Jiangyu Mao Æ Zhang Sun Æ Guoxiong Wu

Clim Dyn (2010) 34:747 761 DOI 10.1007/s00382-009-0628-2 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations in the subtropical high over the western North Pacific and South China Sea Jiangyu Mao Æ Zhang Sun Æ Guoxiong Wu Received: 25 September 2008 / Accepted: 3 July 2009 / Published online: 25 July 2009 Ó Springer-Verlag 2009 Abstract The spatio-temporal variability in summer rainfall within eastern China is identified based on empirical orthogonal function (EOF) analysis of daily rain-gauge precipitation data for the period 1979 2003. Spatial coherence of rainfall is found in the Yangtze Basin, and a wavelet transform is applied to the corresponding principal component to capture the intraseasonal oscillation (ISO) of Yangtze rainfall. The ensemble mean wavelet spectrum, representing statistically significant intraseasonal variability, shows a predominant oscillation in summer Yangtze rainfall with a period of 20 50 days; a 10 20-day oscillation is pronounced during June and July. This finding suggests that the 20 50-day oscillation is a major agent in regulating summer Yangtze rainfall. Composite analyses reveal that the 20 50-day oscillation of summer Yangtze rainfall arises in response to intraseasonal variations in the western North Pacific subtropical high (WNPSH), which in turn is modulated by a Rossby wave-like coupled circulation convection system that propagates northward and northwestward from the equatorial western Pacific. When an anomalous cyclone associated with this Rossby wavelike system reaches the South China Sea (SCS) and Philippine Sea, the WNPSH retreats northeastward due to a reduction in local pressure. Under these conditions, strong monsoonal southwesterlies blow mainly toward the SCS Philippine Sea, while dry conditions form in the Yangtze J. Mao (&) Z. Sun G. Wu State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG), Institute of Atmospheric Physics, Chinese Academy of Sciences, P.O. Box 9804, 100029 Beijing, China e-mail: mjy@lasg.iap.ac.cn Z. Sun Zhejiang Meteorological Observatory, Hangzhou, China Basin, with a pronounced divergent flow pattern. In contrast, the movement of an anomalous anticyclone over the SCS Philippine Sea results in the southwestward extension of the WNPSH; consequently, the tropical monsoonal southwesterlies veer to the northeast over the SCS and then converge toward the Yangtze Basin, producing wet conditions. Therefore, the 20 50-day oscillation of Yangtze rainfall is also manifest as a seesaw pattern in convective anomalies between the Yangtze Basin and the SCS Philippine Sea. A considerable zonal shift in the WNPSH is associated with extreme dry (wet) episodes in the Yangtze Basin, with an abrupt eastward (westward) shift in the WNPSH generally leading the extreme negative (positive) Yangtze rainfall anomaly by a 3/8-period of the 20 50-day oscillation. This finding may have implications for improving extended-range weather forecasting in the Yangtze Basin. Keywords Summer Yangtze rainfall Intraseasonal oscillation Subtropical high The western North pacific 1 Introduction Since Madden and Julian (1971, 1972) discovered 40 50- day eastward-propagating oscillation in the tropics, intraseasonal oscillation (ISO) in the broad period range from 30 to 60 days has been identified in the Asian summer monsoon regime (e.g., Krishnamurti and Subrahmanyam 1982; Murakami and Nakazawa 1984; Wang and Ding 1992). However, this ISO propagates not only eastward but also northward (e.g., Lau and Chan 1986; Wang and Rui 1990). In the western North Pacific, such 30 60-day convective activity even shows northwestward propagation

748 J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations (e.g., Kawamura et al. 1996; Hsu and Weng 2001; Tosu et al. 2005). Another dominant ISO that modulates the activity of the Asian summer monsoon is 10 20-day westward-propagating oscillation (e.g., Krishnamurti and Ardauny 1980; Chen and Chen 1995; Mao and Chan 2005). As an important component of the Asian summer monsoon system, the East Asian summer monsoon (EASM) involves both tropical and subtropical monsoons, encompassing a large area spanning the equatorial regions of Southeast Asia, the South China Sea (SCS), and the East China Sea to regions in eastern China, Japan, and Korea (Lau et al. 1988; Lau and Yang 1996). The term EASM is sometimes restricted to the subtropical monsoon that prevails over eastern China north of 20 N, Korea, and Japan (Chen et al. 2000; Mao and Wu 2006). According to this definition, EASM is dominated by the Meiyu (eastern China) Baiu (Japan) front, which is climatologically one of the major convergence zones within the global atmospheric circulation (Chen et al. 2000). The formation and maintenance of the Meiyu front along the Yangtze Basin in eastern China depends on the convergence between tropical flow from warm oceans and cold air from middle to high latitudes; thus, tropical extratropical interaction is an inherent feature of the East Asian monsoon (e.g., Tanaka 1992; Fukutomi and Yasunari 2002; Hsu 2005). The distribution of summer rainfall in eastern China generally involves three east west bands of heavy rainfall related to the northward migration of the Meiyu front across the Yangtze River (Tao and Ding 1981). However, this northward migration is not uniform: it shows a north south ISO induced by the eastward propagation of the intraseasonal global divergent circulation (Chen and Murakami 1988). Lau et al. (1988) analyzed the seasonal and intraseasonal progression of the monsoon rain zone in eastern China, including the establishment and shift of the Meiyu front, and suggested that the northward progression and rapid transition of these major rainbands may result from the phase lock between the 40- and 20-day intraseasonal monsoon modes. Because the moisture and moisture flux in the SCS western Pacific fluctuate intraseasonally (Chen et al. 1988), opposite-phase intraseasonal variations exists between the SCS and Yangtze Basin, as observed from temporal evolutions in filtered monsoon indices calculated from outgoing longwave radiation and 850-hPa zonal winds (Chen et al. 2000). The opposite-phase variations between the two monsoon elements arise from anomalous circulation associated with the northward-migrating 30 60-day monsoon trough/ridge from the equator to 20 N, and with the westward-propagating 12 24-day monsoon low high along the latitudes between 15 N and 20 N. Tsou et al. (2005) and Mao and Chan (2005) reported that during the extreme