Indianj, phys, 79 (5), 457-471 (2005)
<
^ ^ e v l e w
U P %
V ariability o f Indian m onsoon and its rainfall forecasting
S D e T D atta ^ M D e ' and A|B B hattacharya '*
'D epartm ent o f P h y sics. U niversity o f K alyani, ISJadia, 741 2 35, West B engal. India
^Department o f P h y sics, Serampore C o lleg e, Scra|;ipore 7 1 2 201, W est B engal, India
'U n iv ersity S c ie n c e Instrum entation C:enire, U niversity o f lia ly a n i. Kalyani 741 2 35, W est B engal, India E-m ail ; abb@ klyuni>Jem ct.in
I*
R eceived 6 August 2004, acrepdjpi 18 March 2005
A bstract : The present state o f k n o w led g e on the variability o f Indian moj4soon with sp ecial em phasis on the intra-scasonal and inter-annual
»;iriaiion o f the m onsoon g iv in g im portance to 'm onsoon j e ts ’, seasonal o sc iB a to n s, annual variability and coherent intra-scasonal o sc illa tio n s are (JisciisNcd at length. D ifferent techniques and m odels im plem ented for forecasting^m onsoon rainfall have been thoroughly considered. A link b etw een '.tratosphere and m onsoon rainfall is exam ined. TTie instability characteristics o f m a r o o n disturbances arc outlined Finally, scope o f future investigations are pointed out.
K eyw ords : M o n so o n , rain fall, in stab ility characteristics P\CS No. : 9 4 .1 0 .- s
Plan of the A r t ic le
1. Introduction
I Intra-seasonal and inter-annual variation 2.1. In tra -sea so n a l * m on soon j e t s'
2.2. In tra -sea so n a l o sc illa tio n s o f m on soon 2.3. In ter-annual v a r ia b ility
2.4. C oh eren t in tra -se a so n a l o sc illa tio n s Link between stratosphere an d m onsoon rainfall I- Forecasting of m onsoon rainfall
Instability characteristics o f m onsoon disturbance
»• New forecast m odels fo r In d ian SW monsoon rainfall
^ Conclusions an d scx>pe fo r fu tu re investigations
otresponding Author
1. Introduction
M onsoon is defined as a seasonal shift in wind direction. In a true m onsoon clim ate, seasonal wind shifts typically cause a drastic change in the general precipitation and tem perature patterns. H owever, the m onsoon m ay be associated with dry w eather as well, since the 'wet' m onsoon phase o f warm , m oist air is seasonally replaced by a 'dry’ m onsoon o f cool, dry air.
This phenomenon is the dom inant feature o f low-latitude clim ates stretching from West A frica to the w estern Pacific Ocean. To u n d e rsta n d w hy th e se are m o re fa v o re d a re a s , w e n e e d know ledge o f the driving forces behind m onsoons and th e Earth's w eather in general [1,2].
The annual m onsoon cycle can be described as a result o f th e an n u al v a ria tio n o f in c o m in g so la r ra d ia tio n and th e differential heating at the surface o f land and water. T his has been recognized for hundreds o f years, as W ebster [3] noted in his discussion o f m onsoon dynam ics. Sim ply stated, sections o f the earth’s surface heat and cool at different rates depend on their ability to absorb solar radiation and the tim e o f year. B odies o f water, w hich can absorb sunlight at varying depths (and consequently reflect less back to th e atm osphere), store energy m ore efficiently than land and therefore retain heat longer than a land m ass. D ue to the shallow ness o f the absorbing surfaces.
© 2005 lA C S
4 5 8 S D e, T D a tta , M D e a n d A B B h a tta c h a ry a the land surfaces gain or lose heat at a qu ick er rate. To m aintain
an energy balance, heat is transferred from areas o f surplus to d efic it, and in the case o f a la n d -w a te r d iffe re n tia l, this is accom plished through a p h enom enon know n as the "land-sea breeze". A s the hot air rises over the land, it is replaced by the co o ler air over the water. A t night, how ever, the land cools at a q u ick er rate than the water, so the w ind shifts, blow ing from the land to the w arm er water.
To ensure sustainable agriculture in a region, know ledge o f the local clim ate is essential [4,5]. Clim atic limitations are a strong indicator o f agronom ic potential and can be used to determ ine w h ich c ro p s are best su ite d fo r a re g io n , as ra in fa ll and tem peratures are tw o m ajor variables affecting crop type and yield. So agricultural planning is esp ecially critical in m onsoon regions, w hich experience distinct w et and dry seasons.
T his paper review s the variability o f Indian m onsoon with special em phasis on the intra-seasonal and inter-annual variation o f the m onsoon giving im portance on 'm onsoon je ts', seasonal o s c illa tio n s, an n u al v a ria b ility and c o h e re n t in tra se a so n a l oscillations. D ifferent techniques and m odels im plem ented for forecasting m onsoon rainfall have been th o roughly considered.
A link betw een stratosphere and m onsoon rainfall is exam ined.
T he in stab ility c h a ra c teristic s o f m o n so o n d istu rb an ces are outlined and scopes o f future in vestigations are pointed out.
2. Intra-seasonal and in ter-a n n u al v ariations
The .seasonal m ean and its inter-annual variability arc influenced by the intra-seasonal oscillations (ISO 's) [6] o f the Indian sum m er m onsoon, as the intra-seasonal and inter-annual variations are in flu e n c e d by a co m m o n m o d e o f sp atial v a riab ility . T h e frequency o f chaotic ISO regim es determ ines the seasonal m ean monscx>n and thus sets a lim it on m onsoon predictability. A higher frequency o f active (break) conditions w ithin a m onsoon season causes a strong (w eak) su m m er m onsoon.
T h e p r e d ic ta b ility o f th e In d ia n s u m m e r m o n s o o n is d e te rm in e d on w h eth er th e in te r-a n n u a l su m m e r m o nsoon variability is dom inated by the slow ly varying external forcing o r by internal processes. I f the intra-seasonal anom alies related to the 'active' ('break') condition en h an ce (w eaken) the large- scale m o n so o n flow , a la rg er fre q u e n c y o f 'activ e' ('b reak ') co n d itio n s co u ld resu lt in a stro n g e r (w e a k e r) than norm al seaso n al m ean m o n so o n . T h e IS O 's a re c h a ra c teriz e d by a broadband spectrum w ith p erio d b etw een 10 and 90 days and are not pure sinusoidal oscillations. T h u s the duration o f 'active' an d 'b reak ' c o n d itio n s d u rin g a m o n so o n se a so n co u ld be different. T he num ber o f 'active' and 'break' episodes in a m onsoon season could also b e unequal d u e to th e ir initial phases, as th ese are low f i ^ u e n c y o scillatio n s.
T h e ISO 's an d inter-an n u al v ariab ility o f the Indian sum m er m onsoon are governed by a co m m on m ode o f spatial variability.
Ferranti e ta l [7] reported that the intra-seasonal and inter-annu variability sim ulated by a general circulation model (GriVl)ai alm ost sim ilar. G osw am i et a l [8] used daily observations surface w ind for a period o f 10 years and found that the sp un) structures o f the intra-seasonal and inter-annual variability n sim ilar. In F igure 1, the geographical distribution of the min seasonal activity and the inter-annual variability [9] arccompaicd A nnam alai e ta l (10] used N C E P/N C A R re-analysis for 17 (1979-1995) but could not find a com m on m ode o f vanabilitv which describes intra-seasonal and inter-annual vanabiluv *'i the mons(K)n. P alm er [ 11J introduced the concept oi ’nudifLcj chaos' w here the IS O s could be intrinsically 'chaotic', but tht slow ly varying forcing could 'nudge' them to reside in cnhci one o f the tw o po ssib le eq u ilib ria for a long enough time and thereby in flu en ce the seaso n al m ean. A nnam alai et al ] loi exam ined the statistics o f the IS O 's that are modulated by iht^^
slow ly varying boundary forcing (El N inos and La Ninas). Bm due to the sm all .sample size (num ber o f El N inos and La Nuu w ithin that period), the statistical significance o f th eir results w ere not very certain. A jaya e ta i [5] used daily data oi 42 yeaI^
and proposed a m ethod to separate the contribution o f the ISO\
to the seasonal m ean, from that o f the external forcing. TfK\
fo u n d th a t th e in lr a - s c a s o n a l an d in te r-a n n u a l m onsoon j intraseasonal
20S
40E 60E 60E 100E 120E 140E
F ig u r e J. G e o g r a p h ic a l d istr ib u tio n o f in tr a -se a so n a l and inter
a c tiv ity .
