AMERICAN METEOROLOGICAL SOCIETY Journal of Climate EARLY ONLINE RELEASE This is a preliminary PDF of the author-produced manuscript that has been peer-reviewed and accepted for publication Since it is being posted so soon after acceptance, it has not yet been copyedited, formatted, or processed by AMS Publications This preliminary version of the manuscript may be downloaded, distributed, and cited, but please be aware that there will be visual differences and possibly some content differences between this version and the final published version The DOI for this manuscript is doi: 10.1175/JCLI-D-14-00373.1 The final published version of this manuscript will replace the preliminary version at the above DOI once it is available If you would like to cite this EOR in a separate work, please use the following full citation: Nguyen-Le, D., J Matsumoto, and T Ngo-Duc, 2015: Onset of the rainy seasons in the eastern Indochina Peninsula J Climate doi:10.1175/JCLI-D-14-00373.1, in press © 2015 American Meteorological Society Manuscript (non-LaTeX) Click here to download Manuscript (non-LaTeX): Manuscript.docx Onset of the rainy seasons in the eastern Indochina Peninsula Dzung Nguyen–Le*1, Jun Matsumoto1, 2, Thanh Ngo–Duc3 Department of Geography, Tokyo Metropolitan University, Hachioji, Tokyo, Japan Department of Coupled Ocean-Land-Atmosphere Processes Research, JAMSTEC, Yokosuka, Kanagawa, Japan Department of Meteorology and Climate Change, Hanoi College of Science, Vietnam National University, Hanoi, Vietnam 10 Revised version for submission to: 11 Journal of Climate 12 April 2015 13 14 15 16 17 18 *Correspondence to: 19 Dzung Nguyen-Le 20 Department of Geography, Tokyo Metropolitan University 21 1-1 Minami-Osawa, Hachioji-shi, Tokyo 192-0397, Japan 22 Phone: +81-809-179-1002 23 Fax: +81-42-677-2596 24 E-mail: nguyen-ledung@ed.tmu.ac.jp 25 Abstract 26 The onset dates of rainy season over the eastern Indochina Peninsula (8.5°–23.5°N, 27 100°–110°E) are objectively determined for individual years from 1958–2007 using the 28 empirical orthogonal function (EOF) analysis On average, the onset of the summer rainy 29 season (SRS) determined by EOF1 is May 6, with a standard deviation of 13 days The 30 autumn rainy season (ARS) indicated by EOF2 has a mean onset and standard deviation of 31 September 16 and 12 days, respectively The SRS onset is characterized by evolution of 32 summer monsoon westerlies and the northward propagation of strong convection from the 33 equatorial region Conversely, the withdrawal of the summer monsoon over northeastern 34 Indochina in late summer–early autumn favors the ARS onset Both onsets are strongly 35 associated with intraseasonal oscillation on 30–60 and 10–20 days scales 36 Examination of the precursory signals associated with the early/late onsets of both SRS 37 and ARS implies that ENSO has a significant impact on their year-to-year variations In La 38 Niña years, the subsequent SRS tend to have early onsets Simultaneously, the western North 39 Pacific subtropical high (WPSH) weakens and retreats eastward earlier In contrast, advanced 40 ARS onset generally occurs during an El Niño developing autumn with weakened equatorial 41 easterlies and suppressed convection over the central Indian Ocean from the preceding 42 summer, as evident in weakened Walker circulation However, robust precursory signals in 43 SST are observed only from mid–summer (July–August) An earlier ARS onset is also 44 associated with the development of an anomalous Philippine Sea anticyclone and a westward- 45 extended WPSH from mid-summer However, no coherent correlation is found between the 46 late onset and La Niña 47 48 Keywords: rainy season onset; interannual variability; summer monsoon withdrawal; ENSO; 49 intraseasonal oscillation; sub-tropical high; eastern Indochina Peninsula 50 Introduction 51 The Indochina Peninsula (ICP) is considered to be part of the Asian summer monsoon 52 region that experiences a wet monsoon in summer and a dry season in winter (Figs 1b and 53 1d) The onset of the summer rainy season (SRS) in the ICP, which climatologically occurs in 54 early May, is characterized by rapid