Abstract

The Indian monsoon system (IMS) is among the most complex and important climatic features on land. This study proposes a simple and robust method to investigate large-scale variations and changes in the IMS that accounts for fluctuations in amplitude, onset, and duration of the summer monsoon, including active and break phases, and the postmonsoon season. This study uses 35 years (1979–2013) of daily data from the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR) at 1° resolution and indicates great potential for application to other reanalyses and climate model datasets. The method is based on combined EOF (CEOF) analysis of variables associated with the IMS’s seasonal cycle (precipitation, circulation at 10 m, and temperature and specific humidity at 2 m). The first CEOF (CEOF-1) explains ~40% of the total variance and represents the continental-scale Asian monsoon. The second CEOF (CEOF-2) explains 11% of the variance and characterizes the Indian monsoon variability, including increased precipitation over western, central, and northern parts of India and the monsoon onset and demise over those regions. Thus, CEOF-2 is referred to as the large-scale index for the Indian monsoon system (LIMS). It is shown that LIMS’s amplitude is strongly correlated with the total June–September precipitation over India. LIMS is continuous in time and can be used to evaluate significant postmonsoon rainfall events that often affect the Indian subcontinent. Moreover, LIMS exhibits spectral variance on intraseasonal time scales that are associated with active and break phases of the monsoon during summer and enhanced rainfall in the postmonsoon period.

1. Introduction

The Indian monsoon system (IMS) is among Earth’s most documented and yet enigmatic phenomena. India, with over 1.2 billion inhabitants in 2015, relies on the regularity of the seasonal rains for the overall well-being of its population. Monitoring, predicting, and understanding mechanisms driving variations and changes in the Indian monsoon on subseasonal to multiannual time scales are of great scientific, political, and socioeconomic significance.

Numerous indices have been proposed to investigate the South Asian summer monsoon, and some of the most popular are documented in Wang and Fan (1999). Some indices are more empirical and largely based on precipitation totals, such as the all Indian summer rainfall index (AIRI). This index, defined by the seasonally averaged precipitation over all the Indian subdivisions from June to September, is perhaps the best-known measure of the Indian summer monsoon (Parthasarathy et al. 1994). Rainfall-based methods, while considered objective approaches to evaluate the regularity and intensity of the monsoon, are subject to spatial variability of rain gauge stations and rainfall estimating methods. Another disadvantage of methods based on rainfall only is that they are not very reliable to investigate monsoons in global climate models given the large biases and uncertainties often associated with simulated rainfall patterns (e.g., Dai 2006).

Methods that do not directly rely on precipitation have been proposed to evaluate the Indian monsoon. For instance, Webster and Yang (1992) proposed an index (WY index) to investigate the interannual variability of the intensity of the South Asian summer monsoon based on variations of time-mean zonal wind shear between 850 and 200 hPa. The rationale for this index is that changes in daily wind shear are directly related to the large-scale rainfall over the Indian region during the monsoon season, including persistently wet (“active”) and persistently dry (“break”) phases (Prasad and Hayashi 2007). It has been debated, however, that the WY index and AIRI are often incongruent (Ailikun and Yasunari 1998) and that the former index is strongly related to the El Niño–Southern Oscillation (ENSO) variability (Goswami et al. 1999).

Goswami et al. (1999) argues that the variability of precipitation on seasonal-to-interannual time scales that effectively represents the Indian monsoon covers a large area that encompasses the Indian continent, the northern Bay of Bengal, and parts of southern China [termed extended Indian monsoon rainfall (EIMR)]. These last two regions are not considered in AIRI. Furthermore, according to that study, the interannual variability of the monsoonal circulation is primarily driven by variations in the regional Hadley circulation, which depends on gradients of diabatic heating resulting from variations of precipitation in the EIMR domain. Based on this mechanism, Goswami et al. (1999) proposed an index that describes the monsoon–Hadley circulation and is defined as the meridional wind shear anomaly between 850 and 200 hPa averaged over the EIMR domain. Although the interannual variations of the AIRI and EIMR are similar, there are years of notable exceptions. By recognizing the existence of two nearly independent major precipitation centers in South Asia, one located near the Bay of Bengal and the other in the vicinity of the Philippines, Wang and Fan (1999) identified important mechanisms explaining the differences between the Indian summer monsoon and the Southeast Asian summer monsoon and found the sources of the major discrepancies among AIRI, the EIMR index, and the WY index.

Another relevant issue for agriculture in India is the onset of the monsoon. The Indian monsoon is a distinctive phase in the evolution of the continental-scale Asian monsoon that accompanies significant seasonal variations in large-scale atmospheric and oceanic circulations in the Indo-Pacific region (e.g., Rao 1976). However, the definition of onset and demise of the monsoon is not unique (Rao 1976; Janowiak and Xie 2003) and several mechanisms have been suggested to explain the interannual variability of these dates (Ananthakrishnan 1988; Murakami and Nakazawa 1985; Yanai et al. 1992; Soman and Kumar 1993; Lau et al. 1998; Wu and Zhang 1998; Hsu et al. 1999). Consequently, numerous indices, approaches, and theories have been proposed to estimate the monsoon cycle (e.g., Rao 1976; Yanai et al. 1992; Krishnamurti and Ramanathan 1982; Fasullo and Webster 2003; Janowiak and Xie 2003; Taniguchi and Koike 2006; Pai and Rajeevan 2007; Misra and DiNapoli 2014).

Rao (1976) identifies the southwest monsoon when southeasterly winds dominate the low-level circulation, temperature decreases, and precipitation increases. The onset of the monsoon is subjectively identified when the departure from the climatological mean of one of these variables is observed for a sufficient long period of time. Rao (1976) recognizes that the onset of the monsoon in the interior of the peninsula may not immediately be seen as a remarkable increase of rainfall. Yanai et al. (1992) argues that the monsoon onset over Asia results from the interaction process between the Tibetan Plateau–induced circulation and the circulation associated with the northward migration of the rain belt of precipitation. Krishnamurti and Ramanathan (1982) emphasize the role of the energy exchange from the irrotational to the nondivergent component in the evolution of the monsoon. Pai and Rajeevan (2007) recognize that during the monsoon onset, major changes are observed in the atmospheric wind flow at all levels and that there is noticeable acceleration of cross-equatorial flow across the Somali coast and westerly zonal flow over the equatorial Indian Ocean. Misra and DiNapoli (2014) identify the strong relationships between the strong June–August Somali jet, which affects the Indian monsoon, and the causes of earlier-than-normal onset of the Southeast Asian monsoon.

Janowiak and Xie (2003) defines the onset of wet season of all monsoon regions as the first occurrence of four consecutive pentads during which the individual pentad precipitation accumulation in three of the four exceeds 33% of the climatological rainy season mean precipitation accumulation. Fasullo and Webster (2003) replaced rainfall, which is poorly measured and modeled, with vertically integrated moisture transport to diagnose the onset and withdrawal of the monsoon. However, while conceptually simple and easy to apply to reanalyses and climate model data as well as efficient in identifying the transitions in large-scale monsoon circulation, their index does not capture the synoptic variability and spatial complexity of the monsoon transitions, eventually yielding false onsets when compared to methods normally used by the India Meteorological Department (IMD). Taniguchi and Koike (2006) argue that the rapid enhancement of the low-level wind speed is more adequate to characterize the abrupt beginning of the Indian monsoon than the vertically integrated moisture transport.

The Indian monsoon cycle is also strongly affected by the boreal intraseasonal oscillations (ISO) that often cause active and/or break periods (Krishnamurti and Ardanuy 1980; Lawrence and Webster 2001, 2002; Kripalani et al. 2004; Joseph et al. 2008; Hoyos and Webster 2007). The Madden–Julian oscillation (MJO; Madden and Julian 1994) is considered the most relevant mode of intraseasonal variability in the tropics with known impacts on the Indian summer monsoon (Knutson and Weickmann 1987; Jones et al. 2004; Pai et al. 2011; Joseph et al. 2008; Suhas et al. 2013). Depending on the timing and intensity of these events, they can cause serious economic problems for India. For instance, dry periods in the early phases of the monsoon cycle can seriously affect rain-fed farm systems. Conversely, persistently wet conditions occurring mainly at the end of the wet season when the soil is already saturated can increase the probability of floods and landslides, among other problems. Thus, diagnosing and predicting these events are of primary importance for the nation.

Furthermore, the summer monsoon is not the only period of major concern in continental India. Significant rainfall may occur in the three months following the end of the Indian monsoon (normally from October to December). These events typically affect the northeastern, eastern, and southern parts of India, with substantial impacts to agriculture and water management. The India Meteorological Department classifies this period as the postmonsoon season, also known as the retreating southwest monsoon season or the northeast monsoon season (e.g., Dhar and Rakhecha 1983; Parthasarathy et al. 1994; Singh and Sontakke 1999, and references therein). The postmonsoon season, particularly between October and November, is also considered a transition period between the southwest summer monsoon (from June to September) and the northeast winter monsoon (from December to May). A thorough analysis on the large-scale interannual variability and spatial variation of the wet areas over India during the postmonsoon season is described in Singh and Sontakke (1999).

