Errata: Retraction Volume 34, Issue 12, 3480, Article first published online: 22 August 2014
Analysis of rainfall variation over the central Indian (CI) region, using a high resolution rainfall gridded data set over the Indian longitudes, shows that the recent decade (1998–2009) has large number of drought events, compared with the earlier two decades (1979–1997). Improved understanding of the underlying mechanism responsible for these summer monsoon droughts is important because of their profound socio-economic impact over this region. Our analysis here reveals that most of the recent droughts are associated with extended break events (breaks lasting more than 7 d) and that the process responsible for it may also be the mechanism responsible for these droughts.
Sub-seasonal diagnostics, focusing on such drought-producing break monsoon events, show prominent structures resembling the central Pacific (CP) El Niño warming over central equatorial Pacific Ocean. We then performed sub-seasonal moisture budget diagnostics using newly available European Centre for Medium Range Weather Forecasts (ECMWF) interim reanalysis. The primary aim is to identify the various moist processes through which the break monsoon conditions are initiated and maintained over CI, thus leading to large-scale drought conditions. The budget diagnostic shows that dry advection is the principal moisture process to initiate drought-producing break conditions over CI. The primary processes leading to dry advection are as follows: the enhanced rainfall over western north Pacific (WNP) induces cyclonic circulation anomalies to its northwest as a Rossby wave response, and the northerlies at the poleward of this circulation advect dry air of low moisture content from continental subtropics to south Asia. The present result also indicates that the enhanced radiative cooling over CI plays an important role in maintaining such break conditions. A direct implication of our research is that observational efforts are essential to monitor the three-dimensional moisture distribution over the south Asian monsoon region, for a better understanding of these drought-producing break monsoon events.
The boreal summer seasonal (June to September) mean rainfall associated with the Indian summer monsoon (ISM) is the main source of water for the agrarian society of the Indian subcontinent. Extreme departure from this seasonal mean, such as large-scale droughts and floods, seriously affects agricultural output and regional economies. Although successful prediction of such severe events is still fraught with uncertainties, a proper assessment may provide us a better future outlook on extreme events.
Charney and Shukla (1981), in their seminal research study hypothesized that the anomalous boundary conditions of sea surface temperature (SST), sea ice, snow cover, soil wetness and other land surface conditions have significant influence on seasonal mean circulation and rainfall anomalies. Consistent with this hypothesis, several studies suggested that the mean and interannual variability associated with boreal summer monsoon are influenced by slowly varying boundary conditions. The El Niño and Southern Oscillation (ENSO) associated SST anomalies is now widely regarded as a major influence on ISM rainfall (Webster et al., 1998). Apart from ENSO SST anomalies, the anomalous SSTs in the western Pacific and Indian Oceans (either induced by ENSO or developed independently due to local air–sea interactions) are also believed to be important for the Asian monsoon variability (Shukla, 1975; Ju and Slingo, 1995; Soman and Slingo, 1997; Chandrashekhar and Kitoh, 1998; Yun et al., 2008). With the discovery of the Indian Ocean Dipole (IOD, Saji et al., 1999; Webster et al., 1999), the subsequent studies showed the significant importance of IOD (hence regional SST anomalies) in influencing the ISM rainfall (Behera et al., 1999; Ashok et al., 2001). Annamalai and Liu (2005) also noted that the magnitude of monsoon-related rainfall and circulation anomalies are influenced by regional SST anomalies.
The properties of El Niño exhibited changes in frequency and amplitude before and after the late 1970s (An and Wang, 2000). However during late 1990s and 2000s, El Niño events show different characteristics in terms of location of maximum anomalous SST compared with the canonical El Niño (Latif et al., 1997; Ashok et al., 2007; Kao and Yu, 2009; Kug et al., 2009). Recent studies also argued that in the tropical Pacific, there exists another phenomenon that is distinctly different from the canonical El Niño, characterized by a ‘horseshoe’ spatial pattern flanked by a colder SST on both sides along the equator (Rasmusson and Carpenter, 1982). These studies led to various definitions of a new type of El Niño: the dateline El Niño (Larkin and Harrison, 2005), the El Niño Modoki (Ashok et al., 2007), the central Pacific (CP) El Niño (Kao and Yu 2009) and the warm pool El Niño (Kug et al., 2009). During 2004 summer season, it had a maximum SST anomaly in the central equatorial Pacific (Niño4 region, 160°E–150°W and 5°S–5°N), differing from the conventional El Niño pattern, and Modoki (pseudo-El Niño) was named to represent this phenomenon. The pronounced modification in the structure of the El Niño has implications for its teleconnection patterns. A recent study by Kumar et al. (2006) emphasized that the El Niño events with warmest SST anomalies in the CP basin are more effective in producing the droughts over the Indian subcontinent than events with the warmest SST in the equatorial eastern Pacific (EP).
The Indian subcontinent recently suffered one of the worst droughts during the boreal summer season of 2009. Against the long period average, it produced a monthly rainfall of 53% in June (imposing a deficit of 47%), 96% in July (within a normal limit), 73% in August (a deficit of 27%) and 79% in September (a deficit of 21%), with a seasonal rainfall of 77% (a deficit of 23%). To motivate our focus, Figure 1(b) shows the time series of anomalous rainfall averaged along the central Indian (CI) longitudes (74.5–85°E, 16.5–26.5°N, denoted as CI and shown as a box in Figure 1(a)). The choice of the CI domain follows the conjecture that the climatological mean and variance are relatively homogenous over this region (Figure 1(a) and (c)). Moreover, a close look at the vegetation cover in a normal monsoon of 2001 (using the filled normalized difference of vegetative index, NDVI, figure not shown) over the Indian region reveals a tremendous change over CI compared with other regions, from pre-monsoon onset to post-onset phase (zero values of NDVI in earlier phase indicating poor vegetation and high values in later phase resulting in healthy vegetation). So this suggests that a severe drought or wet condition over this region will have a huge impact on the socio-economic condition of India.
