Indian Summer Monsoon is a major component in the Asian monsoon system. The rainfall obtained during the southwest monsoon (SWM) season from June to September (JJAS) contributes about 70–90% of the annual rainfall in most parts of the country (Pant and Rupa Kumar, 1997) and thus plays an important role in the water resources, agricultural practices and power generation in India. The mean rainfall obtained during the season is about 852 mm with a standard deviation of 85 mm (Kulkarni et al., 2009). During the onset of winter monsoon (also known as northeast monsoon; NEM) the southwesterly winds of the SWM season changes to northeasterly winds and last for 3 months from October to December (OND). In south Asia, only the southeastern peninsular regions of India and Sri Lanka receive NEM rainfall. Average rainfall received in the season is about 312 mm, with a standard deviation of 84 mm (Kripalani and Kumar, 2004). This accounts for 50% of the annual rainfall in the east coast of India (Kumar et al., 2007). The cultivation of rice in Sri Lanka and Tamil Nadu is influenced by NEM and thus has a high significance in relation to the society (Zubair, 2002; Kumar et al., 2007).
The SWM and NEM rainfall over India exhibits large interannual variability (Nityanand and Sontakke, 1999; Singh, 2001). Recent studies by Kar et al. (2001) and Wu et al. (2009) have shown that the dominant variability of SWM rainfall is in the ENSO (2–8 years) band. Several other studies explored the influence of ENSO and its relation to the SWM rainfall (Pant and Parthasarathy, 1981; Rasmusson and Carpenter, 1983; Krishna Kumar et al., 1995; Krishnamurthy and Goswami, 2000; Rajeevan et al., 2004) and identified the decrease in SWM rainfall during El Niño events (Ropelewski and Halpert, 1987; Pant and Rupa Kumar, 1997; Nageswara Rao, 1999). Similar results for warm ENSO events were also obtained by Lau and Nath (2000) using a general circulation model. Kane (1999) found that the El Niño events active during the SWM season are associated with droughts in India. The winter rainfall variability in the tropics and subtropics was also found to exhibit a strong relationship with ENSO (Gartin-Woll, 1999). The interannual variability of NEM rainfall over India showed a strong signal in the ENSO scale with a dominant periodicity in the 4-year period (Nayagam et al., 2009). Dhar et al. (1982) examined the trends and periodicities of NEM rainfall over Tamil Nadu and the peak frequencies corresponding to the time periods of 2.0–2.44 and 3.66–4.40 years were found. Contrary to the SWM season, an increase in the NEM rainfall over the southern parts of India was noticed during El Niño events (Dhar et al., 1982; Suppiah, 1996, 1997). Kane (2000) studied the influence of ENSO on rainfall of different months and found that over peninsular India the effects were mixed with weak/strong rainfall. Revadekar and Kulkarni (2008) predicted the frequency and intensity of extreme NEM precipitation events, 4–6 months in advance using the ENSO index.
Other than the variability in the 2–8-year band, ENSO events are also found to exhibit interdecadal changes before and after 1970 (Wang, 1995). Torrence and Webster (1999) observed interdecadal variations in the frequency and amplitude of ENSO events. This decadal variability has been explained as the interaction of tropical and extratropical oceans in the mean thermocline (Gu and Philander, 1997). By introducing weather noise into a coupled model, Kirtman and Schopf (1998) could simulate the decadal variability of ENSO. Therefore, they denied the role of any external forcing. Ashok et al. (2001) noticed decadal changes in the ENSO–ISMR relationship and similar results for the ENSO–NEM rainfall are observed by Nayagam et al. (2009).