In spite of all efforts in diagnosing the Indian monsoon variability and change, including its onset and demise, a number of lacunas still remain and need to be addressed to properly describe and potentially predict the characteristics of this system on intraseasonal-to-interannual and longer time scales. The objective of this study is to propose a simple and yet robust method to investigate large-scale variations and changes in the IMS that accounts for fluctuations in the amplitude, onset, withdrawal, and length of the wet season, as well as the frequently disrupting wet and dry spells. Furthermore, because significant rainfall affecting large areas of India occasionally occurs after the official end of the monsoon, this study also focuses on diagnosing these postmonsoon events on intraseasonal-to-interannual time scales. The conceptual basis of this method follows the recommendations of Wang and Fan (1999) regarding the spatial specificities of the Asian monsoon cycle and uniqueness of the IMS. Similar methodology has been successfully applied to investigate large-scale variations and changes on multiple time scales in the South American monsoon system using the large-scale index for the South American monsoon (LISAM; da Silva and de Carvalho 2007; Carvalho et al. 2011a,b; Jones and Carvalho 2013). This new method offers an alternative framework to investigate the IMS that is continuous in time, can be applied to reanalyses and climate model data, and can serve as a metric to evaluate variations and changes in the IMS in the present, past, and future climate scenarios. The analysis performed here, however, is based on reanalysis and satellite-derived precipitation and investigates the climate of recent decades.

This study is organized as follows. Precipitation and reanalysis data are presented in section 2. The methodology to calculate and validate the index and identify the onset and demise, amplitude, and intraseasonal-to-interannual variability of the Indian summer monsoon is discussed in section 3. The characterization of the postmonsoon season according to the index is investigated in section 4. Sensitivity tests are discussed in section 5, after definitions and applications. Final conclusions are presented in section 6.

2. Datasets

This study uses 35 years (1979–2013) of daily data from the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR) at 1° spatial resolution (Saha et al. 2010). LISAM (da Silva and de Carvalho 2007), which inspired the present study, uses all variables at 850 hPa. The high-elevation terrain over northern India intersects the 850-hPa surface, especially where precipitation associated with the Indian monsoon is most significant. Thus, to realistically represent the seasonal cycle of the Indian monsoon, the large-scale index discussed here is obtained with the following variables: precipitation Pr, specific humidity q2m and temperature T2m at 2 m, and zonal and meridional winds at 10 m (U10m and V10m, respectively). Other variables such as precipitable water Pw and zonal and meridional winds at 850 hPa (U850 and V850, respectively) are used to investigate mechanisms of variability on intraseasonal time scales.

We use 13 years (1998–2010) of Tropical Rainfall Measuring Mission (TRMM) 3B42V7 data to perform an independent analysis of precipitation over Asia. TRMM 3B42 is a multisatellite precipitation dataset that combines TRMM 2B31 and TRMM 2A12 data products, Special Sensor Microwave Imager, Advanced Microwave Scanning Radiometer, Advanced Microwave Sounding Unit, and Climate Prediction Center infrared observations to estimate precipitation over low to midlatitudes (Huffman et al. 2007). TRMM provides near-global 0.25° resolution precipitation estimates at three-hourly intervals for the period 1998–2014 (Bookhagen and Burbank 2010). Validations are performed with total daily precipitation (mm day−1). Total seasonal precipitation data over India (June–September 1979–2010) available from the India Meteorological Department (Ministry of Earth Sciences, Government of India) are also examined in this study.

3. The large-scale index for the Indian monsoon system (LIMS)

a. Definition

The onset of the southwest Indian monsoon [referred to as simply the Indian summer monsoon (ISM)] is typically characterized by southwesterly winds, decreased temperature relative to the warm months of April and May, and increased rainfall. According to Rao (1976) the dates of onset (withdrawal) can be identified with reference to the sharp increase (decrease) of the 5-day means of rainfall and the changes in the circulation. The name southwest monsoon is used both for the phenomena of rains and southwesterly surface winds and the period during which they occur (Rao 1976). Presently, IMD defines the onset and progress of the monsoon over the country by diagnosing changes in rainfall from stations as well as changes in the wind field and outgoing longwave radiation (OLR). The official onset date is established when the monsoon starts in Kerala. The progressive withdrawal of the monsoon is identified based on criteria that consider variations in precipitation and circulation and incorporate a number of empirical rules (http://www.imd.gov.in/). Given all these definitions, the general consensus is that the date of the onset and withdrawal of the Indian monsoon cannot be uniquely fixed. The large-scale perspective of the ISM presented in this study reflects some of these distinctive aspects. More specifically, this study is concerned with the southwest Indian monsoon when precipitation is well established over western, central, and northern India, which is the most densely populated region in the country (census of India: http://censusindia.gov.in/).

The large-scale monsoon index is obtained by performing combined empirical orthogonal function analysis (CEOF) of the above-mentioned variables (Pr, q2m, U10m, V10m, and T2m) and includes all dates from 1 January 1979 to 31 December 2013. The long-term mean is removed from all variables before the calculation of the CEOF. The domain for which the CEOF is performed extends from 5° to 35°N and 65° to 85°E. Sensitivity tests regarding variables used in the CEOF, domain extent, and resolution are discussed in section 5.

The first CEOF (CEOF-1) explains about 40%, whereas the second CEOF (CEOF-2) explains about 11% of the total variance, and both modes are statistically independent from each other according to North et al. (1982). The respective projections of CEOF-1 onto Pr, U10m, V10m, q2m, and T2m in the domain where the CEOF is calculated are shown as Pearson correlations between CEOF-1 time coefficient and each variable (Fig. 1). Figure 2a further explores the relationships between CEOF-1 and the continental-scale precipitation and shows correlations between the CEOF-1 time coefficient and daily TRMM precipitation (1998–2010) in a broader domain. CEOF-1 characterizes the seasonal intensification of the southwesterly winds over southwest Asia (Figs. 1b,c), which transport large amounts of moisture over the continent (Fig. 1d) during the warm season (Fig. 1e). The monsoonal precipitation is observed over parts of India (Fig. 1a), Bangladesh, Myanmar, Nepal, and southern China (Fig. 2a). Thus, CEOF-1 represents a continental-scale Asian monsoon that runs from the eastern coast of India through the Indochina land bridge, east of 105°E, commonly associated with the Indian monsoon (Goswami et al. 1999; Wang et al. 2003).

Fig. 1.

Correlations between CEOF-1 and seasonal anomalies of (a) Pr, (b) U10m, (c) V10m, (d) q2m, and (e) T2m. Shading indicates statistically significant correlations (at 5% significance level). The effective sample size was estimated considering the serial autocorrelation (Wilks 1995).

Fig. 1.

Correlations between CEOF-1 and seasonal anomalies of (a) Pr, (b) U10m, (c) V10m, (d) q2m, and (e) T2m. Shading indicates statistically significant correlations (at 5% significance level). The effective sample size was estimated considering the serial autocorrelation (Wilks 1995).

Fig. 2.

Correlations between TRMM and (a) CEOF-1 and (b) CEOF-2. Only statistically significant correlations (at 5% significance level) are shown.

Fig. 2.

Correlations between TRMM and (a) CEOF-1 and (b) CEOF-2. Only statistically significant correlations (at 5% significance level) are shown.

The CEOF-2 spatial projections onto Pr, U10m,V10m, q2m, and T2m are shown in Fig. 3 as Pearson correlations between CEOF-2 and these variables. High positive correlations between CEOF-2 and Pr (Figs. 2b and 3a) are observed over the west coast of India (particularly over the state of Maharashtra, where Mumbai is located) extending over central and northern India across the states of Madhya Pradesh, Himachal Pradesh, Uttarakhand, Uttar Pradesh, Bihar, Jharkhand, and West Bengal and thus affecting India’s most densely populated cities. These seasonal variations in precipitation are observed along with the intensification of westerly winds across central and southern India (Fig. 3b) and southerly winds over northwestern and northern India (Fig. 3c). These changes in zonal and meridional circulation characterize the southwesterly circulations associated with the establishment of the southwest monsoon. Accordingly, the buildup of low-level specific humidity (Fig. 3d) is observed over continental India and is closely related to the enhancement of precipitation. Decreasing temperatures south of 25°N (Fig. 3e) accompany changes in circulation that, along with the enhancement of moisture and precipitation, characterize the southwest Indian monsoon as described in Rao (1976). Notice that the pattern of correlation between CEOF-2 and T2m (Fig. 3e) considerably differs from the respective pattern observed for CEOF-1 (Fig. 1e) over the same region; that is, CEOF-1 is associated with positive correlation with T2m over all of India (Fig. 1e), whereas CEOF-2 shows negative correlations in all latitudes equatorward of 25°N (Fig. 3e). The increase in low thick clouds and resulting cloud forcing on albedo during ISM exert an important control on temperature (Hartmann et al.1992) and can explain the pattern of negative correlations with T2m observed with CEOF-2 over southwestern India. Figure 2b further exemplifies the differences between the two modes with respect to daily precipitation. CEOF-2 represents a phase of the Asian monsoon during which precipitation dominates over India compared to other areas. However, the separation between the two orthogonal modes CEOF-1 and CEOF-2 does not imply that both modes are climatically unrelated. The ISM, which is well captured by CEOF-2, is in fact a subset of the continental-scale Asian monsoon, which is in turn identified by CEOF-1.

Fig. 3.

As in Fig. 1, but for CEOF-2.

Fig. 3.

As in Fig. 1, but for CEOF-2.

