Long-term changes in the South China Sea summer monsoon revealed by station observations of the Xisha Islands



[1] The authors depict the long-term changes in the South China Sea (SCS) summer monsoon using observational data of the Xisha Islands. The SCS monsoon is an important component of the Asian monsoon system, and its long-term changes have seldom been explored because of the unavailability of reliable data. The daily Xisha station observations provide an important source of information for understanding the changes in the monsoon. The intensity of the SCS summer monsoon measured by kinetic energy decreased significantly from 1958–1977 to 1978–2004. This change in monsoon was mainly caused by the weakening of the meridional component of lower tropospheric winds, and the weakening in the mean flow was signaled by decreased frequency of strong southerlies (6 m s−1 and above) of the daily winds. The weakening of the monsoon was also associated with increases in sea surface temperature and surface and lower tropospheric air temperatures over SCS, which occurred more frequently when daily surface temperature reaches 29°C and higher. The long-term warming of the lower troposphere was accompanied by cooling at the upper troposphere, destabilizing the local atmosphere. However, from 1958–1977 to 1978–2004, the long-term change in Xisha precipitation tended to decrease; furthermore, in fact, the station precipitation became less variable. Thus besides local air-sea interaction, large-scale atmospheric forcing also plays an important role in causing the long-term change of the Xisha precipitation. Indeed, the warming of Xisha was linked to large-scale warming in the tropics including SCS and was associated with smaller thermal contrast between the Asian continent and the surrounding oceans, which weakened monsoon circulation.

1. Introduction

[2] The South China Sea (SCS) monsoon is an important component of the Asian monsoon system. It is located between the Indian monsoon in the west and the western North Pacific (WNP) monsoon in the east. The Indian summer monsoon, perhaps the monsoon component that has received most research interest [Krishnamurti, 1985; Shukla and Mooley, 1987; Webster, 1987, 2006; Goswami, 2006], is often considered a more typical monsoon phenomenon with a structure of reversed Hadley cell in the latitude-height section and an apparent relationship between cyclonic flow and rainfall patterns. The WNP monsoon, on the other hand, is characterized by its eastward stepwise onset process and a significant contribution of flow confluence to monsoon precipitation [Murakami and Matsumoto, 1994; Murakami et al., 1999; Wu and Wang, 2000, 2001; Wu, 2002; Li and Wang, 2005]. For many, the SCS is part of the Southeast Asian monsoon, which prevails in the Southeast Asian countries and southern China but is distinguished from the East Asian monsoon over northern-eastern China, Japan, Korea, and their adjacent regions [see Chang, 2004; Wang and Li, 2004, and Ding and Chan, 2005 for reviews].

[3] The interaction between the various monsoon components is one of the factors that contributes to the complexity of the SCS monsoon. For example, the subtropical highs over Asia and the western Pacific, whose variability is closely related to the WNP and East Asian monsoons, play an important role in determining the onset and intensity of the SCS monsoon [e.g., He et al., 2001]. The particular location of the SCS monsoon as a transitional zone from the Indian monsoon to the WNP monsoon also equips the local monsoon with many unique features and attracts substantial research interest [for reviews, see Tao and Chen, 1987 and Krishnamurti et al., 1999).

[4] The SCS marks the earliest onset of the Asian summer monsoon [He et al., 1987; Tao and Chen, 1987] although it is also argued that the earliest monsoon onset occurs in the Indo-China peninsula or the Bay of Bengal [Wu and Zhang, 1998; Liu et al., 2002; Wu et al., 2005]. Nevertheless, the onset of SCS monsoon, representing a shift from the winter regime to the summer regime for many regions like southern China, is uniquely characterized by a jump-like process [Lau and Li, 1984; Lau and Yang, 1997]. This unique feature has led to substantial research interest in the onset of SCS monsoon by examining a variety of factors such as the changes in the subtropical western Pacific high, the thermal states of the Indo-China peninsula and Indo-Pacific oceans, and the SCS sea surface temperature (SST) before and after the onset of SCS summer monsoon [Mao et al., 2004; Huang et al., 2007].

