Prediction of the Indian Monsoon
Sulochana Gadgil
4 July 2011
Traditional forecasters Nandiwallas
‘I seek the blessings of Lord Indra to
bestow on us timely and bountiful
Finance Minister Pranab Mukherjee,
budget speech in the Lok Sabha, Feb 2011
•What are we trying to predict?
• Current State of Art: How good are the
•Background about the system responsible
for the monsoon
•Challenge of improving predictions
Note that most of the rainfall occurs during JuneSeptember i.e. the summer monsoon season
• Monsoon Prediction:
• All-India average of the summer monsoon
(June-September) rainfall (ISMR)
(i) long term mean =85.24 cm;
(ii) standard deviation =10% of the mean
The Indian summer monsoon is one of the most
reliable elements of the tropical climate!
ISMR <90% : Drought
ISMR> 110%: Excess rainfall season
90%< ISMR< 110% : normal monsoon
Interannual Variation of the anomaly of ISMR
(as % of the mean); (std dev is about 10% of mean)
Monsoon of 2009 is the one of the top five most severe
droughts during 1876-2009
Variation of ISMR anomaly during 1979-2009
• Despite the monsoon being so reliable, why, is
there so much demand for predictions? The
vagaries of the monsoon have a very large
impact. How large is the impact?
Quantitative assessment of the impact of the
monsoon on GDP and agriculture:
The Indian Monsoon, GDP and Agriculture
Gadgil, Sulochana and Siddhartha Gadgil,
Economic and Political Weekly, XLI, 4887-4895.
• There is a marked asymmetry in the response to
monsoon variability, with the magnitude of the
negative impact of a drought being more than
that of the positive impact of a surplus. Unless
this situation changes, it will not be possible to
maintain the growth rate of food grain production
at an adequate level for ensuring food security.
• The most striking feature we observed is that the
impact of a severe drought on GDP has remained
between 2 to 5% for the last five decades, despite
the marked decrease in the contribution of
agriculture to economy. This is because
agriculture also has an indirect impact on GDP,
since even now the livelihood of over 60% of the
population depends on it.
• We estimated that for a drought of moderate
intensity (ISMR deficit ranging from 10% to 15%), at
current levels of the economy and production, the
impact on GDP at 2006 prices is around Rs. 50,000
crores or more and FGP deficit of around 10 million
tons in food grain production.
• Thus prediction of the interannual variation of ISMR
and particularly for the occurrence/nonoccurrence
of the extremes (i.e. droughts and excess rainfall
seasons) continues to be extremely important even
in the modern era.
How good are the predictions of the
Consider the experience of 2009.
Monthly/seasonal anomalies of all-India rainfall (as
%age of the mean)
• June:
- 47 %
• July:
• August:
• September : -21%
• Seasonal (June-September) : - 23%
• June-August: -36%
Monsoon 2009
The Indian summer
monsoon (June-Sept)
rainfall (ISMR) in 2009
was 77% of its long
period average
i.e. deficit of 23%
• Prediction with atmospheric and
coupled general circulation models
• Predictions for the rainfall over the Indian region
made in April and May made with atmospheric
and coupled models as well as multi-model
ensembles from different centres of the world are
available for the rainfall during June-August 2009
(and not the entire summer monsoon i.e. JuneSeptember) in the public domain.
• Almost all the models predicted for June-August
2009 (for which the observed all-India rainfall was
deficit by 36%) not a deficit (leave alone a severe
drought), but rather above average rainfall over
the Indian region.
• As in 2009, the atmospheric and
coupled models at the major global
centres had failed to predict deficit
rainfall during the last two droughts
i.e. last severe drought of 2002 and
also the drought of 2004.
• Why is the skill so poor? How do we
improve it?
• Improving models, data and data
• Improving models:
• Improving parameterizations which are
known to be important for tropical
systems –such as cumulus
• To improve the simulation of the major
phenomena/modes of importance to
monsoon variability and their links with
the monsoon
• Data Assimilation : Assimilation of which
facets of the ocean-atmosphere system is
• What do we understand about
the Monsoon: the system and its
responsible for the
monsoon rainfall
occurs in
association with
a planetary scale
dynamical systemthe tropical
convergence zone
(TCZ) which
stretches from
over the Indian
monsoon zone
across the Bay and
thence across the
equatorial Pacific.