It appears from Figure 1(b) that the recent decade has a large number of droughts (see Section 'Data and methodology' for a detailed definition) over CI (red line) compared with the earlier two decades; about four droughts occur during the last 10 years (1998–2009) as against the only two events in the early 20 years (1979–1997). Further analysis of the rainfall averaged along all Indian longitudes (over the region 70–100°E, 8–25°N and black line, here after denoted as AI) shows that the earlier two decades have only two prominent droughts. Since the recent decade shows an increase in drought conditions over India, especially over CI, it will be interesting to conduct a detailed study to understand and thus elicit the dominant features during these droughts. Therefore, we here explore on these recent droughts by looking at the various anomalous atmospheric and oceanic features.
The monthly data utilized for this study (Sections 'Seasonal diagnostics', 'Is there any recent drying tendency over the CI region?' and 'Summary and discussion') are anomalies of SST, rainfall and atmospheric circulation data. The anomalies are calculated after removing the monthly climatology. The period of analysis is from 1979 to 2009. We use the NCEP/NCAR reanalysis products (Kalnay et al., 1996), which have a horizontal resolution of 2.5° longitude and 2.5° latitude, to depict the atmospheric circulation. The SST data, with a spatial resolution of 2° grid, are from the improved extended reconstructed version 2 datasets (ER SST, Smith and Reynolds, 2004) created by National Climate Data Center.
A high-resolution daily rainfall data over Indian region is recently available for the period from 1951 to 2009 (Rajeevan et al., 2006). This data has a horizontal resolution of 1° latitude × 1° longitude, and is based on a quality controlled daily rainfall data at 1803 stations well distributed over the country. For this study, we use both monthly (Section 'Seasonal diagnostics') and daily (Section 'Sub-seasonal diagnostics') rainfall data from this. Apart from the high-resolution gridded (HRG) rainfall, this study also uses the monthly precipitation data sets from the Global Precipitation Climatology Project (GPCP version 2.1, Huffman et al., 2009).
In Section 'Sub-seasonal diagnostics', we utilize daily anomalies for all the diagnostics. The daily anomalies are calculated by removing the annual cycle, composed of the time mean and the first three harmonics. Again in Section 'Sub-seasonal diagnostics', bulk of the atmospheric data used come from European Centre for Medium Range Weather Forecasts (ECMWF) interim reanalysis datasets (here after called ERA interim reanalysis). This product covers the data-rich period of 1989–2009 and it is considered to capture well the tropical precipitation and the hydrological cycle based on a system that uses new humidity analysis and improved model physics (Annamalai, 2010). The horizontal resolution is 1.5° and there are 60 vertical levels. More details can be found in Simmons et al. (2007). The daily variables analysed here include winds and specific humidity (q) at all pressure levels. In addition, rainfall, radiation fluxes and latent heat flux are also used.
To assess the robustness of the results from the ERA interim reanalysis, we also use three other daily observational rainfall products. The first one is HRG dataset, as described above and second one is 3-d running mean precipitation estimates from Tropical Rainfall Measuring Mission (TRMM B42 version) Micro Imager (TMI; Huffman et al., 2007) at 1° × 1° resolution for the period from 1998 to 2009. Last one is a proxy data set for rainfall, 2.5° × 2.5° gridded OLR (outgoing longwave radiation) from the advanced very high resolution radiometer (Liebmann and Smith 1996) during the period 1979–2009. In addition to this, to identify and verify whether the monsoon breaks during the recent drought years co-occur with any anomalous boundary forcing such as ENSO, we use the daily SST observations from TMI.
2.2. Diagnostic methods
2.2.1. Identification of events
Following the discussions in the introduction, we categorize the drought and wet conditions (Section 'Seasonal diagnostics') using the boreal summer seasonal rainfall (with HRG datasets) averaged over the CI region (containing 143 grid boxes). The drought (wet) events are considered if the seasonal CI rainfall anomalies are below (above) 15% of the climatological normal. The number of drought years during the period 1979–1997 (1998–2009) is 2 (4, respectively). The number of wet years during these periods is 4 (1, respectively). Figure 1(b) illustrates these details further.
In Section 'Seasonal diagnostics', the categorization of El Niño events are made by the widely used measures of ENSO, Niño3 index (area averaged anomalous SST over the region 170–120°W, 5–5°N) and Niño4 index (over the region 160°E–150°W, 5°S–5°N). The selection of events are such that if either Niño3 SST or Niño4 SST is greater than its corresponding standard deviation during the boreal summer season. There are nine events during this period. Of these, two El Niño events (1982 and 1983) exceed the SST threshold limit only over the Niño3 region. We call these canonical El Niño events as EP events. Similarly there are five events (1991, 1994, 2002, 2004 and 2009) which surpass the threshold limit only over the Niño4 region. These events are hereafter denoted as CP events. Note here that the CP events are quite similar to the El Niño Modoki events in terms of spatial characteristics and amplitude. In addition, there are two El Niño events (1987 and 1997) satisfying the above criteria over both regions. During the earlier period (1979–1997), only one drought event (1987) coincides with the Pacific Ocean warming event with the exception of 1979. In the recent period (1998–2009), three of four drought events occur in association with CP-El Niño events. The only exception is the event of 2000.
In Section 'Sub-seasonal diagnostics', for all the diagnostics on break monsoon events, we focus on the period 15 June to 20 September that represents the established phase of the monsoon over CI. The classification of a break event is such that if for three consecutive days, daily rainfall (OLR) anomalies averaged over CI are below (above) 1 standard deviation. On the basis of the above definition, dates of the monsoon breaks identified from the four products are shown in Table 1. There is general agreement in the identified dates, but there are some clear differences as well, highlighting the uncertainty in observation and reanalysis.