[5] The variability of the SCS monsoon is governed by both local air-sea interaction and remote impact of other large-scale variability of the ocean and the atmosphere [Murakami et al., 1986; He et al., 1992; Ose et al., 1997; Liang and Wu, 2002; Li et al., 2005; Wang et al., 2005; Wu et al., 2005]. Because of the nearly closed ocean domain, fluctuations of the local SST with the monsoon flow over SCS are less significant than those observed in the Arabian Sea and the Bay of Bengal [e.g., Lau and Yang, 1997; Chen et al., 2003], and it is usually difficult to establish a strong relationship in which the monsoon is apparently linked to the local SST. On the other hand, significant relationships have been found between the SCS monsoon and El Niño-Southern Oscillation (ENSO), the western Pacific SST, the Tibetan plateau, and the meridional temperature gradient across tropical Asia [Nitta, 1987; Yanai et al., 1992; Nitta and Hu, 1996; Lau and Yang, 1997; Wu and Zhang, 1998; Yoo et al., 2004; Chen et al., 2007; Huang et al., 2007]. The heating over the Tibetan plateau promotes the reversal of meridional temperature gradient from winter to summer. A warming in the tropical western Pacific shifts the subtropical high eastward, which favors an establishment of cyclonic pattern over SCS and facilitates an early onset of the monsoon. It is also found that the release of strong latent heat over the Indo-China peninsula and the Bay of Bengal plays an important role in triggering the onset of SCS summer monsoon [Liu et al., 2002; Zhang et al., 2002; Wu et al., 2005]. Previous studies have also shown that the SCS monsoon exerts influences on the climate of the adjacent regions [e.g., Huang et al., 2007]. For example, the variations of the pre-rainy season precipitation over southern China, the Meiyu in central-eastern China, the Baiu in Japan, and the Changma in Korea are linked to the activity of SCS monsoon [Huang and Tao, 1992; Kawamura and Murakami, 1998; Ho et al., 2003; Wang et al., 2004].

[6] Although it is realized that the SCS monsoon is an important component of the Asian monsoon system and its fluctuations affect the climate of the nearby countries, to fully understand the monsoon and its variations faces tremendous difficulties. One of the nuisances exists in obtaining reliable data. In this context, the South China Sea Monsoon Experiment [Lau et al., 1998; Lau et al., 2000; Ding et al., 2004; Johnson et al., 2005], which was implemented for May–August 1998, has provided valuable observational data to understand the onset, evolution, and other short-term variability of the monsoon. However, like all field experiments, this experiment is unable to provide data records that are long enough for understanding the long-term changes of the monsoon. Data records of a few years or decades may exist for some places such as islands over SCS, but they are rarely available due to a variety of reasons. Because of these, the long-term variations of the SCS monsoon have seldom been explored. Instead, as reviewed above, previous studies mainly focus on the variability of the monsoon on interannual and shorter timescales. Limited investigations on the interdecadal and longer variability have been carried out such as those by Ding et al. [2002], Peng et al. [2003], and Chan and Zhou [2005]. However, these studies use data from reanalysis products or observations from nearby stations like Hong Kong.

[7] In this study, we analyze the station observations from the Xisha Islands to understand the variations of SCS monsoon. The data cover the time period from 1958 to 2004, which enable us to focus on the long-term changes of the monsoon. The station observations also include daily information, which facilitates our analysis of the more detailed features linking the long-term changes in the SCS monsoon climate to higher-frequency conditions of monsoon weather. To understand the large-scale features associated with the long-term changes of the monsoon, we will also analyze different products of reanalysis. In this study, we recognize the values of long records of the data and the availability of daily data for investigating the SCS summer monsoon.

[8] In the next section, we provide a description of the data to analyze. In section 3, we explore the long-term changes in atmospheric and oceanic observations from Xisha shown in different fields. In section 4, we depict the large-scale circulation features that are associated with the changes in Xisha station observations by analyzing reanalysis products. Finally, we provide a summary of the results and a further discussion in section 5.