28 May11
2 June 11
9 June 11
21 June 11
28 June 11
3 July 11
• Satellite derived Outgoing Longwave
Radiation (OLR) is generally used as
a proxy for rainfall in the tropics.
Regions with deep convective clouds
are characterized by low OLR.
From Srinivasan and Smith 1992
• Almost all the cloud systems that give rain over
the Indian region, are born over the surrounding
warm ocean. Hence the variability of the largescale monsoon rainfall is linked to the variability
of clouding over the equatorial Indian Ocean ,the
Arabian Sea and the Bay of Bengal.
• The variability of monsoon rainfall is also related
to the variability of the clouding over the
equatorial Pacific Ocean. The dominant signal of
the interannual variation of the coupled
atmosphere–ocean system over the Pacific is the
El Nino Southern Oscillation (ENSO)
Interanual Variation of ISMR
• Important Links:
• Linked to El Nino Southern Oscillation –
higher propensity of droughts during El
Nino and of excess monsoon rainfall in La
• Also linked to another mode discovered
recently viz. EQUINOO
Correlation of OLR with ENSO index for JJAS
Note that convection over the entire region (i.e. eq. Indian
Ocean, Indian region, Arabian Sea, Bay of Bengal) is
suppressed (enhanced) during El Nino (La Nina)
EQUINOO :OLR anomalies
July 2002
August 1994
• Equatorial Indian Ocean Oscillation (EQUINOO)*
• When the convection over WEIO (50°–70°E, 10°S–
10°N) is enhanced, convection over EEIO (90°–
110°E, 0°–10°S) is suppressed.
• Associated with this, changes occur in the sea
level pressure gradient and in the zonal
component of the surface wind over the central
equatorial Indian Ocean. Enhancement of
convection over WEIO (EEIO) and suppression
over EEIO (WEIO) is associated with negative
(positive) anomalies of this component of surface
• We call this oscillation as Equatorial Indian
Ocean Oscillation (EQUINOO).
• ----------------------------------------------------------------------• *Gadgil et. al 2003,2004, Ihara et. al 2007
 We use EQWIN an index of EQUINOO,
defined as the negative of the anomaly of
the surface zonal wind averaged over 600E900E:2.50S-2.50N, normalized by its standard
• 1negative (so that positive values of EQWIN
imply favourable for the monsoon),
Correlation of OLR with EQWIN for JJAS
EQUINOO involves convection anomalies of opposite signs over
ISMR in ENSO index-EQWIN plane
Red blobs droughts, brown: severe droughts; Blue blobs :excess;dark
blue:large excess
after Gadgil et al GRL2004
Special Years
1988, 1961: excess monsoon both ENSO
and EQUINOO favourable
: excess monsoon with EQUINOO
favourable but ENSO unfavourable
2009 (as all severe droughts) : both unfavourable
1985: drought with EQUINOO unfavourable
but ENSO favourable
1988: excess monsoon
Atmospheric GCMs run with observed SST
1994: excess monsoon
• Thus the skill of atmospheric models in simulating
droughts/excess rainfall seasons associated with
ENSO is higher than the skill in simulating extremes
associated with EQUINOO. Need to improve
simulation of the link with EQUINOO.
• This was suggested in :
Monsoon prediction – Why yet another failure?
Sulochana Gadgil*, M. Rajeevan and Ravi Nanjundiah
CURRENT SCIENCE, 2005, VOL. 88, p1389-1400
Retrospective predictions with CFS
Note : correct prediction for 1994, but terrible for 2009
1994: Strong Positive Indian Ocean Dipole (IOD) event
Reversal of SST and OLR gradients over the equatorial
Indian Ocean
Cold SST anomaly and suppressed convection over EEIO
highly nonlinear
SST Threshold:27.50 C
(Gadgil , Joseph and
Joshi: Nature 1984)
Variation of temperature profile of EEIO
Note that as
positive phase
predicted by
GFS as well as
CFS. However,
while CFS
excess ISMR,
GFS (like all
over the
Indian region
in CFS
but not
than in
• Need a good assimilation of the vertical
variation of temperature and salinity in the
equatorial Indian Ocean besides a good
model for its evolution. Only if the mixed
layer is realistically incorporated, can a
realistic simulation of the SST anomalies
and hence IOD and EQUINOO is expected.

similar documents