.inmill
( a ) M ean std d ev o f IS O -filtered relative vorticity (10' 1) al
hPa from 1 June to 3 0 S ep te m b er for 2 0 years (1978-97) (b ) Inter-annual std d iv o f s e a so n a l m ean rela tiv e vorticity
I0'« 5*1 ) leased on the sam e 2 0 years.
V ariability o f Indian m o nsoon a n d its ra in fa ll fo r e c a s tin g 4 5 9 varmhiJity is indeed governed by a com m on m ode o f spatial
variiJbility and that the ISO's m ake dom inating contribution to the intcr-annua! variability.
lk)V a given year, the 'active' ('break') days are those for which v^-ind anomalies arc greater (lesser) than the + 1 SD (-1 SD). It has tound that a sm all sh ift o f th e re fe re n c e p o in t does not significantly affect the definition o f 'active’ and 'break' days. In fact, the com posite o f daily precipitation during 'active' days closely corresponds to the canonical pattern o f daily rainfall vailability associated w ith typical 'active' and 'break' condition ot the Indian m onsoon. A typical 'active' condition is associated with strengthening (w eakening) o f the seasonal mean low-level m onsoon circu latio n , en h an cin g (w eak en in g ) the cyclonic vorticity and convective activity in the northern position o f the Tropical C onvergence Zone.
2.1 Intra-seasonal 'monsoon Jets'
Most of the year, the C lim atological winds over the equatorial
Indian Ocean is westerly. In 1973, Wyrtki [ 12] reported that in ApnI-May ('spring') and O cto b er-D ecem b er ('fall'), strong,
sustained w esterly winds generate eastw ard jets in the ocean, nim atological winds in the E qlO are related to the seasonal
north south migration o f the Tropical Convergence Zone (TCZ).
The TCZ is centered on the equator in spring and autumn, leading lo strong westerlies during the.se periods. The semi-annual cycle, the w inds in th e re g io n a lso h a v e s tro n g in tra -se a so n a l oseillaiions (ISO) (13,14). T hese are associated in sum m er with fluctuations o f the Asian sum m er monsoon [15,17) and in winter
with M adden-Julian oscillation (M JO ) ( 18J. A lthough the time resolution o f the TO PEX (Radar) data is not sufficient to capture the effects o f high frequency inU a-scasonal jets, the agreem ent between model dynam ic height and TO PEX sea surface height suggests that the model sim ulation oi'z o n a l currents is reliable {19] This is shown in Figure 2.
The m onsoon je ts have been recently observed using an acoustic current m eter m ounted on a m ooring at 90^E on the equator [20]. The w esterly bursts in sum m er (June-Septem ber) are associated with ISO o f the sum m er monsoon, and in January- March with M JO. To understand the origin o f the m onsoon jets wc have to study the relationship betw een convection, wind stress and currents in the EqlO and then exam ine the role o f equatorial upper ocean dynam ics in the selective response to v^sterly bursts. Sum m er m onsoon ISO ’s is characterized by a bim odal m e rid io n a l stru c tu re o f a tm o sp h eric c o n v e c tio n [ |4 , 17,18]. In the 'active phase' o f the m«mstx>n, there is organized d ie p convection over the North Bay o f Bengal and the G angetic ains. D uring a 'm onsoon break’, convection in this region is ppressed, while there is deep convection over the central and
|s te rn E qlO (E quatorial Indian O cean). T he m onsoon jet celeratcs a few days after the wind stress begins to fall because : westward ZPG (Zonal Pressure G radient), enhanced by the itself, becom es larger than the eastw ard stress. The wind ress is weakly eastw ard or zero, the unbalanced ZPG can give rise to w estw ard currents, as in A ugust o f 2000 and 2002.
Subsequent wind bursts do not generate eastw ard jets. D uring Jim u ary -M arch , the w ind b u rsts in the eastern E q lO are considerably weak than in summer. The equatorial currents are therefore predom inantly w estw ard, although the bursts do give rise to eastward acceleration. The m onsoon je t has the vertical structure expected from theory (21,221; it is about 100 m deep, whereas the WJ (Westerly Jet) can extend upto a depth o f 120 m in the east.
2 .2 Intra’ seasonal oscillations o f monsoon
A ctive and break ep iso d es, are asso ciated w ith en h an ced (decreased) rainfall over central and western India and decreased (enhanced) rainfall over the southeastern peninsula and eastern India (23,24]. The intra-seasonal variations o f rainfall (active- break cy cles) are stro n g ly co u p le d to the in tra -se a so n a l
ZONAL C U R R E N T 1 °S ~1®N 0O-95®E 0 -8 0 m ZONAL WIND S T R E S S 2 ° S -2 ^ N 70-95® E
12 0 9
0 3 0.0 --- 0 .3
20.
10. 0.
F ig u re 2. T im e scries o f (a) lO-day running mean zonal wind .stress (dynes cm ^ : ihin) and m odel daily upper ocean zonal current (cm s”*; bold), (b) 10-day TOPEX sea surface height anomaly (cm; thin) and m odel daily dynam ic height (DHT; dynam ic cm ; bold) difference between eastern and western EqlO.
4 6 0 S D e, T D a tta , M D e a n d A B B h a tta c h a ry a variations o f circulation f 17,25,261- Recent studies [ 16,17,26] have
show n that the spatial structure o f m onsoon ISO is such that they strengthen the seasonal m ean circu latio n in one phase w h ile w e a k e n in g it in th e o p p o s ite p h ase. T h e re fo re , the m onsoon ISO has the potential to m odulate the frequency o f o c c u rre n c e o f L P S (lo w p re s s u re sy s te m ) by a lte rn a te ly enhan cin g and w eakening the zonal w ind shear and low level cyclonic vorticity in the m onsoon trough (M T). D aily N C E P / N C A R (National C enter for Environm ental Prediction / National C enter for A tm ospheric R esearch) reanalysed winds [27] at 850- hPa levels for the period 1 9 5 4 -1993 can be used for studying the larg e-scale intra-seasonal v ariab ility o f c irc u la tio n . P entad C lim ate Prediction C enter M erged A nalysis o f P recipitation (C M A P) for the period 1979-2000 128] can be used to d escribe the large-scale m onsoon intra-seasonal oscillations in rainfall.
T he frequency o f o ccurrence o f the m onsoon synoptic system s as a function o f phase o f the ISO can be determ ined using an index o f IwSO activity. The relative vorticity at 850-hPa represents m o n so o n activ ity q u ite w ell on in tra -se a so n a l tim e scales I 16,17,26|. T herefore, a m onsoon intra-seasonal index (M ISI) can be defined 129] as the 10-90 day filtered relative vorticity at 850-hPa averaged over E - 95^^ E, 12^* N - 22<^ N, w hich represents the core region o f the MT. T he index is constructed each year from 1 June to 30 Septem ber for 40 years and normalized by its ow n standard deviation, A sam ple o f norm alized M ISI for 10 years is show n in Figure 3(a), 3(b). T he total num ber o f LPS during the 40-year period is 510, with a seasonal average o f 12.5.
T he frequency distribution o f genesis o f LPS as a function o f phase o f the ISO can be obtained by putting the genesis dates o f all the LPS during the 40-year period (1954-1993) into bins o f M ISI 129] size 0.25. T his is illustrated in F ig u re 3(b). O ut o f the 510 L PS, 350 occur in the po sitiv e phase o f the ISO and 160 occur in the negative phase. As low s and depressions are the m ain rain-bearing system s o f the m onsoon, the spatial and tem p o ral c lu s te rin g o f L P S is e s s e n tia lly re sp o n sib le fo r increasing (decreasing) rainfall o v er central India during active (break) conditions. We can understand how the ISO achieves this clustering, by exam ining the m odulation o f the low level circulation at 850 hPa by the ISO. It is noted that the vorticity in the m onsoon trough m ay increase (decrease) by 50% during an active (break) spell. T he en h an cem en t o f shear and low level cyclonic vorticity in this region in the po sitiv e phase o f the ISO
increases the probability o f genesis o f LPS. A sim ilar mechanism is responsible for clustering o f tropical cyclones and hurricanes in G u lf o f M exico |3 0 ], eastern Pacific [31J and western Pacific through m odulation o f circu latio n by the M JO .