The Indian Ocean plays a significant role in the evolution of ENSO (Wu and Kirtman, 2004a; Annamalai et al., 2005; Terray et al., 2005). The sea surface temperature (SST) anomalies of Indian Ocean were found to affect ENSO through the modulation of Pacific and Indian Ocean Walker circulation (Wu and Kirtman, 2004a). Wang et al. (2003) showed that an anomalous southeast Indian Ocean (SEIO) anticyclone was found to play a key role in the evolution of Asian-Australian monsoon system during the boreal summer and fall of an El Niño in the developing phase. A strong (weak) SWM is preceded by positive (negative) SST anomalies over the southeastern subtropical Indian Ocean during boreal winter, which is related to the ENSO phenomenon (Terray et al., 2003). This boreal winter SST over the SEIO has been identified as a unique precursor to the Australian summer monsoon (AUSM), tropical Indian Ocean dipole, India Summer Monsoon (ISM) and maritime continental rainfall. Significant spectral peaks at 2 and 4–8-year time scales were found in SEIO index, similar to that of ENSO, ISM or AUSM indices (Terray et al., 2005). The significant role played by southeastern Indian Ocean in the relationship between El Niño and IOD is also discussed by Mathew et al. (2010). They have shown that the development of El Niño in the January to March (July to September) season is accompanied by the warming (cooling) of southeastern Indian Ocean, which is unfavourable (favourable) for the development of IOD. In addition to the relation of Indian Ocean SST and ENSO, the significant role played by Indian Ocean SST and SWM (Meehl and Arblaster, 2002; Wu and Kirtman, 2004b; Terray et al., 2007; Shailendra and Pandey, 2008) and NEM (De and Mukhopadhyay, 1999; Bhanu Kumar et al., 2004; Zubair and Ropelewski, 2006; Kumar et al., 2007; Nayagam et al., 2009) rainfall has also been studied.
Over India, after the climate shift of 1976, a weakening ENSO relationship with SWM rainfall (Kumar et al., 1999) and a strengthening relationship with NEM rainfall was found (Zubair and Ropelewski, 2006; Kumar et al., 2007). The strengthening is due to the positive SST anomalies over the Bay of Bengal and the Arabian Sea. These SSTs strengthen the circulation associated with ENSO (Walker circulation) and favour enhanced NEM rainfall. The collapse in the relationship between SWM and ENSO was explained by the change in the seasonality of the ENSO cycle (Kawamura et al., 2003), Indian Ocean Dipole mode(Ashok et al., 2001), Atlantic Oscillation (Chang et al., 2001; Pal, 2004) and global warming. Using a fully coupled model, Turner et al. (2007) studied the effect of increased greenhouse gases on monsoon–ENSO system and showed that the variations may not be related to climate change. Annamalai et al. (2007) observed that at a decadal time scale, the correlation between the All India rainfall and Nino3 SST varies in magnitude and an increased CO2 concentrations result in the increase of rainfall and its interannual variability. However, the ENSO–monsoon relationship was not found to weaken as the climate warms. Kitoh (2007) also has suggested that a collapse of the ENSO–monsoon relationship could occur as a consequence of internal (i.e. unforced, natural) variability in the climate system.
In the light of the present understanding that the ENSO relationship with rainfall during the SWM (NEM) season is weakening (strengthening), this study aims,
1)To study the spatio-temporal patterns of rainfall in its dominant scale of variability (scale-averaged wavelet power [SAP] of SWM [SAPSWM] and NEM [SAPNEM]), with a special emphasis on the changes that have taken place after mid-1970s. Only a few attempts have been made to understand the relation of ENSO with rainfall at a decadal scale, so the study focuses on the decadal changes of SAPSWM and SAPNEM rainfall.
2)Examine the relationship of rainfall (SAPSWM and SAPNEM) to monthly Indian Ocean SST (SAPSST).
3)To understand the ENSO–NEM rainfall relation and its link with Indian Ocean SST that is less studied. Therefore, the variability and relation of NEM rainfall were given more attention than that of SWM rainfall.
This study has made use of a high-resolution daily gridded rainfall data provided by India Meteorological Department. The data over the Indian region has a spatial resolution of 1° × 1° in latitude and longitude grid boxes and a temporal coverage of 53 years from 1951 to 2003. The details of the development of the data from 1803 stations and the quality controls applied are available in the studies of Rajeevan et al. (2005). The seasonal rainfall is the total rainfall obtained at each grid for a particular season. Thus daily rainfall was transformed to seasonal rainfall at each grid point over the study region. The SAP of JJAS rainfall over India and OND rainfall over peninsular India region south of 19.5°N was extracted, to study the variability of the dominant mode in SWM and NEM rainfall.
