|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 A CLUSTER ANALYSIS OF TROPICAL CYCL ONE TRAJECTORIES IN THE SOUTH INDIAN OCEAN: THE INFLUENCES OF THE EL NINO SOUTHERN OSCILLATION AND THE SUBTROPICAL INDIAN OCEAN DIPOLE By KEVIN D. ASH A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2010
2 2010 Kevin D. Ash
3 To the Captain of the cargo vessel in the Indian Ocean who did not listen to my advice and motivated me to learn more
4 ACKNOWLEDGMENTS I would like to first extend thanks to my parents and my brother for their love and support. I am also grateful to Ann Foster and the United States Geological Survey for employing me during my first year in Gainesville and affording me the opportunity to earn a decent paycheck while in graduate school. I am appreciative as well to the Department of Geography for funding and the opportunity to work as a Teaching Assist ant during my second year I am grateful to the Poehling family and all who graciously supported the establishment of a new fellowship this year in memory of their son, Ryan. Many thanks as well to my thesis committee members Peter Waylen and Tim Fik for their valuable advice. Finally, thank you to my committee chair and advisor Corene Matyas for her dedication and wisdom in challenging and guiding me during the last two years.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 LIST OF ABBREVIATIONS ........................................................................................... 10 ABSTRACT ................................................................................................................... 11 CHAPTER 1 INTRODUCTION .................................................................................................... 13 2 LITERATU RE REVIEW .......................................................................................... 21 Tropical Cyclones of the South Indian Ocean ......................................................... 21 Introduction ....................................................................................................... 21 Brief History and Early Research ..................................................................... 21 Basic Tropical Cyclone Climatology ................................................................. 23 A Climatologically Favorable Environment ....................................................... 24 El Nio Southern Oscillation ................................................................................... 27 Introduction ....................................................................................................... 27 ENSO Influences in the South Indian Ocean .................................................... 28 Tropical Cyclones and El NioSouthern Oscillation ............................................... 31 Subtropical Indian Ocean Dipole ............................................................................ 35 Tropical Temperate Troughs .................................................................................. 37 Conclu sion .............................................................................................................. 39 3 A CLUSTER ANALYSIS OF SOUTH INDIAN OCEAN TROPICAL CYCLONE TRAJECTORIES .................................................................................................... 49 Introduction ............................................................................................................. 49 Tropical Cyclone Data and Study Area Definition ................................................... 50 Cluster Analysis in Tropical Cyclone Research ...................................................... 52 The Clustering Procedure ....................................................................................... 54 Discussion .............................................................................................................. 58 Eastern Formation Region: C1 and C6 ............................................................ 58 Central Formation Region: C2 and C7 ............................................................. 5 9 Western Formatio n Region: C3 and C4 ........................................................... 59 Far West/Mozambique Channel: C5 ................................................................ 60 Conclusion .............................................................................................................. 60
6 4 THE INFLUENCES OF EL NINO SOUTHERN OSCILLATION AND THE SUBTROPICAL INDIAN OCEAN DIPOLE .............................................................. 71 Introduction ............................................................................................................. 71 Data and Methods .................................................................................................. 75 Results .................................................................................................................... 78 Kruskal Wallis ANOVA Res ults ........................................................................ 78 Multiple Comparisons and SST Anomaly Composites ..................................... 78 Discussion .............................................................................................................. 85 Conclusion .............................................................................................................. 89 5 CONCLUSIONS ................................................................................................... 108 APPENDIX TROPICAL CYCLONES ANALYZED IN THE PRESENT STUDY .............................. 113 LIST OF REFERENCES ............................................................................................. 119 BIOGRAPHICAL SKETCH .......................................................................................... 131
7 LIST OF TABLES Table page 1 1 Tropical c yclone close p asse s in the South Indian Ocean ................................. 20 3 1 Seven main cluster s of SIO TCs ........................................................................ 62 3 2 Frequency of TC genesis by month for C1C7. .................................................. 62 4 1 P values for KW ANOVA and ModifiedLevene Equal Variance Tests .............. 92 4 2 Standardized median anomalies for N3.4 N1.2, N4, and SDI ........................... 92 4 3 Multiple c omparison t ests for Ni o3.4 stratified by clusters. ............................. 92 4 4 Multiple comparison t ests for Nio 1.2 stratified by clusters. .............................. 93 4 5 Multiple comparison t ests for Nio 4 stratified by clusters. ................................. 93 4 6 Multiple comparison t ests for SDI stratified by clusters. .................................... 93 4 7 Tropical cyclones by cluster and combined ENSO/SDI phases. ......................... 94
8 LIST OF FIGURES Figure page 1 1 South Indian Ocean region map. ....................................................................... 20 2 1 Annual frequencies of South Indian Ocean tropical cyclones ............................ 41 2 2 Monthly frequencies of South Indian Ocean tropical cyclones .......................... 42 2 3 South Indian Ocean tropical cyclone trajectories ............................................... 43 2 4 Indian Ocean Sea Surface Temperatur es, averaged from 19792008. ............. 44 2 5 Indian Ocean Outgoing Longwave Radiati on, averaged from 19792008. ......... 45 2 6 Difference of El Nio and La Nia SST patterns in the SIO. ............................... 46 2 7 Difference of El Nio and La Nia SLP patterns in the SIO. ............................... 47 2 8 Difference of El Nio and La Nia 500 hPa zonal wind patterns in the SIO. ...... 48 3 1 Dendrogram of S IO TCs when clustered by initial longitude ............................. 63 3 2 Map of five cluster solution of S I O TC s when c lustered by initial longitude. ...... 64 3 3 Dendrogram of central region SIO TC s when clustered by final latitude and longitude. ............................................................................................................ 64 3 4 Dendrogram of western region SIO TC s when clustered by final latitude and longitude. ............................................................................................................ 65 3 5 Dendrogram of eastern region SIO TC s when clustered by final latitude and longitude. ............................................................................................................ 66 3 6 Tropical cyclone trajectories in Cluster 1 (C1) .................................................... 67 3 7 T ropical cyclone t rajectories in Cluster 6 (C6) .................................................... 67 3 8 T ropical cyclone trajectories in Cluster 2 (C2). ................................................... 68 3 9 T ropical cyclone trajectories in Cluster 7 (C7). ................................................... 68 3 10 T ropical cyclone trajectories in Cluster 3 (C3) .................................................... 69 3 11 T ropical cyclone trajectories in Cluster 4 (C4) .................................................... 69 3 12 T ropical cyclone trajectories in Cluster 5 (C5) .................................................... 70
9 4 1 El Nio sea surface temperature index regions.. ................................................ 94 4 2 Subtropical Dipole Index SST regions ............................................................... 94 4 3 Box plots for the Nio 3.4 region stratified by cluster ID. .................................... 95 4 4 Box plots for the Nio 1.2 region stratified by cluster ID. .................................... 96 4 5 Box plots for the Nio4 region stratified by cluster ID. ....................................... 97 4 6 Box plots for SDI stratified by cluster ID. ............................................................ 98 4 7 Composite map of SSTA for Cluster 4 (C4) ....................................................... 99 4 8 Composite map of SSTA for C luster 3 (C3). ..................................................... 100 4 9 Difference of SSTA between C4 and C3. ......................................................... 101 4 10 Composite map of SSTA for Cluster 6 (C6). ..................................................... 102 4 11 Composite map of SSTA for Cluster 1 (C1). ..................................................... 103 4 12 Difference of SSTA between C1 and C6. ......................................................... 103 4 13 Composite map of SSTA for Cluster 2 (C2). ..................................................... 104 4 14 Composite map of SSTA f or Cluster 7 (C7). ..................................................... 105 4 15 Difference of SSTA between C7 and C2. ......................................................... 106 4 16 Composite map of SSTA for Cluster 5 (C5). ..................................................... 107
10 LIST OF ABBREVIATION S ANOVA analysis of variance CA cluster analysis ENSO El Nio Southern Oscillation hPa hectopascals ITCZ inter tropical convergence zone JTWC Joint Typhoon Warning Center m s1 meters per second MSLP mean sea level pressure NCEP National Centers for Environmental Prediction NH Northern Hemisphere NOAA National Oceanic and Atm ospheric Administration RSMC Regional Specialized Meteorological Center SH Southern Hemisphere SICZ South Indian Convergence Zone SIO South Indian Ocean SLP sea level pressure SO Southern Oscillation SST sea surface temperature SSTA sea surface temperature anomaly TC tropical cyclone TTT tropical temperate trough W m2 Watts per square meter WMO World Meteorological Organization WNP Western North Pacific
11 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science A CLUSTER ANALYSIS OF TROPICAL CYCLO NE TRAJECTORIES IN THE SOUTH INDIAN OCEAN: THE INFLUENCES OF THE EL NINO SOUTHERN OSCILLATION AND THE SUBTROPICAL INDIAN OCEAN DIPOLE By Kevin D. Ash May 2010 Chair: Corene Matyas Major: Geography Tropical cyclones (TCs) are a regular feature over the South Indian Ocean (SIO) during the austral warm season from November to April. The storms often pass in close proximity to the islands o f Madagascar, Mauritius, and Reunion, or pass through the Mozambique Channel. In addition to threatening human lives, cyclones are also capable of negative societal impacts in this economically developing region, ruining crops and disrupting the regions busy shipping lanes. It is therefore important to investigate the causes and controls that shape the tracks of the SIO TCs. The goal of this research is to relate variability of TC motion to the oceanic atmospheric circulation. C luster analysis is employ ed to group SIO TC trajectories by their initial and final positions. The clusters are then used to compare group median index values that represent El NioSouthern Oscillation (ENSO) and the Subtropical Indian Ocean Dipole (SIOD) using Kruskal Wallis an alysis of variance (ANOVA) on ranks and post hoc multiple comparisons. ENSO is known to influence TC trajectories in the SIO through modulation of the semi permanent SIO subtropical high. However, though ENSO plays a role in altering the oceanic atmospheric environment within the basin, it
12 is not always simultaneously associated with either a strong or weak subtropical high. The SIOD is known to exert strong influence on precipitation patterns in the SIO region, also through modulation of the SIO subtropical semipermanent high. Both ENSO and SIOD have likewis e been linked to shifts in the austral summer tropical temperate troughs (TTTs) from the African continent to northeast of Madagascar over the western SIO. The results of the multiple comparison tes ts in this study suggest that both ENSO and SIOD are significantly associated with TC trajectories. This significant link between SIOD and SIO TCs has not been documented in the known SIO TC literature, and this study proposes TTTs as the physical mechani sm responsible for the strong eastward component of TC motion when ENSO is in warm phase and SIOD is in negative phase. When ENSO is in cool phase and SIOD is simultaneously in positive phase, TC trajectories tend to show strong westward movement, remain farther north and do not recurve into the mid latitude westerlies. This study is important because the fi ndings suggest that anti phase interactions of ENSO and SIOD frequently associate with anomalous westward or eastward TC trajectories
13 CHAPTER 1 INTRODUCTION Tropical cyclones (TCs) are among the most devastating natural forces on earth. On a global scale, they are responsible for a large portion of casualties and damages attributed to natural hazards (Emanuel, 2005b). TCs are broadly defined as warm core low pressure systems originating over tropical water bodies which derive their energy largely from heat fluxes between the warm ocean surface and the overlying atmosphere (Emanuel, 2003). The ter m tropical cyclone is used throughout this document as a general term encompassing tropical systems of varying levels of intensity including hurricane intensity (> 33 m s1) tropical storm intensity (>17 m s1) and tropical depression intensity (<17 m s1), all using the United States (US) standard 1minute average sustained wind speed. Since the TC data source utilized in this study is the US Joint Typhoon Warning Center (JTWC) best tracks, no further explanation is necessary regarding the differing criteria used by various other meteorological agencies in their respective TC nomenclatur es C hapter 3 provides further details of a more precise definition of a TC as relevant to the analyses presented herein. W hether intense or weak, landfalling or in the open ocean, TCs have great capacity to threaten human life and disrupt society with their combinations of fast winds, heavy rain, and coastal water rise (Pielke and Pielke, 1997) Thus scientists, social policy makers, military commanders, and even investors maintain great interest in improvements of TC prediction both in real time and on longer time horizons TCs follow variable paths, or trajectories, from their development regions over tropical water s, but frequently they exhibit trajectories with some component of westerly and poleward movement. Quite often, TCs that stray into the s ubtropics begin to take
14 recurving poleward trajectories and move with an easterly component, influenced by mid latitude westerly jet streams (Knaff, 2009). T o fully understand TCs is to be able to consistently predict thei r frequency, movement and intensity within useful and acceptable margins of error The ability to predict may then afford greater warning lead time which can save lives or mitigate profit loss (Murnane, 2004) The South Indian Ocean ( SIO) is an important region for TC research an d accounts for about 14% of the average annual global TC activity, a similar percentage to the North Atlantic (Jury, 1993) The peoples of southern Africa, Madagascar, and the Mascarene Islands (Figure 11) are vulnerable to repeated TC strikes and their concomitant hazards (Table 11) Madagascar has endured many devastating events, particularly on the northern half of the island Landfalls of intense TCs Andry and Kamisy (both Saffir Simpson category 4) impacted northern Madagascar in 198384 (Jury et al ., 1993). Successive storm surges up to six meters accompanied three TCs (Daisy, Geralda, and Litanne) that made landfall within the same 100km stretch of coastline early in 1994 (Naeraa and Jury, 1998). These systems destroyed critical shipping and ene rgy infrastructure s, brought flooding rainfall, and caused hundreds of deaths (Chang Seng and Jury, 2010b) Northern Madagascar is also a leading vanillaproducing region and is subject to widespread agricultural losses, as transpired in 2000 and again in 2007 (Brown, 2009). Southern Africa has seen the adverse impacts of TCs as well, though landfalls are much less frequent than in Madagascar. Reason and Keibel (2004) observed that less than 5% of cyclones in the SIO made landfall on the east coast of Afr ica over the period 19542004. Still, Mozambique has endured unusually severe impacts from TCs in a few
15 instances, with widespread flooding from rainfall causing the most casualties and damage. Cyclone Eline in 2000 struck as serious seasonal flooding was already underway and exacerbated the catastrophic flooding across the region, killing hundreds and affecting millions of people (Vitart et al ., 2003; Reason and Keibel, 2004). Disastrous flooding in Mozambique from heavy rainfall also accompanied the pas sages of Cyclone Dera in 2001 and Favio in 2007 (Reason, 2007; Klinman and Reason, 2008). The larger inhabited islands of the Mascarene Archipelago (Mauriti us and Runion) are precariously positioned with respect to SIO TCs TC Dina devastated Runion in 2002 and TC Hollanda likewise wrought widespread damage on Mauritius, though casualties in both cases were few, a testament to the well organized and executed cyclone preparedness and warning systems of these islands (Parker, 1999; Roux et al ., 2004) Whi le these are relatively small islands with small populations, their locations are strategic along busy shipping lanes. SIO TCs are a significant threat to the lives of ship crews and their expensive cargoes, as hundreds of large commercial vessels sail da ily around the Cape of Good Hope and across the region between economic hubs in Europe and the Americas and in South Asia and the Far East (Roberts and Marlow, 2002; Chang Seng and Jury, 2010b) TCs of the SIO have not been scrutinized as closely as those of the Northern Hemisphere or the Australian region over the past thirty years (Jury, 1993). Most recently, improvements in technology and gathering volumes of reliable observations in this economically poor and once datasparse region are allowing for an increase of climatological studies of SIO TCs. G reat leaps forward in weather/climate monitoring
16 and forecasting have followed from a succession of novel inventions in communications, transportation, space b orne remote sensing, and computer science over the past 200 years. The advent of satellite remote sensing particularly revolutionized tropical cyclone science beginning from the launch of the experimental weather satellite TIROS I in 1960, which provided the first pictures of a TC from space (Fritz and Wexler, 1960) Weather satellite images now allow atmospheric scientists to locate and track TCs with greater precision and even estimate their intensities based on cloud patterns (Dvorak, 1975). T hese tec h nologies have not been evenly deployed across the globe, thus the accuracy and lengths of record of TC climatologies are also unequal across the earths oceans. In the SIO, the infrared satellite imagery needed to apply the Dvorak intensity estimation tec hnique has been available since about 1980; therefore, reliable measures of TC positions and intensities begin from that time (Knaff and Sampson, 2009; Chang Seng and Jury, 2010a). Another significant improvement for SIO TC observation was realized in 199 8 with the placement of the Meteosat geostationary satellite over the region (Caroff, 2009; Chang Seng and Jury, 2010a). With a useable record back to 1980, and as TC observation capabilities continue to evolve, there is sufficient data quantity and quali ty to study TC s of the SIO in greater depth. Using these satellite observations, this research focuses on the trajectories or tracks of SIO TCs since 1979. Much scholarly research has previously focused on TC movement in order to improve short term prediction of individual storm tracks, arguably the most important aspect of TC prediction (Chan, 2005) One method is to predict a TCs movement by comparing it to several similar TCs in the historical record in terms of
17 storm location, direction and speed of forward movement, and intensity (Hope and Neumann, 1970 ; Neumann and Randrianarison, 1976; Bessafi et al ., 2002 ). This may be referred to as a statistical analog model or a climatology and persistence model (CLIPER), a method long used and, though i ncreasingly falling into disfavor for short term prediction, still used as of 2005 at the Joint Typhoon Warning Center (JTWC) as a baseline for assessing TC track predictions (Sampson et al ., 2005). The preferred method for short term prediction is now bas ed upon physical equations to model oceanic and atmospheric variables that influence the net directional heading and forward speed of each TC There are two principal factors that account for TC movement. One is th e latitudinal variation of the Coriolis e ffect across a cyclone s wind field inducing poleward advection of positive vorticity which appears in aggregation over time as gentle poleward bending of the TCs track (Chan and Gray, 1982; Elsner and Kara, 1999). This is referred to as beta drift or the beta effect (Chan, 2005). The second and more substantial influence on TC movement is the environmental steering flow. Essentially, a TC vortex is embedded within the larger scale atmospheric wind pattern and carried along with the mean flow (Chan and Gray, 1982; Chan, 2005). Dominant steering features include subtropical ridges of high pressure an d troughs of low pressure. A TCs movement may be approximated using the environmental midtropospheric flow surrounding a TC vortex at 700, 650, or 500 hP a (Chan and Gray, 1982; Wang and Holland, 1996) A weaker and less vertically developed TC is steered more by lower tropospheric flow, whereas a strong and vertically stout TC is steered by deep layer tropospheric flow (Wang and Holland, 1996).
18 It is at this point that this study can be viewed in the proper context Physical numerical modeling of TC movement has proven to be more skillful overall than climatology and persistence (Chan, 2005). Without a doubt, this is invaluable for short term TC predicti on warning, and risk mitigation. However, statistical and climatological methods are still qu ite useful for longer term prediction. Given the vast improvements in TC observation techniques and computing abilities over the past fifty years, statistical a nalysis of a spatially and temporally homogenous global TC dataset can provide a first approximation for prediction of TC frequency and movement on intraseasonal or perhaps even interannual time scales (Klotzbach et al ., 2010) Such statistical predictions of TC frequency have shown promise over the past thirty years, though substantial improvement s are still possible (Hastenrath, 1995; Klotzbach 2007) Much of the improvement already made in seasonal TC frequency prediction owes to greater understanding of the scope and influence of larger scale, low frequency modes of oceanatmosphere variability such as the El NioSouthern Oscillation (ENSO) (Klotzbach et al ., 2010). Research has recent ly turned to ENSO and ot her global and regional oceanatmosphere modes as likely drivers of variability in steering flow on intraseasonal and interannual time scales. Statistical associations linking these modes of variability to identifiable patterns of TC trajectories can spur future research and inform longer term trajectory predictions e ven if the physical mechanisms between the two cannot be explicitly described initially. Furthermore, statistical linkages may be used for longer term disaster planning purposes, if the link ages are strong and stable enough for reliable predictions.