60
50 40 30 20
10
«ji JBiii
- 2liHlI
- 1 0 1Normalised MISI
F ig u r e 3. (b) H istogram o f gen esis ot LPS (lo w and depres.sions) foi ilu Indian m o n so o n r e g io n (50® E - 100® E, Eq- 30® N ) during June to .September for the period o f 1954-1993 as a function o f normalized MISI
The 10-80 day filtered O LR (O utgoing L ong w ave Radiation) anom alies averaged ov er 85-90^* E, 15-20^ N show s the large variability o f con v ectio n associated w ith m onsoon ISO ovci the N orth B ay o f B engal (B O B ). Sea surface tem perature has
highest intra-seasonal variability in B O B and the northern South C h in a S e a (S C S ). C o h e r e n t n o r th w a r d p ro p a g a tio n tit atm osphere-ocean fields associated with m onsoon ISO is seen in all years. To illustrate this in the longitudes o f the western B O B , the period M ay to S ep tem b er 1998 is considered. l()-8() day filtered anom alies o f O L R , w ind speed and Qnet (net hem flux) all show clear propagation o f ISO from 5^ S to the northern boundary o f the Bay o f B engal 132]. T his is illustrated in Figure 4. SST anom alies propagate northw ard in response to northwaicl m oving, alternating p ositive and negative anom alies of Qnet Since the period o f the oscillations is about a m onth in 1998 and 2000 and the northw ard speed is about 1.40 day""* (1 .8 m s~‘), the phase o f the ISO at the eq u ato r is oppe^site to that in the N orth B ay o f B engal. It is noted that this anti-correlation was not c lear in 1999 b e c au se th e 10—2 0 day m onsoon ISO is energetic. In the S outh C h in a Sea, a few episodes o f northward propagation o f co n v ectio n and Q net anom alies are clear to the
122 244 366 488 610 732
D a y s
854 976 1098 1220
Figure 3. (a) Normalized Monsoon Intra-seasonal Oscillation Index (MISI) 129] for 10 years (each year has 122 days starting from 1954). Thin solid line corresponds to ± 1 normalized unit.
V a ria b ility o f In d ia n m o n so o n a n d its ra in fa ll fo re c a stin g 461 north o f 5 0 N fo llo w ed by stro n g n o rth w ard m oving S S T
inomahes as in the B ay o f Bengal.
O L R W S
r r ^ SEP
A U G ■c»r
J U L -
J U N -
MAY
S E P
A U G '
JU L
JU N
► MAY
■*
1 0 S E Q 1 0 N 2 0 N
1 0 S T E Q 10 N 2 0 N
[•imirt 4. T’lnu* latmiclo section s of lO-XO clay filtcied anom alies o f OLK iVV m “), winclspced {in s ‘j, Qnet (W m and SST ( ) averaged over HS ‘>0” \i in the sum m er o f 199K
The Qnet and wSST (Sea Surface 'rem peraturc) anom aly fields have large zonal extent, from the A rabian Sea to the South China Sea and beyond, d iv in g the active and quiescent phases o f the monsoon 11 Cjj. Thus, the estim ation o f Q net is based on a number 01 approximations, its variability is in reasonable agreem ent with that obtained by SR based on B O B buoy data but for one important difference. B uoy data from the northw estern BOB shows that air tem perature in the sum m er m onsoon can be higher than SST, particularly in p eriods w hen the sky is clear and SS'F IS rising.
2 3 Inter-annual variability :
I he Indian su m m e r m o n so o n has v ig o ro u s in tra-seaso n al oscillations in the form o f'a c tiv e ' and w eak (or 'break') spells of monsoon rainfall [33). T h ese activ e and break spells o f the monsoon arc a s s o c ia te d w ith flu c tu a tio n s o f th e tro p ical convergence zo n e (T C Z ) [1 5 ,3 4 -3 6 ]. T h e in tra -se a so n a l
<Lscillations (ISO 's) o f the Indian su m m er m onsoon represent a
^loadband spectrum w ith periods betw een 10 and 90 days but have two preferred bands o f periods [35,37,38], one betw een 10 imd 20 days and the o th e r betw een 3 0 and 60 days. Several modeling studies show that a sig n ifican t fraction o f the inter- annual variability o f the seasonal m ean Indian sum m er monsoon
governed by internal chaotic d y n am ics [39-42]. M ehta and
K rishnam urti [43] exam ined the inter-annual variability o f the 30-50 day m ode with the winds at 850 and 200-hPa for the peritxl 1980-84 using the European C enter for M edium R ange W eather Forecasts (EC M W F) operational analysis. Singh and K ripalani 144] and Singh et a! 1231 used long records o f daily rainfall data over the Indian continent and exam ined the 30-50 day oscillation.
They, however, could not com e to a clear conclusion regarding relationship betw een the ISO s and the inter-annual variability o f the Indian m onsoon rainfall. A hlquisi ct a l |45J studied r^idiosondc observatioas at 12 Indian stations betw een 1951 a^d 1978 and exam ined ISOs with periods longer than 10 days b^t did not try to relate the ISO s with the inter-annual variability o f the monsoon.
I A model proposed by G osw am i |6} de.scribes, how the ISO s ii|llucnce the seasonal mean and inter-annual variability o f the Ifd ian m onsoon. T he model is based on the sim ilarity betw een t i e spatial structure o f the dom inant ISO m ode and that o f the ifter-an n u al variability. The seasonal sum m er m ean (June-
^ p le m b e r , JJA S) precipitation distribution has a m ajor zone o f l^rge precipitation along the m onsoon trough extending to the North Bay o f Bengal and a secondary zone o f precipitation maximum south o f the equator over the warm waters o f the Indian Ocean. These tw o m axim a in the seasonal m ean precipitation represent two favored locations of the T C Z during the sum m er m onsoon season 115,6].
A ctive and break monst^on c o n d itio n s are trad itio n ally defined based on a precipitation criterion 133|. A circulation- based definition o f active and break m onsoons may be useful for various purposes. D uring an active phase o f the Indian m onsoon, typically there is m ore precipitation over central India and a stronger m onsoon trough.
2.4 C oherent Intra-seasonal oscillations :
The Global Experiment o f 1979 [43] revealed that the ISO involve significant m odulation o f SST and turbulent fluxes at the air-sea interface in the Bay o f B engal and equatorial w est P acific.
Accurate esti mation o f surface fluxes were made during the winter o f 1992-93 [46] by C o u p led O cean A tm o sp h ere R esp o n se E x p erim en t (C O A R E ) in the w est P acific. T his led to the unam biguous dem onstration [47] that the slow oscillation o f SST in the C O A R E region is a respon.se to in tra-seaso n al fluctuation o f surface heat, m om entum and buoyancy fluxes associated with the eastw ard propagating, equatorially confined M a d d e n -Ju lia n o s c illa tio n s (M JO )|4 8 1 . T h e e v o lu tio n o f atmo.spheric convection, surface heat fluxes and SST over the Indo-Pacific w arm pool associated w ith M JO has been studied extensively [49-51 ]. T he results o f these studies suggest that at least on intra-seasonal tim e scales, equatorial w arm pcx>l S S T c h a n g e s a re p rim a rily d riv e n by s u rfa c e h e a t flu x . T h e observations from m oored surface buoys in the w estern B ay o f Bengal show co herent 2 0 -4 0 day S S T fluctuations during the sum m er o f 1998 w ith peak to peak range o f upto 2® C» w hich are
4 6 2 S D e, T D a tta , M D e a n d A B B h a tta c h a ry a not captured by the w eekly N ational C enters for E nvironm ental
P rediction (N C E P) SST analysis [52 J. L arge fluctuations in the net surface heal flux Q net are associated w ith the m onsoon ISO.
