans play a major role in enhancing convective rainfall to the VCH; convective rainfall ends over the VCH in late November. Lag-time analyses The SVD analyses are presented using monthly precipitation during the rainy season, April to November, over the VCH and monthly SST in the Pacific and Indian Oceans in the preceding months, from one to three months in advance. The results are displayed in Figures 4 and 5. In general, SST in the Pacific plays a more important role for rainfall variation over the VCH than SST in the Indian Ocean during the rainy season. The results presented in Figures 4 and 5 show that SST in the Pacific is highly related to rainfall variation over the VCH through the rainy season (except August, not shown) from one to three months in advance. The strongest correlations (and highest NSC values) were obtained for April, ober, and November, with coherent SST in the north-western and central to eastern equatorial Pacific and precipitation patterns over the VCH. There are no relationships between precipitation patterns over the VCH and the preceding SST in the Indian Ocean (not shown), with the exception that significant correlations exist between rainfall variation during the last stages of the rainy season (ober and November) and rainfall variation preceding SST by one to three months. Here we only show the significant correlations between precipitation over the VCH and SST patterns in the Pacific and Indian Oceans. Precipitation in April is highly negatively correlated with SST in the central to eastern tropical Pacific (representative of ENSO), and positively correlated with SST in the north-western Pacific one to three months in advance (Figure 4(a)), with NSC values around 0.3. A possible physical mechanism to what is shown in Figure 4(a) is that warm (cold) SST over the central to eastern tropical Pacific corresponding to cold (warm) SST over the north-western Pacific may delay (advance) the SASM onset, and subsequently reduce (increase) rainfall over the VCH. In other words, ENSO has a strong longterm influence on the rainy season onset over the VCH, expressing the advance or delay of the SASM over the Indochina Peninsula. Figure 4(b) shows that rainfall over the central and southern VCH in May is negatively (positively) correlated with the central equatorial (north-western) Pacific SST in February. Different from other months, no more significant lagged relationships were found between SST and precipitation during May. In June (Figure 4(c)), rainfall variation over the VCH is also negatively correlated with SST in the central to eastern equatorial Pacific from one to two months in advance. The one-month lag-time correlation is more significant than the two-month lag-time correlation, as expressed by NSC values of 0.21 and 0.17, respectively. A physical explanation regarding what the heterogeneous correlation maps for June show is that a warm (cold) SST in the central equatorial Pacific can suppress (enhance) the SASM variation through air sea interaction, and subsequently decrease (increase) rainfall over the VCH. In July, precipitation over the VCH is highly correlated with SST in different regions of the Pacific from one to three months in advance. As shown in Figure 4(d), rainfall over the VCH is more highly correlated with SST in the tropical equatorial eastern Pacific three months in advance than with other regions in the Pacific, and the NSC value is 0.2 where lag-times from one to two months in advance result in NSC values of 0.17 and 0.18, respectively. During August, there is no significant lag-time relationship between SST in the oceans and precipitation patterns over the VCH (not shown). Rainfall over the VCH is mainly characterized by cyclones, i.e. storms blowing from outside the SCS (Chen and Yoon, 2000), suggesting that rainfall is influenced by simultaneous synoptic weather patterns rather than by preceding SST variations in both ocean basins. Figure 4(e) shows a positive correlation between precipitation in September over the VCH and SST patterns over the central equatorial Pacific in June, indicating that SST over the central equatorial Pacific can have a longterm influence on the VCH precipitation. Rainfall during this period is associated with cyclonic activities over the SCS and retreat of the SASM southward (Wang and Wu, 1997). Figure 5(a) shows the lagged relationships between precipitation in ober and SST over the Pacific and Indian Oceans up to three months. Precipitation over the VCH is negatively (positively) correlated with SST in the central equatorial (western) Pacific (Figure 5a 1 ) and positively correlated with SST in regions around the equatorial Indian Ocean (Figure 5a 2 ). The high lagged correlation between SST in the Pacific and precipitation