The NOAA extended reconstructed monthly SST was used to study the relation of rainfall with Indian Ocean SST. The spatial resolution in latitude and longitude global grid boxes is 2° × 2° and is available for the period 1854-present (Smith and Reynolds, 2004). The relationship with rainfall is studied using the SAPSST over Indian Ocean (30°N–40°S and 40°E–110°E).
3.1. Wavelet analysis
Wavelet analysis is a common tool for detecting time–frequency variations within a time series. As this analysis extracts the localized variations of power within a time series (Farge, 1992), one can determine both the dominant modes of variability and their variation in time (Torrence and Webster, 1999; Chu and Chen, 2005; Mao and Wu, 2006). The wavelet transform given by Torrence and Compo (1998) was made use in this study.
The wavelet spectrum is defined as the absolute value squared of the wavelet transform. By varying the wavelet scale and translating along the localized time index, one can construct a picture showing both the amplitude of any features versus the scale and how this amplitude varies with time. The fluctuations in power over a range of scales are averaged to obtain the SAP, which, is defined as the weighted sum of the wavelet power spectrum over the chosen scales and the wavelet spectrum averaged over time is called the Global wavelet spectrum.
3.2. Empirical orthogonal functions
Empirical orthogonal function (EOF) analysis reduces the dimensionality of a data and captures maximum variance from the dataset. This is achieved by transforming to a new set of variables, the principal components (PC), which are uncorrelated, and which are ordered so that the first few retain most of the variation present in all of the original variables (Preisendorfer, 1988; Wilks, 1995; Von Storch and Zwiers, 1999; Jollife, 2002).
EOF performed on wavelet spectra is commonly known as wavelet EOF. Here the data matrix is the SAP. The columns represent the time series of wavelet power at each grid and the rows represent the wavelet power in space. The eigenvector captures the maximum amount of variability in the data and the maximum amount of variation is explained by the first PC and the second explains the next largest amount after removal of the first, and so on. The eigenvalue associated with each eigenvector gives the amount of variance explained by each of the eigenvectors. The PC is a time series and is uncorrelated to all other PCs.
The region considered in this study for the analysis of SWM rainfall is shown in Figure 1. The shaded region demarcates the region for NEM rainfall analysis. This study is concentrated on the decadal variability of the ENSO (2–8-year) scale after 1976 as the dominant mode in the SWM and NEM rainfall after 1970s was found to be in 2–8-year period (Figure 2(a) and (b)). Therefore, the SAP of SWM and NEM (hereafter SAPSWM and SAPNEM) rainfall was extracted and the data were subjected to EOF analysis. The PCs of the dominant patterns were obtained and the spatial distribution of PCs is presented as correlation of PCs with its respective SAP. Regions of high correlations are regions of high variability explained by each component. This methodology is similar to that followed in the studies of Mwale and Gan (2005). Thus the PCs of SAPSWM and SAPNEM rainfall were correlated with the respective SAP of seasonal rainfall, in order to understand the variability explained by each component. Similarly, the SAP of monthly SST (hereafter SAPSST) from January to September was extracted and EOF analysis was performed. The PCs thus obtained were correlated with SAPSST of respective months. In order to study the relation of SAPSWM and SAPNEM rainfall with the Indian Ocean SST in the ENSO band, the leading PC of monthly SAPSST from January to May was correlated with the PC1 of SAPSWM rainfall and PC1 of SAPSST for the months from January to September were correlated with PC1 of SAPNEM rainfall. This has been done to find the SAPSST with a predictive skill. The month that has high value of correlation was selected to explain the relation between SST and rainfall. Thus SAPSST for the month of March was found to have high correlation with PC1 of SAPNEM and SAPSWM rainfall.
To find the scale that explains more variability in the ENSO (2–8-year) band, the wavelet power of rainfall was obtained at individual periods such as 2, 3, 4, 5, 6, 7 and 8 years. EOF analysis was then applied to the extracted power corresponding to these periods in order to specifically know the periods that explain more variance in the 2–8-year band. Only the first dominant pattern that explains the spatio-temporal variability to a considerable amount was used for the analysis. The correlation between the PC1 of SAPSWM rainfall for the individual periods and the SAPSWM rainfall was found. The PC1 of individual scale that obtained high correlation with the PC1 of SAPSWM rainfall was selected to explain the variability of SAPSWM rainfall. Similar analysis is extended to SAPNEM rainfall also.