Table 1. Break monsoon dates of the recent drought years as identified from four different rainfall datasets
Break monsoon dates of the recent drought years
5 Jul–17 Jul,
21 Jul–29 Jul
2 Jul–16 Jul,
22 Jul–30 Jul
5 Jul–14 Jul,
25 Jul–30 Jul
2 Jul–15 Jul,
21 Jul–28 Jul
22 Jun–28 Jun,
26 Aug–3 Sep
20 Jun–1 Jul,
26 Aug–3 Sep
21 Jun–28 Jun,
23 Aug–2 Sep
21 Jun–27 Jun,
24 Aug–1 Sep
14 Jun–26 Jun,
24 Jul–10 Aug,
12 Sep–19 Sep
11 Jun–25 Jun,
24 Jul–8 Aug,
12 Sep–18 Sep
13 Jun–23 Jun,
29 Jul–8 Aug,
11 Sep–21 Sep
24 Jul–9 Aug,
12 Sep–20 Sep
Total number of break days
2.2.2. Moisture budget formulation
Here, we provide the formulations for the moisture budget (used in Section 'Sub-seasonal diagnostics'). The nonlinear primitive equation for moisture anomalies (perturbations), when vertically integrated, takes the following form:
where the prime denotes anomalies, the horizontal velocity vector, ω the vertical pressure velocity and the overbar denotes the vertical integration over the troposphere. The term q ′ is in energy units (W m−2), with the latent heat of evaporation (L) absorbed and represents anomalous moisture sink.
Using Yanai and Tomita (1998), Equation (1) can be written as
where E′ represents the anomalous surface latent heat fluxes and P′ the anomalous precipitation. Both E′ and P′ are in energy units (W m−2), with the latent heat of evaporation (L) absorbed. On the right hand side of Equation (2), the first two terms respectively represent the anomalous moisture advection (horizontal) and moisture convergence.
where the net flux at the top of the atmosphere is
and the net flux at the surface is
where subscripts s and t denote surface and top of the atmosphere, and R↑ and R↓ are upward and downward long wave fluxes, and similarly, S↓ and S↑ are for shortwave fluxes. These fluxes are signed in the direction of the fluxes (Su and Neelin, 2002).
In Section 'Sub-seasonal diagnostics', the whole diagnostic (including budget analysis) is carried out with no pre-filtering (band pass), but to evade synoptic noise, 5-d running mean is applied before presentation. Also, the sign convention adopted in the budget diagnosis is such that a positive sign means net moisture convergence into the column and vice versa. Similarly, dry advection out of the column is defined negative and moist advection is positive. For anomalous Fnet, negative (positive) values indicate cooling (warming) in the atmospheric column.
3. Anomalous atmospheric and oceanic characteristics of the recent droughts
3.1. Seasonal diagnostics
The variability of monsoon is largely controlled by the internal dynamics of the atmosphere; however, the slowly varying boundary forcing from underlying oceans and land also plays an important role in affecting the interannual variability of the monsoon. The SST anomalies associated with ENSO are the dominant factor, as already mentioned in the introduction. So in Figure 2, we display the anomalous SST patterns in June to September of the recent drought years. It also provides the SST map of 1987 (pronounced drought during 1979–1997) for comparison. To aid in interpretation and comparison, in all the panels, we show the significant positive SST anomalies (shaded contours, indicating warming) satisfying the criterion that the standardized SST anomalies at those grid points exceed its corresponding standard deviation. The anomalous SST distribution during the recent drought years (Figure 2(a)–(c)) differs from the El Niño of 1987 (Figure 2(d)), such that the warming is mostly confined to the central equatorial Pacific. The associated anomalous warm SST in the CP region extends more meridionally and has weaker zonal SST gradients compared to its counterpart, EP events. Also, this boreal summer warming in the central tropical Pacific distinctly persists through the following boreal winter (figure not shown). So the spatial characteristics of anomalous SST associated with the recent droughts resemble CP El Niño events more than EP El Niño events. But during 1987, the warming extends up to date line, thus assuming the features of both CP and EP events. Note that in 2009, there is maximum SST in the EP, but the warming is not much as pronounced (see the shaded contour) as that of the canonical EP events during boreal summer and hence it falls under the CP El Niño category (Section 'Data and methodology'). Interestingly, the distribution of precipitation anomalies during this year (Figure 3(c)) also reveals that the anomalous surplus rainfall confines to the central and subtropical Pacific. We will discuss this next, in detail. Apart from these features, the Indian Ocean (IO) is anomalously warmer during 2002 and 2009, possibly due to the cloud-free condition caused by the anomalous subsidence. Similar patterns are also observed in the drought event of 1987.
Figure 3(a)–(d) shows the anomalous rainfall and low-level circulation features, for the boreal summer seasons of 2002, 2004, 2009 and 1987, respectively. During these years, there is reduced rainfall over the Indian subcontinent with anomalous low-level easterlies, suggesting a weakened monsoon circulation (Figure 3(a)–(c)). But, the most conspicuous feature during the recent drought years is that while the SST warming (shaded contours) confines to the central tropical Pacific, the associated maximum surplus rainfall appears over the central equatorial Pacific and western north Pacific (WNP). Our motivation for this study originates from this observation that the anomalous high rainfall during these recent drought years ranges from the western to CP with weak rainfall anomalies over the EP, thus resembling the atmospheric response during CP El Niño events as described in Yeh et al. (2009) (see their Figure 2(b)). This is in contrast to the canonical EP events in which the equatorially confined rainfall anomalies are more tightly aligned with the anomalous warm SSTs in the tropical EP and are seen only to the east of 160°E. During the summer season of 1987 (Figure 3(d)), the maximum rainfall is near the date line, with the precipitation band extending throughout the equatorial Pacific.
