The Indian summer monsoon (ISM) is characterised by rather abrupt onset at the southern tip (Kerala) followed by northward progression of the tropical convergence zone (TCZ) and establishment of monsoon at northern locations (Webster et al., 1998; Tomas and Webster, 1997; Goswami, 2005a). The onset of the monsoon and its northward progression, eagerly awaited by farmers as starting of a variety of agricultural activities are linked with this event (Abrol, 1996). The establishment of the ISM through the northward propagating onset phase and the subsequent active and break spells of monsoon are associated with the summer intraseasonal oscillations (ISOs) (Krishnamurti, 1985; Webster et al., 1998). The summer ISOs are manifestations of fluctuations of the TCZ and its repeated northward propagation between the oceanic preferred position between equator and 10°N and the continental position between 15°N and 30°N (Yasunari, 1979; Sikka and Gadgil, 1980; Krishnamurti, 1985; Webster et al., 1998). Any significant trend in the rate of northward progression of the monsoon ISO during the onset in the recent decades as compared with the past may provide valuable input for agricultural planning and water resource management. It may be recalled that the monsoon onset is always associated with one episode of the northward propagating ISO (Goswami, 2005b). In this paper, we report a significant change in the rate of northward propagation of the first episode of monsoon ISO associated with the onset phase of ISM.
The onset of the ISM is a distinct phase of the monsoon annual cycle and represents the beginning of organised convection in the form of the TCZ to be sustained over the monsoon season and is associated with a manifold increase in the kinetic energy of the low-level westerly jet over the Arabian Sea within a span of less than a week (Goswami, 2005b, Figure 2.14). Although there exists a number of definitions of onset of monsoon over Kerala (MOK) (Ananthakrishnan et al., 1967; Ananthakrishnan and Soman, 1988; Lin and Wang, 2002; Fasullo and Webster, 2003; Gadgil and Joseph, 2003; Goswami and Xavier, 2005; Xavier et al., 2007), the most commonly used definition is based on rainfall over a number of stations exceeding a threshold that is sustained for minimum period of time (Ananthakrishnan and Soman, 1988). Based on a similar criterion, the India Meteorological Department (IMD) defines inset dates on different locations (Ananthakrishnan et al., 1967). Difference in such IMD onset dates over two locations, one in the southern tip of India [Thiruvananthapuram (TVR), located at (76.9°E, 8.6°N), earlier known as Trivandrum] and another over central India [Nagpur (NGR), located at (79°E, 21°N)] would give an idea about the speed of propagation of the first episode of monsoon ISO. The difference in onset dates between TVR and NGR (Figure 1) indeed has an increasing trend significant at 10% level (geographical locations of the two stations are shown in inset). It is to be noted that the onset over TVR has little trend in the period of study (1960–2007) and the trend in Figure 1 is primarily contributed by an increasing trend in onset dates over NGR (figure not shown). This observation indicates that the northward progression of ISO during the onset phase may be systematically slowing down. Such a secular slowing trend has important implications on the extended range predictability of the monsoon ISOs.
The average northward propagation of the summer ISO during the monsoon season (June–September) has been estimated in several studies in the past to be approximately 1° latitude per day (Krishnamurti, 1985; Hartmann and Michaelson, 1989). There is, however, considerable event-to-event and year-to-year variability of the speed of northward propagation of monsoon ISO. However, no quantitative study of the speed of propagation of the ISO during the onset phase alone and how it might have changed over the years has so far been made. The objective of the present study is to quantify the same and unravel the factors responsible for this change in speed of northward propagation of the ISO/TCZ during the onset phase.
2. Data sets used
As reliable onset dates from IMD are available after 1960, all our analysis is carried out for data between 1960 and 2007. High-resolution daily precipitation data over India compiled by IMD (Rajeevan et al., 2006) for the required period between 1960 and 2007 are used to investigate the northward propagation of monsoon ISO in precipitation during the onset phase. This is a quality-controlled data set analysed into 1° × 1° latitude–longitude boxes based on daily data from 1803 stations distributed over India. Large-scale circulation data from National Centre for Environment Prediction/National Centre for Atmospheric Research (NCEP/NCAR) reanalysis (Kalnay et al., 1996) and ERA40 (Uppala et al., 2005) are used to examine the link between precipitation and circulation, and for diagnosing the mechanism for change in speed of propagation. Keeping in mind the year-to-year variability of the onset dates and the period of the onset phase, the onset phase is generally confined to May–June months. Therefore the northward propagation and the mean condition during these two months are examined to understand the mechanism of slowdown of northward propagation.
