15 APRIL 2006 ZUBAIR AND ROPELEWSKI 1567

The Strengthening Relationship between ENSO and Northeast Monsoon Rainfall over
Sri Lanka and Southern India
LAREEF ZUBAIR AND C. F. ROPELEWSKI
International Research Institute for Climate Prediction, Palisades, New York

(Manuscript received 9 August 2004, in final form 9 August 2005)

ABSTRACT

Recently, it was reported that the relationship of the Indian southwest monsoon rainfall with El Niño–
Southern Oscillation (ENSO) has weakened since around 1980. Here, it is reported that in contrast, the
relationship between ENSO and the northeast monsoon (NEM) in south peninsular India and Sri Lanka
from October to December has not weakened. The mean circulation associated with ENSO over this region
during October to December does not show the weakening evident in the summer and indeed is modestly
intensified so as to augment convection. The intensification of the ENSO–NEM rainfall relationship is
modest and within the historical record but stands in contrast to the weakening relationship in summer. The
intensification of the circulation is consistent with the warming of surface temperatures over the tropical
Indian Ocean in recent decades. There is modestly intensified convection over the Indian Ocean, strength-
ening of the circulation associated with ENSO (Walker circulation), and enhanced rainfall during El Niño
episodes in a manner consistent with an augmented ENSO–NEM relationship.

1. Introduction The Indian Meteorological Department refers to Oc-

tober to December period as the northeast monsoon
While the boreal summer phase of the Asian mon-
(Dhar and Rakecha 1983) but this season also is re-
soon has been well investigated through studies of the
ferred to as the winter monsoon (e.g., Nageswara Rao
Indian southwest monsoon rainfall, its behavior during
1999) and postmonsoon (e.g., Singh and Sontakke
the other seasons has received less attention (Pant and
1999). In Sri Lanka, the northeast monsoon was used to
Kumar 1997). During the fall, the zone of maximum
designate the November to March period initially and
monsoon rainfall migrates to southern India, Sri Lanka,
thereafter to refer to December to February (Bamford
and the neighboring sea (Fig. 1a). The rainfall during
1922). Since January to March is relatively dry over
this period is of immense societal significance to 150
southern India and Sri Lanka (Fig. 1c), we refer to
million people as it supports the main cultivation sea-
October to December as the northeast monsoon
son known as Maha in Sri Lanka and Rabi in southern
(NEM) consistent with the designation of the Indian
India (Zubair 2002)
Meteorological Department.
The low-level wind field over South Asia is south-
While the summer monsoon rainfall over India usu-
westerly during the boreal summer. Starting in Octo-
ally decreases during El Niño events, quite the opposite
ber, it switches to northeasterly as the winter progresses
occurs in Sri Lanka and Southern India from October
(Fig. 2a). During the summer, rain falls all over South
to December (Dhar and Rakecha 1983; Rasmusson and
Asia but is diminished over (southern) peninsular India
Carpenter 1983; Ropelewski and Halpert 1987; Suppiah
and Sri Lanka. After October, the rainfall is enhanced
1997; Nageswara Rao 1999; Singh and Sontakke 1999;
over peninsular India, Sri Lanka, and the equatorial
Kripalani and Kumar 2004). Recently, a weakening
eastern Indian Ocean and is diminished to the north
of the relationship between the Indian southwest (sum-
(Fig. 2b).
mer) monsoon and the El Niño–Southern Oscillation
(ENSO) was reported (Kumar et al. 1999). To assess
whether this weakening relationship extends to
Corresponding author address: Lareef Zubair, International
Research Institute for Climate Prediction, 130 Monell Building, the NEM, we investigate the relationship between
61 Rt. 9W, Palisades, NY 10964. ENSO and the NEM rainfall in Sri Lanka and southern
E-mail: [email protected] India.