10 1448 T. D. NGUYEN, C. UVO AND D. ROSBJERG Mar Apr Feb Apr (a) Jan Apr (b) Feb May May Jun (c) Apr Jun Figure 4. Heterogeneous correlation map for the first mode in the SVD expansion for precipitation (right panels) and preceding the monthly Pacific SST (left panels). The NSC value is presented on the top right of the left panel. Solid lines represent the positive correlation coefficients and dashed lines the negative ones. Light and dark shading shows correlation coefficients above 0.4 and 0.6 (below 0.4 and 0.6), respectively. over the VCH from one to three months in advance, with corresponding NSC values of 0.28, 0.23, and 0.21, respectively, are displayed in Figure 5a 1.Onthe other hand, a high correlation between precipitation and SST patterns over the equatorial eastern Indian Ocean (the NSC is 0.21) can be found one month in advance (Figure 5a 2 ). Rainfall variation over the VCH in ober is associated with deep convection, withdrawal of the SASM, and the Indian Ocean ITCZ (Goswami and Shukla, 1984; Wang and Wu, 1997;

11 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION Jun Jul May Jul (d) Apr Jul (e) Jun Sept Figure 4. (Continued). Qian and Lee, 2000). A reasonable physical mechanism related to what is shown in Figure 5(a) is that the onset of ENSO influences rainfall over the VCH from one to three months in advance by influencing the variations of SASM. On the other hand, a warm (cold) SST in September over the eastern Indian Ocean can enhance (suppress) convective activities and the ITCZ movement over the southern Indochina Peninsula, and subsequently increase (decrease) rainfall over the VCH in ober. Precipitation in November is positively (negatively) correlated to SST in the western (central tropical) Pacific (Figure 5b 1 ) and positively correlated to SST over the equatorial eastern Indian Ocean (Figure 5b 2 )uptothree months in advance. Rainfall in this month is related to convective activities in the equatorial eastern Indian Ocean and the north-western Pacific (Qian and Lee, 2000; Yoo et al., 2004). A possible physical mechanism associated with heterogeneous maps shown in Figure 5(b) is that a warm (cold) SST over the eastern Indian ocean and north-western Pacific can motivate (suppress) convective activities in relation to heavy rainfall over the VCH at the end of the rainy season. It is apparent that SST pattern variations over the equatorial eastern Indian Ocean and the north-western Pacific have a long-term influence on precipitation over the VCH. We also use SVD to conduct seasonal analyses, as shown in Figure 6. The results indicate that seasonal SST in the two oceans exerts different influence on seasonal rainfall in different areas of the VCH. The tropical central to the eastern Pacific SST is negatively related to precipitation in the southern part of the VCH, while SST in the northern and central southern Indian Ocean is negatively associated with precipitation over the northern and central parts of the VCH. The Indian Ocean seasonal SST plays a more important role in the variation of seasonal rainfall over the VCH than does the Pacific seasonal SST, as expressed by NSC values of 0.2 and 0.18, respectively. The orthogonality imposed by the SVD method can create artificial patterns, which is especially sensitive

12 1450 T. D. NGUYEN, C. UVO AND D. ROSBJERG Sept Aug (a1) Jul Sept Aug (a2) Jul Figure 5. Heterogeneous correlation map for the first mode in the SVD expansion for precipitation (right panels) and preceding the monthly Pacific and Indian Ocean SST (left panels). The NSC value is presented on the top right of the left panel. Solid lines represent the positive correlation coefficients and dashed lines the negative ones. Light and dark shading shows correlation coefficients above 0.4 and 0.6 (below 0.4 and 0.6), respectively. in short time series. To ensure the robustness of our results, a simple test was done by comparing the observed cases with the patterns obtained from the SVD analysis. An extreme ENSO, including El Niño and La Niña events, in 1998 is considered. April precipitation and SST over the Pacific in February (El Niño events) and November precipitation and SST over the Indian Ocean in ober (La Niña events) were checked. As shown

13 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION Nov Sept Nov (b1) Aug Nov Nov Sept Nov (b2) Aug Nov Figure 5. (Continued). in Figure 7, the positive anomalies over the equatorial central to eastern Pacific in February are associated with significant rainfall deficit in April at most sites over the VCH (upper panel). The positive anomalies over the equatorial eastern Indian Ocean in ober are related to abundant rainfall in November at most sites over the VCH (lower panel). The upper panel in Figure 7 is similar to Figure 4(a), obtained by SVD, for precipitation in April and for the Pacific SST in February. The lower panel in Figure 7 is similar to Figure 5(b), obtained from SVD, for November precipitation and the Indian Ocean SST in ober.

