Real-time multivariate indices for the boreal summer intraseasonal oscillation over the Asian summer monsoon region
- 4.2k Downloads
The boreal summer intraseasonal oscillation (BSISO) of the Asian summer monsoon (ASM) is one of the most prominent sources of short-term climate variability in the global monsoon system. Compared with the related Madden-Julian Oscillation (MJO) it is more complex in nature, with prominent northward propagation and variability extending much further from the equator. In order to facilitate detection, monitoring and prediction of the BSISO we suggest two real-time indices: BSISO1 and BSISO2, based on multivariate empirical orthogonal function (MV-EOF) analysis of daily anomalies of outgoing longwave radiation (OLR) and zonal wind at 850 hPa (U850) in the region 10°S–40°N, 40°–160°E, for the extended boreal summer (May–October) season over the 30-year period 1981–2010. BSISO1 is defined by the first two principal components (PCs) of the MV-EOF analysis, which together represent the canonical northward propagating variability that often occurs in conjunction with the eastward MJO with quasi-oscillating periods of 30–60 days. BSISO2 is defined by the third and fourth PCs, which together mainly capture the northward/northwestward propagating variability with periods of 10–30 days during primarily the pre-monsoon and monsoon-onset season. The BSISO1 circulation cells are more Rossby wave like with a northwest to southeast slope, whereas the circulation associated with BSISO2 is more elongated and front-like with a southwest to northeast slope. BSISO2 is shown to modulate the timing of the onset of Indian and South China Sea monsoons. Together, the two BSISO indices are capable of describing a large fraction of the total intraseasonal variability in the ASM region, and better represent the northward and northwestward propagation than the real-time multivariate MJO (RMM) index of Wheeler and Hendon.
KeywordsBoreal summer intraseasonal oscillation Madden-Julian Oscillation Real-time multivariate index Northward propagation Asian summer monsoon Monsoon onset
It has been well recognized that the tropical intraseasonal oscillation (ISO) exhibits prominent seasonal variation (Madden 1986, Wang and Rui 1990; Salby and Hendon 1994; Zhang and Dong 2004; CLIVAR Madden-Julian Oscillation (MJO) working group 2009; Kikuchi et al. 2012). Compared to boreal winter, during boreal summer the main centers of convective variability associated with the ISO are shifted away from the equator to 10–20°N, and the propagation patterns are considerably more complicated. While the boreal winter ISO (also known as the MJO) shows predominantly eastward propagation, the boreal summer ISO (BSISO) also exhibits northward/northeastward propagation over the Indian summer monsoon (ISM) region (Yasunari 1979, 1980; Krishnamurti and Subramanian 1982; Lau and Chan 1986; Wang et al. 2005; Annamalai and Sperber 2005), and northward/northwestward propagation over the Western North Pacific-East Asian (WNP-EA) region (Murakami 1984, Chen and Chen 1993; Kemball-Cook and Wang 2001; Kajikawa and Yasunari 2005; Yun et al. 2009, 2010), often in conjunction with MJO-like propagation along the equator (Lawrence and Webster 2002). Whereas the MJO has been regarded as applicable in all seasons, albeit with generally weaker variability in boreal summer (Madden and Julian 1972, 1994; Wheeler and Hendon 2004; Zhang 2005), the BSISO has been regarded as a specific mode of the tropical ISO that prevails in boreal summer (Wang and Xie 1997). Thus, for many applications it is instructive to consider the tropical ISO as described by two modes, the MJO and BSISO. The MJO dominates during boreal winter (December–April) and the BSISO dominates during boreal summer (June–October) with May and November being transitional months during which either mode may prevail (Kikuchi et al. 2012).