northeastward expansion of low-level southwesterlies 55 from the equatorial Indian Ocean (IO), accompanied by a significant increase in convection 56 (e.g., Matsumoto 1997; Wang and LinHo 2002; Zhang et al 2002; Nguyen-Le et al 2014) 57 The annual reversal of the surface monsoonal winds and topographic effect are the main 58 causes of rainfall seasonality over Southeast Asia (Chang et al 2005) Over the east coast of 59 the ICP between approximately 12°–19°N (ECI), where the terrain is complex because of a 60 narrow lowland plain along the coast and the Truong Son mountain range of Vietnam with 61 elevations reaching higher than 1000 m (Fig 1a), rainfall reaches its maximum in autumn 62 (Cheang 1987; Chen et al 2012; Fig 1e) During the summer, however, a downslope Foehn 63 wind, an effect of the summer monsoon, causes a relatively dry summer in this region 64 (Nguyen-Le et al 2014) In September, the summer monsoon begins to retreat, and the winter 65 monsoon begins around October (Matsumoto 1997), bringing northeasterly surface winds that 66 cause orographic rainfall on the windward coastal plain of the ECI (Yokoi and Matsumoto 2008) 67 Various studies have been conducted over recent decades to understand the interannual 68 variability of the onset dates of the summer monsoon and the SRS, owing to its importance to 69 agricultural activity In particular, its close relationship with El Niño–Southern Oscillation 70 (ENSO) has been investigated (e.g., Ju and Slingo 1995; Lau and Yang 1997; Wu and Wang 71 2000; Zhang et al 2002; Zhou and Chan 2007) Lau and Yang (1997) showed that the delayed 72 (advanced) South China Sea (SCS) monsoon onset might be related to basin-wide warm 73 (cold) events of the Pacific and IO Zhou and Chan (2007) also found that in years associated 74 with a warm (cold) ENSO event or the year after such years, the SCS monsoon tends to have 75 late (early) onsets In the ICP, Zhang et al (2002) revealed that warm sea surface temperature 76 (SST) anomalies in the western Pacific and cold SST anomalies in the central–eastern Pacific 77 in the preceding winter–spring correspond to an early SRS onset However, previous studies 78 have not included reference to the eastern part of the ICP (8.5°–23.5°N, 100°–110°E) 79 Recently, an advanced SRS onset over the Bay of Bengal (BOB) and western Pacific 80 was evident between the periods 1994–2008 and 1979–1993 (Kajikawa et al 2012), and the 81 SCS summer monsoon was also found to exhibit significant decadal variability with an 82 advanced onset after 1993/94 (Kajikawa and Wang 2012) Xiang and Wang (2013) revealed 83 that this advanced onset was mainly represented by a robust decadal shift in the mid-to-late 84 1990s, which they attributed to a mean state change in the Pacific basin characterized by a 85 grand La Niña-like pattern In contrast, onset of the autumn rainy season (ARS) has received 86 much less attention to date, owing to its localized nature Thus, this study aims to reveal the 87 onset of both the SRS and ARS over the eastern ICP on annual and inter-annual timescales 88 Specifically, the onset dates in individual years during 1958–2007 are objectively determined 89 Reanalysis and satellite data are utilized to compose the temporal and spatial structures of 90 atmospheric circulation and convection during the onsets Then, composite analyses are 91 conducted between the early/late SRS (ARS) onset categories to investigate the precursory 92 signals in the preceding winter and spring (summer) The remainder of this paper is organized 93 as follows Section describes the data Section introduces the definition of the rainy season 94 onset Section presents climatological features of large-scale circulation and the 95 intraseasonal oscillation (ISO) activities related to the onsets The precursory signals 96 associated with the interannual variations of these onsets are examined, and their underlying 97 processes are discussed in section Finally, conclusions are given in section 98 99 Datasets 100 The rainfall data used in this study were daily-mean precipitation from the 101 APHRODITE dataset provided by the Research Institute for Humanity and Nature (RIHN) 