Abstract the Indian Ocean Dipole's Influence On

ABSTRACT

THE INDIAN OCEAN DIPOLE’S INFLUENCE ON ATLANTIC TROPICAL CYCLONE ACTIVITY

Alan J. Marinaro, MS Department of Geography Northern Illinois University, 2015 Dr. David Changnon, Director

Improving early tropical cyclone forecasts would assist reinsurance decision makers as they seek information that can minimize risks. Early lead forecasts are based on model variables before December 1 (Year 0) that predict Atlantic tropical cyclone activity (Year +1). The autumn

Indian Ocean Dipole (IOD) has an 8 to 14 month antecedent correlation with the El Niño -

Southern Oscillation (ENSO). ENSO is traditionally the best non-lead and overall predictor of

Atlantic tropical cyclone activity. Analyses were performed over a 30-year period from 1984/85-

2013/14, with some time variation depending on the test. Correlation, spatial, and wavelet analyses were utilized to find associations between the IOD, west and east components of the IOD, and four other variables related to the following season’s ENSO state and tropical cyclone activity.

The prior western pole of the October IOD (WIOD) was demonstrated to have statistically significant r-squared values (i.e. 99% confidence interval) to upcoming tropical storm activity (i.e. explained 25% of the variance), named storm counts (28%), and ENSO (21%). The WIOD has no connection with U.S. hurricane landfalls. Wavelet analysis between October IOD variables and following August-October ENSO data was observed to have the best time-frequency relationship.

Dynamic reasoning for these relationships reside within the idealized biennial IOD-ENSO cycle,

Walker circulation process, and the impact of ENSO on the state of the Atlantic Basin. The

WIOD’s integration into early-lead forecast models could be an advantage for those in the reinsurance industry and other decision makers impacted by Atlantic tropical cyclones.

NORTHERN ILLINOIS UNIVERSITY DEKALB, ILLINOIS

AUGUST 2015

THE INDIAN OCEAN DIPOLE’S INFLUENCE ON

ATLANTIC TROPICAL CYCLONE ACTIVITY

A THESIS SUBMITTED TO THE GRADUATE SCHOOL

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE

MASTER OF SCIENCE

DEPARTMENT OF GEOGRAPHY

Thesis Director: Dr. David Changnon

ACKNOWLEDGEMENTS

This study would not have been possible without many of my colleagues and friends. I would like to thank the Department of Geography staff and faculty members for supporting me during the last two years. I am very appreciative of my thesis advisor, Dr. David Changnon, who helped me become a better person, student, scientist, and professional. I am thankful of Dr. Isaac

Hankes for sharing his expertise in tropical meteorology. Dr. Jie Song and Dr. Walker Ashley were instrumental in motivating me to be more active within the science, being honest in evaluating my research, and making suggestions in order to improve my writing.

TABLE OF CONTENTS

Page

LIST OF TABLES ...... vi

LIST OF FIGURES ...... vii

Chapter

1. INTRODUCTION ...... 1

2. BACKGROUND ...... 6

Atlantic Basin Tropical Cyclone Information...... 6

Definitions and Metrics of Atlantic Tropical Cyclone Activity ...... 12

Oceanic Equatorial Modes of Variability ...... 13

The El Niño - Southern Oscillation ...... 14

The Indian Ocean Dipole ...... 16

ENSO and Indian Ocean Dipole Relations ...... 18

Dynamics of the Indo-Pacific Circulation and Cycle ...... 20

Indo-Pacific Relationship to Atlantic Tropical Cyclone Activity ... 24

Atlantic Tropical Cyclone Related Climatic Modes of Variability ...... 25

The North Atlantic Oscillation...... 26

The Atlantic Multidecadal Oscillation ...... 27

Chapter Page

The Quasi-Biennial Oscillation...... 28

Seasonal Tropical Cyclone Prediction ...... 29

Colorado State University Model ...... 29

Tropical Storm Risk Model ...... 31

Climate Prediction Center Forecast ...... 32

3. DATA ...... 34

Atlantic Basin Tropical Cyclone Metrics ...... 34

Teleconnection Data ...... 34

CSU Autumn and IOD Variables ...... 35

Limitations and Advantages ...... 35

4. METHODS ...... 37

Research Objectives ...... 37

Predictor Variable Discussion...... 37

Canonical Correlation Analysis ...... 38

Time-Frequency Link between ENSO and the ACE index ...... 38

Time-Frequency Link between Autumn Variables and the ACE index ...... 40

Time-Frequency Link between IOD Variables and ENSO ...... 40

Cross-Correlation between Seasonal Tropical Cyclone Counts and the IOD ..... 41

Composite Cumulative ACE Track Maps and Statistics ...... 41

Chapter Page

Spatial Correlation of the WIOD to Tropical Atlantic Basin Climate Variables. 42

5. RESULTS ...... 43

Cross Correlations and Running Correlation (1984/85-2013/14) ...... 43

Time-Frequency Analysis (1984/85-2013/14) ...... 46

Correlation of Named Storms and U.S. Hurricane Landfalls to IOD Variables .. 51

Composite Cumulative ACE Track Analysis (1984-2012) ...... 52

Linking the WIOD to Ocean-Atmosphere Variables ...... 56

6. SUMMARY AND CONCLUSION ...... 61

REFERENCES ...... 65

APPENDIX ...... 75

LIST OF TABLES

Table Page

1. Current Saffir-Simpson scale in use by the NHC ...... 8

2. December variables for CSU model ...... 30

3. April variables for CSU model ...... 30

4. June variables for CSU model ...... 31

5. December variables for TSR model ...... 32

6. Cross-correlations between the autumn-IOD variables and ensuing August-October ENSO state and seasonal ACE index. The yellow shaded boxes of correlation values are significant at 95% confidence. The red value has the best relationship ..... 45

7. Cross-correlations between preceding IOD variables and ensuing tropical season named storm counts. The yellow shaded boxes of correlation values are significant at a 95% confidence interval and red text denotes best correlation ...... 51

8. Cumulative ACE summations and tropical cyclone counts for the maps in Figures 19, 20, and 21 relating to the three October IOD variables ...... 55

9. Abbreviations and acronyms ...... 76

LIST OF FIGURES

Figure Page

1. Time series of August forecasts and observations of ACE. The solid horizontal lines represent the mean of the 2001-2013 observations (125 for ACE) ...... 3

2. Observed number of basin hurricanes (A) and ACE (B) versus insurance losses. The corresponding years are shown as well (e.g. number "05" on the plot indicates the 2005 hurricane season). Vertical dash lines indicate the 2001-2013 average of the number of hurricanes (A) and ACE (B) ...... 4

3. Main development region (MDR) defined by box ...... 6

4. Monthly tropical cyclone track and origin climatology from June to November...... 10

5. Niño region geography in the equatorial Pacific ...... 15

6. El Niño and La Niña episode tropical Pacific sea-surface temperature and wind profiles...... 15

7. Winter and summer monsoon changes and associated precipitation and temperature anomalies...... 16

8. West and east locations of DMI ...... 17

9. The Biennial IOD-ENSO Cycle denoted by seasonal perturbations of monsoons, IOD events, Pacific trade winds, and ENSO base state ...... 21

10. Three steps: How a negative IOD event during the previous fall can trigger a springtime El Niño event ...... 22

11. The Walker circulation’s link to ENSO and Atlantic tropical cyclone activity ...... 25

12. Phase representation of the NAO over the North Atlantic ...... 26

13. AMO phase (-) and AMO phase (+) hurricane landfall climatology off the U.S. East Coast...... 28

viii

Figure Page

14. Niño region 3 wavelet power spectrum. (a) Data in time series format. (b) The wavelet power spectrum of those data. The contours are shown at 25% (blue), 50% (green), 75% (light green), and 95% (red) of the wavelet power scale; each percentage represents a confidence interval. (c) The global wavelet power spectrum ...... 39

15. Running correlations between the WIOD/EIOD versus the ACE index the following year. The red correlation values are significant at a 95% confidence interval ...... 44

16. Similar wavelet power spectrums of seasonal (a.) Atlantic ACE index and (b.) ASO Oceanic Niño Index indicate a time-frequency relationship between the ENSO state and Atlantic tropical cyclone activity. The ellipses (1 and 2) and the arrow denote areas of analogous periodicity...... 47

17. Wavelet power spectrums of (a.) October-November Atlantic Multidecadal Oscillation, (b.) November Eastern Equatorial Sea-Level Pressure, (c.) November North Atlantic 500mb Geopotential Heights, and (d.) ACE Index for following season ...... 48

18. Wavelet power spectrums of (a.) October WIOD, (b.) October Dipole Mode Index (DMI), (c.) October EIOD, and (d.) August-October Oceanic Niño Index for following season. The ellipses (1 and 2) and boxes (3 and 4) denote areas of analogous periodicity ...... 50

19. Maps of the (a) five highest and (b) five lowest years representing preceding DMI values corresponding to ensuing tropical season’s cumulative composite ACE. Scale on right side represents cumulated ACE (104 푘푡2) ...... 53

20. Maps of the (a) five highest and (b) five lowest years representing preceding EIOD values corresponding to ensuing tropical season’s cumulative composite ACE. Scale on right side represents cumulated ACE (104 푘푡2) ...... 53

21. Maps of the (a) five highest and (b) five lowest years representing preceding WIOD values corresponding to ensuing tropical season’s cumulative composite ACE. Scale on right side represents cumulated ACE (104 푘푡2) ...... 54

22. Spatial correlation map between Atlantic basin sea-level pressure and the WIOD. Areas of deep violet denote areas with relevant correlations ...... 57

Figure Page

23. Spatial correlation map between Atlantic basin 700mb relative humidity and the WIOD. Arrows denote areas of 700mb relative humidity with relevant correlations ...... 58

24. Spatial correlation map between tropical Pacific sea-surface temperatures and the WIOD. Blue and violet shaded areas (cold temperatures) denote significant inverse correlation linked to ENSO ...... 60

25. The research links between the WIOD, ENSO, and Seasonal Atlantic ACE index. Blue arrows are known research links. The red arrow is a new connection within the research ...... 63 1

CHAPTER 1

INTRODUCTION

The Atlantic Ocean tropical cyclone season is defined from June 1 to November 30

(AOML 2015a). Hurricanes create large human, socio-economic, and financial costs in the

United States, Caribbean, and Latin America (Pielke Jr et al., 2003; Blake et al., 2011).

Typically, developed countries (i.e. United States) endure the greatest losses due to tropical cyclones (Blake and Gibney 2011). Hurricane Mitch (1998), for example, killed 10,000 people in Central America and caused seven to ten billion dollars (i.e. 2014 inflation adjusted) in damages (Easterling et al., 2000). The Dominican Hurricane (1930) killed approximately forty- seven hundred people and caused ~700 million dollars (i.e. 2014 inflation adjusted) of damage

(Hartwell 1930; Pielke Jr et al., 2003). In the United States, property damages related to strong tropical cyclones have risen dramatically along with the human toll due to the coastal flooding, heavy rainfall, and storm surges.

Rising costs have occurred as a consequence of dense coastal infrastructures and population exposure in the U.S., Central America, and Caribbean countries (Pielke Jr and

Landsea 1998; Boruff and Cutter 2007). The United States has endured seven storms that created losses of ten billion dollars or more since Hurricane Andrew in 1992 (Blake et al., 2011).

Hurricane Katrina (2005) was the first storm to create over 100 billion dollars in damages and recorded the most deaths (1200 people) in the U.S. since 1928 (Blake et al., 2011). Modern-era

2 tropical cyclones can have negative effects on off-shore oil production, shipping, tourism, and damage coastal infrastructures when traversing ocean basins (Pielke Jr and Landsea 1998).