For a detailed understanding of the distinct atmospheric signatures (especially over the western to CP) associated with the recent drought events, we present the anomalous map of (Figure 4(a)–(c)) geopotential height and vorticity (both at 850 hPa). During the recent drought events, there is anomalously suppressed convection over the Maritime Continent region (Figure 3(a)–(c)) with overlying negative vorticity anomalies, depicting an anomalous anticyclone (see contours in Figure 4(a)–(c)). One can also see anomalous lower level westerly wind over the tropical western CP, extending meridionally to 15°N (Figure 3(a)–(c)). The corresponding analysis of the longitude–pressure sections of circulations over the equatorial (averaged over 10°S–5°N, figures not shown) regions indicates that the anomalous CP warming generates a regional Walker circulation with upward motion near the CP and descending limb near the western Pacific and equatorial IO, thus suppressing convection over there. To the north of the anomalous anticyclone over the equatorial western Pacific, there is a well-organized anomalous cyclone over WNP extending from south China Sea to Philippine Sea region, followed by an anomalous anticyclone further north of it (Figure 3(a)–(c) and Figure 4(a)–(c)). Associated with these, there is anomalous low over WNP region and a high north of it, as seen from the map of low-level 850-hPa geopotential anomaly (see shaded contours in Figure 4(a)–(c)). During 2009, the anomalous high is slightly northeast compared with other years. The aforementioned features are in agreement with recent studies (Weng et al., 2009; Chen and Tam, 2010; Pradhan et al., 2011). However, during the summer drought of 1987, the above anomalous features over WNP are not apparent.
As noted above, the recent drought events show a broad area of positive vorticity anomalies from the south China Sea extending into the central tropical Pacific (Figure 4(a)–(c). A recent study by Pradhan et al. (2011) infers that CP warming or El Niño Modokis have a strong tendency to be associated with increase in tropical cyclone frequency during boreal summer. Earlier observational and modelling studies (Kumar and Krishnan, 2005; Mujumdar et al., 2007) also indicate that the CP warming increases the typhoon activity over WNP. So our results here are also consistent with their findings.
We noted earlier that during the recent droughts, there are regions of decreased and increased precipitation (Figure 3(a)–(c)) from the equatorial western Pacific to mid-tropics. Also there are alternating meridional pattern comprising of an anomalous anticyclone over the equatorial western Pacific; a cyclonic anomaly over WNP and an anticyclone further northward. These anomalous features (in low-level circulation and rainfall) can be interpreted as meridional dispersion of Rossby waves associated with the warmer SST anomalies in the CP region (Li and Wang, 2005; Lau and Wang, 2006), as per the Matsuno–Gill (Matsuno, 1966; Gill, 1980) theory. This also explains the meridional pattern of low and high rainfall regions observed over the equatorial western Pacific. The enhanced precipitation over WNP is then expected to subsequently induce subsidence over the Indian subcontinent. According to an earlier study by Kumar et al. (2006), the ‘westward shifted’ Pacific Ocean warm events drive more intense sinking over the Indian region, initiating severe drought condition over the Indian landmass. However, our present analysis indicates that the convection variability over WNP can serve as an important component that mediates the ENSO–monsoon teleconnection dynamics. More detailed discussion of WNP causing drought conditions over the Indian subcontinent follows in the next section (Section 'Moist process'), where we extend our discussion on the recent droughts to sub-seasonal time scales.
3.2. Sub-seasonal diagnostics
The interannual variations in monsoon rainfall are relatively low; the interannual standard deviations being around 10% of the summer mean rainfall. Within the boreal summer season, rainfalls fluctuates on intraseasonal timescales of a few days to weeks, giving rise to active and break spells. The duration and frequency of the active/break spells contribute to the seasonal mean and thus modulate the interannual variability (Goswami and Ajayamohan, 2001). During the break spells, the trough moves to the foothills of the Himalayas and dry conditions prevail over central and northwest India. While breaks are intrinsic to the monsoon system, prolonged dryness or extended breaks (breaks lasting between 3 and 7 d or more) during the peak rainy season often results in droughts (Ramamurthy, 1969). So it is the extended break events that often have large impacts that particularly affect agriculture or water supply (Webster et al., 1998). In this regard, the studies further show that these extended breaks contribute to substantial reduction in agricultural output and growth of gross domestic product (Gadgil and Gadgil, 2006). So, proper analyses are requisite to have a better understanding on such drought-producing break monsoon events, and also for a quantitative assessment of its impact on the agriculture production and economy. In this section, we attempt to carry out additional analysis on sub-seasonal timescales, focusing on the break monsoon events of the recent drought years. In particular, our effort is to understand the interaction between moist physics and circulation causing such break monsoon events.
3.2.1. Composite evolutions of the break events
Figure 5(a)–(d) shows the break monsoon composite rainfall maps, averaged over the total number of break monsoon days, from all the data products (see Table 1 for the total number of break days for the recent drought years and Section 'Moisture budget formulation' for more details on the identification of break events). From Table 1, one can easily note that the recent droughts are associated with break events of prolonged duration of about 5–12 d (extended breaks). The HRG and TRMM observations show considerable rainfall suppression over CI and along the Western Ghats, with a maximum of about 8 mm day−1 (Figure 5(a) and (b)). The other data sets show similar signatures (Figure 5(c) and (d)) and this robustness in various data products (observational and reanalysis data outputs) is remarkable. However, there exists large assortment in magnitude and different shading intervals in Figure 5(c) and (d) further suggest this. As an example, OLR anomalies over CI region are very weak (about one-quarter of magnitude observed in HRG) compared to HRG observations (Figure 5(a) and (c)). The corresponding rainfall anomalies from ERA interim reanalysis show it to be about half of the magnitude as that of HRG (Figure 5(a) and (d)). Additionally, OLR anomalies over the tropical WNP (115–150°E, 10–22°N) and central equatorial Pacific are comparatively weaker than TRMM observations. It is important to note here that the present research does not intend to focus on the intercomparison of rainfall products; instead, our interest is to identify various moist processes through which the dryness is developed over CI region and maintained through the break monsoon period, leading to large-scale drought conditions.