© 2006 American Meteorological Society

1568 JOURNAL OF CLIMATE VOLUME 19

FIG. 1. (a) Sri Lanka and the Indian states that are included at least partially in (southern) peninsular India.
Elevation is indicated as a gray shade with values below 500 m masked out. The Western Ghats mountain range
runs parallel to the western coast. The Eastern Ghats is a smaller and discontinuous range parallel to the eastern
coast. (b) A representation of peninsular India that includes the subdivisions of Coastal Andhra Pradesh (22),
Rayalseema (24), South Interior Karnataka (28), Coastal Karnataka (26), Kerala (29), and Tamil Nadu (25). (c)
Monthly rainfall climatologies for Sri Lanka, Tamil Nadu, peninsular India, and all India for the period of record.

2. Data 90°–150°W). 2) Velocity potential (␹) and circulation

fields were obtained from the National Centers for En-
The following monthly datasets were used in the vironmental Prediction–National Center for Atmo-
analysis: 1) Sea surface temperature (SST) reconstruc- spheric Research (NCEP–NCAR) reanalysis (Kalnay
tions from 1869 to 2003 were obtained from Kaplan et et al. 1996). 3) Rainfall data for 1871 to 2002 for Sri
al. (1998). The Niño-3 ENSO index is defined as the Lanka were obtained from its Department of Meteo-
average SST anomaly over the Pacific region (5°S–5°N, rology for 16 stations (details in Zubair et al. 2003) and
15 APRIL 2006 ZUBAIR AND ROPELEWSKI 1569

FIG. 2. (a) The mean low-level wind field at 925 mb during October to December for
1958–2002 (m s⫺1). (b) The mean monthly rainfall during October to December for 1979 to
2004 (mm month⫺1).

for subdivisions in India (Fig. 1b) were obtained from and Table 1) and we concentrate on the ENSO-NEM
the Indian Institute of Tropical Meteorology (Pant and rainfall relationship in this region. The Sri Lanka and
Rupa Kumar 1997). 4) Outgoing longwave radiation peninsular India rainfall shows correlation with the si-
(OLR) data from 1974 onward were obtained from the multaneous ENSO index of Niño-3 that is significant at
Climate Prediction Center of the National Oceanic and the 99% level from 1871 to 2002 (Table 2). There are
Atmospheric Administration. (5) Climate Anomaly sharp regional variations in the ENSO–NEM rainfall
Monitoring System–OLR Precipitation Index (CAMS– correlations in India and only in Tamil Nadu and adja-
OPI) rainfall estimates based on combination of satel- cent regions is the relationship consistent over the pe-
lite and station observations were also obtained from riod of record. The correlation of Niño-3 with Sri Lan-
the Climate Prediction Center (Janowiak and Xie kan rainfall is stronger than that with Tamil Nadu rain-
1999). fall, which in turn is stronger than that with the
peninsular India rainfall (Table 2). As the Sri Lankan
3. Analysis rainfall has the strongest NEM–ENSO relationships,
Peninsular India and Sri Lanka receive a large frac- we use it as a representative rainfall index to examine
tion of their annual rainfall during the NEM (Fig. 1c the influence of global fields. The simultaneous corre-
1570 JOURNAL OF CLIMATE VOLUME 19