phases of the 30 60-day oscillation, the convective anomaly in the SCS is out of phase with that in the Yangtze Basin, suggesting a physical linkage between ISOs of the summer monsoon in the Yangtze Basin and the region from the SCS to the western North Pacific. Previous studies have reported the influence of the subtropical high in the troposphere on the surrounding weather and climate (e.g., Huang and Yue 1962; Tao and Zhu 1964; Chang et al. 2000a, b). In fact, the western North Pacific subtropical high (WNPSH) in the middle and lower troposphere, and the South Asian high in the upper troposphere, are important components of the EASM system (Tao and Chen 1987), because the WNPSH separates the Meiyu front from the SCS monsoon trough that extends from the northern Indochina Peninsula to the Philippine Sea. The meridional displacement and zonal shift of the WNPSH ridge influence the location at which the southeasterlies on the southern flank of the WNPSH converge with the southwesterlies from the Somali cross-equatorial flow, thereby affecting EASM rainfall (Tao and Chen 1987). Given that the low-level southwesterly jet at the northwestern edge of the WNPSH is responsible for moisture transport over East Asia, Lu (2001) examined interannual variations in the summer WNPSH. A pronounced variability in the WNPSH was found to exist at its western extent, with a zonal shift in the WNPSH being related to convective intensity in the warm pool. Lu emphasized that this zonal shift in the WNPSH plays a crucial role in influencing the transport of water vapor and resultant rainfall in East Asia. In investigating the cause of extreme floods in eastern China during 1998, Zhu et al. (2003) suggested that the floods were related to the 30 60-day ISO activity in the western North Pacific, where the monsoon trough and subtropical anticyclone appeared as an anticlockwise propagation with enhanced and suppressed convective anomalies in a 30 60-day period. In a study of the 1991 extreme floods in the Yangtze Basin, Mao and Wu (2006) found that the active and break sequences of rainfall in the middle and lower parts of the basin were controlled mainly by a 15 35-day oscillation, with an anomalous low-level cyclone (anticyclone) appearing alternately over the northern SCS and the Philippine Sea, leading to a northeastward (southwestward) shift of the western North Pacific subtropical anticyclone over the SCS, and hence producing a lower-tropospheric divergence (convergence) over the Yangtze Basin. This ISO was associated with a dipole anomaly in the upper troposphere, characterized by an anomalous cyclone (anticyclone) over eastern China and an anomalous anticyclone (cyclone) over the northern Tibetan Plateau, resulting in a southwestward shrinking (northeastward expansion) of the South Asian anticyclone and the formation of a convergence (divergence) over eastern China.

J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations 749 The above case studies show that the ISO of summer Yangtze rainfall is regulated by intraseasonal variations in the subtropical anticyclone within the lower troposphere, and that extreme floods in the Yangtze Basin appear to be related to strong ISO activity. Does such a situation occur in all summers? What characteristic periods of ISOs of summer Yangtze rainfall are there in most years? These questions are to be addressed in the present study. Because intraseasonal variability in Yangtze rainfall is associated with low-level circulation anomalies over the western North Pacific and SCS, it is also possible that intraseasonal variations in the WNPSH are modulated by these anomalies. Lau and Chan (1986) and Chen and Murakami (1988) noted that intraseasonal convection anomalies propagate northward and northwestward in the western North Pacific and SCS. Hsu and Weng (2001) further examined the spatial and temporal characteristics of ISO in the western North Pacific, as well as the nature of its northwestward propagation, and found that enhanced (suppressed) convection is coupled with the low-level cyclonic (anticyclonic) anomaly. This coupled circulation convection system shows a Rossby wave-like structure and propagates northwestward and/or westward toward southeastern China; the positive feedback between the anomalous circulation and convection results in rapid enhancement of the system. Northwestward propagation of the coupled circulation convection system in the western North Pacific is thought to result from moisture convergence at the northwestern corner of the convection (Hsu and Weng 2001), within which moisture is transported by the strengthened southwesterlies associated with enhanced cyclonic circulation. Because a potentially unstable atmosphere exists in the region northwest of the area of deep convection (arising from moisture convergence), convection continues to operate with the help of near-surface convergence and upward motion. Consequently, the deep convection propagates northwestward, triggered in part by surface friction, as near-surface moisture convergence arising from surface friction is located to the northwest of the deep convection. During the course of northwestward propagation, ocean atmosphere interaction plays an important role in supplying the energy required to sustain the circulation and convection. Thus, this convectively coupled Rossby wave-like mechanism may help to explain intraseasonal variations in the WNPSH, which is modulated by the northwestwardpropagating ISO in the western North Pacific. Tsou et al. (2005) demonstrated that the combined effect of surface frictional-diabatic heating and vorticity advection causes the 30 60-day convection and circulation to simultaneously develop and propagate northwestward in the western North Pacific and SCS. Therefore, the objective of the present study is to systematically examine the behavior of the predominant ISO of summer Yangtze rainfall based on long-term daily precipitation data and to identify modulation of the WNPSH by the ISO system in the western North Pacific and SCS. Section 2 describes the data and methods used in this study, while the spatio-temporal variability associated with summer Yangtze precipitation is outlined in Sect. 3. The spatial structure and propagation of the ISO that regulates summer Yangtze rainfall are examined in Sect. 4. Section 5 considers intraseasonal modulation of the WNPSH and its impact on Yangtze rainfall. Finally, a summary and discussion are given in Sect. 6. 