102 and the Meteorological Research Institute of Japan Meteorological Agency (MRI/JMA; 103 Yatagai et al 2012) This dataset was created primarily from rain gauge data recorded from 104 1951–2007, and covers the entire monsoon Asia region (60°E–150°E, 15°S–55°N) on 0.25°× 105 0.25° grids However, because of concerns regarding the quality of observation during the 106 First Indochina War, only data from1958–2007 (50 years) were analyzed 107 The National Center for Environmental Prediction/National Center for Atmospheric 108 Research (NCEP/NCAR) reanalysis data (Kalnay et al 1996) from the same period on 2.5° × 109 2.5° grids were used to analyze atmospheric circulation associated with the rainy season 110 onset Additionally, daily-mean outgoing longwave radiation (OLR) from 1979–2007 on 1.0° 111 × 1.0° grids were obtained from the National Oceanic and Atmospheric Administration 112 (NOAA) Outgoing Longwave Radiation–Daily Climate Data Record (NOAA 2014a) Finally, 113 monthly-mean SST and mean sea level pressure (SLP) on 1.0° × 1.0° and 5.0° × 5.0° grids for 114 the period 1958–2007 were obtained from the Met Office Hadley Centre’s HadISST1 and 115 HadSLP2 datasets, respectively (Rayner et al 2003; Allan and Ansell 2006) 116 117 Determination of the rainy season onset date 118 In order to objectively determine the onset of the rainy season in the eastern ICP, we 119 proposed a method based on principal components of the first two leading empirical orthogonal 120 function modes (EOF1, EOF2) of standardized rainfall data (Fig 2) Prior to standardization, 121 actual daily rainfall is firstly normalized by cubic-root transformation, making its frequency 122 distribution closer to the normal distribution than that of the original data (Stidd 1953) 123 The eigenvector pattern of EOF1, accounting for 48.9% of total variance, is definitely 124 positive over the entire studied region with only low loadings along the ECI (Fig 2a) Its 125 principal component (PC1) is observed to generally switch signs from negative (positive) to 126 positive (negative) around April (October; Fig 2c), illustrating that EOF1 is associated with 127 wetter conditions during summer–autumn and drier conditions during winter–spring 128 Although this positive pattern may seem incongruous, it is due to the substantially higher 129 amount of rainfall in the ECI during the summer than during the distinct dry season occurring in 130 February–March (Fig 3b) Additionally, the ECI occasionally receives tropical cyclone-based 131 in from mid-summer (July; Nguyen-Thi et al 2012) Therefore, positive signals in EOF1 over 132 this area can be understood because there is a relative contrast between summer and winter 133 Meanwhile, EOF2 contributes 10.8% of the total variance, and has the strongest signal 134 in the ECI (Fig.2b), showing that rainfall in the area experiences a maximum in autumn, 135 corresponding to the annual distribution of PC2 (Fig 2d) Since its rainfall peak generally 136 appears in late September–October (Wang and LinHo 2002), the southern part of the 137 peninsula also has positive loadings of EOF2 On the other hand, EOF2 is negative in the 138 northern section, implying that the beginning of the drier period there occurs simultaneously 139 with the outbreak of rainfall in the ECI However, it could therefore be considered that EOF2 140 only reflects the withdrawal of the SRS in the northeastern ICP It is of note that Fig 2d 141 suggests that the annual cycle of PC2 is very similar to the seasonal distribution of rainfall in 142 the ECI, which reaches its maximum intensity in autumn and has double peaks occurring 143 around September–November and May–June (Wang and LinHo 2002) Meanwhile, the 144 northern ICP exhibits a typical summer monsoon rainy season with a continuous rainfall 145 increase from late spring (March; Fig 3b) It is thus considered that EOF2 represents this 146 atypical bimodal annual cycle of rainfall, particularly along the ECI 147 Consequently, results from EOF analysis capture most of the spatial and temporal 148 characteristics of rainfall in the eastern ICP reasonably well, with EOF1 (EOF2) representing 149 the