D uring the active o r convective phase o f the ISO, the sky is cloudy and surface w inds are strong, leading to negative Qnet w hile the calm or quiescent phase is m arked by clear skies and light w inds, giving large positive Qnet. SR show s that the intra- seasonal SST changes can be understood to the first order o f approxim ation as a response to oscillations o f Qnet. The root- m ean-square differences betw een T M l (T R M M M icrow ave Im ager) fields and three-day buoy data in this region are about 0.6® C for SST and 1.3 m s for w ind speed [53].
3. Link between stratosphere and monsoon rainfall
The Quasi-Biennial O.scillation (Q B O ) in the m ean zonal wind is the dom inant feature o f the tropical stratosphere (20-30 km).
T he easterly and w esterly w ind regim es alternate regularly with a period varying from about 24 to 30 m onths. T he discovery o f the QBO in tropical stratospheric w inds stim ulated the search for strato sp h eric-tro p o sp h eric lin k s.In the late 1970‘s som e evidence for link betw een the IM R (Indian M onsoon R ainfall) and stratospheric zonal w inds w as reported [54-56]. Thapliyal [56] reported that the m ean January circulation features at the 50-hPa levels are able to indicate the deficient m onsoon rainfall in w esterly Q B O years. T hapliyal (57] extended this study by exam ining data o f m ean m onthly global circulation features in the low er stratosphere for the period 1965-80. H e found that in winters o f the easterly Q B O years, a subtropical ridge is situated over the northern hem isphere around 20° N. Such w inters are follow ed by norm al m onsoon activity. T hus, T hapliyal [56,57]
suggested that the January circulation features at 50-hP a levels could indicate the deficient (norm al) m onsoon rainfall in westerly (easterly) Q B O years.
M ukherjee et a l [58] found a significant (at the 5% level) sim ultaneous correlation o f -♦-0.39 betw een the m onsoon rainfall and m ean zonal w ind fo r Jun e-A u g u st at 30 hPa level using w ind data for Balboa (9° N , 80° W ) for the period 1951-82.
An im portant feature o f the Q B O is the dow nw ard phase propagation. T he wind reversal first appears above 30 km and propagates dow nw ard at a speed o f about 1 km /m onth. U sing this fact, Bhalm e e ta l [59] related the January 10-hPa zonal wind anom alies at B alboa w ith IM R and found a correlation o f 0.52 during the period 1958-85. T hey found that IM R tends to be less (m ore) than norm al during an easterly (w esterly) anomaly.
The results o f T hapliyal [57] and B halm e et a l [59] m ay appear to be contradictory; how ever, it m ust b e noted that the results are based at different levels (5 0 -h P a and 10-hPa).
A t the beginning o f 1990's, additional evidence has been found on the links betw een strato sp h eric w inds and IM R. T he India M eteorological D ep artm en t [60] uses 16 param eters in an o p e ra tio n a l lo n g r a n g e - f o r e c a s tin g m o d e l. T w o o f th e se
param eters are related to the stratosphere, nam ely the 50-hPa w ind pattern in w inter and the 10-hPa zonal wind pattern in January.
P arthasarathy et a l [61] found a relationship between the tropical zonal stratospheric w inds at the above 3 stations and rainfall in 5 h om ogeneous regions o f India. T hey also found that the standardised w ind anom alies averaged separately lor the dry and w et rainfall years are characterised by opposite sig n s at all th e s e 3 s ta tio n s , th u s c o n firm in g the stron«
association betw een them . Singh [62] found that the stronger are the Q B O easterlies at 10-hPa levels in January, the large? is the area over the country, likely to suffer from dry conditions Thus, the Q B O seem s to play an im portant role in the interannual fluctuations o f IM R.
Further, K ripalani e ta l [63] related the N orthern Hemisphere 50-hPa geopotential heights w ith IM R. T hey found significant p o sitiv e c o rre la tio n s alo n g 1 0 °-2 0 ° N d u rin g January and February. H ow ever, the m axim um relationship is seen during M arch, w ith high p o sitiv e correlation over the Canadian sectoi and high negative over the E ast A sian sector.
M eehl [64] suggested that the coupled interaction between ocean and atm osphere contributes to a m echanism that produces a biennial com ponent o f interannual variability in the tropical Indian and Pacific regions. H is prem ise was that a wet (dry) m onsoon year w ould be follow ed by a dry (w et) monsoon year H e show ed that the ocean heat content provides the one-yeai tim e scale m em o ry n ec e ssa ry fo r the biennial mechanism W hether this m echanism has any relationship with strato.spheric Q B O needs a separate investigation.
4. Forecasting of monsoon rainfall
In India, agricultural and industrial econom y largely depends u pon m o n so o n an d so its fo re c a s tin g is very im portant.
M eteorological scientists have reco rd s o f earlier monsoons, cycles o f clim ates, m easuring instrum ents to m easure speed of w ind, hum idity, and tem peratures etc. B ut these tools are not sufficient for forecasting o f m onsoon.
J u ly a n d A u g u s t a r e th e m a jo r r a in y m o n th s over N orthw estern (N W ) parts o f India. Investigations by De and B isw as [65] and Pillai and D e [66] have show n that end o f July/
A ugust rainfall is im portant in deciding the seasonal rainfall of the year. By the end o f A ugust w e get a clear indication of the perform ance o f m onsoon o f the year. F requency o f breaks is also m axim um in July and August. As such the correct prediction o f w eekly rainfall in July and A ugust can be a useful input tor the long-range prediction also.
In 1977, India and U .S.S.R . jo in tly carried out research and they found o u t that the so u th -w est m onsoon is depend on so m any a c tiv itie s g o in g o n in A n tra c tic an d P acific oceans.
Df.Vasant Gow arikar, Dr.Thapaliyal and others [60] worked out a
V a ria b ility o f In d ia n m o n so o n a n d its ra in fa ll fo r e c a s tin g 4 6 3 jflodule o f 16 p aram eters to study m onsoon. W ith the study o f
16 param eters and a v a ila b le sta tistic s o f e a rlie r m onsoon, scientists started forecasting o f m onsoon since 1990 and m ore or less the forecasting w as satisfactory o r close to accuracy. In March 2003, m eteorologists changed som e param eters but main parameters rem ained sam e. T h is study tells us about long-term forecasting, w hile study o f satellite pictures gives inform ation on m ovement o f clouds, cyclo n es, and m easurem ents o f wind, temperatures, hum idity give picture o f local changes.
Blanford [67] w as the first to attem pt a forecast o f m onsoon based on the hypothesis that "varying ex ten t and thickness o f the Himalayan snow s exercise a great and prolonged influence on the clim ate conditions and w eather o f the plains o f northw est India". T he su ccess o f B lan fo rd 's ten tativ e forecasts during 1882-85 encouraged the m eteorologists to start operational LR P (Ltmg Range Forecasting) o f m onsoon rainfall covering the whole of India in 1886. Since then the L R F o f m onsoon has becom e an important operational task o f the IM D (India M eteorological Department). T h e fo re c a s ts a fte r 1895 w ere b ased on (i) Himalayan snow cover, O ctober to M ay, (ii) ’local peculiarities' of pre-m onsoon w eather in India and (iii) 'local peculiarities' over Indian O cean and A ustralia [68].
Sir Gilbert W alker [69-71], the D irector G eneral o f IM D , has initiated extensive studies o f w orldw ide variation o f w eather elements such as pressure, tem perature, rainfall ctc.^ with the aim of developing an o b jectiv e m ethod for L R F o f m onsoon rainfall over India by expan d in g the w orks o f H ildebrandsson [72!, Lockyer and L o ck y er [73] and oth ers w ho had draw n attention to the g lobal-scale o scillatio n s in surface pressure.
These studies led W alker to identify three large-scale pressure seesaw p atte rn s; tw o in th e N o rth e rn H e m isp h e re (N orth Atlantic Oscillation (N A O ) and N orth Pacific Oscillation (NPO)) and one in the Southern H em isphere (Southern O scillation (SO)).