2 Data and methods 2.1 Data The basic data analyzed in this study are daily averaged rain-gauge precipitation in China, as provided by the National Meteorological Information Center, China Meteorological Administration (CMA). For eastern China (defined here as 20 40 N and 100 125 E), data are available from 368 observation stations. To highlight the intraseasonal variability, the time series from each station for the summer season (defined here as the period from 1 May to 31 August) is smoothed using a 5-day running mean to remove high-frequency fluctuations. Because raingauge precipitation data are unavailable over oceanic areas, daily outgoing longwave radiation (OLR) data (Liebmann and Smith 1996) from the National Oceanic and Atmospheric Administration (NOAA) are used to represent large-scale tropical convective activity (e.g., Waliser et al. 1993) over tropical oceans and land. The primary atmospheric circulation data are extracted from National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis products (Kalnay and Coauthors 1996), which contain daily-mean wind, geopotential height, and other variables. Both OLR and NCEP/NCAR reanalysis data have a horizontal resolution of 2.5 latitude 9 2.5 longitude. Continuous OLR records are available from 1979; consequently, we consider the 25 summers from 1979 to 2003. Based on theoretical studies of the response of the tropical atmosphere to diabatic heating (Webster 1972; Gill 1980), diabatic heating is thought to play an important role in the evolution of ISO. Thus, daily diabatic heating (apparent heating) is calculated as the residual of a largescale heat budget, and is vertically integrated from the surface to 300 hpa (Yanai et al. 1973). 2.2 Methods To identify the dominant ISO of Yangtze rainfall, it is necessary to define a reference time series that well

750 J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations represents the evolution of rainfall in the Yangtze Basin during each summer. In this approach, a time series at a given station is generally selected to represent the situation in the Yangtze Basin. Alternatively, a domain could be chosen subjectively to include stations along the Yangtze River, with the area-averaged rainfall over this domain being defined as the Yangtze rainfall index. However, these methods may not necessarily reflect temporal variations in rainfall throughout the entire Yangtze Basin. Therefore, in the present study we applied an empirical orthogonal function (EOF) analysis to 5-day running means of raingauge precipitation anomalies in eastern China for the period May August from 1979 to 2003. The time mean state for each station, defined as the climatological daily means for the period May August, is subtracted from daily rainfall data before EOF analysis, similar to the method used by Hsu et al. (1999). The climatological daily means are derived from daily rainfall data from 1979 to 2003 according to calendar dates. Because the principal component (PC) of an EOF mode shows how the spatial pattern of this mode oscillates over time, we use the corresponding PC of the dominant mode as the reference time series for each summer; the spatial pattern of this dominant mode provides a better reflection of large, coherent rainfall variations throughout the entire Yangtze Basin (see Fig. 1b). Wavelet analysis is a useful tool for decomposing a time series into the time frequency domain (e.g., Torrence and Compo 1998). It is also a powerful bandpass filtering technique, as the wavelet transform is based on a known response function (the wavelet function). Thus, the time series of the ISO component can be reconstructed by the inverse transform over a range of scales. Following Mao and Wu (2006), we use wavelet analysis to identify the dominant ISO modes of Yangtze rainfall and to isolate ISO components. Likewise, the corresponding ISO signals are extracted from other meteorological fields such as OLR and 850-hPa winds. The wavelet basis function selected here is the sixth-order derivative of a Gaussian function (Torrence and Compo 1998), because abrupt temporal variations occur in Yangtze rainfall. Composite analysis is applied to several variables (e.g., OLR and 850-hPa winds) to identify the spatial structure and temporal evolution of the ISO. The statistical significance of composite fields is estimated based on Student s t-test. 3 Spatio-temporal variability in summer Yangtze rainfall 3.1 Spatial pattern of rainfall and temporal evolution EOF analysis is an objective and straightforward approach to identifying the spatio-temporal variability of summer precipitation in eastern China. Figure 1 shows the spatial patterns of the first two EOFs of the 5-day running-mean daily precipitation anomalies for the period May August from 1979 to 2003, with the first mode (EOF1) and second mode (EOF2) accounting for 11.74 and 9.58% of the total variance, respectively. Each of these two modes is separated from the subsequent mode based on the test of statistical significance proposed by North et al. (1982). EOF1 (Fig. 1a) shows a north south-oriented dipole pattern with opposite rainfall anomalies to the south and north of the Yangtze River, which captures the seasonal northward migration of the heavy rainfall belt. The corresponding principal component (PC1) time series shows year-to-year differences in seasonal (as well as intraseasonal) change in monsoon rainfall within eastern China (Fig. 2); the seasonal change is modulated by intraseasonal fluctuations during each summer. In contrast, EOF2 (Fig. 1b) seems to possess a tripole-like structure dominated by positive rainfall anomalies throughout the entire Yangtze Basin; negative anomalies are confined mainly to coast areas of southern China, with minor anomalies in northern China. In fact, EOF2 captures mainly the rainfall variability in the Yangtze Basin, which is the familiar Fig. 1 Spatial patterns of the first two leading EOFs of the 5-day running-mean daily rainfall anomalies over eastern China, for a EOF1 and b EOF2. The loadings (unit mm day -1 ) were scaled by one standard deviation of the corresponding PC. Solid and dashed contours indicate positive and negative loadings, respectively. The percentage at the top of each panel refers to the variance explained by the mode

