SPATIAL VARIABILITY OF DAILY RAINFALL OVER ORISSA, INDIA, DURING THE SOUTHWEST SUMMER MONSOON SEASON

INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Clmatol. 3: 1867 1887 (3) Publshed onlne n Wley InterScence (www.nterscence.wley.com). DOI: 1.1/joc.974 SPATIAL VARIABILITY OF DAILY RAINFALL OVER ORISSA, INDIA, DURING THE SOUTHWEST SUMMER MONSOON SEASON M. MOHAPATRA, a U. C. MOHANTY b, * and S. BEHERA b a Inda Meteorologcal Department, Bhubaneswar, Orssa-751, Inda b Centre for Atmospherc Scences, Indan Insttute of Technology, Delh, New Delh-1116, Inda Receved 18 February 3 Revsed 17 September 3 Accepted 4 September 3 ABSTRACT Southwest summer monsoon ranfall over Orssa, a state of eastern Inda, shows characterstc spatal and temporal varablty, due to the nteracton of basc westerly flow wth orography and the synoptc-scale monsoon dsturbances developng over the northern Bay of Bengal and movng west-northwestwards along the monsoon trough. The am of ths study s to fnd out the man features of the spatal varablty of daly monsoon ranfall over Orssa and assocated synoptc systems. Emprcal orthogonal functon (EOF) analyss s a good tool to flter out the man components from any nose, and ths was appled to daly monsoon ranfall (June September) data of 31 unformly dstrbuted statons over a perod of years (198 1999). The assocaton of synoptc systems wth sgnfcant EOFs has been confrmed by analysng daly synoptc systems over Orssa and ts neghbourhood durng the same perod. The frst three EOFs n S-mode may be attrbuted to good monsoon ranfall n assocaton wth low-pressure systems/cyclonc crculaton (LPSC), lke lows, depressons, cyclonc storms and cyclonc crculaton extendng up to the md-tropospherc level over the northwest (NW) Bay/NW and the adjonng northeast (NE) Bay, over Gangetc West Bengal (GWB) and over Orssa/NW and the adjonng west central (WC) Bay. The fourth and ffth EOFs may be attrbuted to a weak monsoon condton beng assocated only wth a monsoon trough wthout any embedded system and LPSC over the NE and the adjonng NW Bay respectvely. Based on rotaton of sgnfcant EOFs n T-mode, whch gves better regonalzaton, Orssa conssts of fve homogeneous regons of daly monsoon ranfall: () eastern Orssa, () western Orssa, () northwest Orssa, (v) north Orssa and (v) northeast Orssa. Eastern Orssa gets hgher ranfall than the other regons, due to LPSC over NW Bay/NW and the adjonng NE Bay, western Orssa ranfall s due to LPSC over Orssa/NW and the adjonng WC Bay; lkewse, northwest Orssa ranfall s due to LPSC over GWB, north Orssa ranfall s due to LPSC over the NE and the adjonng NW Bay, and northeast Orssa ranfall s due only to a monsoon trough wthout any sgnfcant embedded system over Orssa and adjonng land/sea areas. Copyrght 3 Royal Meteorologcal Socety. KEY WORDS: EOF analyss; daly monsoon ranfall; regonalzaton; Orssa (Inda) 1. INTRODUCTION The southwest summer monsoon ranfall (June September) over Inda s characterzed by a sem-permanent monsoon trough that extends from west Pakstan to Burma across northwest Inda and the northern Bay of Bengal, by low-pressure systems and cyclonc crculatons extendng up to md-tropospherc level (LPSC), whch frequently develop over the northern Bay of Bengal and move n a west-northwesterly drecton across Orssa or Gangetc West Bengal, and by a westerly trough that occasonally moves n an easterly drecton across the central parts of Inda. However, there s varaton n monsoon ranfall over Inda, both n space and tme, due to nteracton of the basc monsoon flow and synoptc-scale systems lke LPSC. Also, there s nteracton of convecton wth the Western Ghats, Eastern Ghats, other hll peaks and, above all, the Hmalayas. * Correspondence to: U. C. Mohanty, Centre for Atmospherc Scences, Indan Insttute of Technology, Delh, Hauz Khas, New Delh 1116, Inda; e-mal: mohanty@cas.td.ernet.n Copyrght 3 Royal Meteorologcal Socety

1868 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA Therefore, the monsoon ranfall over Inda s hghly complex n nature. There have been many studes on spatal varablty of ranfall over Inda, e.g. Rakhecha and Mandal (1981), Hastenrath and Rosen (1983), Rasmusson and Carpenter (1983), Prasad and Sngh (1988), Krpalan et al. (1991) and Majumdar (1998). Orssa State, a meteorologcal subdvson of Inda, les on the east coast of Inda, adjacent to the northern Bay of Bengal. The monsoon ranfall over Orssa s more complex than that over all Inda and s sgnfcantly dfferent from that of Inda due to the larger nfluence of LPSC developng over the northern Bay of Bengal and sgnfcant nteracton between convecton and basc flow due to the vared physography of Orssa, whch ncludes the Eastern Ghat hll ranges. Hence, to understand the daly ranfall varablty and to predct the daly ranfall over Orssa, t s essental to fnd out the man features of daly ranfall over the regon by flterng out any nose. As emprcal orthogonal functon (EOF) analyss s one of the best technques for ths purpose, ths was appled to daly monsoon ranfall over Orssa. In vew of the sgnfcance of precptaton n a clmatologcal/meteorologcal context, the majorty of studes nvolvng applcaton of EOF/prncpal component analyss (PCA) have utlzed ranfall data to study the spato-temporal varablty of the precptaton regme over varous regons (e.g. Murata, 199; Drosdowsky, 1993; Sumner et al., 1993). The daly ranfall over a regon was studed by Sumner et al. (1993) to fnd out the spatal organzatons of daly ranfall over Mallorca, Span, based on data for 4 years. A detaled dscusson of the prncpal components and common factor methodologes can be found n Morrson (1976), Mathur (1976) or Harman (1976). When appled to atmospherc flow parameters, EOF analyss leads to the concept of map typng (Rchman, 1981) and the specfcaton of characterstc modes of varaton n the data (Barnston and Lvezey, 1987). Studes on spatal varablty of monsoon ranfall over Inda usng EOF/PCA nclude those of Bed and Bndra (198) on seasonal (June September) monsoon ranfall of 7 statons durng 1911 7; of Hastenrath and Rosen (1983) on annual total ranfall of 31 meteorologcal subdvsons of contguous Inda durng 19 7, as estmated by the Inda Meteorologcal Department (IMD); of Rasmusson and Carpenter (1983) on seasonal area-weghted monsoon ranfall of 31 meteorologcal subdvsons calculated