Hurricanes that make landfall decimate lives, destroy communities, and create an aftermath environment vulnerable to crime, disease, and social unrest (Cutter et al., 2006). The toll on the societies, economies, and governments of the United States, Mexico, and Caribbean countries is a significant problem that warrants the need for seasonal tropical cyclone predictions of basin wide tropical cyclone activity (Landsea et al., 1994; Camargo and Zebiak 2002;

Camargo et al., 2007).

The first person to attempt Atlantic basin seasonal prediction of tropical cyclones was

Dr. William Gray in 1984. He integrated teleconnection climatology with respect to the El Niño -

Southern Oscillation and the Quasi-Biennial Oscillation (QBO) with little to no lead time (Gray

1984). Since 1984, a ground-swell of research by academics, insurance companies, and governments has expanded the body of work and lead times (i.e. period prior to tropical storm season, June 1.) of such forecasts (Camargo et al., 2007).

The Atlantic Basin is difficult to forecast based on the increased variability of tropical cyclone activity compared to other ocean basins (Frank and Young 2007). Atlantic basin tropical cyclone forecasts have become more accurate overtime, but large forecast failures still exist including those during the Atlantic basin tropical cyclone seasons of 2004 (number of tropical cyclones above the climatological mean) and 2013 (number of tropical cyclones below the climatological mean). Annual forecast failures since 2001 are shown in Figure 1 where two metrics of seasonal tropical cyclone activity deviate from the actual observed values during

3 poorly forecasted years (i.e., 2004 and 2013). The tropical cyclone season of 2013 endured 39% of the normal climatological Accumulated Cyclone Energy index (ACE) along with zero instances of major hurricanes the entire tropical cyclone season (Blake 2014).

Figure 1. Time series of August forecasts and observations of ACE. The solid horizontal lines represent the mean of the 2001-2013 observations (125 for ACE). (from Fig. 1 in Barmpadimos et al., 2014)

Insurance and reinsurance companies traditionally use seasonal tropical cyclone forecast models to project economic losses due to damage and mortality prone insured properties, corporate interests, and people. Insurance companies incorporate predictions of the observed seasonal ACE index (Figure 2A) because it is the best proxy for monetary losses according to a statistically significant Kendall's Tau correlation coefficient of 0.58

(Barmpadimos et al., 2014). Kendall's Tau correlation coefficient, related to the observed seasonal total number of basin hurricanes compared to monetary losses (Figure 2B), is only 0.47

(Barmpadimos et al., 2014).

Figure 2. Observed number of basin hurricanes (A) and ACE (B) versus insurance losses. The corresponding years are shown as well (e.g. number "05" on the plot indicates the 2005 hurricane season). Vertical dash lines indicate the 2001-2013 average of the number of hurricanes (A) and ACE (B). (from Fig. 8 in Barmpadimos et al., 2014)

The 2013 tropical cyclone season postmortem raised a number of issues for forecasters, such as the need to assess the forecast variables that models utilize (Hamilton 2013). The 2013 seasonal forecast model failure led atmospheric scientists to examine ways to improve Atlantic tropical cyclone models. Recent literature about the Indian Ocean Dipole (IOD), a sea-surface temperature dipole around the Equatorial Indian Ocean, has demonstrated the ability to precede

ENSO events (e.g. El Niño and La Niña) with eight to fourteen months of lead (Loschnigg et al.,

2003; Izumo et al., 2010). ENSO is the number one mode of variability that correlates best with

Atlantic tropical cyclone activity (Camargo et al., 2007). Understanding the dynamic link

5 between the IOD and ENSO equatorial processes should theoretically lead to more valuable seasonal Atlantic tropical cyclone forecasts with earlier lead-times.

Reinsurance companies continue to express a need for earlier forecasts due to governmental regulatory boundaries and set pre-allocation of company resources. Late issued seasonal tropical cyclone forecasts are of little benefit to the reinsurance industry because regulation delays the enactment of new policy. (Elsner et al., 2006; Barmpadimos et al., 2014).

Although reinsurance companies would prefer that forecasts for the June 1 – November 30 tropical cyclone season be released by January 1; there is still added value to incorporate seasonal forecasts into decision making if issued no later than June 1 (Elsner et al., 2006). The seasonal forecasts issued later in the season (June 1 and after) have implications on some

"marginal" underwriting procedure outcomes where insurance decisions are undecided. The late-season forecast leads to a yes or no decision to approve or deny underwriting procedures only if the forecast is outside the current climatology (Barmpadimos et al., 2014). Earlier and more accurate forecasts could guide the private sector to make important decisions with respect to economic, agricultural, energy, and especially insurance sectors of the global market

(Barmpadimos et al., 2014). Reinsurance companies that have lost billions due to tropical cyclone disasters would benefit from models with robust lead times. From a cost-benefit analysis point-of-view, an improved earlier forecast would help reinsurance companies delegate the balance between the costs of mitigation or direct losses from tropical hazards (Kunreuther

2006). The most important reason to improve statistical tropical cyclone models is to save lives and prepare for natural disasters. Ultimately, much of the discussion related to the private sector should be secondary to the vulnerable human component which natural disasters exploit.

CHAPTER 2

BACKGROUND

Atlantic Basin Tropical Cyclone Information

The Atlantic Basin tropical cyclone season starts on June 1 and lasts until November 30 with the climatological peak in tropical cyclone activity on September 10. Tropical cyclones begin as tropical disturbances over the Gulf of Mexico, North Atlantic, or Caribbean Sea. The most robust tropical cyclones often begin as an African Easterly Wave or tropical wave, off the coast of western Africa, and mature as they move over the Cape Verde Islands. These storms then move west into the main development region (Figure 3), which is an area bounded by 10°N-

20°N and 60°W-20°W (Goldenberg et al., 2001).

Figure 3. Main development region (MDR) defined by box. (adapted from Goldenberg et al., 2001)

An African Easterly Wave (AEW) starts out as an inverted trough of low-pressure without a closed circulation. These westerly propagating waves can eventually develop into a tropical cyclone if cyclogenesis evolves primarily over the intertropical convergence zone

(Burpee 1972). An AEW's ability to avoid shear, dry air intrusions, and the Saharan Air Layer also increases its chances at becoming a tropical cyclone (Dunion and Velden 2004). AEWs become tropical cyclones around 60% of the time. Most major hurricanes originate off the coast of western Africa as tropical waves (Landsea 1993). All fully barotropic tropical entities are

Carnot engines that originate in the tropics where sea-surface temperatures are 27°C or higher

(AOML 2015a; AOML 2015b, Gray 1979).

A typical tropical cyclone features convergence at sea-level, upward vertical motion at mid-level, and divergence above the storm (AOML 2015a). Low sea-level pressure anomalies are favorable for the strengthening of tropical cyclones (Knaff 1997). The environment surrounding the storm must be moist because dry air inhibits development (Prospero and Lamb

2003; Dunion and Velden 2004; Wong and Dessler 2005). Warm sea-surface temperatures act as a vehicle for latent heat flux that can intensify tropical cyclones (DeMaria and Kaplan 1994,

Goldenberg et al., 2001). Elevated vertical wind over the Atlantic basin can weaken a storm by disorganizing convection (Gray 1984, Aiyyer and Thorncroft 2006). The general circulation at

500mb guides tropical cyclone tracks, specifically the Bermuda High and an extension of the high over the southeastern U.S. The extension of the Bermuda High increases the probability of landfall within the Gulf of Mexico and Caribbean (Kossin et al., 2010). A weaker and more easterly shifted Bermuda High is associated with storms that make landfall on the east coast of the U.S. or recurve and become extra-tropical over the Atlantic Ocean (Kossin et al., 2010).

Classification of tropical cyclones is dependent on the maximum sustained wind speeds of a tropical cyclone. A tropical cyclone can be classified as a tropical depression (≤ 17 m/s [38 mph]), tropical storm (18–32 m/s [39-73 mph]), or hurricane (33 m/s [74 mph] or higher)

(AOML 2015a). Hurricanes can be placed into five categories by wind, central pressure, storm surge height, and damage (Simpson and Saffir 1974). The updated Saffir-Simpson (Table 1) scale currently used by the National Hurricane Center (NHC). Wind thresholds per storm category have been modified slightly and central pressure has been dropped on the updated

Saffir-Simpson scale (NHC 2015a). In climatology, the reference to "Major Hurricane" is a hurricane classified by Category 3, 4, and 5 (Table 1). A tropical storm or hurricane center which traverses over a body of land is known as a landfall by definition.

Table 1

Current Saffir-Simpson scale in use by the NHC.

Separate months within the tropical cyclone season exhibit different spatial characteristics with respect to origination point, track trajectory, and storm frequency (Blake and

Gibney 2011; NHC 2015b). June and July observe very few tropical disturbances based in the

Gulf of Mexico, Caribbean Sea, and U.S. East Coast (Figure 4). Many meteorologists call these

"homebrew" storms (non-AEW) due to their close origin and track location. June only averages

0.6 tropical storms and 0.1 hurricanes without any of them being a major hurricane while July averages 1.1 tropical storms, 0.4 hurricanes, and 0.1 major hurricanes (Blake and Gibney 2011).

Little to any land falling storms occur in June and July. Starting in August and September, the

African Easterly Jet becomes an increasingly important atmospheric feature during the most active part of the season. August is spatially dictated by Caribbean Sea disturbances and AEWs

(Figure 4) which make it ~200% more active compared to July (Blake and Gibney 2011). August and September are the most robust of all the months where the majority of tropical cyclone activity is dictated by AEWs (Hopsch et al., 2007). September contains the highest number of ocean recurve tracks compared to the rest of the months (NHC 2015b). Even with ocean recurve storms, all Atlantic basins are very active overall during this period (Figure 4). August averages

3.3 tropical storms, 1.6 hurricanes, and 0.7 major hurricanes (Blake and Gibney 2011). August is the second most active month during the tropical cyclone season (Blake and Gibney 2011).

September, the most active month, averages 4.0 tropical storms, 2.6 hurricanes, and 1.3 major hurricanes (Blake and Gibney 2011). One-third (i.e. the highest ratio of all the months) of all tropical storms become major hurricanes. August landfall probabilities increase from July where early season Gulf of Mexico and U.S. East Coast landfalls are possible. September is the most active landfall month where all basins in the West Atlantic are under threat (Figure 4).

Figure 4. Monthly tropical cyclone track and origin climatology from June to November. (from NHC 2015b)

A sharp delineation exists between the frequencies of AEWs from the beginning to the end of October, as the African Easterly Jet moves south and weakens approaching boreal winter

(Burpee 1972). October ends with Caribbean Sea "homebrew" activity where Florida, the northeast U.S. coast, and Nova Scotia, Canada become the climatologically best areas for landfalls (Figure 4). The third most active tropical cyclone season month is October, which averages 2.0 tropical storms, 1.1 hurricanes, and 0.4 major hurricanes (Blake and Gibney 2011).

November averages 0.7 tropical storms, 0.5 hurricanes, and 0.13 major hurricanes (Blake and

Gibney 2011) (Figure 4).