J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations 751 Meiyu mode (Lau et al. 1988). The principal component (PC2) time series clearly shows intraseasonal variability in Yangtze rainfall and interannual differences in this variability (Fig. 2). Large-amplitude ISOs are clearly discerned during the summers of, for example, 1979, 1991, and 1998. To validate the stability of the spatial patterns, EOF analysis is also applied to the climatological daily rainfall time series, as performed by Lau et al. (1988). That is, before EOF analysis, the climatological rainfall at each station for each day during May August was generated as the mean value for the period 1979 2003. The resulting spatial patterns are similar to those shown in Fig. 1, indicating that EOF1 and EOF2 do indeed represent two major spatial patterns of rainfall variability in eastern China. 3.2 Identification of the dominant ISO of summer Yangtze rainfall Because EOF2 captures the spatial coherence of variability in Yangtze rainfall, the time-varying amplitude of this pattern is reflected in the corresponding PC2 time series for each summer. PC2 is therefore selected as the reference Fig. 2 Principal component time series for PC1 (dashed lines) and PC2 (solid lines) corresponding to EOF1 and EOF2, respectively, for the period May August from 1979 to 2003. The unit is dimensionless

752 J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations time series with which to isolate the dominant intraseasonal signals of Yangtze rainfall. To confirm that PC2 is a suitable index in this regard, for comparison we calculated the area-averaged daily rainfall over the region where the loadings are greater than 1 mm day -1 in the EOF2 mode (Fig. 1b), although such an area-averaged rainfall index may be affected by the choice of domain. Figure 3 shows examples from selected years to illustrate the temporal evolutions of PC2 and the area-averaged rainfall index. Similar variations are found between the two time series for each summer. Wavelet analysis is applied to the corresponding PC2 for each summer from 1979 to 2003. Prior to this analysis, PC2 for each summer is normalized by its standard deviation to eliminate interannual variability. The wavelet power spectra for five individual summers are presented as examples in Fig. 3. For the 1979 summer, wavelet spectrum is statistically significant at the 95% confidence level for periods between 20 and 30 days, while for the 1982 summer the spectrum was most significant at a period of 20 50 days. As reported by Mao and Wu (2006), the dominant ISO period during the 1991 summer was 15 35 days. During the 1993 summer, a pronounced spectrum was observed in a band of \20 days, whereas in 2002 the significant spectrum covered a much wider frequency band from 15 to 60 days. In summary, for the years 1979, 1982, 1991, and 2002, the dominant ISO periods were largely greater than 20 days, while for the year 1993 the dominant ISO period was less than 20 days. Varying dominant periods were found for the other 20 analyzed summers based on significant power spectra (not shown). It is clear that the dominant ISO periods of Yangtze summer rainfall vary from year-to-year. To obtain the characteristic ISO periods for most of the analyzed years, we averaged the wavelet power spectra of 25 individual summers to generate an ensemble mean spectrum (Fig. 4), following Lawrence and Webster (2001). Note that the ensemble mean only includes those summer spectra statistically significant at the 95% confidence level. Wavelet-derived spectra were selected because they provide an accurate estimate of the total variance within a period band (Lawrence and Webster 2001), in which short-term fluctuations are analyzed with a narrow window and long-term fluctuations are analyzed Fig. 3 Left-hand panels area-averaged daily rainfall (histogram mm day -1, scale on the left-hand ordinate) over the Yangtze Basin (32.5 36 N, 110 125 E) for selected years (shown in the upper left of each panel), and the corresponding PC2 (solid line, dimensionless, scale on the right-hand ordinate) and 20 50-day filtered PC2 (dotdashed line, dimensionless, scale on the right-hand ordinate). The numbers 1, 3, 5, and 7 in the central panel shows the phases of the 20 50-day oscillation (see Sect. 3 for details). Right-hand panels wavelet power spectrum (shading) of PC2 shown in the left-hand panels. Solid contours encloses regions with confidence level [95% for a red-noise process with a lag-1 coefficient of 0.90. Dashed lines indicate the cone of influence outside which edge effects become important (Torrence and Compo 1998). The spectrum values corresponding to the shades of gray are indicated in the scale bar

J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations 753 Fig. 4 Ensemble mean wavelet power spectrum. Wavelet analysis of the corresponding PC2 was first completed for each individual year. Next, the 25 individual wavelet spectra for the period 1979 2003 were averaged to generate the ensemble mean spectrum. The spectrum values corresponding to the shades of gray are indicated in the scale bar with a wide window. The ensemble mean spectrum shows maximum power at periods mainly between 20 and 50 days from late May to early August, with maxima occurring around the mean period of 35 days, although spectral values are also appreciable at periods above 50 days. Such a characteristic ISO period (of 20 50 days) is consistent with the results of Lau et al. (1988), who identified 40- and 20-day oscillations in the rainfall climatology of eastern China based on an EOF analysis of 10-day mean rainfall records. The above findings suggest that 20 50-day oscillation is a major agent in regulating summer Yangtze rainfall. Moreover, spectral values at periods of 10 20 days are pronounced from mid-june to late July, indicating that this may be another characteristic period for these months. In addition to 1993 (Fig. 3), the 10 20-day oscillation also occurred in other years. For example, a 12 24-day oscillation was dominant during the summer of 1989 (Chen et al. 2000). For the remainder of this study we focus on the 20 50-day oscillation of Yangtze rainfall; the 10 20-day mode will be explored in a separate study. To reveal the origin of the 20 50-day Yangtze rainfall oscillation, the behavior and thermodynamic structure of this ISO mode are investigated in terms of low-level circulation, OLR, and diabatic heating. The ISO signals associated with these variables are obtained by applying the 20 50-day wavelet filter for each summer. The 20 50- day filtered PC2 time series show that several cycles with different amplitudes occurred in each summer from 1979 to 2003. As in Mao and Wu (2006), a strong ISO cycle is defined as a cycle with positive and negative anomalies (or wet and dry periods), each with a peak amplitude greater than half of one standard deviation from zero. Based on this criterion, 44 strong cycles were selected for composite analyses. Each strong cycle is further divided into eight phases (see Fig. 3). Phase 1 represents the minimum value of the 20 50-day filtered PC2 for the dry period, while phase 3 represents the transition from the dry to the wet period. Phase 5 is the maximum value of the 20 50-day filtered PC2 for the wet period, and phase 7 is the transition from the wet to the dry period. Phases 2, 4, 6, and 8 occur at the time when the oscillation reaches half of its maximum or minimum value. Thus, the structure and propagation of the 20 50-day ISO are examined based on its life cycle, which is determined using a phase compositing technique, as in Mao and Chan (2005). 