19 The goal of this study is to enhance understanding of South Indian Ocean (SIO) TC activity by testing the hypotheses that sea surface temperature anomalies (SST A ) in the equatorial Pacific Ocean and/or the tropical and subtropical Indian Ocean (IO) are contemporaneously associated with particular types of SIO TC trajectories. This is important because SSTA can be used as a proxy measure of the strength of the atmospheric circulation around the sem i permanent SIO subtropical high, thus potentially providing insight into forcing mechanisms for the above mentioned TC trajectories. Chapter 2 will review relevant literature regarding known oceanic atmospheric global and regional pattern changes l inked to the SST anomalies (SSTA ) in the above stated regions, discuss how these changes relate to TCs in other ocean basins, and set a framework for how these may influence SIO TC trajectories. Chapter 3 explains the cluster analysis used here to assign a sample of 191 SIO TCs to groups according to their trajectories and then discusses the defining characteristics of each group. Chapter 4 then uses the clustering structure of Chapter 3 to test the above stated hypotheses using SST composites and statistical t ests of indices representing the El Nio Southern Oscillation (ENSO) and Subtropical Indian Ocean Dipole (SIOD). The conclusions and potential applications of this research are then summarized in Chapter 5, and results will show significant links between ENSO, SIOD, and SIO TC trajectories. It is hoped that the results presented here will contribute to a better understanding of the relationship between SIO TC tracks, ENSO, and regional modes of variability internal to the IO. The implications of skillful monthly, multi monthly, or even up to seasonal TC trajectory forecasts would be beneficial for long term disaster planning and risk mitigation for those vulnerable to TC impacts in the region.
20 Table 11 Tropical Cyclone Close Passes in the South Indian Ocean. Number of times d uring the period 1979 2008 that a tropical cyclone passed within 200 kilometers (km) of the coast of the four main countries. Note that Reunion is an overseas department of France. Country TC Passes <= 200 km Madagascar 58 Mauritius 49 Reunion 37 Mozambique 25 Figure 11. South Indian Ocean Region Map. Notable t ropical cyclones Dera (2001) in green and Dina (2002) in purple. Notable cyclone seasons for TC impacts include: 1984 in yellow, 1994 in orange, 2000 in blue, and 2007 in red.
21 CHAPTER 2 LITERATURE REVIEW Tropical Cyclones of the South Indian Ocean Introduction In order to set the context for the analyses in Chapters 3 and 4, this chapter reviews the existing relevant literature on previous tropical cyclone (TC) research in the South Indian Ocean (SIO). A brief note on the beginning of TC research in the basin will be followed by an updated basic TC climatology leading into a literaturebased climatological description of typical favorable c onditions for TCs in the basin. There will also be a substantial section reviewing the literature on the known oceanic atmospheric effects of El Nio Southern Oscillation (ENSO) within the SIO and how these effects are thought to influence TC activity in the SIO. The chapter will also outline the basic understanding of the other oceanic atmospheric pheno menon that will be tested here, namely the Subtropical Indian Ocean Dipole (SIOD). Finally, the concluding portion of this chapter will summarize and link together the most important concepts in preparation for the data exploration and analyses in Chapters 3 and 4. Brief History and Early Research The term cyclone was first applied in reference to the tropical systems of the Indian Ocean. Sir Henry Piddington coined the word while stationed in India during the first half of the 19th century to describe the rotating nature of the wind fields of the violent storms observed in the Bay of Bengal (Emanuel, 2005a). In the SIO, TC research was likewise pur sued in earnest by the latter half of the 19th century through British colonial resources. These early research efforts were conducted in great part by Charles Meldrum inaugural director of the Royal Alfred Observatory and founding
22 member of the Meteorol ogical Society of Mauritius (Buchan, 1901; Visher, 1922). Through collection s of observations of wind direction and velocity, pressure, humidity, and cloud types both from Mauritius and from ocean vessel reports, he approximated TC paths across the SIO an d advised ships at sea in TC avoidance tactics (Buchan, 1901) Unfortunately, for all of Meldrums curiosity, ingenuity, and ability to synthesize information, his burgeoning TC research program did not withstand the changing winds of global commerce. The opening of the Suez Canal in 1869, coupled with the rise of steam ships beginning in the 1850s, shifted important shipping lanes between Britain, India, and Asia away from Mauritius to the Red and Arabian Seas and the Bay of Bengal (Anderson, 1918; Pearson, 2003) The shift was deleterious to Meldrums work, and the number of reporting ocean vessels in the SIO dropped from 787 in 1878 to 283 by 1900 (Ward, 1902). His work bears mention here because it was not completely lost to the annals of science. Me rging Meldrums TC data and German records (now apparently lost) early 20th century American geographer and climatologist Stephen Visher surmised that about twelve TCs occur annually in the SIO between 40E and 100 E, an accurate estimation that remains v alid over the long term up to the present ( Visher, 1922) Very little research was published in English in the peer reviewed literature on SIO TCs after Visher until the early 1990s. This is likely attributable to the relatively low economic status of the SIO region and a coincident lack of access to the newer communications and remote sensing technologies that were being pioneered in Northern Hemisphere weather/climate research. One important exception was
23 Neumann and Randrianarisons (1976) work on shor t term statistical prediction of TC motions in the SIO. This paper while focused more on methodology and forecast verification than climate scale research, also included an updated climatological SIO TC narrative. The authors estimated that roughly ten TCs traverse the region each year, with the majority occurring between December and April. M ore recent paper s gave varying accounts of the baseline SIO TC climatology specific to their respective time periods and study area bou ndaries (Jury, 1993; Bessafi and Wheeler, 2006; Ho et al ., 2006; Kuleshov et al ., 2008) Therefore to provide appropriate context for this study a short descriptive SIO TC climatology is offered in the following subsection. Basic Tropical Cyclone Climato logy For the present study, an updated thirty year climatology of SIO TC counts was derived using the best track archives from the TC forecasting and monitoring arm of the United States military, the Joint Typhoon Warning Center (JTWC) (Chu et al ., 2002). The archive is updated annually and is freely available at http://www.usno.navy.mil/NOOC/nmfc ph/RSS/jtwc/best_tracks/shindex.html The period 19792008 is used here, following the recommendation of Knaff and Sampson (2009) that JTWC Southern Hemisphere (SH) TC data are most accurate, consistent, and suitable for scientific research from about the year 1980 and forward. A total of 33 9 SIO TCs of any intensity occurred during this period of record, not counting those in the Australian region generally east of 90 100E (Figure 21) It should be noted here that because austral summer spans successive calendar years, annual TC counts begin anew each July. The twenty nineyear annual mean number of SIO TCs was 11.7 with a standard deviation of 2.4 The minimum was 6 in 198283 and the maximu m was 17 in 199697, with a median of 12 TCs. E stimates of total annual
24 frequency discussed abo ve were 12 in the 1920s and 10 in the 1970s, both within one standard deviation of the mean given here for the more recent period. TCs formed in the SIO in every calendar month during the thirty year period, though the regions TC season typically spans austral summer from November through April (Figure 2 2) The peak months were January through March, during which 56% of all SIO TCs occurred, and on average, there were two cyclones in each of those months. Storms forming between June and September were usually weak and short lived. October and May TCs, while not numerous, occasionally wer e of hurricane intensity. As the present work focuses on SIO TC tracks, a track map of the SIO TCs analyzed in Chapter 4 is provided for geographical context (Figure 23) A Climatologically Favorable Environment In the North Atlantic basin, tropical easterly waves are important foci for nascent TCs (Landsea et al ., 1998). While tropical easterly waves also exist in the SIO, they typically propagate slower, are less pronounced in structure, and are not the primary source of tropical disturbances that undergo TC genesis (Parker and Jury, 1999). The primary sources of tropica l disturbances that become TCs in the SIO are n ortherly and westerly wind surges that penetrate from near the equator southward and locally enhance convergence and vorticity along the inter tropical convergence zone (ITCZ) (Jury et al ., 1994; Jury et al ., 1999). In austral summer, the ITCZ is a region in the SIO where persistent easterly and southeasterly trade winds associated with the semi permanent South Indian subtropical high pressure ridge converge with northerly and westerly monsoon outflow winds that cross the equator and are associated with the massive continental boreal high pressure systems of central and east Asia (Jury et al ., 1994). This area of sustained convergence is characterized by relatively persistent
25 convective clusters, a portion of which may intensify and become more symmetrically organized when a favorable thermodynamic environment exists (Gray, 1998). The position and orientation of the ITCZ varies within and between the seasons, and peaks in convective activity within the ITCZ gen erally coincide with peaks in TC activity. In October during the monsoon transition, the ITCZ sets up close to the equator (Jury et al ., 1994). It migrates gradually southwestward to its most poleward configuration at the height of austral summer, and re turns quickly to the north during the monsoon transition months of March and April (Jury et al ., 1994). The orientation of the SIO ITCZ during austral summer is generally southwest to northeast, extending from the Mozambique Channel to Sumatra. The boundary is approximately coincident with the average 28C SST (sea surface temperature) isotherm and rarely shifts south of 15S (Figures 2 4 and 25) ( Liebmann and Smith, 1996; Parker and Jury, 1999; Reynolds et al ., 2002 ). This configuration allows for the uplift of the moisture laden boundary layer and subsequent latent heat release through condensation, a principal energy source for TCs. Jury (1993) offered a synopsis of climatological features that are associated with variability of TC frequency in the SIO. G re ater TC activity is likely when easterly wind anomalies at 200 hPa are centered near 15S inducing enhanced anticyclonic shear This suppresses unfavorably strong jet stream westerlies south of 20S and allows for a moderate background easterly flow from the equator southward to near 10S. The results are favorable levels of vertical wind shear and diverg ent flow across the region, which provide a favorable environment of outflow mechanisms for developing TCs
26 ( Merrill, 1988) When upper level troughs intrude to near 15S, these favorable wind shear and diverg ent patterns are greatly reduced over the tropical SIO In the lower troposphere, more TCs occur when both the northern and southern Hadley cells are simultaneously strong er than norm al (Jury, 1993). This allows for strong convergence of Indian monsoon outflow and brisk subtropical trade winds in the 10S 15 S zone, result ing in increased convection, low level cyclonic vorticity and upper level poleward outflow (Love, 1985) Shanko and Camberlin ( 1998) also observed that when SIO TCs are active the northeasterly Indian monsoon outflow bypasses east Africa and flows across the equator into the SH leading to reduced onshore flow and resultant drought in Ethiopia. When the Hadley cell s are weak, Indian monsoon outflow is deflected west over east Africa bringing seasonal rains while SIO TC activity is reduced. Not surprisingly, positive SST anomalies in the SIO are likewise associated with an increase in TC frequency (Jury et al ., 1 999 ; Leroy and Wheeler, 2008; Chang Seng and Jury, 2010a). The focus of much previous SIO TC research was largely on their frequency and/or regional environmental features that influence their frequency. The basic conditions for formation are not substantial ly different from other ocean basins, and the basic climatology of the region is now relatively well established. Intraseasonal to i nterannual scale studies were restricted in the past because the historical TC data of the region are not of high quality. Since the 1970s, this situation has been improving and more recent literature focused on potential relationships between SIO TCs and the most important tropical low frequency source of oceanatmosphere variability ENSO, which itself requires a thorough d iscussion.
27 El Nio Southern Oscillation Introduction ENSO is well known as the leading mode of interannual oceanatmosphere variability in the tropics with important teleconnections influencing a significant proportion of interannual variability in extrat ropical regions as well (Trenberth et al ., 2005). The Southern Oscillation (SO) was so termed by Walker and Bliss during the 1920s to describe the inverse relationship between the relatively high surface pressure readings over the eastern South Pacific Ocean and typically low surface pressure readings near the maritime continent and eastern Indian Ocean (Julian and Chervin, 1978). The El Nio phenomenon was known originally as an annual occurrence of war ming sea surface temperatures off the coast of Peru and Ecuador, with onset near or just after the austral summer solstice coincident with a seasonal slacking of southeast trades (Wyrtki, 1975). Subsequently, the term El Nio has become associated with ex cessive warm anomalies of SST off the west coast of South America and extending westward along the equator to the international dateline, a result of reduced equatorial easterly winds and concomitant Ekman pumping induced cold upwelling (Trenberth, 1991). The linkages of these atmospheric and oceanic phenomena were first explained in detail by Bjerknes (1969), who suggested that the equatorial zonal wind circulation (the so called Walker Circulation) was thermally driven by equatorial SSTs and that variabil ity in these could then affect the strength of the meridional Hadley circulations. Additional important pieces of the ENSO puzzle were discovered in the 1970s. Kidson (1975) quantitatively identified the SO signal in the global pressure and precipitation fields, and Wyrtki (1975) described the reduced sloping of sea level across the
28 equatorial pacific as a response to weakened southeasterly trades during El Nio. In 1976, Trenberth (1976) suggested the importance of moisture convergence and enhanced conv ection over the expanding warm pool near the dateline in strengthening the Hadley cell w hile weakening the Walker cell. As the basic understanding of the coupled oceanic atmospheric nature of ENSO evolved, it became apparent that it constitutes an integral part of the global climate system. Namias (1976) noticed low pressure anomalies and a southward dip of the jet stream off the coast of California during El Nio, and both van Loon and Madden (1981) and Horel and Wallace (1981) provided further evidence t hat ENSO modulates pres sure, wind patterns, rainfall, and temperature on a global scale. Philander (1985) first applied the term La Nia to those conditions which ar e approximately opposite to El Nio, chiefly characterized by strong equatorial easterlies and a tongue of cold SST anomalies extending from the coast of South America westward in the equatorial Pacific Ocean to near the dateline with a warm pool near the maritime continent (Trenberth, 1991) The known oceanic atmospheric teleconnection links between ENSO and the SIO are discussed in the next section. ENSO Infl uences in the South Indian Ocean A plethora of research exists on the impacts of ENSO on SST and atmospheric circulation variability in the SIO in all seasons. However, because this study is focused on TCs of the SIO special attention is given here to the influences of ENSO within the SIO during austral warm season from October to April. Particular mention is made of changes in ocean atmospheric variables known to be important in TC formation and movement.
29 It is very well established that E l Nio (La Nia) is associated with an increase (decrease) of SST s in the western tropical SIO. During warm ENSO events from October until February anomalous easterly lower tropospheric winds are present in the tropical eastern SIO west to near 90E, coincident with raised mean sea level pressures (MSLP) over subtropical eastern SIO and Australia due to the shift of the downward branch of the Walk er circulation ( Pan and Oort, 1983; Gutzler and Harrison, 1987; Harrison and Larkin, 1998; Reason et al ., 2000; Larkin and Harrison, 2001; Lau and Nath, 2003; Yoo et al ., 2006). These easterlies force a downwell ing, westwardpropagating oceanic Rossby wave, which coupled with equatorial Ekman transport induces warming of SST and deepening of the thermocline in the tropical western SIO (Chambers et al ., 1999; Klein et al ., 1999; Xie et al ., 2002). This warm pool in the western SIO is most often coincide nt with positive SST anomalies (SST A ) in the equatorial Pacific Ocean near 130W, which is within the commonly used Nio3.4 index region (Pan and Oort, 1983; Nicholson, 1997). The positive SST A in the western SIO would support a greater probability of TC genesis, while the easterlies in the eastern SIO would not favor TCs due to anticyclonic wind curl and cool SST A With a warm pool over the western SIO during ENSO warm events convection is enhanced and MSLP is lowered ( Oort and Yienger, 1996; Reason et al ., 2000; Trenberth and Caron, 2000; Larkin and Harrison; 2001; Lau and Nath, 2003). Positive zonal wind anomalies are noted in the western tropical and subtropical SIO (van Loon and Rogers, 1981; Meehl, 1987; Karoly, 1989; Reason et al ., 2000; Yoo et al ., 2006). These wester lies are a vital feature of El Nio as related to SIO TC trajectories, as they
30 would likely influence a more easterly movement away from southern Africa and Madagascar. The proclivity for westerly anomalies over the western SIO durin g El Nio has been very well researched in relation to southern African rainfall. The tropical temperate troughs (TTTs) of southern Africa are known to shift northeastward over Madagascar when the SIO subtropical high weakens This allows mid latitude up per troughs to intrude equatorward and provide outflow for tropical convection, with cloud bands extending southeastward into the central SIO (Lindesay et al ., 1986; Mason and Jury, 1997; Cook, 2000; Tyson and PrestonWhyte, 2000; Nicholson, 2003; Pohl et al ., 2009; Manhique et al ., 2009) The importance of ENSO and the TTTs for TC trajectories are key components of this paper, and the TTT phenomenon will be discussed again in another section. Many authors have confirmed that the La N ia spatial pattern is largely characterized by opposite signed anomalies of SST, MSLP, and winds that are present during El Nio (Wolter, 1987; Reason et al ., 2000; Larkin and Harrison, 2001). Notably, Wolter (1987) found that during La Nia the southeasterly trades are stronger than average in the SIO and the monsoon northeasterlies across the Bay of Bengal and South China Sea are also simultaneously stronger. The stronger northeast monsoons penetrate southward across the equator in the central and eastern SIO and deflect back eastward under Coriolis forcing, increasing lower tropospheric westerlies in the eastern SIO during La Nia from October February ( Hastenrath, 2000; Larkin and Harrison, 2001). The enhanced moisture flux convergence over this region is associated with increased cloudiness and convection (Wolter, 1987). These associations suggest a
31 broad favorable area for TC activity across the tropical SIO during La Nia, with persistent southeasterly trade winds and a strong subtropical high likely to influence TC mo vement toward the west. Farther south in the SIO, there are weakened subtropical and midlatitude westerlies associated with La Nia (Reason et al ., 2000). Again, this suggests a poleward retreat of the troughs that would potentially influence recurvature in a TC Using NOAA SST and NCEP reanalysis data, difference maps of SST, SLP, and 500 hPa zonal winds were constructed to illustrate the changes in spat ial patterns that accompany El Nio and La Nia in the SIO (S ee Figures 26, 2 7, and 28 ) (Kalnay et al ., 1996; Reynolds et al ., 2002). Using the Nio3.4 region index and based on the definitions of El Nio and La Nia periods as defined by the National Oceanic and Atmospheric Administration (NOAA), El Nio years are 198283, 1986 87, 198788, 199192, 199495, 1997 98, 200203, 200405, and 200607. La Nia years are 198485, 198889, 199596, 199899, 19992000, 200001, and 200708. This information is available at http://www.cpc.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml These ENSO TC associations have been explored in the peer reviewe d literature and the next sec tion will discuss the findings. Tropical Cyclones and El NioSouthern Oscillation Many volumes and articles are available on the connection between ENSO and TCs, and the relationship therein is particularly well established in the Northern Hemisphere ( NH) H owever, it is of note that where the global ENSO TC literature has been reviewed the SIO is often an afterthought, if not completely omitted from consideration ( Gray and Sheaffer, 1991; Landsea, 2000; Chu, 2004; Camargo et al ., 2009). Furthermore, studies that include the SIO in aggregation of hemispheric or
32 global ENSO and TC sign als are difficult to interpret for any one basin. For example, Frank and Young (2007) found global numbers of TCs increase during ENSO, with positive contributions to this increase from all TC regions except the North Atlantic. However, this study haphaz ardly considered the entire SH as one ocean basin and therefore the results are very difficult to relate to more specific regional findings. Since the focus of this work is the SIO, this section surveys and integrates literature relevant to the ENSOTC as sociation only in that region. Having reviewed the larger scale SIOE NSO relationship in the previous section, this section will present the known links between these ENSO influenced changes and TCs There is no basinwide correlation between ENSO and TC genesis frequency in the SIO, just as in the North and South Pacific regions (Jury, 1993; Revell and Goulter, 1986; Lander, 1994). It is instead t he spatial patterns of TC genesis and movement that exhibit link s to ENSO. TC genesis is more active over the western SIO during El Nio (Evans and Allan, 1992; Xie et al ., 2002). Favorable conditions are generated when strong easterlies appear near Sumatra early in the austral warm season, exciting a we stward propagating oceanic Rossby wave that couples with off equatorial Ekman transport to induce an unusually deep warm pool centered near 15S, 60E (Jury et al ., 1999; Reason et al ., 2000; Xie et al ., 2002). At the height of austral summer, the TTTs of ten shift from the African continent to near the vicinity of the warm pool, resulting in an intersection of the South Indian Convergence Zone (SICZ) and the SIO Intertropical Convergence Zone (ITCZ) This combination of favorable thermodynamics and a lift ing mechanism leads to more convective activity in the western tropical SIO ( Lindesay et al ., 1986; Mason and Jury, 1997; Cook, 2000; Xie et al ., 2002; Ho et al ., 2006 ;
33 Manhique et al ., 2009). During an ENSO warm event c ooler SSTs in the eastern SIO and anticyclonic shear from equatorial lower tropospheric easterlies account for a reduction of TC genesis, particularly in early summer (Ho et al ., 2006; Camargo et al ., 2007c; Kuleshov et al ., 2009). Conversely, during an ENSO cool event the western SIO suf fers reduced TC genesis while the eastern SIO experiences more, with positive SST anomalies, midtropospheric relative humidity, and increased cyclonic vorticity contributing to the eastern shift in genesis (Wolter, 1987; Camargo et al ., 2007c; Kuleshov et al ., 2009) The tracks of SIO TCs have also been studied, and i n terms of track density there is some disagreement in the literature. One group of authors have consistently conveyed their finding of a partition at 85E with greater (reduced) TC activity in the western ( eastern) SIO during El Nio years and the reverse relationship during La Nia (Kuleshov and de Hoedt, 2003; Kuleshov et al ., 2008 ). Ho et al (2006) found a similar split with their cut point at 75E however Chang Seng and Jury (2010a) observed in SST comp ositing that those years of greater TC activity between 50E 70E exhibit ed a L a Nia signal in the eastern Pacific. The contrasting results could be attributed to the use of different data sets, different definitions of El Nio /La Nia years, and different intensity thresholds for the definition of a TC. In any case, the fact that varying results have been obtained lends support to the notion that east to west differences in yearly TC activity within the SIO cannot be clear ly established based solely on ENSO. As noted in the previous section, wind anomalies were found to be more westerly to the south of 10S across the SIO during strong warm ENSO events especially over the western SIO ( van Loon and Rogers, 1981; Meehl, 1987; Karoly, 1989; Cook, 2000;
34 Reason et al ., 2000; Xie et al ., 2002 ; Yoo et al ., 2006 ). Also, the zonal steering flow averaged over 850200 hPa across the tropical and subtropical SIO was found to be more westerly (easterly) during El Nio (La Nia) (Vitart et al ., 2003; Ho et al ., 2006). This was hypothesized to lead to a greater risk of landfall for Mozambique during La Nia, when steering flow is highly zonal (Vitart et al ., 2003; Reason and Keibel, 2004). Ho et al (2006) suggested from track density m aps that westerly steering flows could increase the incidences of recurving TCs just east of Madagascar during El Nio Similar recurvature trends can be inferred from track density maps in Camargo et al (2007c). Given the associations already noted abov e between ENSO phase and the presence of westerly winds near Madagascar, the notable shift of TTTs over the western SIO during warm ENSO events, and the observed tendency for recurving TCs during warm ENSO phases, it is perplexing why no link has been expl ored between the three. Several very recent studies have used outgoing longwave radiation ( OLR ) and rainfall anomalies to bolster the long suggested strong links between the location of TTTs and ENSO phase (Fauchereau et al ., 2009; Pohl et al ., 2009; Manhique et al ., 2009). There seems to be a gap in the literature here where the westerly wind anomalies and negative OLR convective anomalies at the source of these maritime based El Nio TTTs could be tested for associations with SIO TCs. It s tands to reason that during El Nio TCs may often develop and move southeastward as integral parts of the TTTs, or perhaps TCs tracking through the SIO could interact with the troughs and be shunted south and eastward.