Importantly, the BSISO is the dominant source of short-term climate variability in the Asian summer monsoon (Webster et al. 1998) and global monsoon (Wang and Ding 2008). It is known to affect summer monsoon onsets (Wang and Xie 1997; Kang et al. 1999), the active/break phases of the monsoon (Annamalai and Slingo 2001; Goswami 2005; Hoyos and Webster 2007; Ding and Wang 2009), and the monsoon seasonal mean (Krishnamurthy and Shukla 2007, 2008). It is also a possible source of seasonal climate predictability for precipitation (Wang et al. 2009a; Lee et al. 2010) and extratropical atmospheric circulation (Ding and Wang 2005; Lee et al. 2011; Wang et al. 2012). Two different periodicities of the BSISO have been identified: periods of 30–60 days (e.g., Wang et al. 2005) and 10–20 days (e.g., Kikuchi and Wang 2010). The wet and dry spells of the BSISO strongly influence extreme hydro-meteorological events, major driving forces of natural disasters, and thus the socio-economic activities in the World’s most populous monsoon region (Lau and Waliser 2005).
There have been several attempts to define indices for the BSISO mainly using Eigen techniques (Lau and Chan 1986; Waliser et al. 2004; Annamalai and Sperber 2005; Kikuchi et al. 2012 and others). Most recently, Kikuchi et al. (2012) reviewed existing approaches and proposed a bimodal ISO index that includes a BSISO mode with prominent northward propagation and large variability in off-equatorial monsoon trough regions, and a MJO mode with predominant eastward propagation along the equatorial zone. Their index successfully identifies the summer and winter component of the tropical ISO, but still has some limitations to capture the BSISO particularly over the Bay of Bengal and WNP-EA region (Fig. 6 in Kikuchi et al. 2012). In addition, as the Kikuchi et al. (2012) index was defined using 25- to 90-day filtered data, complications arise when trying to apply it in real time and to the output of global numerical forecast models. Ideally we would like an index or indices of the BSISO that involve no time filtering, and thus no smearing of information across different observation days or different forecast lead times.
Section 2 describes the method to define the BSISO indices proposed in this study. Basic characteristics of the BSISO captured by the indices are presented in Sect. 3. Section 4 describes the composite life cycle of the BSISO modes and fractional variance of OLR and U850 anomalies captured by the indices. How to apply the indices for real-time monitoring is discussed in Sect. 5. Summary and discussion are given in Sect. 6.
2 Definition of BSISO indices
The data used include the daily Advanced Very High Resolution Radiometer (AVHRR) OLR with 2.5° horizontal resolution from the National Oceanic and Atmospheric Administration (NOAA) polar orbiting satellites (Liebmann and Smith 1996) and daily horizontal wind at 850 and 200 hPa from NCEP/Department of Energy (DOE) Reanalysis II (Kanamitsu et al. 2002) with 2.5° horizontal resolution.
2.2 Process to define BSISO indices
Summary of results from different EOF analyses
Explained variance of EOF1 and EOF2 (%)
Fraction of leading PCs variance in 30- to 60-day band
Mean Coh2 of PC1 and PC2 in 30- to 60-day band
OLR′ and U850′
OLR′, U850′, and V850′
OLR′, U850′, and U200′
3 Basic Characteristics of the BSISO Indices
Based on our analysis, two BSISO indices are proposed in this study: BSISO1 comprising the 1st and 2nd MV-EOF modes, and BSISO2 comprising the 3rd and 4th modes. As will be shown, BSISO1 represents the canonical northward propagating BSISO over the ASM region addressed in many previous studies with a 30–60 day quasi-oscillating period (e.g., Annamalai and Sperber 2005; Wang et al. 2005; Kikuchi et al. 2012). BSISO2, on the other hand, mainly captures the northward and northwestward propagating BSISO with periods of both around 30 days and 10–20 days (e.g., Kikuchi and Wang 2010).
3.1 BSISO1: the canonical northward propagating BSISO component
Figure 2 shows the spatial structures and PCs of the first two MV-EOF modes. To display the full horizontal wind vector, the associated meridional wind field at 850 hPa (V850) was obtained by regressing V850 anomalies, normalized by their area-averaged temporal standard deviation (2.651 m s−1), against each PC. The spatial structure of the first mode (EOF1) displays mostly an east–west seesaw pattern in OLR, while EOF2 shows more of a quadrupole pattern characterized by a north to south dipole over the ISM region, and an oppositely-directed (south to north) dipole over the WNP-EA region. The U850 component of the EOF spatial structures is in approximate quadrature with the OLR component, with westerly anomalies occurring to the north and east of the positive OLR anomalies, and vice versa for the easterlies. Similar between EOF1 and EOF2 is the characteristic northwest to southeast slope of their patterns. As will be discussed further, this similarity is the first suggestion that EOF1 and EOF2 should be treated together as a pair to form what we call BSISO1.