The average annual number of storm types from 1981 to 2010 for the Atlantic Basin include: 12.1 tropical storms, 6.4 hurricanes, and 2.7 major hurricanes. The annual number of tropical storms, hurricanes, and major hurricanes counts for the Basin has increased from 1851 to

2010 (Vecchi and Knutson 2008; Blake and Gibney 2011). Tropical cyclone data has been adjusted for missing storms, modern observation techniques, aircraft reconnaissance, and the remote sensing era. The number of tropical cyclones were measured at a rate of +1.6 storms/century from 1878 to 2006 (Vecchi and Knutson 2008). Many studies argue that the increasing linear trend is nothing more than a function of natural variability on a decadal scale

(Pielke Jr et al., 2005; Nyberg et al., 2007; Klotzbach and Gray 2008) while others believe it could be a signal related to global climate change (Emanuel 2005; Trenberth and Shea 2006;

Holland and Webster 2007).

Trends in major hurricanes have increased overtime from 1981 to 2006 according to

(Elsner et al., 2008; Vecchi and Knutson 2008) while others report a decrease from 1944 to 1991

(Landsea 1993) and 1896-1995 (Bove et al., 1998), respectively. Bove et al. (1998) also report a

12 downturn in landfalls for both hurricanes and major hurricanes from 1896-1995 as did Solow and

Moore (2002) from 1900-1998. These studies demonstrate temporal variability with respect to the frequency and landfalling of tropical cyclones. Teleconnections (modes of atmospheric or oceanic variability) are linked to the frequency of tropical cyclones with a specific strength, landfall area, storm track, and origin of a tropical cyclone (Gray 1984; Shapiro 1989; Elsner et al., 2000; Kossin et al., 2010; Colbert and Soden 2012).

Definitions and Metrics of Atlantic Tropical Cyclone Activity

The Accumulated Cyclonic Energy (ACE) index is a metric that takes into account the duration, intensity, and frequency of tropical cyclones that reach tropical storm strength or higher

(Bell et al., 2000). The ACE is a measure of tropical cyclone activity over any period or any ocean basin but can act as a measure for any sole cyclone. Calculating the ACE index is done by summating the wind speed squared over time (the duration of the tropical cyclone season) during

6-hour intervals (Bell et al., 2000) using the HURDAT database (Jarvinen et al., 1984). The

Power Dissipation Index (Emanuel 2005) formula is nearly the same as the ACE with the exception of cubing the wind speed instead of squaring it. Seasonal basin wide damage potential in the Atlantic is best correlated with the ACE index. The Net Tropical Cyclone Index (Gray et al., 1994) and Power Dissipation Index (Emanuel 2005) are comparable to the ACE index.

Gray et al. (1994) introduced the Net Tropical Cyclone (NTC) index as a measure of seasonal tropical cyclonic activity. Gray et al. (1994) defines the formula for seasonal NTC using six counts of different tropical cyclone metrics (i.e. named storms, hurricanes, major hurricanes, named storm days, hurricanes days, and major hurricane days) and the base period mean of these six counts over 30-year period. The percentages within the NTC formula are

13 determined by dividing seasonal counts of the six metrics over their base period means. If 19 named storms occurred in 2012 and the base period mean is 12.1 named storms, then the seasonal contribution of named storms is 157% (e.g. %NamedStorms). This process is repeated for each of the six metrics. The final NTC is calculated by summating the six percentages relating to each metric then dividing by six (e.g. NTC = [%NamedStorms + %Hurricanes +

%MajorHurricanes + %NamedStormDays + %HurricaneDays + %MajorHurricaneDays] ÷ 6).

Repetitious parameters within the formula could possibly exaggerate a particular storm's contribution to the NTC decreasing index accuracy. Although the NTC and ACE have completely different formulas, the two indices explain 92% of each other from 2001 to 2013

(Barmpadimos et al., 2014).

Oceanic Equatorial Modes of Variability

The American Meteorological Society defines a teleconnection as a link between changes in weather variables over separate regions of the Earth (AMS 2015). Teleconnections are cyclic with respect to direct (+) and inverse (-) correlation over different time scales where a connection occurs between two or more separate regional weather variables (AMS 2015). The time scale variability of teleconnections can vary from a matter of weeks to multiple decades.

Bjerknes (1969) discussed teleconnections based on sea-surface temperature patterns over the equatorial pacific that were related to the difference of sea-level pressure between Tahiti and Darwin, Australia (Walker 1924). The pressure differential was enough to modify the trade winds near the equatorial Pacific. Bjerknes (1969) noticed ocean warming events due to upwelling along the equatorial coast of South America spreading out into the central Pacific

Ocean associated with a weakening of trade winds in the winter of 1957-58. A similar process regarding trade wind reversals and sea-surface temperature gradients was noted by Fletcher

(1945) within the Indian Ocean; Bjerknes documented this event in the winter of 1963-64. The ocean-atmosphere coupling between the Indian Ocean and Pacific Ocean is driven by the Walker circulation over the equator (Bjerknes 1969).

The El Niño - Southern Oscillation

In the early 1980's, four sea-surface temperature domains (Figure 5) from 80°W to 160°E across the equator (from 5°N to 5°S) were discovered to be related to ENSO (Barnston et al.,

1997). These domains would later be known as Niño regions 1.2, 3, 3.4, and 4. Niño region 3.4 had the most significant relationship to the original "Walker" Southern Oscillation, therefore was used as a platform to categorize El Niño or La Niña events. An El Niño (La Niña) event (i.e.

ENSO event) is defined by the National Oceanic and Atmospheric Administration as five contiguous three month running-mean periods of Niño region 3.4 sea-surface temperature anomalies at or above 0.5°C (at or below -0.5°C) (CPC 2015b). La Niña and El Niño equatorial

Pacific sea-surface temperature and wind profiles are provided in Figure 6. Periods in the ENSO record that do not meet the La Niña or El Niño criteria are simply referred to as Non-ENSO events. ENSO has the most inter-annual variability (i.e. explains 30% of the global climate system) out of all global teleconnections. Cycles between El Niño and La Niña events occur every two to seven years (Minobe 2000).

Figure 5. Niño region geography in the equatorial Pacific. (from CPC 2015a)

Figure 6. El Niño and La Niña episode tropical Pacific sea-surface temperature and wind profiles. (from JISAO 2015)

Global climate conditions related to La Niña and El Niño for June to August and

December to February periods vary greatly due to changes in circulation patterns and monsoons.

ENSO events are of most concern because they drive the global variability and distribution of temperature and precipitation patterns across the globe (Trenberth et al., 1998). Ropelewski and

Halpert (1987) and Halpert and Ropelewski (1992) classify spatiotemporal ENSO event climate anomalies with respect to precipitation and temperature (Figure 7).

Figure 7. Winter and summer monsoon changes and associated precipitation and temperature anomalies. (after NOAA 2015)

The Indian Ocean Dipole

Nicholls (1978) and Brier (1978) found a connection between the QBO and ocean- atmospheric coupling over the Indian Ocean. Interannual variability studies of Indian Ocean sea-

17 surface temperatures indicate a dipole is likely present over the Indian Ocean around the equator

(Nicholls 1989; Meyers 1996). The Indian Ocean Dipole (IOD) is a teleconnection based on an empirical orthogonal function analyses of basin wide sea-surface temperature and surface winds

(Saji et al., 1999; Webster et al., 1999). The primary principal component accounted for 30% of the variability of Indian Ocean sea-surface temperatures. The primary principal component was related to ENSO (Saji et al., 1999). The secondary principal component describes 12% of Indian

Ocean sea-surface temperatures and can be related to the dipole pattern. Ocean-atmosphere coupling and interannual variability are both present regarding the IOD (Saji et al., 1999;

Webster et al., 1999; Cai et al., 2013). The IOD is quantified by an index called the dipole mode index (DMI). The DMI is the difference between the sea-surface temperature regimes (bounds) residing approximately around the equatorial coast (Figure 8) of East Africa (i.e. 50°E-70°E,

10°S-10°N) and western coast of Sumatra (i.e. 90°E-110°E, 10°S-0°S) (Saji et al., 1999). A positive (negative) DMI event corresponds to above (below) normal sea-surface temperatures inside the western box and below (above) normal sea-surface temperatures inside the eastern box.

Figure 8. West and east locations of DMI. (from BOM 2015a)

Different DMI events feature a plethora of regional climate anomalies that impact India,

Indonesia East Africa, and Australia. Positive (negative) DMI events, similar to 1994 (1996), feature above (below) normal rainfall over East Africa, an active (inactive) Indian summer monsoon, and higher (lower) probabilities of drought over Indonesia and Australia (Behera et al.,

1999; Latif et al., 1999; Saji et al., 1999; Webster et al., 1999). Before 1999, the descriptive variability of Pacific (ENSO) and Atlantic Ocean sea-surface temperatures were thought to modulate Australian and East African rainfall production (Smith 1994; Fontaine and Janicott

1996). The reoccurring frequency of IOD events should remain the same over time, but more positive DMI events are likely to be endured than negative DMI events (Cai et al., 2013).

ENSO and Indian Ocean Dipole Relations

The IOD and ENSO have similar equatorial-ocean properties, subsurface thermoclines, pole related sea-surface temperature setups, and Walker circulation links (Vinayachandran et al.,

2002; Wu and Kirtman 2004; Webster and Hoyos 2010). Saji and Yamagata (2003) found IOD and ENSO events similarly occur every two to seven years by using wavelet analysis method on both the DMI and ENSO region 3.4 time series data. Many believe the properties of the IOD are similar to that of ENSO but the IOD is independent in relation to the interannual variability of

ENSO (Saji et al., 1999; Webster et al., 1999; Saji and Yamagata 2003; Izumo et al., 2010).

A statistical connection between ENSO and the IOD occurs in the boreal autumn where

Niño region 3.4 sea-surface temperatures correlate (inversely) significantly with the DMI (Saji and Yamagata 2003). Kug and Kang (2006) discovered the western pole of the DMI (mean

October-November sea-surface temperature anomaly leading the next year) negatively correlates with ENSO (mean Niño region 3.4 November-January sea-surface temperature anomaly).

Global climate models have linked the processes of the IOD to ENSO (Wu and Kirtman

2004). The influence between the IOD and ENSO relationship has created two groups of thought. The first group believes the IOD influences ENSO (Wu and Kirtman 2004; Terray and

Dominiak 2005; Kug and Kang 2006; Izumo et al., 2010), while the second group believes

ENSO describes the variability of the IOD (Latif et al., 1999; Okumura and Deser 2010; Wang and Wang 2014; Zhao and Nigam 2015). Webster and Hoyos (2010) relate ENSO and the IOD to a cycle in which both influence each other to some degree through dynamic processes.

Meyers et al. (2007) explains how the IOD and ENSO possess little to no dynamic linkage over the equatorial ocean. Zhao and Nigam (2015) applied an extended empirical orthogonal function analysis with the influence of ENSO removed. They concluded that the IOD was not a dipole, and the eastern pole is a manifestation of ENSO influence. Kug and Kang (2006) created a variant mean sea-surface temperature anomaly index around the same area (i.e. shifted 5 degrees to the west; 55°E-75°E, 10°S-10°N) as the western pole of the DMI. The new western index

(Kug and Kang 2006) was calculated for all November to January periods from 1958 to 2001.

The new western index was highly correlated (+0.62) with November to January periods of Niño region 3.4 temperature anomalies one year later (Kug and Kang 2006). A possible reason why the IOD may not be effective in predicting ENSO going into the late fall and winter is related to the decay of the east IOD pole’s sea-surface temperature pool (Izumo et al., 2010; Zhao and

Nigam 2015).