SRS (ARS) over most locations (along the ECI) The changes in sign from negative to 150 positive of PC1 and PC2 correspond closely to the advent of these rainy seasons Therefore, 151 we defined the onset for each individual year during 1958–2007 by analyzing the PC using a 152 modification of the definition given by Zhang et al (2002) Specifically, the onset timing of 153 the SRS (ARS) is defined as the day on which PC1 (PC2) satisfies the following conditions: 154 the PC begins to be positive and persists continuously for seven days; within 20 consecutive 155 days, the number of days with positive PC exceeds 14 days These thresholds are determined 156 after conducting many sensitivity tests to reflect the apparent seasonal changes in rainfall 157 from dry to wet conditions As a result, the mean onset of the SRS was determined to occur 158 around May with a standard deviation of 13 days (Table 1) This is consistent with findings 159 from previous studies such as Matsumoto (1997), Wang and LinHo (2002), and Zhang and 160 Wang (2008), who showed that the climatological SRS onset over the ICP occurs during the 161 first weeks of May Figure 3b suggests that rainfall along the east coast of the northern and 162 southern ICP begins to exceed mm day-1 after the onset Analyzing the rainfall index 163 represented by the daily area-averaged rainfall from 30 stations in central ICP, Zhang et al 164 (2002) showed that the monsoon onset over Indochina during 1951–1996 was on May on 165 average, with a standard deviation of 12 days This result is significantly correlated with ours 166 (the non-parametric Spearman rank correlation is 0.46), exceeding the 99% confidence level 167 In the case of ARS, the mean onset and standard deviation are September 16 and 12 168 days, respectively (Table 2) Although these onset dates are only defined following PC2, 169 rainfall along the ECI is observed to increase remarkably, whereas it decreases rapidly along 170 the northeastern coast after onset (Fig 3b) To the authors’ knowledge, this study is the first 171 to research the interannual variability of the ARS onset in Indochina However, Nguyen-Le 172 and Matsumoto (2015) used rain gauge data in the ECI recorded from 1979–2007 to show 173 that the ARS begins on average in the 51st pentad (September 8–12) with a standard deviation 174 of three pentads The discrepancy is related to the differences in the studied period, the use of 175 different datasets, and the onset criterion Nevertheless, a significant (99%) correlation (r = 176 0.58) is detected among the results As previously mentioned, this onset is simultaneous with 177 the withdrawal of the SRS in the northeastern ICP, and the climatological result is in good 178 agreement with Matsumoto (1997) and Zhang and Wang (2008) According to Matsumoto 179 (1997), the SRS retreats from the northeastern ICP in early to mid-September Zhang and 180 Wang (2008) also suggested that the SRS in the ICP begins equatorward withdrawal in mid- 181 September Additionally, the ARS onset time series are well (95%) correlated (r = 0.3) with 182 the Indian summer monsoon withdrawal dates given by Syroka and Toumi (2004) 183 Statistical differences in the actual rainfall amount over the entire studied region 184 between consecutive pentads around the onset were also examined (Figs and 5) Here, 185 “pentad” is defined as follows: the onset day is denoted as day 0, and the “−” and “+” signs 186 denote time prior to and after the onset day, respectively Thus, the successive pentad means 187 are computed as “onset 0” representing the average from day to +4 and “onset −1” denoting 188 the average from day ˗5 to day ˗1 Over most of the ICP region, there is a significant increase 189 in rainfall during the SRS onset (Fig 4f) While rainfall intensifies considerably along the 190 ECI and southeastern ICP at the ARS onset pentad, the opposite behavior was observed over 191 the northern section (Fig 5f) Note that because of the discrepancy between the actual and 192 standardized rainfall, the highest actual rainfall and its increase are detected between 193 15°N−18°N, primarily because of the significantly higher