35 In addition, many authors have noted the insufficiency of ENSO phase aggregated to the seasonal time scale as the sole predictor of convective variability over Africa and extending into the SIO ( Waylen and Henworth, 1996; Mason and Jury, 1997; Fauchereau et al ., 2009; Pohl et al ., 2009; Manhi que et al ., 2009 ; Chang Seng and Jury, 2010a) The work presented in this paper builds on this notion to demonstrate a link between ENSO, TTTs, and TCs in tandem with a regional mode of SIO SST variability. A more integrated understanding of the Pacific teleconnection and the local SST will also allow consider ation of these relationships at the subseasonal temporal scale rather than the more typical seasonaggregated paradigm The other important mode of variability for SIO TC trajectories that must be accounted for in tandem with ENSO is the SIOD, which will be discussed in the following section. Subtropical Indian Ocean Dipole As demonstrated in the previous section, ENSO is undoubtedly link ed to SIO oceanatmospheric variability including TC frequency and trajectories However, a significant portion of this variability cannot be solely attributed to ENSO. There are modes of variability internal to the SIO that must also be taken into account, and these w ould be expected to interact with the ENSO phenomenon to potentially affect SIO TC tracks. As will be demonstrated in Chapter 4, the most relevant phenomenon in relation to TC trajectories in the SIO is the Subtropical Indian Ocean Dipole (SIOD). This phenomenon has been independently quantified as the second EOF of SIO SST A at least three times in the literature, though not always identified explicitly as the SIOD ( Behera et al ., 2000; Behera and Yamagata, 2001; Huang and Shukla, 2007; Leroy and Wheeler, 2008 ). I t has been implicated as equal in importance with ENSO in understanding warm season rainfall variability in bot h Africa and Australia ( Reason,
36 2001; England et al ., 2006). The pattern peaks during austral summer and is characterized in the positive (negative) mode by warm (cool) SST a nomalies centered about 25S 40S and 55E 75 E, with cool (warm) anomalies to the west of Australia centered about 18S 30S and 85E 105E (Behera and Yamagata, 2001; Suzuki et al ., 2004; Huang and Shukla, 2008). Most importantly, the SIOD pattern has been shown to occur largely independent of ENSO phase (Behera et al ., 2000 ; Reason, 2001). The SIOD is known in positive mode to enhance precipitation over southeastern Africa and southwestern Australia (Behera and Yamagata, 2001; Reason, 2001; England et al ., 2006). In particular, when the warm pole is shifted west closer to southern Africa the potential for heavy precipitation is further increased, owing to anomalous onshore southeasterlies advecting humid maritime air over the sub continent (Reason, 200 1; Reason, 2002). The mechanism for cooling in the eastern pole is strengthened southeast trades and resultant increased ocean surface evaporation and mixing, while the western warm pole develops concurrently with increased southward Ekman transport of warm SSTs from the tropical SIO combined with decreased northward Ekman transport of cool SSTs from the midlatitudes (Behera and Yamagata, 2001) In negative mode, the SIOD is associated with drier conditions over far southeastern Africa, as well as in sout hwestern Australia. The SST A poles are generally reversed, with warm anomalies stretching from the eastern subtropical SIO back to the west northwest toward Madagascar in association with slackened southeasterly trades. The western pole cools south and s outheast of Madagascar in conjunction with enhanced latent and sensible heat loss from enhanced westerlies and
37 the advection of drier, cooler air masses from high er latitudes ( Behera and Yamagata, 2001) Reason (2002) noted the strong similarity between t he positive SIOD SST and wind anomaly patterns and the atmospheric anomalies associated with tropical temperature troughs (TTTs) This is important because it raises the possibility of local SST A influence in the variability of TTTs, which are often attri buted to remote forcing mechanism s from ENSO as discussed previously in this chapter. The potential association between SIO TCs and SIOD has not been explored explicitly. In Leroy and Wheeler (2008), the second EOF of IndoPacific region SST A is used as o ne of several predictors of TC activity in the western SIO. This pattern is identified by the authors as the tropical Indian Ocean Dipole of Saji et al (1999) and Webster et al (1999), though the visual loading s strongly suggest the SIOD pattern of Behera and Yamagata (2001) and later Suzuki et al (2004). Thus, there is precedence, albeit inadvertent, for a significant association between SIO TCs and SIOD. As discussed earlier in the chapter, it stands to reason that an eastward shift of TTTs over the western SIO could in theory influence TCs to take a poleward or even eastward turn. Therefore, if SIOD is also associated with the occurrence of TTTs, and if the SIOD occurs largely independent of ENSO (Fauchereau et al ., 2003; Washington and Preston, 2006) then an interactive consi deration of ENSO and SIOD is necessary to associate one or both with the variability in TTTs and/or in SIO TC trajectories. A more detailed discussion of the literature on TT Ts follows in the next section. Tropical Temperat e Troughs While it is well known that recurving TCs often are steered by and interact with atmospheric troughs, the synoptic patterns of the troughTC relationship as potentially altered by ENSO and/or SIOD have not been fully explored in the SIO The present
38 study is aimed at this gap in the literature, and TTTs are implicated as an important steering mechanism for SIO TCs. In this section, a summary is provided o f the extant literature on TTTs. Lower tropospheric c onvergence zones characterized by posi tively tilted poleward moisture plumes or cloud bands were an early discovery after the advent of weather satellite remote sensing in the 1960s. Streten (1973) noted the existence of these diagonal cloud bands in each of the SH oceans, located equatorwar d and to the west of the semi permanent subtropical highs. This region of enhanced convection is sometimes referred to as the South Indian Convergence Zone (SICZ), and is only prominent during austral summer (Cook, 2000; Todd et al ., 2004). Over southern Africa and into the western SIO, these are referred to as tropical extratropical cloud bands or tropical temperate troughs (TTTs). They have been very well known for many years as the principal source of summer precipitation over southern Africa (Harangozo and Harrison, 1983; Lindesay et al ., 1986; Mason and Jury, 1997). Their structure over the African continent is in the coupling of a thermally induced tropical low often located over southeast Angola or southwest Zambia near 20S with an upper level wes terly transient wave passing south of Africa (Harangozo and Harrison, 1983; Lyons, 1991; Mason and Jury, 1997; Todd and Washington, 1998). TTTs do not always reside over Africa, but sometimes shift north and eastward over Madagascar and into the SIO ( Maso n and Jury, 1997; Washington and Todd, 1999). TTTs represent important extrusions of heat and moisture into the upper troposphere which are then redistributed into the higher latitudes ( Todd et al ., 2004).
39 The relevance of TTTs to SIO TC trajectories has been mentioned here in earlier sections. The longitudinal placement of TTTs has been linked to phases of ENSO, with the warm phase representing a shift of sustained convection over the western SIO (Lindesay et al ., 1986; Mason and Jury, 1997; Cook, 2000; Tyson and PrestonWhyte, 2000; Nicholson, 2003; Fauchereau et al ., 2009; Pohl et al ., 2009; Manhique et al ., 2009). Similar eastward shifts have been observed during negative SIOD events (Reason, 2002; Pohl et al ., 2009). These studies demonstrated anomalous westerlies over Madagascar and into the tropical/subtropical SIO in accompaniment of the eastwarddisplaced TTTs. Previous studies of TCs suggested a possible link be tween recurving SIO TCs and El Nio and between westward moving TCs and La Nia (Jury et al ., 1999; Xie et al ., 2002; Vitart et al ., 2003). The longitudinal shifting of the TTTs, in conjunction with the vacillation of the SIO trade winds, are likely related to monthly or perhaps even seasonal t rends in the directional motion of TCs through the control mechanisms of ENSO and/or SIOD. Conclusion This literature review has surveyed the SIO TC literature and discussed two of the most important known sources of intraseasonal to interannual variabilit y in the tropical and subtropical SIO: ENSO and SIOD. ENSO is known to exert a strong influence on oceanic atmospheric variables that are linked to SIO TC frequency and trajectories. However important ENSOs role may be in the variability of either TTT or TCs it is not sufficient alone to explain changes in convection and the associated wind patterns that steer TCs The SIOD has been shown to exert influence on convective patterns in the absence or in opposition to the influence of ENSO. T here has not been a comprehensive study of potential associations between SIOD and SIO TC activity The
40 only known possible link in the literature between the two was in Leroy and Wheeler ( 2008) wherein the second EOF of IndoPacific SSTs displayed an SIOD pattern bu t was not clearly identified as such by the authors Additionally, the southern African TTTs and the socalled SICZ (which are both linked to ENSO and SIOD) have not been linked with variability of TC trajectories, though their coincident seasonal natures, geographic locations, and atmospheric structures strongly suggest a meaningful association. The rest of this study explores the identified gaps in the literature of the relationships between SIO TC trajectories, ENSO, and SIOD The presence/absence of TTTs and associated weakness /strength of the SIO trade winds and subtropical high are proposed as key mechanisms in these relationships In the next chapter, cluster analysis will be utilized to classify SIO TC tracks according to their genesis and final locations in order to test whether patterns of local (SIOD) and/or remote (ENSO) SST A associate with the different types of TC trac ks.
41 Figure 21. Annual f requencies of South Indian Ocean t ropical cyclones in blue. Five y ear running m ean in red. Period of record is 1979 80 to 200708 during which a total of 339 storms occurred.
42 Figure 22. Monthly f re quencies of South Indian Ocean tropical c yclones Period of record is 197980 to 200708 during which a total of 3 39 storms occurred.
43 Figure 23. South Indian Ocean Tropical Cyclone Trajectories. Trajectory map of the tropical cyclones included in this study.
44 Figure 24. Indian Ocean Sea Surface Temperatures averaged from 19792008. From top, October, December, February, A pril. Orange shading represents SSTs greater than 26C and red shading represents SSTs greater than 28 C.
4 5 Figure 25. Indian Ocean Outgoing Longwave Radiation, averaged from 19792008. From top, October, December, February, April. Color Key: Blue 180 200 W m2, Green 200220 W m2, Orange 220240 W m2, Red 240 260 W m2, Pink 260280 W m2, Clear >280 W m2.