While the N A O and N P O are essen tially regional in nature, the SO has since been recognized as a p h enom enon with global- scale in flu en ces. T h e S O w as la te r lin k e d to th e o cean ic phenomenon c a lle d El N ifio in th e e a st-e q u a to ria l P acific characterized by w arm ing o f the sea surface along the Peru coast;
this led to the theory o f W alker C irculation [7 4 1. W alker also succeeded in rem o v in g the su b jectiv ity in the earlier forecast methods by introducing th e co n cep t o f correlation in the field of LRF o f m onsoon rainfall. W ith these pioneering contributions to the field o f L R F, S ir G ilb ert W alker has been m ost closely identified w ith the early attem p ts at m onsoon forecasting in India and his findings are relev an t even today.
Walker [75] also attem pted L R F fo r sub-regions o f India by dividing the co u n try into 3 h o m o g en o u s regions nam ely (i) Northeast India, (ii) P e n in su lar India and (iii) N orthw est India.
I^egression form ulae w ere d ev elo p ed separately for these three regions, w hich had b een su b seq u en tly revised several tim es 1^8]. After that very little progress w as m ade in L R F o f m onsoon
rainfall until the early eighties w hen several studies have re e sta b lish e d th e stro n g link b e tw e e n the m o n so o n ra in fa ll variability and E N SO (El N ino Southern Oscillation) using better data sets.
Pre-m onsoon surface p ressu re a n d therm al fie ld s o ver India M onsoon being the result o f land-sea heating contrast involving large-scale seasonal reversal o f pressure, tem perature and w inds, rqany studies have been carried out to identify useful predictors b ^ e d on the pressure and therm al fields during an teced en t v e n te r and p re -m o n so o n seaso n s. P arth a sa ra th y et a l [76]
d ^ e lo p e d a predictor param eter, which they called West C entral Iijdia (W C l) p re -m o n s o o n (M a rc h -A p ril-M a y ) p re s s u re , r ^ r e s e n te d by the m ean o f sea level pressure (SL P) at six s ^ tio n s (Jodhpur, A h m edabad, B om bay, Indore, S agar and Aptola) located in the core region o f high correlation, w hich sl|ow ed a CC o f -0 .6 3 (significant at 1% level) with the A ISM R (i^ll India Sum m er M onsoon R ainfall) during 1951-80. E arlier, P|uthasarathy et a l\7 7 ] found that the mean surface tem peratures at, th e se s ta tio n s d u rin g M A M se a so n a lso sh o w e d h ig h correlation (0.6) with m onsoon rainfall for the period 1951-80.
M ooley and P aolino [78], using m axim um and m inim um tem perature data for the period 1901-75, revealed that a predictor based on M ay m inim um temp>eratures over western Indian region has good potential for LRF. K rishna K um ar [79] identified tw o predictors b ased on th e m inim um tem peratures d u rin g M arch over Ea.st Peninsular India and during M ay over w est central India.
The operational L R F model [601 o f the India M eteorological D epartm ent also uses three m inim um tem perature param eters representing northern, central and east coastal areas o f India.
Pre-m onsoon 5 00 - hP a R idge location over India
Banerjee e ta l [80] identified the m ean latitudinal location o f the 500-hPa ridges along 75*^ E in April over India. This is considered to be one o f the m ost im portant predictors. The m id-tropospheric anticyclone over southern India m igrates from 11.5^ N in January to its northernmost position o f 28.5^^ N during July. From October, the ridge starts shifting back southw ard. M ooley et a l [81]
found a CC (Correlation Coefficient) o f 0.71 (significant at 0.1 % level) betw een the April ridge location and A ISM R during 1939- 84; a m ore northw ard location indicates better perform ance o f the m onsoon and vice versa. It is conjectured that the northw ard an d s o u th w a rd d is p la c e m e n ts o f th is m id - tr o p o s p h e r ic anticyclonic circulation are related to the seasonal m arch o f the solar rad iatio n and the associated diabatic heat source. T he anom alies in the seasonal evolution o f the m id-tropospheric circulation, as m easured by the April ridge location, can be taken to be a good precursor o f the slow ly varying planetary-scale circulation. A delayed northw ard displacem ent o f the ridge is considered to indicate large-scale anom alous descending m otion
4 6 4 S D e, T D atta, M D e a n d A B B h a tta c h a ry a over the Indian region [82]. In a detailed diagnostic study using
daily locations o f the 500 hPa ridge during the pre-m onsoon m onths o f M arch-M ay for the period 1967-87, K rishna Kum ar et al [83] found that the ridge location in M arch show ed a CC o f - 0.47 with A ISM R, while in April it show ed a CC o f +0.63. They also found that the negative correlation o f the M arch ridge was m ore dom inant with the m onsoon rainfall o f the peninsular India, w hile the positive co rrelatio n o f the A pril rid g e was m ore dom inant with the m onsoon rainfall o f northern India. The difference between the tw o ridge locations (April m inus M arch) show s a CC o f 0.73 with A ISM R. T hough the 500-hPa ridges has show n c o n sisten tly s ig n ific a n t re la tio n w ith m onsoon ra in fa ll in re cen t y e a rs, th e s u b je c tiv ity in v o lv e d in th e determ ination o f its location im poses som e lim itation on its reliability.
U pper tropospheric winds over India
MonsiH^n circulation over India involves m arked changes in the upper tropospheric wind field. K eeping this in view, many studies have show n that the upper air w inds during their pre
m onsoon transition phase can provide a useful predictor. Verma and K am te [84] and Joseph [85) have identified the association betw een Indian m onsoon rainfall and 200-hP a m eridional wind com ponents for the m onth o f M ay, and indicated its potential for prediction o f the seasonal rainfall. P arthasarathy et al |8 6 | have further investigated the relatio n sh ip betw een m eridional wind index (arithmetic average o f 200 hPa meridional com ponent o f wind for M ay at Bombay, Delhi, M adras, N agpur and Srinagar) and A ISM R by using an extended data set for the period 1964- 88 and found a CC o f -0.72 (significant at 0.1 % level).
EN SO indicators
G iving the im portance to the E N S O phenom enon on the clim ate variability in the tropics and ov er several o ther regions o f the globe, m any predictors have been developed representing the strengths o f both its atm o sp h eric co m p o n en t, the S outhern Oscillation (SO), and its cx:eanic com ponent, the El Nino. Walker [75] developed an index o f SO based on a co m b in atio n o f pressure, tem perature and rainfall. S ubsequent w orkers have d e v e lo p e d s e v e ra l o th e r in d ic e s o f S O u s in g d if f e r e n t com binations o f stations m ainly based on pressure data [87,88].
T he m ost com m only used S outhern O scillation Index (SOI) as a m easure o f the strength o f the W alker C irculation across the Pacific, is taken as the norm alized differen ce betw een the SL P (Sea Level Pressure) anom alies at Tahiti and D arw in.
H ow ever, th e CC*s b etw een v ario u s S O In d ices during p re c e d in g w in te r and s p rin g s e a so n s and A IS M R are not s u f f ic ie n tly h ig h fo r L R F as th e r e la tio n s h ip d e v e lo p s sim ultaneously and even follow s the m onsoon season [89,90].
Shukla and Paolino [91] found that the CC o f A ISM R with Darwin SL P during w inter (D ec-Jan-Feb) w as positive w hile that during
spring (M arch-A pril-M ay) was negative. B ecause o f this change o f sign in the C C s, the C C o f the w inter to spring tendency (M A M -D JF) o f SLP at D arw in with AISM R during 1901 -81 w as found to be m uch h igher (-0.46, significant at 0.1% level) than the C C s o f in d iv id u al seaso n s. T h u s, the w inter to sprin«
tendency is considered to be a reliable precursor for the nature o f SO during the m onsoon and later m onths.
The planetary-scale tropical S L P anom alies associated w uh the SO occur in conjunction w ith the episodes o f large-scale sea surface tem perature (SST ) anom alies (El-N ino/I.a Nina) m the tropical Pacific [82,92,93]. It has generally been observed that warm er SSTs in central and eastern parts o f equatorial Pacitk are associated with low er m onsoon rainfall [94,95]. Parlhasarathv and Sontakkc [96], using CO ADS SST data during 1951-80 hiiVL- identified three im portant regions in the Pacific Ocean whose SSTs have show n significant relationships with the AISMR.