This study also concentrates on the rainfall variability of the NEM season, therefore, the relation between the SAPSST (March) with NEM rainfall was studied for the periods 1951–1975 and 1977–2001. This analysis helps to quantify the effect of Indian Ocean SST on NEM rainfall, before and after the climate shift of 1976. For this analysis, SAPSST for the period 1951–1975 was also extracted and subjected to EOF analysis and the PC obtained is correlated with NEM rainfall.
4.1. SWM and NEM rainfall variability over India
The SAPSWM and SAPNEM rainfall were subjected to EOF analysis and the PCs that explain the maximum decadal variability of the ENSO scale are retained. Leading two PCs of SAPSWM and SAPNEM rainfall explain 81 and 79% of variance, respectively.
4.1.1. Spatial and temporal patterns of SWM rainfall
As the study utilizes the scale-averaged (2–8-year band) power, the variability is obtained in a scale higher than the 2–8-year periodicity. Therefore, the spatial patterns presented here represent the decadal variations of ENSO scale variability. For SAPSWM rainfall, 57.5% of the variability is explained by PC1.
The spatial distribution of correlation of PC1 of SAPSWM rainfall with the SAPSWM rainfall (Figure 3(a)) has positive relation over peninsular India and parts of central India: these regions are expected to get an increased amount of rainfall due to decadal variations of ENSO mode. Northeastern states also have positive values of correlation. A negative relation is observed over the northwest India with a correlation of − 0.6 and a maximum correlation of − 0.8 is observed over the subdivision of east Uttar Pradesh (25–27°N and 79–84°E). Negative correlation is also found in the northwest–southeast (24–29°N and 72–88°E) direction along the monsoon trough region. During the SWM season, the monsoon depressions move along the troughs and bring immense rainfall over the country. Rainfall is minimum along the mean position of trough, but when the trough is over Bihar and east Uttar Pradesh, the latent instability of the air mass enhances rainfall in the trough, even without a low (Rao, 1976). Therefore, a negative correlation over the trough region implies that the rainfall over the trough region is affected adversely. Thus the decadal variation of ENSO mode has a negative influence on SAPSWM rainfall over the monsoon trough region.
The spatial pattern of second PC (PC2 of SAPSWM rainfall) does not have an organized region of correlation (not shown) and therefore is not considered in this study. The correlation between SAPSWM rainfall and the PC1 of SAPSWM rainfall at individual scales was found to have high correlation with 2-, 3- and 4-year periods. Therefore, the power over the 2–4-year period was extracted and correlation of PC1 of SAPSWM rainfall for the 2–4-year period and SAPSWM rainfall is presented (Figure 3(b)). The PC1 of SAPSWM rainfall shows the decadal variability of the 2–4-year period. The patterns in Figure 3(a) and (b) are similar, which indicate that the decadal variation of SAPSWM rainfall is explained by the decadal variations of the 2–4-year periodicity.
The temporal pattern (PC1) of SAPSWM rainfall shows the decadal variations of the 2–8-year period. PC1 explains the variability of dry years prior to 1987, as its values are negative from 1977 to 1986. During this period the subdivisions experienced 3–4 dry years (Figure 4). Wet (dry) years are defined as years that received rainfall greater (less) than one standard deviation from long-term mean of the season. The variability of the year 1987 (dry year) was not represented by PC1. The PC1 of SAPSWM rainfall is positive from 1987 to 2000, with a maximum value in 1996. Except in 1983, the positive values of PC1 are consistent with the higher rainfall received over the region. Therefore, PC1 accounts for the decadal variability of SAPSWM rainfall over India.