2. Data

[9] The main data analyzed in this study are the daily observations from the Xisha Islands, which is (a group of islands) located over the central-northern South China Sea. The data were observed at the Yongxing Island, the largest island of the group of coral islands. The Yongxing Island, with an area of about 2 km2, is located at 16°50′N, 112°20′E. However, the station of observation over Yongxing is officially referred to as Xisha station. The Xisha observations include regular surface observations (winds, temperature, precipitation, clouds, etc.) of four times a day and soundings (temperature, winds, geopotential height, etc.) twice daily. They began from July 1957 with few missing data. Furthermore, the site of observation has never been changed since the observations were operated. Thus the Xisha observations are considered important for both scientific research and weather-climate operations. The long record of the data is especially valuable for climate studies such as the study of the climatological seasonal cycle by Wu and Liang [1998]. The raw data of the product, covering the period of 1958–2004, analyzed in this study were provided by the Hainan Meteorological Bureau, China Meteorological Administration. The observed SST data (for 1961–1999) was provided by the South China Sea Branch, the State Oceanic Administration of China.

[10] The National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis [Kalnay et al., 1996; NNRA] and the European Centre for Medium-range Weather Forecast reanalysis [Uppala et al., 2005; ERA-40] are also used in this study to depict the broad-scale features of atmospheric circulation that are associated with the variations of monsoon revealed by the Xisha station observations. This is important for better understanding and explaining the local climate variation and for assessing the reanalysis products with valuable station data. We also analyze the precipitation analysis of Chen et al. [2002], who have provided precipitation reconstruction data over land and oceans (PREC) in a resolution of 2.5° × 2.5° latitude/longitude and precipitation reconstruction over land only (PREC/L) in a resolution of 0.5° × 0.5° latitude/longitude, for the time period from 1948 to present. The data sets contain gauge-based global precipitation observations. The high-resolution PREC/L data also include information over islands and, for the United States, they compare well with the NOAA’s Climate Prediction Center Unified Precipitation data [Higgins et al., 2000] except for high mountainous regions (M. Chen, personal communication, 2007). For the overlapped time period since 1979, the PREC is similar to the Climate Prediction Center Merged Analysis of Precipitation [Xie and Arkin, 1996].

3. Long-Term Changes in Xisha Observations

[11] Kinetic energy is a useful parameter for measuring the intensity of monsoon [Webster and Yang, 1992; Xu and Chan, 2002]. Perhaps it depicts more appropriately the dynamical aspect of monsoon than the variability of monsoon precipitation. To understand the variability of monsoon climate over the Xisha Islands, we analyze the yearly changes in kinetic energy calculated from the surface winds of the observatory station (see Figure 1). The most striking feature of the Xisha kinetic energy is its long-term change characterized by a substantial reduction from 1978 (Figure 1a). This feature implies a long-term weakening of the SCS summer monsoon (see more discussions in the next section). In spite of a few exceptions in the 1960s and 1990s, the normalized values of kinetic energy are mostly positive during 1958–1977 but are negative during 1978–2004. The means are 0.73 for 1958–1977 and −0.54 for 1978–2004. As seen in Table 1 later, the unnormalized mean values for the two periods are 9.78 and 6.41 m2 s−2, respectively.

Figure 1.

Normalized fields of (a) JJA surface kinetic energy, (b) meridional wind, and (c) zonal wind over Xisha. In Figures 1a and 1b, the solid lines measure the values of normalized means for respective periods of time (1958–1977 and 1978–2004). In Figure 1c, the dashed lines denote the values of normalized standard deviations for the respective time periods.

Table 1. Means and Standard Deviations (SD) for Periods 1958–2004, 1958–1977, and 1978–2004a
Fields1958–2004 Means1958–1977 Means1978–2004 MeansMean DIFF1958–2004 SD1958–1977 SD1978–2004 SDSD DIFF
  • a

    DIFF represents the differences of period II minus period I. Shown are the values for kinetic energy (K), total cloud amount (N), precipitation (P), and others. The values significant at the 99% confidence level are shown in bold and italics except the 97% significance for the standard deviation of SST-Ts.

K, m2 s−27.859.786.413.372.652.301.87−0.43
v, m s−13.614.173.190.980.740.500.590.09
u, m s−11.341.331.340.010.630.470.740.27
Ts, °C28.8428.6728.970.300.350.330.32−0.01
P, mm/day594.61615.96578.80−37.16254.99313.95205.91108.04
SST, °C29.4829.1929.710.520.400.330.30−0.03
SST-Ts, °C0.660.510.780.