These three regions are: (i) 14"-26°N. 128^-140°E; (ii) 14^-20 N, 176”-160°W and(iii) 14^N-10"S, MS^’-KXl^W, whoseMAM-DJI tendencies in SST shc^wed C C s o f 0.4, -0.51 and -0.52 respectively with A ISM R. T he intensities o f El N ino events are generally assessed on the basis o f the average SSTs over three Nino regions in the Pacific Ocean, widely known as N INO 1+2, NINOJ and N IN 0 4 . A m o n g these, only N 1 N 0 4 SST (MAM-DJF) show s statistically significant C C (-0.54 during 195 1-80) with AISMR.
C ross-equatorial flo w
The ob.servational studies o f Saha [97], Pisharoty [98] and C adet and R everdin [99] and m odeling studies o f Washington et a l 1100] and Shukla [ 101 ] have established the importance ol the role o f cross equatorial flow over Indian O cean and moistuiv flux from both Indian O cean and A rabian Sea regions in Indimi m onsoon rainfall, m ainly over the w est coast. However, there is som e difference o f opinion regarding the relative dominance ul fluxes from Indian O cean and A rabian Sea. The East African low -level je t is one o f the m ost im portant m anifestations of the c ro s s -e q u a to ria l flo w in v o lv in g la rg e -s c a le m o istu re and m om entum tran sp o rt [ 102]. C ro ss-eq u ato rial flow develops during the onset phase o f the m onsoon season and its strength, if predictable, can p rovide an im portant predictor for AISMR H astenrath [103] suggested the use o f SST 's over the Arabian Sea and Parthasarathy and Sontakke [96] attem pted to represent the strength o f cross-equatorial flow in the Indian Ocean region by N ouvelle-A galega SL P difference as well as S S T s. However, none o f these observ atio n s could find practical utility in the developm ent o f L R F schem es for A ISM R . T hus, in spite of the know n physical link o f cross-equatorial flow w ith mon.soon rainfall, there has been very little progress in identifying useful predictors based on it. F u th er w orks are needed to find the p o ssib le lin k b e tw e e n c ro ss-e q u a to ria l flo w and monsoon rainfall.
V a ria b ility o f In d ia n m o n so o n a n d its ra in fa ll fo r e c a s tin g 4 6 5 N o rth ern H em ispheric su rfa ce a ir tem perature
Many efforts have been m ade to link the N orthern H em ispheric (NH) mean surface tem perature ano m aly with the strength o f the Indian sum m er m onsoon by using the long period data on hemispheric m ean surface tem p eratu res. Verma et al |104]
jjentified the N H w in te r su rfa c e air te m p e ra tu re anom aly
(January+February) as an im portant predictor for LR F o f AISMR.
ITiis parameter showed the CC o f 0.56 during 1951 >80 maintaining
its significance even during the later years, and is recognized as
one of the m ost im portant predictors.
Eurasian/Himalayan S now co ver
The earlier attem pts m ade by B lanford and W alker brought out the association o f g reater H im alayan snow cover with deficient monsoon rainfall o v er India d u rin g 1880-1920. Subsequently, the reported snow accum ulation show ed very large variability and the relationship w ith the m onsoon rainfall was found to be opposite com pared to the earlier four decades 182). Though this led to the dropping o f snow cover as a predictor by the IM D in 1950, the recent availability o f m ore reliable satellite estim ates of snow cover extent revived interest on this param eter and several studies have show n a statistically significant inverse co n elation between the E urasian/H im alayan w inter snow cover and AISMR 1105-108J. The spatial extent o f E urasian/H im alayan snow cover during the winter is considered to be an im portant slowly varying b o u n d a ry c o n d itio n fo r th e s u b s e q u e n t d e v e lo p m e n t o f m onsoon circulation and th erefo re is a potential predictor with strong physical link [ 109J. Parthasarathy and Yang [ 110| examined the relationship betw een E urasian snow cover (1974-1992) and AISMR and found a C C o f -0.47 for the m onth o f February.
Satellite data, unfortunately, are not available for long periods and also suffer from several inhom ogeneities 1111J, which need to be clarified to develop a useful predictor representing .satellite derived snow cover.
Quasi - B iennia I O scillation
The Quasi-Biennial O scillation (Q B O ) in the m ean zonal wind of the tropical stratosphere (20-30 km ), w ith easterly and westerly wind regimes alternating regularly with a period varying from 24 to 30 m onths, has been found to have strong association with the perform ance o f m o n so o n o v er India. M u kherjee et al [58]
lound a sig n ifican t sim u lta n e o u s C C o f + 0 .3 9 betw een the monsoon rainfall and zonal w ind (JJA ) at 30-hP a using wind data of Balboa (9'» N, 80^ W ) during 1951-1982. Taking cue from the fact that the w ind reversal first ap p ears above 30 km and propagates dow nw ard at a sp eed o f about 1 km /m onth, B halm e al [112] related th e Jan u ary 10 hPa zonal w ind anom alies at Balboa with A IS M R an d fo u n d a C C o f 0 .52 (significant at 1%
level) during 1958-85. A ISM R tends to be less (more) than normal during easterly (w esterly) anom aly. T hough very prom ising, the 10 hPa data from B alb o a are n o t reg u larly reported after 1988
and efforts are being m ade by som e groups to use 10 hPa w ind data from other tropical stations like A.scension Island (8°S, 14°24'W) and Singapore (1 ^40^N, 104°E) having reasonably long- period data.
Spatial patterns o f p red icto r-ra in fa ll relationships
The .spatial distribution o f the relationship o f any predictor with Inidian sum m er m onsoon rainfall can be .studied by com puting its C C 's with the sub-divisional m onsoon rainfall. M ost o f the pi^diclors are known to show the highest CC's over northw estern I central India, and the low est CC's over northeast and extrem e athern parts o f the Peninsula. T he .spatial patterns o f C C 's o f (>sl o f the predictors with the sub-divisional rainfall have been m d to be strikingly sim ilar to tho.se o f AISM R. T hese patterns g e n e ra lly c o n s is te n t w ith th e p re d o m in a n t first E O F ipirical Orthogonal Functions) mode structure o f the m onsoon nfall over India [82] and therefore essentially reflect the spatial berence o f A ISM R.
^rifhniqucs used in LR F studies
M bst o f the studies on L R F o f Indian mon.soon rainfall are based on e m p iric a l o r s ta tis tic a l te c h n iq u e s . T h e s e s ta tis tic a l techniques range from sim ple correlation analysis to advanced procedures such as canonical co n e la tio n analysis and neural networking.
A lm ost all the predictors identified so far have been based on correlation analysis. T hough correlation is a very useful diagnostic tool in bringing out the as.sociation betw een various m eteorological fields, it is highly .sensitive to the data w indow over w hich it is calculated, both in term s o f the position and the length o f the w indow in the lim e dom ain. This im poses certain lim itations on the reliability o f the predictors.
T he most com m only used statistical technique for L R F o f monsoon rainfall is the linear regression analysis. A lai ge num ber o f regression m odels (sim ple as well as m ultiple) have been proposed so far 1113]. T he predictors for the m odel are eith er subjectively chosen rep resen tin g various im portant fo rcin g s on the m onsoon or entered into the schem e by follow ing .some objective criteria. Both approaches have their ow n lim itations;
the subjective selection m ay not optim ize the variance explained w hile the o bjective selection is highly sensitive to the d ata w indow and m ay re su lt in o v e rfittin g to th e d a ta sa m p le [68,96,114]. As the regression m odels tend to acquire sam ple- specific characteristics, th eir reliability is better assessed by testing on as large an independent data set as possible.
A u to -re g re ssiv e in te g ra te d m o v in g a v e ra g e (A R IM A ) m odels w ere also used to forecast th e A ISM R as well as the m onsoon rainfall o v er N o rth w est-ln d ia and P en in su lar India, and w ere reported to, have show n m arginally better forecast skill o v e r the m ultiple regression m odels [115]. H ow ever, the
4 6 6 5 D c\ T D a tta , M D e a n d A B B h a tta c h a ry a auto-correlations in A ISM R during the period 1871-1990 are
statistically insignificant [ 116]. In view o f this, the applicability o f A R IM A m odels for m onsoon rainfall forecasting is doubtful.