Figure 4a exemplifies the relationships between the two time coefficients CEOF-1 and CEOF-2 from 1984 to 1995. A relatively short period is displayed in Fig. 4a for the sake of clarity. Several interesting aspects emerge from the behavior of the time coefficients. The seasonal cycles of CEOF-1 and CEOF-2 evidence the association of these modes with monsoon regimes such that positive and negative values of these modes are related to the wet and dry phases of the monsoon, respectively. Moreover, CEOF-2 clearly exhibits a much sharper transition from dry (negative) to wet (positive) phases compared to CEOF-1, which is considered a typical characteristic of the ISM (e.g., Rao 1976; Wang 2006; Taniguchi and Koike 2006). Additionally, CEOF-2 lags CEOF-1 by about 56 days (standard deviation of approximately 7 days), indicating that the continental-scale monsoon identified with CEOF-1 begins early in the season (on average from mid-April to early May), likely in phase with the enhancement of the large-scale continent–ocean differential heating. Positive correlations between CEOF-1 and T2m over the entire domain (Fig. 1e) are consistent with this hypothesis. The transition from the wet to dry (positive to negative) phase occurs, on average, about 5 days later for CEOF-1 than CEOF-2 (standard deviation of 12 days). Another important difference between the two time coefficients is that the seasonal cycle of CEOF-1 is essentially represented by a single harmonic, whereas the respective CEOF-2 cycle is not. A semiannual cycle in the time coefficient reveals the complex transition of the India monsoonal cycle from wet to dry conditions between October and December, also known as the postmonsoon period. The annual and semiannual cycles in the CEOF-2 time coefficient are clearly evident in the wavelet power spectrum (Fig. 4b). High (low) amplitudes of the seasonal cycle can be identified as an increase (decrease) in the wavelet power spectrum. For instance, the amplitude of CEOF-2 is remarkably low in 1987 (high in 1988 and 1990), resulting in low (high) spectral variances at the frequency associated with the annual cycle in these years. When the annual and semiannual cycles are removed from CEOF-2, the increase in the wavelet spectral variance on interannual time scales, which are typically associated with ENSO, becomes evident (Fig. 4c). Additionally, CEOF-2 exhibits significant variations on synoptic-to-subseasonal time scales (Figs. 4b,c). The differences in the wavelet spectrum observed at time scales near the annual cycle (Fig. 4c) likely result from the “leakage of the energy” (in this case toward the strong annual and semiannual cycles) commonly associated with discrete wavelet transforms (Peng et al. 2009). The wavelet power spectrum of CEOF-2 was obtained using the Morlet wavelet base function [see Torrence and Compo (1998) for details on the wavelet transform]. We removed the annual and semiannual cycles before performing other analyses with LIMS using Fourier filters.

Fig. 4.

(a) Example of the temporal variation of CEOF-1 (blue) and CEOF-2 (LIMS; red); (b) wavelet power spectrum of LIMS obtained with Morlet wavelet base function. The period associated with the annual and semiannual cycles are indicated for reference. (c) Wavelet power spectrum of LIMS after removing the annual and semiannual cycles. Red contours show statistically significant values at 5% level. Shading represents values above one standard deviation [see Torrence and Compo (1998) for details about statistical significance].

Fig. 4.

(a) Example of the temporal variation of CEOF-1 (blue) and CEOF-2 (LIMS; red); (b) wavelet power spectrum of LIMS obtained with Morlet wavelet base function. The period associated with the annual and semiannual cycles are indicated for reference. (c) Wavelet power spectrum of LIMS after removing the annual and semiannual cycles. Red contours show statistically significant values at 5% level. Shading represents values above one standard deviation [see Torrence and Compo (1998) for details about statistical significance].

These results indicate that CEOF-2 identifies the most important large-scale features of the ISM when the monsoonal circulation and precipitation are well established over western, central, and northern India. Moreover, the coefficient captures the postmonsoon rains (as it will be shown in detail in section 4) and the transition to the dry period. Given the representation of the entire cycle of the Indian monsoon, CEOF-2 time coefficient will be referred to as the large-scale index for the Indian monsoon system (LIMS) and will be the main focus of this study.

b. Onset, withdrawal, and amplitude of the ISM

In this study, the onset, withdrawal, and amplitude of the ISM are defined based on LIMS. We exemplify our methodology by arbitrarily showing the behavior of LIMS in 2011. Prior to the calculation of onset and demise dates, LIMS is smoothed with a 5-day running mean. The onset is then defined as the date when the smoothed index transitions from negative to positive values (Fig. 5). Likewise, the withdrawal date is defined when the 5-day running mean transitions from positive to negative values. The 5-day running mean criterion is consistent with the empirical definition of the ISM discussed in Rao (1976). LIMS identifies a median onset date on 11 June, upper quartile on 15 June, and lower quartile on 6 June. The median onset agrees with the climatological onset date for the active ISM over west-central-northern India established by the IMD (see also Janowiak and Xie 2003). The LIMS median withdrawal date is on 6 October, upper quartile on 12 October, and lower quartile on 27 September. Likewise, the median demise date agrees fairly well with the mean withdrawal dates for east-central India. According to the IMD in 2011 the onset of the monsoon over western and central India occurred between 4 and 11 June, and the withdrawal from northern India was approximately on 13 October. Our methodology identified the onset on 6 June and withdrawal on 9 October.

Fig. 5.

Example of LIMS seasonal variation during 2011. The onset and withdrawal dates for the 2011 ISM are indicated in the figure.

Fig. 5.

Example of LIMS seasonal variation during 2011. The onset and withdrawal dates for the 2011 ISM are indicated in the figure.

Figure 6 illustrates the interannual variation of the onset and withdrawal dates (top) and duration (bottom) of the ISM. The onset and withdrawal (or demise) dates determined according to Fasullo and Webster (2003) from 1979 to 2000 (FW) are included for comparison as FW essentially describes variations in circulation and moisture flux and may correlate well with LIMS. Correlations between LIMS and FW onset, demise, and duration are approximately equal to 0.37, 0.06, and 0.30, respectively. LIMS clearly shows later withdrawal dates and thus longer ISM seasons compared to FW. The differences between the two methods are expected, and they are largely attributed to the smoothing of the synoptic and intraseasonal signals in FW, which are identified and accounted for using LIMS. This mechanism seems particularly important to account for fluctuations in the demise dates, likely caused by ISO (e.g., Pai et al. 2011). According to LIMS, the median duration of the monsoon is equal to 115 days and the upper quartile and lower quartile are equal to 126 and 106 days, respectively. The interannual variation of the ISM duration is shown in Fig. 6 (bottom) along with the indication of strong and moderate El Niño (EN) and La Niña (LN) years (according to the oceanic Niño index; http://ggweather.com/enso/oni.htm). Although ENSO does not completely explain the interannual variability of ISM duration, with exception of the 1997/98 strong event, there is a clear relationship between short ISM and EN events, and this relationship seems consistent throughout the three decades.

Fig. 6.

Interannual variability of the ISM (top) onset and withdrawal dates and (bottom) duration according to LIMS. Dates for the onset, demise, and duration according to Fasullo and Webster (2003) (FW-onset, FW-demise, and FW-duration, respectively) are included for comparison. Moderate and strong EN and LN episodes are indicated in (bottom). LIMS onset dates exclude false onsets. Dashed single (double) lines indicate LIMS (FW) median duration.

Fig. 6.

Interannual variability of the ISM (top) onset and withdrawal dates and (bottom) duration according to LIMS. Dates for the onset, demise, and duration according to Fasullo and Webster (2003) (FW-onset, FW-demise, and FW-duration, respectively) are included for comparison. Moderate and strong EN and LN episodes are indicated in (bottom). LIMS onset dates exclude false onsets. Dashed single (double) lines indicate LIMS (FW) median duration.

To objectively investigate the interannual variability of the ISM, this study defines monsoon amplitude as the integral of the positive values of the smoothed LIMS from the onset to the demise dates. Thus, the LIMS amplitude is a proxy for the intensity of the monsoon over the Indian subcontinent, which, according to the index, is represented by the covariability of circulation, precipitation, temperature, and specific humidity. The relationship between LIMS amplitude and precipitation over India is investigated using the total seasonal (June–September) precipitation provided by the India Meteorological Department (see section 2) and is shown in Fig. 7. To be comparable, precipitation and LIMS amplitudes are represented as the departure from the mean (ΔLIMS% and ΔPrec%, respectively). The correlation between LIMS and total seasonal (June–September) precipitation in India is ~0.86 and is statistically significant at the 1% significance level. Figure 7 (top) shows the interannual variation in ΔLIMS% and ΔPrec% between 1979 and 2010 (period when ΔPrec% is available as of the date of this study). The correspondence between the two indices and the ability of LIMS to detect extremely dry and wet ISM conditions are evident (Fig. 7, top). Years with moderate and strong EN events are also indicated in Fig. 7 (top), reinforcing that EN impacts the total seasonal precipitation (and therefore LIMS amplitude) by controlling the duration of the monsoon season (Fig. 6, bottom). Interestingly, the 1997/98 episode does not seem to have affected the total precipitation over India as observed for other strong EN years such as the 1982 or 1987 events. According to the records of precipitation from the India Meteorological Department, the total rainfall during the rainy season (June–September) in 1997 was slightly above normal (+2.2% departure from the climatological mean). We emphasize, however, that the amplitude is calculated between the onset and demise according to LIMS, which can extend beyond (or be shorter than) the typical June–September season. Early and/or late onsets and demises, which are mostly related to the occurrence of ISO early or late in the season, modify the length of the wet season in a few days, or even weeks, and can weaken the correlations between LIMS and June–September precipitation over India.