4.1.2. Spatial and temporal patterns of NEM rainfall
The PC1 of SAPNEM rainfall has positive correlation over the subdivisions in the eastern parts of peninsular India. A maximum correlation of 0.7 is observed over the subdivisions of Tamil Nadu and Andhra Pradesh. A region of positive values is also seen north 13°N latitude, over Western Ghats (Figure 5(a)). The region has correlation value as high as 0.8, along 16.5°N latitude. The second PC (PC2) correlates negatively with most of the regions in south peninsula including north Kerala. PC2 explains the variance, which was left unexplained by PC1 (Figure 5(b)). The maximum correlation is noticed over the subdivision of North interior Karnataka. It is noticeable that the variance of first PC is thrice that of the second. The values obtained for the respective PCs are 60.8 and 18.2%. The correlation between the PC1 of SAPNEM rainfall at individual scales and the PC1 of SAPNEM rainfall has shown that the decadal variations of the biennial scale dominate and that corresponding to the 4-year period comes second (Figure 6(a) and (b)). As the correlation map of SAPNEM rainfall with PC1 and PC2 is spatially coherent with the correlation maps of SAPNEM rainfall at individual scales 2 and 4 years, respectively, the spatial pattern obtained (Figure 5(a) and (b)) represents the decadal variations of 2–4-year mode in NEM rainfall.
The temporal patterns of the two PCs of SAPNEM rainfall are shown in Figure 7. The subdivisions of south peninsular India experienced a decrease in rainfall with 4–5 dry years during the period 1977–1989, when the PC has a negative phase. Positive values are observed during the period 1990–1998 and the PC1 of SAPNEM rainfall has a positive peak in 1994. The positive values are consistent with the higher rainfall received over the core region of NEM rainfall. The subdivisions of Rayalaseema, Tamil Nadu, Coastal Andhra Pradesh and Kerala have received more rainfall during the period with 3–4 wet years. All the dry years (1984, 1988, 1989 and 2000) of NEM rainfall in the study period are well represented by PC1, but the wet years 1977 and 1987 are not captured by PC1, as the excess rainfall was contributed mainly by the subdivisions of Karnataka. The below normal rainfall from 1980 to 1989 and above normal from 1990 to 1999 of NEM are explained by PC1. Therefore, the decadal variability of ENSO and the variability of NEM rainfall are closely related. The PC1 is negative for both SWM and NEM season during the 1980s, whereas in the 1990s they are in the positive phase. Therefore, PC1 accounts for the dominant variability of NEM rainfall over India and explains the decadal changes of ENSO scale (biennial) variability during the period 1977–2001.
The variation of PC2 is opposite to that observed for PC1 upto 1983, after which PC1 has more amplitude than PC2. The maximum positive value of PC2 is observed in 1977 and the minimum in 1988. The positive (negative) values of PC2 during the period 1977–1983 (1984–1993) are consistent with the above (below) normal rainfall received over the subdivisions of coastal Karnataka, South interior Karnataka and North Interior Karnataka. Both the PCs could not explain the variability of 1988 (wet year) and 1995 (below normal). Other than these 2 years, the variability of all the years during the period 1977–2001 is well explained by the two PCs. Thus the two PCs explain the spatial and temporal variability of rainfall over the peninsular Indian region during the OND season. The decadal variation of PC2 but is opposite to that of the PC1 during the period 1977–1993, but have the same sign as that of PC1 from 1994. So a combined effect is obtained after 1994. Thus through EOF analysis, the spatial and temporal patterns of the decadal variability of ENSO mode in the Indian rainfall are identified.
4.2. Relationship with Indian Ocean SST
The PC1 of monthly SAPSST from January to September was obtained and its spatial distribution is presented as correlation with SAPSST for the respective months. In order to study the relation of SAPSST with SAPSWM and SAPNEM rainfall the correlation between their leading PC was found out. The SAPSST of a month that shows better relationship with SAPNEM and SAPSWM rainfall was selected. The criterion for the selection of a particular month is the highest value obtained in the correlation of PC1 of monthly SAPSST with PC1 of SAPSWM and SAPNEM rainfall. Figure 8 shows the correlation between PC1 of SAPSST of different months with the PC1 of SAPNEM and SAPSWM rainfall. Only the correlation prior to their respective monsoon season was only considered. Thus for both the SAPSWM and SAPNEM rainfall, the higher correlation is obtained with March SAPSST. Therefore, the influence of Indian Ocean is studied using SAPSST for the month of March.