Dynamics of the Indo-Pacific Circulation and Cycle

Several studies, both observational (Terray and Dominiak 2005; Kug and Kang 2006) and modeled (Loschnigg and Webster 2000; Loschnigg et al., 2003; Wu and Kirtman 2004;

Izumo et al., 2010), were able to create a better understanding of the relationships between the tropical Indian and Pacific Oceans. The Indo-Pacific circulation is observed to be approximately a biennial cycle expounded on by Loschnigg et al. (2003) and depicted (Figure 9) by Webster and Hoyos (2010). Climate variability (i.e. ENSO and IOD events cycle every 2-7 years at different magnitudes) over the equatorial oceans and corresponding circulation patterns do not always follow the same temporal restraints in regards to idealized biennial IOD-ENSO cycle.

The following timeline describes the first year (e.g. a strong summer Indian Ocean monsoon, fall negative IOD, weak spring trade winds, and late spring into summer El Niño event) of the biennial cycle (Izumo et al., 2010).

Figure 9. The Biennial IOD-ENSO Cycle denoted by seasonal perturbations of monsoons, IOD events, Pacific trade winds, and ENSO base state. (from Webster and Hoyos 2010)

Beginning in the summer, the Indian Ocean monsoon is observed to be strong relative to normal followed by 850mb winds from west to east across the equatorial Indian Ocean (Izumo

2010; Webster and Hoyos 2010). A negative IOD event initializes due (Figure 10(a)) to the upwelling of warm water off the west coast of Sumatra (i.e. region of east IOD pole) by virtue of the monsoon. Conversely, the waters parallel to the east coast of tropical Africa are colder than average. Into the autumn, a low-pressure cell is seeded over the eastern pole of the IOD where upward vertical motion (i.e. convection) helps amplify Walker circulation (Wu and Kirtman

2004; Izumo et al., 2010). The eastern pole of the IOD normalizes with respect to sea-surface temperatures (Izumo et al., 2010).

Figure 10. Three steps: How a negative IOD event during the previous fall can trigger a springtime El Niño event. (adapted from Izumo et al., 2010)

The winter endures weakened equatorial Pacific Ocean trade winds because of the normalization of the eastern pole (Figure 10(b)) of the IOD (Izumo et al., 2010). The shift to

23 weaker trade winds allows for Kelvin waves associated with convection to transfer thermal energy eastward across the equatorial Pacific. Warm sea-surface temperatures upwell off the

Peruvian coastline after the underwater transit of warm water via the positively tilted west to east orientated Pacific Ocean thermocline (Izumo et al., 2010). The weak trade winds will eventually spread the warm water from the east section of the equatorial Pacific toward the central section

(Izumo et al., 2010). El Niño now establishes itself with a low-pressure cell over the equator near the warm pool at 160°E altering the Walker circulation (Loschnigg et al., 2003; Izumo et al.,

2010). Spring brings a chance of further development from a weak ENSO event to a mature one

(Figure 10(c)). Bjerknes (1969) feedback, a positive feedback between easterly wind stress and the sea-surface temperature anomaly differential across the equatorial Pacific, can enhance El

Niño. Although the El Niño may have strengthened, the heat content provided from around the thermocline is slowly starting to become depleted by the end of spring (Zhu et al., 2014).

The second year is temporally similar to the first year with respect to the dynamics but explained for the positive IOD and La Niña. The weak summer monsoon is forced by the decay of the El Niño event (Loschnigg et al., 2003; Izumo et al., 2010). A positive IOD event forms in the autumn because the weak monsoon allows cold (warm) water to propagate toward the eastern

(western) pole of the IOD (Loschnigg et al., 2003; Izumo et al., 2010). Trade winds, in the winter, increase as a response to the differential thermal gradient across the equatorial Indian

Ocean (Izumo et al., 2010). The positive IOD breaks down (i.e. the eastern pole), a low-pressure cell develops around 145°W, while high pressure dominates the eastern tropical Pacific Ocean.

Strong trade winds prevail from winter into spring (Loschnigg et al., 2003; Izumo et al., 2010).

The thermocline is strongly tilted positive allowing for cold water along and underneath it to

24 upwell into the Niño regions (Figure 5). La Niña conditions become present and the Walker circulation strengthens (Izumo et al., 2010).

From the late spring into the summer, the Walker circulation is quite robust during La

Niña events (Loschnigg et al., 2003; Izumo et al., 2010). It is strong enough that trade winds can drive cold sea-surface temperatures further west than expected. La Niña starts to fade because of western Pacific atmospheric cooling (An and Wang 2001). This cooling is due to sea-surface temperatures migrating too far west into the basin near the low-pressure cell at 145°W. A negative feedback mechanism causes the cutoff of upward vertical motion over the west Pacific; the depletion of heat and humidity (cool sea-surface temperatures) which provides instability for the west Pacific (An and Wang 2001). The sub-surface water temperatures eventually decay to the point where a non-ENSO event exists. The cycle is now complete.

Indo-Pacific Relationship to Atlantic Tropical Cyclone Activity

The IOD state in the previous autumn allows for the potential forward diagnosis of late summer and autumn Atlantic basin tropical cyclone season activity through ENSO prediction.

ENSO is connected to the Atlantic Ocean Basin because of the Walker circulation (Walker 1924;

Bjerknes 1969). El Niño (La Niña) events feature a low-pressure (high-pressure) cell that resides over the equatorial pacific coast of South America which is associated with vertical ascent

(descent). The relationship between ENSO and tropical cyclone activity (Figure 11) in the

Atlantic is related to an increase (decrease) in vertical wind shear brought on by enhanced

(depressed) upper-level westerlies from a weakened (strengthened) Walker Circulation due to El

Niño (La Niña) (Gray 1984; Shapiro 1987; Aiyyer and Thorncroft 2006; Klotzbach 2011). The

Walker circulation over the Atlantic creates subsidence (upward vertical motion) over the majority of the basin when an El Niño (a La Niña) event is present allowing the domain to be favorable for higher (lower) than normal sea-level pressure (Chu 2004). The probability of observing two or more Atlantic hurricane landfalls increases by ~136% during La Niña periods versus El Niño periods (Bove et al., 1998). Landsea et al. (1999) noticed the mean intensity of tropical cyclones during La Niña periods was slightly stronger than El Niño periods.

Figure 11. The Walker circulation’s link to ENSO and Atlantic tropical cyclone activity. (adapted from BOM 2015b)

Atlantic Tropical Cyclone Related Climatic Modes of Variability

The teleconnections discussed are composed of either atmospheric or oceanic variables and referenced as climate modes. The use of climate modes such as the North Atlantic

Oscillation (NAO), El Niño - Southern Oscillation (ENSO), Atlantic Multidecadal Oscillation

(AMO), and Quasi-Biennial Oscillation (QBO) are common place within tropical cyclone statistical modeling.

The North Atlantic Oscillation

The North Atlantic Oscillation is the difference of sea-level pressure between the Azores and Iceland (Walker 1924; Rogers 1984). The positive (negative) mode of the NAO is related to a stronger than normal low (high) pressure over Iceland, which allows for a more zonal

(meridional) jet-stream flow across the North Atlantic Ocean. A negative NAO is responsible for blocking over Greenland which generally implies jet-stream troughing over the eastern U.S. and

Western Europe whereby the opposite is true with a positive NAO. (Figure 12). The NAO is usually assessed on daily to monthly timescales.

Figure 12. Phase representation of the NAO over the North Atlantic.

The modern NAO index is calculated using empirical orthogonal function analysis transposed against 500mb height anomalies (Barnston and Livezy 1987). In Elsner et al. (2000), a bootstrap statistical modeling procedure was applied to the NAO value distribution curve with respect to major hurricane frequency in Gulf of Mexico (from TX to AL coast) and U.S. East

Coast (from NC to ME coast). NAO values around zero have the highest likelihood of a major hurricane tracking near or land falling on the Gulf of Mexico coast, while NAO values near positive one have the best probability of a major hurricane tracking around or hitting the U.S.

East Coast (Elsner et al., 2000).

The Atlantic Multidecadal Oscillation

The Atlantic Multidecadal Oscillation is an oceanic teleconnection related to the sea- surface temperatures in the Atlantic basin with a ~60 to 70 year cycle (Schlesinger and

Ramankutty 1994). The AMO is defined as the area weighted sea-surface temperature average bounded by 0°N-70°N and land masses (i.e. The Americas to the west, Europe and Africa to the east) surrounding the Atlantic Ocean (ESRL 2015a). During the AMO's positive phase, the sea- surface temperatures are higher relative to normal. The higher heat content powers the Carnot engine that is a tropical cyclone. Positive AMO periods see higher frequencies of all tropical cyclone types (i.e. major hurricane, hurricane, and tropical storm) while negative AMO periods

(Figure 13) inhibit tropical cyclone genesis (DeMaria and Kaplan 1994; Shapiro and Goldenberg

1998; Goldenberg et al., 2001; Elsner et al., 2006; Trenberth and Shea 2006).

Figure 13. AMO phase (-) and AMO phase (+) hurricane landfall climatology off the U.S. East Coast. (from Goldenberg et al., 2001)

The Quasi-Biennial Oscillation

The Quasi-Biennial Oscillation is a stratospheric teleconnection defined by the mean zonal equatorial wind around the 40mb level of the atmosphere with approximately a twenty- eight month cycle (Baldwin et al., 2001). The QBO can be classified by its positive (west) and negative (east) phase, where the direction represents a westerly or easterly wind regime at the equator (Angell 1986). The importance of the QBO becomes clearer as these upper-level equatorial waves propagate downward into the troposphere, modulating extra-tropical waves

(Baldwin et al., 2001). Currently, the Climate Prediction Center (CPC) uses the 30mb pressure level for the QBO calculation versus the defined 40mb pressure level; the change is negligible with regards to QBO values or cycle period. Tropical cyclone activity is amplified during the

QBO west phase and impeded in the east phase (Gray 1984; Shapiro 1989; Camargo and Sobel

2010). Atlantic Basin tropical cyclone activity, out of all the global basins, has the best overall

29 relationship with the QBO (Camargo and Sobel 2010). A La Niña (El Niño) like atmospheric state allows for a lack (surplus) of vertical wind shear associated with QBO west (east) periods which increase (decrease) tropical cyclone activity (Gray el al. 1992).

Seasonal Tropical Cyclone Prediction

The first modern manuscript to discuss seasonal tropical cyclone prediction was written by Nicholls (1979). Nicholls wanted to predict the Austral cyclone tropical cyclone season

(October to May). Mean winter (June to August) sea level pressure over Darwin (ENSO related) was used as an antecedent variable for activity during the early (October to December), middle

(January to February), late (March to May), and entire tropical cyclone season using the 1950 to

1975 period of record. The best relationship time periods came in the early season and during the entire season with -0.48 and -0.40 significant regressed correlation coefficients (p < 0.05), respectively. This was the first time large-scale circulation patterns (i.e. sea-level pressure over

Darwin) had been linked to tropical cyclone activity for a specific ocean basin.

Colorado State University Model

In 1984, Dr. William Gray, a Colorado State University atmospheric scientist, created the first Atlantic basin tropical cyclone seasonal forecast. The forecast was based on physical links between the climatic modes of ENSO and the QBO that correlate to tropical cyclone activity

(Gray 1984). Gray and eventually Klotzbach, would make these seasonal tropical cyclone forecasts publicly known as the CSU Model (Gray 1984; Gray et al., 2006).

CSU model variables are classified by location (latitudinal and longitudinal bounds), time period, and mean weather variable over the location. These CSU model variables are used to create December (Table 2), April (Table 3), and June (Table 4) issued forecasts (Klotzbach and Gray 2009, 2014a, and 2014b). The variables themselves are known to have robust to moderate correlations with the Net Tropical Cyclone index the following tropical cyclone season. The European Centre for Medium-Range Weather Forecasts (ECMWF) Niño region forecasts (Table 3, # 4; Table 4, #3) are used to gauge future ENSO conditions.