46 Figure 26 Difference of El Nio and La Nia SST p atterns in the SIO
47 Figure 27. Difference of El Nio and La Nia SLP p atterns in the SIO.
48 Figure 28 Difference of El Nio and La Nia 500 hPa zonal wind p atterns in the SIO.
49 CHAPTER 3 A CLUSTER ANALYSIS O F SOUTH INDIAN OCEAN TROPICAL CYCLO NE TRAJECTORIES Introduction The literature review in the previous chapter demonstrated that very little research has considered the potential relationship between the Subtropical Indian Ocean Dipole (SIOD) and South Indian Ocean (SIO) tropical cyclones (TCs). Other potential interac tions between El NioSouthern Oscillation (ENSO), SIOD, tropical temperate troughs (TTTs), and TC trajectories have likewise not been explored. Copious volumes have been written on many of these topics with reference to African or Australian rainfall, an d it seems much of that research could be extended and applied with respect to TCs. As noted previously, tropical cyclones are not scarce in the SIO, but occur in similar abundance to hurricanes of the North Atlantic Unfortunately, the difficulty in conducting oceanatmosphere research in the Southern Hemisphere (SH) is that reliable observations are few and far between However, remote sensing technology began filling in the vast SH observational voids during the 1960s and particularly since the lat e 1970s. It is now accepted in the scientific community that sufficient satellitederived measurements and estimates of oceanatmosphere phenomena in the SH exist beginning from about 19791980 ( Uppala et al ., 2006; Knaff and Sampson, 2009) Giv en thirty years of consistent SH data, opportunity exists to attempt more comprehensive explanation of the links between SIO TCs, the synoptic scale oceanatmospheric patterns in which their life cycles are embedded and larger scale sources of regional and global im port ance such as SIOD and ENSO. The balance of the work presented here is devoted to filling in the gaps in the literature on the links between
50 SIOTC tr ajectories ENSO, and SIOD The approach described herein is to first reduce dimensionality of the TC data by classification of the storms into a manageable yet meaningful structure by way of cluster analysis This is intended to facilitate statistical analyses of the TCs and interpretation of the findings in relation to known physical mechanisms governi ng their directions of motion guided by ENSO and/or SIOD influenced synoptic patterns In Chapter 4, analysis of variance (ANOVA) will be applied to the clusters to compare the median values of ENSO and SIOD indices. Global composite maps of sea sur face temperature anomalies (SSTA ) will also be constructed to complement the statist ical results. Tropical Cyclone Data and Study Area Definition The Joint Typhoon Warning Center (JTWC) best track dataset for the SH provided the latitude/longitude fixes of SIO TCs at six hour intervals, which are deemed reliable and suitable for scientific and statistical research from about 1980 onward (Knaff and S ampson, 2009). A subsequent desirable effect of using the recent thirty year period from 1979 2008 is to research principally the teleconnections operating after the known Pacific Ocean climate shift of 1976 77, as it has been shown that spatial patterns of ENSO SIOD interaction differ on multi decadal time scales (Trenberth, 1990; Zinke et al ., 2004). Yet, permanent geostationary weather satellite coverage over the entire SIO was not realized until May 1998 (Caroff, 2009; Chang Seng and Jury, 2010a). Th is means that estimates of TC intensity may not be reliable prior to 1998, and that weak and/or short lived TCs may not be accurately accountedfor during this period. Thus, only SIO TCs which attained a maximum lifetime intensity of at least 30 m s1 (or 60 knots ma ximum 1 minute sustained wind) are included in the analysis
51 The decision to include only relatively well organized TCs is intended to alleviate concerns about TC fix data quality, which are most error prone in analyses of weaker, less organized tropical systems (Yip et al ., 2006). For each of the 191 TCs in this study, all fixes provided in the JTWC best track data are utilized, regardless of the estimated intensity. This means that a portion of the fixes near each storms beginning and end p oints likely include position estimates made when the systems were not yet well developed, or near the end of their life cycle when they were undergoing extratropical transition. The inclusion of all fixes is justified in order to construct a meaningful classification of TC trajectory types over their full life cycles from cyclogenesis to cyclolysis. W eak TCs do not typically exhibit a well defined, axisymmetric vertical vortex structure, thus their movements in relation to the larger background steering flow can be quite aberrant. As this study is primarily concerned with TCs whose movement behaves more or less in concert with the mean steering flow, the decision not to consider weak storms is upheld. The spatial extent of the study region is restricted according to the eastern most boundary of the regional SIO TC forecasting agency Official TC forecast responsibilities and data archives for the World Meteorological Organization (WMO) are assigned to the La Runion Regional Specialized Meteorological Ce nter (RSMC) for the Indian Ocean south of the equator and eastward from the Africa n continent to 90E (Caroff, 2009) East of 90E in the SIO, the responsibility is assigned to WMO RSMC Perth in Western Australia. The intended region of study here is the western SIO including southern Africa, Madagascar, and the Mascarene archipelago, but not the Australia region. Thus, the artificial 90E boundary is applied here to further truncate
52 the number of storms in the analysis leaving out all TCs which formed in the SIO east of 90E and never crossed that meridian into RMSC La Runions area of responsibility. Finally, a bri ef comment is necessary regarding the use of JTWC TC data here instead of WMO La Runion TC data. T he latter dataset does not contain complete records of TCs that began well east of 90E and later crossed the boundary on a westerly tack. A WMO SH TC bes t track archive was recently assembled which incorporated La Runion and Perth data and eliminated the anthropogenic boundary effect; however these data are not available to this author ( Kuleshov et al ., 2008). Ho et al (2006) addressed dataset concerns by performing the same analysis on both the JTWC and La Runion datasets, with spatially and temporally consistent results. Thus, the JTWC data are utilized here without further deliberation. A tota l of 191 SIO TCs are included in the cluster analysis ( See Appendix) Cluster Analysis in Tropical Cyclone Research Cluster analysis (CA) is a widely known quantitative classification and data exploration method that has been frequently employed in geophysical research over the last fifty years (Gong and Richm an, 1995). The most common use of CA is as an exploratory tool to reduce highly heterogeneous data sets to a few manageable and internally similar groups such that variability is maximized between groups while simultaneously minimized within groups (Lattin et al ., 2003). In CA, there is no foreknowledge of an optimum number of clusters in the data, such that the final cluster solution may result in a number of groups that does not necessarily reflect the natural modality of the data. A well struct ured clustering solution, whether it reveals the optimum natural modality of the data or not, can aid a researcher to understand
53 mechanisms whi ch shape differences and similarities of places, events, or phenomena (Wilks, 2006). Classification schemes, both subjective and objective, have been applied to both extratropical and tropical cyclone trajectories. Many have successfully employed CA to sift through the complex and chaotic nature of these transient entities and group observations into recognizable and interpretable archetypes (for examples see Wolter, 1987; Blender et al ., 1997; Trigo et al ., 1999; Gaffney et al ., 2007; Manhique et al ., 2009) Of special interest here are those which have classified TCs based on their trajectories. These types of analyses have been performed for TC trajectories in the western North Pacific (WNP), North Atlantic, and eastern North Pacific. Harr and Elsberry (1991) qualitatively identified straight moving, recurving north, and recurving south TCs in the western North Pacific (WNP). In the same basin, Elsner and Liu (2003) used four geographic variables in a kmeans clustering algorithm and identified three types of TC paths: recurving, north oriented, and straight moving. Camargo et al (2007a) employed a regression mixture model to further categorize WNP TCs into seven clusters, taking into account geographic location as well as forward speed and the actual shape of each TCs entire trajectory Most recently, Choi et al (2009) used a fuzzy clustering approach to i dentify four different types of TC trajectories that make landfall on the Korean peninsula. In the North Atlantic, Elsner (2003) also employed kmeans clustering to identify three types of hurricane tracks, two of which often make landfall in the United States. Camargo et al (2008) built upon the regression mixture model clustering from their WNP study and found three clusters of TC trajectories in the eastern North Pacific. In
54 e ach of the TC clustering studies listed above, the respective clusters were used to draw conclusions about each trajectory types potential for landfall, their seasonality, and relationships to prominent low frequency sources of atmospheric variability such as ENSO. Following from these analyses in other oceans, the goal here is to use CA to complete the first known classification of TC trajectories in the SIO. T he categorization scheme is subsequently used to identify which trajectory types associate closely with ENSO and SIOD The Clustering Procedure The clustering algorithm implemented in this study is a twostage agglomerative hierarchical approach using a Euclidean distance measure and the group average linkage met hod. Stage one delineates TC genesis regions within the SIO while stage two examines the direction of storm motion for TCs within each genesis region. Agglomerative hierarchical clustering begins with each observation as its own group and then iterativel y joins the two closest groups until there is one group that includes every observation (Lattin et al ., 2003). The criteria for determining the closest proximity group is the average Euclidean distance between all available pairs in any two groups being c ompared (Wilks, 2006). The statistical software package NCSS is utilized to compl ete the clustering procedures. The variable utilized to perform the initial clustering procedure is based upon pr evious research demonstrating that El Nio (La Nia) is assoc iated with more TCs forming west (east) of about 75E to 85E (Ho et al ., 2006; Camargo et al ., 2007c; Kuleshov et al ., 2009). This is theoretically justified in that a favorable pool of anomalously warm SST s and deeper than normal thermocline are present in the western SIO during strong ENSO warm events (Xie et al ., 2002; Camargo et al ., 2007c),
55 along with unfavorable negative SST anomalies, stronger lower tropospheric easterlies, and elevated mean sea level pressure (MSLP) in the eastern SIO (Pan and Oort, 1983; Reason et al ., 2000; Larkin and Harrison, 2001) It follows then that a zonal stratification in the initial longitude points echo es the established regional physical characteristics of ENSO in relation to TC genesis Thus, the longi tudes of the initial center fixes for each TC are clustered in the first stage of the analysis. In hierarchical CA, one diagnostic of the strength of a clustering structure is the cophenetic correlation, which is related to the dendrogram. A dendrogram is a tree diagram that illustrates the hierarchy of linkages carried out by the clustering algorithm (Wilks, 2006). Cophenetic correlation indicates how well the visual structure and dissimilarity index of a dendrogram reflects the clustering structures and dissimilarity measures in the actual data matrix (Romesburg, 1984). If the cophenetic correlation is near 0.8 or greater, the clustering structures as depic ted in the dendrogram are more likely to reflect real clusters in the data. If the cophenetic cor relation is significantly less than 0.8, the clustering structures as visualized in the dendrogram are more likely exaggerations of reality and the results of the CA are not suitable for further analysis (Romesburg, 1984). The cophenetic correlation corresponding to the dendrogram for the first stage of the CA is 0.7 7 which indicates that visual inspection of the dendrogram should allow for discernment of an appropriate number of clusters (Figure 3 1) It is important here to discuss the reasoning behind the chosen number of clusters. One subjective method for choosing the number of clusters is by visually identifying a cut line in the tree diagram where the average dissimilarity between the groups jumps
56 considerably between grouping iterations (Wilks, 20 06). In Figure 31 this jump occurs at approximately 810 on the dissimilarity index, with groups 25 all agglomerating near this point. Therefore, the cut line is placed such that there are five clusters. G roup 1 is small and clearly different from all other groups. Group 2 is likewise separated from groups 35. With five clusters tentatively identified, these are then mapped using ArcGIS to visualize the geographic pattern (Figure 3 2) The map of the five cluster solution helps to justify the choice of that number. In CA there is not always a clear optimum number of groups, especially with larger data sets, and t he number of clusters extracted is often partly a function of the researcher's goal for the analysis (Wilks, 2006). Since this study seeks to identify potential links between ENSO, S IOD, and TC trajectories, it makes sense to incorporate previously accepted knowledge on the effects of ENSO in the SIO TC genesis regions. Such incorporation is accomplished w ith the five cluster solution. Three of the five groups contain approximately 89% of the TCs and cover the region within which ENSO influences on genesis are thought to be the strongest. Group 3 roughly coincides with the regi on in the west ern SIO where TC genesis is more likely during a warm ENSO event, while group 5 coincides with the region in the eastern SIO where genesis is more likely during a cool ENSO event ( Ho et al ., 200 6 ; Camargo et al ., 2007c; Kuleshov et al ., 2009) Recall that Ho et al (2006) placed the boundary separating these regions at about 75E, but Kuleshov et al (2008) found the boundary to be 85E. Therefore, group 4 is the transition region where the see saw influences of ENSO phases on TC genesis are not well defined. The other 11% of observations from the remaining two groups are located at the far western and eastern bounds of the
57 study region. Very little has been published about ENSO and TCs forming in the Mozambique Channel, thus group 2 coinci des with a region where the ENSO influence on TCs may be subsidiary to other influences. F inally, group 1 is clearly exceptional, including four TCs with unusually long distance tracks. One of the principal goals of this study is to investigate the direct ions of the TC trajectories in relation to ENSO and SIOD. ENSO is generally thought to influence more westerly tracks during a strong cold phase and more easterly tracks during a strong warm phase (Vitart et al ., 2003). It is also hypothesized that a stro ng positive SIOD (with anomalously strong trade winds) should influence more westerly tracks, while a strong negative SIOD (weaker trades and eastward shift of TTTs) should influence more easterly tracks (Behera and Yamagata, 2001; Reason, 2002). Thus, a second stage of clustering is undertaken to differentiate the cases where storms that form in a similar geographic location follow very different tracks. Only the three large clusters across the main development regions of the tropical western and central SIO are utilized for this second CA as Group 1 is too small, and TCs within Group 2 were close to land. This second CA is performed for each group separately so that the TCs within groups 3, 4, and 5 are clustered by the latitude and longitude of their final positions (Figures 3 3, 3 4, and 35) The importance of this second round of clustering is to mete out storms that move mostly westward and threaten land from those that track more eastward and remain at sea. T he same agglomerative hierarchical method is implemented for the second CA as was described for the first, with a Euclidean distance measure and group average linkage. A total of seven clusters will be discussed in the rest of the study: two motion based clusters derived from each of the or iginal groups 3,
58 4, and 5, and one from the original cluster 2. The seven clusters are ranked by their size and then designated simply as C1 through C7 ( Table 31). Discussion The clustering procedure yields three large groups in the three main genesis re gions of the SIO (western, central, and eastern) that generally move west or southwest. 60% of the TCs in the study are contained within one of these three groups, confirm ing that a majority of TCs in the SIO move with a southwesterly component This is not surprising given that average background conditions absent an ENSO or SIOD influence would allow for such a motion generally around the northwest fringes of the subtropical high. S econdary groups are also found for each of the three main genesis regions which are likely candidates for association with departures from the average flow regime, such as happen with opposing phases of ENSO or SIOD Approximately 27% of the TCs examined belong to one of these groups Seven of the eight clusters are described below to complete this chapter. Eastern Formation Region: C1 and C6 The largest group overall is C1 with 40 TCs. These form in the eastern SIO east of 90E and track to the west and southwest before weakening or recurving south and eastward in the area south of 15S and east of 75E (Figure 3 6 ) The peak months of formation for these TCs are February April with a secondary peak in November, and t his type of SIO TC rarely passes within 200 kilometers (km) of po pulated land masses. Although also forming in the eastern SIO, group C6 has 16 TCs and is distinguished from C1 by longer distance tracks with more westward trajectories, with a few in this group crossing west of 60 E and passing within at least 200 km of land (Figure 3 7 )
59 These unique storms occur most frequently during the SIO peak months of January March. Central Formation Region: C2 and C7 Occurring most often during January March with a secondary peak in November, C2 is also a common type with 39 TCs in the group. These TCs develop in the central genesis area between 75E and 85E and track westward or southwestward with a concentration of final points south of 20S between 50E and 75E (Figure 38 ) A second concentration of final points exists near northern Madagascar and within the Mozambique Channel. During the period 19792008, 16 of the 39 TCs either made landfall or passed within 200 km Madagascar, Mozambique, or the Mascarene Islands C7 also forms in the central genesis region and is far less common with only 14 TCs. These storms occur in most every month during the austral summer but show a unique tendency to occur more in October when TCs in other parts of the basin are rare. They track g enerally southward and southeastward, ending south of 15S and east of 80E, and do not threaten any populated regions of the SIO (Figure 3 9 ) Western Formation Region: C3 and C4 In the western genesis area, 39 TCs comprise C3; this group has the highest frequency of making land fall or passing close by with 33 out of the 39 impacting Madagascar Mozambique, or the Mascarene Islands. These TCs occur very frequently from December to March, with very few forming during the transition months of November or April. C3 TCs form in an arc from between 6S and 12S and 60E to 70E bending south and west to between 10S and 18S and 55E and 60E (Figure 3 10) C3 final points are heavily concentrated around Madagascar, probably in large part because of deleterio us interactions between the TC vortex and the mountainous
60 Malagasy terrain which exceeds 2500 m eters at its heights. A second concentration of TC final positions are found south of the Mascarene Islands between 27S 33S and 50E 60E. C4, the counterpar t to C3 in the western genesis area, shows a tendency to develop near 12S between 60E 72E in Jan. Feb (Figure 3 1 1 ) Movement is southward and southeastward and C4 TCs seldom pass within 200 km of the principal populated islands of the Mascarene archipelago Trajectory end points are generally south of 20S and east of 63E. These two groups represent the greatest opportunity to apply knowledge of the effects of ENSO and SIOD on TC tracks. If the likelihood of C3 versus C4 TC trajectory types could be predicted one month or three months in advance with some confidence, this could be invaluable information for disaster planning purposes in the inhabi ted regions of the western SIO. Far West/ Mozambique Channel: C5 The last group is located at the far western edge of the SIO (Figure 3 1 2 ) 17 TCs in C5 formed near Madagascar or in the Mozambique Channel. Given their close proximity to land at genesis, it is not surprising that they impact populated regions at a high rate with 14 of 17 tracking within 2 00 km of one of the major inhabited places in the region. TCs in this group that form north of about 12S exhibit a southwestward and then southward movement, but the remainder of the TCs in the group move mostly to the south and east, which is known to be the common trajectory for TCs with genesis in the Mozambique Channel (Reason, 2007). These occur mainly at the height of austral summer in Jan. Feb. Conclusion In this chapter, a two stage hierarchical CA of TC initial longitudes and final latitudes an d longitudes was used to identify eight types of TC t rajectories in the SIO
61 Similar research methods have been utilized in the literature in other active TC basins, but such a classification for the SIO is not found in the peer reviewed literature. The results in the first stage corresponded well to the known geographic patterns of TC genesis in relation to ENSO. The second stage of CA parsed out three additional groups by splitting the main genesis region clusters based on the geographic locations of t he ir final positions. The overarching goal of this study is primarily to investigate the influences of ENSO and SIOD on TC trajectories in the SIO. The cluster solution presented above fits in with changes in the geographic distribution of a favorable env ironment for TC genesis associated with ENSO. This provides confidence that the clustering solution reflects at least a portion of the underlying physical mechanisms behind the variability of TC genesis and trajectories in the region. Therefore, in the next chapter the groups described above are used as a basis for further inquiry into ENSO and S IOD as sources of TC trajectory variability in the SIO. The method for testing these associations will be to compare the seven groups via ANOVA between median values of ENSO and SIOD indices. Composite maps for each group will also allow for qualitative identification of SST A anomalies in key regions of the subtropical SIO and the tropical Pacific.
62 Table 31. Eight cluster solution for South Indian Ocean tropic al cyclones ranked by the number of TCs in each group. The group average initial and final longitudes final latitude are displayed in decimal degrees east and south. Movement direction is a generalization to dichotomize the T Cs from common genesis regions (western, central, eastern SIO, and Mozambique Channel) The number of TC passing within 2 00 kilomete rs of land is given in the far right column. Cluster Designation Number of TCs Initial Longitude Genesis Region Final Latitude Fin al Longitude Movement Direction Close Pass C1 40 95.60 Eastern 24.73 83.33 Southwest 2 C2 39 78.83 Central 23.76 56.65 Southwest 16 C3 39 62.98 Western 26.95 49.89 Southwest 33 C4 22 65.38 Western 26.55 69.32 Southeast 3 C5 17 44.85 Channel 27.06 48.61 Southeast 14 C6 16 99.01 Eastern 20.20 55.51 West 7 C7 14 81.91 Central 20.89 86.59 Southeast 0 Sum / Average 187 75.51 24.31 64.27 75 Table 32. Frequency of TC genesis by month for C1 C7. Clusters Sept. Oct. Nov. Dec. Jan. Feb. Mar. Apr. May C1 0 1 6 4 4 6 11 7 1 C2 0 0 7 3 11 8 5 4 1 C3 0 0 0 9 13 7 7 1 2 C4 0 0 1 3 7 6 3 2 0 C5 0 0 0 0 5 8 3 1 0 C6 0 1 2 1 4 3 5 0 0 C7 1 3 1 0 2 2 2 2 1
63 Figure 31 Dendrogram of Sou th Indian Ocean tropical cyclones when clustered by their initial longitude The vertical red line indicates the cut point for a five cluster solution. The cophenetic correlation is 0.77.
64 Figure 3 2 Map of five cluster solution of South Indi an Ocean tropical cyclones when clustered by their initial longitude. Each point represents the latitude/longitude position at which the system was first identified as a tropical cyclone. Figure 33 Dendrogram of South Indian Ocean tropical cyclones in central region when clustered by their final latitude and longitude. The vertical red line indicates the cut point for a two cluster solution. The cophenetic correlation is 0.72.
65 Figure 34 Dendrogram of South Indian Ocean tropical cyclones in western region when clustered by their final latitude and longitude. The vertical red line indicates the cut point for a two cluster solution. The cophenetic correlation is 0.71
66 Figure 35 Dendrogram of South Indian Ocean tropical cyclones in eastern region when clus tered by their final latitude and longitude. The vertical red line indicates the cut point for a two cluster solution. The cophenetic correlation is 0.71
67 Figure 36 T ropical cyclone trajectories in Cluster 1 (C1): eastern region and south motion Figure 37 Map of tropical cyclone trajectories in Clu ster 6 (C6): eastern region and west motion.
68 Figure 38 Map of tropical cyclone trajectories in Clu ster 2 (C2): central region and southwest motion. Figure 39 Map of tropical cyclone trajectories in Cluster 7 (C7): central region and southeast motion.
69 Figure 310. Map of tropical cyclone trajectories in Clu ster 3 (C3): western region and southwest motion. Figure 31 1 Map of tropical cyclone trajector ies in Cluster 4 (C4) : western region and southeast motion.
