Meteorol Atmos Phys 95, 53–61 (2007) DOI 10.1007/s00703-006-0197-4 Printed in The Netherlands
Disaster Preventive Research Institute, Kyoto University, Gokasho, Uji, Kyoto, Japan
Active, weak and break spells in the Indian summer monsoon V. S. Prasad and T. Hayashi With 7 Figures Received June 22, 2005; revised October 10, 2005; accepted March 2, 2006 Published online: July 31, 2006 # Springer-Verlag 2006
Summary Active weak and break phases of the Indian summer monsoon for the period 1958–2002 are studied using ERA -40 data. The criteria for identifying the break are proposed and tested using the 850 hPa level horizontal wind shear. Independent datasets such as All-India Rainfall, NOAA Outgoing Long-wave Radiation and C M A P rainfall datasets are used for the verification of the proposed criteria.
1. Introduction The Indian summer monsoon (I S M ) rainfall oscillates in the form of ‘‘active’’ and ‘‘weak’’ spells. These active and weak spells of the monsoon are associated with the tropical convergence zone (TCZ ; Sikka and Gadgil, 1980). The inter-annual variability in the ISM is mainly due to the occurrence of long, dry spells in the latter (Krishnamurti and Bhalme, 1976). Apart from the above intraseasonal variations, which occur every year, a special situation traditionally known as ‘‘break monsoon’’ occurs (Gadgil and Joseph, 2003). These breaks are periods when the monsoon trough is located close to the foot of the Himalayas, which leads to a striking decrease in rainfall over most of India but increase along the Himalayas and parts of northeast India and the southern On leave from the National Center for Medium Range Weather Forecasting, A-50 Institutional Area, Phase-II, Sector 62, Noida, U.P., India.
peninsula (Rao, 1976). An exhaustive survey of the observed characteristics of the breaks in the IS M was conducted by Ramamurthy (1969) for the 1888–1967 period and by De et al (1998) for the 1968–1996 period. They identified break as a day in which the trough of low pressure was not seen on the surface chart over India and in which easterlies were practically absent (hereafter, R þ D break). Although interruption of rainfall is recognized as the most important feature of these ‘‘breaks’’, the criterion used by the Indian Meteorological Department (IM D ) for identifying break is the synoptic situation associated with such a rainfall anomaly and is defined clearly by Rao (1976). Recently, there are many studies of breaks identified on the basis of different criteria over regions differing in spatial scale. Webster et al (1998) use the term ‘‘break (active) spells’’ to denote weak (strong) spells of convection and 850 hPa zonal winds over a large-scale region (65–95 E, 10–20 N). On the other hand, Goswami and Mohan (2000) define breaks on the basis of the strength of the 850 hPa wind at a single grid point 15 N, 90 E. Anamalai and Slingo (2001) use the term ‘‘break’’ to denote weak spells of daily average rainfall. Krishnan et al (2000) defines the break days as days with positive Outgoing Long-wave Radiation (O L R ) anomalies exceeding 10 Wm2 over the region 73–82 E, 18–28 N.
54
V. S. Prasad and T. Hayashi
Gadgil and Joseph (2003) in their study defined the break using the rainfall distribution in the monsoon zone and the called it ‘‘rainbreak.’’ They identified breaks on the basis of average daily rainfall of the western (<2.5 mm=day) and eastern (<7.5 mm=day) parts of the monsoon zone. There is diversity in the duration of breaks, as well as their frequency of occurrence, identified in different studies because the definition of breaks is based on different criteria. In addition, there is difference in the definition of break in these studies and that of the traditional approach where breaks are defined as an event with specified characteristics such as surface pressure pattern, and these breaks do not occur every year. However, all other types of break, except rainbreak and R þ D break, occur three or four times a season. In several studies (e.g., Annamalai and Slingo, 2001; Goswami and Mohan, 2000), the active spells and breaks are identified as crests= troughs of a specific mode (e.g., 30–50 day) of an appropriate field. This suggests that the breaks in these studies are weak spells of intraseasonal variations in the monsoon, which occur every year. Sikka and Gadgil (1978), in a case study involving the 1972 and 1973 monsoon seasons, showed that day-to-day meridional difference between the zonal winds along 80 E between 28 and 20 N at 900 hPa, the top of the boundary layer, was related to All-India Rainfall. According to them, if the difference was negative (positive), it would be related to the above (below normal) rainfall on an All-India basis related to cyclonic (anti-cyclonic) relative vorticity at the top of boundary layer. Thus, it is possible to develop a diagnostic criterion to identify active= break situations using wind shear. Further, it is directly related to the pressure pattern and may thus be close to the traditional method. It also has economical significance, as it can be used as a forecasting tool using G C M outputs. Thus, in this study, diagnostic criteria are derived from the variability in the horizontal zonal wind shear at the 850 hPa level to retrospectively assess the active, weak and break phases of the I S M . This offers several advantages as this procedure depends on fields that are better observed and modeled than rainfall and is also found to be related to both anomaly of rainfall distribution and the synoptic situation.