Fig. 7.

(top) Interannual variation and (bottom) scatterplot of LIMS amplitude (departure from the mean; %) and June–September total precipitation over India (departure from the mean; %). Years with significant drought and flood events [according to the IMD (http://www.imd.gov.in/)] and moderate-to-strong EN years are also indicated in (top). Linear regression between the Indian June–September precipitation (dependent variable) and LIMS amplitude (independent variable) is shown in (bottom).

Fig. 7.

(top) Interannual variation and (bottom) scatterplot of LIMS amplitude (departure from the mean; %) and June–September total precipitation over India (departure from the mean; %). Years with significant drought and flood events [according to the IMD (http://www.imd.gov.in/)] and moderate-to-strong EN years are also indicated in (top). Linear regression between the Indian June–September precipitation (dependent variable) and LIMS amplitude (independent variable) is shown in (bottom).

The scatterplot between ΔLIMS% and ΔPrec% (Fig. 7, bottom) shows the linear relationship between the two indices and the strong association between negative and positive departures of ΔLIMS% and dry and wet conditions over India, respectively. It is worth mentioning that the correlation between ΔLIMS% and ΔPrec% is larger for precipitation calculated over all of India rather than in any other subregions identified by the India Meteorological Department. Correlation between ΔLIMS% and ΔPrec% over northwestern India, central India, southern peninsular India, and northeastern India during June–September is approximately equal to 0.74, 0.72, 0.54, and 0.13, respectively. The total June–September precipitation over India correlates well with precipitation over northwestern India (~0.84), central India (~0.81), and southern peninsular India (~0.63) but poorly with northeastern India (~0.21).

c. Intraseasonal variability and active and break phases of ISM

LIMS exhibits statistically significant variation on synoptic to intraseasonal time scales, as indicated by its wavelet power spectrum (Fig. 4b). During the summer, these variations are strongly related to active and break phases of the ISM. To demonstrate these relationships LIMS was bandpass filtered to retain only intraseasonal variations between 10 and 70 days (LIMS10–70). The range of the intraseasonal band is slightly shifted toward higher frequencies compared to previous studies, which usually consider periods between 25 and 80 days to be associated with active and break phases (Waliser et al. 2003; Lawrence and Webster 2002; Hoyos and Webster 2007). This was done to account for the 50% response of the filter at the cutoff limits of the band and to avoid aliasing from the strong semiannual cycle associated with LIMS.

Figure 8 shows lag composites of bandpass-filtered (10–70 days) intraseasonal anomalies of Pw and 850-hPa circulation when LIMS10–70 is above its 75th (defined as the Indian monsoon active phase) and below its 25th (defined as the Indian monsoon break phase) percentiles. Lag composites 10 days prior to and 10 days after the active (break) phases are shown in the left (right) column of Fig. 8. Active and break phases exhibit nearly symmetric spatial characteristics for all lags, consistent with Singh et al. (2015). These events are clearly related to the northward propagation of convection accompanied by changes in circulation that are typically associated with boreal ISO (Krishnamurti and Ardanuy 1980; Lawrence and Webster 2001, 2002; Kripalani et al. 2004; Joseph et al. 2008; Hoyos and Webster 2007). Cyclonic (anticyclonic) intraseasonal anomalies begin to enhance (suppress) convective activity over southern India 10 days prior to the peak of wet (dry) regimes (Figs. 8a,b). At lag 0, Pw increases (decreases) over India and anomalies above (below) 5 mm (−5 mm) are observed in parts of central and northern India (Figs. 8c,d). These anomalies are observed along with westerly (easterly) anomalies over large portions of central and southern India and easterly (westerly) anomalies over northern India, along the Himalayan arc. The westerly (easterly) anomalies over central India during active (break) phases of the ISM have been documented in previous studies (e.g., Hoyos and Webster 2007). Anomalous anticyclonic (cyclonic) circulation and easterly (westerly) winds dominate over most of the Indian subcontinent 10 days after the peak of the active (break) phases, forcing suppression (enhancement) of the monsoonal convection, as indicated by the negative (positive) Pw anomalies (Figs. 8e,f).

Fig. 8.

Lag composites of (left) active and (right) break phases of the ISM of Pw and 850-hPa winds. All variables are 10–70-days bandpass filtered: (a),(b) 10 days prior (lag −10 days), (c),(d) on the date (lag 0), and (e),(f) 10 days after the event (lag +10 days). Active (break) phases were defined when LIMS10–70 index is above (below) its 75th (25th) percentiles. Gray area shows the Himalayas and the Tibetan Plateau.

Fig. 8.

Lag composites of (left) active and (right) break phases of the ISM of Pw and 850-hPa winds. All variables are 10–70-days bandpass filtered: (a),(b) 10 days prior (lag −10 days), (c),(d) on the date (lag 0), and (e),(f) 10 days after the event (lag +10 days). Active (break) phases were defined when LIMS10–70 index is above (below) its 75th (25th) percentiles. Gray area shows the Himalayas and the Tibetan Plateau.

The contrasting patterns of precipitation during wet and dry phases of ISM obtained with TRMM (1998–2012) are shown in Fig. 9. In the wet regime (active ISM phase) precipitation enhances over western and northern India and the foothills of the Himalayas and over the Bay of Bengal (Fig. 9a) and weakens equatorward (Fig. 9c). Conversely, ISM break periods are characterized by dry conditions over most of the Indian subcontinent but enhanced convection over the central-eastern Himalayan foothills and northeastern parts of India (Fig. 9b) and in the equatorial Indian Ocean (Rajeevan et al. 2010). The east–west dichotomy in the Himalayan foothills precipitation pattern in contrasting phases of the IMS (Fig. 9c) has been identified in Vellore et al. (2014). The magnitude of the anomalies depends on the percentiles of LIMS10–70 adopted in the composites; that is, the larger the percentile, the stronger the magnitude of the anomalies.

Fig. 9.

Composites of TRMM precipitation (1998–2012) during the (a) wet (active) monsoon phase, (b) dry (break) monsoon phase, and (c) the difference between the two composites (mm).

Fig. 9.

Composites of TRMM precipitation (1998–2012) during the (a) wet (active) monsoon phase, (b) dry (break) monsoon phase, and (c) the difference between the two composites (mm).

Intraseasonal variations in the ISM early in the season are of great concern for farmers that rely on the onset of the monsoon to decide the timing of planting. The failure of the monsoon most strongly impacts small farms and regions without irrigation. Thus, detecting and monitoring long periods with dry conditions early in the wet season are useful to investigate climate variations and changes in ISM. False onsets can be identified with LIMS during seasons when the smoothed index becomes steadily positive for a few days (false onset) and then returns to negative values (dry gaps). Such persistent negative values, which can last for more than one week soon after the onset, have been observed during the period of analysis and indicate that the monsoon failed to self-maintain. Figure 10 identifies the occurrence of these events as the duration of the false onset (consecutive wet days after the onset) and subsequent dry gaps (consecutive days in dry conditions after the onset). According to this methodology, false onsets (determined according to the 5-day moving average—see section 3b) were observed in 8 years, between 1979 and 2013. Although not all interannual variability of ISM is due to the failure of the monsoon early in the season, most years with false onsets were associated with precipitation deficits over India. The duration of false onsets and subsequent breaks varied between 5 and 19 and 3 and 19 days, respectively. Even though some of these years coincided with the start of a moderate-to-strong El Niño (1982, 1987, and 1992), ENSO alone clearly does not explain the interannual variability of these events. It is worth noting that the number of false onsets depends on the length of the moving average. Shorter lengths of the moving averages result in more false onsets, and vice versa. The criterion adopted here was chosen to be consistent with the method discussed in Rao (1976) to determine the onset of ISM. Interestingly, if false onset dates are considered as the actual onset dates, correlations between the onsets obtained with the FW index and LIMS (Fig. 6, top) increase to 0.84. LIMS identifies synoptic-to-intraseasonal variability that is smoothed in the FW index.

Fig. 10.

Interannual variability of false onsets identified according to LIMS. Striped bars show the persistence of wet days after the start of the false onset; dark bars indicate the number of subsequent (and consecutive) dry days before the actual monsoon starts. Only years with false onsets are displayed (1979–2013 ISM seasons).

Fig. 10.

Interannual variability of false onsets identified according to LIMS. Striped bars show the persistence of wet days after the start of the false onset; dark bars indicate the number of subsequent (and consecutive) dry days before the actual monsoon starts. Only years with false onsets are displayed (1979–2013 ISM seasons).