The spatial distribution of correlation of PC1 of SAPSST with SAPSST (March) is presented here (Figure 9). Negative correlation with high values (−0.8) is found over north Arabian Sea and SEIO. South Arabian Sea, Bay of Bengal and south central Indian Ocean have positive correlation with maximum value of 0.6. The spatial pattern obtained here represents the decadal variability of SAPSST over Indian Ocean.
4.2.1. Relation between SWM and Indian Ocean SST
The correlation was carried out with the SAPSST (March) over Indian Ocean and the PC1 of the SAPSWM rainfall. The pattern (Figure 10(a)) shows high positive correlation (0.6) over south Arabian Sea, Bay of Bengal and negative correlation (−0.8) over SEIO. The negative correlation of north Arabian Sea and positive correlation over south central Indian Ocean have maximum correlation values of − 0.6 and 0.6, respectively. The large negative correlation over the SEIO indicates that an increase of SST over will negatively affect the rainfall over India during the SWM season. The increase in SST over the central Indian Ocean, Bay of Bengal and south Arabian Sea leads to an increase in SWM rainfall. As a large region of the Indian Ocean is having a positive correlation, an increase in SST over the adjoining Indian Ocean favours the rainfall over the country. This pattern is similar to that of Figure 9.
4.2.2. Relation between NEM and Indian Ocean SST
The correlation between the SAPSST over Indian Ocean and the PC of the SAPNEM rainfall is shown in Figure 10(b). The March SST over Indian Ocean better explains the relation between SAPSST and SAPNEM rainfall in the ENSO scale. Negative correlation is observed over southeast and southwest of Indian Ocean and also over north Arabian Sea. Positive correlation is found over central Arabian Sea, Bay of Bengal and south central Indian Ocean. Large regions of negative correlation observed over SEIO have a maximum value of − 0.8, near Australia. The strong relation between the SST over Indian Ocean and NEM rainfall in the ENSO scale is quite evident. An increase in SST over the Arabian Sea and Bay of Bengal increases the NEM rainfall. A cooling of the SEIO region has been more favourable for NEM rainfall over India in the recent decades. This relation can be identified a few months prior to the NEM rainfall.
Similar correlation pattern is obtained for both the SWM and NEM seasons (Figure 10(a) and (b), respectively) when the PC1 of SAPSWM and SAPNEM rainfall are correlated with SAPSST (March). This indicates that the PCs have similar variability and periodicity embedded in it. Therefore, spatial regions of Indian Ocean that influence the variability of SAPSWM and SAPNEM rainfall are the same, but the strength of the correlation is more for SAPSWM rainfall than for SAPNEM rainfall.
Figures 9 and 10 also give similar correlation patterns. An increase in SST over SEIO leads to a decrease in rainfall over India, whereas an increase in SST over the central Indian Ocean, Bay of Bengal and south Arabian Sea favours the rainfall of the following monsoons (SWM and NEM). Moreover, as the PCs of SAPSST (March) (figure not shown) and rainfall (SAPSWM and SAPNEM) are similar, they have same variability. The PCs have decadal variation: Figure 9 gives the decadal variations of SAPSST for the month of March. The variation in SAPSST is obtained a few months prior to the SWM and NEM seasons. So an understanding of the patterns of SAPSST over Indian Ocean can help to know the strength of the following monsoon rainfall over India. Thus the SAPSST (March) is useful to predict the following monsoon rainfall over India.
4.2.3. Wavelet analysis of SST over SEIO
As a large region over the SEIO is found to have high correlation with PC1 of SAPNEM rainfall, an index is made (85°–110°E and 5°–40°S) over the region and a wavelet analysis is performed in order to understand its dominant frequency during the period of 1951–2003. It shows that the wavelet power in the 2–4-year scale is significant at 90% level (Figure 11). In recent decades (after 1976), the maximum power is observed in the biennial scale and significant power is also noticed at 4-year period.
The variability of the SEIO SST is similar to that of NEM where, the first PC explains the biennial mode and the second explains 4-year mode of variability. Also for SWM, the dominant mode in the ENSO band is observed in the 2–4-year period. The relation of SWM with SEIO has already been discussed by Terray et al. (2003). The result obtained here clearly explains the relationship of NEM and SWM with Indian Ocean SST at a decadal scale.