2. Data The E C M W F Reanalysis (E R A ) datasets with a spatial resolution of 2.5 deg 2.5 deg for the period 1957–2002 is the primary dataset used in this study (Gibson et al, 1997). The precipitation datasets obtained from the All-India Rainfall (A IR ) and sub-divisional rainfall (Parathasarathy et al, 1994) series constructed from the a weighted average of land surface observations throughout India from 1871–1999 and the C P C Merged Analysis of Precipitation (CM AP ) (Xie and Arkin, 1997) 1979–2000 satellite and gauge merged precipitation analyses are used for verification purposes. The daily outgoing long-wave radiation (O L R ) data used for the period 1975–2001 (except 1978) is from the Advanced Very High Resolution Radiometer (AV H R R ) instrument (Libmann and Smith, 1996). The year 1978 is excluded due to unavailability of the O L R data for most of that year.
3. Circulation index Wang et al (2001) developed a dynamical index for the I S M based on Horizontal Wind Shear (HWSI ). It is defined as the difference in the 850-hPa zonal winds between a southern region of 5–15N, 40–80E (Zone 1) and a northern region of 20–30N, 70–90E (Zone 2). Similar circulation index was also used by Syroka and Toumi (2002) for the study of the withdrawal features of the ISM . Further, Syroka and Toumi (2004) found that the monthly mean H W S I and inter-annual variations in monthly mean A I R are well correlated. The regions are chosen to be dynamically consistent with the convective heating over the summer Indian continent (Wang and Fan, 1999). H W S I also captures both the variability in the position and the intensity of the monsoon trough through a first-order approximation of the relative vorticity and the strength of the low-level Somali jet into the I S M domain. Further, Prasad and Hayashi (2005) studied variations in H W S I and its relation to large-scale rainfall over the Indian region. They showed that the day-to-day variations in the HWSI display remarkable agreement with the O L R variations averaged over the Indian region and also correlate well with those in of the C M A P data for the period 1979–2001.
Active, weak and break spells in the Indian summer monsoon
55
Fig. 1. 1958–2001 composite annual cycle (solid line) along with 1 standard deviation (dashed lines) for each day (a) 850-hPa Zone-1 zonal wind, (b) same for Zone-2, and (c) H W S I
Further, Gadgil and Joseph (2003) in their rainbreak study clearly pointed out large anomalies in 850 hPa zonal wind composites precisely over Zone 1 and Zone 2 regions. Figure 1a–c depicts the mean annual cycle along with the standard deviation of HWSI and the mean zonal wind component for Zone 1 and 2, respectively. It can be clearly seen that Zone-1 zonal winds have a strong seasonal cycle with a peak in the I S M period and also have relatively less variation (Fig. 1a). At the same time, Zone-2 zonal winds have very a low value throughout the year with large variations during the ISM period (Fig. 1b). It is due to the fact that strong westerlies generally prevail over Zone 1 during the monsoon period and, at the same time, winds in Zone 2 oscillate between westerlies and easterlies depending on the position of the monsoon trough. During the active monsoon period, weak easterlies prevail over Zone 2, while during break periods, the monsoon trough shifts northwards and Zone 2 winds become westerly. Thus, this index depends on the position of the monsoon trough, the strength of the westerlies in the south and the strength and direction of the winds in the north. HWSI is therefore both physically sensible and a practical tool for studying monsoon intensity, and along with Zone 2 zonal winds, it can distinguish the break and active phases of the IS M .