14 1452 T. D. NGUYEN, C. UVO AND D. ROSBJERG Figure 6. Heterogeneous correlation map for the first mode in the SVD expansion for seasonal precipitation (right panels) and seasonal Pacific and Indian Oceans SST (left panels). The NSC value is presented on the top right of the left panel. Solid lines represent the positive correlation coefficients and dashed lines the negative ones. Light and dark shading shows correlation coefficients above 0.4 and 0.6 (below 0.4 and 0.6), respectively. Feb Apr.98 Nov.98 Figure 7. Rainfall anomalies (right panels) and SST anomalies (left panels) in Solid (dashed) line indicates positive (negative) SST anomalies, light and dark shading indicates anomalies >1.0 C and 2.0 C, respectively, in SST pattern and >0.5 and 1.0 standard deviation, respectively, in rainfall pattern. SUMMARY AND CONCLUSIONS We have used SVD to illustrate the relationships between standardized monthly tropical SST anomalies in both the Pacific and Indian Oceans and standardized monthly precipitation anomalies over the VCH throughout the early to late period of the rainy season (April to November), considering both simultaneous and lag relationships. The results obtained from the leading mode of SVD show different relationships between the tropical Pacific and the Indian Ocean SST and precipitation patterns over the VCH during different months of the rainy season. Rainfall variations over the VCH are most sensitive to SST changes in April, ober, and November. Changes in SST can significantly influence the onset and withdrawal of monsoon during the early and late rainy seasons, respectively. In June, July, August, and September,

15 RELATIONSHIP BETWEEN THE INDIAN AND PACIFIC OCEANS SST AND THE VCH PRECIPITATION 1453 the analyses reveal that rainfall variations over the VCH are weakly correlated with SST anomalies over both basins. In April, a very significant relationship between SST in the tropical central to eastern equatorial Pacific and precipitation over the VCH was found. During this month, rainfall variation over the VCH is controlled mainly by deep convective activities over the equatorial eastern Indian Ocean and the SCS. ENSO apparently affects the synoptic activities through large-scale circulation. In May, when the rainy season over the VCH gets started, no significant relations between SST in the oceans and precipitation patterns were found. Precipitation over the VCH is largely generated by active cumulus convection and other disturbances controlled by south-easterly winds converging on the equatorial Indian Ocean westerly wind over the southern Indochina Peninsula. In June, the full SASM is set up. Precipitation over the VCH is then associated with convective cloudiness and the SASM variation. Rainfall variations over the VCH are negatively correlated with SST anomalies in both the southern Pacific and northern Indian Oceans, suggesting that SST in both oceans influences convection and the SASM variation through air sea interactions related to the Walker and local Hadley circulations. Through July to September, no significant correlations can be found in precipitation patterns, contrary to what is seen for SST patterns in both oceans. Rainfall over the VCH is mainly characterized by cyclones, storms, and other disturbances blowing from the SCS. The strong SASM may affect SST variations in both oceans in this period, rather than the SASM being influenced by SST patterns. In ober, precipitation over the VCH is highly related to ENSO and positively correlated with the equatorial eastern Indian Ocean SST. During this month, rainfall variation is associated with the ITCZ and convection over the equatorial eastern Indian Ocean. The equatorial eastern Indian Ocean SST is highly related to ENSO through air sea interactions. In November, a significant positive correlation can be found between precipitation over the VCH and SST over the equatorial eastern Indian Ocean and the north-western Pacific. To investigate the extent to which precipitation over the VCH can be predicted from SST in both oceans, we analysed the lag-time relationships between precipitation and SST. The results revealed that the Pacific SST is well correlated with rainfall one to three months in advance throughout the rainy season, except in August. For the Indian Ocean, significant lag-time relationships (one and two months) were found between the Indian Ocean SST and the VCH precipitation in ober and November. High lag-time correlations between