In Figure 5(e), we also display the composite SST anomaly (averaged all over the break monsoon days) using the available TMI data. It suggests the persistence of warm SST anomalies along the central equatorial Pacific and also the preference of drought-producing monsoon breaks to occur simultaneously with CP El Niño events. All these further illustrate the co-occurrence of CP El Niño events during the recent drought years, as described in Section 'Seasonal diagnostics'
Basically, deep tropical convection is characterized by low cloud temperatures and small OLR values, whereas regions of large OLR indicate the absence of cloud cover (Figure 5(c)). The suppressed convection allows increased radiative heat loss from the atmosphere. On the other hand, over cloudy region, radiative cooling is comparatively less. As clouds and moisture have a tendency to trap outgoing radiation, anomalous low OLR leads to radiative heating of the atmosphere. This cooling (warming) tends to strengthen the descending (ascending) motion over that region, which suppresses (enhances) precipitation in cloud-free (cloudy) areas (Raymond, 2000). More detailed discussion on this anomalous radiative feedback is given later.
Now we present the space–time composite evolution of break monsoon events for further illustrating that the break events over CI are not of local origin but are part of large-scale perturbations involving other regional Asian monsoon heat sources in combination with anomalous boundary forcing. For the rest of the diagnostics in this section, we use only ERA interim reanalysis. The preparation of the composites is from the extended breaks identified from ERA interim rainfall, so as to be consistent with our subsequent discussions (Section 'Moist process') using moisture budget estimates.
Figure 6 shows the space–time evolution of composite rainfall anomalies starting from 15 d before the peak break monsoon phase, with ‘day 0’ being referred to as peak break phase over CI. So hereafter, the ‘day 0’ represents the maximum amplitude of negative rainfall anomalies. In addition, we also provide the vertically integrated moisture flux to diagnose the moisture transport and to understand the atmospheric response to regional rainfall anomalies. Note here that vertically integrated moisture flux mainly replicates low-level flow patterns because of abundant moisture there. At day −15, negative rainfall anomalies cover the central equatorial IO which extends to the southern peninsular parts of India by day −10 (Figure 6(a) and (b)). During the same time period, there are two active convection (positive anomalies) centres, one over tropical WNP and another over the central equatorial Pacific near 160°E. In association with these rainfall or diabatic heating anomalies, there are weak easterly wind anomalies over the equatorial IO together with a divergence centre over the Maritime continent; while westerly anomalies occupy the equatorial western Pacific. These westerlies are interpreted as partly due to Rossby wave response to the positive rainfall anomalies over the central equatorial Pacific and also partly due to Kelvin wave response to the negative rainfall anomalies over the Maritime continent. In its northward migration, the rainfall anomalies over the equatorial IO intensify and occupy the entire Indian landmass by day −5 (Figure 6(c)). At day 0, negative rainfall anomalies are diagonally oriented from CI to the western end of the equatorial Pacific, with a maximum of about −140 W m−2 over the region extending from northern Bay of Bengal to CI region (Figure 6(d)). During this period, an anticyclone covers the south Asian monsoon region encompassing the Indian landmass with prominent easterly anomalies. By this time, the positive rainfall anomalies over WNP strengthen further, while it is less prominent over the equatorial CP. Another centre of rainfall maximum (positive) appears over the eastern IO (EIO, 70–110°E, 10°S–5°N) which gets weaker by day 10 (Figure 6(d)–(f)). We do similar diagnostics using OLR and TRMM and the above features in rainfall are reproducible (figures not shown). In earlier section (Section 'Seasonal diagnostics'), we noted similar anomalous features in seasonal timescales (though less pronounced) possibly suggesting an interpretation that during these drought-producing breaks, there occurs a superposition of sub-seasonal variability and anomalous boundary forcing. Moreover, studies illustrate that both sub-seasonal and interannual variabilities share the same spatial structure and the former one has the potential to influence the seasonal mean monsoon (Sperber et al., 2000; Goswami and Ajayamohan, 2001).
The prevalence of strong low-level westerlies (Figure 6(a)–(d)), during and prior to the break monsoon conditions, has implications on the El Niño conditions over the central equatorial Pacific. Such strong low-level winds or westerly wind events (WWEs) can trigger equatorial jets and Kelvin waves in the ocean that may influence the air–sea interaction in the region. Furthermore, these WWEs can cause the extension of warm SST anomalies to the central equatorial Pacific, thus influencing the El Niño evolution (Goswami, 2005; Saith and Slingo, 2006).
Other studies focusing on the dynamics of Indian monsoon break conditions indicate a smooth transition from break to active phase within a few pentads (Annamalai and Sperber, 2005; Rajeevan et al., 2006). But our diagnostics show here the persistence of break condition over CI from −5 to 10 d and a weak northward migration of positive rainfall anomalies from the eastern IO during 0–10 d (Figure 6(c)–(f)). Therefore the composite picture here suggests the influence of other factors such as rainfall changes over the WNP, EIO and central equatorial Pacific, in conjunction with the El Niño-related anomalous boundary forcing. So these are the factors that need to be considered in order to address our focus on the recent drought conditions over CI. It is important to note here that this study differs from the past studies (e.g. Annamalai and Sperber, 2005), which use band pass filtered data retaining only intraseasonal signals. Here, our composites based on raw anomalies containing the low frequency component further ensure to highlight the anomalous boundary forcing effect.