As described above, the break situation is synoptically defined by the strengthening of southwesterlies in the north and shifts in the position of the monsoon trough. In another situation, H W S I weakens without winds in Zone 2 becoming strengthened in a westerly direction, and this situation is called the weak monsoon. In these weak monsoon conditions, the strength of zonal winds in Zone 1 is reduced without much change in winds in Zone 2. It is important to distinguish this situation from the break situation because in such a situation there is no increase in rainfall in the northeastern part of India and rainfall is reduced over the entire country. To identify the different phases of the ISM using these criteria, the year 1979 is selected as a typical example. This year is selected because it has the largest number of active and break days as classified by traditional methods and also because C M A P precipitation analysis is available from 1979 onwards. Figure 2 depicts the H W S I along with zonal winds for Zone 1 and 2 for the IS M period of the year 1979. A strong westerly wind persists in Zone 1 throughout the season with a dip in late August to early September, whereas there is a considerable amount of variation in the case of Zone 2. In the first week of July, zonal winds in Zone 2 become strong westerly winds and this may be due to the shift
56
V. S. Prasad and T. Hayashi
Fig. 2. Daily averages of 850-hPa zonal winds for Zone 1 and 2 along with H W S I for June– September 1979
of the monsoon zone towards the foothills of the Himalayas and thus can be treated as a break monsoon period as per the traditional I M D definition. A similar situation is repeated from 15–27 August. From 25 July onwards, the westerlies in Zone 2 are gradually replaced by easterly winds, signifying the beginning of the active monsoon period. From 27 August to 3 September, a slight increase Zone 2 winds and a decrease in Zone 1 winds is seen. This may be related to the weak monsoon period. This classification is then verified using C M A P pentad rainfall data. The occurrence of widespread rain in the pentad beginning 25 July is clearly observed. The typical break type of rainfall distribution with an increase in rainfall along the foothills and southern tip of peninsula and a decrease in the rest of country can be observed in the pentad rainfall beginning 19 August (Fig. 3). Then rainfall for the pentad beginning 24 August covers some break and weak days and thus a general decrease in the rainfall is seen throughout the country including the foothills. Thus, it is possible to identify three different phases of the I S M using zonal winds over Zone 1, Zone 2 and H W S I .
4. Definition of active, break and weak phases of the monsoon Fig. 3. Xie-Arkin rainfall for the pentad beginning (a) 25 July 1979, (b) 19 August 1979, and (c) 24 August 1979
The following approach adopted for identifying active, weak or break phase days is to isolate
Active, weak and break spells in the Indian summer monsoon
57
consecutive days in the I S M during the period 1958–2002.
5. OLR , circulation and rainfall patterns in the ISM phases
(i) In the first step, daily climatology and standard deviation of HWSI and that of zonal winds in Zone 1 and Zone 2 are constructed. (ii) In the next step, days with an H W S I value greater than the mean plus standard deviation are isolated as active days of the I S M season. (iii) Then, weak monsoon days are identified as days with a low H W S I index and low zonal winds in Zone 2. Similarly, break days are identified as days with a low HWSI index but relatively strong westerly winds in Zone 2. Here, the mean plus standard deviation and the mean minus standard deviation are used as objective criteria for identifying strong and weak phases, respectively.