4.2.4. Relation between NEM rainfall and PC of March SAPSST
As the focus of this study is also on the NEM rainfall variability, the influence of March SAPSST on the NEM rainfall is studied. The correlation between the NEM rainfall index (averaged over 6.5°–19.5°N–72°–87°E) and the PC1 of SAPSST (January to September) was performed. The maximum correlation of NEM rainfall is obtained with PC1 of SAPSST (March) (Figure 12), implying the influence of SAPSST over Indian Ocean on the NEM rainfall. This indicates that the decadal changes of ENSO over Indian Ocean are influencing the NEM rainfall or the variation of March SST over Indian Ocean in the ENSO scale is directly linked to the NEM rainfall. In order to identify whether the relation between NEM rainfall and SAPSST (March) over Indian Ocean has undergone any changes after the climate shift of 1976, the PC1 of SAPSST for the period 1951–1975 and 1977–2001 was correlated with NEM rainfall index. The results show that the magnitude of correlation has increased from 0.1510 (1951–1975) to 0.4385 (1977–2001), which is significant at 95% level. The PC1 of SAPSST (March) is correlated with NEM rainfall, in order to understand the spatial pattern over peninsular Indian region. From the correlation map (Figure 13), it is clear that prior to 1976, the PC1 of SAPSST (March) is less correlated with NEM rainfall, whereas after 1976, the relation has strengthened and high correlation observed over the subdivisions of south peninsular India that receive NEM rainfall. Therefore, after 1976, the decadal variations of Indian Ocean March SST (SAPSST) have a greater role in the NEM rainfall. As the relation is found 6 months prior to the NEM season, this result can be of use in the prediction of NEM rainfall.
5. Summary and discussion
The dominant frequencies and the spatial variability of the NEM and SWM rainfall have been studied using wavelet and EOF analyses.
1)The dominant signal in the Indian rainfall is found to be in the ENSO band (2–8 years). Therefore, the power of rainfall in the ENSO band (SAPSWM and SAPNEM) has been extracted and its spatial and temporal variability have been studied. The leading EOF of SAPSWM explains a variability of 57.5%, whereas for the SAPNEM, the first pattern covers the core rainfall region and its PC explains 60.8% of total variance. The temporal pattern (PCs) of the rainfall (SAPSWM and SAPNEM) is consistent with the above (below) normal rainfall received over the region. The individual period that explains the maximum variability in the ENSO scale for the SWM and NEM rainfall was also identified. For SAPSWM, maximum power is observed in the 2–4-year period, whereas for SAPNEM, it is in the biennial mode. As the study utilizes SAP, the decadal variability of 2–4-year period is observed in the spatial and temporal patterns of rainfall.
2)As the dominant periodicity of rainfall is found to be in the ENSO scale, monthly Indian Ocean SST is filtered and averaged in the 2–8-year period. The relation of filtered rainfall with filtered SST was studied and found that the variability of SAPSWM and SAPNEM rainfall is well represented by the March SAPSST. The PC of SAPSWM and SAPNEM is well correlated with the SAPSST over the SEIO. A warm south Arabian Sea and Bay of Bengal along with a cold SEIO favours the Indian rainfall. Wavelet analysis of the index over the SEIO helped to understand that the variability of SST is also in the 2–4-year period, as that of Indian rainfall. The study also identified that the PC1 of SAPSST for the month of March, over the Indian Ocean is well correlated with the NEM rainfall. Through wavelet analysis the dominant biennial mode of SEIO is also highlighted.
3)Finally, the influence of Indian Ocean SST in the ENSO scale with the NEM rainfall, before and after the climate shift of 1976 is studied and the relation is found to have increased after 1976. As the relation is noticed few months prior to the occurrence of NEM rainfall, this PC along with other predictors may help to improve statistical models to predict the NEM rainfall.
This study focused on the dominant variability of Indian rainfall, which is in the ENSO band. Studies on the other frequencies are also important and will help to understand the spatio-temporal pattern and to bring out the significance of the global Oceans in modulating the rainfall over India. All these results point to the strong air–sea interaction. A better understanding would be possible with numerical studies using coupled Ocean–Atmosphere models.