To further exemplify these relationships and demonstrate the associations between intraseasonal variability and ISM amplitude, as well as to illustrate the importance of false onsets, this study examines in detail four extreme seasons: 1982 and 1987, which were related to extremely dry conditions (e.g., Krishnamurti et al. 1989; Kumar et al. 2006), and 1994 and 2013, which were associated with excessive rainfall. Figure 11 shows LIMS and LIMS10–70 during these seasons and also indicates the LIMS climatological and actual onset and withdrawal dates (see section 3b). According to LIMS, one peculiar characteristic of both the 1982 and 1987 monsoon seasons was their failed onset (Fig. 11); that is, seasonal changes in circulation, moisture, and precipitation would have occurred in approximate agreement with the LIMS climatology, but the monsoon failed early in the season, retrograding to conditions that are typical of the premonsoon season. As discussed before, these breaks in the ISM are characterized by negative values of LIMS and LIMS10–70. According to these indices, in 1982 the false onset occurred approximately at the expected climatological date of onset and monsoonal conditions were observed for about 13 days. Subsequently, negative LIMS values were observed for about 13 days after the false onset (dry gaps), indicating that the monsoon failed to self-maintain.

Fig. 11.

Examples of LIMS (black) and LIMS10–70 (red) indices in (top) very dry (1982 and 1987) and (bottom) very wet (1994 and 2013) ISM years. Dotted–dashed lines indicate the climatological and dark dashed lines the actual onset and withdrawal dates. Thick dark blue line shows LIMS climatology.

Fig. 11.

Examples of LIMS (black) and LIMS10–70 (red) indices in (top) very dry (1982 and 1987) and (bottom) very wet (1994 and 2013) ISM years. Dotted–dashed lines indicate the climatological and dark dashed lines the actual onset and withdrawal dates. Thick dark blue line shows LIMS climatology.

Figure 12 shows a composite of the intraseasonal anomalies (10–70 days) of Pw and 850-hPa circulation during the 13-day period of the false onset (Fig. 12a) and illustrates the enhancement of monsoonal conditions (i.e., westerly anomalies and increase in Pw) over India during 12–24 June and the subsequent weakening of the ISM (~25 June–5 July) before the true onset of the monsoon was established according to LIMS (Fig. 12b). Precipitation and specific humidity were also consistent with Pw during this period (not shown), corroborating the false establishment of the monsoon. After the dry period, the 1982 ISM recovered to nearly normal conditions before undergoing an early withdrawal, well characterized by the negative anomalies in LIMS10–70 that also contributed to the negative departure (~−14.5%) in the seasonal precipitation. The 1982 ISM season was likely affected by the onset of a strong El Niño event.

Fig. 12.

Intraseasonal (10–70 days) anomalies of precipitable water during (a) false onset of 12–24 June 1982, (b) dry gap of 25 June–6 July 1982, (c) false onset of 10–15 June 1987, and (d) dry gap of 16–29 June 1987.

Fig. 12.

Intraseasonal (10–70 days) anomalies of precipitable water during (a) false onset of 12–24 June 1982, (b) dry gap of 25 June–6 July 1982, (c) false onset of 10–15 June 1987, and (d) dry gap of 16–29 June 1987.

Likewise, the 1987 drought, one of the most dramatic on record (−19.4% departure in precipitation over India; Parthasarathy et al. 1994), was also characterized by a failed onset according to LIMS likely associated with intraseasonal oscillations. Monsoonal conditions lasted for about one week followed by 15 days of break, as indicated by the fluctuation of LIMS and LIMS10–70 indices (Fig. 11, top right). The false onset (wet period) was shorter than in 1982 (10–15 June, according to LIMS) but with strong impacts over India (Fig. 12c). The dry gap was relatively longer (~16–29 June) than the wet period (Fig. 12d), causing dramatic impacts in extensive areas of India. Later, the monsoon intensified, but the underlying basic atmospheric state, likely driven by the onset of the 1987 strong El Niño, was such that the ISM was weak for most of the season. An early withdrawal (about 20 days earlier than the LIMS climatology) was also observed in 1987, resulting in the fourth-largest negative precipitation departures over continental India (−19.4%) and the largest over northwestern India (−43.9%) between 1901 and 2010 according to the IMD.

In contrast, the 1994 (Fig. 11, bottom left) and 2013 (Fig. 11, bottom right) monsoons were very active throughout the season, as indicated by the strong positive values of LIMS and LIMS10–70 indices and the high amplitude of the ISM (Fig. 7). Although the ISM exhibited some signal of weakening after the onset of the monsoon in both years, LIMS remains positive despite small negative oscillations of LIMS10–70, indicating that ISM did not retrograde to premonsoon conditions, nor did it experience significant breaking periods as observed in the 1982 and 1987 seasons. The 1994 ISM was associated with precipitation departure from the mean of about +12.5% over India (Fig. 7). According to Guan and Yamagata (2003), a persistent anomalous anticyclonic circulation over Japan, Korea, and eastern and northeastern parts of China caused an abnormally hot and dry summer in 1994 in these regions and very wet conditions west of the Bay of Bengal. The 2013 monsoon season was also very peculiar. It exhibited a pronounced transition to strong ISM conditions early in the season (Fig. 11, bottom right), as indicated by high amplitudes of LIMS and LIMS10–70 indices. This intraseasonal event in early June of 2013 had a particularly significant impact in the north Indian state of Uttarakhand, a sacred and touristic region of India. Between 16 and 17 June, when the LIMS index was greater than its 99th percentile, torrential rains caused flash floods that triggered landslides and were responsible for the deaths of thousands of pilgrims and tourists (Kala 2014). LIMS also indicates that the 2013 summer monsoon was longer than average, which could be explained by an intraseasonal disturbance that occurred at the end of the ISM cycle. These four examples illustrate the practicality of LIMS in identifying the characteristics of ISM and relationships with precipitation and its extremes over India.

4. The postmonsoon season

As discussed in the introduction, during the Indian postmonsoon season (IPM), intense convective activity, largely organized into mesoscale convective systems, is observed over the Maritime Continent, which is regarded as the main heat source for monsoonal circulation on both sides of the equator (Chang and Lau 1980, 1982; Compo et al.1999; Chang et al. 2005). These events may increase moisture and precipitation and modify circulation over east-central and northern India with patterns of anomalies that project onto CEOF-2 (Fig. 3), thus causing LIMS to exhibit positive values. These conditions are responsible for the secondary maximum in the LIMS wavelet power spectrum (Fig. 4b).

The IPM is identified with LIMS as positive amplitudes in the index after the end of the wet monsoon phase (Fig. 4a). Since these events are characterized with LIMS, they will be hereafter referred to as IPMLIMS. A comparison between LIMS and LIMS10–70 indicates that the IPMLIMS is intermittent and largely explained by intraseasonal variability and, therefore, can be identified when LIMS exhibits positive anomalies after the monsoon withdrawal, as observed in Fig. 11 for the years previously discussed. Thus, the interannual variability of IPMLIMS can be described by the count of independent events associated with positive values of LIMS after the end of the monsoon (Fig. 13). This analysis indicates that approximately 75% of these IPMLIMS events occur between October and December, with a peak in December (~30%), as previously characterized by the India Meteorological Department (Fig. 13a). Only about 8% of these events (between 1979/1980 and 2012/2013 winter seasons) occur in February and no event is observed from March to August. Using TRMM (from 1998 to 2010) precipitation data, we show that daily precipitation averaged over all IPMLIMS events corresponds to a large fraction of the total seasonal precipitation between October and December (Fig. 13c) over central and northern parts of India and Nepal, which include high elevations of the central Himalayas (Fig. 13d). In some areas, this contribution can be as high as 2.5–3 times the average October–December mean daily precipitation. The persistence of these events also varies considerably (Fig. 13b). The median persistence is 11 days, the lower quartile 6 days, and the upper quartile 25 days, with the frequency distribution skewed toward high values (Fig. 13b). Notice, however, that persistence was calculated using a smoothed 5-day moving average of LIMS and may vary if another length of the moving window is considered in the analysis.

Fig. 13.

Winter postmonsoon events based on LIMS: (a) monthly frequency (%), (b) persistence of events (days), (c) daily rainfall averaged during October–December based on TRMM (mm day−1), and (d) fraction of the mean daily rainfall (October–December) corresponding to the postmonsoon events (%).

Fig. 13.

Winter postmonsoon events based on LIMS: (a) monthly frequency (%), (b) persistence of events (days), (c) daily rainfall averaged during October–December based on TRMM (mm day−1), and (d) fraction of the mean daily rainfall (October–December) corresponding to the postmonsoon events (%).

Convective activity over India in the postmonsoon season depends on several mechanisms, including the organization of meso- to synoptic-scale systems (Yang et al. 2015), the MJO, the development of tropical depressions and cyclones, and the interactions among these systems (Duvel 2015). Tropical cyclones are not uncommon over the Bay of Bengal between October and December, and these events are normally associated with excessive rainfall over southeastern India (e.g., Kumari et al. 2014). The MJO influence on precipitation regimes over India during the boreal summer has been well documented (e.g., Wang 2006). However, the peak in the frequency of the MJO is observed in the boreal winter (Jones et al. 2004), and, therefore, the oscillation can potentially increase IPMLIMS events during this season. Moreover, Duvel (2015) showed that the MJO modulates the number of tropical depression initiations in the south Indian Ocean, and these systems often landfall in different parts of India. Some of these systems may increase moisture and modify regional circulation for a few days with patterns that resemble monsoonal conditions, thus leading to positive amplitudes of IPMLIMS.