The O L R value is a good proxy for inferring rainfall activity associated with tropical convection. Usually, long-term averaged O L R data are considered a good surrogate for the rainfall. Thus, to verify whether the criteria can identify the different phases of the ISM correctly, O L R composites based on active, weak and break days are prepared. Figure 4 shows these composites along with mean seasonal O L R for the 1975–2001 (excluding 1978) period. A local t-test was applied to each O L R composite to assess the statistical significance of results and the results only at 95% significance level are shaded. It is observed that in the case of seasonal composite, the lowest O L R values (<200 Wm2 ) occur in the Bay of Bengal around 90 E, 15 N. In the case of active composite, this pattern elongates westwards
Fig. 4. O L R patterns for 1975–2001(excluding 1978) in Wm2 (a) seasonal mean, (b) active composite, (c) weak composite (d) break composite (in case of composites, statistically significant results at 95% level are indicated by shading)
58
V. S. Prasad and T. Hayashi
towards the Indian mainland and covers most of the Indian region. This indicates widespread rainfall activity. In the case of weak and break phases, O L R is higher than 240 Wm2 over a large part of India. The seasonal pattern of low O L R over the Bay of Bengal shifts towards the northeastern part of India in the break period and southeastward in the weak period. Thus, Fig. 4 shows a strong signature of the boreal summer Maddan-Jullian Oscillation (MJ O ; Maddan and Julian, 1994) with the active period characterized by the low OL R in the northwest-southeast tilting band across India through Indonesia, and the break period characterized by the low O L R in the central Indian Ocean basin at 95% confidence level. But recent studies of Wang et al (2005) showed that the antecedent of active=break cycles emerges in the western equatorial Indian Ocean itself. In view of these results, the relationship between the active=break cycle and MJO should be evaluated more critically. Thus, the proposed index for June–September period is subjected to Lomb-Scraggle periodigram (Scraggle, 1982) analysis to identify dominant modes of oscillations. Figure 5 depicts such analysis for the good monsoon (1978; 1981; 1988; 2001) and poor monsoon (1979; 1982; 1987; 2002) years. Here, a good (poor) monsoon year is related to above (below) normal rainfall in the monsoon season on an All India basis in that year. It clearly shows
that, in case of poor monsoon years (thin lines) low frequency oscillations (10–20 day period) are insignificant and they are dominated by oscillations in the MJO scale. But in the case of good monsoon years high-frequency 10–20 period oscillations are also significant. Thus, the MJ O may be key in the occurrence of active and break phases of I S M and short intraseasonal variability is concurrent to it. It also indicates that the major difference between good and poor monsoon seasons is negligible=absent of low-frequency oscillations in the latter. It agrees with Krishnamurti and Bhalme (1976) finding that more long dry spells (breaks) occurs in a poor monsoon year than in a good monsoon year. Similarly, to study the circulation characteristics, composite patterns of 850-hPa wind vectors based on I S M phases are prepared and then compared with those of the seasonal pattern. In the active composite, strong westerlies occur south of 20 N and strong easterlies occur to the north implying large cyclonic vorticity (Fig. 6b). In the weak composite, a general decrease in the strength of the westerlies is observed throughout the Indian region (Fig. 6c). Whereas in the case of the break composite, a shift of the westerly region north of 20 N can be observed (Fig. 6d). It is also found that the Ramamurthy (1969) and I M D criterion for breaks, i.e. westerlies over the entire region with no easterlies to the
Fig. 5. Lomb-Scraggle periodigram analysis for the H W S I and for the good monsoon (1978, 1981, 1988, and 2001; thick line) and poor monsoon year (1979, 1982, 1987, and 2002; thin line)
Active, weak and break spells in the Indian summer monsoon
59
Fig. 6. 850-hPa zonal wind (m=s) composite for (a) seasonal mean, (b) active composite, (c) weak composite, (d) break composite (in case of composites, Statistically significant results at 95% level are indicated by shading)
north of the normal position of the monsoon trough, is satisfied from the surface to 700 hPa (not shown). The same is also found in the case of weak spells. Further correlation analysis is made between the number of active, weak and break days, and compared with seasonal and monthly rainfall over the Indian region and the sub-Himalayan West Bengal Sub-division (Tables 1 and 2, respectively). This subdivision is picked up as representative of Table 1. Correlation Coefficients (CCs) for All-India Rainfall (A I R ) for months June–September and June–September (JJAS) totals, and the number of active, weak and break days A I R vs
JJAS
Jun.
Jul.
Aug.
Sep.