4 Spatial structure and propagation of the ISO 4.1 OLR and diabatic heating Figure 5 shows the composite evolutions of the 20 50-day filtered OLR and diabatic heating fields. In the driest phase (phase 1, Fig. 5a), a significant positive OLR (negative heating) anomaly in the Yangtze Basin is accompanied by a significant negative OLR (positive heating) anomaly to the south, in a band that extends from the northern SCS to the Philippine Sea. Another positive OLR (negative heating) anomaly exists around the equatorial region. The OLR anomaly is out of phase with the diabatic heating anomaly, as diabatic heating at low latitudes is dominated by convective heating (Hsu and Weng 2001). Of note are the opposite-phase variations in convective anomalies between the Yangtze Basin and the SCS, consistent with the findings of Chen et al. (2000). Tsou et al. (2005) reported similar results, with the strongest 30 60- day convection extending from the SCS to the western North Pacific, and with a dry belt accompanied by anticyclonic vorticity on its northern side. However, their dry belt around equatorial areas was located westward of the dry belt observed in the present study. Note that with the exception of the anomaly in the Yangtze Basin, the remanent spatial pattern, consisting of the anomaly centers in the northern SCS Philippine Sea and in the equatorial region, resembles the wave-like OLR and diabatic-heating

754 J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations Fig. 5 Composite evolutions of the 20 50-day filtered OLR (color scale Wm -2 ) and diabatic heating (contours Wm -2 ) during an ISO cycle of Yangtze rainfall. Phases 1 8 are shown. Contour interval is 10 W m -2. The regions in which the OLR anomaly is statistically significantly at the 95% confidence level are indicated by closed dot-dashed contours (a) (b) (e) (f) (c) (g) (d) (h) structure given in Hsu and Weng (2001, see their fig. 2f) which was derived from the lagged regression coefficients against the area-averaged (0 20 N, 120 160 E) filtered OLR index. These features indicate a connection between the 20 50-day oscillation of Yangtze rainfall and the wavelike convective anomaly system in the western North Pacific and SCS. As the positive OLR anomaly migrates northward from around the equator to north of 5 N and strongly intensifies, simultaneous weakening occurs in the negative OLR anomaly in the northern SCS and the positive OLR anomaly in the Yangtze Basin (Fig. 5b). This major convective suppression in the western North Pacific represents an abrupt westward shift of the WNPSH (as discussed in Sect. 5 below). Note also that compared with the last phase (phase 1), the negative OLR anomaly in the northern SCS decays locally rather than moving further northward. By phase 3, this negative OLR anomaly almost vanishes from the coast of southeastern China and the East China Sea (Fig. 5c), whereas the dominant positive OLR anomaly in the western Northern Pacific is further enhanced, with its westward expansion being more evident than its northward migration. As the positive OLR anomaly continues to expand northwestward and approaches the coast of the SCS, a negative OLR anomaly rapidly develops in the Yangtze Basin (Fig. 5d), forming a seesaw pattern between the Yangtze Basin and the SCS. This finding indicates the response of Yangtze rainfall to the arrival in the SCS of suppressed convection. Note that another negative OLR anomaly appears around the equator, suggesting that the wave-like convective anomaly system originates from the equatorial western Pacific. Subsequently, the newly developed negative OLR anomaly shows a large expansion, whereas the positive OLR anomaly shrinks toward the northern SCS; consequently, both the negative convection anomaly in the Yangtze Basin and the positive convection anomaly in the northern SCS reach peak intensities during phase 5 (Fig. 5e). From phase 5 to phase 6, the negative

J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations 755 OLR anomaly in the western North Pacific shifts from the equator to north of 10 N, reflecting an abrupt eastward retreat of the WNPSH. As a result, the positive OLR anomaly in the northern SCS shows a marked decrease in size, simultaneous with a weakening in convective activity within the Yangtze Basin (Fig. 5f). Yangtze convection is further suppressed with increasing convective activity in the SCS Philippine Sea, developing as a significant heat source (Fig. 5g, h). The above evolutions suggest that the 20 50-day oscillation of Yangtze rainfall is linked to a wave-like convective anomaly system in the western North Pacific and SCS. As this system propagates northward and northwestward from the equatorial region, a positive (negative) convective anomaly occurs alternately in the SCS Philippine Sea. As a response, a negative (positive) convective anomaly appears in the Yangtze Basin, resulting in a seesaw pattern in convective anomalies between the Yangtze Basin and the SCS Philippine Sea. Thus, the origin of the 20 50-day oscillation of Yangtze rainfall lies in the equatorial western Pacific. 4.2 OLR and low-level circulation Hsu and Weng (2001) suggested that the northwestward propagation of the coupled circulation convection system in the western North Pacific is the result of a positive feedback between anomalous circulation and convection. To enable an examination of the impact of such a coupled wave-like system on the Yangtze rainfall anomaly, Fig. 6 shows the composite evolutions of the 20 50-day filtered OLR and 850-hPa wind fields. During phase 1 (Fig. 6a), an anomalous cyclone, accompanied by the negative OLR anomaly, is observed over the northern SCS, with anomalous northeasterlies on the northwestern side of this cyclone prevailing over the Yangtze Basin, resulting in suppressed convection. Associated with the positive OLR anomaly around the equatorial western Pacific is a weak Fig. 6 Composite evolutions of the 20 50-day filtered OLR (color scale Wm -2 ) and 850- hpa winds (vectors ms -1 ) during an ISO cycle of Yangtze rainfall. Phases 1 8 are shown. Open circles indicate grid points where the filtered wind anomalies are significantly different from zero at the 95% level (Student s t-test) in at least one of the wind components (zonal or meridional). Centers of anomalous cyclones and anticyclones are marked by C and A, respectively. The thick solid line from Q (35 N, 110 E) to P (0, 150 E) in a indicates the cross-section Q P along which lagged regression is plotted in Fig. 7 (a) (b) (c) (e) (f) (g) (d) (h)