3.2.2. Moist process
Here, we perform sub-seasonal moisture budget diagnostics using the ERA interim reanalysis in order to identify the various moist processes causing the break monsoon conditions over CI, thus leading to large-scale drought conditions. Note here that the analysed circulation fields and the moist budgets from the ERA interim reanalysis are influenced by moist convective parameterization employed in the forecast model of the reanalysis system (Annamalai et al., 1999).
Figure 7 shows the composite evolution of budget estimates from Equation (2), averaged over the relevant regions (CI, AI, EIO and WNP), for the entire life cycle of break monsoon (starting from day −15 to day 15). For brevity, plot showing the time evolutions over the central equatorial Pacific is not provided here, but discussed. As expected, precipitation (P′) and moisture divergence () are dominant terms governing the moisture budget over the all relevant domains. As an example, at day 0, negative (positive) precipitation anomalies over CI and AI (WNP and EIO) are balanced by moisture divergence (convergence). Quantitatively, over CI domain, moisture divergence accounts for about 90% of the rainfall anomalies.
Horizontal advection () is generally more in phase with the moisture tendency term () than with precipitation anomalies. Over the open oceans of the WNP (Figure 7(c)), EIO (Figure 7(d)) and central equatorial Pacific (figure not shown), evaporation (− E′) and horizontal advection cancel each other during most part of the break monsoon life cycle. However, in CI (Figure 7(a)) and AI (Figure 7(b)), a different picture emerges. Here the horizontal advection is large and negative before the break monsoon rainfall peak (day 0), completely explaining the drying tendency (especially over CI) during the period from day −10 to day 0. Notably, the accumulation of dry air well before the commencement of breaks over CI and AI implies the pivotal role of dry advection as the initiator in causing the break monsoon conditions. More specifically over CI, the dry advection peak leads the established break monsoon phase (with associated maximum in both moisture divergence and rainfall anomalies) by about 5–7 d. Also note that, quantitatively, the contribution by horizontal advection to simultaneous rainfall is almost negligible here, re-emphasizing the importance of dry advection as a precursor.
A question of relevance here, what causes the initiation and maintenance of this dry advection? Recalling our earlier discussion, we notice pronounced westerly anomalies over the equatorial western Pacific during the period from day −15 to day −5 (Figure 6(a)–(c)). This tends to promote cyclonic vorticity over WNP owing to Ekman pumping and subsequently leads to in situ rainfall increase. One can see a prominent cyclonic circulation over the WNP region, as a Rossby wave response to the positive rainfall anomalies (Figure 6(a)–(c)). As a result, the northerlies at the poleward flank of this cyclone will advect dry air (from the north) towards the Indian plains and northern Bay of Bengal. A composite map of horizontal advection (figure not shown) at day −5 clearly manifests this dry air intrusion. Additionally, the northern Arabian Sea anticyclonic circulation anomalies (Figure 6(c)–(e)) advect dry air from higher latitudes to CI through its northerly component. Our present results are consistent with the earlier findings of Annamalai and Sperber (2005). In their study based on the mutual interaction among the regional Asian monsoon heat sources associated with intraseasonal (30–50 d mode) disturbances, they suggest that the enhanced convection over WNP forces descending Rossby waves to the west, promoting rainfall suppression over continental India. So, one possible interpretation here is that the circulation anomalies forced by convective anomalies over both the Maritime Continent and CP influence the active/break phases over WNP and CI. The positive convective anomalies over WNP then forces Rossby descent to the west and initiate break conditions over CI.
After the initiation, the breaks persist for about 10 to 12 d (from day −5 to day 8, see Figure 7(a) and (b)) with anomalous features of anticyclonic circulation covering the entire Indian subcontinent (Figure 6(c)–(e)) and with associated moisture divergence (Figure 7(a) and (b)). The divergence is strengthened and maintained by mainly two factors. Firstly, the positive feedback between anomalous circulation and negative rainfall anomalies over India can strengthen it. Secondly, radiative cooling may play some role as discussed next.
Over the CI region, anomalous Fnet from Equation (3) shows a steady radiative cooling from day −8 onwards, with a peak value of −60 W m−2 at day 0 (Figure 8(a), red line). This cooling dissipates slowly around day 10. Similar signatures are noticed over AI also (Figure 8(a), green line). Over WNP (Figure 8(a), blue line), anomalous Fnet is positive from the beginning of the life cycle, in response to the pronounced rainfall over there (black line in Figure 7(c)). The warming steadily intensifies from day −10 onwards to day 3 and once the rainfall anomalies fade away by day 8; it results in net radiative cooling. On the other hand, over EIO, reduced rainfall (black line in Figure 7(d)) results in net radiative cooling from day −15 to day −5 (Figure 8(a), yellow line). But once the rainfall is enhanced and established, anomalous radiative warming starts growing from day −3 onwards to day 10.
For further interpretation of the cooling over the CI, Figure 8(b) shows the shortwave and longwave radiation balance at the surface and top of the atmosphere. At the surface, the reduction in cloud cover (Figure 7(a)), during day −8 to day 10, promotes stronger absorption of net shortwave radiation flux (, thin line with open circles in Figure 8(b)) and subsequent upwelling of net longwave flux (, thin line), both of them roughly balancing each other. During the same period, the TOA radiation budget shows increase in both net shortwave (see dashed line with triangle in Figure 8(b)) and longwave (OLR, dashed line) emissions, both of them balancing the net radiative cooling (see red line). The enhanced emission is due to the fact that suppressed cloud-free conditions lead to less trapping of outgoing radiation, as described earlier. So over CI, radiative cooling contributes to the total diabatic cooling.