Table 2

December variables for CSU model. (adapted from Table 4 in Klotzbach and Gray 2009)

Predictor 1) October-November SST (55-65°N, 10-60°W) (+) 2) November 500mb geopotential height (67.5-85°N, 10-50°W) (+) 3) November sea-level pressure (7.5-22.5°N, 125-175°W) (+)

Table 3

April variables for CSU model. (adapted from Table 2 in Klotzbach and Gray 2014a)

Predictor 1) January-March Atlantic SST (5-35°N, 10-40°W) (+) 2) March sea-level pressure (20-40°N, 20-35°W) (-) 3) February-March sea-level pressure (5-20°S, 85-125°W) (+) 4) ECMWF 1 March SST Forecast for September Nino Region 3 (5°S-5°N, 90-150°W) (-)

Table 4

June variables for CSU model. (adapted from Table 2 in Klotzbach and Gray 2014b)

Predictor 1) April-May SST (15-55°N, 15-35°W) (+) 2) April-May 200mb u-wind (0-15°S, 150°E-120°W) (+) 3) ECMWF 1 May SST Forecast for September Nino Region 3 (5°S-5°N, 90-150°W) (-) 4) May sea-level pressure (20-40°N, 30-50°W) (-)

Overtime, the variables of the CSU model were modified when new physical links within the atmospheric circulation process were found. The statistical approach uses multilinear regression (Gray et al., 1992) between antecedent variables (climate modes or bounded boxed weather variables) and the predictands (metrics). Variables can change per each CSU model issuing period. Seasonal metrics include tropical (named) tropical storm days and counts, hurricane days and counts, major hurricane days and counts, NTC index, and ACE index information.

Tropical Storm Risk Model

Tropical Storm Risk (TSR) was a U.K. government funded venture in 1998 to further research tropical cyclone prediction. The project involved scientists, meteorologists, climatologists, and statisticians from private and public entities (TSR 2014). The basis for the

TSR tropical cyclone forecasts come from the research of Mark Saunders and Adam Lea

(Saunders and Lea 2005, 2008; Klotzbach et al., 2011). TSR has issued Atlantic tropical cyclone forecasts for every year since 2000 in the months of December to August (Camargo et al., 2007,

Klotzbach et al., 2011). The TSR Model is similar to that of the CSU Model, as it uses that same regression techniques and general predictands, except it also tries to quantify land-falling tropical cyclones (Klotzbach et al., 2011).

Unlike the Gray-Klotzbach models, the variables do not change throughout the TSR model issue period from month to month. The same TSR model variables (Table 5) are used for

December, April, and June issued forecasts (Saunders and Lea 2014). The forecasted July-

September TSR variables (Table 5) are a product of a regression against the highly correlated antecedent El Niño-Southern Oscillation-Climatology and Persistence (CLIPER) model. The

TSR model forecasts sea-surface temperatures for Niño region 3.4 and 925mb zonal winds

(Table 5) (Mack 2013). The July-September TSR variables are forecasted by proxy from

CLIPER model runs for the months corresponding to forecast issue dates.

Table 5

December variables for TSR model. (adapted from Saunders and Lea 2014)

Predictor 1) July-September 925mb u-wind (7.5-17.5°N, 30-100°W) 2) July-September SST (10-20°N, 20-60°W)

Climate Prediction Center Forecast

The Climate Prediction Center has released Atlantic tropical cyclone outlooks with cyclone predictions since 2002. Late May and early August are the only times in which these

33 outlooks are released to the public. The outlooks are probabilistic in nature, given by ranges of expected tropical storms, hurricanes, major hurricanes, and ACE index values expected for the upcoming Atlantic tropical cyclone season. The forecast is mostly based off the AMO, ENSO,

QBO, the inter-annual mode of tropical convection, and other teleconnection based predictors

(Klotzbach et al., 2011). The inter-annual mode is measured by 200mb Velocity Potential between 30°N and 30°S, which links upper-level divergent circulation with equatorial and subtropical convective fluctuations (Chelliah and Bell 2004; Bell and Chelliah 2006).

Essentially, CPC outlooks are based off human inferences of trends and teleconnection climatology related to tropical cyclone activity.

CHAPTER 3

DATA

Atlantic Basin Tropical Cyclone Metrics

The updated version of the monthly ACE index (1985-2014), used in the Maue (2011) dataset, is calculated in reference to the Bell et al. (2000) formula. Tropical cyclone data from the HURDAT database (Jarvinen et al., 1984) was applied to calculate the ACE index each season (Maue 2011). The forecast verification file (CSU 2015) at CSU contains calculated seasonal counts of named storms (tropical storms, hurricanes, and major hurricanes) and corresponding per diem metrics. The counts will be acquired for correlation analysis.

Teleconnection Data

The Oceanic Niño Index measures the trimonthly average of sea-surface temperature anomalies in Celsius for Niño region 3.4 (CPC 2015b). The ENSO metric will be employed to assess descriptive relationships between the AMO, IOD, ACE index, two December forecast

CSU variables (Table 2, #2 and #3), and tropical cyclone activity. Two time series are developed with Oceanic Niño Index data for the August to October (1985 to 2014) and

September to November (1984 to 2013) periods. The August to October index (Year +1) is used to gauge ENSO during the congruent tropical cyclone season within the Atlantic Ocean. The

September to November index (Year 0) is utilized as an ENSO related autumn variable to compare to the following tropical cyclone season. The AMO index (ESRL 2015a, Schlesinger

35 and Ramankutty 1994) is a measure of Atlantic sea-surface temperature anomalies over the

Atlantic Ocean. The AMO is compiled into time series values of the mean October-November

AMO index (Year 0) from 1984 to 2013 and employed as an autumn variable.

CSU Autumn and IOD Variables

NCEP Reanalysis monthly data (Kalnay et al., 1996) were downloaded from the ESRL

(2015b) web resource for the selected period, geographical area, and weather variable for variables #2 and #3 on Table 2. Autumn variable #2 is defined by mean November geopotential heights across the North Atlantic. Autumn variable #3 is composed of mean October-November sea-level pressure over the east tropical Pacific. The autumn variables are put into time series format from 1984 to 2013. The Indian Ocean Dipole (Saji et al., 1999; Webster et al., 1999) is represented by the western (WIOD) and eastern (EIOD) poles of the Dipole Mode Index.

Geographically bounded NCEP Reanalysis data values (Kalnay et al., 1996) for the EIOD and

WIOD (Figure 8) were gathered using the ESRL (2015b) web resource. Three time series (1984-

2013) are then created using October sea-surface temperature data for the WIOD, EIOD, and the differenced index known as the DMI (i.e. WIOD minus EIOD). The IOD relevant variables

(Year 0) are related to the upcoming Atlantic tropical cyclone season activity (Year +1) through

ENSO (Gray 1984; Izumo et al., 2010; Webster and Hoyos 2010; Klotzbach 2011).

Limitations and Advantages

NCEP Reanalysis data have been used in many refereed studies since Kalnay et al.

(1996) made them available to the public. The limitations include: available data coverage of the

Southern Hemisphere before the modern satellite era (1948 to ~1979); lack of precipitation data

36 over the tropical Pacific (CCSP SAP 1.3, 2008). The advantages include the ability to determine physical processes, climate variability, droughts, monsoons, and modes of variability such as

ENSO (CCSP SAP 1.3, 2008). The latter is important in this study because the IOD connects to the physical forcing of ENSO regarding favorable or unfavorable conditions for tropical cyclone development across the Atlantic Basin (Izumo et al., 2010). The ACE index is able to quantify basin wide tropical cyclone activity and it is preferred by reinsurance companies (Barmpadimos et al., 2014) over the NTC. The ACE index is not a measure of tropical cyclones that have reached land. Powell and Reinhold (2007) refer to the ACE index as a metric that fails to describe the spatial distribution of tropical storms, hurricanes, and major hurricanes. The ACE index metric is biased toward the inflationary component of long-duration tropical cyclones (Bell and Chelliah 2006). Despite this drawback, the ACE index is still the primary metric used by public and private institutions alike.

CHAPTER 4

METHODS

Research Objectives

The primary goal of this research is to determine whether the autumnal IOD state (Year

0) has an antecedent relationship to Atlantic seasonal tropical cyclone activity (Year +1).

Secondary objectives include demonstrating that the IOD variables are more robust than the variables used in previous models and forecasts, describing the connection between the prior autumnal IOD state and the subsequent Atlantic tropical season’s ENSO, and understanding the link between ENSO and seasonal tropical cyclone activity in the Atlantic basin. The primary goal will be achieved by logical procession of all the secondary objectives.

Predictor Variable Discussion

Two December variables (Table 2, #2 and #3) will be used within the data analysis.

December variable #1 (Table 2, #1) will be omitted because it shares sea-surface temperature domain with the AMO (Klotzbach and Gray 2009). December variable #2 relates the antecedent

October-November North Atlantic 500mb circulation with slower 200mb zonal winds speeds over the Atlantic Basin during the tropical cyclone season allowing for less upper-level shear

(Klotzbach and Gray 2009). December variable #3 links higher than normal November 500mb geopotential heights over the northeastern tropical Pacific to the development of La Niña conditions correlating to less shear and lower sea-level pressure in the Atlantic Basin (Klotzbach

38 and Gray 2009). The AMO and ENSO will be used for prediction of tropical cyclone activity the following summer into fall. Both were chosen due to the dynamical connection with Atlantic tropical cyclone activity (Gray 1984; Klotzbach 2011). TSR variables will not be included in this study due to lack of availability regarding the CLIPER model. The IOD related variables include the dipole mode index and both poles (east and west) of the IOD are defined as the three other variables.

Canonical Correlation Analysis

An analysis of the west and east IOD poles for each month during the August to

December period (Year 0) will be correlated to the following season’s ACE index (Year +1).

Correlation analysis of prior autumn variables (Year 0) will be performed with the following season’s ACE index and August-October Oceanic Niño Index (Year +1). The significant relationships will be described and compared to results using other methodologies. All analyses will be completed over a thirty year period starting in 1984/85 and ending in 2013/14 (i.e. Year

0/Year +1).

Time-Frequency Link between ENSO and the ACE index

The wavelet analysis (Torrence and Compo 1998) will demonstrate what time intervals

(i.e. approximate 2 to 8 years periodicity) a dataset (e.g. annual Niño region 3 sea-surface temperatures) frequents, as shown by the global wavelet variance values in Figure 14(c).

Wavelet analysis accomplishes this by comparing the Niño region 3 sea-surface temperature time series to the randomized integral of white noise noise via the Monte Carlo method. The two- dimensional wavelet power spectrum pattern (Figure 14(b)) is used to assess the power of time-

39 frequency relationships over the period (year) and length (~1870 to 1996) of the time series

(Torrence and Compo 1998). The interactive wavelet web resource is used to plot graphics similar to Figure 14 for all time-frequency analyses of data (Torrence and Compo 2000).

Figure 14. Niño region 3 wavelet power spectrum. (a) Data in time series format. (b) The wavelet power spectrum of those data. The contours are shown at 25% (blue), 50% (green), 75% (light green), and 95% (red) of the wavelet power scale; each percentage represents a confidence interval. (c) The global wavelet power spectrum. (Torrence and Compo 1998)

The shaded contour area depicted by 75% and 95% confidence intervals (i.e. yellow- green and red shading) indicates a frequency and time in which Niño region 3 sea-surface temperature data moderately to strongly differentiates itself from first integration of white noise created from a Monte Carlo process (Torrence and Compo 1998). Datasets with similar two- dimensional wavelet spectrum features (i.e. contoured confidence interval areas and global variance) have the highest probability of being correlated if physical reasoning favors causation.