A sample of the PCs associated with the leading two EOFs is provided in Fig. 2c. Consistent with the ordering of EOF modes, the variations of PC1 can be seen to be slightly larger than those of PC2. During the years shown it can be seen that PC1 and PC2 mostly vary on the intraseasonal timescale, with PC1 often leading PC2 by about a quarter cycle. This provides further evidence supporting our inclusion of the first two EOF modes as a pair in BSISO1.
3.2 BSISO2: The ASM pre-monsoon and onset component
In this study we also consider the third and fourth MV-EOF modes and argue that they capture the northward and northwestward propagating component of the BSISO, particularly during the pre-monsoon and monsoon onset period (Wang and Xie 1997; LinHo and Wang 2002). Different to the first two modes, the OLR and U850 patterns in EOF3 and EOF4 tend to be more in phase over the ISM and WNP region with a southwest-northeast tilted horizontal structure (Fig. 3). That is, the westerlies are shifted to correspond more closely with the regions of negative OLR and the easterlies with positive OLR. The structure of EOF4 resembles the ‘climatological ISO’ of Kang et al. (1999; their Fig. 4e) and the fast annual cycle pattern of the pre-monsoon and onset period of Linho and Wang (2002; their Fig. 4a). Consistent with this resemblance, Fig. 4 shows that PC3 and PC4 have maximum variance from late May to early July, corresponding to the pre-monsoon and onset period. We are thus confident that the third and fourth MV-EOF modes identified in this study are linked with a previously-described physical mode of the climate system.
The time variation of PC3 and PC4 is indicated for some sample years in Fig. 3c, indicating that these modes generally have a shorter time scale than PC1 and PC2 (Fig. 2c). Power spectra of PC3 and PC4 (Fig. 5c, d) confirm that the bulk of their variance is concentrated at around 30 days (PC3) and in the 10- to 20-day range (PC4). Further, PC3 and PC4 have high coherence in the 10- to 20-day range and at ~30 days, with approximately 90° phase difference (Fig. 6b). Analysis of the lag correlations between the two PCs shows that PC3 tends to lead PC4 by about 3–4 days for variability in the 10- to 20-day period range, and at a lag of 7–8 days for variability in the 20- to 50-day range (Fig. 7b). These relatively high coherence values, their similar horizontal tilts, and similar climatological seasonal cycles lead us to conclude that the third and fourth MV-EOF modes should be treated together as a pair, which we call BSISO2.
Interestingly, BSISO2 is not well correlated with the eastward-propagating MJO as measured by RMM1 and RMM2. While PC4 has no significant correlation with either RMM1 or RMM2, PC3 is correlated with RMM1 having a maximum correlation of 0.47 when PC3 leads RMM1 by about 3 days, but not significantly correlated with RMM2. Thus the linkage between BSISO2 and the eastward MJO is very weak.
4 Composite Life Cycles and Fractional Variance
Further understanding of the structure and patterns of variability captured by BSISO1 (EOFs 1 and 2) and BSISO2 (EOFs 3 and 4) can be achieved by constructing composites in a similar fashion to what was done for the MJO by Wheeler and Hendon (2004). Given the strong lead-lag behavior of PC1 and PC2, it is convenient to diagnose the state of BSISO1 as a point in the two-dimensional phase space defined by PC1 and PC2. Using the same argument, we diagnose BSISO2 as a point in the two-dimensional phase space defined by PC3 and PC4. In this analysis we use the normalized versions of the PCs, calculated by dividing by their 1981–2010 May–October standard deviations, to construct each phase space.