Thermodynamic energy considerations imply that the increased atmospheric radiative loss over CI needs to be compensated by adiabatic descent (Holton, 1982), which achieves here partly through the anomalous east–west circulation between WNP and CI (discussed earlier) and partly through the local Hadley circulation between EIO and CI (as discussed in Sikka and Gadgil, 1980; Krishnan et al., 2000; Annamalai, 2010). Accordingly, one can note a close correspondence between the reduced rainfall and cooling (enhanced rainfall and warming) over CI (WNP and EIO), with these features persisting until day 10 and with the precursory convective signal from WNP occurring from day −15 onwards (Figure 7(a)–(d) and Figure 8(a)). The induced subsidence over CI will in turn inhibit the local convection and rainfall. In response to the suppressed convection and high pressure anomalies, the divergent flow from CI region will enhance. In this regard, a study by Neelin and Held (1987) also suggests that the net radiative forcing at the top of the atmosphere is directly proportional to the low-level divergence. So the pronounced radiative cooling over CI is expected to influence low-level circulation anomalies, which in turn amplifies the negative rainfall anomalies over there. Therefore, this cloud–radiation feedback in turn further strengthens the anomalous divergent circulation over the Indian subcontinent and this feedback continues to operate until the radiative cooling loses its intensity. The implication is that the cloud–radiation interactions are also important for maintaining these drought-producing monsoon breaks.
4. Is there any recent drying tendency over the CI region?
Recently, Yeh et al. (2009) suggest a profound change in the characteristics of El Niño in recent years, with more frequent occurrence of warm events over the central equatorial Pacific. Our analysis here also indicates the same for the recent decade (1998–2009), with the simultaneous occurrence of CP events and Indian droughts (see Section 'Data and methodology' on the categorization of El Niño events). For understanding this further, we examine the anomalous SST patterns during the current decade (1998–2009) and decades prior to it (1979–1997). Figure 9 shows the first leading pattern of the tropical Pacific SST variability for the two periods and it represents the overall strength of the ENSO events. The associated temporal pattern correlates highly with the fluctuations in the all ENSO indices (Niño3.4, Niño4 and Niño3 indices). During the period 1979–1997, the dominant SST mode shows a characteristic anomalous SST maximum over the eastern equatorial Pacific (Figure 9(a)). On the other hand, the anomalous SST in the recent decade (Figure 9(b)) shifts towards the dateline and central equatorial Pacific. In Figure 10, we plot the 11-year sliding standardized means of CI rainfall (red line, from HRG) and Niño4 SST anomalies (blue line) during the boreal summer period of 1979–2009. Niño3.4 SST anomalies (black line, averaged over the region 170–120°W, 5°S–5°N) are also included for showing robustness of the result in other Niño indices. Low-frequency variations in CI rainfall and in both the Niño indices show a clear resemblance till late 1980s. But it diverges largely thereafter (during mid-1990s), suspecting a breakdown of ENSO–monsoon relationship (Kumar et al., 1999). Quite interestingly, both indices show a clear association in the recent decade (after 1997), indicating a recuperation of this relationship. This change reflects the recent decrease in the monsoon rainfall which is manifested in the higher number of drought events in the recent decade (Figure 1(b)). A closer look at Figure 1(b) also reveals that during 1995–1999 rainfall over CI hovers around zero, indicating normal rainfall events.
We further examine the diabatic heating and large-scale circulation patterns during the boreal summer season of 1979–1997 and 1998–2009, in order to discern the anomalous atmospheric features between these two periods. Figure 11(a) and (b) shows the simultaneous correlation between summer precipitation (from GPCP) and CI rainfall (using HRG) for the periods 1979–1997 and 1998–2009, respectively. Note that CI index is multiplied by −1 and so negatively correlated. A pattern resembling the atmospheric response during CP El Niño events appears during the recent decade (Figure 11(b)) with surplus rainfall over WNP and the tropical CP, as shown in the earlier sub-section (Figure 3(a)–(c)). There is suppressed rainfall over the western end of the west Pacific also. The associated large-scale circulation depicts low-level westerlies over the tropical western to CP with a well-organized large-scale anomalous cyclonic circulation, north of it (Figure 12(b)). One can also see a pronounced low over the South China Sea as shown by the distribution of 850-hPa geopotential anomaly. In contrast in the earlier decades (1979–1997), the anomalous atmospheric signals are rather weak over the tropical Pacific. In order to understand the sensitivity of tropical anomalous Walker circulation during these periods (1979–1997 and 1998–2009), we also examine the divergent circulation (using 200-hPa velocity potential) patterns (Figure 13). During 1979–1997, the correlation coefficient between the CI rainfall index and velocity potential shows weaker relationship over the Indian subcontinent, with an eastward shift in the El Niño-induced subsidence pattern (Figure 13(a)). In contrast it is stronger and significant (at 80% level) in the recent decade (1998–2009, see Figure 13(b)). So this study, within the constraints of the duration of the datasets, suggests a possibility that the recent change in the El Niño characteristic and the consequent modifications in the large-scale circulation anomalies, favouring intensification of convective activity over the WNP region (Figures 11(b) and 12(b)), may have contributed to the recent strengthening of the ENSO–monsoon relationship.
5. Summary and discussion
Using a high resolution rainfall gridded data over the Indian longitudes, we analyse the interannual variability of the ISM Rainfall. An examination of the time series of anomalous rainfall averaged along CI longitudes reveals that the recent decade (1998–2009) witnessed a large number of droughts compared with the earlier two decades (four events in the present decade against two events in the earlier two decades, Figure 1(b)). In this study, we focus on these recent droughts, by carrying out seasonal and sub-seasonal diagnostic studies, for an improved understanding of the fundamental mechanisms.