TABLE 1. Annual rainfall, mean NEM rainfall overall, and that sodes to determine the probabilities expected by
during El Niño and La Niña episodes for 1958–2002. NEM rainfall chance. Here, El Niño conditions are associated with
is given within parentheses as a percent of annual rainfall.
wetter conditions, Neutral conditions with near-normal
Mean NEM NEM rainfall and La Niña with drier conditions. Where there
Annual NEM rainfall rainfall is a perfect association, the Heidke skill score yields a
Region rainfall rainfall El Niño La Niña score of 100, a perfectly negative relationship the score
Sri Lanka 1853 780 (42%) 885 718 would be ⫺50, and when there is only random associa-
Peninsular India 1170 345 (29%) 382 307 tion, the score tends to 0. The Heidke skill score for the
Kerala (29) 2784 479 (17%) 511 429 association between ENSO and Sri Lanka NEM rain-
Tamil Nadu (25) 923 458 (49%) 499 405
Coastal Karnataka (26) 3407 247 (7%) 279 235
fall during 1958–80 was 9 and during 1981–2002 was 25.
South Interior 873 208 (24%) 231 193 The latter score is somewhat higher than those typically
Karnataka (28) experienced in operational precipitation forecasts for
Rayalseema (24) 743 219 (29%) 228 204 the United States. These figures show that the predict-
Coastal Andhra 1006 383 (38%) 451 328 ability of NEM rainfall based on ENSO has certainly
Pradesh (22)
not decreased and, in fact, has been enhanced during
the latter period. The contingency table also shows that
the relationship during La Niña appears to be strength-
lation between NEM rainfall and SST brings out high
ening as much as that during El Niño. In much of the
correlations in the equatorial eastern Pacific Ocean
remaining analysis, we report on the behavior during El
reminiscent of the ENSO signature (Fig. 3a). We also
Niño episodes alone for brevity.
characterize the large-scale atmospheric circulation as-
As reported previously by Kumar et al. (1999), the
sociated with ENSO through examination of the veloc-
relationship between the Indian southwest monsoon
ity potential to provide a basis for comparison with the
rainfall and Niño-3 has dropped below the 99% signifi-
earlier work on the summer monsoon described in Ku-
cance level since 1980 (Fig. 3b). The relationship be-
mar et al. (1999).
tween Sri Lanka NEM rainfall and Niño-3 largely
The contingency table (Table 3) compares the influ-
stayed significant at the 95% level through the last 140
ence of ENSO on NEM rainfall for the early (1958–80)
yr and has surpassed the 99% significance level since
and later (1981–2002) periods. During both periods, El
1980. The Tamil Nadu NEM rainfall too maintains its
Niño conditions are associated with a tendency toward
(smaller) correlation with ENSO over the historical
wetter conditions and La Niña conditions are associ-
record increasing to 95% significance levels in recent
ated with a tendency toward drier conditions. Here we
decades. The peninsular India NEM rainfall had a weak
use the Heidke skill to assess the strength of these as-
relationship with Niño-3 for a hundred years since 1871
sociations (Wilks 1995). The Heidke score (S) is given
but this relationship has strengthened to the 95% sig-
by S ⫽ 100 (C–E ) / (N–E ), where C is the number of
nificance level in the late 1970s.
correct forecasts, E is the number of correct forecasts
The rise in ENSO–NEM correlation over peninsular
expected by chance, and N is the total number of fore-
India is derived particularly from the northern subdivi-
casts. In the computation of the Heidke score, we used
sions where there has been an increase in correlations
the observed occurrences of rainfall and ENSO epi-
since 1980 (Fig. 4). These three subdivisions coastal
Andhra Pradesh (22), Rayalseema (24), and south in-
TABLE 2. The correlations of NEM rainfall for Sri Lanka, pen- terior Karnataka (28) receive 39%, 29%, and 24% of
insular India, and subdivisional India with simultaneous Niño-3
their respective annual rainfall during the NEM. The
for the full record and for the two decades before and after 1980
are tabulated. Correlations that are significant at 90%, 95%, and low-level circulation over peninsular India has intensi-
99% are shown in italics, bold italics, and bold, respectively. fied in the easterly direction in recent decades (Fig. 3c)
and this shall lead to enhanced orographic rainfall
Region/period 1871–2002 1958–80 1981–2002 (Sarker 1966) to the windward side of the Western
Sri Lanka 0.51 0.47 0.64 Ghats mountain range (Rayalseema and south interior
All India ⫺0.09 ⫺0.29 0.30 Karnataka) and the Eastern Ghats mountain range
Peninsular India 0.23 0.30 0.44
(Coastal Andhra Pradesh). The correlation for the sub-
Kerala (29) 0.21 0.28 0.22
Tamil Nadu (25) 0.38 0.41 0.44 divisions along the western coast (Kerala and Coastal
Coastal Karnataka (26) 0.01 0.08 0.02 Karnataka) has remained at past levels (Table 2).
South Interior Karnataka (28) 0.08 0.45 0.50 These subdivisions lie to the lee of the Western Ghats
Rayalseema (24) 0.01 0.21 0.24 mountain range during the NEM and do not garner
Coastal Andhra Pradesh (22) ⫺0.06 0.33 0.37
orographically induced rainfall. These findings suggest
15 APRIL 2006 ZUBAIR AND ROPELEWSKI 1571