4.1 BSISO1 composite life cycle
It is interesting to note from Fig. 8a that Phase 5 (271 cases) and 7 (304 cases) are the most favorable for the strong initial events while Phase 3 (184 cases), 4 (160 cases), and 6 (189 cases) are the least favorable. This may indicate that the northwest-southeast tilted rainband from the Indian subcontinent to the equatorial Western Pacific tends to strengthen from Phase 3–5 then weaken. The amplitude then increases in Phase 7, associated with the cyclonic Rossby gyre over the WNP-EA region, which then propagates northwestward with weakened amplitude. It may also be associated with a slowdown of ISO propagation in Phases 5 and 7. The phase preference for the strong initial events needs further investigation.
4.2 BSISO2 composite life cycle
Figure 8b depicts the PC3 and PC4 phase space composite curves of the BSISO2 index. The life cycle of BSISO2 is shorter than BSISO1, with all curves reaching the centre of the phase space within 30 days. The average interval time between phases is 3.2 days. In general, the BSISO1 and BSISO2 tend to decay quicker than the MJO (see Fig. 7 of Wheeler and Hendon 2004), implying that the BSISO may have lower predictability than the MJO.
4.3 Fractional Variance
5 Application to real-time monitoring
Examples shown in this section demonstrate that the BSISO1 and BSISO2 indices are capable of identifying and monitoring prominent BSISO events over the entire ASM region.
6 Summary and Discussion
Given the extreme importance of the BSISO, we have made an effort to define new indices to assist in real-time monitoring and forecast applications of the BSISO. The BSISO indices proposed in this study were designed to better represent fractional variance and the observed northward/northwestward propagating ISO over the ASM region than the RMM index. Albeit its excellence in measuring the equatorial eastward propagating MJO, the RMM index is limited in its ability to capture ISO activity during boreal summer when it is furthest from the equator.
After considerable sensitivity tests, our chosen method to define the new BSISO indices uses the MV-EOF analysis of daily mean OLR and U850 anomalies over the ASM region (10°S–40°N, 40°–160°E) from May to October for the 30 years 1981–2010. The OLR and U850 anomalies are obtained from removing the slow annual cycle as well as the effect of interannual variation through subtracting the running mean of the last 120 days. We do not apply any other time filtering. We identify the first four MV-EOF modes as important for representing the BSISO over the ASM region that account for 19.4 % of total daily variance of the combined OLR and U850 anomalies. Based on our analysis, two BSISO indices are proposed: BSISO1 comprises the first two MV-EOF modes, and BSISO2 consists of the 3rd and 4th modes.
BSISO1 represents the canonical northward and northeastward propagating ISO over the ASM region during the entire warm season from May to October with quasi-oscillating periods of 30–60 days in conjunction with the eastward propagating MJO (e.g., Annamalai and Sperber 2005; Wang et al. 2005; Kikuchi et al. 2012). EOF1 and EOF2 can be treated together as a pair to form what we call BSISO1 because of the following three reasons. First, EOF1 and EOF2 exhibit similarity in their spatial structure characterized by a northwest to southeast slope. Second, the mean seasonal cycle of variance of PC1 is similar to that of PC2 with strong variance throughout the May to October period. Third, PC1 and PC2 have significant coherence in the 30- to 60-day range with a 90° phase difference indicating that PC1 leads PC2 by a quarter cycle. Analysis of lag correlation coefficients between the PCs further reveals that PC1 tends to lead PC2 by about 13 days with a maximum correlation of 0.34 for non-filtered data, and 0.45 for 30- to 60-day filtered data.
BSISO 2, on the other hand, mainly captures the northward and northwestward propagation components of the BSISO, particularly during pre-monsoon and monsoon onset period. Different to the first two modes, the OLR and U850 patterns in EOF3 and EOF4 tend to be more in phase over the ISM and WNP region with a southwest-northeast tilted horizontal structure. The structure of EOF4 resembles the ‘climatological ISO’ of Kang et al. (1999) and the fast annual cycle pattern of the pre-monsoon and onset period of LinHo and Wang (2002). Consistently, PC3 and PC4 have maximum variance from late May to early July and the bulk of their variance is concentrated at around 30 days (PC3) and in the 10- to 20-day range (PC4). With high coherence in the 10- to 20-day range and at ~30 days, PC3 tends to lead PC4 by about 3–4 days for variability in the 10- to 20-day period range, and at a lag of 7–8 days for variability in the 20- to 50-day range. These relatively high coherence values, their similar horizontal tilts, and similar climatological seasonal cycles lead us to conclude that third and fourth MV-EOF modes should be treated together as a pair, which we call BSISO2. BSISO2 has a close relationship with the onset of the ASM. Considering only monsoon onsets that occur when the BSISO2 amplitude is outside the unit circle, 68 % (70 %) of the onset dates for the Indian (South China Sea) monsoon occur in Phases 2–4.