3. The observed slowdown
Here, we estimate the trend of the northward propagation speed of the monsoon ISO during the onset phase from IMD daily rainfall data. As there is considerable year-to-year variability in the propagation speed, we estimate the average propagation speed in two periods, one early period and one later period. For this purpose, the IMD daily rainfall is passed through a 20–70 day band pass Lanczos filter (Duchon, 1979) to highlight the low-frequency sub-seasonal component of precipitation that is associated with the TCZ movement. A reference time series is created by averaging the filtered data over southern India between 6°N and 12°N and 75°E and 85°E between 15 May and 30 June for all (48) years. The reference time series is then normalised by subtracting long-term mean and then dividing by standard deviation. The lag–lead composites of filtered precipitation between 15 May and 30 June with respect to the peak phases (when standardised anomaly > + 1.0) of the reference time series are constructed which goes up to a lag of 15 days and lead of 15 days. This process essentially picks up the spatial pattern of rainfall before, on and after the peak phase of rainfall. So, for each active peak identified from reference time series, there are spatial patterns for 31 days (from lag − 15 to lead + 15). An average pattern is then created by averaging each of the 31 patterns from the total number of peak phases. This average pattern for each of the 31 days is the composite evolution pattern of all ISO phases selected from the reference time series. Dividing the period of study into two equal parts, the lead/lag composites are created for two periods, namely the first 24-year period (1960–1983) and for the latest 24 years (1984–2007). The composites averaged between 72°E and 85°E for different lead/lag as a function of latitude for the two periods are shown in Figure 2 (top and bottom panels, respectively). The overall slowdown of northward propagation in the second period (as shown by the solid line) compared with that in the first period, primarily comes from a slowdown north of 18°N. Although the progress in this latitude belt was almost instantaneous in the earlier period, it is much slower in recent years. The average propagation speed across the continent has reduced from approximately 2° latitude per day to about 1° latitude per day, a significant reduction.
4. Mechanism for the slowdown
What is responsible for this slowdown of the TCZ during the onset phase of ISM? Are the changes in the regional climate responsible for this slowdown linked with large-scale change in global climate? To find answers to these questions, we examine the critical background mean circulation and thermodynamic parameters responsible for the northward propagation of the TCZ. The northward propagation of the east–west oriented rain band (i.e. the TCZ) occurs due to the atmospheric response of the latent heating associated with the rain band producing anomalous low-level moisture convergence slightly north of the original band leading to northward movement of the heat source and the rain band (Jiang et al., 2004; Goswami, 2005b, Figures 2.11 and 2.15). In the absence of mean background flow with vertical easterly shear of zonal winds, a heat source produces atmospheric response having low-level convergence collocated with the heat source. In the presence of the background easterly shear, atmospheric response produces a barotropic vorticity with maximum north of the heat source. The barotropic vorticity forces frictional boundary layer convergence also to be north of the original heat source thereby moving the heat source northward. Thus, a critical magnitude of easterly shear is required for northward propagation. Theoretical studies have shown (Jiang et al., 2004; Wang, 2005) that the process of generating anomalous moisture convergence to the north of the convective heat source is controlled by the north–south gradient of low-level background humidity close to the equator, whereas away from the equator, it is governed by easterly vertical shear of background zonal winds. Vertical shear of zonal wind between 200 and 850 hPa during May–June and averaged between 60°–90°E and equator–10°N from NCEP/NCAR reanalysis as well as from ERA-40 show a weakening trend (Figure 3, top panel) significant at 5% level. Notwithstanding the small bias in the easterly shear over the region from ERA-40 compared with that from NCEP/NCAR reanalysis, almost identical trends in the two data sets provide confidence in realism of the decreasing trend in the easterly shear. Similarly, the gradient of mean low-level humidity across the equator, represented by the difference between May and June 850-hPa humidity between a north box (60°–90°E; equator–10°N) and a south box (60°–90°E; 10°S–equator), show a clear and significant (at 5% level) decreasing trend in both data sets (Figure 3, bottom panel). Thus, it is clear that changes in regional climate in the form of decreasing trend in easterly shear and decreasing trend in north–south gradient of low-level humidity are responsible for the observed slowing down trend of northward propagation of the monsoon ISO during the onset phase.