Power spectrum scales vary with different variable types and associated units, but this does not impede the analysis of comparing analogous periodicity between two variables. The analysis will attempt to confirm if the Annual Atlantic ACE index and Oceanic Niño Index from August to

October are related temporally from 1985 to 2014.

Time-Frequency Link between Autumn Variables and the ACE index

Wavelet analysis (Torrence and Compo 1998) will be used to determine if the four autumn variables (1984 to 2013) have similar time-frequency signatures compared to the following year’s Atlantic Basin tropical season ACE index (1985 to 2014). The same method used from the previous analysis is applied to autumn variables.

Time-Frequency Link between IOD Variables and ENSO

Wavelet analysis is adapted to find the connection between the October values of the

DMI, EIOD, and WIOD the August to October Oceanic Niño Index. The connection with ENSO is also diagnosed because it is the mechanism whereby the IOD modulates Atlantic tropical cyclone activity. An investigation regarding the difference between the WIOD and EIOD with

41 respect to ENSO will be carried out. Analyses will be completed over a 30-year period starting in 1984/85 and ending in 2013/14.

Cross-Correlation between Seasonal Tropical Cyclone Counts and the IOD

Correlations between the October values of EIOD, WIOD, and DMI crossed by subsequent tropical cyclone season storm counts of tropical storms, hurricanes, and major hurricanes will be calculated. U.S. landfalling hurricanes and per diem counts of tropical storms, hurricanes, and major hurricanes will also be assessed against the preceding October EIOD,

WIOD, and DMI.

Composite Cumulative ACE Track Maps and Statistics

Cumulative ACE track maps are composed of five years of summative ACE from tropical cyclones spatially using a longitudinal (x) and latitudinal (y) Atlantic Basin Cartesian grid. Each individual tropical cyclone exerts a certain amount of cyclonic energy to the mapping grid dependent on storm track and magnitude. Storm tracks come and go throughout the season and leave there cyclonic energy imprint cumulatively upon this mapping product. The areas where intense tropical cyclones occur and tropical cyclone tracks frequently cross, contain the highest values of cumulative ACE. Instead of using just one seasonal composite of ACE, five seasons of cumulative ACE will be composited for this analysis. The parings of the five highest valued years and five lowest valued years associated with three IOD variables (DMI, EIOD, and

WIOD) will be utilized to construct six cumulative ACE track map composites. The datasets included in the analysis are the three October IOD indices. Additionally, cumulative ACE statistics will be given per each five-year composite from the 1985 to 2012 Atlantic tropical

42 cyclone seasons. The MATLAB code supplied for the cumulative ACE track composite maps is publicly available through Kerry Emanuel’s MIT website (http://eaps4.mit.edu/faculty/Emanuel

/products).

Spatial Correlation of WIOD to Tropical Season Atlantic Basin Climate Variables

Spatial correlation maps are created to demonstrate links between the prior October

WIOD (1984-2013) and following tropical season’s climate variables (1985-2014). The variables of interest include sea-level pressure, 700mb relative humidity over the Atlantic Basin along with sea-surface temperatures over the equatorial Pacific. These maps are created through the ESRL web resource (ESRL 2015c).

CHAPTER 5

RESULTS

Cross Correlations and Running Correlation (1984/85-2013/14)

The autumnal WIOD and EIOD have different relationships when being compared to

ACE values of subsequent annual Atlantic tropical seasons. All correlations of +/- 0.35 (0.45) are statistically significant at the 95% (99%) confidence interval for a sample size of 30 samples.

The WIOD correlation to the ACE index rises from 0.36 in August to a peak of 0.50 in October then falls significantly in November and December (Figure 15). The EIOD has a weak relationship with the latter ACE index. A peak in correlation also occurs with the ACE in

October (-0.17), then the correlation collapses like the WIOD. This provides validity to Izumo et al. (2010) regarding the EIOD breakdown during the late autumn and winter proceeding the formation of an ENSO event. The results also favor the Zhao and Nigam (2015) monopole theory of the IOD. In this case, the October WIOD explains 25% of subsequent ACE season variability as opposed to the EIOD, which explains 3% (Figure 15). The disintegration between the next season’s ACE index and preceding running October to December EIOD and WIOD values are likely a result of the proxy relationship between the following boreal summer ENSO state and seasonal Atlantic ACE (Gray 1984; Klotzbach 2011).

Figure 15. Running correlations between the WIOD/EIOD versus the ACE index the following year. The red correlation values are significant at a 95% confidence interval.

During a negative (positive) IOD event, decay was due to the easterlies (westerlies) collapsing allowing for the EIOD sea-surface temperature pole to normalize (Izumo et al., 2010).

After the EIOD normalization, the westerlies (easterlies) strengthen allowing for the development of El Niño (La Niña) (Izumo et al., 2010). In conclusion, this particular running correlation analysis implies a link between the October WIOD state and the upcoming seasonal

ACE index. Tropical cyclone activity (related to ACE) in the Atlantic basin is driven by the shear and vertical motion brought on by ENSO through the Walker circulation (Gray 1984,

Aiyyer and Thorncroft 2006).

The statistically significant autumn variables for the following season’s Atlantic ACE include variables A, C, and E in Table 6. The August to September Oceanic Niño Index for the

45 following season is statistically significant to the variables B, E, and G in Table 6. The antecedent October WIOD (variable E) achieves the best link with both the subsequent year’s

ACE and ENSO state with correlations of 0.50 and -0.46, respectively. These results expound on the important dynamic links between the prior October WIOD, ensuing ENSO state, and subsequent Atlantic tropical cyclone activity. The prior October DMI was statistically significant to the ensuing August-October ENSO at the 95% confidence interval. The EIOD did not exhibit any significant relationships. The October-November AMO and November 500mb North

Atlantic height index (i.e. a CSU predictor) both had statistically significant relationships with next tropical season’s ACE index but not ENSO. This analysis concluded with the WIOD or west pole of the IOD being the dominant pole in the correlations to the ACE index and ENSO, likely again because of the EIOD decay (Table 6).

Table 6

Cross-correlations between the autumn-IOD variables and ensuing August-October ENSO state and seasonal ACE index. The yellow shaded boxes of correlation values are significant at 95% confidence. The red value has the best relationship.

Time-Frequency Analysis (1984/85-2013/14)

Gray (1984) and Klotzbach (2011) ensure a dynamic connection (through upper-level shear and vertical motion) exists between tropical cyclone activity and the state of ENSO. If

ENSO and the ACE index are related, then they should have similar power spectrums over thirty-year. Similarity is exhibited between the August-October Oceanic Niño Index (Figure

16(a)) and the seasonal ACE index (Figure 16(b)) power spectrums. The marked and shaped areas (e.g. ellipses, boxes, and arrows) denote similarities between different datasets. The analysis was used to confirm variable time-frequency relationships of the ACE index and ENSO state during the Atlantic tropical cyclone season.

The power spectrums of three autumn variables in Figures 17(a), 17(b), and 17(c) are compared to the ensuing seasonal Atlantic ACE index (Figure 17(d)) over a thirty year period.

The October-November AMO and November North Atlantic 500mb geopotential heights share a

~8 year periodicity variance with ensuing ACE index. November eastern equatorial sea-level pressure has a peak periodicity, shown by the variance, of around 15 to 16 years. Judging areas of analogous periodicity would lead one to conclude that the October-November AMO (Figure

17(a)) and November North Atlantic 500mb geopotential heights (Figure 17(b)) power spectrums are weak to moderate analogs of ACE index. Little tangible time-frequency results materialized when analyzing the wavelet results of the autumn variables.

Figure 16. Similar wavelet power spectrums of seasonal (a.) Atlantic ACE index and (b.) ASO Oceanic Niño Index indicate a time-frequency relationship between the ENSO state and Atlantic tropical cyclone activity. The ellipses (1 and 2) and the arrow denote areas of analogous periodicity. (after Torrence and Compo 1998)

Figure 17. Wavelet power spectrums of (a.) October-November Atlantic Multidecadal Oscillation, (b.) November Eastern Equatorial Sea-Level Pressure, (c.) November North Atlantic 500mb Geopotential Heights, and (d.) ACE Index for following season. (after Torrence and Compo 1998)

The October EIOD, IOD, and WIOD power spectrum profiles are contrasted with ensuing August-October ENSO (Oceanic Niño Index) conditions over the tropical Pacific. The primary reasoning for this analysis is to demonstrate that the IOD is coupled with ENSO where the IOD leads 8-14 months as expressed by Izumo et al. (2010). The possible time-frequency relationships between October IOD variables and ENSO could provide a unique way to present the interaction regarding the two teleconnections.

All of the October IOD variables in Figures 18(a), 18(b), and 18(c) have very similar power spectrums, cycle every two to four years, and match well with the August-October

Oceanic Niño Index (Figure 18(d)) for the upcoming season. Each of the thirty-year power spectrums contain circled or boxed areas of analogous periodicity as a means to show commonality between IOD datasets and August-October Oceanic Niño Index. The EIOD and

DMI power spectrums each share two of four analogous periodicity areas (Figures 18(b) and

18(c)). The WIOD shares three of four areas of analogous periodicity with the latter Oceanic

Niño Index (August-October). The leading October WIOD retains the best correlation (Table

6(E)) and most robust matching power spectrum with the 10 to 12 month lagged August-October

ENSO state (Figures 18(a) and 18(d)).

Figure 18. Wavelet power spectrums of (a.) October WIOD, (b.) October Dipole Mode Index (DMI), (c.) October EIOD, and (d.) August-October Oceanic Niño Index for following season. The ellipses (1 and 2) and boxes (3 and 4) denote areas of analogous periodicity. (after Torrence and Compo 1998)

Correlation of Named Storms and U.S. Hurricane Landfalls to the IOD variables

Atlantic tropical cyclone season named storms, hurricanes, and major hurricanes, along with corresponding per diem metrics, are correlated to the preceding October IOD variables from

1984/85 to 2013/14 (Table 7). The data source for named storm counts and days are from CSU

(2015). None of the IOD variables had significant correlations with U.S. hurricane landfalls. The

October WIOD preceding the next year’s tropical season had statistically significant relationships with every count type except U.S. hurricane landfalls. The only other significant relationship exists between hurricane days and the preceding tropical season’s October DMI.

(Table 7)

Table 7

Cross-correlations between preceding IOD variables and ensuing tropical season named storm counts. The yellow shaded boxes of correlation values are significant at a 95% confidence interval and red text denotes best correlation.

Composite Cumulative ACE Track Analysis (1984-2012)

The lowest and highest five values of each October IOD variable are compiled into mapped composites of ACE. All The WIOD, EIOD, and DMI values are assumed to be from the month of October of the previous year (1984 to 2011). For example, the highest five values of the DMI occur during the years of 1991, 1994, 1997, 2002, and 2006. A cumulative composite is made up of Atlantic tropical seasons one year after the highest five years of DMI values. The cumulative ACE is composited (Figure 19(a)) using the Atlantic tropical seasons of 1992, 1995,

1998, 2003, and 2007. These composite seasons represent the DMI’s influence on the spatiality and intensity of tropical cyclones over the Atlantic Basin. The lowest five values of the DMI correspond to the tropical seasons of 1985, 1996, 1997, 1999, and 2011. The highest DMI and lowest DMI composite maps (Figures 19(a) and 19(b)) have a very similar ACE imprint.

Slightly more tropical activity occurred over the Yucatan Peninsula and eastern part of the

Caribbean Sea on the highest DMI map (Figure 19(a)). The highest (lowest) DMI maps (Figure

19) are similar to the lowest (highest) EIOD maps (Figure 20) because the EIOD is negatively correlated with the DMI.