[12] It is speculated that the significant change in kinetic energy shown in Figure 1a is associated with robust features of long-term changes in other parameters. In the rest of this section, we further depict these features by analyzing dynamical and thermodynamical fields for both atmospheric and oceanic components. Investigation into the 2-m surface winds indicates that the change in kinetic energy shown in Figure 1a is mainly explained by the long-term change in the meridional wind component (Figure 1b) instead of the zonal wind. With a correlation coefficient of 0.93, there exists a close relationship between the variability of kinetic energy and the meridional wind. Indeed, in terms of positive and negative signs, there is a nearly one-to-one corresponding relationship between the two on interannual timescales. The meridional wind also shows a significant long-term reduction from 1978, which is specified by the normalized mean values of 0.77 for 1958–1977 and −0.56 for 1978–2004. As it will be seen in Table 1 later, such a change reflects the substantial weakening of the southerly wind.

[13] On the other hand, the zonal component of surface winds over Xisha does not exhibit a long-term change in its intensity (Figure 1c; also see Table 1). Compared with the meridional wind component, the normalized zonal wind appears positively and negatively almost equally over the entire period of time. However, the zonal wind, which is insignificantly correlated to the meridional wind and kinetic energy, tends to be more variable during the latter period (1978–2004) than during the earlier period, as indicated by the different values of standard deviation of 1.17 and 0.74.

[14] We now examine the variability of surface air temperature (Ts), the SST, and the difference between SST and Ts from the Xisha observations (see Figure 2). Both Ts and SST are higher in the latter period than in the earlier period. The increases in the means are 0.3°C for Ts and 0.52°C for SST (see Table 1). Given that temperatures vary relatively smaller in SCS compared with the Arabian Sea, the Bay of Bengal, and the central Pacific, these values reflect large changes in temperatures. In fact, the apparent increases in the temperatures occurred in 1977, 1 year before the changes in kinetic energy and meridional wind shown in Figure 1. However, the difference between SST and Ts increases significantly, from negative to positive, in 1978 (Figure 2c). Analysis of the monthly values shows that SST is always higher than Ts. Thus Figure 2c indicates that the rise of SST outpaces that of Ts during the second period. For SST-Ts, the unnormalized mean difference between 1978–2004 and 1958–1977 reaches 0.27°C (Table 1).

Figure 2.

Normalized fields of (a) JJA surface temperature (Ts), (b) SST, and (c) temperature difference SST-Ts at Xisha. The solid lines measure the values of normalized means for respective periods of time.

[15] We also quantitatively compute the surface latent heat flux over Xisha, using the observed station wind speed and specific humidity, by applying the transfer coefficients of air-sea flux obtained by Jiang et al. [2004] based on Xisha observations. Our computation is similar to but is more direct than the approach of Wang and Li [1993] who calculated specific humidity from temperature in a simple tropical atmosphere model. It is found that latent heat flux decreased from 94 W m−2 during 1958–1977 to 88 W m−2 during 1978–2004. More significant drop occurred from the 1970s to the 1980s. According to the wind-induced surface heat flux theory [Emanuel, 1987], precipitation should decrease during the latter period when local heat flux becomes smaller.

[16] Figure 3 shows the difference in the vertical profile of temperature over Xisha between 1978–2004 and 1958–1977. It is seen that the escalation of temperature from the first period to the second period occurs only at the lower troposphere, below the 500-mb level. The largest increase appears at the surface (1000 mb). At the mid-upper troposphere (500 mb and above), temperatures decrease from 1958–1977 to 1978–2004, especially at 300 mb. Thus the tropospheric atmosphere becomes relatively unstable from the earlier period to the latter period.

Figure 3.

Vertical profile of the difference (1978–2004 minus 1958–1977) of JJA temperature observed over the Xisha Islands. Units are in 0.1 degree Celsius.

[17] Figure 4 shows the variability of total cloud amount and precipitation over the Xisha Islands. One may anticipate that more cloud amount and precipitation appear during the latter period when the atmosphere is more unstable. However, Figure 4 indicates that the cloud amount decreases from the first period to the second period. In fact, the reduction in clouds occurred in the mid-1970s instead of 1978. The main feature of the change in Xisha precipitation is that the precipitation becomes less variable during the second period, especially since the 1980s. This feature can be substantiated by the standard deviation of 1.23 for the earlier period and 0.81 for the latter period. In short, on the timescales studied, the warming in Xisha SST is consistent with the decrease in the overlying total clouds and precipitation (see Table 1). Considering the domain-closed feature of SCS, the changes in large-scale atmospheric circulation patterns should play a critical role in controlling the variability of the local SST. We will discuss this feature in more detail later.