756 J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations anomalous anticyclone that can be identified by highly anomalous easterlies. The above observations indicate that the extreme reduction in Yangtze rainfall is associated with a wave-like dipole flow pattern characterized by an anomalous cyclone over the northern SCS and an anomalous anticyclone east of the maritime continent. The cyclonic circulation and convection anomalies then dissipate around the coast of southeastern China, whereas the anticyclonic circulation and dry belt rapidly intensify and migrate a large distance northward (Fig. 6b). The diminishment of cyclonic circulation and convection anomalies may result from increased static stability associated with adiabatic cooling (Tsou et al. 2005), reduced solar radiation due to the cloud radiation effect (e.g., Kemball-Cook and Wang 2001), and land surface depletion of the moisture supply (Hsu and Weng 2001) after the convection has fully developed. By phase 3, the anticyclonic circulation has become a dominant system, showing a zonal structure (Fig. 6c). Because of the strengthening and northwestward extension of the anticyclonic circulation, the Yangtze Basin comes under the influence of anomalous southwesterlies, leading to anomalous active convection (Fig. 6d). A weak anomalous cyclone, accompanied by a negative OLR anomaly, occurs around the equatorial western Pacific (Fig. 6e), with anomalous westerlies in the equatorial region. Such a flow pattern is similar to the analytical solution for a symmetric heating source about the equator (Gill 1980), which may reflect the emission of Rossby waves from equatorial active convection (Gill 1980; Kemball-Cook and Wang 2001). However, the anomalous easterlies that represent the equatorial Kelvin wave response are not evident. This is consistent with the finding of Hsu and Weng (2001), who showed that the anomalous low-level circulation developed in response to equatorial convective heating does not resemble a Kelvin-Rossby wave packet. Because these features are to some extent different from theoretical solutions (Gill 1980) and modeling results (Wang and Xie 1997), Hsu and Weng (2001) suggested that the northwestward-propagating coupled circulation convection system has a Rossby wave-like structure. Returning to the present study, the anomalous anticyclone to the north shrinks toward the northern SCS in such a way that the Yangtze Basin is strongly affected by the southwesterly anomalies, causing the strongest convective anomaly. Note that the positive OLR anomaly that accompanies this anomalous anticyclone reaches its strongest intensity, possibly reflecting a positive feedback between anomalous circulation and convection (Hsu and Weng 2001). The anomalous cyclone, together with the anomalous anticyclone, forms a clear dipole pattern in the western North Pacific and SCS. This dipole pattern of the coupled circulation convection system is similar to that identified by Hsu and Weng (2001), indicating a Rossby wave-like behavior. It turns out that the seesaw pattern in convective anomalies between the Yangtze Basin and the SCS Philippine Sea depends on the location of the anomalous anticyclone. Consequently, the convective anomaly in the Yangtze Basin weakens as the anomalous anticyclonic circulation undergoes a strong decay (Fig. 6f), while the anomalous cyclone rapidly develops and migrates northward, which may result from positive feedback via circulation convection interaction (Hsu and Weng 2001). Note that the strong westerly anomalies associated with this cyclone occur mainly on the southern side of the negative convection anomaly (diabatic heating), indicating that the northwestward-moving anomalous cyclone may represent a Rossby wave response to asymmetric convective heating away from the equator (Gill 1980). Wang and Xie (1997) attributed the enhancement of the northwestward-moving Rossby wave to the vertical shear of monsoonal flow. Subsequently, the anomalous cyclone migrates steadily northward and expands northwestward to the north SCS (Fig. 6f, h), showing the opposite patterns in both OLR and winds to those in phases 3 and 4. Also evident is an anomalous anticyclone located northeast of the anomalous cyclone, forming a wave train emanating from the SCS into the extratropical North Pacific. This wave train, similar to that reported by Kawamura et al. (1996) and Tsou et al. (2005), involves a change in the vertical structure of the ISO from baroclinic in the tropics to barotropic in the extratropics. As noted by Kawamura et al. (1996), this lower-tropospheric anomalous cyclone induced by the convective anomaly in the SCS and western North Pacific acts as an origin of the wave train dispersion. Note also that both the anomalous cyclone during phases 7 8 and anomalous anticyclone during phases 3 4 intensify and show slow propagation, possibly resulting from the enhanced positive feedback between convection and circulation anomalies (Hsu and Weng 2001). The entire evolutions from phase 1 to phase 8 show that the 20 50-day oscillation of Yangtze rainfall can be ascribed to a Rossby wave-like coupled circulation convection system that propagates northwestward and northward from the equatorial western Pacific toward the northern SCS, leading to the alternating development of anomalous cyclones and anticyclones over the SCS Philippine Sea. Thus, the northeasterly (southwesterly) anomalies on the northern side of the anomalous cyclone (anticyclone) result in deficient (excessive) rainfall in the Yangtze Basin. The northward and northwestward propagation of the convectively coupled Rossby wave-like system in the