The maps (Figures 21(a) and 21(b) of the five highest and five lowest tropical seasons with respect to the WIOD contain very different spatial distributions of cumulative tropical cyclone energy. The lowest five WIOD composite (Figure 21(b)) does not contain many tropical cyclone tracks or points of vigorous ACE. The highest five WIOD composite (Figure 21(a)) displays accumulated ACE throughout the entire basin. The difference between the WIOD map composites is further evidence that the WIOD is an antecedent to ACE activity in the Atlantic

Basin during the tropical cyclone season.

Figure 19. Maps of the (a) five highest and (b) five lowest years representing preceding DMI values corresponding to ensuing tropical season’s cumulative composite ACE. Scale on right side represents cumulated ACE (104 푘푡2).

Figure 20. Maps of the (a) five highest and (b) five lowest years representing preceding EIOD values corresponding to ensuing tropical season’s cumulative composite ACE. Scale on right side represents cumulated ACE (104 푘푡2).

Figure 21. Maps of the (a) five highest and (b) five lowest years representing preceding WIOD values corresponding to ensuing tropical season’s cumulative composite ACE. Scale on right side represents cumulated ACE (104 푘푡2).

Summed counts of tropical cyclones and cumulative values of ACE (Table 8) were created for all six of the aggregated ACE maps (Figures 19-21). Again, the EIOD and IOD values do not exhibit noticeable differences between the five highest (positive) and five lowest

(negative) values of tropical cyclone counts or ACE (Table 8, grey). The differences in tropical cyclones counts and ACE between the five lowest (Table 8, blue) and five highest (Table 8, yellow) WIOD seasons is 538 ACE units, 26 tropical cyclones, and mean ACE per cyclone is 4.3

ACE units. (Table 8)

Table 8

Cumulative ACE summations and tropical cyclone counts for the maps in Figures 19, 20, and 21 relating the three October IOD variables.

Linking the WIOD to Ocean-Atmosphere Variables

The next analysis will focus solely on confirming the WIOD’s association dynamically with climate variables over the Atlantic Ocean during the August to October period when tropical cyclone activity is at climatological peak. The WIOD was chosen because it has been the most robust variable in the fall preceding the next year’s tropical cyclone season. Maps of spatial correlation have been examined to show the connection between ENSO and climate variables that favor tropical development in the Atlantic Basin. The spatial correlation maps will focus on the preceding October WIOD (Year 0) relationships with mean values of August to

October ocean-atmosphere variables (Year +1) from 1984/85 to 2013/14. The ocean-atmosphere variables include: Atlantic basin sea-level pressure, 700mb relative humidity, and sea-surface temperatures over the ENSO regions. The spatial correlation map (Figure 22) between the sea- level pressure and the antecedent WIOD depicts an inverse relationship. If western Indian Ocean sea-surface temperatures are high (low), then sea-level pressure over the Atlantic Basin should be low (high). The high values of WIOD relate to 850mb convergence, upward vertical velocities and convection over the Atlantic basin associated to La Niña’s Walker circulation illustrated in Figure 11 (Chu 2004). The opposite is also true regarding this relationship; Low values of the WIOD are associated with an El Niño driven Walker circulation. The spatial correlation of the October WIOD with respect to Atlantic basin sea-level pressure is statistically significant (at 95% and 99% confidence interval) over much of the area between 10°N and 30°N.

Figure 22. Spatial correlation map between Atlantic basin sea-level pressure and the WIOD. Areas of deep violet denote areas with relevant correlations. [Image provided by the NOAA/ESRL Physical Sciences Division, Boulder, CO, from their Web site (http://www.esrl.noaa.gov/psd/).]

The WIOD has a positive statistically significant relationship (95% confidence interval) with 700mb relative humidity around the main development region, South Atlantic, and Bay of

Campeche supporting development of AEWs and “home grown” tropical cyclones occur (Figure

23). The increase in relative humidity, in most cases, helps to bolster the strength of tropical cyclones (Gray 1968). Unfortunately, this study cannot conclude that 700mb relative humidity can act as a solely dependent process of IOD-ENSO climate dynamics. Discerning between the

Atlantic low-pressure cell of the Walker circulation (i.e. as a result of IOD-ENSO climate dynamics) and decadal Atlantic see-surface temperature cycles (AMO) as the reasoning behind the degree of 700mb relative humidity levels over the Atlantic basin would be a difficult process.

Figure 23. Spatial correlation map between Atlantic basin 700mb relative humidity and the WIOD. Arrows denote areas of 700mb relative humidity with relevant correlations. [Image provided by the NOAA/ESRL Physical Sciences Division, Boulder, CO, from their Web site (http://www.esrl.noaa.gov/psd/).]

Sea-surface temperatures over the Pacific ENSO regions maintain a significant inverse correlation (95% and 99% confidence intervals) with the prior October WIOD (Figure 24). The

Indo-Pacific relationship is well demonstrated by the confirmation of La Niña conditions associated with high values of the WIOD (Figure 24). La Niña conditions through the Walker circulation yield weaker upper level winds and less shearing over the Atlantic basin. Low sea- level pressure and more upward vertical motion (Chu 2004) is observed over the Atlantic while high pressure and downward vertical motion exist within the cell over the Peruvian Coast. The appearance of warm sea-surface temperatures (Figure 24) over the eastern Indian Ocean (i.e. low-pressure cell) and high values of the WIOD confirm the Indo-Pacific leg (e.g. Figure 11) of the Walker circulation.

The link between ENSO and the Atlantic basin tropical cyclone activity involves two different meteorological phenomena: upper level wind shear and upward vertical motion. Slower than normal 200mb upper level winds associated with a La Niña event make is easier for convection to take place. The convective cloud tops that have been sheared by fast upper-level winds are a detriment to a developing or mature tropical cyclone. Upward vertical velocity serves as the lifting mechanism for the development of convection over the basin. The Atlantic basin processes during El Niño events are the opposite during a La Niña events (i.e. descent and strong upper-level wind shear).

Figure 24. Spatial correlation map between tropical Pacific sea-surface temperatures and the WIOD. Blue and violet shaded areas (cold temperatures) denote significant inverse correlation linked to ENSO. [Image provided by the NOAA/ESRL Physical Sciences Division, Boulder, CO, from their Web site (http://www.esrl.noaa.gov/psd/).]

CHAPTER 6

SUMMARY AND CONCLUSION

The United States, Caribbean, Mexico, and Central American have experienced damaging and life-threatening storm surges, inland flooding, and wind damage associated with strong tropical cyclones. Billions of dollars in losses can occur (e.g. Hurricane Mitch in 1998) and hundreds of lives (e.g. Hurricane Katrina in 2005) can be taken with just one strong cyclone.

Tropical cyclones in the current era impinge on economic activity such as tourism, oil drilling, and coastal development. Destructive storms can further exacerbate socio-economically depressed coastal areas and cities that already have a social vulnerability component (Cutter et al., 2006). The economic and human consequences of Atlantic tropical cyclone activity accosts the need for improved and earlier Atlantic basin tropical cyclone predictions.

The first attempts at forecasting tropical cyclones were documented by Nicholls (1979) and Gray (1984) in the South Pacific and the Atlantic basins, respectively. Both researchers used teleconnections such as the QBO or Southern Oscillation as antecedents for tropical cyclone activity. Since 1984, many different outlets such as CSU, TSR, and CPC have continued to create and improve models. These models are only effective just before and during the tropical cyclone season. Early lead time forecasts have improved slightly over time, but many decision makers are seeking forecasts issued the previous November or December prior to the tropical cyclone season. The reinsurance sector has expressed the need for tropical cyclone forecasts to be issued by January 1; Internal processes and government regulations make it difficult for this

62 sector to adjust the cost-benefit analysis in time to change mitigation tactics and insurance policy

(Elsner et al., 2006; Barmpadimos et al., 2014).

The over-forecasted tropical season of 2013 was expected to have high counts of storms and an above average ACE index. Forecast failures by most meteorological outlets for the 2013

Atlantic tropical cyclone season support the demand for improved seasonal modeling. The failure of the seasonal forecast models forced the climatological community to think about ways to improve such forecasts.

The Indian Ocean Dipole’s ability to influence ENSO is of great interest for two reasons. Primarily, ENSO events explain variability in tropical cyclone activity over the Atlantic basin better than any other teleconnection. Secondly, the IOD is predictive of ENSO with an eight to fourteen month lead. If the reasoning is true, then the IOD could possibly be predictive of Atlantic tropical cyclone activity with about a year’s lead. The essence of this research was to document the new connection between the IOD (e.g. specifically the WIOD) and Atlantic basin tropical cyclone activity (Figure 25).

Almost all the analyses demonstrated the October WIOD (western pole of the IOD) as a valid predictor with strong relationships (r-values) to ENSO (-0.46), number of named storms

(0.53) and hurricanes (0.49), and the ACE index (0.50) relative to other lead predictor variables.

Albeit, the WIOD has demonstrated skill in portending tropical cyclone activity; The WIOD did not have a significant correlation to landfalling U.S. hurricanes (0.03). The October DMI and

EIOD did not have significant correlations to ensuing Atlantic tropical cyclone activity because of a breakdown (Izumo et al., 2010) of sea-surface temperature anomalies over the eastern pole

63 of the IOD. Methodologies including Pearson correlations, Wavelet power spectrums, spatial correlation, counts of named storm type correlations, composite cumulative ACE tracks, and composite ACE statistics all support this connection. The autumn Indian Ocean Dipole can be predictive of Atlantic seasonal tropical cyclone activity through Walker circulation dynamics, the idealized biennial IOD-ENSO cycle, and ENSO’s influence on Atlantic tropical cyclone activity.

The early predictive lead time of the WIOD, connection with ENSO the following year, and

ENSO’s concurrent influence on Atlantic tropical cyclone activity should be of great interest to the reinsurance sector. The WIOD’s inclusion into early developed statistical tropical storm forecasts could benefit decision makers in government, meteorological outlets, and the private sector with respect to long-term planning efforts and ability to minimize risks.

Figure 25. The research links between the WIOD, ENSO, and Seasonal Atlantic ACE index. Blue arrows are known research links. The red arrow is a new connection within the research.

Future work could include empirical orthogonal function analysis of the tropical Indian

Ocean. Indian Ocean related principal component indices would be utilized to find further relationships with Atlantic tropical cyclone activity. Pacific Basin tropical cyclone activity can be evaluated by finding early lead relationships with IOD variables. Downscaled results from future climate models (out to 100 years) could be used to identify changes in the IOD-ENSO relationship overtime. The IOD-ENSO relationship could also be used to make a predictive model for boreal North American winter temperature distribution and precipitation anomalies.

REFERENCES

Aiyyer, A. R., & Thorncroft, C. (2006). Climatology of vertical wind shear over the tropical Atlantic. Journal of Climate, 19(12), 2969-2983.

AMS. (2015). Teleconnection - AMS Glossary. Retrieved from

An, S. I., & Wang, B. (2001). Mechanisms of locking of the El Nino and La Nina mature phases to boreal winter*. Journal of Climate, 14(9), 2164-2176.

Angell, J. K. (1986). On the variation in period and amplitude of the quasi-biennial oscillation in the equatorial stratosphere, 1951-85. Monthly Weather Review, 114(11), 2272-2278.