G ow ariker ef f l/160] developed parametric and multiple power regression (M PR) models with 15 predictors for L R F o f A ISM R, w hich w ere later, m odified to include 16 predictor param eters.
T he param etric m odel is qualitative and indicates either the m onsoon rainfall to be excess or deficient, depending upon the proportion o f favourable/unfavorable param eters out o f the total o f 16 param eters.
T hapliyal [68] has developed dynam ic stochastic transfer (DwST) m odels for the p red ictio n o f A IS M R as w ell as the m onsoon rainfall over peninsular and northw estern India. In this m ethod, the position o f 500 hPa sub-tropical ridge over India has been considered as an input for the dynam ic transfer c o m p o n e n t, c o u p le d w ith a s to c h a s tic tr a n s f e r s y s te m represented by an A R IM A process and the m onsoon rainfall as the output.
Thapliyal [115] evaluated the relative perform ance o f multiple regression, M PR , A R IM A and D S T m odels and found that the D S T m odel has the h ig h est accu racy am o n g those m odels.
H ow ever, this m odel considers only one p red icto r as input and it is d esira b le to d e v e lo p D S T m o d e ls in v o lv in g m u ltip le p a ra m e te rs re p re se n tin g v a rio u s fo rc in g s o f th e m o n so o n system .
D ue to the spatial variability o f m onsoon rainfall over India, the forecasting o f country wide m ean rainfall is o f lim ited practical utility. K eeping this in view, K rishna K um ar [117] attem pted L R F o f sub-divisional m onsoon rainfall by canonical correlation analysis (CCA ) technique, using m onsoon rainfall data from 29 m eteorological sub-divisions in India as the predictand data set and the SST patterns (M A M -D JF ) o v er the P acific O cean and m inim um tem peratures (M arch and M ay) over India as predictor data set during the period 1958-87. H e found that the spatial extent and the m agnitudes o f useful skill scores for sub-divisional L R F are much larger than those obtained with m ultiple regression analysis [118]. Thus, CCA technique appears to be a prom ising m ethod for L R F o f the spatial patterns o f m onsoon seasonal rainfall.
Com pared to the num ber o f studies using em pirical m ethods, the studies on the seasonal prediction o f m onsoon rainfall using general circulation m odels (G C M s) are very few. T h is m ay be partly because o f the lack o f skill in the sim ulation o f m onsoon rainfall over the Indian subcontinent by m ost o f the G C M s [ 119J.
A lso , th ere are m ark ed d iffe re n c e s fo u n d in the m on so o n precipitation sim ulated by differen t G C M s (W CRP, 1992). T he fact that the sim ulation o f the sum m er m onsoon rainfall o v er the Indian region is very sensitive to the initial conditions [120]
also create serious problem s in this context. Ju and Slingo [121]
a n d S o m an and S lin g o [1 2 2 ] h a v e d e m o n s tra te d th a t the
seasonal integrations o f G C M s could sim ulate strong and weak m o n so o n c irc u la tio n s b ased on th e S S T distributions over tropical Pacific and Indian O ceans.
S ecular variations in the A lS M R -p re d ic t o r relationships M ost o f the predictors show ed insignificant C C ’s with AISMR till 1950 and the C C 's becam e significant only around the yeai 1951. H ow ever, there are som e exceptions like Darwin SLP tendency and P acific SST, w hich show ed significant CC’.s for relatively shorter periods aro u n d 1891 and 1921. T he C C \ of som e predictors have even changed their sign during the earlv part o f this century.
Parthasarathy et a l [123] have exam ined the relationship betw een B om bay S L P and A ISM R with an extended data set tor the pieriod 1847-1990 and found turning points in the CC*s around the years, 1870, 1900 and 1940. T hey attributed these turning points to d e lin e a te b etw een tw o a ltern atin g regim es in the m onsoon circulation, identified earlier as 'm eridionar and 'zonal' types by Fu and F le tc h e r [124J. Partha.sarathy et al [12.^1 concluded that the Indian sum m er m onsoon had passed through two meridional (1871-1900; 1941 -90) and two zonal (1847-1870.
1901-1940) circulation regim es during the pa.st 150 years. The>
also found th at the relatio n sh ip betw een B om bay SLP and A ISM R becom es d om inant only w hen the E N SO variance m Bom bay pressure is high and falls apart when the ENSO vanance is small.
It appears that during the last 3 or 4 decades, the HNSO phenom enon has played a dom inant role in the clim ate variahilit>
in general and m onsoon variability in particular. During ihis
period, all the three facets o f atm osphere-land-ocean system seem to have hieen strongly coupled. It is not clear how and why this cou p lin g w as not d om inant in the earlier decades A nnam alai [125] suggests that the presence o f decadal-scalc o sc illa tio n s in th e p re d ic to rs th e m se lv e s m ay possibly be responsible for the instability in the relationship between AISMR and its p re d ic to rs. T h e slid in g C C 's also su g g e st that (he p red ictab ility o f the m o nsoon il.self m ay be having secular v a ria tio n s , w h ich p ro b a b ly is o n e o f th e re a so n s for the tem porary 'lull’ in L R F research im m ediately after Walker's time.
5. Instability characteristics of monsoon disturbances In m any cases the onset o f Indian M onsoon is sudden and the o n s e t p h a s e is a s s o c ia te d w ith s o m e fro m o f transient disturbances. In m ost o f the cases the disturbances originate in the A rabian Sea and very rarely they originate in the Bay of B engal. O nce the m o nsoon sets, its fu rth er progress takes place d u e to ra in b e a rin g s y s te m s lik e m o n so o n tro u g h , lows, d epressions, m id trop o sp h eric cy clones, etc. T hese synoptic scale system s are con sid ered as perturbations em bedded in the b asic m o n so o n cu rren t. M an y attem p ts h av e been made to explain the m onsoon disturbances o f the B ay o f Bengal in terms
dynamic instability. S om e scattered studies are available for he s \ i n the A rabian Sea. T h e onset phase differs in each hin after the o n se t a lo w -p ressu re area develops in the vrabian Sea, w hich later dev elo p s into cyclonic storm [ 126]. To ivestigaie the instability d istu rb an ces for the form ation o f the vchmic storm» w e h ave to first analy se the grid point values ot ,nd and tem perature at a p a rtic u la r reso lu tio n obtained at jndard isobaric levels for a particular pericxl. This is considered
^ i)ic basic state for th e fo rm atio n o f the system .
Hie system w ind has a strong m eridional and vertical shear
^j so longitudinally averag ed zonal How is considered to find s impact in the m eridional plane. To get the instability, we have , m eridional g radient o f ab so lu te vorticity and iicniial vorticity for v erify in g the necessary conditions o f (totropic, b a ro c lin ic and c o m b in e d b a ro tro p ic -b a ro c lm ic istjbility for each day. Since the baroclinicity is involved verticcil luiification should be taken into account. T he required static abilit) from surface to 100 hPa is co m puted from the area-
naged tem perature using the relatio n [ 126|
a - - R lP [ d T ld P - R T /C p P ] ,
; rr IS static stability, P is pressure, 7 is tem perature, R is IS constant, and Cp is sp e c ific heat at co n sta n t pressure.
cMce, the static stability is varying only in the vertical directitm.
v‘\i. com pulations are c a rrie d out tt> verify the necessary uKiiTK>ns o f barotropic, b aroclinic and com bined barotropic- iioclmic instability in the m eridum al plane /.c.,
/ J - f / y y - O ,
P /i)“ ( a ’ f/p ) p. = 0 and , (t U y y - f ^ - { a - ^ U i > ) p - 0 .
here p IS Coriolis param eter, U is zonal w ind and/j^ is C oriolis itc Heie, subscriptsp and y im plies differentiation with resp>ect
that variable. In this case, -plane approxim ation is assum ed
■d so/;, and p values co rresp o n d to 0.2 5 ° N, w hich is the nter latitude o f the latitudinal belt u n d er consideration. The ijcctory o f the o b se rv e d c y c lo n ic storm lies in th e N orth abian Sea. So tw o areas in the longitudinal zone are considered,
. t irsi from 39° E to 81 ° E (big area) and second from 51 ° E to E (small area). T h u s, all th e in p u t p a ra m e te rs and the mpuiations are carried o u t in th e b ig as well as in the sm all -i Zonal wind shear in creases from big area to sm all area for (he days. W ind sh ear in creases and static stability decreases (he lower lev els w ith tim e. T h e n ec e ssa ry co n d itio n for
«^'Jtropic instability fl2 7 J is sa tisfied fro m lo w er to upper
’P“sphere. It is tru e also fo r the case o f necessary condition trombined b a r o tro p ic -b a ro c lin ic in s ta b ility (1261. T h e igniiude o f m eridional g rad ien t o f absolute vorticity (;0 -
Vat ia h ility o f In d ia n m o n so o n a n d
its ra in fa ll fo r e c a s tin g
4 6 7 as well as rmtential v o rticity (y. p ) in c r e a s e s fro m 3 1 st M a y to 3rd June. T his is illustrated in F igures 5 (a ) and 5 (b ).