What is responsible for the decreasing trend in the easterly shear? The decreasing trend in the shear of zonal winds comes mainly from the trend in zonal winds at 200 hPa, as there is no significant trend in the 850-hPa winds (Figure 4, top panel). Further, strong correlation between May–June zonal winds at 200 hPa and May–June (area averaged) Equatorial Pacific (5°S–5°N, 150°E–80°W) sea surface temperature (SST) (Figure 4, bottom panel) indicates that the upper level winds are strongly related to the El Nino and Southern Oscillation (ENSO) phenomenon. It appears that increasing trend in the frequency of El Ninos (and fewer La-Ninas) in the Pacific could be associated with global warming (e.g. Trenberth and Hoar, 1997; Mason, 2001) that leads to a systematic weakening of easterly shear over the Indian monsoon region through eastward shift and weakening of the east–west Walker circulation (Vecchi et al., 2006; Zhang and Song, 2006).
To demonstrate the Pacific linkage more clearly with respect to the trend in Walker circulation, a combined empirical orthogonal function (CEOF) analysis of the anomalous zonal wind (U wind) and the pressure velocity (dp/dt) for ten standard levels from the NCEP data averaged over the equatorial belt (5°S–5°N) was carried out for the period 1960–2007. The first CEOF and the corresponding principal component (PC1) are shown. The increasing trend in PC1 (Figure 5, bottom panel) represents the increasing influence of the anomalous Walker circulation with ascending motion around the dateline and descending motion over the central and eastern Indian Ocean indicated by CEOF1 (Figure 5, top panel). It may be noted that over the central Indian Ocean (70°–100°E), the shear is positive (westerly). Climatological mean during the period over this region is negative (easterly). Thus, in the earlier 23-year period, the anomalous Walker circulation made the mean easterlies stronger and during the later 23 years, the Walker circulation made it weaker. This is exactly what has resulted in the trends in Figure 5 (bottom panel). While shift of the Walker circulation is an equatorial teleconnection mechanism through which the monsoon easterlies could be affected, local processes such as the Tibetan plateau elevated heat source can also influence the north–south gradient of the tropospheric temperature (TT) and influence the easterly shear. Indeed, the meridional gradient of TT (delTT) as defined by Xavier et al. (2007) is well correlated with the easterly shear during May–June (figure not shown). It may be noted (Goswami and Xavier, 2005; Xavier et al., 2007) that May–June is the transition time for delTT from negative to positive and represents the timing of the Indian monsoon onset. This, in turn, is linked with the El Nino strength through the extra-tropical atmospheric bridge described in Goswami and Xavier (2005). However, there is no trend in delTT. Therefore, the trend in easterly shear (Figure 3, top panel) is likely to be due to the shift of Walker circulation associated with the strengthening of the El Ninos.
The trend in the north–south gradient of 850-hPa humidity is essentially due to a significant trend in northern box (Figure 6(a)) as the same in the southern box is not significant (Figure 6(b)). The decreasing trend in humidity over the northern box during May–June is notable in the backdrop of an increasing trend in surface humidity (Figure 6(c)) as well as that in the sea surface temperature (Figure 6(d)) in the northern box. A significant trend in horizontal convergence of moisture (primarily due to meridional advection, Figure 6(e)) appears to be the primary cause for the decreasing trend in humidity in the northern box. This is because the May–June vorticity at 850 hPa over the box is largely anticyclonic and has no significant trend (Figure 6(f)), and hence, the boundary layer Ekman pumping does not contribute to the trend of humidity over the box.
5. Conclusions and discussions
Unravelling a new facet of regional climate variability, here we show that the northward progression of the ISO/TCZ during the onset phase of ISM has significantly slowed down during recent years compared with the period between '70s and ‘80s. The slowdown of northward propagation is dynamically consistent with the observed weakening of May–June mean easterly vertical shear in the region and weakening of north–south gradient of low-level humidity across the equator. Evidence is provided to support the hypothesis that weakening of easterly shear is due to eastward shift of the Walker circulation associated with strengthening of the El Ninos.
Understanding change of regional climate such as this slowdown of progression of the Indian monsoon and its linkages with slowly varying coupled ocean–atmosphere global phenomena such as the ENSO, provides basis for developing predictive tools for regional climate. While the significant year-to-year variation of the speed of propagation of the initial episode of monsoon ISO after onset over Kerala limits the predictability, the physically consistent significant trend in progression of the monsoon may provide a predictable component that could be factored in the predictive models for progression of the ISM.