70 Figure 31 2 Map of tropical cyclone trajectories in Cluster 5 (C5): Mozambique Channel region.
71 CHAPTER 4 THE INFLUENCES OF EL NINO SOUTHERN OSCILLATION AND THE SUBTROPICAL INDIAN OCEAN DIPOLE Introduction As discussed in Chapter 1 tropical cyclones (TCs) are a recurring phenomenon in the South Indian Ocean (SIO) during austral summer. These damaging and life threatening systems are of particular concern for the inhabitants on the southwest fringe of the basin, especially in Madagascar and Mozambique where social and economic factors may serve to amplify human vulnerability to environmental hazards such as TCs (Brown, 2009; Silva et al ., 2010). One of the utmost aims of this work is to contribute to scientific understanding of the relationships between SIO TC movements, El Nio Southern Oscillation (ENSO) and the Subtropical Indian Ocean Dipole (SIOD) It would be useful in the future if the physical risk from TC strikes could be better understood, such that a developing nation like Madagascar might be able to plan one or perhaps three months in advance when a favorable pattern for TC strikes is anticipated. A survey of previous research in Chapter 2 found that ENSO has a noticeable teleconnection signal in the SIO. During the warm phase, positive sea surface temperature anomalies (SSTA ) develop in concert with a deepened thermocline to enhance convection and potential TC genesis northeast of Madagascar and west of about 75E 85E (Chambers et al ., 1999; Jury et al ., 1999; Klein et al ., 1999; Reason et al ., 2000; Xie et al ., 2002; Ho et al ., 2006; Camargo et al ., 2007c; Kuleshov et al ., 2009). During a cool ENSO event con vection along the Intertropical Convergence Zone (ITCZ) is enhanced east of about 75E 85E and TC genesis is more frequent owing to positive SST A increased midtropospheric relative humidity, and lower tropospheric
72 shear from near equatorial westerlies ( Wolter, 1987; Larkin and Harrison, 2001; Ho et al ., 2006; Camargo et al ., 2007c; Kuleshov et al ., 2009). These documented ENSO TC genesis relationships were accounted for in the first stage of the cluster analysis (CA) of SIO TC trajectories, which was det ailed in Chapter 3. The first clustering algorithm separated 187 TC trajectories into four principal groups by their initial longitudes: three in the main development region east of 52E to near 110E (western, central, eastern) and one in the far western SIO near Madagascar and within the Mozambique Channel. The western development region is favored with strong El Nio conditions, the eastern region is favored with strong La Nia conditions, and the central region represents an area of variable ENSO infl uence. A second round of clustering then subdivided each of the three main formation regions into two based on the latitude and longitude of each TCs final position. Using this method the results of the CA allow for comparative analyses of groups that share a geographic development region but differ in their directions of motion. ENSO is known to associate with wind anomalies in the SIO that influence TC trajectories. During El Nio westerly anomalies are present over the tropical and subtropical SIO (van Loon and Rogers, 1981; Meehl, 1987; Karoly, 1989; Reason et al ., 2000; Ho et al ., 2006; Yoo et al ., 2006). It has been suggested that these westerlies steer TCs to recurve to the east away from Madagascar, (Jury et al ., 1999; Xie et al ., 2002; Vitart et al ., 2003; Ho et al ., 2006). La Nia is associated with easterly steering anomalies stemming from a strong SIO subtropical high, and TCs exhibit straighter westward trajectories during the cool ENSO phase (Wolter, 1987; Vitart et al ., 2003). D espite these findings however, sole consideration of ENSO to predict seasonal
73 steering flow, and therefore landfalls, has not sufficiently accounted for spatiotemporal variations of the TC trajectories (Vitart et al ., 2003; Klinman and Reason, 2008). This study addresses the ENSO shortcomings through consideration of the SIOD as a second phenomenon associated with significant variability of the SIO subtropical high and therefore TC steering flow. The SIOD represents the second empirical orthogonal function of Indo Pacific SST A and often occurs in the absence of or in opposition to the ENSO teleconnection (Behera and Yamagata, 2001; Fauchereau et al ., 2003 ; Washington and Preston, 2006). During the positive phase, SST A are warm in the western subtropical SIO and cool in the eastern subtropical SIO, similar to La Nia. During the negative phase, anomalies are generally r eversed, with a spatial SIO SSTA signature not unlike El Nio (Behera and Yamagata, 2001; Suzuki et al ., 2004; Huang and Shukla, 2008) S IOD can be used to infer the strength and direction of the TC steering flow anomalies because the SST A are mainly the result of Ekman transport and latent and sensible heat flux between the ocean surface and the planetary boundary layer (Behera and Yamagat a, 2001; Reason, 2002). This study is unique in that SIOD has not previously been tested for association with SIO TC trajectories. SIOD and ENSO are also implicated in the development and evolution of austral summer tropical temperate troughs (TTTs). These tropical extratropical cloud bands are the most important synoptic feature related to precipitation across southern Africa ( Harangozo and Harrison, 1983; Lindesay et al ., 1986; Mason and Jury, 1997) During El Nio and/or a negative SIOD event, the principal tropical convective source of TTTs very often shifts over far eastern Africa, Madagascar, or into the western SIO in conjunction with westerly upper waves that propagate northward to near 10S ( Lyo ns,
74 1991; Mason and Jury, 1997; Todd and Washington, 1999; Cook, 2000; Reason, 2002; Fauchereau et al ., 2009; Manhique et al ., 2009 ; Miyasaka and Nakamura, 2010). Thus, given the known relationships between ENSO and SIO TC tracks, it is possible that cert ain couplings of ENSO and SIOD could show strong relationships to certain SIO TC trajectory types. Previous studies have established that El Nio is associated with positive zonal steering wind anomalies, but they have not tested for nor discussed the rel ationship in a broader synoptic context that includes the eastward shift of TTTs as the physical mechanism by which TC recurvature may occur with either or both El Nio and negative SIOD. Additionally, the research presented herein also addresses the inadequacy of the seasonal scale in TC trajectory trend prediction by testing monthly data. Using the trajectory clusters from Chapter 3 to stratify TCs by their initial and final positions, the relationships between the groups and both ENSO and SIOD will be t ested using monthly scale SSTA indices via nonparametric analysis of variance (ANOVA) The results will for the first time explicitly link patterns of TC tracks to intraseasonal variability in tropical Pacific SST s and subtropical S IO SST s. An ample the oretical framework to understand the mechanisms by which ENSO and SIOD exert steering influence on SIO TCs through longitudinal shifts of TTTs exists in the literature on rainfall variability in southern Africa. From this broader perspective, intraseasonal or perhaps even interannual forecasts of TC trajectories that account for both ENSO and SIOD phases could be utilized for improved predictions of the number of TCs likely to threaten the southwestern rim of the SIO.
75 Data and Methods The seven main clusters (C1 C7) from Chapter 3 are used now to analyze and compare the groups SST anomaly patterns in the important regions relating to ENSO and SIOD. These clusters were identified through an analysis of TC genesis and lysis points for 191 TCs that reached peak intensity of at least 30 ms 1 and passed west of 90E. The four TCs in C8 that formed very near Australia have been omitted from the analysis as they do not associate strongly with any particular ENSO or SIOD pattern. SST monthly index values i n numerical format were acquired online respectively for SIOD and ENSO at http://www.jamstec.go.jp/res/ress/behera/iosdindex.html and http://www.cpc.noaa.gov/data/indices/ Each TC is assigned the monthly values of the index values corresponding to the month(s) spanning each TCs life cycle. If a storms life cycle bridged two months, those two months values were averaged. Then, using the TC clusters as treatment groups, analysis of variance ( ANOVA) is performed for three ENSO SST A indices and for SIOD. As not every SST A variable was found to approximate a normal distribution (results not shown here) a nonparametric rank test to c ompare the medians of the seven clusters, the Kruskal Wallis Test (KW) with ties adjustment, is carried out using the software package NCSS (Kruskal and Wallis, 1952; Higgins, 2004). The null hypotheses for the tests are that the median values of the inde x values for the Nio or SIOD regions are not significantly different across all seven TC clusters. The alternative hypotheses are that there exists a significant difference in at least one pair of TC clusters in their median values of the Nio or SIOD in dices To ensure that the results of the KW tests are appropriately interpreted, ModifiedLevene Equal Variance Tests are applied in each case (Brown and Forsythe, 1974). If the groups do not differ
76 significantly in the ir variances, the results of the KW tests are accepted and investigated further where appropriate. Given a rejection of the null hypothesis for the KW tests and given that equality of variance is satisfied, it is then necessary to perform multiple comparisons amongst the groups to determine which TC trajectory groups are significantly different with respect to contemporaneous SST anomalies in the Nio and SIOD regions Since the data here are not assumed to approximate a normal distribution, Dunns rank sums procedure is used to test statis tical significance in multiple comparisons (Dunn, 1964). Both a standard Z test and a test using a Bonferroni correction are employed in the multiple comparisons. The decision of how best to account for experiment wide error rate is often subjective and dependent on the nature of the data being analyzed (Cabin and Mitchell, 2000). Thus, the results will be presented in tables reporting both liberal and conservative significance levels with respect to the probabilities of Type I errors As complementation for the ANOVA results, global SST A composite maps are constructed for each cluster and plotted in ArcGIS with overlaid SST A centers of action for the Nio and SIOD regions. This allows for qualitative examination of the SST A patterns to help visualize their geospatial signatures, keeping the formal statistical inferences from the Nio and SIOD indices firmly in mind. The dataset used here is the National Oceanic and Atmospheric Administration (NOAA) Optimum Interpolation (OI ) SST Version 2 which is available online (Reynolds et al ., 2002). The data are weekly SST means with 1.0 degree latitude by 1.0 degree longitude resolution, and the temporal coverage is from October 1981 to the present. A n SST average is extracted base d on the initial and final dates for each respective TCs life cycle. The spatial scope
77 of the data is global and the temporal scale is approximately the duration of each storms life cycle. Composites are then averaged according to cluster membership, such that one global SST composite exists for each of the seven clusters. To account for seasonality, monthly averages for each month from October to May were also downloaded from the same source (Reynolds et al ., 2002) and then composited over the period O ct. 1981 to May 2008. To obtain anomaly composites for each TC, the 26 year monthly composite was subtracted from each of the 175 individual TC event composites available during the study period. For example, TC Jayas life cycle spanned parts of March and April in 2007. Therefore, the final anomaly composite map for Jaya is the SST composite for that TCs life cycle minus the combined Mar ch April 26 year composite. For the KW ANOVA tests, three SST regions are outlined in the Pacific for ENSO (see Figur e 41): Nio 1+2 (0 10 S, 90W 80W), Nio 3.4 (5N 5S 170W 120W), and Nio4 (5N 5S 160E 150W). The Nio 3.4 (N3.4) index is commonly used alone to represent ENSO events, while the Nio1+2 (N1.2) region is sensitive to warm events and Nio4 (N4) is more sensitive to cold events (Trenberth, 1997; Hanley et al ., 2003). ENSO index data is utilized for all 1 87 TCs in the seven principal clusters. Two SST regions are also outlined for the SIOD (Figure 4 2): the Subtropical Dipole Index west pole (SDI west, 37S 27S, 55E 65E), and the Subtropical Dipole Index east pole (SDI east, 28S 18S, 90E 100E) The Subtropical Dipole Index (SDI) is calculated by subtracting the east pole from the west pole (Behera and Yamagata, 2001). SDI data is not yet available for 2008, therefore six TCs from 2008 were omitted, leaving a total of 181 TCs for the SDI tests.
78 Resu lts Kruskal Wallis ANOVA Results The results of the KW tests suggest that sea surface temperatures in the tropical Pacific and the subtropical Indian Oceans are significantly and contemporaneously associated with variability of SIO TC trajectories (Tables 4 1 4 3, 4 4, 4 5, and 46 ; Figures 43 4 4, 4 5, and 46 ) The null hypotheses of no significant differences in the median values of N3.4, N1.2, N4, and SDI across all TC clusters are and N3.4 and SDI are =0.001. The results of the Modified Levene Equal Variance tests for all groups are a failure to reject the null hypothesis of homoscedasticity, again at the 95% confidence level. Therefore, there is sufficient evidence to compare the individual clusters to dete rmine which groups significantly differ in association with N3.4 N1.2, N4, and SDI. Multiple Comparisons and SST Anomaly Composites Even on qualitative inspection of group median values for N3.4, N1.2, N4, and SIOD (Table 42), C4 appears to be associated with El Nio and negative SIO D while C6 and C3 appear to be highly associated with La Nia and positive SIOD This relates well with previous work in the basin in that a strong La Nia group (C6) originates in the east and has v ery westward oriented trajectories, while a strong El Nio group (C4) originates in the west and is characterized by south and eastward motion (Evans and Allan, 1992; Vitart et al ., 2003; Ho et al ., 2006; Kuleshov et al ., 2008 and 2009). Noticeably however, C3 forms in the western region but is apparently more associated with La Nia, in contrast to C4. In order to provide statistical support for this perceived relationship, and to explore further associations, KW multiple comparison z value tests
79 are per formed based on median values of N3.4, N1.2, N4, and SIOD across all seven clusters. Test results show that there is statistical evidence to support the claim that C4 is significantly more associated wit h El Nio as compared to the six other groups and based on all three Nio index regions (Tables 4 3 4 4, and 45 ) The sea surface temperature anomaly (SST A ) composite for C4 (Figure 47) corroborates this finding well, with positive SST A evident from the South Amer ican littoral west to 170E. Also present is a SSTA pattern in the SIO suggestive of the negative SIOD which exhibits warm anomalies from Madagascar eastward along 20S into the eastern ocean (Behera and Yamagata, 2001; Suzuki et al ., 2004). Indeed, the SIOD multiple comparisons (Table 46 ) show that C4 is significantly more associated with the negative SIOD phase than all other clusters. El Nio is well known to associate with westerly wind anomalies over the SIO during austral warm season (van Loon and Rogers, 1981; Meehl, 1987; Karoly, 1989; Reason et al ., 2000; Yoo et al ., 2006) Likewise, TTTs are associated with anomalous westerlies over the same geographic area which extend from the boundary layer up to 500 hPa and have a longitudinal signature that can extend across the entire S IO ( Todd and Washington, 1999). Reason (2002) found these TTTs to vary in association with SIOD, and several authors found that the deepest convection coincident with TTTs shifts over the SIO in association with E l Nio (Cook, 2000; Pohl et al ., 2009; Fauchereau et al ., 2009; Manhique et al ., 2009). Given the high significance of both El Nio and negative SIOD for C4, and the propens ity of TTTs to prevail over Madagascar and the western SIO during both of these phenomena, the conclusion is
80 that C4 type TCs occur when ENSO is in warm phase and SIOD is in negative phase. The longitudinally shifted and latitudinally amplified TTTs then explain the mechanism by which TCs in C4 are swept southeastward away from Madagascar. H aving established the oceanatmospheric connections for C4, it is logical now to compare with the cluster that forms in the same region yet threaten s land with far great er frequency, C3. This group is significantly different in N3.4, N1.2, and N4, and SDI from C4, which strongly suggests that C3 occurs with La Nia and/ or a positive SIOD (Tables 43 4 4 4 5 and 46 ). The SSTA map (Figure 48) suggests a similar conclusion, with cool anomalies in the equatorial Pacific and warm anomalies south and south east of Madagascar. According to the multiple comparison tests for N3.4, N1.2, N4, and SDI C3 is not significantly different from any other group other than C4 and this suggests that the phasing of La Nia and the positive SIOD is not as prevalent in this group as was described for C4 with ENSO in warm phase, SIOD in negative mode, and TTT s consequently shifted north and east into the tropical SIO T he known regional patterns during La Nia should support more westward moving TCs that would threaten Madagascar with enhanced easterly and southeasterly trades across the tropical SIO (Wolter, 1987; Reason et al ., 2000; Larkin and Harrison, 2001; Vitart et al ., 2003 ). In accordance with positive SIOD, warm SSTA off of southern Africa increase convection over the far southern reaches of the continent coincident with the southwestward shifted TTTs (Reason, 2002; Todd and Washington, 1999). For storms in C3 that reached south of 25S, interaction with upstream continental TTTs would explain late recurvature. A difference map of C4 SST A minus C3 SST A (Figure 4 9), coupled with the Z values in the multiple compar isons, support the assertion that
81 both ENSO and SIOD phases should be accounted for to gain a fuller understanding of the mechanisms under which TC motion is more likely to mirror the C4 type with TTT interaction and southeastward movement or the C3 type with stronger trades and westward movement The trajectory type that develops in the eastern region and tracks far westward, group C6, also is stron gly associated with the positive SIOD phase and La Nia. Readily identifiable La N i a and positive SIOD signal s are apparent both in the high significance of C6 in the multiple comparisons (Tables 43, 4 4 4 5 and 46 ) and in the SSTA map (Figure 4 10) The composite map shows equatorial cool anomalies in the Pacific as well as coherent warm anomalies off the southeast coast of Africa into the southwest IO. Cool anomalies also extend eastward from Madagascar along 15S 25S across the SIO. This is cr ucial because these cool SST A are largely induced by increased sea to air latent and sensible heat exchange and Ekman transport associated with the strong easterly and southeasterly winds (Behera and Yamagata, 2001). Therefo re, the presence of these SSTA allow s inference of anomalously strong easterly trade winds across the SIO north of 25 S which c ould serve to steer TCs on long duration westward zonal tracks, as in Vitart et al (2003). T herefore, w hen La Nia and positive SDI occur simultaneously the TCs of this group remain at lower latitudes because the strong subtropical SIO high and associated easterlies preclude repeated intrusions of midlatitude transient waves which in turn fail to foster eastward propagation of convective elements associated with the Angola/Zambia thermal trough. Thus, the conclusions of Vitart et al (2003) can be improved upon.