from 31 representatve statons durng 1875 1979; of Shukla (1987) on seasonal monsoon ranfall of 31 meteorologcal subdvsons durng 191 7 of Prasad and Sngh (1988) on area-weghted seasonal moonsoon ranfall of 31 subdvsons durng 191 8; and of Gregory (1989) on area-weghted seasonal ranfall of 9 meteorologcal subdvsons durng 1871 1985 as calculated from ranfall over 36 statons representng 36 dstrcts of contguous Inda (Parthasarathy et al., 1987). Applcatons of EOF/PCA n the medum-range scale have been made by Gadgl and Iyengar (198), Krpalan et al. (1991) and Majumdar (1998). Krpalan et al. (1991) used pentad ranfall data durng the monsoon season over 5 blocks of.5.5 lattude longtude grds for the perod 191 8, whereas Gadgl and Iyengar (198) used the 5 year mean pentad ranfall data (total 73) pertanng to 53 statons n pennsular Inda, based on data for 191 5 as compled by Ananthakrshnan and Pathan (1971). Majumdar (1998) used weekly ranfall departures from normal durng the southwest monsoon perod for all 35 meteorologcal subdvsons from 1977 to 1986, as calculated by the IMD. The frst EOF of all the abovementoned studes ndcates that the ranfall pattern remans the same throughout Inda except for the northeast and the extreme southeast of Inda. Ths pattern may represent extreme seasonal ranfall leadng to all-inda flood/drought years. There s a strong spatal couplng between monsoon ranfall over northwest and central Inda, ncludng Orssa, whereas these regons have lttle affnty wth northeastern and southeastern Inda. Comparng other sgnfcant EOFs n the above studes, there s a characterstc pattern over a zonal band extendng from Orssa and adjonng regons to northwest Inda, opposte to that over the rest of Inda. Ths zonal band s also the regon of the monsoon trough and movement of low-pressure systems lke depressons, etc., developng over the Bay of Bengal. Whereas Orssa and Gangetc West Bengal consttute a separate homogeneous regon accordng to Prasad and Sngh (1988), Orssa, Chhatshgarh and east Madhya Pradesh consttute a homogeneous regon accordng to Gregory (1989). All the above studes, though gvng a broad pattern of monsoon ranfall over Inda and the role of the monsoon trough and monsoon depressons, could not determne objectvely the spatal varablty of ranfall wth respect to the regon of formaton and movement of synoptc dsturbances lke LPSC over the Bay of Bengal. Unlke the above-mentoned studes, whch used pentad, weekly, monthly and seasonal ranfall ths study uses daly ranfall, whch wll help n objectvely lnkng the spatal varablty of monsoon ranfall Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1869 over Orssa to the LPSC and monsoon trough. It wll also help n classfcaton of Orssa nto dfferent homogeneous regons. Ths study wll be useful for further analyss and for desgnng statstcal forecast models for short-range predcton of monsoon ranfall over Orssa.. DATA AND METHODOLOGY Orssa State, a meteorologcal subdvson of Inda, les on the east coast of Inda adjacent to the northern Bay of Bengal (Fgure 1(a)). Fgure 1(a) descrbes the surface sobarc pattern, the basc monsoon flow at.9 km above mean sea level (a.m.s.l.) over the Indan regon, the locaton of the mean poston of the monsoon trough and the regons of ranfall maxma and mnma wth respect to the monsoon trough durng the representatve month of July. The tracks of cyclonc dsturbances developng over the Bay of Bengal durng the representatve monsoon month of July clearly demonstrate the west-northwesterly movement of these systems along the monsoon trough (IMD, 1979). Hence, the basc flow n the lower levels extendng up to about 6 km a.m.s.l. s westerly over the regon south of the monsoon trough. As westerly wnds are relatvely dry and contnental over Orssa, because of the long path over the land mass from the west coast to Orssa, these are less ran-bearng. Hence, Orssa does not get any apprecable amount of ranfall n the absence of any synoptc-scale monsoon dsturbances over the northern Bay. In the presence of dsturbances lke LPSC over the NW Bay, there s an nteracton between the basc westerly flow, whch s a relatvely dry, contnental wnd, and the monsoon dsturbance, leadng to maxmum convergence n the southwest sector of the system. As Orssa les n the southwest sector of the system over the northwest Bay and neghbourhood, where the maxmum number of these systems develops durng the monsoon season, Orssa gets maxmum ranfall. Many studes, lke those of Rao and Rajaman (197, 1975), Raghavan (1973), Rajaman and Rao (1981) and Pathan (1993), have confrmed that the southwest sector of the westwards-movng monsoon depresson gets maxmum ranfall due to maxmum convergence. In addton to the nteracton between the basc westerly flow of the monsoon and monsoon dsturbances lke LPSC,there s also an orographc nteracton due to Eastern Ghat, whch extends from southwest to northeast n south Orssa. Also, there s an orographc nteracton due to smaller hll peaks n dfferent parts of Orssa. The dfferent physographcal regons and dstrcts of Orssa are shown n Fgure (a). Physographcally, Orssa 35 N 3 998 5 1 15 1 1 (a) 14 16 18 Araban sea 11 998 Ranfall Mnmum Monsoon Trough Ranfall Maxmum Wnd Speed = 1 kt Wnd Speed = 5 kt 1 Orssa Bay of Bengal 1 14 7 75 8 85 9 95 E 16 18 11 3N 5N N 15N (b) 6 4 5 7 75E 8E 85E 9E 95E 1. North East Bay (NE Bay). North West Bay (NW Bay) 3. West Central Bay (WC Bay) 4. North Coastal Andhra Pradesh (NCAP) 5. Orssa (ORS) 6. East Madhya Pradesh and Chatshgarh (EMPC) 7. Jharkhand (JKD) 8. Gangetc West Bengal (GWB) 9. Bangladesh (BDS) 3 8 9 1 Fgure 1. (a) Mean sea level pressure (hpa) pattern and mean wnd (knots) at.9 km durng representatve month of July (source: Inda Meteorologcal Department) and (b) regons of low pressure systems/cyclonc crculaton (LPSC) under consderaton Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

187 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA conssts of broadly four regons. () coastal plan; () southwest hlly regon of Eastern Ghat; () northern upland; and (v) central rver basn. There are a number of hll peaks n the Eastern Ghat regon and northern upland. Though the Eastern Ghat hll ranges extend from Tamlnadu State n the southwest to Orssa n the northeast, beng parallel to the east coast of Inda they are more promnent n south Orssa. The eastern part of Koraput and the nteror part of Ganjam dstrcts le on the eastern sde of Eastern Ghat. The dstrcts mentoned n Fgure (a) are 13 undvded dstrcts of Orssa. Though, recently, these dstrcts have been subdvded to form a total of 3 dstrcts, the present study has been based on these 13 dstrcts; Balasore and Cuttack consttute north coastal Orssa, Pur and Ganjam consttute south coastal Orssa, Kalahand, Bolangr, Koraput and Phulban consttute south nteror Orssa and the remanng dstrcts consttute north nteror Orssa. Owng to all the above-mentoned factors, the daly monsoon ranfall over Orssa s hghly complex n nature. The basc objectve of ths study s to fnd out the man features of the spatal varablty of daly (a) 7 8 6 7 4 1 1 5 11 1 3 13 9 18 8 84 86 88 Southwest hlly regon Coastal Plan Northern upland Central rver basn Axs of Eastern Ghat Hll peaks 1. Sundergarh. Keonjhar 3. Mayurbhanj 4. Sambalpur 5. Dhenkanal 6. Bolangr 7. Phulaban 8. Kalahand 9. koraput 1. Ganjam 11. Pur 1. Cuttack 13. Balasore (b) 31 3 19 3 9 8 17 16 18 7 15 5 6 4 13 14 11 5 3 1 4 9 1 1 1 8 7 6 1. Balasore. Chandabal 3. Barpada 4. Udala 5. Rarangapur 6. Paradeep 7. Cuttack 8. Jenapur 9. Anandapur 1. Keonjhar 11. Champua 1. Kamakshyanagar 13. Pallahara 14. Anugul 15. Rajkshorenagar 16. Panposh 17. Jharasuguda 18. Sambal pur 19. Ambobhona. Pur 1. Bhubaneswar. Phulban 3. Bolangr 4. Gopalpur 5. Sorada 6. Mohana 7. Paralakhemund 8. Gunupur 9. Gudr 3. Bsam Cuttack 31. Nawarangapur Fgure. (a) Physographcal map of Orssa and (b) selected ran gauge statons of Orssa Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1871 monsoon ranfall over Orssa by flterng out the nose and to correlate these man features of ranfall varablty wth the broad-scale synoptc systems lke LPSC. Ths study uses EOF analyss appled to the daly monsoon ranfall data from 31 statons n Orssa for a perod of years (198 99). As the monsoon season (June September) conssts of 1 days, the data matrx used n the study s a matrx of 31 44 statons by days form. Also, an attempt s made to fnd out dfferent homogeneous regons of daly monsoon ranfall over Orssa by rotatng the sgnfcant EOFs. On a real-tme bass, daly ranfall s recorded at about 6 statons n Orssa. However, a contnuous record of daly ranfall durng the monsoon season for the perod of years (198 99) s only avalable for 31 statons. These 31 statons, whch are almost unformly dstrbuted n Orssa, have been selected for ths study (Fgure (b)). The necessary qualty control of the data has been carred out and the mssng data, though very few n number for these 31 statons, have been flled n by consderng the ranfall at surroundng statons. The correlaton between the daly average ranfall over Orssa based on real-tme data and the daly average ranfall over Orssa based on the data from 31 selected representatve statons n Orssa s found to be.93, whch s hghly sgnfcant. Hence, the daly ranfall over Orssa can be well represented by these 31 statons..1. Modes of analyss For the present study, the parameter (ranfall) s fxed, and hence there are two possble modes of analyss: space (S) and tme (T) modes (Rchman, 1986). In S-mode analyss, the varables are the statons and the observatons are the values of daly ranfall. The EOF loadng matrx then contans the correlatons of each varable or staton wth each component. These can be plotted on a map to depct the spatal pattern of each component. Ths matrx can also be multpled by the orgnal data matrx to form the component ampltudes or score matrx, whch contans the tme seres of each component. Ths s usually standardzed to have zero mean and unt varance. The procedure followed n ths study s based on the algorthm proposed by Von Storch and Hannoschock (1984). Accordng to ths algorthm, each row s assumed to be a sample (here, a staton) and each column a tme varable (here, the daly ranfall) n the data matrx. The zero-averaged data set s calculated from the data matrx and then the sngular value decomposton of the problem (.e. the decomposton of the matrx nto a sum of rank-one matrces formed as an outer product of the bass vectors for the rows and columns wth a premse that the bass vectors are orthogonal wth any specfed metrc) s determned from the zero-averaged data set. The large egenvalue (EV) matrx s then rewrtten to a vector and the approprate normalzaton s also appled. Rchman (1981) has noted that the S-mode analyss, whch groups statons wth smlar temporal varatons, may not be the best method of map typng, and suggests the use of a T-mode analyss. In T-mode analyss, the data matrx s transposed so that each ndvdual tme s a varable and each ranfall staton an observaton. Ths analyss produces components wth loadngs on the ndvdual days and ampltudes on the observatons or spatal varables. A potental problem wth ths form of analyss s the manner n whch the smlarty between the spatal patterns s measured. The measure of smlarty mostly used n EOF analyss s the covarance or correlaton coeffcent. In ths study, the covarance matrx has been used for T-mode analyss... Rotaton of EOFs It has been found that the EOFs form a predctable set of patterns, due largely to the orthogonalty constrant. The frst EOF maxmzes the explaned varance by havng generally large loadngs on all varables, and subsequent components are dpoles or more complex patterns. In addton to ths doman dependence, Rchman (1986) has shown that unrotated EOFs suffer from a number of further defcences. These can be remeded by rotatng the EOFs. The most mportant dffculty nvolved n rotaton of EOFs s that no generally applcable crteron exsts to determne the number of EOFs to be rotated, and ths number s sgnfcant, snce the loadngs on the rotated EOFs depend on the number of EOFs retaned for rotaton. Also, for unrotated components, the sgnfcant components represent the sgnal and the remanng hgher order EOFs the nose. Thus, t s mportant to know the number of EOFs that are sgnfcantly dfferent from the nose and that Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

187 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA can be explaned by physcal processes. The other major dffculty s that the best method of rotaton also depends to some extent on the partcular data set beng analysed. There are dfferent technques to determne the sgnfcance of the EOFs. Most of these technques can be vewed as domnant varance rules. Kaser (1958) suggests the Kaser Guttman test, whch s based on the assumpton that egenvectors that have egenvalues less than unty for a correlaton matrx explan less varance than uncorrelated whte nose. The scree test (Cattel, 1966) looks for a cut-off n the dfference between successve EVs or a break n slope n EV sequence, wth the EVs representng nose decreasng n geometrc progresson. Craddock and Flood (1969) suggest that the break s more dstnct n a plot of log(ev) aganst egenvector number. Presendorfer et al. (1981), Overland and Presendorfer (198) and Prensendorfer (1988) have used Monte Carlo smulatons wth ndependent random sequences to determne confdence levels for the EVs. The procedure nvolves repeatedly generatng sets of vectors of ndependent Gaussan random numbers wth the same dmenson and sample sze as the data beng analysed and then computng the EVs of ther dsperson matrces. The Monte Carlo EV curves are plotted for 95% and 5% levels of confdence. The normalzed EVs are calculated from the orgnal data EVs and then plotted aganst egenvector number. Sgnfcant data EVs are those that le above the Monte Carlo curve correspondng to the 95% level of confdence. The above procedure s applcable to a data set of smaller dmenson. When the dmenson of the data set s relatvely large (around 1 or above), the expense of fndng the EVs for the Monte Carlo test accordng to the above rule can be excessve, and asymptotc theory of EVs of large symmetrc random covarance matrces may be appled (Presendorfer et al., 1981). In ths theory, there are so many EVs that we only work wth fractonal amounts of total EVs, e.g. the hghest 95% of them. The detaled calculaton and the average asymptotc EVs for selected choces of the fractonal ndex are gven by Presendorfer et al. (1981). A potental problem wth the above two rules s that data may not be approxmately Gaussan, e.g. ranfall varable, whch s the precondton for these rules. The resamplng procedure wll not smulate accurately the physcal process that generated them and the results of the test may be msleadng. Hence, Presendorfer et al. (1981) have suggested, the above two tests should not necessarly elmnate any of the hgher components, whch may be found to be sgnfcant under other tests, especally f the EOF can represent the evoluton of a geophyscal process. As the present study deals wth a large data set, the asymptotc theory has been appled. North et al. (198) have used samplng theory to establsh error lmts for the EVs and have suggested that egenvectors whose EVs overlap form degenerate multplets and should not be splt when truncatng the egenvector sequence. The rotaton of EOFs (whether orthogonal or oblque) attempts to produce a smple structure. Rchman (1986, 1987) also argues for the rotaton of components and for the establshment of a smple structure n par-wse plots between the man EOFs so that each EOF used should, as far as possble, have hgh assocatons wth some orgnal axes and low assocatons wth others. In practce, f a smple structure exsts n the data then t s seen n the EOFs wth large loadngs on a few varables and near zero loadngs elsewhere. In the S-mode analyss, ths results n a clusterng or regonalzaton of the spatally dstrbuted varables, whch s useful for solatng specfc areas for further detaled study. Rotaton of T- mode components produces a clusterng of the temporally dstrbuted varables,.e. a compostng of smlar anomaly maps. Here, all the above technques are appled to fnd out the sgnfcant EOFs n both S- and T-modes. The sgnfcant EOFs n both the modes are rotated by the varmax method. The results from both methods are examned n ths study..3. Sgnfcant EOFs and physcal processes To fnd out the physcal processes that can be lnked to the sgnfcant EOFs, the percentage contrbutons of dfferent synoptc systems to the seasonal total monsoon ranfall over dfferent statons n Orssa durng 198 99 are calculated. For ths purpose, the frequences of days wth LPSC lke lows, depressons, cyclonc storms and cyclonc crculaton extendng up to the md-tropospherc level (cycr) over the dfferent regons (Fgure 1(b)) NW Bay, NW and the adjonng NE Bay, NE and the adjonng NW Bay, NW and the adjonng WC Bay, WC and the adjonng NW Bay, WC Bay off north coastal Andhra Pradesh (NCAP), Bangladesh, Orssa and the adjonng meteorologcal subdvsons lke GWB, east Madhya Pradesh, Chhatshgarh, NCAP and Jharkhand, etc. are found out from dfferent weather reports publshed by the Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1873 IMD. A day s consdered as an LPSC day for a regon f a low/depresson/cyclonc storm/cycr s found n the synoptc chart accordng to the 3 UTC observaton. Also, the days wth an all-inda break(weak) monsoon (AIBM) condton and monsoon trough wthout any sgnfcant embedded systems lke LPSC over the abovementoned regons are found out from dfferent weather reports publshed by the IMD. The past 4 h ranfall recorded at 3 UTC on the day followng the day of occurrence of each type of synoptc systems above s consdered to fnd out the total ranfall due to the dfferent synoptc systems durng 198 99. The percentage contrbuton to the seasonal monsoon ranfall over each staton from a synoptc system s calculated from the rato of total ranfall over that staton due to all the days of occurrence of that synoptc system and the total ranfall over that staton durng the whole perod (198 99). The spatal patterns of percentage contrbutons over dfferent statons n Orssa due to dfferent synoptc systems/combnaton of systems are analysed to fnd out the pattern of percentage contrbuton most smlar to the spatal pattern of each sgnfcant EOF. Also, the correlaton between the loadngs of each sgnfcant EOF (31 loadngs of each EOF correspondng to 31 statons) and the percentage contrbutons to seasonal monsoon ranfall over the 31 correspondng statons due to the dfferent synoptc systems are calculated. The synoptc systems/combnaton of synoptc systems for whch percentage contrbutons to seasonal monsoon ranfall show the hghest correlaton wth the loadngs of a sgnfcant EOF may be attrbuted to that sgnfcant EOF. 3.1. Mean daly ranfall dstrbuton 3. RESULTS AND DISCUSSION The mean daly monsoon ranfall dstrbuton n Orssa based on the data for 198 99 s gven n Fgure 3(a). The mean daly ranfall vares from 6. mm at Gopal Pur (Ganjam dstrct) to 11.5 mm at Pallahara (Dhenkanal dstrct n the northern upland). In general, the mean daly ranfall s less (< 8. mm) on the eastern sde of Eastern Ghat and the adjonng areas of south coastal Orssa. It s hgher ( 1 mm) on the western sde of Eastern Ghat, the western sde of the central rver basn and many parts of the northern upland except the Keonjhar dstrct. Ths pattern of dstrbuton may be due to the fact that Orssa generally gets ranfall n the monsoon season due to the LPSC developng over the northern Bay of Bengal wth the monsoon trough from the system extendng n a west-northwesterly drecton. As the basc monsoon flow becomes westerly n ths stuaton, the eastern and western sdes of Eastern Ghat become lee and wndward sdes respectvely wth respect to the basc westerly flow. Ths results n hgher ranfall over the western sde and less ranfall over the eastern sde of Eastern Ghat. The dmnshed ranfall actvty over some parts of the Keonjhar dstrct and the western part of Sambalpur may be attrbuted to smlar reasons due to the hll peaks to the west of these regons. The standard devatons n the daly monsoon ranfall over dfferent statons n Orssa are gven n Fgure 3(b). The standard devatons are generally less (<16 mm) over the eastern sde of Eastern Ghat n the southwest hlly regon and the adjonng areas of south coastal Orssa. It s also less over the northern upland, beng <16 mm over parts of the Keonjhar dstrct. It s hgher over the western sde of the Eastern Ghat. There s a regon of hgher standard devaton extendng from the central part of coastal Orssa n the southeast towards the Sambalpur dstrct n the northwest. The hgher varablty n ths regon may be due to the fact that: () the ranfall over Orssa s generally assocated wth the LPSC developng over the northern Bay wth the monsoon trough from the system extendng n a west-northwesterly drecton; and () the frequency of these LPSCs s hghly varable n tme and space wthn the season and also from year to year. It s also found that, over the eastern sde of Eastern Ghat, where the mean daly ranfall s small, the standard devaton s also small. There are about 1 days out of 44 days for any ndvdual staton durng whch the ranfall s nl, though there are only 36 days when all the statons under consderaton have receved nl ranfall smultaneously. Hence, the mean and standard devaton n daly ranfall are also calculated for each staton consderng only those days wth ranfall. It s found that the spatal patterns of mean and standard devaton are almost same as that shown n Fgure 3(a) and (b) respectvely. Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

1874 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA (a) Mean 1 1 (b) SD 8 1 8 8 1 4 16 1 8 >4 mm 1 8 >1 mm >8 mm, <1 mm <8 mm 16 16 >, <4 mm >16, < mm <16 mm Fgure 3. Daly monsoon ranfall (mm) over Orssa. (a) mean; (b)standard devaton (SD) 3.. Interstaton correlaton n daly monsoon ranfall over Orssa The correlaton coeffcent (CC) between daly ranfall recorded at dfferent statons durng 198 99 (44 days) s found to be postve for all the statons, except that the daly ranfall recorded over Paralakhemund (GNJ dstrct) has nsgnfcant negatve correlaton wth daly ranfall over Sambalpur, Jharsuguda (both n Sambalpur dstrct), Pallahara (Dhenkanal dstrct) and Panposh (Sundergarh dstrct). It also ndcates that the correlaton between ranfall over statons on the eastern sde of Eastern Ghat n the southwest hlly regon and adjonng areas of south coastal Orssa and ranfall over statons on the western sde of Eastern Ghat, adjonng the central rver basn and the western part of northern upland s relatvely lower. The nterstaton CC.3,.4 and.5 are llustrated n Fgure 4(a) (c). From Fgure 4(a), t can be concluded that the daly ranfall dstrbuton over most parts of Orssa durng the monsoon season s homogeneous and maybe assocated wth broad-scale features of the monsoon crculaton. The spatal coherence n the daly ranfall actvty over dfferent statons gradually decreases when consderng the hgher threshold values of CC lke.4 and.5, leadng to dfferent sub-homogeneous regons. Consderng CC >.4, there are two dfferent homogeneous regons, vz. () north Orssa and () the eastern sde of Eastern Ghat. Consderng CC >.5, the dfferent homogeneous regons are () the eastern sde of Eastern Ghat, () the western sde of Eastern Ghat, and () north coastal Orssa and the adjonng area. However, the regons based on the threshold values of nterstaton CC are not well defned and dstnct from each other. 3.3. Unrotated EOFs n S-mode analyss The results of the EOF analyss ndcate that 16 EOFs are requred to explan 8% of the total varance of the ranfall (Table I). The scree test and the plot of log(ev) ndcate a break n slope near EOF number 4 (Fgure 5(a) (b)). Accordng to asymptotc theory (Presendorfer et al., 1981), fve EOFs are sgnfcant (Fgure 5(c)). These frst fve EOFs explan about 54% of the total varance n daly monsoon ranfall over Orssa. Hence, an attempt has been made to nterpret the frst fve EOFs ndvdually to fnd out the large-scale features of the monsoon crculaton assocated wth monsoon ranfall over Orssa. The plots of frst fve EOFs are gven n Fgure 6. The loadngs of EOF 1, whch explan about 8.8% of total varance, are postve throughout the state (Fgure 6(a)). Though the ampltude vares n space, the values of EOF 1 are unform and consstent n nature. Also, there s a regon of maxma extendng from the Cuttack Pur dstrcts coasts n the eastsoutheast towards Sambalpur dstrct n the west-northwest. Ths patternrepresentsan actve/vgorous monsoon Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1875 (a) CC.3 (b) CC.4 (c) CC.5 Fgure 4. Interstaton correlaton coeffcents (CC) of daly monsoon ranfall over Orssa. (a) CC >.3; (b) CC >.4; (c) CC >.5 condton over Orssa. The spatal dstrbutons of percentage contrbuton of LPSC over dfferent regons, attrbuted to dfferent EOFs, are shown n Fgure 7. It s found that the spatal pattern of the total percentage contrbuton from LPSC over NW Bay and NW and the adjonng NE Bay to monsoon ranfall over dfferent statons (Fgure 7(a)) and the spatal pattern of EOF 1 (Fgure 6(a)) are smlar. Also, the CC between the loadngs of EOF 1 over dfferent statons and the percentage contrbutons of LPSC over NW Bay and NW and the adjonng NE Bay to monsoon ranfall over correspondng statons under consderaton s found to be.8, whch s hghly sgnfcant. The CCs >.35 and >.3 are sgnfcant at the 95% and 9% levels of confdence respectvely. Hence, EOF 1 may be assocated wth the LPSC over the NW Bay/NW and the adjonng NE Bay wth the monsoon trough from the system extendng n a west-northwesterly drecton across GWB or north Orssa. The above pattern s possble, as Orssa les n the southwest sector (sector of maxmum low-level convergence) of the LPSC over NW Bay or NW and the adjonng NE Bay. The contrastng feature on the eastern and western sdes of the Eastern Ghat wth loadngs on the eastern sde beng less than.1, though postve (not shown n fgure), s possble wth ths LPSC, as wth these systems the western and eastern sdes of Eastern Ghat behave as wndward and lee sdes respectvely wth basc monsoon flow as westerles. The loadngs of EOF, whch explan about 8.3% of the total varance, are negatve or nsgnfcantly postve over coastal Orssa and the adjonng areas. These are postve over nteror Orssa (Fgure 6(b)). Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

1876 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA 4 (a) 3 1 1 4 7 1 13 16 19 5 8 31 (b) 1 1 4 7 1 13 16 19 5 8 31 1 1 8 (c) 6 Normalsed egenvalue curve 4 1 Asymptotc curve 1 4 7 1 13 16 19 5 8 31 Fgure 5. (a) EVs (b) log(ev) and (c) normalzed EV curve versus asymptotc curve for average EVs of EOFs n S-mode The postve loadng s maxmum over the Sambalpur dstrct and the negatve loadng reaches a maxmum just along the coast. Ths type of ranfall dstrbuton s assocated wth below-normal actvty over coastal Orssa and slowly ncreasng actvty towards nteror Orssa. It s found that both the spatal patterns of EOF (Fgure 6(b)) and the percentage contrbuton of ranfall over dfferent statons due to the LPSC over GWB (Fgure 7(b)) are smlar. Also, the correlaton between the loadngs of EOF and the percentage contrbutons of ranfall over the correspondng statons due to the LPSC over GWB s.86, whch s hghly sgnfcant. Hence, the EOF may be assocated wth the LPSC over GWB and ts neghbourhood. The loadngs of EOF 3, whch explan about 7.9% of total varance, are postve over south Orssa, SBP dstrct and some areas of PRI and DNK dstrcts (Fgure 6(c)). These are negatve over the rest of Orssa. The postve loadng s hgher on the western sde of Eastern Ghat and the negatve loadng reaches a maxmum over the northernmost parts of Orssa, coverng the Balasore and Mayurbhanj dstrcts. Ths type of ranfall actvty, lke defcent ranfall over the northern part and excess ranfall over the southern part, may be attrbuted to the LPSC over Orssa/NW and the adjonng WC Bay wth the monsoon trough extendng from the system towards the west-northwest. The spatal dstrbuton of percentage contrbutons to seasonal ranfall Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1877 (a) EOF 1 (b) EOF..4....... (c) EOF 3.4.. (d) EOF 4......4... (e) EOF 5..4.... >.4 >., <.4 >, <. <, >.. <. Fgure 6. Unrotated EOFs of daly monsoon ranfall over Orssa n S-mode: (a) spatal pattern of EOF 1 ; (b) spatal pattern of EOF ; (c) spatal pattern of EOF 3 ; (d) spatal pattern of EOF 4 ; (e) spatal pattern of EOF 5 over dfferent statons n Orssa due to the LPSC over Orssa/NW and the adjonng WC Bay (Fgure 7(c)) and the dstrbuton of loadngs of EOF 3 (Fgure 6(c)) are smlar. Also, the CC between them s.86, whch s hghly sgnfcant. Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

1878 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA (a) 5 5 3 (b) 1 1 1 3 35 3 5 5 35 3 1 5 5 (c) 1 1 (d) 15 16 1 3 3 3 1 16 15 15 15 3 35 3 4 35 5 (e) 1.5 1.5 1.5 1.5 1 1 1.5 1.5 Fgure 7. Percentage contrbuton to total monsoon ranfall by dfferent systems attrbuted to EOFs n S-mode: (a) LPSC over NW Bay and NW and the adjonng NE Bay; (b) LPSC over GWB; (c) LPSC over Orssa and NW and the adjonng WC Bay; (d) only monsoon trough wthout any embedded system over Orssa and ts adjonng land/sea areas; (e) LPSC over NE and adjonng NW Bay The loadngs of the EOF 4, whch explan about 4.5% of the total varance, are negatve over a regon extendng from the Cuttack Pur dstrcts coasts and the adjonng area of Balasore n the southeast towards the Sambalpur dstrct n the west-northwest across the Dhenkanal dstrct (Fgure 6(d)). These are postve over the rest of Orssa. The postve loadngs gradually ncrease from both sdes of the regon of negatve loadngs. Ths type of ranfall dstrbuton s attrbuted to a weak monsoon stuaton over Orssa when only the monsoon trough exsts wthout any sgnfcant embedded systems over Orssa or ts neghbourng land/sea areas. Skka and Gadgl (198) have also dscussed the shftng of the maxmum cloud zone durng dfferent Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1879 Table I. The varance explaned by the frst 16 EOFs n S-mode EOF number Varance (%) Cumulatve total varance (%) 1 8.8 8.8 8.3 37.1 3 7.9 45. 4 4.6 49.6 5 4. 53.6 6 3.5 57.1 7 3. 6.1 8.8 6.9 9.5 65.4 1.4 67.8 11.3 7.1 1. 7.3 13. 74.3 14 1.9 76. 15 1.9 78.1 16 1.9 8. phases of the monsoon. Comparng the spatal pattern of percentage contrbuton of only the monsoon trough wthout any sgnfcant embedded systems over Orssa and ts adjonng land/sea areas to the seasonal ranfall (Fgure 7(d)) wth Fgure 6(d), both the patterns are smlar, thus supportng the above argument. Also, the CC between the loadngs and the percentage contrbuton to seasonal ranfall over correspondng statons from the above-mentoned synoptc system s.36, whch s sgnfcant. The loadngs of EOF 5, whch explan about 4% of the total varance, are postve over the regon, extendng from the eastern sde of Mayurbhanj dstrct n the northeast towards Bolangr and Kalahand dstrcts n the southwest (Fgure 6(e)). There s a negatve loadng n the rest of Orssa. Ths type of dstrbuton represents weak monsoon actvty over Orssa as a whole. It may be attrbuted to LPSC over NE and the adjonng NW Bay wth the monsoon trough from the system extendng northwestwards across Bangladesh. The LPSC over NE and the adjonng NW yelds a systematc pattern of weak monsoon ranfall over Orssa as a whole, even though t yelds good ranfall n some parts of Orssa. Comparng Fgure 6(e) wth Fgure 7(e), the spatal dstrbuton of percentage contrbutons of LPSC over NE and the adjonng NW Bay to seasonal ranfall over Orssa and the spatal pattern of EOF 5 are smlar. Also, the correlaton between loadngs and percentage contrbutons over correspondng statons due to ths system s.33, whch s sgnfcant at the 9% level of confdence. As there are a few studes devoted to EOF analyss wth daly ranfall, the study by Krpalan et al. (1991) usng pentad ranfall s compared wth the present study. Another study by Majumdar (1998) on weekly ranfall departure from normal s based on a smaller perod of data of 1 years and the study has a bas towards defcent ranfall. Comparng the results wth those obtaned by Krpalan et al. (1991), t s found that n both cases the frst EOF s assocated wth a strong/actve monsoon, ndcatng the predomnant role of the LPSC over NW Bay/NW and the adjonng NE Bay wth the monsoon trough from ths system extendng to northwest Inda for ranfall actvty over Inda and Orssa; the varance explaned by EOF 1 of the present study (8.8%) s hgher than the varance explaned by EOF 1 determned by Krpalan et al. (1991;.4%). Ths ndcates a more sgnfcant role of the LPSC over NW Bay/NW and the adjonng NE Bay on Orssa ranfall than that on all-inda ranfall. Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

188 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA 3.4. Rotated EOFs n S-mode analyss To ascertan the smple structure, par-wse plots of loadngs n both unrotated and rotated EOFs have been nspected. It s found that rotaton clearly succeeds n yeldng strong, smple structures. However, rotaton of the frst four EOFs yelds the smplest structure. The ntra-eof correlatons also ndcate that the rotated EOFs for rotaton of the frst four EOFs are most uncorrelated to each other. However, the rotatons are carred out for the frst two, three, four and fve EOFs to fnd out the sgnfcance. Consderng the values of EOFs >., the sopleths are drawn and based on these sopleths, daly ranfall affnty areas are determned for rotaton of the frst two, three, four and fve EOFs. The results of regonalzaton are shown n Fgure 8. Though the rotaton of the frst two EOFs yelds two dstnct regons wthout any overlappng, the regonalzaton does not cover most parts of Orssa (Fgure 8(a)). The rotaton of the frst three EOFs, though, yelds three regons (Fgure 8(b)): the regon () overlaps wth the southern parts of the regons () and (). The regonalzaton based on rotaton of the frst four EOFs (Fgure 8(c)) gves four dstnct regons wth mnmum area of overlappng. The regon of overlappng conssts of some parts of Bolangr and Kalahand dstrcts. The rotaton of the frst fve EOFs (Fgure 8(d)) yelds a more complcated structure, wth more regons of overlappng. Hence, the rotaton of four EOFs yelds dstnct regonalzaton wth a mnmum area of overlappng and (a) EOFs (b) 3 EOFs (c) 4 EOFs v (d) 5 EOFs v v v v v v v Fgure 8. Regonalzaton defned by. loadng sopleths n rotated S-mode EOFs; (a) rotaton of frst two EOFs; (b) rotaton of frst three EOFs; (c) rotaton of frst four EOFs; (d) rotaton of frst fve EOFs Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1881 dstnct geophyscal processes assocated wth dfferent regons. The lmtaton of ths regonalzaton s that t could not dstngush the western and eastern sdes of Eastern Ghat, whch show dfferent ranfall characterstcs n assocaton wth synoptc dsturbances, lke the LPSC over Orssa and the adjonng land/sea regons. Consderng the mean loadngs of the frst four unrotated and rotated EOFs over four dfferent regons obtaned by the above process (Table II), t s found that regon () s characterzed by a sgnfcantly hgher postve mean loadng of rotated EOF 1. Hence, regon () gets hgher ranfall than the other regons due to the LPSC over NW Bay/NW and the adjonng NE Bay. Regon () s characterzed by sgnfcantly hgher postve mean loadng of rotated EOF ; hence, compared wth the other regons, t gets more ranfall due to the LPSC over GWB. Regon () s characterzed by a sgnfcantly hgher postve mean loadng of rotated EOF 3 and, hence, gets more ranfall than the other regons due to the LPSC over Orssa/NW and the adjonng WC Bay. Regon (v) s characterzed by a sgnfcantly hgher postve mean loadng of rotated EOF 4 and, hence, gets more ranfall than the other regons due to the monsoon trough wthout any sgnfcant embedded systems over Orssa and the adjonng land/sea areas. 3.5. Unrotated EOFs n T-mode The scree test and the plot of log(ev) ndcate a break n slope near EOF number 5 (Fgure 9). The frst fve EOFs explan about 75% of the total varance n daly ranfall (Table III). The spatal dstrbutons of the frst fve unrotated EOFs n T-mode are depcted n Fgure 1. It s found that the values of EOF 1 (Fgure 1(a)), whch explan about 3.9% of total varance, are postve for all statons n Orssa except the eastern sde of Eastern Ghat. It s hgher over the western sde of Eastern Ghat, the adjonng areas of the central rver basn and the western sde of the northern upland. Lke EOF 1 of the S-mode, t may be assocated wth the composte anomaly of ranfall dstrbuton due to the LPSC over NW Bay and NW and the adjonng NE Bay wth the monsoon trough extendng from the system to a west-northwesterly drecton across GWB/north Orssa. The values of EOF (Fgure 1(b)), whch explan about 18.5% of the total varance, are postve over south Orssa except for some areas of the Pur dstrct. It s also postve over the adjonng areas of the northern upland and the central rver basn. The postve values are hgher over the western sde of Eastern Ghat and the adjonng areas of the central rver basn. The negatve values are hgher over north coastal Orssa and adjonng areas of the Mayurbhanj dstrct. Ths may be assocated wth the composte anomaly of ranfall due to the LPSC over Orssa/NW and the adjonng WC Bay wth the monsoon trough extendng west-northwestwards across the centre of the system lke the EOF 3 n S-mode. The values of EOF 3 (Fgure 1(c)), whch explan about 14.% of the total varance, are negatve over the regon extendng from Cuttack Pur dstrcts coasts n the east towards the Kalahand dstrct n the west. It s postve elsewhere, beng at a maxmum over the northernmost part of Orssa. Ths may be assocated wth the composte anomaly pattern of ranfall dstrbuton over Orssa due to the LPSC over GWB wth the monsoon trough extendng west-northwestwards from the centre of the system. So, t corresponds to EOF n S-mode. The values of EOF 4 (Fgure 1(d)), whch explan about 5.6% of the total varance, are postve over northeast Orssa and most parts of Eastern Ghat except the easternmost sde. The postve values are hgher over the extreme northeast of Orssa and some parts of the Sambalpur and Bolangr dstrcts. The negatve values are orented over the regons () extendng along coastal Orssa and () extendng from Cuttack Pur Table II. Mean values of the frst four EOFs n S-mode a Unrotated Rotated EOF1 EOF EOF3 EOF4 EOF1 EOF EOF3 EOF4 Regon ().1.17.1..34..4.1 Regon ().1.1.11.3.1.3.6.3 Regon ().1.11.1.1.3.3.9.6 Regon (v).16.3..6..1.9.36 a Mean values >. are n bold. Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

188 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA dstrcts coasts n the southeast towards the Sundergarh dstrct n the northwest. Lke EOF 4 n S-mode, ths may be attrbuted to the monsoon trough wthout any embedded system over Orssa and ts adjonng land/sea areas, but representng the composte anomaly pattern. The values of EOF 5 (Fgure 1(e)), whch explan about 3.7% of the total varance, are postve over a regon n north Orssa coverng parts of Dhenkanal, Keonjhar and Mayurbhanj dstrcts. It s also postve over the western sde of Eastern Ghat and the adjonng areas of Sambalpur dstrct. The values are negatve elsewhere. Lke EOF 5 n S-mode, ths pattern s attrbuted to the composte anomaly pattern of ranfall over Orssa due to the LPSC over NE and the adjonng NW Bay wth the monsoon trough extendng west-northwestwards across Bangladesh. 3.6. Rotated EOFs n T-mode The par-wse plots of loadngs n both unrotated and rotated EOFs suggest that rotaton of the frst fve EOFs yelds the smplest structure. The ntra-eof correlatons also ndcate that the rotated EOFs for rotaton of the frst fve EOFs are most uncorrelated to each other. However, the rotaton s carred out for the frst three, four, fve and sx EOFs. Consderng the values of rotated EOFs equal to., the sopleths are drawn and based on these sopleths, daly ranfall affnty areas are determned for varous rotatons. The results of regonalzaton are shown n Fgure 11. The regonalzaton due to rotaton of the frst three EOFs n T-mode does not cover most parts of Orssa (Fgure 11(a)). Also, some parts of the Sambalpur and Sundergarh dstrcts are the overlappng regon for () and (). The rotaton of the frst four EOFs results n four regons (Fgure 11(b)). However, regons () and (), () and (), and () and (v) overlap wth each other. The regon of overlappng ncreases from that of rotaton of the frst three EOFs. Consderng the rotaton of the frst fve EOFs, a smpler structure s obtaned wth mnmum area of overlappng (Fgure 11(c)). Consderng the rotaton of the frst sx EOFs, t s found that the structure becomes more complcated, wth a number of overlappng regons (Fgure 11(d)). Hence, the rotaton of the frst fve EOFs yelds the smplest structure and the regonalzaton based on ths rotaton s more apprecable. 6 5 (a) 4 3 1 1 4 7 1 13 16 19 5 8 31 1 (b) 1 4 7 1 13 16 19 5 8 1 Fgure 9. (a) EV and (b) log(ev) of EOFs n T-mode Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

RAINFALL VARIATIONS IN ORISSA DURING SUMMER MONSOON 1883 (a) EOF 1 (b) EOF...... (c) EOF 3. (d) EOF 4........... (e) EOF 5... >. >, <. <, >. <. Fgure 1. The spatal pattern of unrotated EOFs n T-mode: (a) EOF 1 ;(b)eof ;(c)eof 3 ;(d)eof 4 ;(e)eof 5 The mean values of the frst fve unrotated and rotated EOFs over dfferent regons obtaned by the above process are gven n Table IV. It s found that, lke S-mode regonalzaton, each of the dfferent regons obtaned by rotaton of the frst fve EOFs s domnated by a sngle major rotated EOF. Regon () s characterzed by a sgnfcantly hgher postve mean value of rotated EOF 1 and, hence, gets more ranfall than other regons wth the LPSC over NW Bay/NW and the adjonng NE Bay. Regon () s characterzed by a sgnfcantly hgher postve mean value of rotated EOF and, hence, gets more ranfall than the other regons Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)

1884 M. MOHAPATRA, U. C. MOHANTY AND S. BEHERA Table III. Varance and cumulatve varance explaned by the frst 16 EOFs n T-mode EOF number Varance (%) Cumulatve varance (%) 1 3.9 3.9 18.5 51.4 3 14. 65.6 4 5.6 71. 5 3.7 74.9 6.6 77.5 7.4 79.9 8. 8.1 9 1.9 84. 1 1.6 85.6 11 1.5 87.1 1 1. 88.3 13 1.1 89.4 14 1.1 9.5 15 1. 91.5 16.9 9.4 wth the LPSC over Orssa/NW and the adjonng WC Bay. Regon () s characterzed by a sgnfcantly hgher postve mean value of rotated EOF 3 and, hence, gets sgnfcantly hgher ranfall than the other regons wth the LPSC over GWB. Regon (v) s characterzed by a sgnfcantly hgher postve mean value of rotated EOF 5 and, hence, gets sgnfcantly hgher ranfall than the other regons wth the LPSC over NE and the adjonng NW Bay. Regon (v) s characterzed by a sgnfcantly hgher postve mean value of rotated EOF 4 and, hence, gets sgnfcantly hgher ranfall than the other regons wth only the monsoon trough wthout any embedded system over Orssa and ts adjonng land/sea areas. Comparng the regonalzaton due to rotaton of sgnfcant EOFs n both S- and T-mode (Fgure 8(c) and Fgure 11(c)), t s found that the area of overlappng regons s less n the case of regonalzaton based on T-mode rotaton. Whereas T-mode rotaton successfully dfferentates the ranfall pattern on the western and eastern sdes of Eastern Ghat, the S-mode rotaton fals to do so. Consderng the mean ranfall dstrbuton (Fgure 3(a)), t s a fact that the ranfall dstrbutons on the eastern and western sdes of Eastern Ghat are dfferent from each other. Though the regon on the eastern sde of Eastern Ghat s not a separate homogeneous regon when consderng the loadngs >., t behaves lke a separate homogeneous regon when consderng the equal sgns of loadngs (ether all postve or all negatve) over dfferent statons for all the sgnfcant EOFs. Prasad and Sngh (1988) regonalzed Inda by assumng that the regon over whch the Table IV. Mean values of the frst fve EOFs n T-mode a Unrotated Rotated EOF1 EOF EOF3 EOF4 EOF5 EOF1 EOF EOF3 EOF4 EOF5 Regon ().11..1.7.9.34.3.4.3.1 Regon ().5.8.3..8.5.6.4.3 Regon ().6.19.1.5.3.5..47.7.8 Regon (v)...1.5.19.1.1.1.6.3 Regon (v).1.3.9.6.19.9..1.41.5 a Mean values >. are n bold. Copyrght 3 Royal Meteorologcal Socety Int. J. Clmatol. 3: 1867 1887 (3)