The seasonal distribution (Section 'Seasonal diagnostics') of Pacific SST anomalies in the boreal summer of the recent droughts confines mainly to the central equatorial Pacific (Figure 2(a)–(c)). Corresponding rainfall anomalies (Figure 3(a)–(c)) are anomalously high in the central equatorial Pacific and WNP while it is suppressed over the Indian longitudes and Maritime Continent. During the recent drought years, the CP El Niño warming modifies the regional Walker circulation over the tropical Pacific, which in turn forces large-scale circulation anomalies favourable for the intensification of convective activity over the WNP region. The strong ascending motion over WNP then induces subsidence over Indian subcontinent.
We then examined sub-seasonal climate anomalies, associated with the recent droughts, following the supposition that the prolonged dryness or extended break often leads to drought conditions over CI. These diagnostics (Section 'Sub-seasonal diagnostics') also produce pronounced SST warming over the central tropical Pacific as that of its seasonal counterpart, implying the preference of the drought-producing monsoon breaks to occur simultaneously with CP El Niño events. The poleward propagation of negative rainfall anomalies to Indian and BoB latitudes from day −15 to day −5 (Figure 6(a)–(c)) and subsequent quadrapole rainfall structure at the peak break phase (day 0, see Figure 6(d))) are reminiscent of Madden-Julian Oscillation during boreal summer (Annamalai and Slingo, 2001; Seo et al., 2007). We then performed sub-seasonal moisture budget diagnostics, using the newly available ERA interim reanalysis, so as to advance our understanding on such drought-producing monsoon breaks over CI. This intends to illustrate the importance of moist and thermodynamic processes (involving the interaction between moist physics and large-scale regional circulation with additional feedbacks from cloud–radiation interaction) in initiating and maintaining such break events.
On the basis of the budget diagnostics, we draw the following inferences. Over CI, the injection of anomalous dry air commences approximately 10 d before the break monsoon peak phase (negative rainfall maximum over there, see Figure 7(a)). There are two main sources for this dry air advection. Firstly, the enhanced rainfall over WNP induces cyclonic circulation anomalies to its northwest as a Rossby wave response and the northerlies at the poleward of this circulation advect dry air of low moisture content from continental subtropics to south Asia, particularly to Indian plains and northern Bay of Bengal (Figure 6(a)–(c)). Additionally, the northern hemisphere anticyclonic circulation anomalies advect low moisture air from north to CI (Figure 6(c)–(e)), through its northerly component. The present result also indicates that the cloud–radiation feedback effects, through net radiative flux, play an important role in maintaining the break monsoon conditions (Figure 8). The suppressed convection over CI produces increased net radiative heat loss (primarily as a result of the increase in emission of shortwave and longwave fluxes) from the atmosphere, which needs to be compensated adiabatically by subsidence. The subsidence over CI is achieved through the anomalous equatorial east–west circulation (linking WNP and CI) and the regional Hadley circulation (between EIO and CI). Anomalous subsidence over CI further inhibits local convective activity, which in turn influences low-level divergence further amplifying the negative rainfall anomalies.
At seasonal timescales, the implied descent over CI induced from WNP convective anomalies acts as an additional forcing, further contributing to drought conditions over CI. The results emerged from the present analysis point out that the nature of linkage between CP El Niño events and ISM is intervened by convection changes over the WNP and appear to be rather different from the classical teleconnection mechanism involving the anomalous changes in the equatorial Walker cell (Kumar et al., 2006).
The sub-seasonal budget diagnostics performed here identify the role of dry advection as a precursor signal in causing drought-producing break monsoon conditions over CI. The significance of the moisture advection term from Equation (2) demands that the current models need to capture the details in regional variation in moisture field during the boreal summer period, apart from the details in location and intensity of diabatic heating anomalies along the equatorial and subtropical Pacific. This indicates further the need to capture the details of the moist physics that heavily depend on the physical parameterizations employed. An immediate implication of our budget diagnostics is that observational efforts are necessary to monitor the three-dimensional moisture distribution over the Asian summer monsoon region that would help in better understanding and modelling of severe monsoon conditions. We wish to note here that our budget estimates and subsequent inferences on dry air precursor are from ERA interim reanalysis only. Therefore, the present results need to be verified with other reanalysis products, for consistency.
We often argue that the prospect of predicting the seasonal mean summer monsoon would have been improved if the statistics of the sub-seasonal variability are strongly modulated (or constrained) by the slowly varying boundary forcings such as ENSO. The present study points out the occurrence of drought-producing break monsoon events during anomalous boundary forcing years. So, this may have some implications on seasonal prediction improvement. However, modelling studies so far indicate that the sub-seasonal variability over the ISM region is not sufficiently influenced by slowly varying SST changes associated with ENSO (Goswami, 2005). Moreover, simulation of sub-seasonal variability and our understanding of its interaction with other variability are in their infancy (Sperber and Annamalai, 2008). Furthermore, according to a recent study by Turner and Annamalai (2012), even the best models still show considerable difficulty in simulating the South Asian monsoon and its variability on a range of time scales (from synoptic to interannual). As we are beginning to understand more about processes driving the monsoon, its seasonal evolution and modes of variability, this gives us a major opportunity for a more detailed research by building better models and ultimately reducing the uncertainty in the monsoon rainfall simulation.
K. P. Sooraj sincerely thanks Dr Rajeevan of National Aerospace Research Laboratory, India, for kindly providing the high-resolution gridded rainfall datasets for performing analysis on the recent monsoon years. He expresses his sincere gratitude to Dr Jan Hafner and Dr H Annamalai of International Pacific Research Centre, Hawaii, for fruitful discussions on the moisture diagnostics. He also thanks Dr K. Ashok for his suggestions during the progression of this work. This work was supported by the 2011 Postdoc Development Program of Pusan National University, and the Korea Meteorological Administration Research and Development Program under Grant CATER 2012–3071.