precipitation over the VCH and SST patterns in both oceans were observed especially in the months of ober and November. The relationships identified between monthly precipitation and SST in the Pacific and Indian Oceans may provide the basis of a predictive scheme. Establishment of such a scheme would require further analysis of the level of skill that can be achieved in prediction of precipitation over the VCH. The results obtained in this work will be used toward the development of models for monthly precipitation forecast, and it is expected that such forecasts will be significantly useful for agricultural planning in the central highlands of Vietnam. ACKNOWLEDGEMENTS This work was financed by the Water SPS Subcomponent 1.3 of the DANIDA program support to Water Resources University, Vietnam. The authors thank the National Hydrometeorology Institute of Vietnam for providing the precipitation data set. The authors also thank the anonymous reviewers for constructive comments on the original manuscript. REFERENCES Barnett TP Interaction of the monsoon and Pacific trade wind system at inter-annual time scales. Monthly Weather Review 112: Bendat JS, Piersol ES Radom Data-Analysis and Measurement Procedures. John Wiley: New York; 525. Bretherton CS, Smith C, Wallace JM An intercomparison of methods for finding coupled patterns on climate data. Journal of Climate 5: Chambers DP, Tapley BD, Stewart RH Anomalous warming in the Indian Ocean coincident with El Niño. Journal of Geophysical Research 104(2): Chen TC, Weng SP Interannual and intraseasonal variations in monsoon depressions and their westward-propagating predecessors. Monthly Weather Review 127: Chen TC, Yoon JH Interannual Variation in Indochina summer monsoon rainfall: Possible mechanism. Journal of Climate 13: Chung C, Nigam S Asian summer monsoon ENSO feedback on the Cane-Zebiak model ENSO. Journal of Climate 12: Ding Y, Wang Q, Yan J Some aspects of climatology of the summer monsoon over the South China Sea. In Atmospheric Circulation to Global Change, IAP Chinese Academy of Sciences. China Meteorological Press: Beijing; Fennessy M, Shukla J. 1994a. GCM simulations of active and break monsoon period. In Proceedings of the International Conference on Monsoon Variability and Prediction, Trieste, WMO/TD 619, Fennessy M, Shukla J. 1994b. Simulation and predictability of monsoons. In Proceedings of the International Conference on Monsoon Variability and Prediction, Trieste, WMO/TD 619, Goswami BN Interannual variations of Indian summer monsoon in a GCM: External conditions versus internal feedbacks. Journal of Climate 11: Goswami BN, Shukla J Quasi-periodic oscillations in a symmetric general circulation model. Journal of the Atmospheric Sciences 41: Goswami BN, Krishnamurthy V, Annamalai H A broad scale circulation index for the interannual variability of the Indian summer monsoon. Quarterly Journal of the Royal Meteorological Society 125: He H, McGiness JW, Song Z, Yanai M Onset of Asian monsoon in 1979 and the effect of the Tibetan Plateau. Monthly Weather Review 115: Hendon HH Indonesian rainfall variability: Impacts of ENSO and local air-sea interaction. Journal of Climate 16: Hsu H Relationship between tropical heating and global circulation: interannual variability. Journal of Geophysical Research 99: Ju J, Slingo JM The Asian summer monsoon and ENSO. Quarterly Journal of the Royal Meteorological Society 121:

16 1454 T. D. NGUYEN, C. UVO AND D. ROSBJERG Kinter JL, Miyakoda K, Yang S Recent change in the connection from the Asian monsoon to ENSO. Journal of Climate 15: Kirtman BP, Shukla J On the influence of the Indian summer Monsoon on ENSO. Quarterly Journal of the Royal Meteorological Society 126: Klein SA, Soden BJ, Lau NC Remote Sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge. Journal of Climate 12: Kumar KK, Rajagopalan B, Cane MA On the weakening relationship between the Indian monsoon and ENSO. Science 284: Lanzante JR A rotated eigenanalysis of the correlation between 700-mb heights and sea surface temperatures in the Pacific and Atlantic. Monthly Weather Review 112: Lau N, Nath MJ A modeling study of the relative roles of the tropical and extratropical SST anomalies in the variability of the global atmosphere-ocean system. Journal of Climate 7: Lau NC, Nath MJ The role of the atmospheric bridge in linking tropical Pacific ENSO events to extratropical SST anomalies. Journal of Climate 9: Lau KM, Yang S Climatology and interannual variability of the