Figure 4.

Normalized fields of (a) JJA total cloud amount and (b) precipitation observed over Xisha. In Figure 4a, the solid lines measure the mean values of normalized cloud for respective periods of time (1958–1977 and 1978–2004), and in Figure 4b, the dashed lines denote the standard deviations of normalized precipitation for the respective time periods.

[18] Table 1 further quantifies the time means and standard deviations of the various fields for periods 1958–1977 and 1978–2004, as well as the differences between the two periods. The table indicates that the decreases in kinetic energy (K), meridional wind (v), and total cloud amount (N) and the increases in SST, Ts, and their difference (SST-Ts) from the first period to the second period are all significant, exceeding the 99% confidence level based on the Student t test. For SST-Ts, the change in its standard deviation is also significant at the 97% confidence level. It is also seen that the monthly mean values of the zonal and meridional winds are positive (westerly and southerly flows) for both periods. Thus the southerlies over Xisha, which are much stronger than the local westerlies, decrease significantly from 1958–1977 to 1978–2004.

[19] In addition, Table 1 shows that the precipitation over Xisha decreased from 1958–1977 to 1978–2004. However, like the slight increase in the zonal wind, the decrease in the precipitation rate is insignificant. Instead, the changes in the standard deviations of both zonal wind and precipitation are statically significant, which have been previously seen in Figures 1c and 4b.

[20] The availability of daily Xisha observations has also enabled us to analyze detailed features of the link of climate to weather conditions. Figure 5 shows the histograms of daily values of zonal and meridional wind components over Xisha. Figure 5a reveals that, from 1958–1977 to 1978–2004, all easterlies become less frequent and weak westerlies (weaker than 4 m s−1) appear more frequently. Both of these changes tend to increase the westerly component of the winds. However, strong westerlies (stronger than 4 m s−1) become less frequent during the latter period, offsetting the increase in westerly component caused by the changes in the easterlies and the weak westerlies. For the meridional wind component, an increase in frequency from 1958–1977 to 1978–2004 appears in weak southerlies (up to 6 m s−1). Strong southerlies of 6 m s−1 and above occur less frequently during 1978–2004. Thus the decrease in meridional wind from the first period to the second period shown in Figure 1b is mainly caused by the decrease in the frequency of strong southerlies although the small increase in weak northerlies (up to −3 m s−1) also supplies to the change in the mean meridional wind. Within this context, the decrease in the frequency of strong southerlies plays a larger role than the increase in the frequency of weak southerlies in contributing to the decrease in the mean meridional wind (see Figure 1b).

Figure 5.

Histogram depicting the frequency of occurrence of (a) JJA daily zonal wind and (b) meridional wind of Xisha, over different ranges of wind speed, for periods 1958–1977 and 1978–2004. Units are in percentage.

[21] Figure 6 shows the histograms of daily temperature, cloud amount, and precipitation. From 1958–1977 to 1978–2004, the frequency of Xisha surface temperature of 29°C and higher increases clearly but the frequency of lower temperature decreases (Figure 6a). That is, the increase in temperature from the first period to the second period shown in Figure 2 is mainly caused by the high-temperature bands. Figure 6b indicates that the decrease in total cloud amount shown in Figure 4a is due to the reduction in cloud amount when the coverage of cloud is 90% and more. The frequency distribution of precipitation exhibits mixed signals (Figure 6c). Rainfall is less frequent in the bands of 0.1–1 and 5–10 mm per day during 1978-2004 compared with 1958–1977. However, it is more frequent in other bands, especially in the category of trace amount of precipitation and when precipitation is heavier than 50 mm per day.

Figure 6.

Histogram depicting the frequency of occurrence of (a) JJA daily surface temperature, (b) total cloud amount, and (c) precipitation of Xisha, over different data ranges, for periods 1958–1977 and 1978–2004. Units are in percentage.