J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations 757 western North Pacific and resultant 20 50-day oscillation of Yangtze rainfall are clearly evident from the lagged regression of OLR anomalies (Fig. 7). The lagged regression charts in the figure are presented in a northwest southeast orientation (indicated by the line Q P in Fig. 6a) and a north south orientation. The regressed OLR anomalies correspond to one standard deviation of the ISO index. Positive and negative OLR anomalies are found to occur alternately to the south and north of 25 N over time, with distinct opposite-phase variations in OLR anomalies between the SCS Philippine Sea and the Yangtze Basin. The disturbances originating from the equatorial region propagate northwestward (Fig. 7a) and northward (Fig. 7b) as far as around 25 N rather than moving continuously into the Yangtze Basin. The physical mechanism responsible for the northward and northwestward propagation of the 20 50-day oscillation can be understood with reference to the findings of Hsu and Weng (2001) and Tsou et al. (2005). Hsu and Weng (2001) suggested that moisture convergence at the northwestern corner of the convection causes a potentially unstable atmosphere, favoring the northwestward propagation of the coupled circulation convection system in the western North Pacific. In turn, Tsou et al. (2005) showed that the combined effect of surface frictional-diabatic heating and vorticity advection result in the development of 30 60-day convection and circulation, and its subsequent northwestward propagation in the western North Pacific and SCS. Both Hsu and Weng (2001) and Kemball-Cook and Wang (2001) emphasized the role of air sea interactions in fostering northwestward propagation of intraseasonal convective anomalies in the western North Pacific. Although in the present study we do not show the behavior of air sea interactions related to the 20 50-day oscillation in terms of surface downward shortwave radiative flux, sea surface temperature, surface latent heat flux, or low-level moisture, it is expected that similar interactive processes between the ocean surface and overlying air (i.e., similar to those described by Kemball-Cook and Wang 2001) occur to generate the instability necessary for convection. A positive surface downward shortwave radiative flux anomaly exists in the subsidence region to the north or northwest of the convection, causing a positive skin temperature anomaly that results in turn in low pressure at the surface and low-level moisture convergence. Thus, the atmosphere becomes potentially unstable and convection moves further northward and northwestward. 5 Intraseasonal variations in the WNPSH The most prominent feature in Fig. 6 is that the anomalous cyclone (anticyclone) accompanied by enhanced (suppressed) convection migrates northward and northwestward from the equator to the northern SCS, generating opposite temporal variations in the convective anomalies located in the Yangtze Basin and the northern SCS Philippine Sea. This seesaw phenomenon reflects intraseasonal variations in the WNPSH. With the aim of revealing the intraseasonal modulation of the subtropical high by the 20 50-day oscillation in the western North Pacific and SCS, Fig. 8 shows the composite filtered and unfiltered 850-hPa winds for the two extreme phases. As suggested by Mao and Wu (2006), the ridgeline is superimposed on the wind field to highlight the position and main body of the WNPSH, as Yangtze rainfall is sensitive to the position of the WNPSH ridgeline. During the driest phase of Yangtze rainfall, an anomalous cyclone is found to arrive at the northern SCS in the form of a convectively coupled Rossby wave-like system that propagates northward and northwestward (Fig. 8a), leading to a reduction in local pressure because the anomalous cyclone is concomitant with the low-pressure anomaly. Thus, the SCS monsoon trough deepens, while the WNPSH retreats a large distance northeastward, as clearly identified from the unfiltered wind field. The westernmost point of the WNPSH ridgeline is located east of 130 E; thus, strong monsoonal southwesterlies run mainly toward the SCS, converging with southeasterlies from the southern flank of the ridgeline in the SCS Philippine Sea. However, eastern China is dominated by an anticyclonic circulation with a ridgeline located around 30 N, resulting in dry conditions in the Yangtze Basin, with a strongly divergent flow pattern. Fig. 7 Lagged regression of OLR anomalies (W m -2 ) onto the 20 50 day ISO index of Yangtze rainfall from day -25 to day 25 with a lag of 1 day along a line Q P (as shown in Fig. 6a) and b 115 E. Shading indicates OLR anomalies significantly different from zero at the 95% confidence level

758 J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations Fig. 8 Composite unfiltered (streamlines) and filtered (arrows) 850-hPa winds (m s -1 ) for the a driest and b wettest ISO phases of Yangtze rainfall. Solid lines represent the ridgeline of the western North Pacific subtropical high (a) (b) In contrast, during the wettest phase of Yangtze rainfall (Fig. 8b) the subtropical anticyclone extends southwestward to the east of the SCS, due to the arrival of an anomalous anticyclone; the westernmost point of the WNPSH ridgeline reaches around 120 E. As described above, this anomalous anticyclone takes the form of a convectively coupled Rossby wave-like system emitted by suppressed equatorial convection. Because the subtropical anticyclone covers a much larger area east of the Philippine Sea, the tropical monsoon southwesterlies veer to the northeast over the SCS and then converge toward the Yangtze Basin, resulting in wet conditions with the support of moisture influx from the SCS. Moreover, the southwestward-extending WNPSH tends to produce a larger pressure gradient to its northwest (Lu 2001), resulting in heavy precipitation within the Yangtze Basin. During the decaying phases of the inactive ISO period (phases 1 and 2), the WNPSH extends a large distance to the southwest because of the rapid growth and northwestward propagation of the equatorial anomalous anticyclone (see Fig. 8a). As a result, the westernmost point of the 850- hpa ridgeline reaches west of 125 E and south of 20 N during phase 2 (not shown). Consequently, from phases 2 to 5 the WNPSH ridge continually intrudes westward and migrates northward. The opposite situation occurs from the decaying phase of the active ISO period to the developing phase of the inactive ISO period (phases 5 through 8 and back to 1) due to the development and northwestward propagation of the equatorial anomalous cyclone, with the WNPSH ridge located mainly north of 20 N. The above results suggest that the 20 50-day oscillation of Yangtze rainfall occurs in response to intraseasonal variations in the WNPSH, which in turn is modulated by an anomalous cyclone (anticyclone) that propagates northward and northwestward from the equatorial western Pacific in the form of a convectively coupled Rossby wavelike system. Such intraseasonal modulations of the WNPSH over a 20 50-day cycle are also found at other isobaric surfaces in the lower and middle troposphere: we obtained similar results when examining the wind and geopotential height fields at 700 and 500 hpa. It is interesting to note that an abrupt westward extension of the WNPSH, with large amplitude, appear from phases 1 to 2, and an eastward retreat from phases 5 to 6. However, the wettest phase (phase 5) in the Yangtze Basin coincides with the time when the lower-tropospheric