Latitude
F i f » u r e 5. (a ) and (h) N e c e s sa r y e o n d iiio n Itjr e o in b in e d b a ro tro p ic- burocliiiic in stability
T he negative zone is found prom inently betw een 15° N to 20° N. The zero line around 15° N indicates the possible preferred latitude for the form ation of' the system . M eridional shear is considered to be the prim ary m echanism for the form ation o f the system in the w hole troposphere.
In the w'hole troposphere, irujiisoon disturbance is found to be b aro tro p icaliy unstable. T h e latitudinal belt o f n eg ativ e m eridional grad ien t o f ab so lu te vo rticity is reduced but its m agnitude is increased with tim e. T his is true for the case o f m eridional gradient o f potential vorticity also.
6. New' fo re c a st m odels fo r In d ia n S W m on so o n ra in fa ll T he 2002 forecast for the Indian .south-west m onsoon by the India M eteorological D epartm ent prom pted severe criticism over the validity o f the ex istin g 16- p aram eter pow er reg ressio n statistical mcxiel for L ong R ange Fi^recast (LRF). IM D ’s forecast perform ance betw een 1988 and 2002 is presented in Figure 6. As evident from the figure, the character o f the m onsoon season o f July 2002 had been 'unique' and the nature o f the anom aly by either hindsight or retrofitting has not yet been pinpointed 1128].
As a backlash, the IM D attem p ted to d ev elo p a new set o f L R F models. The new ly adopted 8- param eter pow er regression model w as first used for the Ipng-range forecasts for the 2003 SW mon.soon rainfall. IM D also sim ultaneously used a 10- param eter
4 6 8 S D e, T D a n a , M D e a n d A B B h a tta c h a ry a p ow er regression and probabilistic m odels for the long range
forecast. A com parative study for the new 8> and 10- paiam eter
F 'igu rc 6 . T h e p e rfo rm a n c e o f o p e r a tio n a l fo r e c a s ts [1 2 8 ) b e tw e e n I‘>88 and 2 0 0 2 (sou rce . IM D . N ew D elh i)
m odels as com pared to the early 16- p a ra m e te r m odel w as conducted (128] for 1996-2002, as show n in Table 1.
T a b le 1. A com parative chart for the 8- and 10- param eter m o d els with 16- parameter m odel (Sou rce . IM D . N e w D elh i)
Year 8- p aram eter 10- p ara m ete r 16- param eter
1 9 9 6 + ^ + 3 7
1 9 9 7 - 1 I -1- 10
1 9 9 8 - 4 - 2 4 6
1 9 9 9 - 5 - 6 - 12
2 0 0 0 -f 3 4- 3 - 7
2 0 0 1 + 1 + 3 - 7
2 0 0 2 - 17 - 14 20
N ote ; A ctual m inus forecast (%)
W ith these data, it has now becom e possible to issue the long range forecast in tw o stages, first in m id- A pril using data upto M arch and an update in m id-July u sing data upto June.
7. Conclusions an d scope fo r fu tu re investigations
D ifferent structures o f monscx>n d istu rb an ces are well know n o v e r I n d ia a n d n e ig h b o u r in g S e a s , b u t th e in s t a b i l i t y characteristics o f such d istu rb an ces o v er A rabian Sea and Bay o f B engal arc not very clearly know n. D ifferen t attem pts have been m ade by the scien tists to d eterm in e the role o f barotropic, baroclinic and com bined barotropic-baroclinic instability on the initial form ation o f such distu rb an ces o v er A rabian Sea. F urther w orks arc required in this line.
T h e 'stro n g ' m o n so o n y e a rs are c le a rly a sso ciated w ith higher frequency o f o ccu rren ce o f'a c tiv e ' con d itio n s and 'w eak' m onsoon years are clearly asso ciated w ith h ig h er frequency o f o ccurrence o f 'break' co n d itio n s. T h u s, the frequency o f intra
seasonal regim es determ ines w hether it will be a 'strong* or Indian sum m er m onsoon. As the seasonal m ean monsoon influenced partly by external forcing and partly by the iSo even if the strong (w eak) m onsoons are not clearly charaeteruecj by higher frequency o f the ’active' ('break') conditions, the n,!e o f ISO's could not be ruled out. T h e fact that even after separai,n„
the contribution o f the external forcing, strong (weak) monsoons are clearly characterized by higher frequency o f ’active’ ( break • co n d itio n s in d icates that th e IS O ’s play a dom inant role n*
determ ining the seasonal m ean. A s the IS O ’s are inumsicali chaotic, the prediction o f inter-annual variations of the Indiir.
su m m e r m o n so o n b e c o m e s d iffic u lt an d w ill have ?o probabilistic in nature.
T he grow th o f erro rs in the transitions from active to hie.iK is governed by the low freq u en cy 30-60 day oscillations ol ihf m onsoon H adley circulation [ 129). The significance of the stud\
IS that they are not lim ited only to the Indian sum m er mons<u>fi IS O ’s but rep resen ts a fundam ental property o f tropica! imim
seasonal variability in general. F or exam ple, they arc applicable to the eastw ard prop ag atin g M addcn-Julian O s c i l l a t i o n s 11 <i |
in its convectively coupled regim e o v er the Indian f)ccan and w e s te rn P a c if ic . T h o u g h e x te n d e d ra n g e prediction convectively active co n d itio n s m ay rem ain to be difliciili, the conceptual and m odeling support can predict diy spells up three w eeks in advance.
N avone and C eccatto 1131J have used ’feed-forw ard’ neui ii netw ork technique for the prediction o f Indian monsoon Kuntel!
with tw o p redictors (500 hPa ridge location and Darwin Si P
tendency from Jan u ary to A pril). T hey reported that the ncm.
netw orks could m ake a better use o f the predictive infoini:iii« n available in the pred icto r data. H ow ever, further work needs i be d o n e to c o n c lu s iv e ly e s ta b lis h th e su p e rio rity o< th(
advanced statistical tool o v er o th er m ethods.
It has been noted th a t there have been alternating penod'^
extending to 3-4 decad es w ith less and m ore frequent weak
m onsoons o v er India. F o r ex am ple, the 44-y ear period J 921-64 w itnessed ju s t th ree d ro u g h t years; d u rin g such epoch.s, the
m onsoon w as found to be less correlated with the E N S O Dui
the other periods like that o f 1965-87, w hich had as m a n y 10
dro u g h t years o u t o f 23, th e m onsoon w as found to be strongly linked to the E N S O . T he problem is an interesting one and to be
considered as a future problem . IM D has adopted a two-stage forecast p rocess, the first one on 16 A pril and the second yn update to be m ade available in m id- July. H ow ever, IMD has noi got yet the forecasting m odels peer-review ed which is definitely
in line for en su rin g req u isite in p u t p aram eters needed foi statistical m odels from space satellite data o f the In d ia n
R esearch O rg an isatio n 's (IS R O ) satellites such as 'Clim atesat.
'M etsat' and 'O ccan sat' fo r fu rth e r fo recastin g o f the Indian
m o nsoon presen tly and in future.