year-round rainfall, and the reduced 194 bimodality of its seasonal distribution there (Fig 3b–c), instead of between 11°N−15°N, 195 where the largest positive signals of EOF2 (Fig 2b) and the standardized rainfall and its 196 tendency (figure not shown) are found 197 In addition to the pentad-to-pentad changes, we also compared the 10-day rainfall 198 difference prior to and after the onset day (figure not shown) Similar results are obtained with 199 the most significant rainfall increase occurring on the onset day, suggesting that our definition 200 reasonably captures the rainy season onset over the eastern ICP 201 202 Climatological features of the rainy season onset over the eastern ICP 203 4.1 Onset of the SRS 204 To show the atmospheric circulation and convection related to the SRS onset, evolution 205 of composite OLR, streamlines at 850 hPa and the western Pacific subtropical high (WPSH) 206 denoted by the 500 hPa geopotential high exceeding 5860 gpm around the onset are 207 composed (Fig 6) The OLR threshold of 230 Wm−2 was used to indicate strong convection 208 In the preceding pentad -3 and -2, low OLR is only detected over the southern BOB, Sumatra, 209 and the equatorial western Pacific Nitta et al (1992) suggested that blocking by topography 210 causes convection to pile up around Sumatra In terms of low-level flow, the studied region is 211 located along the boundary between two wind systems: mid-latitude westerlies and the 212 southeasterly trade wind related to the WPSH (Matsumoto 1997; Zhang et al 2002) 213 Zhang et al (2004) reported that pre-onset strong latent (sensible) heating in the ICP 214 (Indian subcontinent) is favorable for triggering a cyclonic vortex or trough on the BOB, 215 which is represented as a cyclonic circulation over Sri Lanka, and helps the WPSH to split 216 over the BOB (Figs 6a–b) At pentad -1, accompanied by the cyclone circulation around 5°S, 217 the formation and deepening of the BOB trough make the in-between tropical westerlies 218 accelerate and rapidly expand northeastward from the equatorial IO to the eastern BOB and 219 the south-western coast of the ICP (Fig 6c) This implies that the summer monsoon onset 220 over the eastern BOB is generally earlier than over the eastern ICP, which is consistent with 221 the climatological onset date there of May (Mao and Wu 2007) Meanwhile, the mid- 222 latitude westerlies notably weaken and retreat northward Simultaneously, the WPSH and 223 associated southeasterly trade wind also retreat eastward At pentad 0, the tropical westerlies 224 converge with the southeasterlies over the eastern ICP, favoring a strong enhancement of 225 convection there Particularly, anomalous southerlies over the southern ICP, which are related 226 to the Rossby wave response to the extreme negative phase of the ISO on a 10–20-day variation 227 (10–20DV) over Sumatra (Wen and Zhang 2005, 2007), are beneficial to the northward 228 progression of tropical convection from the equatorial ocean Consequently, convection over 229 the ICP reaches its strongest phase, which appears as an abrupt northward propagation of strong 230 convection from low latitudes (Fig 6d) After onset, at pentad +1 and +2, the low-level 231 prevailing winds over the entire ICP are clearly identified as monsoon westerlies (Figs 6e-f) 777 778 Fig Latitude–time section of 5-day mean (b) actual precipitation amount (mm day-1) and 779 (c) standardized precipitation along the eastern Vietnam–China boundary and east coast of the 780 ICP (blue dots in the geographic map (a)) Red and purple lines indicate the mean SRS and 781 ARS onset date over the eastern ICP, respectively 36 782 783 Fig (a), (c), (e), (g), (i) Evolution and (b), (d), (f), (h), (j) differences between two 784 consecutive pentads of pentad-mean precipitation (mm day-1) over the eastern ICP (8.5°– 785 23.5°N, 100°–110°E) centered on SRS onset dates The contour interval is mm/day; shaded 786 areas in (b, d f, h, j) indicate a significant difference above the 95% confidence level 37 787 788 Fig Same as Fig but for the ARS onset date Contour interval is mm/day 38 789 790 Fig Composite evolutions of pentad mean OLR (grey shaded, Wm-2), geopotential height 791 at 500 hPa (red contour; gpm), and wind streamlines at 850 hPa (black contour) from three 792 pentads before SRS onset dates to two pentads after SRS onset dates Only OLR values lower 793 than 230 Wm-2 and geopotential heights exceeding 5860 gpm are plotted; the interval of the 794 purple contours is gpm The missing areas represent the below ground position of 850 hPa 39 795 796 Fig (a), (c) Longitude–time section along 10°–20°N; and (b), (d) latitude–time section 797 along 100°–110°E of 10–20DV and 30–60DV of OLR, respectively Day is the SRS onset 40 798 799 Fig Same as Fig but for the ARS onset 41 800 801 Fig Same as Fig but for the ARS onset 42 802 803 Fig 10 The bimonthly early-minus-late SRS onset composite of: (a)–(c) 850 hPa wind 804 vector, and (d)–(f) SST, in the preceding December–January, February–March, and April– 805 May, respectively In (a)–(c), the reference difference wind-vector arrow is m s-1, and in 806 (d)–(f) the contour interval is 0.2 °C Shaded areas indicate a significant difference above the 807 95% confidence level 43 808 809 Fig 11 (a)–(c) Composite bimonthly 500 hPa geopotential height indicated by 5865 gpm 810 (thin) and 5860 gpm (thick) contours for early (solid) and late (dashed) SRS onset years; and 811 (d)–(f) the bimonthly early-minus-late SRS onset composite of OLR in the preceding 812 December–January, February–March, and April–May, respectively In (d)–(f), the contour 813 interval is Wm-2; shaded areas indicate that a significant difference above the 95% 814 confidence level 44 815 816 Fig 12 Difference in the mean standard deviation of filtered OLR (Wm-2) on (a) 10–20- and 817 (b) 30–60-day time scales in March (early minus late SRS onset) The contours denote 818 climatological values during 1979–2007; hatched areas indicate that the difference is 819 significant above the 95% confidence level 45 820 821 Fig 13 The monthly early-minus-late SRS onset composite of: (a) TT, (b) vertical shear of 822 zonal wind, and (c) 850 hPa specific humidity in March Shaded areas indicate that the 823 difference is significant above the 95% confidence level 46 824 825 Fig 14 The bimonthly early-minus-late ARS onset composite of: (a)–(c) 850 hPa wind 826 vector; and (d)–(f) SST in the preceding May–June, July–August, and September–October, 827 respectively In (a)–(c), the reference difference wind-vector arrow is m s-1; in (d)–(f) the 828 contour interval is 0.2 °C Shaded areas indicate a significant difference above the 95% 829 confidence level 47 830 831 Fig 15 (a)–(c) The composite bimonthly 500 hPa geopotential height indicated by 5865 gpm 832 (thin), and 5860 gpm (thick) contours for early (solid) and late (dashed) SRS onset years; and 833 (d)–(f) bimonthly early-minus-late SRS onset composite of OLR in the preceding May–June, 834 July–August, and September–October, respectively In (d)–(f), the contour interval is Wm-2 835 Shaded areas indicate a significant difference above the 95% confidence level 48 836 837 Fig 16 Difference in the mean standard deviation of filtered OLR (Wm-2) on: (a) 10–20-; and 838 (b) 30–60-day time scales in August (early minus late ARS onset) Contours denote 839 climatological values during 1979–2007; hatched areas indicate a significant difference above 840 the 95% confidence level 49 841 842 Fig 17 Monthly early-minus-late ARS onset composite of: (a) TT, (b) vertical shear of zonal 843 wind; and (c) 850 hPa specific humidity in August Shaded areas indicate that a significant 844 difference above the 95% confidence level 50 ... variability of the rainy season onset in eastern Indochina, through its modification of the 517 timing of onset/ withdrawal of the summer monsoon, and the strength and shape of the WPSH 518 In general,... captures the rainy season onset over the eastern ICP 201 202 Climatological features of the rainy season onset over the eastern ICP 203 4.1 Onset of the SRS 204 To show the atmospheric circulation... area, indicating the 278 withdrawal of the wet season over the northeastern ICP Note that this switch in direction of the 279 prevailing wind leads to the formation and intensification of an