Active Weak Break Break þ weak
0.48þ 0.61 – –
0.67þ 0.21 – –
0.28 0.25 0.59þ 0.60
0.34 0.35 0.52þ 0.56
0.51þ 0.52þ 0.42þ 0.57þ
CCs significant at the 95% and 99% confidence levels are denoted by and þ , respectively
Table 2. Correlation Coefficients (CCs) for Sub-Himalayan West Bengal Rainfall (foothills) for months June–September and June–September (JJAS) totals, and the number of active, weak and break days Sub-Himalayan West Bengal rain vs
July
August
September
Break Weak Break þ weak
0.27 0.14 0.24
0.33 0.11 0.17
0.11 0.21 0.42
CCs significant at the 95% and 99% confidence levels are denoted by and þ , respectively
rain over the foothills. Table 1 shows that AI R for the season and for the peak season (July and Aug) significantly negatively correlated with the number of break and weak days rather than active days. This agrees with the finding of Sikka (1980). However, for the initial month of June, it is positively correlated with the active days, and for September, it is equally positively correlated with the active days and negatively with weak and break days. Similarly, the correlations between
60
V. S. Prasad and T. Hayashi
weak days and A I R are not very high in June, July and August when compared with those of break plus weak days. The number of break days and foothills rain are positively correlated, whereas the number of weak days are slightly negatively correlated (Table 2). Thus, circulation and proxy rainfall composite pattern and correlation analysis clearly highlighted the climatologically observed features of the ISM phases. 6. Comparison of H W S I breaks and breaks based on other criteria The breaks identified on the basis of the present method are then compared with those of R þ D breaks and also with those of the rainbreaks. These two methods (R þ D breaks and rainbreaks) are only considered for comparison as they are based on basic station and raingauge observations. To ascertain the overlap with the breaks identified in these various studies, only breaks of duration of two days and longer are considered (table not
shown). For the 1958–1989 period, there are about 242 H W S I breaks, 235 rainbreaks and 223 R þ D breaks in total. The major difference is in 1967, 1974 and 1987, in which the HWSI and rainbreaks agrees well compared to R þ D breaks. On the whole, there is good agreement between all three breaks with varying degrees of overlap, particularly for long-duration breaks. For the years 1959, 1962, 1967 and 1968, the results are similar to the case of HWSI and R þ D breaks. Most of the differences occurred in breaks of short duration. Some differences may be due to differences in classifying extreme weak and break situations. In both cases, rainfall over the monsoon zone decreases, but in the case of a break, it shifts towards the foothills. As shown in the above discussion of the 1979 illustration, H W S I classifies 15–26 August as a break period and 27–31 August as a weak period. Figure 7 depicts the composite of O L R and 850-hPa-level winds of both of these break (a and c) and weak (b and d) periods. It can be clearly observed that these
Fig. 7. Daily means of O L R (Wm2 ) and 850-hPa-level winds (m=s) for break (a) and (c), and weak (b) and (d) spells during August 1979
Active, weak and break spells in the Indian summer monsoon
are entirely different situations. In the break, the main rainfall zone (O L R as proxy) shifts towards the foothills and the northeast region along with monsoon circulation. However, in the case of weak composite, there is general decrease in the rainfall throughout the country and major rainfall activity shifts south of 5 N. Secondly, in the weak composite, the absence of monsoon circulation can be seen with northwesterly domination over the northeast and the foothills. Note that August 15–31 is a break based on the criteria used by De et al (1998) as well as by Gadgil and Joseph (2003). This clearly shows the efficacy of the chosen criteria on the basis of the HWSI Index in the identification of breaks. Thus, it can be concluded that the H W S I criteria are reasonable for active=break studies of the ISM . 