[22] It should be pointed out that there exists consistency in the long-term differences in frequency distribution between the various parameters shown in Figures 5 and 6. For example, during the latter period (1978–2004), a reduction in the frequency of strong southerly wind decreased evaporation, which increased the occurrence of high SST. The feature that shows that decreases in frequency occurred only in 90–100% for cloud coverage and mainly in the band of 0.1–1 mm per day for precipitation (light precipitation) suggests the importance of stratus clouds whose reduction enhanced shortwave radiation and increased SST.

[23] Since we will analyze the large-scale atmospheric circulation patterns in the NNRA and ERA reanalyses associated with the long-term changes in Xisha observations, it is interesting to know how the reanalysis data are similar to the station observations. In Figure 7, several fields including the zonal and meridional wind components, Ts, and precipitation in the Xisha observations are compared with those in the reanalysis and PREC/L products. We use the grid point of 17.5°N, 112.5°E for the reanalyses and the box-centered 16.75°N, 112.25°E for PREC/L precipitation. Although these locations are not the same as Xisha, they are the closest points of the observatory station (16.83°N, 112.33°E). As seen from Figures 7a and 7b, there is a large agreement between the Xisha observations and the reanalyses for the wind components. For the zonal wind, the coefficients of correlation in which the linear terms have been removed are 0.74 between Xisha and NNRA and 0.77 between Xisha and ERA. For the meridional wind, they are 0.77 for Xisha-NNRA and 0.62 for Xisha-ERA, indicating a better match between Xisha and NNRA. More importantly, as seen from Figure 7b, the meridional wind components in both NNRA and ERA drop consistently (with the Xisha observation) in 1978.

Figure 7.

Comparisons of normalized JJA fields for (a) zonal wind, (b) meridional wind, and (c) surface temperature between Xisha, NNRA, and ERA, and for (d) precipitation between Xisha and precipitation analysis PREC/L.

[24] Figure 7c shows weaker relationships for Ts, R = 0.58 for Xisha-NNRA, and R = 0.4 for Xisha-ERA. Surprisingly, the correlation coefficient is only 0.18 between NNRA and ERA, indicating a substantial discrepancy in the temperature near Xisha between the two reanalysis products. For precipitation, there is a very close relationship (R = 0.88) between the Xisha observation and the 0.5° × 0.5° (latitude/longitude) precipitation analysis PREC/L in which island observations are included [Chen et al., 2002]. However, there is no apparent relationship (R = 0.2) between the Xisha precipitation and the 2.5° × 2.5° PREC precipitation of the closest grid point at 16.25°N, 111.25°E.

4. Associated Large-Scale Features

[25] As shown in the last section, there remain discrepancies between Xisha observations and reanalysis products, and between different reanalysis products. Previous studies [Yang et al., 2002; Inoue and Matsumoto, 2004] have already discussed the data problem for Asian climate studies. In particular, Yang et al. [2002] have pointed out the potential problem in NNRA due to an encoding error that causes discontinuity in data quality over the eastern Asian continent before and after 1968. Thus we examine the large-scale features associated with the long-term changes in Xisha observations in both NNRA and ERA.

[26] Figure 8 shows the difference patterns in winds and geopotential height between 1978–2004 and 1958–1977. Let us first focus on NNRA (Figures 8a and 8b). Over the Asian continent, a strong differential high appears, centered in the extratropics of East Asia. The high weakens with altitude significantly. At the lower troposphere, the high occupies a large part of East and Southeast Asia and is accompanied by weaker southwest monsoon flow over Indo-China peninsula and eastern China. It causes anomalous divergence over southern Asia but convergence over the equatorial regions. At the upper troposphere, the northwesterly flow over Indonesia associated with the cyclonic patterns over the southern Indian and Pacific oceans causes anomalous divergence over the equatorial regions including the southern SCS and part of Indonesia. That is, compared with the ones in 1958–1977, the larger values of geopotential high appear over East Asia and anomalous low-level divergence dominates over southern Asia during 1978–2004. Also, in NNRA, the trade winds weaken over the tropical central-eastern Pacific, leading to weaker local Walker circulation during 1978–2004.

Figure 8.

Differences in JJA winds (vectors; in meter per second) and geopotential height (contours; in meter) between 1978–2004 and 1958–1977 for (a and b) NNRA and (c and d) ERA. Values are shown for both levels of (a and c) 200 mb and (b and d) 850 mb.