J. Mao et al.: 20 50-day oscillation of summer Yangtze rainfall in response to intraseasonal variations 759 WNPSH reaches its most westerly position (Fig. 8). Note that considerable zonal displacement of the WNPSH occurs during phase 2, immediately before the beginning of the active period (phase 3). This finding suggests that the change in the subtropical high occurs one phase (1/8-period of a 20 50-day oscillation) earlier than the initial active ISO period of Yangtze rainfall. Therefore, the considerable westward extension of the WNPSH leads to an extreme, positive Yangtze rainfall anomaly by three phases (3/8- period of a 20 50-day oscillation). Likewise, an abrupt eastward retreat of the WNPSH occurs during phase 6, with a relative phase lag of 3/8-period of the 20 50-day oscillation with regard to the extreme negative rainfall anomaly (phase 1). This phase lag further demonstrates the close relationship between Yangtze rainfall and the location of the WNPSH at an intraseasonal timescale, which may have implications for extended-range weather forecasts in the Yangtze Basin and which may assist in improving the predictability of intraseasonal rainfall events. 6 Summary and discussion As an extension of Mao and Wu s (2006) study, we examined the dominant ISOs controlling Yangtze summer rainfall from the climate-mean perspective, based on raingauge precipitation, OLR, and NCEP/NCAR reanalysis data for the period 1979 2003. EOF analysis was performed to identify spatio-temporal variability in summer precipitation within eastern China. Because the second leading mode (EOF2) captures the spatial coherence of variability in Yangtze rainfall (of great interest in the present study is intraseasonal behaviors in this area), the corresponding principal component (PC2) for each summer was used as the reference time series to isolate the dominant intraseasonal signal of Yangtze rainfall. The wavelet transform was then applied to the reference time series. Because the different dominant periods are found in different summers, the ensemble mean spectrum was calculated as the average of the statistically significant wavelet power spectra of 25 individual summers, to obtain the characteristic ISO periods for most of the analyzed years. The ensemble mean spectrum shows a predominant oscillation of Yangtze summer rainfall with a characteristic period of 20 50 days, suggesting that the 20 50-day oscillation is a major agent in regulating summer Yangtze rainfall. The spectral values at periods of 10 20 days from middle June to late July are also pronounced, implying that 10 20-days may be another characteristic period during these months. Composite analyses reveal that the 20 50-day oscillation of Yangtze rainfall shows a distinct seesaw in convective anomalies between the Yangtze Basin and the northern SCS Philippine Sea, with anomalous convection in the SCS Philippine Sea resulting from a wave-like coupled convection circulation system that propagates northward and northwestward from the equatorial region. Thus, the 20 50-day oscillation of Yangtze rainfall can be ascribed to a convectively coupled Rossby wave-like system emitted by suppressed and active equatorial convection, with its origin in the equatorial western Pacific. When the Rossby wave-like system propagates toward the northern SCS, an anomalous cyclone (anticyclone) appears alternately over the SCS Philippine Sea, thereby modulating the northeastward (southwestward) shift of the WNPSH. During the dry episode in Yangtze rainfall, an anomalous cyclone arrives at the SCS Philippine Sea, leading to a reduction in the actual local pressure because the anomalous cyclone is concomitant with the low-pressure anomaly. Consequently, the WNPSH retreats a large distance northeastward, with strong monsoonal southwesterlies running mainly toward the SCS Philippine Sea. However, an anticyclonic circulation, with a ridgeline around 30 N, dominates eastern China, resulting in dry conditions with a strongly divergent flow pattern in the Yangtze Basin. In contrast, during the wet episode of an ISO in Yangtze rainfall, the subtropical anticyclone extends southwestward to the east of the SCS, due to the arrival of an anomalous anticyclone. Thus, the southwesterlies of the tropical monsoon veer to the northeast over the SCS and then converge toward the Yangtze Basin, resulting in wet conditions with the support of moisture influx from the SCS and heavy precipitation in the Yangtze Basin. The 20 50-day oscillation of Yangtze rainfall arises in response to intraseasonal variations in the WNPSH, which in turn is modulated by an anomalous cyclone (anticyclone) that propagates northward and northwestward from the equatorial western Pacific. A considerable zonal displacement of the WNPSH is associated with the extreme dry (wet) period in the Yangtze Basin, with the abrupt eastward (westward) shift in the WNPSH generally leading the extreme negative (positive) Yangtze rainfall anomaly by a 3/8-period of the 20 50-day oscillation. This finding indicates that the phase lag may help in improving extended-range weather forecasting in the Yangtze Basin. It should be mentioned that the focus of this study is modulation of the WNPSH by the tropical ISO that occurs in the western North Pacific and SCS. In fact, this tropical ISO may also be influenced by intraseasonal disturbances originating from mid-latitude regions. As demonstrated by Lu et al. (2007), the westward-propagating wave train in the upper troposphere over the mid-latitude North Pacific may affect ISOs of convection and lower-tropospheric circulation in the subtropical and tropical western North