Here we examine the relationship between IPMLIMS and the MJO using the MJO index described in Jones and Carvalho (2011a). This index is derived from CEOF analysis of bandpass-filtered anomalies (20–200 days) of daily averaged zonal wind at 850 and 200 hPa, which describe the MJO-related circulation, and OLR (Liebmann and Smith 1996), which characterizes the MJO convective activity. The wide 20–200-day band is applied after removing the seasonal cycle and adequately represents the MJO because it minimizes the problem of successive MJO events being artificially merged into each other (Matthews 2000). Details about the criteria to calculate the index and applications can be obtained in Jones and Carvalho (2011b, 2012). This method is comparable to Matthews (2000). Here the MJO index is used to identify days when the oscillation was active, regardless of its phase.

The interannual variability in the number of days with positive IPMLIMS (after the withdrawal of the monsoon) and the frequency (percent) of simultaneous days classified as active MJO according to the MJO index are shown in Fig. 14. The total number of tropical cyclones (any category), depressions, and named storms that made landfall in India and were recognizable with IPMLIMS based on the date of the event and the magnitude of the index are also indicated.1 Interesting properties and characteristics of IPMLIMS events emerge from Fig. 14. For instance, the multiannual variability in the IPMLIMS is noticeable and is indicated by the decrease in the number of days with positive IPMLIMS in the early twenty-first century, or more specifically between 2000 and 2008 (Fig. 14a). Moderate to strong warm (striped bars) and cold (checkered bars) ENSO phases are also shown in Fig. 14a suggesting no systematic relationship between this mode and IPMLIMS in these decades. The correspondence between active IPMLIMS winters and days with active MJO is shown in Fig. 14b. The MJO was active in more than 80% of all IPMLIMS active days in 10 winters (~30% of all seasons) and more than 65% of the days in 14 winters (~41% of all seasons). Tropical storms, depressions, and cyclones that affected India and were linked to positive IPMLIMS occurred in 76% of all seasons, and some of these systems may have been associated with the MJO (Duvel 2015). In some seasons, however, they seemed to be the dominant factor affecting IPMLIMS. For instance, in 2010 the high amplitude of LIMSIPM was mostly explained by Cyclone Jal, which made landfall over southeastern India in early November 2010 and claimed 54 lives (Kumari et al. 2014). The highest amplitudes of IPMLIMS during 2010 were observed concomitantly with the landfall of the cyclone and persisted until several days after the event. In 1995 two very severe cyclonic storms affected different parts of India and Bangladesh during the month of November, but most of the IPM variability was not explained by the MJO or the tropical cyclone activity. All these events projected well onto LIMS as they were associated with a sudden increase in moisture, precipitation, and cyclonic circulation across the Indian subcontinent that, to some extent, resembled monsoonal conditions.

Fig. 14.

(a) Interannual variability of the number of days observed with positive LIMS in the postmonsoon season. Striped (checkered) bars indicate warm (cold) moderate and strong ENSO phases. Numbers indicate the total number of tropical disturbance (depressions, cyclones, and severe and very severe tropical storms) that affected India and occurred simultaneously with positive LIMS in the postmonsoon season. Named storms that affected India and are identified with LIMS are also indicated (https://en.wikipedia.org/wiki/North_Indian_Ocean_tropical_cyclone); (b) percentage of days with positive LIMS in the postmonsoon season (Fig. 13a) associated with MJO active days.

Fig. 14.

(a) Interannual variability of the number of days observed with positive LIMS in the postmonsoon season. Striped (checkered) bars indicate warm (cold) moderate and strong ENSO phases. Numbers indicate the total number of tropical disturbance (depressions, cyclones, and severe and very severe tropical storms) that affected India and occurred simultaneously with positive LIMS in the postmonsoon season. Named storms that affected India and are identified with LIMS are also indicated (https://en.wikipedia.org/wiki/North_Indian_Ocean_tropical_cyclone); (b) percentage of days with positive LIMS in the postmonsoon season (Fig. 13a) associated with MJO active days.

5. Sensitivity tests

Several sensitivity tests were performed to identify the best method to adequately represent the Indian monsoon with the multivariate LIMS index. The focus of these analyses was on the following aspects: 1) CEOF domain, 2) type of reanalysis (and resolution), 3) inclusion of upper-air variables from CFSR, 4) exclusion of Pr from the analysis, 5) replacement of q2m and low-level circulation (U10m and V10m) by vertically integrated moisture flux (using CFSR), and 6) use of a single variable (Pr) in the EOF analysis.

CEOF analysis, like any other EOF analysis, is domain sensitive. The first set of tests was performed by progressively increasing the domain using CFSR. The best domain was decided based on the most realistic onset and withdrawal dates provided by the method. As expected, large domains weaken the signal of the Indian monsoon and emphasize the continental-scale Asian monsoon resulting in longer summer monsoon seasons. On the other hand, very small domains result in very few grid points to perform CEOF analysis, which is also undesirable. The chosen domain of 5°–35°N, 65°–85°E produced the most realistic ISM onsets and withdrawal dates.

A second test was performed by calculating CEOF with the same set of variables as before, using a similar domain but with the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (2.5° × 2.5° spatial resolution; Kalnay et al. 1996). The first two CEOF modes identified the large-scale Asian monsoon (CEOF-1) and the Indian monsoon (CEOF-2) systems, and the latter showed high spectral variance on the annual and semiannual cycles and on synoptic to intraseasonal time scales. However, the time coefficients indicated unrealistic early onsets and thus ISM with longer duration. This result suggested that coarse spatial resolution of the NCEP–NCAR reanalysis does not adequately resolve the peculiarities of the ISM as a subset of a much larger Asian monsoon system. Another factor could be the relatively poor representation of precipitation in the NCEP–NCAR reanalysis. More studies are necessary to properly address these issues.

A third test using CFSR included 200-hPa zonal winds (in addition to the other variables used to define LIMS) to account for seasonal changes in the upper-level jet and zonal wind shear as suggested in Webster and Yang (1992). Likewise, this procedure resulted in much earlier onsets of the ISM. Seasonal changes in upper-level circulation seem to lead seasonal changes in specific humidity and precipitation, which can explain these differences.

A fourth test was performed by removing Pr from the analysis. Pr is usually subject to biases and large uncertainty in reanalyses and climate models (Saha et al. 2010; Kalnay et al. 1996). The first two CEOF modes performed without Pr are very similar, which is encouraging, but without precipitation, the onset of the monsoon is earlier than expected based on the climatology. Even considering the reanalysis precipitation biases, the fact that precipitation lags seasonal changes in circulation improves the identification of the onset. However, an index calculated without including Pr can still be useful to diagnose intraseasonal variability once the monsoon is already established.

A fifth test was performed by replacing q2m, U10m, and V10m with integrated moisture flux and maintaining Pr as input variables in the CEOF analysis. Again, this method resulted in early onsets, which can be explained by the dominant role of the winds in the magnitude of the moisture flux. Seasonal changes in circulation in mid- and upper levels of the atmosphere seem to lead seasonal changes in rainfall. The set of surface-level variables (q2m, T2m, U10m, and V10m) exhibits seasonal changes that are more consonant with precipitation and are thus recommended for the construction of the LIMS. Taniguchi and Koike (2006) pointed out that moisture transport represents the abrupt transition of atmospheric conditions associated with ISM but often fails to adequately identify the onset of the monsoon rainfall.

Last, another test was performed by calculating the EOF analysis with CFSR precipitation (Pr) as a single variable (same EOF domain). The first EOF explains about 17% of the total variance and exhibits a pattern of correlation with TRMM (used as an independent variable) that is similar to CEOF-1. The second EOF explains about 7% of the total variance and shows a pattern of correlation with TRMM that considerably differs from LIMS and exhibits high correlation with precipitation south of India in the Arabian Sea. Neither EOF-1 nor EOF-2 captures the postmonsoon events, but EOF-1 exhibits high spectral variance on intraseasonal time scales.

6. Conclusions

This study proposes a new index designed to describe and quantify complex large-scale characteristics and temporal variations of the ISM. The main purpose of this study is to provide a framework that is simple and yet robust and can be applied to reanalyses and global climate models to investigate variations and changes in the ISM on intraseasonal to interannual time scales. Another major purpose of LIMS is to allow a temporally continuous investigation of IMS, including postmonsoon activity. This approach’s premise, based on secular observations, assumes that monsoon regimes can be well characterized by the spatiotemporal covariability in precipitation, circulation, specific humidity, and temperature at low levels of the atmosphere over India. The combined strong seasonal cycles of these five variables are well captured by LIMS and are the basis to identify the onset and withdrawal of the monsoon and its intraseasonal-to-interannual variability over the Indian subcontinent, The median onset (demise), determined when a 5-day moving average of LIMS becomes progressively positive (negative), is consistent with the observed onset (demise) of the monsoon over central-northwestern India. Therefore, LIMS identifies the ISM when it is well established over subcontinental India.

This study demonstrates that the monsoon amplitude, which can be interpreted as the combined increase in seasonal amplitudes of all above-mentioned variables, can be inferred by the integral of the positive values of LIMS during the wet season. This metric compares very well (correlation of 0.86) with the observed total precipitation over India from June to September and indicates that this simple approach is valuable and robust in characterizing interannual-to-decadal variations in the ISM. The residual difference can be, to a large extent, due to the interannual variability in the date of onset and withdrawal of the monsoon and implications for the computation of the total amplitude. These results demonstrate that LIMS is useful to identify climate variations and changes in the IMS using reanalyses and global climate models.