[27] Figure 8b also shows that, during the latter period, the anomalous northerly flow along the East Asia coast associated with the anomalous high to the north merged with the anomalous northward cross-equatorial flow over Indonesia, which enhanced local moisture convergence and thus precipitation. The enhancement of near-equatorial convection weakened the local summertime meridional cell and thus suppressed the convection over SCS and northward (also see Figure 10 later), providing an additional mechanism for explaining the decrease in cloud amount and precipitation over Xisha during 1978–2004. Together with the weaker large-scale atmospheric circulation (smaller latent heat flux from the ocean), the reduction in cloud amount increased SST because of larger radiative heat flux into the ocean.

[28] The features described above for NNRA also generally appear in ERA, but noticeable differences exist. For example, the increase in lower tropospheric height over East Asia, the low-level convergence and high-level divergence over the equatorial regions, and the weakening of the Walker circulation over the tropical Pacific during 1978–2004 can all be found in Figures 8c and 8d. However, these features are weaker than those in NNRA (compared with Figures 8a and 8b). Over East Asia, the lower tropospheric high decreases rapidly with altitude and becomes low (see Figure 8c). The weakening of the Walker circulation over the central-eastern Pacific is also less apparent.

[29] From 1958–1977 to 1978–2004, the tropics become warmer. This feature can be seen from both NNRA and ERA-40 (Figure 9) in spite of the differences in amplitude and regional features between the two products. For this study, the most interesting feature in Figure 9 is the warming over the Indian and western Pacific oceans and the cooling over the Asian continent. These ocean warming and land cooling mark weaker thermal contrast between the continent and the oceans, leading to smaller temperature gradient across East-Southeast Asia including the Xisha Island during the latter period. Such a long-term decrease in thermal contrast and temperature gradient is more obvious in NNRA than in ERA-40. However, in both reanalysis products, the signals of temperature appear in the entire troposphere. The result, especially that shown in Figure 9b, is consistent with the finding of Zhang et al. [2004] who analyzed the difference in surface temperature between 1980–1993 and 1962–1976 in NNRA and in independent terrestrial air temperature data set (see their Figures 4b and 10d). Note that the warming of the tropical oceans is also associated with the weakening of the tropospheric atmospheric circulation as seen in Figure 8 previously. Such association is consistent with the result of the study of Xu et al. [2007] who linked a steady decline of East Asian summer monsoon to the warming in SCS and the cooling in central South China.

Figure 9.

Differences in JJA air temperature (in degree Celsius) between 1978–2004 and 1958–1977 for (a and b) NNRA and (c and d) ERA, for (a and c) 200 mb and (b and d) 850 mb. Values significantly exceeding the 95% confidence level are shaded.

[30] We further analyze the large-scale features of the long-term precipitation change by examining the difference in precipitation patterns. We depict the difference (1978–2004 minus 1958–1977) in precipitation from the reconstructed precipitation data set [Chen et al., 2002] in resolution of 2.5° × 2.5° latitude/longitude in Figure 10a (PREC; over land, island, and oceans) and in resolution of 0.5° × 0.5° in Figure 10b (PREC/L; over land and island only). Although the Xisha station precipitation is closely related to the precipitation of the nearest grid points only in PREC/L but not in PREC, the precipitation in PREC is also analyzed here to depict the large-scale features of precipitation patterns, assuming that the large-scale patterns are more reliable than local features in the analysis product. The main features shown in Figure 10a include the long-term decrease in precipitation over tropical Asia (30°N and 30°S) including SCS where Xisha is located, the Bay of Bengal, the Arabian Sea, the tropical Indian Ocean, and the far western Pacific. They also comprise the increase in precipitation over the tropical Southern Hemisphere in and near Indonesia. That is, from 1958–1977 to 1978–2004, precipitation decreases over a large part of southern Asia including Southeast Asia. This decrease in precipitation is clearly consistent with the weakening in the thermal contrast between the Asian continent and the surrounding oceans shown in Figure 9. The higher-resolution product (Figure 10b) also shows a long-term decrease in precipitation over Southeast Asia including the Philippines and other regions.

Figure 10.