Southeast Asian summer monsoon. Advances in Atmospheric Sciences 14: Lau NC, Nath MJ Atmosphere ocean variations in the Indo- Pacific sector during ENSO episodes. Journal of Climate 16: Li C, Yanai M The onset and interannual variability of the Asian summer monsoon in relation to land-sea thermal contrast. Journal of Climate 9: Matsumoto J Seasonal transition of summer rainy season over Indochina and adjacent monsoon region. Advances in Atmospheric Sciences 14: Prohaska J A technique for analyzing the linear relationships between two meteorological fields. Monthly Weather Review 104: Qian WH, Lee DK Seasonal march of Asian summer monsoon. International Journal of Climatology 20(11): Reynolds RW, Rayner NA, Smith TM, Stokes DC, Wang W An improved in Situ and Satellite SST analysis for climate. Journal of Climate 15: Saha K, Sanders F, Shukla J Westward propagating predecessors of monsoon depressions. Monthly Weather Review 109: Shukla J Interannual variability of monsoon. In Monsoons, Fein JS, Stephens PL (eds). Wiley and Sons: New York; Sikka DR Some aspects of large-scale fluctuations of summer monsoon rainfall over India in relation to fluctuations in planetary and regional scale circulation parameters. Proceedings of the Indian Academy of Sciences-Earth and Planetary Sciences 89: Soman MK, Slingo JM Sensitivity of Asian summer monsoon to aspects of sea surface temperature anomalies in the tropical Pacific Ocean. Quarterly Journal of the Royal Meteorological Society 123: Torrence C, Webster PJ Interdecadal changes in the ENSOmonsoon system. Journal of Climate 12: Uvo CT, Repelli CA, Zebiak SE, Kushnir Y The Relationships between tropical Pacific and Atlantic SST and northeast Brazil monthly precipitation. Journal of Climate 11: Wainer I, Webster PJ Monsoon-ENSO interaction using a simple coupled ocean-atmosphere model. Journal of Geophysical Research 101: Walker GT Correlation in seasonal variations of weather, IV, a further study of world weather. Memoirs of the Indian Meteorological Department 24: Wallace JM, Smith C, Bretherton S Singular value decomposition of wintertime sea surface temperature and 500-mb heights anomalies. Journal of Climate 5: Wallace JM, Zhang Y, Lau KH Structure and seasonality of interannual and interdecadal variability of the geopotential height and temperature fields in northern hemisphere troposphere. Journal of Climate 6: Wang B, Wu R Peculiar temporal structure of the South China Sea summer monsoon. Advances in Atmospheric Sciences 14: Wang B, Wu R, Li T Atmosphere-warm ocean interaction and its impacts on Asian-Australian monsoon variations. Journal of Climate 16: Webster PJ, Yang S Monsoon and ENSO selectively interactive systems. Quarterly Journal of the Royal Meteorological Society 118: Webster PJ, Yang S, Wainer I, Dixit S Processes involved in monsoon variability. In Physical Processes in Atmospheric Models, Sikka DR, Singh SS (eds). Wiley Eastern: New Delhi; Wu R, Kirtman BP. 2004a. Impacts of the Indian Ocean on the Indian summer monsoon ENSO relationship. Journal of Climate 17: Wu R, Kirtman BP. 2004b. The tropospheric biennial oscillation of the monsoon-enso system in an interactive ensemble coupled GCM. Journal of Climate 17: Wyrtki K Indonesian throughflow and associated pressure gradient. Journal of Geophysical Research 92: Xie A, Chung YS, Liu X, Ye Q The interannual variability of the summer monsoon onset over the South China Sea. Theoretical and Applied Climatology 59: Yanai M, Li C, Song Z Seasonal heating of the Tibetan Plateau and its effects on the evolution of the Asian summer monsoon. Journal of the Meteorological Society of Japan 70: Yasunari T Impact of Indian monsoon on the coupled atmosphere ocean system in the tropical Pacific. Meteorology and Atmospheric Physics 44: Yoo SH, Ho CH, Yang S, Choi HJ, Jhun JG Influences of tropical western and extratropical Pacific SST on east and southeast Asian climate in the summers of Journal of Climate 17: Yu L, Rienecker MM Mechanisms for the Indian Ocean warming during the El Ninõ. Geophysical Research Letters 26: Yu JY, Mechoso CR, McWilliams JC, Arakawa A Impacts of the Indian Ocean on the ENSO cycle. Geophysical Research Letters 29(8): 1204, /2001GL Zhang Y, Li T, Wang B, Wu G Onset of Asian summer monsoon over Indo-china and its interannual variability. Journal of Climate 15: Zhang Z, Zohnny CLC, Ding Y Characteristics, evolution and mechanisms of the summer monsoon onset over Southeast Asia. International Journal of Climatology 24: