Links Between the Southern Oscillation Index and Hydrological Hazards on a Tropical Paci®C Island1

Links between the Southern Oscillation Index and Hydrological Hazards on a Tropical Paci®c Island1

James P. Terry,2 Rishi Raj,3 and Ray A. Kostaschuk4

Abstract: River ¯oods and hydrological droughts (low stream water resources) are a recurrent problem in different parts of Fiji,causing disruption and hardship for many rural communities. These extremes in ¯uvial behavior are associated with large seasonal variability in rainfall,generated by intense tropical storms in the wet season and prolonged rain failure in the dry season. Such conditions are linked to the in¯uence of El NinÄ o/Southern Oscillation (ENSO) in the Southwest Paci®c. The Southern Oscillation Index (SOI) is the climatic measure of the strength of ENSO activities and shows good correspondence with (1) tropical cyclone ¯ood magnitude and (2) critically low stream discharges after a 2-month time lag,in two of Fiji's main river systems. If ENSO conditions become more frequent or sustained in the future as some climate models sug- gest,then the SOI will be a useful tool for projecting in advance the severity of hydrological hazards,which can assist in disaster mitigation and management.

On the interannual timescale,the El and a high incidence of storms to different NinÄ o/Southern Oscillation (ENSO) is our parts of the tropical Paci®c. As sea surface planet's most powerful climatic phenomenon temperatures rise off the western coast of the (Hilton 1998). At intervals of about 5±7 yr, Americas,a tongue of warm water stretches there is a disturbance in the Walker atmo- back along the equator. Rainfall becomes abun- spheric circulation over the Paci®c Ocean and dant in this new low-pressure region,whereas a weakening of the southeast trade winds. the western Paci®c suffers rainfall failure and This causes a large pool of warm ocean water drought. The strength of ENSO activity is usually centered around New Guinea to surge expressed as the Southern Oscillation Index eastward across the equatorial Paci®c (SOI),which is a measure of monthly atmo- (Congbin Fu et al. 1986). This disturbance spheric pressure differences between Tahiti to the Paci®c Ocean and atmospheric system, and Darwin (see Ropelewski and Jones 1987, called El NinÄ o,lasts for more than a year and Allan et al. 1991). brings droughts,prolonged wet conditions, ENSO and Stream Behavior 1 Financial support provided by the University of the ENSO clearly has the potential to affect South Paci®c and the Natural Sciences and Engineering stream behavior in a major way through its Research Council of Canada. Manuscript accepted 13 October 2000. in¯uence on Paci®c-wide climatic patterns. In 2 Department of Geography,University of the recent years there has been a concerted focus South Paci®c,Suva,Fiji (fax: 679 301 487; E-mail: in hydrological studies to explore possible terry [email protected]). links between ENSO and various aspects of 3 Hydrology Section,Public Works Department, river ¯ow,including seasonality and extremes P.O. Box 3740,Samabula,Fiji (E-mail: [email protected]). 4 Department of Geography,University of Guelph, (e.g.,Moss et al. 1994,Chiew et al. 1998, Ontario,N1G 2W1,Canada (E-mail: rkostasc@ Waylen and Laporte 1999). Indications have uoguelph.ca). emerged in Costa Rica,eastern Australia, New Zealand,and parts of the Paci®c U.S. mainland that valid relationships between Paci®c Science (2001),vol. 55,no. 3:275±283 : 2001 by University of Hawai`i Press stream ¯ow and SOI,or other ENSO mea- All rights reserved sures such as sea surface temperatures,can be

275 276 PACIFIC SCIENCE . July 2001 established (Kahya and Dracup 1993,Moss faces the predominant trade winds and there- et al. 1994,Cayan et al. 1999,Krasovskaia fore receives much more precipitation than et al. 1999). the northwest side,which is rain-shadowed For the Fiji Islands in the tropical South- by interior highlands reaching elevations over west Paci®c,the most extreme high and low 1300 m. river ¯ows (i.e.,¯oods and droughts [here The climate is distinctly seasonal. The drought refers to hydrological drought: di- typical wet season extends from November to minished stream water resources after pro- April and the dry season from May to Octo- longed no or low rainfall]) are linked to the ber. Tropical cyclones are a common feature response of island ¯uvial systems to ENSO- of the wet months. From 1970 to 2000,40 induced tropical cyclones and periods of cyclones traversed Fiji island waters. These prolonged rain failure and are serious hy- severe storms often produce extreme rain- drological hazards because of their impacts on falls,especially over the interior of Viti Levu the physical and human environment. These because the volcanic mountains force oro- impacts include channel erosion and siltation, graphic lifting of the spiraling rain bands. destruction of property,damage to agricul- Cyclone Gavin in March 1997,for example, ture (threatening food security),and risks to produced a deluge. The highest-elevation human life and health (R.R.,J.P.T.,and J. weather station at 760 m received a maximum Rokovada,1998,paper presented to the 27th 24-hr rainfall of 610 mm,at 10-min in- Annual Conference of the South Paci®c tensities reaching 150 mm/hr (Terry and Applied Geoscience Commission,Suva,Fiji; Raj 1999). Widespread soil saturation and Terry and Raj 1999). Such impacts place rapid surface and subsurface runoff cause a a dif®cult socioeconomic burden on our high degree of hydrological short-circuiting. resource-limited island nation: the costs to Rivers respond with rapid and very large the Fiji Government of the 1997±1998 increases in discharge,and dangerous over- drought and 1999 ¯oods amounted to F$100 bank ¯oods are the result. million and F$40 million,respectively. Im- In the dry season,an uneven distribution proved mitigation and management of hy- of rain days and prolonged lack of moisture drological hazards is therefore a priority in can occur. Rainfall seasonality is more pro- Fiji,but this requires a better understanding nounced for the island's leeward northwest of the links between these hazards and ENSO region,which receives only 20% of the an- than currently exists. nual total in the dry months,compared with The aim of this study therefore was to ex- 33% for the windward side (Terry and Raj amine the relationship between the strength 1998). Rivers in the northwest may experi- of the ENSO signal measured by the SOI and ence very low stream base¯ows as a result, the magnitude of extreme high and low dis- which represents a severe depletion of avail- charges in rivers on the main island of Fiji. able water resources because the rural popu- The potential use of the SOI as a predictor lation depends mainly on surface water for for hydrological hazards in the future was domestic needs and watering livestock (Terry then considered. and Raj in press). Human suffering and economic hardship are the result. In years without strong ENSO activity, Fiji Climate and ENSO In¯uences Fiji's rainfall bene®ts from convection along Fiji's main island Viti Levu (Figure 1) is a the low-pressure South Paci®c Convergence mountainous volcanic island 17.5 south of Zone (SPCZ) that extends diagonally from the equator in the Southwest Paci®c. The near the Solomon Islands,across to Samoa, tropical latitude and in¯uence of the nearby the Cook Islands,and beyond (Salinger et al. warm Southern Equatorial Ocean Current 1995). At the start of El NinÄ o events,con- gives Viti Levu a wet/dry tropical climate. A vective storms and tropical cyclones affecting clear geographical distribution in rainfall is the islands are generated as the eastward- observed. The southeast side of the island migrating pool of warm sea surface tem- The Southern Oscillation Index and Hydrological Hazards . Terry et al. 277

Figure 1. Location of the Rewa and Ba Rivers on Viti Levu island,Fiji. peratures passes across northern Fiji waters. Both produced damaging ¯oods in many Then,as El NinÄ o conditions develop fully,an places (Terry and Raj 1999). The ¯oods were equatorward shift in the SPCZ (Hay et al. then followed by one of the worst droughts to 1993,Vincent 1994) away from the Fiji group af¯ict Fiji this century (Fiji Meteorological leads to prolonged dry conditions. The ex- Service 1998). Rain failed across successive ceptional El NinÄ o of 1997±1998 was a good dry seasons and the intervening wet season example of the variability in climatic and hy- (when precipitation is normally reliable),and drological impacts in Fiji. The event began river ¯ows reached record lows (Public with a succession of severe tropical cyclones: Works Department,Hydrology Section, Gavin in March 1997 and June in May 1997. unpubl. river ¯ow records). 278 PACIFIC SCIENCE . July 2001

Figure 2. Hydrograph for the Rewa River at Navolau during tropical cyclones Gavin and June in 1997.

materials and methods slopes. It also has a recorded history of major ¯oods caused by tropical cyclones: ¯ooding In this study,correlation analysis was used to has occurred 15 times during tropical cy- examine the relationship between SOI and clones between 1970 and 2000 (i.e.,one ¯ood extreme river behavior on Viti Levu. SOI every 2 yr on average),and the ¯uvial system data were obtained from the Fiji Meteoro- can be described as a ¯ood-dominated re- logical Service. The Rewa and Ba Rivers gime. The river hydrograph in Figure 2 (Figure 1) were selected for investigation be- illustrates the characteristics of the high- cause of the availability of reliable long-term magnitude ¯ood peaks produced by tropical ¯ow records from hydrological gauging sta- cyclones Gavin and June at the start of the tions operated by the Hydrology Section of 1997±1998 El NinÄ o. the Public Works Department,and because In earlier work we (Kostaschuk et al. in these are the main rivers on the wet (wind- press) found that tropical cyclones occur ward) and dry (leeward) sides of the island. more often in Fiji waters when the SOI is Topography is steep and rugged in the upper negative,due to the in¯uence of the Southern parts of both study basins,which rise to Oscillation on the usual spatial distribution of 1360 m at the Mount Victoria divide,but cyclone origin,with more storms spawned becomes more hilly nearer the coast. farther to the north and east (closer to Fiji) The Rewa River in the wet zone has the during El NinÄ os (Revell and Goulter 1986, largest drainage basin,covering 2900 km 2. Basher and Zheng 1995). However,cyclone Vegetation is mostly dense rain forest,with frequency in Fiji waters did not show a good subsistence cultivation of root crops and link to temporal patterns in the SOI. Here, cattle pasturing on ¯oodplains and adjacent therefore,we compare ¯ood peak magnitude The Southern Oscillation Index and Hydrological Hazards . Terry et al. 279

Figure 3. Peak ¯ood discharge (Q) in the Rewa River at Navolau during tropical cyclones versus SOI for the month that the cyclone occurred.

(rather than timing) produced by tropical curve has the best ®t to the data (r À0:57, cyclones in the Rewa River with the SOI P > 0.05,Table 1). For the analysis of low (Figure 3). ¯ows in the Ba River,dif®culties arise when The Ba River in the north of Viti Levu is comparing SOI and monthly means of dis- Fiji's third largest river system,covering 930 charge because of the large in¯uence of sea- km2. It is the most important river in the dry sonality on river behavior. Thus,poor cor- zone,but experiences very low base¯ows in relations were derived between SOI versus periods of drought. Savanna grassland vege- monthly Q (r 0.17),prior 3-month mean of tates the steep highland terrain,whereas the SOI versus monthly Q (r 0.21),SOI versus more gentle coastal slopes are widely used for Q for the month with the lowest ¯ow each growing sugarcane. Figure 4 compares tem- year (r 0.18),and lowest monthly SOI in poral patterns in the SOI with mean dis- year versus lowest monthly Q in year charge (Q) in the Ba River. (r 0.23). However,once ¯ow seasonality is removed by plotting 13-month running means,low ¯ow trends in the Ba River show results good correspondence with El NinÄ o events, For the Rewa River,Figure 3 indicates that although lagging behind by approximately 2 peak river discharge produced by tropical cy- months (r 0.73, P > 0.001,Figure 5). clones is negatively related to the strength of the SOI (i.e.,tropical cyclones that occur discussion during El NinÄ o events are more likely to produce larger ¯oods). A curve-®tting exer- Despite the fact that it remains dif®cult to cise showed that an exponential regression predict the occurrence of tropical cyclones 280 PACIFIC SCIENCE . July 2001

Figure 4. A comparison of the temporal patterns in SOI and Ba River ¯ow (Q) behavior at Toge,1980±2000.

TABLE 1 SOI Relationship with Peak Flood Discharge in the Rewa River

Regression Curve Type Equation Correlation

Linear Q À531:94SOI 3529:70 r À0:49 Polynomial Q À129:33SOI2 À 670:03SOI 3719:7 r À0:54 Exponential Q 3295.50eÀ0:154SOI r À0:57

near the Fiji Islands (as it does elsewhere),our vulnerable sections of the rural community work suggests that the strength of the SOI at who live on low-lying areas near ¯oodplains the time of future storms can give some in advance of the expected ¯ood wave. This is warning of the size of the ¯ood in the Rewa important because in the future ¯oods may River,which has the largest and most popu- become more severe due to increasing tropi- lated river basin in Fiji. Because tropical cy- cal cyclone intensity in the Paci®c region if clones rarely develop very close to Fiji,but the climate changes toward more sustained instead tend to generate nearer the equator El NinÄ o±like conditions,as many climate and then approach on a southerly track over 2 scientists now predict (see Holland 1997, or 3 days,this delay will be useful in alerting Trenberth and Hoar 1997,Timmermann The Southern Oscillation Index and Hydrological Hazards . Terry et al. 281

Figure 5. Correlation between 13-month running means of Ba River ¯ow (Q) and 5-month running means of SOI with a 2-month lag.