82 While TCs in group C6 develop in the eastern SIO and track far westward, TCs in the counterpart group C1 also develop in the east but recurve into t he mid latitudes in the central or eastern SIO The SST A composite map (Figure 4 11 ) for C1 does not depict a strong ENSO warm or cool SST A pattern. However, the data exhibit a northwest southeast warm SSTA in the central and southeast subtropical IO which strongly resembles the negative SIOD mode. A map of the differences in the C1 and C6 composites (Figure 41 2 ) show significant differences in SSTA patterns in the ENSO and SIOD regions. In particular, the presence of warm SST A in the central and eastern tropical /subtropical SIO and cool SSTA in the southwest sbutropical SIO indicates weakened trade winds and more frequent frontal intrusions. This would mean stronger cold advection and anomalous westerlies allowing for greater sea to air sensible and latent heat flux to the south. Also, reduced Ekman transport from the usual trades to the north, in conjunction with increased air to sea heat flux from warm advection ahead of the cold fronts, would allow for warm SST A patterns associated with negative SIOD and/or El Nio. Therefore, with troughs moving farther north into the central/eastern SIO, this would influence C1 TCs to recurve more abruptly into the subtropics than C6 TCs T he multiple comparisons (Ta ble s 4 3 4 4 4 5 and 46 ) lend credence to the SSTA pattern s and provid e support to the assertion that group C1 is more of a negative SIOD related phenomenon and perhaps less dependent on warm/cool ENSO phase. This is in contrast to the eastern counterpart group C6 which occurs when ENSO is in cool phase and SIOD is in warm (positive) phase. The fact that C1 is composed of generally recurving TCs but does not have a strong El Nio signal supports the notion
83 that significant alteration in the TC steering wind patterns via a shift or weakening of the subtropical high can happen without strong ENSO influence (Behera and Yamagata, 2001; Fauchereau et al ., 2003; Washington and Preston, 2006). The TC cluster with genesis in the central region and movement to the west and southwest is C2. C2 is not significantly different from other groups in the ENSO tests save for C4 and only significantly different from C6 in the SDI multiple comparisons (Tables 43 4 4 4 5 and 46 ) However, Figure 46 should n ot be interpreted to mean that C2 shows a strong relationship to with SIOD that is opposite in sign to the strong relationship that is apparent between C6 and SIOD It is a strong relationship compared to a weak one, not a strong positive relationship com pared to a strong negative relationship. The SSTA composite map for C2 displays an easily recognizable La Nia pattern in the equatorial Pacific east of 180 (Figure 4 1 3 ) which is expected given the more east to west motion in the group that should accompany stronger SIO trade winds (Wolter, 1987; Reason et al ., 2000; Larkin and Harrison, 2001). Though not statistically significant according to Tables 43, 4 4, or 45 the ENSO indices also indicate a possible weak association with La Nia (Table 42). However, there is not an easily identifiable pattern in the C2 SST A map that would indicate a trade wind pattern shift associated with the SIOD, congruent with earlier dis cussion of the weak statistical SDI signal. Since TCs of type C2 sometimes threaten inhabited places in the SIO, it is important again to compare the C2 SST A anomaly map with another group (C7) which shares the central genesis region but is characterized by TCs that track more eastward through the open SIO. C2 and C7 are not statistically different when compared simply
84 by the index values for ENSO and SDI. There are signs of an El Nio pattern in the SST A map in the N1.2 region for C7, however there is not a strong contiguous warm SST A from South America to the date line (Figure 414). The difference map of C7 min us C2 SSTA (Figure 4 15) assists in identification of the equatorial Pacific as a region of importance, with warm SSTA off of South America and somewhat incoherently spread westward to the date line. Overall, there is not a significant statistical nor a strong geospatial difference between C2 and C7 that would allow for concise interpretation (Tables 43, 4 4, 4 5, and 46) The SSTA patterns in the SIO that might typically point to a positive/negative SIOD appear to have been disrupted by the TCs in both C2 and C7, with cold anomalies present along the TC tracks likely linked to cold water upwelling in the wake of the TCs. Finally, group C5 whic h formed mainly near or within the Mozambique Channel did not show any signific ance in the multiple comparison tests with respect to ENSO or SDI. An SSTA composite map was constructed for C5, but no coherent ENSO or SIOD patterns are evident (Figure 41 6 ) This is somewhat surprising in that these storms are located in close proximity to the region where TTTs often occur. The lack of a clear signal may be a function of the relatively small number of storms that form in this region From 19792008, only 40 of 339 (11.8%) TCs of any intensity in the western SIO formed in this region, and only 17 of those are included in the present study Therefore, it can only be said from this study that no strong ENSO or SIOD signal was found when only more intense TCs of this region that occurred during the period 1979 2008 were considered. An in depth analysis of these storms on a caseby case basis could yield a
85 better understanding of their relationship with TTTs as suggested by Mavume et al ( 2009). Discussion In this chapter, KW ANOVA tests were performed to compare the median values of N3.4, N1.2 N4 and SDI across the seven main TC trajectory clusters from Chapter 3. All ENSO indices and the SDI showed significant differences when all groups were tested tog ether Given that Vitart et al (2003) suggested a basinwide influence of E l Nio /La Nia on TC steering flow, the ENSO resul t s (particularly for N3.4) were not unexpected. Likewise, given the association between SST A and the strength of the trade winds in the SIO, it was not surprising that the SIOD was implicated in TC t rajectory variability at high significance level s comparable to ENSO (Behera and Yamagata, 2001; Suzuki et al ., 2004; Huang and Shukla, 2007) The link between SIOD and TC trajectories shown here represents a new finding not seen in previous literature. To explore the TC SIOD ENSO links further, multiple comparisons were made to find where the significance exists between clusters and SST A composite maps were constructed to aid interpretation of geospatial patterns Group C4 (western genesis, south motion) was found to associate highly with both El Nio and negative SIOD in comparison to all other clusters. The proposed physical mechanism behind this association is the northeastward shift of TTTs from southern Africa into the western SIO that often accompanies both E l Nio and negative SIOD (Mason and Jury, 1997; Todd and Washington, 1999; Cook, 2000; Reason, 2002; Fauchereau et al ., 2009; Manhique et al ., 2009) Anomalous westerly winds from the lower troposphere up to 500 hPa accompany the TTTs and are deep enough to steer TCs in the western SIO south and
86 eastward away from Madasgascar. The eastward shift of TTTs during El Nio and negative SIOD has been well documented in previous literature relating to African rainfall, but has not previously been linked to TC trajectories in the SIO. TCs that form in the western SIO and tend to move west or southwestward (group C3) were found to be significantly different t han group C4 and are more associated with La Nia and the positive phase of SIOD. Both La Nia and positive SIOD have been previously linked to stronger east and southeasterly trade winds in the tropical/subtropical SIO (Wolter, 1987; Behera and Yamagata, 2001; Vitart et al ., 2003) This finding is important to discern the ocean ic atmospheric environment of TCs that form near Madagascar but track east ward over the ocean from those that form in the same region and often make landfall. Consideration of ENS O phase alone is not sufficient to capture shifts in the subtropical /tropical TC steering flow in the SIO associated with the subtropical high and TTTs. The phase of the SIOD should be considered as well, as SST A in this region can provide additional expl anation for the variability of TC trajectories This would be particularly useful when ENSO is neutral and SIOD is either strong positive or negative. Likewise, when ENSO is in warm (cool) phase and SIOD is concurrently in negative (positive) phase, thes e interact to produce relatively persistent eastward (westward) TC steering environments. For example, Klinman and Reason (2008) noted that TC Favio in February 2007 followed a track south of Madagascar and then northwest making landfall in Mozambique. Th e system followed a highly unusual path given its occurrence during a weak El Nio year, and the authors noted that such a TC did not fit neatly into the ENSO TC track paradigm of Vitart et al (2003) wherein Mozambique is under threat for
87 landfall during La Nia. Applying the results from this study the SIOD was strongly posi tive in Feb. 2007, which is associated with more zonally oriented westward TC tracks. Therefore, in this case consideration of both ENSO and SIOD phases allows the track of Favio to fit within a more robust framework. A similar dichotomy was found for groups C 6 and C 1 in the eastern SIO. TCs that form in the eastern SIO and follow westward trajectories (C6) were significantly more associated with La Nia and positive SIOD than any other group. The physical mechanism is a nomalously strong southeast and easterly trade winds which are known to accompany both of these pat terns (Wolter, 1987; Behera and Yamagata, 2001; Vitart et al ., 2003) Group C1 is charac terized by TCs that also form in the eastern region, except these tend to recurve into the subtropics and remain well east of the inhabited western rim of the SIO. This group was found to be strongly associated with negative SIOD yet not strongly associated with any ENSO phase. The TTTs, though most often studied in the context of southern African rainfall, may extend even into the central and eastern subtropical SIO during negative SIOD (Todd and Washington, 1999; England et al ., 2006) Thus once agai n in the eastern SIO a more complete understanding of the oceanic atmospheric differences between low latitude westward moving TCs and those that recurve into the midlatitudes can be achieved by considering the SIOD phase in tandem with the ENSO phase. A r eview of previous literature revealed that the degree of association between ENSO and SIOD is not fully understood. The SIOD was noticed in EOF analysis of IndoPacific SSTs as the second EOF behind a leading EOF that was apparently strongly linked to ENS O ( Behera and Yamagata, 2001; Suzuki et al ., 2004). It does not
88 necessarily follow that the underlying cause(s) associated with EOF2 is unrelated to the underlying cause(s) associated with EOF1 (Fauchereau et al ., 2003; Washington and Preston, 2006) The spatial patterns of SST A can only be interpreted as orthogonal centers of variability not symptoms of two independent and noninteractive causes In this study, N3.4 and SDI have a contemporaneous Spearmans rank correlation of 0.1665 (p value=0.0251) and neither N1.2 nor N4 are significantly correlated (not shown) Thus, there is statistical significance in their simultaneous association, but not to a level where one can be said to drive the other. Future research should explore the relationship between ENSO forcing and internal SIO forcing more comprehensively, as interpretable lag lead relationships may be found. It has already been suggested from coral reef analysis that the spatial patterns of SST and sea level pressure (SLP) re sulting from ENSO SIOD interactions has changed since the 1970s (Zinke et al ., 2004). These changes likely produce alterations in TC trajectories, and it is important to try and determine if socially and economically vulnerable societies can expect increased or decreased exposure to negative TC impacts. G iven that the results described above suggest the importance of considering both ENSO and SIOD a contingency table (Table 47 ) is constructed stratifying the seven clusters by TCs that occurred in four ca tegories of ENSO/SIOD interaction: positive ENSO with positive SIOD (E+S+) positive ENSO with negative SIOD (E+S ) negative ENSO with positive SIOD (E S+) and negative ENSO with negative SIOD (E S ) Fishers Exact Test is known to provide exact p valu es for contingency table tests thereby remaining robust in the presence of small samples (Higgins, 2004). Therefore, using Fishers Test the groups from the same TC genesis regions as defined in the
89 earlier cluster analysis are tested for association wit h the interactive ENSO/SIOD categories. The results of the Fishers Test indicate that TC trajectories and type of interaction between ENSO and SIOD are significant for the western and eastern regions of the study area. For the western region, a highly si gnificant association is found between the type of TC trajectories and the type of interaction between ENSO and SIOD (p value=0.00002). Out of 22 (36) TCs in C4 (C3) 16 (4) occurred when ENSO was positive and SIOD was negative. In contrast, one (13) out of 22 (36) occurred when ENSO was negative (La Nia) and SIOD was positive. For the eastern region, a significant association again exists between the type of TC trajectories and the type of interaction between ENSO and SIOD (pvalue=0.025). Out of 39 ( 16) TCs in C1 (C6), 13 (0) occurred when ENSO was positive and SIOD was negative, whereas when ENSO was negative and SIOD was positive the ratios were 10/39 (7/16). No such significant association exists in the central region between C2 and C7 (pvalue=0. 195). These results further support the notions that type C4 storms are highly associated with an anti phasing of ENSO (warm) and SIOD (negative) wherein conditions are very favorable for TC trough interaction, and that type C6 storms are also associated with the opposite anti phasing of ENSO (cool) and SIOD (positive) wherein conditions are unfavorable for trough interaction across the entire SIO. Conclusion The KW ANOVA and subsequent tests described in this chapter have allowed comparisons between the T C trajectory clusters from Chapter 3 in terms of their associations with three ENSO indices and a SIOD index. There is statistical evidence that ENSO phase is an important factor for TC trajectories in the SIO, a finding that is in
90 agreement with previous research (Vitart et al ., 2003). There is also statistical evidence that SIOD phase is an equally important factor for TC trajectories in the SIO TTTs are important producers of rainfall on the western rim of the SIO, and their most intense tropical convective clusters often shift over the western SIO during E l Nio and/or the negative SIOD phase. A broad literature base exists on the topic because TTTs contribute substantially to warm season rainfall in southern Africa. However, TTTs as synoptic features integrally linked with tropical convection have not been linked to TC trajectories in the SIO The significance of ENSO and SIOD on SIO TC trajectories is logically explained in this study through the existing framework from the TTT literature. The ENSO/SIOD interaction is an important consideration which is not found in the existing literature. Previous research suggests that SIOD fluctuations may develop independent of ENSO phase (Behera and Yamagata, 2001; Fauchereau et al ., 2003; Washington and Preston, 2006) ENSO alone does not account for the observed variability of SIO TC trajectories, but when ENSO and SIOD are considered in tandem a better accounting can be made for changes in strength and position of the SIO subtropical high and the TTT s of the South Indian Convergence Zone (SICZ). Thus, when ENSO and SIOD are in anti phase there can be higher confidence of the direction and magnitude of anomalous TC steering winds in the SIO. The proposed importance of ENSO/SIOD interaction was then tested, and the resul ts from the western and eastern regions give further weight to the assertion that anti phasing of ENSO and SIOD is significantly associated with both anomalously westward and eastward moving TCs The statistical inferences are linked to physical mechanisms of TC steering flow because the ENSO and SIOD indices give
91 approximations on the strength and extent of SIO southeast trades versus anomalous low middle tropospheric westerlies associated respectively with the SIO subtropical high and the tropical temperate troughs. The following chapter will conclude this study by summarizing the important results and pointing to future directions in SIO TC research.
92 Table 41 P values for KW ANOVA and ModifiedLevene Equal Variance Tests. All tes ts evaluated with alpha level at 0.05 and red text indicates statistical significance at the given alpha level. KW ANOVA pvalues are corrected for ties Nio 3.4 Nio 1.2 Nio 4 SDI KW ANOVA 0.0002 0. 0051 0. 0017 0.0002 Mod. Levene 0.7516 0. 1114 0. 0901 0.1048 Table 42 Standardized median anomalies for Nio3.4 (N3.4) sea surface temperature, Nio 1.2 (N1.2), Nio4 (N4) and Subtropical Dipole Index (SDI). Pos i tive values of N1.2, N3.4 and N4 indicate a trend toward El Nio conditions, while negative values indicate a trend toward La Nia conditions. SDI is positive when the southwest IO is warmer than average while the southeast IO is cooler and SDI is negative when the inverse pattern occur s. Cluster ID Nio 3.4 Nio 1.2 Nio 4 SDI C1 0.02 0.2 0 0. 05 0.19 C2 0.23 0. 31 0.1 6 0.06 C3 0.41 0. 29 0. 16 0.34 C4 0.71 0.5 4 0. 68 0.70 C5 0.08 0. 09 0.0 4 0.32 C6 0.58 0.5 4 0. 58 0.77 C7 0.03 0.2 1 0. 10 0.14 Table 43. KW Multiple Comparison Z Value Tests for Nio3.4 region stratified by clusters. Text shaded in red or blue indicates significant differences of median Nio3.4 values at alpha level of 0.05. Text shaded only in red indicates significance using a Bonferroni correcti on to account for experiment wide accumulated probability of Type I errors. Nio 3.4 C1 C2 C3 C4 C5 C6 C7 C1 0.0000 0.9546 1.8630 2.9438 0.2625 2.0139 0.2301 C2 0.0000 0.9027 3.7362 0.4777 1.2830 0.9189 C3 0.0000 4.5029 1.1811 0.5944 1.5750 C4 0.0000 2.6551 4.1913 2.0765 C5 0.0000 1.4921 0.4086 C6 0.0000 1.8231 C7 0.0000
93 Table 44. KW Multiple Comparison Z Value Tests for Nio1.2 region stratified by clusters. Text shaded in red or blue indicates significant differences of median Nio1.2 values at alpha level of 0.05. Text shaded only in red indicates significance using a Bonferroni correction to account for experiment wide accumulated probability of Type I errors. Nio 1.2 C1 C2 C3 C4 C5 C6 C7 C1 0.0000 0.4709 0.6646 3.1539 1.7057 0.9658 1.9593 C2 0.0000 0.1925 2.7421 1.3345 1.3192 1.6126 C3 0.0000 2.5787 1.1846 1.4660 1.4727 C4 0.0000 1.0631 3.4173 0.6690 C5 0.0000 2.2379 0.3175 C6 0.0000 2.4431 C7 0.0000 Table 45. KW Multiple Comparison Z Value Tests for Nio4 region stratified by clusters. Text shaded in red or blue indicates significant differences of median Nio4 values at alpha level of 0.05. Text shaded only in red indicates significance using a Bonferroni correction to account for experiment wide accumulated probability of Type I errors. Nio 4 C1 C2 C3 C4 C5 C6 C7 C1 0.0000 0.5731 0.8373 2.8491 0.0272 2.1490 0.3357 C2 0.0000 0.2625 3.3199 0.4709 1.7067 0.7486 C3 0.0000 3.5429 0.6755 1.5065 0.9394 C4 0.0000 2.3175 4.2364 1.9070 C5 0.0000 1.8476 0.2670 C6 0.0000 2.0219 C7 0.0000 Table 46 KW Multiple Comparison Z Value Tests for Subtropical Dipole Index (SDI) region stratified by clusters. Text shaded in red or blue indicates significant differences of median SDI values at alpha level of 0.05. Text shaded only in red indicates significance using a Bonferroni correction to account for experiment wide accumulated probability of Type I errors. SDI C1 C2 C3 C4 C5 C6 C7 C1 0.0000 0.6145 1.8983 2.1023 1.4886 3.0697 1.0640 C2 0.0000 1.2843 2.6152 1.0129 2.5881 0.6123 C3 0.0000 3.6928 0.0106 1.5729 0.3405 C4 0.0000 3.0512 4.4799 2.6093 C5 0.0000 1.3277 0.3018 C6 0.0000 1.5845 C7 0.0000
94 Table 47 Tropical cy clones by cluster and ENSO/SDI phases. Red (blue) text signifies positive (negative) phase of ENSO or SDI. Western region in orange (Fishers test p value 0.00002). Eastern region in green (pvalue 0.025). Central region in black (pvalue 0.195). The SDI is not available after 2007, hence the sample size is only 181. C1 C2 C3 C4 C5 C6 C7 Total ENSO / SDI 6 9 9 3 4 5 4 40 ENSO /SDI 13 8 4 16 5 0 5 51 ENSO/ SDI 10 10 13 1 5 7 2 48 ENSO/SDI 10 11 10 2 2 4 3 42 Total 39 38 36 22 16 16 14 181 Figure 41. El Nio sea surface temperature index regions. Nino1+2 is outlined in red, Nio 3.4 in hatching, and Nio4 in yellow. Figure 42 Subtropical Dipole Index SST regions. Dipole West ( SDI West ) in orange, and S ubtropical Dipole East (SDI_East) in green.
95 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 1 2 3 4 5 6 7Cluster IDsNino-3.4 Index Figure 43. Box plots of standardized sea surface temperature anomalies for the Nio3.4 region stratified by cl uster ID.
96 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 1 2 3 4 5 6 7Cluster IDsNino-1.2 Index Figure 44. Box plots of standardized sea surface temperature anomalies for the Nio1.2 region stratified by cluster ID.
97 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 1 2 3 4 5 6 7Cluster IDsNino-4 Index Figure 45. Box plots of standardized anomalies for the Nio4 region stratified by cluster ID.
98 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 1 2 3 4 5 6 7Cluster IDsSDI Figure 46 Box plots of standardized anomalies for the Indian Ocean Subtropical Dipole Index (SDI) stratified by cluster ID.
99 Figure 47. Composite map of sea surface temperatur e anomalies for Cluster 4 (C4), western region, south motion.
100 Figure 48. Composite map of sea surface temperature anomalies for Cluster 3 (C3), western region, southwest motion.
101 Figure 49. Differe nce of sea surface temperature anomalies between C4 and C3.
102 Figure 410. Composite map of sea surface temperature anomalies for Cluster 6 (C6), eastern region, west motion.
103 Figure 411. Composite map of sea surface temperature anomalies for Cluster 1 (C1), eastern region, south m otion. Figure 412. Difference of sea surface temperature anomalies between C1 and C6.