Another advantage of LIMS is its spectral variance on subseasonal time scales. Extreme wet and dry conditions during the ISM have been associated with intraseasonal variations. Depending on the timing of these events, they can influence the onset and demise of the monsoon with great impact to agriculture. Moreover, these events have been largely associated with major water-related disasters in the country. We show that LIMS can be used to identify significant intraseasonal events not only during the summer monsoon but also in the postmonsoon season. Examples are shown to validate these hypotheses. Additionally, this versatile and temporally continuous index helps to identify relationships with other indices and atmospheric coupled modes, such as the MJO and ENSO, among many others not explored in this study.

Other potential application of LIMS includes the forecast of ISM onset and demise as well as active and break phases of the monsoon. Jones et al. (2012) successfully applied LISAM (da Silva and de Carvalho 2007), which is conceptually similar to LIMS, to investigate the forecast skill of the South American monsoon in the National Centers for Environmental Prediction (NCEP) Climate Forecast System version 2 (CFSv2) reforecasts. Similar studies could be performed for the Indian monsoon using LIMS. Additionally, LIMS could be applied in real time using forecasts from global models [e.g., NCEP Global Forecast System (GFS)].

This study demonstrates that all phenomena whose large-scale dynamics project well onto monsoon-like features, even during the dry season, will be well characterized by LIMS. Here a number of applications are presented, but we expect that LIMS can be tested in other studies and for multiple purposes.

Acknowledgments

The authors thank Dr. Bodo Bookhagen for numerous fruitful discussions and the anonymous reviewers for thoughtful suggestions. This study was supported by the Climate and Large-Scale Dynamics Program from the National Science Foundation (AGS-1116105 and AGS-1053294). F. Cannon thanks the support of NASA Headquarters under the NASA Earth and Space Science Fellowship Program (13-EARTH13F-26). CFSR was provided by the Research Data Archive at the National Center for Atmospheric Research, Computational and Information Systems Laboratory, Boulder, CO (available online at http://rda.ucar.edu/datasets/ds093.0). The availability of precipitation data from the India Meteorological Department is much appreciated.