FIG. 3. (a) The correlation of October to December sea surface temperatures with simul-
taneous Sri Lanka rainfall for 1958–2002. (b) Sliding correlations with a 31-yr window of
Niño-3 with simultaneous Sri Lankan, Tamil Nadu, and peninsular India NEM rainfall are
shown. A similar sliding correlation of all-India southwest monsoon rainfall and simultaneous
Niño-3 is also shown with its sign changed to positive to facilitate comparison. Correlation
values at 0.36 and 0.46 are significant at 95% and 99% levels and these thresholds are shown.
(c) The difference between 1981–2002 and 1958–80 in the mean October to December low-
level circulation at the 925 mb (m s⫺1).

that changes in the monsoon circulation in recent de- monsoon season over the recent decades reflect large-
cades is strengthening the zonal component of the low- scale changes in the character of the atmospheric cir-
level flow that in conjunction with regional orography culation anomaly patterns associated with ENSO. In-
results in stronger ENSO–NEM relationships in parts deed, their analysis shows significantly different global
of peninsular India. 200-hPa velocity potential anomaly patterns in the lat-
Kumar et al. (1999) argue that the decreases in the ter versus the earlier period. In contrast, El Niño com-
ENSO–precipitation relationships during the summer posite anomalies for the October to December season
1572 JOURNAL OF CLIMATE VOLUME 19

TABLE 3. Contingency tables for the ENSO association with a show only subtle changes in the velocity potential
composite index of areally weighted Sri Lanka and Indian NEM anomaly patterns and a slightly enhanced magnitude
rainfall for (top) 1958–80 and (bottom) 1981–2002. Wet, normal,
between the earlier and later periods (Fig. 4). This is
and dry were defined based on tercile boundaries with wet being
100 mm above normal, dry being 130 mm below normal, and consistent with the relatively stable and slightly en-
normal being values in between. Normal rainfall was taken as hanced relationships between these anomaly fields and
values within half a standard deviation of the mean rainfall. El ENSO-related precipitation over peninsular India and
Niño and La Niña phases were identified as seasons where the Sri Lanka during the October to December season.
average Niño-3 values were greater than 0.5°C and less than
Note that the interpretation of the changes in the 200-
⫺0.5°C, respectively.
hPa velocity potential should be limited to large-scale
1958–80 features alone (Rasmusson et al. 1999). Examination of
Wet Normal Dry lower level vertical velocity shows unambiguously that
there is convection over peninsular India and Sri Lanka
El Niño 3 1 2
Neutral 4 3 3 during the NEM.
La Niña 1 3 3 We note, however, that part of the enhanced magni-
1981–2002
tude of the anomalies during the later period may be
due to the overall tendency for slightly stronger circu-
Wet Normal Dry
lations features in the NCEP–NCAR reanalysis (Kal-
El Niño 4 1 2 nay et al. 1996) because of the introduction of satellite
Neutral 3 4 2
data in 1979. The later period includes the two strongest
La Niña 1 2 3
ENSOs of the twentieth century, 1982/83 and 1997/98,

FIG. 4. Composites of the October to December velocity potential anomalies (␹ ⫻ 10⫺6 m2

s⫺1) at 200 mb for El Niño seasons before and after 1980 are shown as (a) before 1980: 1963,
1965, 1968, 1969, 1972, 1977, and (b) after 1980: 1982, 1986, 1987, 1991, 1994, 1997, 2002.
15 APRIL 2006 ZUBAIR AND ROPELEWSKI 1573

FIG. 5. Correlations of Sri Lanka northeast monsoon rainfall with 200-hPa velocity
potential (␹) for (a) 1958–80 and (b) 1981–2002.