104 Figure 41 3 Composite map of sea surface temperature anomalies for Cluster 2 (C2), central region, southwest motion.
105 Figure 414. Composite map of sea surface temperature anomalies for Cluster 7 (C7), central region, south motion.
106 Figure 415. Difference of sea surface temperature anomalies between C7 and C2.
107 Figure 41 6 Composite map of sea surface temperature anomalies for Cluster 5 (C5), Mozambique Channel region.
108 CHAPTER 5 CONCLUSIONS This study tested the hypotheses that tropical cyclone (TC) trajectories in the South Indian Ocean (SIO) are influenced by oceanic atmospheric variability associated with El Nio Southern Oscillation (ENSO) and /or the internal sea surface temperature (SST) dipoles of the subtropical SIO The significance of t his work is largely in testing SIOD for association with SIO TC trajectories to build upon the previously known, but insufficient association with ENSO. The results of Fisher s Test in Chapter 4 hone in on the most specific conclusion from this study: SI OD and ENSO together have the greatest explanatory power for highly exceptional types of TC trajectories such as C4 and C6. These TCs represent easily identifiable departures from normal which occurs when ENSO and SIOD are in anti phase. In the case of C4, the TCs track southeastward away from land, whereas in the case of C6, TCs track much farther westward than usual and can threaten populated regions of the SIO. I t is paramount to understand their potential influence on the tracks o f SIO TCs which regularly threaten and disrupt the lives of the regions inhabitants during the austral warm season. The broad steering mechanisms of TCs are now well understood, as related to short term interactions between a TC and its surrounding envir onment. The goal here was to extend and ap ply this understanding across a large sample of events to relate the sample of TCs to larger scale synoptic regimes known to vary according to ENSO and SIOD Hierarchical cluster analysis (CA) was employed to group TCs into four main genesis regions: western (55E 73E), central (74E 86E), eastern (88E 104E) and Mozambique Channel (west of 55E) Subsequently, these were grouped by the geographic location at which the storms life cycle s ended, which resulted in seven main
109 groups of SIO TC trajectories one in the Mozambique Channel and two in each of the other three genesis regions This structuring allowed for groups from the same TC genesis region but with disparate trajectory directions to be tested in r elation to ENSO and SIOD, which are known to associate with changes in the S IO atmospheric circulation. Using nonparametric Kruskal Wallis analysis of variance (KW ANOVA), inferential testing was performed for ENS O using the Nio3.4, Nio 1.2, and Nio4 indices and for the Subtropical Indian Ocean Dipole (SIOD) using the Subtropical Dipole Index (SDI). The ANOVA results and follow on multiple comparisons indicated statistically significant differences between groups. Group C4 (western genesis, south and east motion) was more associated with El Nio and the negative mode of SIOD, while group C3 (western genesis, west and southwest motion) was more associated with La Nia and the positive mode of SIOD Within the eastern genesis region, group C6 (west mo tion) was more associated with La Nia and negative SIOD, while group C1 (south motion) occurred more often during El Nio and positive SIOD. The findings were crossvalidated qualitatively through SST anomaly (SST A ) composite maps which depicted easily discernible spatial SST A patterns in/near the indexed regions for Nio3.4 and SDI It was not surprising that ENSO phase was implicated by the data as significantly affecting TC trajectories, as this was found before in the SIO and other ocean basins. Ho wever, ENSO, while a necessary factor for consideration, is not sufficient to account for variability in SIO TC trajectories. The i dentification in this study of the SIOD as significantly relating to TC trajectories in the SIO is a new finding. SIOD is k nown to
110 associate with strong (weak) subtropical east and southeasterly trade winds and a strong (weak) parent SIO subtropical high during a positive (negative) event. Therefore, during a positive event TCs follow more westward trajectories than normal, s wept along within the strong trade winds, as seen in groups C6 and C3 In contrast, weakened trades during a negative SIOD event allow for extratropical influences on the TC trajectories, as in groups C4 and C1 which display more poleward and eastward movement than their counterparts. Based on previous literature relating to variability in African rainfall, the physical mechanisms proposed to influence this increased poleward motion for groups C4 and C1 are tropical temperate troughs (TTTs). During a negative SIOD event the trade winds weaken and shift these TTTs northeast over northern Mozambique and Madagascar and into the western SIO with the deepest convection sometimes focusing over the warm waters of the western SIO. Concurrent with the weakened subtropical high, midlatitude troughs penetrate farther equatorward and may interact with the eastward shifted continental African cyclonic circulation to extend anomalous westerlies into the SIO. These atypical winds from the TTTs can extend throughout t he troposphere deep enough to influence the track of a TC to turn poleward or eastwar d. Therefore, this study is unique in suggest ing a link between increased (decreased) tendencies of TC recurvature in the SIO during negative (positive) SIOD events thro ugh the mechanism of TTTs. The TTTs are also known to display very similar patterns during El Nio and La Nia with the negative (positive) SIOD patterns mirroring El Nio (La Nia). However, it has been established that the SIOD can occur largely independent of ENSO. Using the CA structure and stratifying by coincidences of El Nio (La Nia) with both positive and
111 negative SIOD events, it was found that the frequency of recurving (zonal) moving TCs is enhanced (reduced) when SIOD is negative (positive) and ENSO is simultaneously in warm (cool) phase. This finding builds on the work of Vitart et al (2003), who admitted that while important ENSO does not account for times when SIO TC steering flow varies independent of the ENSO signal The results from t his study suggest SIOD as a useful additional measure to account for the strength of the subtropical high through the SST A forced by anomalous trade winds. Furthermore, if SIOD is useful in accounting for anomalous TC tracks, then it should also be useful in consideration of vertical wind shear across the SIO. The extent and magnitude of vertical wind shear should in theory relate to the phase and magnitude of SIOD. There are many more unanswered questions to be addressed. Future work should consider SIOD alongside ENSO when assessing probabilities of TC strikes. However, other relationships should be explored as well. The North Atlantic Oscillation (NAO ) is known to affect the circulation and moisture fluxes associated with the convergent zones over southern Africa (McHugh and Rogers, 2001). Chang Seng and Jury (2010b) also suggested a link between the boreal winter circulation and SIO TC activity via o utflow from the Asian monsoon system. These inter hemispheric exchanges may affect the position and structure of the TTTs and thus whether they may translate eastward over the western SIO. Much more investigation should focus on the causes of variability in the SIO subtropical high, as it is a key mechanism through which TCs either remain at low latitudes or recurve into the temperate westerlies. SIOD may be characterized as a proxy measurement for the trade winds associated with the high, and though use of the Subtropical Dipole Index (SDI) showed statistically significant
112 results in this study it would be fruitful to represent the broader environmental atmospheric flow with an index not limited to such small centers of action. Reason (2001) noted that rainfall variability in southern Africa is sensitive to the geographi c location of the warmest SSTA in the SIOD, meaning that either a larger area could be used to index the SIOD (as in Suzuki et al ., 2004) or that another i ndex could represent warm SSTA closer to the African coast. Any improved accounting of the position and/or strength of the SIO subtropical high could potentially improve the ability to predict broad trends in SIO TC steering flow. It is important, especially for the developing nations along the western rim of the SIO, to be able to withstand the impacts of natural disturbances, including TCs. If understanding of teleconnections and the internal variability of the Indian Ocean can be improved, it would be beneficial to forecast whether a TC that might form one or three months in the future will be more or less likely to approach inhabited places. This information would not specify the exact point of landfall, for example, but could be given in probabilistic terms. G iven the formation of a hypothetical TC if some threshold probability were to be exceeded (meaning high confidence in a more westerly or easterly movement), then early disaster plans could be enacted. Crops could be harvested two weeks in advance, while food supplies and other necessary resources could be strategically relocated. The findings and explanations presented in this study can serve to spur further investigation of SIO TC trajectories, as the direction of their movement is of great importance in determining their ultimate societal impacts
113 APPENDIX TROPICAL CYCLONE S ANALYZED IN THE PRESENT STUDY Name Season Gen Month Lys Month Vmax Ivan 1979 3 3 60 Albine 1980 11 12 100 Viola/ Claudette 1980 12 12 110 Hyacinthe 1980 1 1 70 Jacinthe 1980 2 2 100 Fred 1980 2 2 95 Kolia 1980 2 3 60 Laure 1980 3 3 100 Florine 1981 1 1 105 Johanne 1981 3 3 80 Olga 1981 4 4 120 None 1982 10 10 85 Armelle 1982 11 11 80 Chris/ Damia 1982 1 1 120 Justine 1982 3 3 75 Karla 1982 4 5 100 Bemany 1983 11 12 65 Dadafy 1983 12 12 65 Elinah 1983 1 1 65 Naomi 1983 4 4 60 Oscar 1984 10 10 70 Andry 1984 12 12 130 Bakoly 1984 12 12 90 Fanja 1984 1 1 75 Jaminy 1984 2 2 100 Daryl 1984 3 3 75 Kamisy 1984 4 4 100 Celestina 1985 1 1 65 Ditra 1985 1 2 70 Gerimena 1985 2 2 65 Kirsty 1985 3 3 105 Helisaonin 1985 4 4 110 Delifina 1986 1 1 110 Costa 1986 1 1 70 Erinesta 1986 1 2 115 Gista 1986 2 2 85 Honorinina 1986 3 3 110
114 Name Season Gen Month Lys Month Vmax Jefotra 1986 3 4 105 Alison/ Krisostoma 1986 4 4 75 Billy_Lila 1986 5 5 95 Alinina 1987 1 1 75 Daodo 1987 3 3 75 Elizabeta 1987 4 4 75 Calidera 1988 1 1 65 Doaza 1988 1 2 115 Gwenda/ Ezenina 1988 2 2 90 Filao 1988 2 3 80 Gasitao 1988 3 3 130 Adelinina 1989 10 11 75 Barisaona 1989 11 11 100 Calasanjy 1989 1 1 75 Edme 1989 1 1 115 Firinga 1989 1 2 90 Leon/ Hanitra 1989 2 3 125 Gizela 1989 2 2 65 Jinabo 1989 3 3 65 Krisy 1989 3 4 105 Pedro 1990 11 11 65 Alibera 1990 12 1 135 Baomavo 1990 12 1 85 Cezera 1990 2 2 80 Dety 1990 1 2 95 Edisoana 1990 2 3 100 Gregoara 1990 3 3 110 Alison 1991 1 1 65 Bella 1991 1 2 130 Debra 1991 2 3 90 Elma 1991 2 3 60 Fatima 1991 3 4 90 Alexandra 1992 12 12 105 Harriet/ Heather 1992 2 3 120 Farida 1992 2 3 120 Jane_Irna 1992 4 4 120 Aviona 1993 9 10 65
115 Name Season Gen Month Lys Month Vmax Colina 1993 1 1 95 Edwina 1993 1 1 110 Finella 1993 2 2 75 Jourdanne 1993 4 4 125 Konita 1993 4 5 90 Alexina 1994 11 11 60 Cecilia 1994 12 12 85 Daisy 1994 1 1 95 Pearl/ Farah 1994 1 1 95 Geralda 1994 1 2 145 Hollanda 1994 2 2 105 Ivy 1994 2 2 100 Litanne 1994 3 3 130 Mariola 1994 3 3 90 Nadia 1994 3 4 120 Odille 1994 3 4 105 Albertine 1995 11 12 115 Dorina 1995 1 1 100 Gail 1995 1 2 75 Ingrid 1995 2 3 100 Josta 1995 3 3 65 Kylie 1995 3 3 85 Marlene 1995 3 4 125 Agnielle 1996 11 11 150 Bonita 1996 12 1 135 Hubert/ Coryna 1996 1 1 75 Doloresse 1996 2 2 75 Edwige 1996 2 2 95 Flossy 1996 2 3 115 Hansella 1996 4 4 95 Itelle 1996 4 4 140 Jenna 1996 4 5 60 Antoinette 1997 10 10 65 Melanie/ Bellamine 1997 10 11 125 Chantelle 1997 11 12 65 Daniella 1997 12 12 120 Phil 1997 12 1 85
116 Name Season Gen Month Lys Month Vmax Fabriola 1997 12 1 60 Pancho/ Helinda 1997 1 2 125 Gretelle 1997 1 1 115 Iletta 1997 1 1 75 Josie 1997 2 2 90 Karlette 1997 2 2 65 Lisette 1997 2 3 75 Rhonda 1997 5 5 100 Selwyn 1998 12 1 65 Anacelle 1998 2 2 115 Victor/ Cindy 1998 2 2 90 Elsie 1998 3 3 90 Gemma 1998 4 4 70 Alda 1999 1 1 65 Damien/ Birenda 1999 1 2 80 Davina 1999 3 3 110 Frederic/ Evrina 1999 3 4 140 Astride 2000 12 1 65 Babiola 2000 1 1 90 Connie 2000 1 2 120 Leon/ Eline 2000 2 2 115 Felicia 2000 2 2 65 Hudah 2000 3 4 125 Ando 2001 12 1 120 Bindu 2001 1 1 100 Charly 2001 1 1 105 Dera 2001 3 3 90 Evariste 2001 4 4 75 Bessi/ Bako 2002 11 12 75 Dina 2002 1 1 130 Eddy 2002 1 1 75 Francesca 2002 1 2 115 Guillaume 2002 2 2 120 Hary 2002 3 3 140 Ikala 2002 3 3 110 Dianne/ Jery 2002 4 4 105 Kesiny 2002 5 5 65
117 Name Season Gen Month Lys Month Vmax Boura 2003 11 11 75 Crystal 2003 12 12 90 Ebula 2003 1 1 65 Gerry 2003 2 2 105 Hape 2003 2 2 90 Japhet 2003 2 3 115 Kalunde 2003 3 3 140 Manou 2003 5 5 75 Beni 2004 11 11 100 Cela 2004 12 12 65 Jana 2004 12 12 80 Elita 2004 1 2 65 Frank 2004 1 2 125 Gafilo 2004 3 3 140 Helma 2004 3 3 65 Oscar/ Itseng 2004 3 3 115 Juba 2004 5 5 65 Arola 2005 11 11 65 Bento 2005 11 12 140 Chambo 2005 12 1 105 Ernest 2005 1 1 100 Gerard 2005 2 2 60 Hennie 2005 3 3 65 Adeline/ Juliet 2005 4 4 125 Bertie/ Alvin 2006 11 11 115 Boloetse 2006 1 2 100 Carina 2006 2 3 130 Bondo 2007 12 12 135 Clovis 2007 12 1 65 Dora 2007 1 2 115 Favio 2007 2 2 120 Gamede 2007 2 3 105 Humba 2007 2 3 80 Indlala 2007 3 3 120 Jaya 2007 3 4 110 Ariel 2008 11 11 60 Bongwe 2008 11 11 65
118 Name Season Gen Month Lys Month Vmax Fame 2008 1 2 85 Gula 2008 1 2 90 Hondo 2008 2 2 130 Ivan 2008 2 2 125 Jokwe 2008 3 3 110 Kamba 2008 3 3 115
119 LIST OF REFERENCES Anderson DE. 1918. The Epidemics of Mauritius, with a Descriptive and Historical Account of the Island. HK Lewis: London. Behera SK, Salvekar PS, Yamagata T. 2000. Simulation of Interannual SST Variability in the Tropical Indian Ocean. Journal of Climate 13: 34873499. Behera SK, Yamagata T. 2001. Subtropical SST dipole events in the southern Indian Ocean. Geophysical Research Letters 28 : 327 330. Bessafi M, LasserreBigorry A, Neumann CJ, Pignolet Tardan F, Lee Ching Ken, M. 2002. Statistical Prediction of Tropical C yclone Motion: An Analog CLIPER Approach. Weather and Forecasting 17: 821 831. Bessafi M, Wheeler MC. 2006. Modulation of South Indian Ocean Tropical Cyclones by the MaddenJulian Oscillation and Convectively Coupled Equatorial Waves. Monthly Weather R eview 134 : 638 656. Bjerknes J. 1969. Atmospheric Teleconnections from the Equatorial Pacific. Monthly Weather Review 97 : 163 172. Blender R, Fraedrich K, Lunkeit F. 1997. Identification of cyclonetrack regimes in the North Atlantic. Quarterly Journal of the Royal Meteorological Society 123 : 727 741. Brown ML. 2009. Madagascars Cyclone Vulnerability and the Global Vanilla Economy. The Political Economy of Hazards and Disasters Eds. Jones EC, Murphy AD. Altamira Press: New York. Brown MB, Forsythe AB. 1974. Robust Tests for the Equality of Variances Journal of the American Statistical Association 69 :364 367. Buchan A. 1901. Charles Meldrum. Nature 65: 9 11. Cabin RJ, Mitchell RJ. 2000. To Bonferroni or Not to Bonferroni: When a nd How Are the Questions. Bulletin of the Ecological Society of America 81: 246 248. Camargo SJ, Robertson AW, Gaffney SJ, Smyth P, Ghil M. 2007a. Cluster Analysis of Typhoon Tracks. Part I: General Properties. Journal of Climate 20: 36353653. Camargo SJ, Robertson AW, Gaffney SJ, Smyth P, Ghil M. 2007b. Cluster Analysis of Typhoon Tracks. Part II: LargeScale Circulation and ENSO. Journal of Climate 20: 36543676.
120 Camargo SJ, Emanuel KA, Sobel AH. 2007c. Use of a Genesis Potential Index to Diagnos e ENSO Effects on Tropical Cyclone Genesis. Journal of Climate 20: 48194834. Camargo SJ, Robertson AW, Barnston AG, Ghil M. 2008. Clustering of eastern North Pacific tropical cyclone tracks: ENSO and MJO effects. Geochemistry Geophysics Geosystems 9 : Q06V05. Camargo SJ, Sobel AH, Barnston AG, Klotzbach PJ. 2009. The influence of natural climate variability on tropical cyclones and seasonal forecasts of tropical cyclone activity. Global Perspectives on Tropical Cyclones, 2nd edition World Scientific: New York; in press. Caroff P. Operational practices for best track elaboration at RSMC La R eunion. Presented 5 May 2009 at the International Best Track Archive for Climate Stewardship (IBTrACS) Worksho p, NOAA National Climatic Data C enter, Asheville, NC. Chambers DP, Tapley BD, Stewart RH. 1999. Anomalous warming in the Indian Ocean coincident with El Nio. Journal of Geophysical Research 104 : 3035 3047. Chan JCL. 2005. The Physics of Tropical Cyclone Motion. Annual Review of Flui d Mechanics 37: 99 128. Chan JCL, Gray WM. 1982. Tropical Cyclone Movement and Surrounding Flow Relationships. Monthly Weather Review 110 : 13541374. Chang Seng DS, Jury MR 2010a. Tropical cyclones in the SW Indian Ocean. Part 1: inter annual variabi lity and statistical prediction. Meteorology and Atmospheric Physics, in press. Chang Seng DS, Jury MR 2010 b Tropical cyclones in the SW Indian Ocean. Part 2: structure and impacts at the event scale. Meteorology and Atmospheric Physics in press. Choi K S, Kim BJ, Choi CY, Nam J C. 2009. Cluster analysis of Tropical Cyclones making landfall on the Korean Peninsula. Advances in Atmospheric Sciences 26: 202210. Chu JH, Sampson CR, Levine AS, Fukada E. 2002. The Joint Typhoon Warning Center tro pical cyclone best tracks, 1945 2000. Naval Research Laboratory Reference Number NRL/MR/7540 0216. Chu P S. 2004. ENSO and tropical cyclone activity. Hurricanes and Typhoons: Past, Present, and Future. Eds. Murnane RJ Liu K B. Columbia University Pres s: New York.
121 Cook KH. 2000. The South Indian Convergence Zone and Interannual Rainfall Variability over Southern Africa. Journal of Climate 13: 3789 3804. Dunn OJ. 1964. Multiple Comparisons Using Rank Sums. Technometrics 6 : 241 252. Dvorak VF. 1975. Tropical Cyclone Intensity Analysis and Forecasting from Satellite Imagery. Monthly Weather Review 103 : 420 430. Elsner JB. 2003. Tracking Hurricanes. Bulletin of the American Meteorological Society 84: 353356. Elsner JB, Kara AB. 1999. Hurricanes of the North Atlantic Oxford University Press: New York; 488. Elsner JB, Liu K B. 2003. Examining the ENSO typhoon hypothesis. Climate Research 25: 43 54. Emanuel K. 2003. Tropical Cyclones. Annual Review of Earth and Planetary Sciences 31: 75 104. Ema nuel K. 2005a Divine Wind: The History and Science of Hurricanes Oxford University Press: New York; 285. Emanuel K. 2005b. Increasing destructiveness of tropical cyclones over the past 30 years. Nature 436 : 686688. England MH, Ummenhofer CC, Santoso A. 2006. Interannual Rainfall Extremes over Southwest Western Australia Linked to Indian Ocean Climate Variability. Journal of Climate 19: 19481949. Evans JL, Allan RJ. El Nio/Southern Oscillation Modification to the Structure of the Monsoon and Tro pical Cyclone Activity in the Australasian Region. International Journal of Climatology 12: 611623. Fauchereau N, Trzaska S, Richard Y, Roucou P, Camberlin P. 2003. Seasurface temperature covariability in the Southern Atlantic and Indian Oceans and its connections with the atmospheric circulation in the Southern Hemisphere. International Journal of Climatology 23 : 663677. Fauchereau N, Pohl B, Reason CJC, Rouault M, Richard Y. 2009. Recurrent daily OLR patterns in the Southern Africa/Southwest Indian Ocean region, implications for South African rainfall and teleconnections. Climate Dynamics 32: 575 591. Frank WM, Young GS. 2007. The Interannual Variability of Tropical Cyc lones. Monthly Weather Review 135 : 35873598.
122 Fritz S, Wexler H. 1960. Cloud Pictures from Satellite TIROS I. Monthly Weather Review 88: 79 87. Gaffney SJ, Robertson AW, Smyth P, Camargo SJ, Ghil M. 2007. Probabilistic clustering of extratropical cycl ones using regression mixture models. Climate Dynamics 29: 423 440. Gong X, Richman MB. 1995. On the Application of Cluster Analysis to Growing Season Precipitation Data in North America East of the Rockies. Journal of Climate 8 : 897931. Gray WM, Sheaffer JD. 1991. El Nio and QBO influences on tropical cyclone activity. Teleconnections linking worldwide climate anomalies Eds. Glantz MH, Katz RW, Nicholls N. Cambridge University Press: Cambridge, UK; 535. Gutzler DS, Harrison DE. 1987. The Structure and Evolution of Seasonal Wind Anomalies over the Near E quatorial Eastern Indian and Western Pacific Oceans. Monthly Weather Review 115 : 169 192. Hanley DE, Bourassa MA, OBrien JJ, Smith SR, Spade ER. 2003. A Quantitative Evaluati on of ENSO Indices. Journal of Climate 16 : 1249 1258. Harangozo SA, Harrison MSJ. 1983. On the Use of Synoptic Data in Indicating the Presence of Cloud Bands over Southern Africa. South African Journal of Science 79: 413414. Harr PA, Elsberry RL. 1991. Tropical Cyclone Track Characteristics as a Function of LargeScale Circulation Anomalies. Monthly Weather Review 119 : 14481468. Harrison DE, Larkin NK. 1998. El NioSouthern Oscillation Sea Surface Temperature and Wind Anomalies, 19461993. Reviews of Geophysics 36 : 353 399. Hastenrath S. 1995. Recent Advances in Tropical Climate Prediction. Journal of Climate 8 : 15191532. Hastenrath S. 2000. Zonal Circulations over the Equatorial Indian Ocean. Journal of Climate 13: 27462756. Higgins JJ. 2004. An Introduction to Modern Nonparametric Statistics Brooks/Cole: Pacific Grove, CA; 366. Ho CH, Kim J H, Jeong J H, Kim HS, Chen D. 2006. Variation of tropical cyclone activity in the South Indian Ocean: El NioSouthern Oscillation and MaddenJulian Oscillation effects. Journal of Geophysical Research 111 : D22101.