7. Conclusions Objective criteria are developed to monitor the different phases of the ISM and tested. The different phases so categorized are found to be related not only to the A I R fall distribution but also to the circulation characteristics. This is useful for I S M season classification. The criteria may be useful for medium range weather forecasting using GCM outputs. Acknowledgments This study is supported by the Japan Society for the Promotion of Science through the FY2003 Postdoctoral Fellowship program. The first author expresses his thanks to the Head of NCMRWF for the support and encouragement that he provided for conducting this work. The authors thank the Editor and the two anonymous reviewers for constructive comments that led to improvements in the final manuscript. The ECMWF ERA-40 data used in this project were provided by ECMWF or obtained from the ECMWF data server. References De US, Lele RR, Natu JC (1998) Breaks in the Southwest monsoon. Indian Meteorological Department, Report No 1998=3 Gibson JK, Kallberg P, Uppala S, Hernandez A, Nomura A, Serrano E (1997) ERA description. ECMWF Reanalysis Project Report Series 1, 72pp Krishnamurti TN, Bhalme HN (1976) Oscillations of a monsoon system, part 1. Observational aspects. J Atmos Sci 33: 1937–1954 Krishnan R, Zhang C, Sugi M (2000) Dynamics of breaks in the Indian summer monsoon. J Atmos Sci 57: 1354–1372
61
Libmann B, Smith CA (1996) Description of a complete (interpolated) outgoing longwave radiation data set. Bull Amer Meteor Soc 77: 1275–1277 Madden RA, Julian PR (1994) Obsetvations of the 40–50 day tropical oscillations: a review. Mon Wea Rev 112: 814–837 Parthasarthy B, Munot AA, Kothawale DR (1994) All India monthly and seasonal rainfall series; 1871–1993. Theor Appl Climatol 49: 217–224 Prasad VS, Taiichi Hayashi (2005) Large-Scale summer monsoon rainfall over India and its relation to 850 hPa wind shear; hydrological processes (GAME special issue) (accepted) Ramamurthy K (1969) Monsoon of India: Some aspects of the ‘‘break’’ in the Indian southwest monsoon during July and August. Forecasting Manual 1–57 No IV 18.3: Indian Meteorological Department Rao YP (1976) Southwest monsoon. Meteorological Monogr, Synoptic Meteorol No. 1=1976, Indian Meteorological Department, Delhi, India Scraggle JD (1982) Statistical aspects of spectral analysis of unevenly spaced data. Astrophys J 263: 835–853 Sikka DR, Sulochana G (1978) Large-scale rainfall over India during the summer monsoon and its relationship with the lower and upper tropospheric vorticity. Ind J Meteor Hydrol Geophys 29: 219–231 Sikka DR, Sulochana G (1980) On the maximum cloud zone and the ITCS over India longitude during the southwest monsoon. Mon Wea Rev 108: 1840–1853 Sulochana G, Joseph PV (2003) On breaks of Indian monsoon. Proc Indian Acad Sci (Earth Planet Sci) 112(4): 529–558 Syroka J, Toumi R (2002) Recent lengthening of the south Asian summer monsoon season. Geophys Res Lett 29 (DOI: 10.1029=2002 GL015053) Syroka J, Toumi R (2004) On the withdrawal of the Indian summer monsoon. QJR Meteorol Sci 130: 989–1008 Wang B, Fun Z (1999) Choice of South Asian summer monsoon indices. Bull Amer Meteor Soc 80: 629–638 Wang B, Wu R, Lau KM (2001) Interannual variability in the Asian Summer monsoon: contrast between the Indian and western North Pacific–East Asian monsoons. J Climate 14: 4073–4090 Wang B, Webster PJ, Teng H (2005) Antecedents and self induction of active-break South Asian monsoon as unraveled by satellites. Geophys Res Lett (DOI: 32=1029= 2004GL020996) Wu GX, Zhang YS (1998) Tibetan Plateau forecasting and timing of the monsoon onset over South Asia and South China Sea. Mon Wea Rev 126: 913–927 Xie PP, Arkin PA (1997) Global precipitation: a 17-year monthly analysis based on gauge observations, satellite estimates and numerical model outputs. Bull Amer Meteor Soc 78: 2539–2558 Corresponding authors’ address: V. S. Prasad (E-mail:
[email protected]) and T. Hayashi, Disaster Preventive Research Institute, Kyoto University, Gokasho, Uji, Kyoto, Japan