The composite life cycles of BSISO1 and BSISO2 well demonstrate the circulation patterns associated with each index. The BSISO1 circulation cells are more Rossby wave like, whereas the circulation associated with BSISO2 is more elongated and front-like. We further demonstrate that the BSISO indices proposed in this study can be applied to real-time monitoring of the BSISO similar to the real-time monitoring of the eastward propagating MJO with the RMM index.
Although the BSISO indices are derived in the ASM domain, it is also indicative of the ISO in the North American monsoon region. This is because when intraseasonal convective anomalies occur in the equatorial western Pacific, they move not only northwestward toward the Philippine Sea but also eastward along the Intertropical Convergence Zone toward the North American monsoon region. As shown by Wang et al. (2005) and Moon et al. (2012), the ISO over the Mexican monsoon tends to be out of phase with the ISO over the southern Bay of Bengal. In terms of BSISO1, the peak in intraseasonal convective activity over the Mexican monsoon area occurs in Phase 1. This makes it also somewhat possible to monitor the North American monsoon by using BSISO1.
We thank the organization and members of the WWRP/THORPEX/WCRP YOTC MJO Task Force for their insights. We also thank two anonymous reviewers for their valuable comments that helped to improve our manuscript. JYL, BW, XF, and DW acknowledge support from the NOAA/MAPP project Award number NA10OAR4310247, AMDT1. BW and XF acknowledge support from Climate Dynamics Program of the National Science Foundation under award No AGS-1005599. This study has been also supported by APEC Climate Center and International Pacific Research Center, which is in part supported by JAMSTEC, NOAA (NA09OAR4320075) and NASA (NNX07AG53G). DW’s contribution was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. ISK were supported by the National Research Foundation of Korea (NRF) Grand Funded by the Korean Government (MEST) (NRF-2009-C1AAA001-2009-0093042). This is the SEOST publication number 8744 and IPRC publication number 912.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
- Lau WKM, Waliser DE (eds) (2005) Intraseasonal variability of the atmosphere-ocean climate system. Springer, HeidelbergGoogle Scholar
- Liebmann B, Smith CA (1996) Description of a complete (interpolated) outgoing longwave radiation dataset. Bull Am Meteorol Soc 77:1275–1277Google Scholar
- LinHo, Wang B (2002) The time-space structure of the Asian-Pacific summer monsoon: a fast annual cycle view. J Clim 15:2001–2019Google Scholar
- Moon JY, Wang B, Ha KJ, Lee JY (2012) Teleconnections associated with Northern Hemisphere summer monsoon intraseasonal oscillation. Clim Dyn. doi: 10.1007/s00382-012-1394-0
- Murakami M (1984) Analysis of deep convective activity over the western Pacific and southeast Asia. Part II: seasonal and intraseasonal variations during northern summer. J Meteorol Soc Jpn 62:88–108Google Scholar
- Wang B, Ding Q (2008) The global monsoon: major modes of annual variation in tropical precipitation and circulation. Dyn Atmos Oceans 44:165–183Google Scholar
- Wang H, Wang B, Huang F, Ding Q, Lee JY (2012) Interdecadal change of the boreal summer circumglobal teleconnection (1958–2010). Geophys Res Lett 39:L12704. doi: 10.1029/2012GL052371
- Yasunari T (1979) Cloudiness fluctuations associated with the northern hemisphere summer monsoon. J Meteorol Soc Jpn 57:227–242Google Scholar
- Yasunari T (1980) A quasi-stationary appearance of 30 to 40 day period in the cloudiness fluctuation during the summer monsoon over India. J Meteorol Soc Jpn 58:225–229Google Scholar