123 Hope JR, Neumann CJ. 1970. An Operational Technique for Relating the Movement of Existing Tropical Cyclones to Past Tracks. Monthly Weather Review 98: 925933. Horel JD, Wallace JM. 1981. Planetary Scale Atmospheric Phenomena Associated with the Southern Oscillation. Monthly Weather Review 109 : 813 829. Huang B, Shukla J. 2007. Mechanisms for the Interannual Variability in the Tropical Indian Ocean. Part II: Regional Processes. Journal of Climate 20: 29372960. Huang B, Shukla J. 2008. Interannual variability of the South Indian Ocean in observations and a coupled model. Indian Journal of Marine Sciences 37 : 13 34. Julian PR, Chervin RM. 1978. A Study of the Southern Oscillation and Walker Circulation Phenomenon. Monthly Weather Review 106 : 1433 1451. Jury, MR. 1993. A Preliminary Study of Climatological Associations and Characteristics of Tropical Cyclones in the SW Indian Ocean. Meteorology and Atmospheric Physics 51 : 101 115. Jury MR, Pathack B, Wang B, Powell M, Raholijao N. 1993. A Destructive Tropical Cyclone Season in the SW Indian Ocean: January February 1984. South African Geographical Journal 75: 53 59. Jury MR, Parker B, Waliser D. 1994. Evolution and Variability of the ITCZ in the SW Indian Ocean: 198890. Theoretical and Applied Climatology 48 : 187194. Jury MR, Pathack B, Parker B. 1999. Climatic Determinants and Statistical Prediction of Tropical Cyclone Days in the Southwest Indian Ocean. Journal of Climate 12: 17381746. Kalnay M, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelews ki C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D. 1996. The NCEP/NCAR 40 Year Reanalysis Project. Bulletin of the American Meteorological Society 77: 437 471. Karoly DJ. 1989. Southern Hemisphere Circulation Features Associated with El NioSouthern Oscillation Events. Journal of Climate 2 : 12391252. Kidson JW. 1975. Tropical Eigenvector Analysis and the Southern Oscillation. Monthly Weather Review 103 : 187196. Kim HM, Webster PJ, Curry JA. 2009. Impact of Shifting Patterns of Pacific Ocean Warming on North Atlantic Tropical Cyclones. Science 325 : 77 80.
124 Klein SA, Soden BJ, Lau N C. 1999. Remote Sea Surface Temperature Variations during ENSO: Evidence for a Tropical Atmospheric Bridge. Journal of Climate 12: 917932. Klinman MG, Reason CJC. 2008. On the peculiar storm track of TC Favio during the 20062007 Southwest Indian Ocean tropical cyclone season and relationships to ENSO. Meteorology and Atmospheric Physics 100 : 233 242. Klotzbach PJ. 2007. Recent development s in statistical prediction of seasonal Atlantic basin tropical cyclone activity. Tellus 59A : 511518. Klotzbach PJ, Barnston AG, Bell G, Camargo SJ, Chan JCL, Lea A, Saunders M, Vitart F. 2010. Seasonal Forecasting of Tropical Cyclones. Ed. Guard C. Gl obal Guide to Tropical Cyclone Forecasting, 2nd edition World Meteorological Organization, in press. Knaff JA. 2009. Revisiting the maximum intensity of recurving tropical cyclones. International Journal of Climatology 29 : 827837. Knaff JA, Sampson CR. 2009. Southern hemisphere tropical cyclone intensity forecast methods at the Joint Typhoon Warning Center, Part I: control forecasts based on climatology and persistence. Australian Meteorological and Oceanographic Journal 58: 1 7. Kruskal WH, Wallis WA. 1952. Use of Ranks in OneCriterion Variance Analysis. Journal of the American Statistical Association 47 : 583 621. Kuleshov Y, de Hoedt G. 2003. Tropical Cyclone Activity in the Southern Hemisphere. Bulletin of the Australian Meteorological and Oceanographic Society 16: 135 137. Kuleshov Y, Qi L, Fawcett R, Jones D. 2008. On tropical activity in the Southern Hemisphere: Trends and the ENSO connection. Geophysical Research Letters 35: L14S08. Kuleshov Y, Ming FC, Qi L, Chouaibou I, Hoareau C, Roux F. 2009. Tropical cyclone genesis in the Southern Hemisphere and its relationship with the ENSO. Annales Geophysicae 27: 2523 2538. Lander MA. 1994. An Exploratory Analysis of the Relationship between Tropical Storm Formation i n the Western North Pacific and ENSO. Monthly Weather Review 122: 636 651.
125 Landsea, CW. 2000. El NioSouthern Oscillation and the seasonal predictability of tropical cyclones. El Nio: Impacts of Multiscale Variability on Natural Ecosystems and Soci ety Eds. Diaz HF, Markgraf V. Cambridge University Press: Cambridge, UK. Landsea CW, Bell GD, Gray WM, Goldenberg SB. 1998. The Extremely Active 1995 Atlantic Hurricane Season: Environmental Conditions and Verification of Seasonal Forecasts. Monthly Weather Review 126 : 11741193. Larkin NK, Harrison DE. 2002. ENSO Warm (El Nio) and Cold (La Nia) Event Life Cycles: Ocean Surface Anomaly Patterns, Their Symmetries, Asymmetries, and Implications. Journal of Climate 15: 11181140. Lattin JM, Carrol l JD, Green PE. 2003. Analyzing Multivariate Data Brooks/Cole: Pacific Grove, CA; 556. Lau N C, Nath MJ. 2003. AtmosphereOcean Variations in the IndoPacific Sector during ENSO Episodes. Journal of Climate 16 : 3 20. Leroy A, Wheeler MC. 2008. Statistical Prediction of Weekly Tropical Cyclone Activity in the Southern Hemisphere. Monthly Weather Review 136 : 36373654. Liebmann B, Smith CA. 1996. Description of a Complete (Interpolated) Outgoing Longwave Radiation Dataset. Bulletin of the Ameri can Meteorological Society 77 : 12751277. Lindesay JA, Harrison MSJ, Haffner MP. 1986. The Southern Oscillation and South African rainfall. South African Journal of Science 82 : 196198. Love G. 1985. Cross Equatorial Interactions during Tropical Cyclogenesis. Monthly Weather Review 113 : 14991509. Lyons SW. 1991. Origins of Convective Variability over Equatorial Southern Africa during Austral Summer. Journal of Climate 4 : 23 39. Manhique AJ, Reason CJ C, Rydberg L, Fauchereau N. 2009. ENSO and Indian Ocean sea surface temperatures and their relationships with tropical temperate troughs over Mozambique and the Southwest Indian Ocean. International Journal of Climatology in press. Mason SJ, Jury MR. 1997. Climatic variability and change over southern Africa: a reflection on underlying processes. Progress in Physical Geography 21: 23 50. Mavume AF, Rydberg L, Rouault M, Lutjeharms JRE. 2009. Climatology and Landfall of Tropical Cyclones in the SouthWest Indian Ocean. Western Indian Ocean Journal of Marine Science 8 : 15 36.
126 McHugh MJ, Rogers JC. 2001. North Atlantic Oscillation Influence on Precipitation Variability around the Southeast African Convergence Zone. Journal of Climate 14: 36313642. Meehl GA. 1987. The Annual Cycle and Int erannual Variability in the Tropical Pacific and Indian Ocean Regions. Monthly Weather Review 115 : 27 50. Merrill, RT. 1988. Environmental Influences on Hurricane Intensification. Journal of the Atmospheric Sciences 45: 16781687. Miyasaka T, Nakamura H. 2010. Structure and Mechanisms of the Southern Hemisphere Summertime Subtropical Anticyclones. Journal of Climate, submitted for publication. Murnane RJ. 2004. Introduction. Hurricanes and Typhoons: Past, Present, and Future. Eds. Murnane RJ, Liu K B. Columbia University Press: New York Naeraa M, Jury MR. 1998. Tropical Cyclone Composite Structure and Impacts over Eastern Madagascar During January March 1994. Meteorology and Atmospheric Physics 65 : 43 53. Namias J. 1976. Some Statistical and Synoptic Characteristics Associated with El Nio. Journal of Physical Oceanography 6 : 130 138. Neumann CJ, Randrianarison EA. 1976. Statistical Prediction of Tropical Cyclone Motion over the Southwest Indian Ocean. M onthly Weather Review 104: 76 85. Nicholson SE. 1997. An Analysis of the ENSO Signal in the Tropical Atlantic and Western Indian Oceans. International Journal of Climatology 17 : 345375. Nicholson SE. 2003. Comments on The South Indian Convergence Zone and Interannual Rainfall Variability over Southern Africa and the Question of ENSOs Influence on Southern Africa. Journal of Climate 16: 555 562. Oort AH, Yienger JJ. 1996. Observed Interannual Variability in the Hadley Circulation and Its Connection to ENSO. Journal of Climate 9 : 2751 2767. Pan YH, Oort AH. 1983. Global Climate Variations Connected with Sea Surface Temperature Anomalies in the Eastern Equatorial Pacific Ocean for the 195873 Period. Monthly Weather Review 111 : 12441258. Parker BA, Jury MR. 1999. Synoptic environment of composite tropical cyclones in the SouthWest Indian Ocean. South African Journal of Marine Science 21: 99 115.
127 Parker D. 1999. Criteria for Evaluating the Condition of a Tropical Cyclone Warning System. Disast ers 23: 193216. Pearson M. 2003. The Indian Ocean: Seas in History Series Ed. Scammell G. Routledge: London. Philander SGH. 1985. El Nio and La Nia. Journal of the Atmospheric Sciences 42: 26522662. Pielke RA Jr, Pielke RA Sr. 1997. Hurricanes: Their Nature and Impacts on Society John Wiley: New York; 279. Pohl B, Fauchereau N, Richard Y, Rouault M, Reason CJC. 2009. Interactions between synoptic, intraseasonal and interannual convective variability over Southern Africa. Climate Dynamics 3 3 : 10331050. Reason CJC. 2001. Subtropical Indian Ocean SST dipole events and southern African rainfall. Geophysical Research Letters 28 : 2225 2227. Reason CJC. 2002. Sensitivity of the Southern African Circulation to Dipole SeaSurface Temperature Patterns in the South Indian Ocean. International Journal of Climatology 22 : 377393. Reason CJC. 2007. Tropical cyclone Dera, the unusual 2000/01 tropical cyclone season in the southwest Indian Ocean and associated rainfall anomalies over Southern Africa. Meteorology and Atmospheric Physics 97: 181 188. Reason CJC, Allan RJ, Lindesay JA, Ansell TJ. 2000. ENSO and Climatic Signals Across the Indian Ocean Basin in the Global Context: Part I, Interannual Composite Patterns. International Journal of Climatology 20 : 12851327. Reason CJC, Keibel A. 2004. Tropical Cyclone Eline and Its Unusual Penetration and Impacts over the Southern African Mainland. Weather and Forecasting 19 : 789 805. Reynolds RW, Rayner NA, Smith TM, St okes DC, Wang W. 2002. An Improved In Situ and Satellite SST Analysis for Climate. Journal of Climate 15: 16091625. Roberts SE, Marlow PB. 2002. Casualties in dry bulk shipping (19631996). Marine Policy 26: 437 450. Romesburg HC. 1984. Cluster Analys is for Researchers Wadsworth: North Carolina; 334.
128 Roux F, ChaneMing F, LasserreBigorry A, Nuissier O. 2004. Structure and Evolution of Intense Tropical Cyclone Dina near La Runion on 22 January 2002: GB EVTD Analysis of Single Doppler Radar Obser vations. Journal of Atmospheric and Oceanic Technology 21 : 15011518. Saji NH, Goswami BN, Vinayachandran PN, Yamagata T. 1999. A dipole mode in the tropical Indian Ocean. Nature 401 : 360 363. Sampson CR, Goerss JS, Schrader AJ. 2005. A concensus track forecast for southern hemisphere tropical cyclones. Australian Meteorological Magazine 54: 115119. Shanko D, Camberlin P. 1998. The Effects of the Southwest Indian Ocean Tropical Cyclones on E thiopian Drought. International Journal of Climatology 18: 13731388. Silva JA, Eriksen S, Ombe ZA. 2010. Double exposure in Mozambiques Limpopo River Basin. The Geographical Journal in press. Streten NA. 1973. Some Characteristics of SatelliteObser ved Bands of Persistent Cloudiness Over the Southern Hemisphere. Monthly Weather Review 101 : 486 495. Suzuki R, Behera SK, Iizuka S, Yamagata T. 2004. Indian Ocean subtropical dipole simulated using a coupled general circulation model. Journal of Geophysical Research 109 : C09001. Todd M, Washington R. 1998. Extreme Daily Rainfall in Southern African and Southwest Indian Ocean Tropical Temperate Links. South African Journal of Science 94: 64 70. Todd M, Washington R. 1999. Circulation anomalies associated with tropical temperate troughs in southern Africa and the south west Indian Ocean. Climate Dynamics 15: 937951. Todd MC, Washington R, Palmer PI. 2004. Water Vapour Transport Associated with Tropical Temperate Trough Systems ov er Southern Africa and the Southwest Indian Ocean. International Journal of Climatology 24 : 555 568. Trenberth KE. 1976. Spatial and temporal variations of the Southern Oscillation. Quarterly Journal of the Royal Meteorological Society 102 : 639 653. Trenberth KE. 1990. Recent Observed Interdecadal Climate Changes in the Northern Hemisphere. Bulletin of the American Meteorological Society 71: 988 993.
129 Trenberth KE. 1991. General Characteristics of El NioSouthern Oscillation. Teleconnections linkin g worldwide climate anomalies Eds. Glantz MH, Katz RW, Nicholls N. Cambridge University Press: Cambridge, UK; 535. Trenberth KE. 1997. The Definition of El Nio. Bulletin of the American Meteorological Society 78: 27712777. Trenberth KE, Caron JM. 2000. The Southern Oscillation Revisited: Sea Level Pressures, Surface Temperatures, and Precipitation. Journal of Climate 13: 43584365. Trenberth KE, Stepaniak DP. 2001. Indices of El Nio Evolution. Journal of Climate 14: 16971701. Trigo IF, Davies TD, Bigg GR. 1999. Objective Climatology of Cyclones in the Mediterranean Region. Journal of Climate 12: 16851696. Tyson PD, PrestonWhyte RA. 2000. The Weather and Climate of Southern Africa. Oxford University Press: Oxford, UK; 396. Uppala SM, Kallberg PW, Simmons AJ, Andrae U, Da Costa Bechtold V, Fiorino M, Gibson JK, Haseler J, Hernandez A, Kelly GA, Li X, Onogi K, Saarinen S, Sokka N, Allan RP, Andersson E, Arpe K, Balmaseda MA, Beljaars ACM, Van De Berg L, Bidlot J, Bormann N, Caires S, Chevallier F, Dethof A, Dragosavac M, Fisher M, Fuentes M, Hagemann S, Holm E, Hoskins BJ, Isaksen L, Janssen PAEM, Jenne R, Mcnally AP, Mahfouf JF, Morcette JJ, Rayner NA, Saunders RW, Simon P, Sterl A, Trenberth KE, Untch A, Vasiljevic D, V iterbo P, Woollen J. 2006. The ERA 40 re analysis. The Quarterly Journal of the Royal Meteorological Society 131 : 29613012. v an Loon H, Madden RA. 1981. The Southern Oscillation. Part I: Global Associations with Pressure and Temperature in Northern Wi nter. Monthly Weather Review 109: 11501162. van Loon H, Rogers JC. 1981. The Southern Oscillation. Part II: Associations with Changes in the Middle Troposphere in the Northern Winter. Monthly Weather Review 109 : 11631168. Visher SS. 1922. Tropical Cyclones in Australia and the South Pacific and Indian Oceans. Monthly Weather Review 50 : 288295. Vitart F, Anderson D, Stockdale T. 2003. Seasonal Forecasting of Tropical Cyclone Landfall over Mozambique. Journal of Climate 16 : 39323945. Wa ng Y, Holland GJ. 1996. Tropical Cyclone Motion and Evolution in Vertical Shear. Journal of the Atmospheric Sciences 53 : 3313 3332.
130 Ward RDC. 1902. Current Notes on Meteorology. Science 15 :435 437. Washington R, Todd M. 1999. Tropical Temperate Links in Southern African and Southwest Indian Ocean SatelliteDerived Daily Rainfall. International Journal of Climatology 19 : 16011616. Washington R, Preston A. 2006. Extreme wet years over southern Africa: Role of Indian Ocean se a surface temperatures. Journal of Geophysical Research 111 : D15104. Waylen PR, Henworth S. 1996. A Note on the Timing of Precipitation Variability in Zimbabwe as Related to the Southern Oscillation. International Journal of Climatology 16 : 11371148. Webster PJ, Moore AM, Loschnigg JP, Leben RR. 1999. Coupled oceanatmosphere dynamics in the Indian Ocean during 199798. Nature 401 : 356 360. Wilks DS. 2006. Statistical Methods in the Atmospheric Sciences Elsevier: Boston; 627. Wolter K. 1987. The S outhern Oscillation in Surface Circulation and Climate over the Tropical Atlantic, Eastern Pacific, and Indian Oceans as Captured by Cluster Analysis. Journal of Applied Meteorology 26 : 540558. Wyrtki K. 1975. El Nio The Dynamic Response of the Equatorial Pacific Ocean to Atmospheric Forcing. Journal of Physical Oceanography 5 : 572 584. Yip CL, Wong KY, Li PW. 2006. Data Complexity in Tropical Cyclone Positioning and Classification. Data Complexity in Pattern Recognition. Eds. Basu M, Ho TK. Springer Verlag: Heidelberg. Yoo S H, Yang S, Ho C H. 2006. Variability of the Indian Ocean sea surface temperature and its impacts on AsianAustralian monsoon climate. Journal of Geophysical Research 111 : D03108. Xie S P, Annamalai H, Schott FA, McCreary, JP Jr. 2002. Structure and Mechanisms of South Indian Ocean Climate Variability. Journal of Climate 15: 864 878. Zinke J, Dullo W C, Heiss GA, Eisenhauer A. 2004. ENSO and Indian Ocean subtropical dipole variability is recorded in a coral record off southwest Madagascar for the period 1659 to 1995. Earth and Planetary Science Letters 228: 177 194.
131 BIOGRAPHICAL SKETCH Kevin Ash was born in Tulsa, Oklahoma in 1978 and lived there until the age of ten when his family moved to Oklahoma City. He graduated from Westmoore High School in 1997 and later finished a Bachelor of Arts in g eography at the University of Oklahoma in May 2004. Kevin then worked for Weathernews, Inc. in Norman, OK for nearly four years, hired originally for his geography background. After two years at Weathernews, he moved into a position in which he provided weather forecasts and route recommendations to commercial cargo vessels. Though he quite enjoyed the work, Kevin decided to leave and pursue a Masters degree in the fall of 2008 after being accepted into the Department of Geography at the University of Florida. After graduating in May 2010, he plans to pursue a PhD i n g eography.