which should also add to increasing the magnitude of is a small, but clear, increase in the magnitudes of the
the anomaly fields. However, these results hold even if NEM precipitation–velocity potential correlation fields
we construct composites of anomalies for the 1980 to during the latter period. These modest shifts in the cor-
2002, which leave out the 1982/83 and 1997/98 episodes. relation structure are consistent with the observed in-
Yang et al. (2002) note that there is greater uncertainty crease in the ENSO correlations with NEM rainfall.
in the NCEP–NCAR reanalysis for the period before The central tropical Indian Ocean has warmed by
1968. A repeat of our analysis for the earlier period between 0.1° and 0.2°C during the NEM period (not
restricted to 1968 to 1980 shows no large differences shown) when comparing the 1981–2002 period with that
from the results presented here. of 1958–80 period. It is not clear whether the increased
Kumar et al. (1999) found that the correlations be- sea surface temperatures are at all related to the slight
tween the all-India southwest monsoon rainfall and the increase in the ENSO–precipitation correlations during
200-hPa velocity potential was strong between 1958 and the NEM. However, the OLR shows enhanced cloud
1980 and weakened considerably and shifted southeast- cover over the region during more recent El Niño epi-
ward between 1981 and 1997. To assess whether there sodes (Fig. 6). The analysis of evidence for cloud cover
have been similar changes in the NEM, we constructed from OLR data is complicated by the fact that it is only
correlations of the Sri Lankan NEM rainfall with the available from 1974. The OLR composites during El
200-hPa velocity potential fields for two periods, 1958 Niño episodes are based on data from the early and
to 1980 and 1981 to 2002 (Fig. 5). The patterns of the later halves of the record. Note also that the composites
correlations structure over the tropical Indian Ocean of OLR during El Niño episodes during the 1981–2002
region do not show significant changes between the two period is significantly more negative over Sri Lanka and
periods and, in particular, show no evidence of the southern India than that during the only El Niño epi-
weakening in correlations that was evident for the sum- sode between 1974 and 1980 (1977).
mer monsoon in Kumar et al. (1999). In contrast, there Examination of vertical velocity fields (not shown)
1574 JOURNAL OF CLIMATE VOLUME 19

FIG. 6. The composite outgoing longwave radiation (W m⫺2) during El Niño events for the
two halves of the observational record from 1974–2002: (a) 1974–89 and (b) 1990–2002.

also shows a slight intensification of the vertical veloc- mented the strengthening of the relationship between
ities in the latter period over the region receiving high northeast monsoon rainfall in southern peninsular In-
NEM rainfall. The enhanced convection is likely to act dia and Sri Lanka after the 1980s. The ENSO–NEM
to strengthen the zonal circulation associated with rainfall correlation that had been significant to the 95%
ENSO leading to increased cloudiness and precipita- level in Sri Lanka has exceeded the 99% significance
tion over Sri Lanka and southern peninsular India dur- level in recent decades; the ENSO–NEM rainfall rela-
ing El Niño episodes. tionship in Tamil Nadu has recently attained a 95%
significance level. Even though these ENSO–NEM
4. Discussion rainfall relationships have strengthened lately, it
In contrast to reports on the weakening of the ENSO– is within the range that has been observed in the
southwest monsoon rainfall relationship, we have docu- past 130 yr. The modulation of OLR fields in the El
15 APRIL 2006 ZUBAIR AND ROPELEWSKI 1575