et al. 1999; P. Whetton,R. Jones,K. Hen- supply (e.g.,carting by truck) before resource nessy,R. Suppiah,K. Walsh,and W. Cai, scarcity becomes critical. Elsewhere in Fiji's 2000,paper presented to the Paci®c Islands dry zone,there may be different lag times Climate Change Conference,Rarotonga, between low SOI and low river base¯ow. Cook Islands). In this scenario,projections of This needs to be investigated so that pro- the likely ¯ood hazard magnitude shall be- jected stream water resources in times of rain come more crucial for the success of disaster failure can be assessed on a catchment-by- reduction programs in Fiji. catchment basis. For future droughts that are related to the ENSO phenomenon (rather than random dry acknowledgments years),forecasts of the depletion of surface water resources in the Ba River can be made The Fiji Meteorological Service is thanked with a high degree of con®dence 2 months for providing the SOI record. This paper was ahead on the basis of SOI values. The Ba originally presented to the Paci®c Islands River has the largest catchment in the Conference on Climate Change,Climate drought-prone northwest of Viti Levu and Variability,and Sea-Level Rise,organized by supports a substantial population of sugarcane the South Paci®c Regional Environmental farmers. Such forecasts will be valuable for Programme,17±21 April 2000,Rarotonga, government planning of alternative water Cook Islands. The Department for Inter- 282 PACIFIC SCIENCE . July 2001 national Development of the British Govern- Paci®c. School of Social and Economic ment provided generous funding through the Development,University of the South University of the South Paci®c for J.P.T. to Paci®c,Suva,Fiji. attend this meeting. The written paper was Holland,G. J. 1997. The maximum potential prepared while J.P.T. visited the University intensity of tropical cyclones. J. Atmos. of Sydney School of Geosciences and thanks Sci. 54:2519±2541. are extended to several colleagues there for Kahya,E.,and J. A. Dracup. 1993. U.S. their constructive comments on the work. stream¯ow patterns in relation to the El Two referees also provided useful suggestions NinÄ o/Southern Oscillation. Water Resour. on the original manuscript. Res. 29:2491±2503. Kostaschuk,R.,J. Terry,and R. Raj. The Literature Cited impact of tropical cyclones on river ¯oods in Fiji. Hydrol. Sci. J. (in press). Allan,R. J.,N. Nicholls,P. D. Jones,and I. J. Krasovskaia,I.,L. Gottschalk,A. Rodrõ Âdguez, Butterworth. 1991. A further extension and S. Laporte. 1999. Dependence of the of the Tahiti-Darwin SOI,early ENSO frequency and magnitude of extreme events and Darwin pressure. J. Clim. ¯oods in Costa Rica on the Southern 4:743±749. Oscillation Index. Pages 81±89 in L. Basher,R. E.,and X. Zheng. 1995. Tropical Gottschalk,J.-C. Olivry,and D. Reed,eds. cyclones in the Southwest Paci®c: Spatial Hydrological extremes: Understanding, patterns and relationships to Southern predicting,mitigating. International Asso- Oscillation and sea surface temperature. ciation of Hydrological Sciences. Publica- J. Clim. 8:1249±1260. tion 255. Cayan,D. R.,K. T. Redmond,and L. G. Moss,M. E.,C. P. Pearson,and A. I. Riddle. 1999. ENSO and hydrologic ex- McKerchar. 1994 The Southern Oscilla- tremes in the western United States. J. tion index as a predictor of the probability Clim. 12:2881±2983. of low stream¯ows in New Zealand. Water Chiew,F. H. S.,T. C. Piechota,J. A. Dracup, Resour. Res. 30:2717±2723. and T. A. McMahon. 1998. El NinÄ o/ Revell,C. G.,and S. W. Goulter. 1986. Southern Oscillation and Australian rain- South Paci®c tropical cyclones and the fall,stream¯ow and drought: Links and Southern Oscillation. Mon. Weather Rev. potential for forecasting. J. Hydrol. 204: 114:1138±1145. 138±149. Ropelewski,C. F.,and P. D. Jones. 1987. An Congbin Fu,H. F. Diaz,and J. O. Fletcher. extension of the Tahiti-Darwin Southern 1986. Characteristics of the response of Oscillation Index. Mon. Weather Rev. the sea surface temperature in the central 115:2161±2165. Paci®c associated with warm episodes of Salinger,M. J.,R. E. Basher,B. B. Fitzharris, the Southern Oscillation. Mon. Weather J. E. Hay,P. D. Jones,J. P. MacVeigh,and Rev. 114:1716±1738. I. Schmidely-Leleu. 1995. Climate trends Fiji Meteorological Service. 1998. Annual in the south-west Paci®c. Int. J. Climatol. weather summary. Climatological Divi- 15:285±302. sion,Nadi Airport,Fiji. Terry,J. P.,and R. Raj. 1998. Hydrological Hay,J.,J. Salinger,B. Fitzharris,and R. drought in western Fiji and the contribu- Basher. 1993. Climatological seesaws in the tion of tropical cyclones. Pages 73±85 in Southwest Paci®c. Weather Clim. 13:9±21. J. P. Terry,ed. Climate and environmental Hilton,A. C. 1998. The in¯uence of El change in the Paci®c. School of Social and NinÄ o±Southern Oscillation (ENSO) on Economic Development,University of the frequency and distribution of weather- South Paci®c,Suva,Fiji. related disasters in the Paci®c islands ÐÐÐ. 1999. Island environment and land- region. Pages 57±71 in J. P. Terry,ed. scape responses to 1997 tropical cyclones Climate and environmental change in the in Fiji. Pac. Sci. 53:257±272. The Southern Oscillation Index and Hydrological Hazards . Terry et al. 283

ÐÐÐ. The 1997±98 El NinÄ o and drought Trenberth,K.,and T. J. Hoar. 1997. El Nin Ä o in the Fiji Islands. In J. Gladwell,ed. Hy- and climate change. Geophys. Res. Lett. drology and water management in the 24:3057±3060. humid Tropics: Hydrological research is- Vincent,D. G. 1994. The South Paci®c sues and strategies for water management. Convergence Zone (SPCZ): A review. UNESCO-IHP (in press). Mon. Weather Rev. 112:1949±1970. Timmermann,A.,J. Oberhuber,A. Bacher, Waylen,P.,and M. S. Laporte. 1999. Flood- M. Esch,M. Latif,and E. Roeckner. 1999. ing and the El NinÄ o±Southern Oscillation Increased El NinÄ o frequency in a climate phenomenon along the Paci®c coast of model forced by future greenhouse warm- Costa Rica. Hydrol. Processes 13:2623± ing. Nature (Lond.) 398:694±697. 2638.