REFERENCES

REFERENCES
Ailikun
,
B.
, and
T.
Yasunari
,
1998
: On the two indices of Asian summer monsoon variability and their implications. Extended Abstracts, Int. Conf. on Monsoon and Hydrologic Cycle, Kyongju, South Korea, Korean Meteorological Society, 222–224.
Ananthakrishnan
,
R.
,
1988
:
The onset of the southwest monsoon over Kerala: 1901–1980
.
Int. J. Climatol.
,
8
,
283
296
, doi:.
Bookhagen
,
B.
, and
D. W.
Burbank
,
2010
:
Toward a complete Himalayan hydrological budget: Spatiotemporal distribution of snowmelt and rainfall and their impact on river discharge
.
J. Geophys. Res.
,
115
,
F03019
, doi:.
Carvalho
,
L. M. V.
,
C.
Jones
,
A. E.
Silva
,
B.
Liebmann
, and
P. L.
Silva Dias
,
2011a
:
The South American monsoon system and the 1970s climate transition
.
Int. J. Climatol.
,
31
,
1248
1256
, doi:.
Carvalho
,
L. M. V.
,
A. E.
Silva
,
C.
Jones
,
B.
Liebmann
,
P. L.
Silva Dias
, and
H. R.
Rocha
,
2011b
:
Moisture transport and intraseasonal variability in the South America monsoon system
.
Climate Dyn.
,
36
,
1865
1880
, doi:.
Chang
,
C. P.
, and
K. M.
Lau
,
1980
:
Northeasterly cold surges and near-equatorial disturbances over winter MONEX area during December 1974. Part II: Planetary-scale aspects
.
Mon. Wea. Rev.
,
108
,
298
312
, doi:.
Chang
,
C. P.
, and
K. M.
Lau
,
1982
:
Short-term planetary-scale interactions over the tropics and midlatitudes during northern winter. Part I: Contrasts between active and inactive periods
.
Mon. Wea. Rev.
,
110
,
933
946
, doi:.
Chang
,
C. P.
,
Z.
Wang
,
J.
McBride
, and
C. H.
Liu
,
2005
:
Annual cycle of Southeast Asia–Maritime Continent rainfall and the asymmetric monsoon transition
.
J. Climate
,
18
,
287
301
, doi:.
Compo
,
G. P.
,
G. N.
Kiladis
, and
P. J.
Webster
,
1999
:
The horizontal and vertical structure of East Asia winter monsoon pressure surges
.
Quart. J. Roy. Meteor. Soc.
,
125
,
29
54
, doi:.
Dai
,
A.
,
2006
:
Precipitation characteristics in eighteen coupled climate models
.
J. Climate
,
19
,
4605
4630
, doi:.
da Silva
,
A. E.
, and
L. M. V.
de Carvalho
,
2007
:
Large-scale index for South America monsoon (LISAM)
.
Atmos. Sci. Lett.
,
8
,
51
57
, doi:.
Dhar
,
O. N.
, and
P. R.
Rakhecha
,
1983
:
Foreshadowing northeast monsoon rainfall over Tamil Nadu, India
.
Mon. Wea. Rev.
,
111
,
109
112
, doi:.
Duvel
,
J. P.
,
2015
:
Initiation and intensification of tropical depressions over the southern Indian Ocean: Influence of the MJO
.
Mon. Wea. Rev.
,
143
,
2170
2191
, doi:.
Fasullo
,
J.
, and
P. J.
Webster
,
2003
:
A hydrological definition of Indian monsoon onset and withdrawal
.
J. Climate
,
16
,
3200
3211
, doi:.
Goswami
,
B. N.
,
V.
Krishnamurthy
, and
H.
Annmalai
,
1999
:
A broad‐scale circulation index for the interannual variability of the Indian summer monsoon
.
Quart. J. Roy. Meteor. Soc.
,
125
,
611
633
, doi:.
Guan
,
Z.
, and
T.
Yamagata
,
2003
:
The unusual summer of 1994 in East Asia: IOD teleconnections
.
Geophys. Res. Lett.
,
30
,
1544
, doi:.
Hartmann
,
D. L.
,
M. E.
Ockert-Bell
, and
M. L.
Michelsen
,
1992
:
The effect of cloud type on Earth’s energy balance: Global analysis
.
J. Climate
,
5
,
1281
1304
, doi:.
Hoyos
,
C. D.
, and
P. J.
Webster
,
2007
:
The role of intraseasonal variability in the nature of Asian monsoon precipitation
.
J. Climate
,
20
,
4402
4424
, doi:.
Hsu
,
H. H.
,
C. T.
Terng
, and
C. T.
Chen
,
1999
:
Evolution of large-scale circulation and heating during the first transition of Asian summer monsoon
.
J. Climate
,
12
,
793
810
, doi:.
Huffman
,
G. J.
, and Coauthors
,
2007
:
The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales
.
J. Hydrometeor.
,
8
,
38
55
, doi:.
Janowiak
,
J. E.
, and
P.
Xie
,
2003
:
A global-scale examination of monsoon-related precipitation
.
J. Climate
,
16
,
4121
4133
, doi:.
Jones
,
C.
, and
L. M. V.
Carvalho
,
2011a
:
Will global warming modify the activity of the Madden–Julian oscillation?
Quart. J. Roy. Meteor. Soc.
,
137
,
544
552
, doi:.
Jones
,
C.
, and
L. M. V.
Carvalho
,
2011b
:
Stochastic simulations of the Madden–Julian oscillation activity
.
Climate Dyn.
,
36
,
229
246
, doi:.
Jones
,
C.
, and
L. M. V.
Carvalho
,
2012
:
Spatial–intensity variations in extreme precipitation in the contiguous United States and the Madden–Julian oscillation
.
J. Climate
,
25
,
4898
4913
, doi:.
Jones
,
C.
, and
L. M. V.
Carvalho
,
2013
:
Climate change in the South American monsoon system: Present climate and CMIP5 projections
.
J. Climate
,
26
,
6660
6678
, doi:.
Jones
,
C.
,
L. M. V.
Carvalho
,
R. W.
Higgins
,
D. E.
Waliser
, and
J. K. E.
Schemm
,
2004
:
Climatology of tropical intraseasonal convective anomalies: 1979–2002
.
J. Climate
,
17
,
523
539
, doi:.
Jones
,
C.
,
L. M. V.
Carvalho
, and
B.
Liebmann
,
2012
:
Forecast skill of the South American monsoon system
.
J. Climate
,
25
,
1883
1889
, doi:.
Joseph
,
S.
,
A. K.
Sahai
, and
B. N.
Goswami
,
2008
:
Eastward propagating MJO during boreal summer and Indian monsoon droughts
.
Climate Dyn.
,
32
,
1139
1153
, doi:.
Kala
,
C. P.
,
2014
:
Deluge, disaster and development in Uttarakhand Himalayan region of India: Challenges and lessons for disaster management
.
Int. J. Disaster Risk Reduct.
,
8
,
143
152
, doi:.
Kalnay
,
E.
, and Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
, doi:.
Knutson
,
T. R.
, and
K. M.
Weickmann
,
1987
:
30–60 day atmospheric oscillations: Composite life cycles of convection and circulation anomalies
.
Mon. Wea. Rev.
,
115
,
1407
1436
, doi:.
Kripalani
,
R. H.
,
A.
Kulkarni
,
S. S.
Sabade
,
J. V.
Revadekar
,
S. K.
Patwardhan
, and
J. R.
Kulkarni
,
2004
:
Intraseasonal oscillations during monsoon 2002 and 2003
.
Curr. Sci.
,
87
,
325
331
.
Krishnamurti
,
T. N.
, and
P.
Ardanuy
,
1980
:
The 10 to 20-day westward propagating mode and “breaks in the monsoons.”
Tellus
,
32A
,
15
26
, doi:.
Krishnamurti
,
T. N.
, and
Y.
Ramanathan
,
1982
:
Sensitivity of the monsoon onset to differential heating
.
J. Atmos. Sci.
,
39
,
1290
1306
, doi:.
Krishnamurti
,
T. N.
,
H. S.
Bedi
, and
M.
Subramaniam
,
1989
:
The summer monsoon of 1987
.
J. Climate
,
2
,
321
340
, doi:.
Kumar
,
K. K.
,
B.
Rajagopalan
,
M.
Hoerling
,
G.
Bates
, and
M.
Cane
,
2006
:
Unraveling the mystery of Indian monsoon failure during El Niño
.
Science
,
314
,
115
119
, doi:.
Kumari
,
N. P. A.
,
S. B.
Kumar
,
J.
Jayalakshmi
, and
K. K.
Reddy
,
2014
:
Raindrop size distribution variations in JAL and NILAM cyclones induced precipitation observed over Kadapa(14.4N, 78.82E), a tropical semi-arid region of India
.
Indian J. Radio Space Phys.
,
43
,
57
66
.
Lau
,
K. M.
,
H. T.
Wu
, and
S.
Yang
,
1998
:
Hydrologic processes associated with the first transition of the Asian summer monsoon: A pilot satellite study
.
Bull. Amer. Meteor. Soc.
,
79
,
1871
1882
, doi:.
Lawrence
,
D.
, and
P. J.
Webster
,
2001
:
Interannual variability of intraseasonal convection and the Asian monsoon
.
J. Climate
,
14
,
2910
2922
, doi:.
Lawrence
,
D.
, and
P. J.
Webster
,
2002
:
The boreal summer intraseasonal oscillation and the South Asian monsoon
.
J. Atmos. Sci.
,
59
,
1593
1606
, doi:.
Liebmann
,
B.
, and
C. A.
Smith
,
1996
:
Description of a complete (interpolated) outgoing longwave radiation dataset
.
Bull. Amer. Meteor. Soc.
,
77
,
1275
1277
.
Madden
,
R. A.
, and
P. R.
Julian
,
1994
:
Observations of the 40–50-day tropical oscillation—A review
.
Mon. Wea. Rev.
,
122
,
814
837
, doi:.
Matthews
,
A. J.
,
2000
:
Propagation mechanisms for the Madden-Julian oscillation
.
Quart. J. Roy. Meteor. Soc.
,
126
,
2637
2651
, doi:.
Misra
,
V.
, and
S.
DiNapoli
,
2014
:
The variability of the Southeast Asian summer monsoon
.
Int. J. Climatol.
,
34
,
893
901
, doi:.
Murakami
,
T.
, and
T.
Nakazawa
,
1985
:
Transition from the Southern to Northern Hemisphere summer monsoon
.
Mon. Wea. Rev.
,
113
,
1470
1486
, doi:.
North
,
G. R.
,
T. L.
Bell
,
R. F.
Cahalan
, and
F. J.
Moeng
,
1982
:
Sampling errors in the estimation of empirical orthogonal functions
.
Mon. Wea. Rev.
,
110
,
699
706
, doi:.
Pai
,
D. S.
, and
M.
Rajeevan
,
2007
: Indian summer monsoon onset: Variability and prediction. National Climate Center Research Rep. 4/2007, 27 pp.
Pai
,
D. S.
,
J.
Bhate
,
O. P.
Sreejith
, and
H. R.
Hatwar
,
2011
:
Impact of MJO on the intraseasonal variation of summer monsoon rainfall over India
.
Climate Dyn.
,
36
,
41
55
, doi:.
Parthasarathy
,
B.
,
A. A.
Munot
, and
D. R.
Kothawale
,
1994
:
All-India monthly and seasonal rainfall series: 1871–1993
.
Theor. Appl. Climatol.
,
49
,
217
224
, doi:.
Peng
,
Z. K.
,
M. R.
Jackson
,
J. A.
Rongong
,
F. L.
Chu
, and
R. M.
Parkin
,
2009
:
On the energy leakage of discrete wavelet transform
.
Mech. Syst. Signal Process.
,
23
,
330
343
, doi:.
Prasad
,
V. S.
, and
T.
Hayashi
,
2007
:
Large-scale summer monsoon rainfall over India and its relation to 850 hPa wind shear
.
Hydrol. Processes
,
21
,
1992
1996
, doi:.
Rajeevan
,
M.
,
S.
Gadgil
, and
J.
Bhate
,
2010
:
Active and break spells of the Indian summer monsoon
.
J. Earth Syst. Sci.
,
119
,
229
247
, doi:.
Rao
,
Y. P.
,
1976
: Southwest Monsoon. Meteor. Monogr.: Synoptic Meteor., No. 1, India Meteorological Department, 367 pp.
Saha
,
S.
, and Coauthors
,
2010
:
The NCEP Climate Forecast System reanalysis
.
Bull. Amer. Soc
,
91
,
1015
1057
, doi:.
Singh
,
N.
, and
N. A.
Sontakke
,
1999
:
On the variability and prediction of rainfall in the post‐monsoon season over India
.
Int. J. Climatol.
,
19
,
309
339
, doi:.
Singh
,
P.
,
V.
Vasudevan
,
J. S.
Chowdary
, and
C.
Gnanaseelan
,
2015
:
Subseasonal variations of Indian summer monsoon with special emphasis on drought and excess rainfall years
.
Int. J. Climatol.
,
35
,
570
582
, doi:.
Soman
,
M. K.
, and
K. K.
Kumar
,
1993
:
Space–time evolution of meteorological features associated with the onset of Indian summer monsoon
.
Mon. Wea. Rev.
,
121
,
1177
1194
, doi:.
Suhas
,
E.
,
J. M.
Neena
, and
B. N.
Goswami
,
2013
:
An Indian monsoon intraseasonal oscillations (MISO) index for real time monitoring and forecast verification
.
Climate Dyn.
,
40
,
2605
2616
, doi:.
Taniguchi
,
K.
, and
T.
Koike
,
2006
:
Comparison of definitions of Indian summer monsoon onset: Better representation of rapid transitions of atmospheric conditions
.
Geophys. Res. Lett.
,
33
,
L02709
, doi:.
Torrence
,
C.
, and
G. P.
Compo
,
1998
:
A practical guide to wavelet analysis
.
Bull. Amer. Meteor. Soc.
,
79
,
61
78
, doi:.
Vellore
,
R. K.
,
R.
Krishnan
,
J.
Pendharkar
,
A. D.
Choudhury
, and
T. P.
Sabin
,
2014
:
On the anomalous precipitation enhancement over the Himalayan foothills during monsoon breaks
.
Climate Dyn.
,
43
,
2009
2031
, doi:.
Waliser
,
D. E.
,
W.
Stern
,
S.
Schubert
, and
K. M.
Lau
,
2003
:
Dynamic predictability of intraseasonal variability associated with the Asian summer monsoon
.
Quart. J. Roy. Meteor. Soc.
,
129
,
2897
2925
, doi:.
Wang
,
B.
,
2006
: The Asian Monsoon. Springer, 787 pp.
Wang
,
B.
, and
Z.
Fan
,
1999
:
Choice of South Asian summer monsoon indices
.
Bull. Amer. Meteor. Soc.
,
80
,
629
638
, doi:.
Wang
,
B.
,
S. C.
Clemens
, and
P.
Liu
,
2003
:
Contrasting the Indian and East Asian monsoons: Implications on geologic timescales
.
Mar. Geol.
,
201
,
5
21
, doi:.
Webster
,
P. J.
, and
S.
Yang
,
1992
:
Monsoon and ENSO: Selectively interactive systems
.
Quart. J. Roy. Meteor. Soc.
,
118
,
877
926
, doi:.
Wilks
,
D. S.
,
1995
: Statistical Methods in the Atmospheric Sciences: An Introduction. Academic Press, 467 pp.
Wu
,
G. X.
, and
Y. S.
Zhang
,
1998
:
Tibetan Plateau forcing and the timing of the monsoon onset over South Asia and the South China Sea
.
Mon. Wea. Rev.
,
126
,
913
927
, doi:.
Yanai
,
M.
,
C.
Li
, and
Z.
Song
,
1992
:
Seasonal heating of the Tibetan Plateau and its effects on the evolution of the Asian summer monsoon
.
J. Meteor. Soc. Japan
,
70
,
319
351
.
Yang
,
X.
,
J.
Fei
,
X.
Huang
,
X.
Cheng
,
L. M. V.
Carvalho
, and
H.
He
,
2015
:
Characteristics of mesoscale convective systems over China and its vicinity using geostationary satellite FY2
.
J. Climate
,
28
,
4890
4907
, doi:.