Niño episodes during the latter half of the available REFERENCES

record supports the evidence based on rainfall for the
Bamford, A. J., 1922: Ceylon rainfall. Quart. J. Roy. Meteor. Soc.,
slight strengthening of ENSO–NEM relationship. 48, 206–207.
If we consider peninsular India as a single entity, Dhar, O. N., and P. R. Rakecha, 1983: Foreshadowing northeast
there has been a remarkable rise in the correlation of monsoon rainfall over Tamil Nadu, India. Mon. Wea. Rev.,
the northeast monsoon rainfall and ENSO that is be- 111, 109–112.
yond the historical range. The rainfall over South Inte- Janowiak, J., and P. Xie, 1999: CAMS–OPI: A global satellite–
rior Karnataka, Rayalseema, and Coastal Andhra rain gauge merged product for real-time precipitation moni-
toring applications. J. Climate, 12, 3335–3342.
Pradesh that lie on the windward side of the Western
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
Ghats mountain range show relatively enhanced corre- analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
lation with ENSO. This regional rise may be attributed Kaplan, A., M. A. Cane, Y. Kushnir, A. C. Clement, M. B. Blu-
to a slight enhancement of the low-level circulation in menthal, and B. Rajagopalan, 1998: Analyses of global sea
recent decades during the NEM that leads to relative surface temperature: 1856–1991. J. Geophys. Res., 103,
increase in the orographic component of rainfall. 18 567–18 589.
Kumar et al. (1999) suggest that the weakening of the Kripalani, R. H., and P. Kumar, 2004: Northeast monsoon rainfall
variability over south peninsular India vis-à-vis the Indian
relationship between ENSO and summer Indian mon-
Ocean dipole mode. Int. J. Climatol., 24, 1267–1282.
soon rainfall might be ascribed, in part, to a shift in the
Kumar, K. K., B. Rajagopalan, and M. A. Cane, 1999: On the
mean position of the Walker circulation in the recent weakening relationship between the Indian Monsoon and
period compared to the earlier record. Indeed, their ENSO. Science, 284, 2156–2159.
analysis shows a dramatic shift in the structure of the Nageswara Rao, G., 1999: Variations of the SO relationship with
correlations between monsoon rainfall and velocity po- summer and winter monsoon rainfall over India: 1872–1993.
tential as well as dramatic changes in velocity potential J. Climate, 12, 3486–3495.
Pant, G. B., and K. R. Kumar, 1997: Climates of South Asia.
anomaly composites during El Niño in comparing the
Wiley, 320 pp.
two periods. Replicating their analysis for the NEM, we
Rasmusson, E. M., and T. H. Carpenter, 1983: The relationship
do not find similar evidence for large shifts in the char- between eastern equatorial Pacific sea surface temperature
acter of the correlations between monsoon rainfall and and rainfall over India and Sri Lanka. Mon. Wea. Rev., 110,
ENSO nor do we find large shifts in the character of the 354–384.
ENSO composites except for larger magnitudes in the ——, M. Chelliah, and C. F. Ropelewski, 1999: The observed cli-
anomaly fields. mate of the 20th century. Modeling the Earth’s Climate and its
Variability, W. R. Holland, S. Joussaume, and F. David, Eds.,
The sea surface temperatures over the Indian Ocean
Elsevier, 1–138.
have warmed in recent decades in the central and west- Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional
ern Indian Ocean and this is likely to modulate both the scale precipitation patterns associated with the El Niño/
monsoon circulation and the influence of ENSO. The Southern Oscillation. Mon. Wea. Rev., 115, 1606–1626.
warming in the Indian Ocean follows a pattern remi- Sarker, R. P., 1966: A dynamical model of orographic rainfall.
niscent of ENSO response of the Indian Ocean SSTs. Mon. Wea. Rev., 94, 555–572.
This warming is consistent with the intensified convec- Singh, N., and N. A. Sontakke, 1999: On the variability and pre-
diction of the rainfall in the post-monsoon season over India.
tion over the Arabian Sea extending to Sri Lanka since
Int. J. Climatol., 19, 300–309.
1980s.
Suppiah, R., 1997: Extremes of the southern oscillation phenom-
enon and the rainfall of Sri Lanka. Int. J. Climatol., 17, 87–
Acknowledgments. The IRI was established by agree- 101.
ment between the NOAA Office of Global Programs Wilks, D. S., 1995: Statistical Methods in Atmospheric Sciences.
and Columbia University. The support by a grant from Academic Press, 467 pp.
NOAA (NA03OAR4310161) is acknowledged. Use of Yang, S., K.-M. Lau, and K.-M. Kim, 2002: Variations of the East
data from the Indian Institute of Tropical Meteorology Asian jet stream and Asian–Pacific–American winter climate
anomalies. J. Climate, 15, 306–325.
and Sri Lanka Department of Meteorology, NCEP–
Zubair, L., 2002: El Niño–Southern Oscillation influences on rice
NCAR, and NOAA/CPC is also acknowledged. Com- production in Sri Lanka. Int. J. Climatol., 22, 242–250.
ments by Anji Seth and Joshua Qian were valuable. ——, S. C. Rao, and T. Yamagata, 2003: Modulation of Sri Lanka
Suggestions by the reviewers and the editors improved rainfall by the Indian Ocean Dipole. Geophys. Res. Lett., 30,
the paper considerably. 1063, doi:10.1029/2002GL015639.