Differences in JJA precipitation (in millimeter per day) between 1978–2004 and 1958–1977 (a) for PREC of resolution of 2.5° × 2.5° latitude/longitude over land and oceans and (b) PREC/L of resolution of 0.5° × 0.5° latitude/longitude over land and islands only. Positive values are shaded.

[31] It should be pointed out that, although we have established a relationship between the smaller ocean-land thermal contrast, weaker atmospheric circulation, and overall reduction of large-scale precipitation pattern for 1978–2004, the change in large-scale ocean-land thermal contrast cannot effectively explain the change in regional precipitation such as the regional feature within southern Asia. The change in local precipitation is associated with not only wind speed but also wind directions in atmospheric cyclonic or anticyclonic patterns. It is also linked closely to local surface latent heat flux as discussed previously.

5. Summary and Discussion

[32] In this study, we have depicted the long-term changes in the South China Sea (SCS) summer monsoon as shown in the observations of the Xisha Islands and have explained the associations of these changes with large-scale atmospheric and oceanic patterns. The SCS monsoon is an important component of the Asian monsoon system, and its long-term changes have seldom been explored because of the unavailability of reliable data. The long-recorded, daily observations of the Xisha station are also an important source of information for understanding the link between monsoon climate and monsoon weather.

[33] The intensity of the SCS summer monsoon measured by kinetic energy decreases significantly from 1958–1977 to 1978–2004. This decrease in monsoon intensity is mainly caused by weakening of the meridional component of lower tropospheric winds. The weakening in seasonal mean flows is reflected by less frequent occurrence of strong southerlies (6 m s−1 and over) of the daily winds. The weakening of the monsoon is also associated with warming at ocean surface and lower tropospheric atmosphere over SCS during the latter period. Specifically, surface warming occurs when daily temperature of 29°C and higher appears more frequently.

[34] The long-term warming of the lower troposphere is accompanied by cooling at the upper troposphere, which indicates that the atmosphere became more locally unstable. However, from 1958–1977 to 1978–2004, the long-term change in Xisha precipitation tended to decrease; in addition, in fact, the station precipitation became less variable. Thus besides local air-sea interaction, large-scale atmospheric forcing also plays an important role in causing the long-term change of the Xisha precipitation. Indeed, the warming of Xisha is clearly linked to large-scale warming in the tropics especially in tropical oceans including the SCS and to cooling in the Asian continent. These changes in temperature weaken the thermal contrast between the Asian continent and the surrounding oceans, and the reduction in land-sea thermal contrast lessens the intensity of monsoon circulation.

[35] Several data fields in the NCEP-NCAR and ERA-40 reanalysis and other products have been compared against the observations of Xisha station. Relatively, the wind fields in the reanalyses and the precipitation in a high-resolution precipitation product (PREC/L) are compared well with the station observations. Overall, the long-term change in the Xisha climate may be part of the so-called climate shift that occurred in the mid-1970s on the global scale [Graham, 1994; Miyakoda et al., 2003]. The Xisha station data confirm that the climate shift in the reanalysis products over the SCS regions is not an artifact although the reanalysis data are different from the station observations in details.

[36] In this study, we have only discussed the long-term change in the SCS summer monsoon before and after the mid-1970s. Changes in climate on such long timescale should also exist in other seasons, and the changes in different seasons should be characterized by different features and be explained by different mechanisms [e.g., Hu et al., 2003; Xu et al., 2007]. It is interesting to conduct further analyses to understand these seasonally dependent features of long-term changes of the SCS monsoon using Xisha observations.

[37] As reviewed in section 1, the onset of SCS monsoon is one of the most active research areas in studying the regional monsoon. This monsoon onset marks not only unique features of the monsoon itself but also signals as precursors of the central-southern China precipitation and the Indian summer monsoon rainfall [Yang and Gutowski, 1992]. Therefore it will also be interesting to understand how the onset date of SCS monsoon changes before and after 1977–1978 and how this change is linked to the long-term changes in other monsoon components.


[38] This work was partially supported by the National Natural Science Foundation of China (grant 90211010) and the bilateral program of the China Meteorological Administration and the U.S. National Oceanic and Atmospheric Administration. The Meteorological Bureau of Hainan Province of China provided the Xisha observational data. We thank C.-P. Chang and two anonymous reviewers for their comments, which are helpful in improving the overall quality of the manuscript.