AOML. (2015a). Frequently Asked Questions: When is hurricane season?. Retrieved from

AOML. (2015b). Tropical Cyclone Climatology. Retrieved from

Baldwin, M. P., Gray, L. J., Dunkerton, T. J., Hamilton, K., Haynes, P. H., Randel, W. J., ... & Takahashi, M. (2001). The quasi‐biennial oscillation. Reviews of Geophysics, 39(2), 179- 229.

Barmpadimos, I., Linford T., and Mitchell-Wallace K. (2014). Seasonal hurricane forecast skill and relevance to the (re)insurance industry, SCOR Global P&C - Technical Newsletter

Barnston, A. G., & Livezey, R. E. (1987). Classification, seasonality and persistence of low- frequency atmospheric circulation patterns. Monthly Weather Review, 115(6), 1083-1126.

Barnston, A. G., Chelliah, M., & Goldenberg, S. B. (1997). Documentation of a highly ENSO‐ related sst region in the equatorial pacific: Research note. Atmosphere-Ocean, 35(3), 367- 383.

Behera, S. K., Krishnan, R., & Yamagata, T. (1999). Unusual ocean‐atmosphere conditions in the tropical Indian Ocean during 1994. Geophysical Research Letters, 26(19), 3001-3004.

Bell, G. D., Halpert, M. S., Schnell, R. C., Higgins, R. W., Lawrimore, J., Kousky, V. E., Tinker, R., Thiaw, W., Chelliah, M., & Artusa, A. (2000). Climate assessment for 1999. Bulletin of the American Meteorological Society, 81(6), s1-s50.

Bell, G. D., & Chelliah, M. (2006). Leading tropical modes associated with interannual and multidecadal fluctuations in North Atlantic hurricane activity. Journal of Climate, 19(4), 590-612.

Bjerknes, J. (1969). Atmospheric teleconnections from the equatorial pacific 1. Monthly Weather Review, 97(3), 163-172.

Blake, E. S. (2014). The 2013 Atlantic Hurricane Season: The Quietest Year in Two Decades. Weatherwise, 67(3), 28-34.

Blake, E. S., & Gibney, E. J. (2011). NOAA Technical Memorandum NWS NHC-6: The Deadliest, Costliest, and Most Intense United States Tropical Cyclones from 1851 to 2010 (and Other Frequently Requested Hurricane Facts).

Blake, E. S., Landsea, C., & Gibney, E. J. (2011). The Deadliest, Costliest, and Most Intense United States Tropical Cyclones from 1851 to 2010 (and Other Frequently Requested Hurricane Facts). NOAA/National Weather Service, National Centers for Environmental Prediction, National Hurricane Center.

BOM. (2015a). About the Indian Ocean Dipole. Retrieved from

BOM. (2015b). The three phases of the El Niño–Southern Oscillation (ENSO). Retrieved from

Boruff, B. J., & Cutter, S. L. (2007). The Environmental Vulnerability of Caribbean Island Nations. Geographical Review, 97(1), 24-45.

Bove, M. C., Zierden, D. F., & O'Brien, J. J. (1998). Are Gulf landfalling hurricanes getting stronger?. Bulletin of the American Meteorological Society, 79(7), 1327-1328.

Brier, G. W. (1978). The quasi-biennial oscillation and feedback processes in the atmosphere- ocean-earth system. Monthly Weather Review, 106(7), 938-946.

Burpee, R. W. (1972). The origin and structure of easterly waves in the lower troposphere of North Africa. Journal of the Atmospheric Sciences, 29(1), 77-90.

Cai, W., Zheng, X. T., Weller, E., Collins, M., Cowan, T., Lengaigne, M., Yu, W., & Yamagata, T. (2013). Projected response of the Indian Ocean Dipole to greenhouse warming. Nature Geoscience, 6(12), 999-1007.

Camargo, S. J., & Zebiak, S. E. (2002). Improving the detection and tracking of tropical cyclones in atmospheric general circulation models. Weather and Forecasting, 17(6), 1152-1162.

Camargo, S. J., Barnston, A. G., Klotzbach, P. J., & Landsea, C. W. (2007). Seasonal tropical cyclone forecasts. Wmo Bulletin, 56(4), 297.

Camargo, S. J., & Sobel, A. H. (2010). Revisiting the influence of the quasi-biennial oscillation on tropical cyclone activity. Journal of Climate, 23(21), 5810-5825.

CCSP SAP 1.3. (2008). Reanalysis of Historical Climate Data for Key Atmospheric Features: Implications for Attribution of Causes of Observed Change. R. M. Dole (Ed.). US Climate Change Science Program.

Chelliah, M., & Bell, G. D. (2004). Tropical multidecadal and interannual climate variability in the NCEP-NCAR reanalysis. Journal of Climate, 17(9), 1777-1803.

Chu, P. S. (2004). ENSO and tropical cyclone activity. Hurricanes and typhoons: Past, present, and potential, 297-332.

Colbert, A. J., & Soden, B. J. (2012). Climatological variations in North Atlantic tropical cyclone tracks. Journal of Climate, 25(2), 657-673.

CPC. (2015a). Tropical Cyclone Climatology. Retrieved from

CPC. (2015b). Historical El Niño/ La Niña episodes (1950-present). Retrieved from

CSU. (2015). CSU Forecast Verification File. Retrieved from

Cutter, S. L., Boruff, B. J., & Shirley, W. L. (2006). Social Vulnerability to Environmental Hazards. Hazards, Vulnerability, and Environmental Justice, 115-132.

DeMaria, M., & Kaplan, J. (1994). Sea surface temperature and the maximum intensity of Atlantic tropical cyclones. Journal of Climate, 7(9), 1324-1334.

Dunion, J. P., & Velden, C. S. (2004). The impact of the Saharan air layer on Atlantic tropical cyclone activity. Bulletin of the American Meteorological Society, 85(3), 353-365.

Easterling, D. R., Meehl, G. A., Parmesan, C., Changnon, S. A., Karl, T. R., & Mearns, L. O. (2000). Climate extremes: observations, modeling, and impacts. Science, 289(5487), 2068-2074.

Elsner, J. B., Liu, K. B., & Kocher, B. (2000). Spatial variations in major US hurricane activity: Statistics and a physical mechanism. Journal of Climate, 13(13), 2293-2305.

Elsner, J. B., Murnane, R. J., & Jagger, T. H. (2006). Forecasting US hurricanes 6 months in advance. Geophysical Research Letters, 33(10).

Elsner, J. B., Kossin, J. P., & Jagger, T. H. (2008). The increasing intensity of the strongest tropical cyclones. Nature, 455(7209), 92-95.

Emanuel, K. (2005). Increasing destructiveness of tropical cyclones over the past 30 years. Nature, 436(7051), 686-688.

ESRL. (2015a). Climate Timeseries AMO (Atlantic Multidecadal Oscillation) Index. Retrieved from

ESRL. (2015b). Create a monthly/seasonal mean time series from the NCEP Reanalysis Dataset. Retrieved from

ESRL. (2015c). Linear Correlations in Atmospheric Seasonal/Monthly Averages. Retrieved from

Fletcher, R. D. (1945). The general circulation of the tropical and equatorial atmosphere. Journal of Meteorology, 2(3), 167-174.

Fontaine, B., & Janicot, S. (1996). Sea surface temperature fields associated with West African rainfall anomaly types. Journal of Climate, 9(11), 2935-2940.

Frank, W. M., & Young, G. S. (2007). The interannual variability of tropical cyclones. Monthly Weather Review, 135(10), 3587-3598.

Goldenberg, S. B., Landsea, C. W., Mestas-Nuñez, A. M., & Gray, W. M. (2001). The recent increase in Atlantic hurricane activity: Causes and implications. Science, 293(5529), 474- 479.

Gray, W. M. (1968). Global view of the origin of tropical disturbances and storms. Monthly Weather Review, 96(10), 669-700.

Gray, W. M. (1979). Hurricanes: Their formation, structure and likely role in the tropical circulation. Meteorology over the Tropical Oceans, 77, 155-218.

Gray, W. M. (1984). Atlantic seasonal hurricane frequency. Part I: El Nino and 30 mb quasi- biennial oscillation influences. Monthly Weather Review, 112(9), 1649-1668.

Gray, W. M., Sheaffer, J. D., & Knaff, J. A. (1992). Inﬂuence of the stratospheric QBO on ENSO variability. Journal of the Meteorological Society of Japan, 70, 975-995.

Gray, W. M., Landsea, C. W., Mielke Jr, P. W., & Berry, K. J. (1994). Predicting Atlantic basin seasonal tropical cyclone activity by 1 June. Weather and Forecasting, 9(1), 103-115.

Gray, W. M., Klotzbach, P. J., & Landsea, C. W. (2006). Extended range forecast of Atlantic seasonal hurricane activity and US landfall strike probability for 2006. Fort Collins, Colorado State University, Department of Atmospheric Science. (May 2006)

Halpert, M. S., & Ropelewski, C. F. (1992). Surface temperature patterns associated with the Southern Oscillation. Journal of Climate, 5(6), 577-593.

Hamilton, J. (2013). 'Forecast Bust' Why 2013 Hurricane Predictions Were So Wrong. All Things Considered. Philadelphia, PA: National Public Radio. Retrieved from

Hartwell, F. E. (1930). The Santo Domingo Hurricane of September 1 to 5, 1930. Monthly Weather Review, 58, 362.

Holland, G. J., & Webster, P. J. (2007). Heightened tropical cyclone activity in the North Atlantic: natural variability or climate trend?. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 365(1860), 2695-2716.

Hopsch, S. B., Thorncroft, C. D., Hodges, K., & Aiyyer, A. (2007). West African storm tracks and their relationship to Atlantic tropical cyclones. Journal of Climate, 20(11), 2468- 2483.

Izumo, T., Vialard, J., Lengaigne, M., de Boyer Montegut, C., Behera, S. K., Luo, J. J., Cravatte, S., Masson, S., & Yamagata, T. (2010). Influence of the state of the Indian Ocean Dipole on the following year’s El Niño. Nature Geoscience, 3(3), 168-172.

Jarvinen, B. R., Neumann, C. J., & Davis, M. A. S. (1984). A tropical cyclone data tape for the North Atlantic Basin, 1886–1983: Contents, limitations. and uses. Tech. Memo. NWS NHC 22, National Oceanic and Atmospheric Administration.

JISAO. (2015). Pacific Decadal Oscillation (PDO) Graphics. Retrieved from

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woolen, J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgens, W., Janowiak, J., Mo, K. C., Ropelewski, C., Wang, J., Leetma, A., Reynolds, R., Jenne, R., & Joseph, D. (1996). The NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society, 77(3), 437-471.

Klotzbach, P. J. (2011). El Niño-Southern Oscillation's impact on Atlantic basin hurricanes and US landfalls. Journal of Climate, 24(4), 1252-1263.

Klotzbach, P. J., & Gray, W. M. (2008). Multidecadal variability in North Atlantic tropical cyclone activity. Journal of Climate, 21(15), 3929-3935.

Klotzbach, P. J., & Gray, W. M. (2009). Extended range forecast of Atlantic seasonal hurricane activity and landfall strike probability for 2010. Fort Collins, Colorado State University, Department of Atmospheric Science. (December 2009)

Klotzbach, P. J., & Gray, W. M. (2014a). Extended range forecast of Atlantic seasonal hurricane activity and landfall strike probability for 2014. Fort Collins, Colorado State University, Department of Atmospheric Science. (April 2014)

Klotzbach, P. J., & Gray, W. M. (2014b). Extended range forecast of Atlantic seasonal hurricane activity and landfall strike probability for 2014. Fort Collins, Colorado State University, Department of Atmospheric Science. (June 2014)

Klotzbach, P. J., Barnston, A. G., Bell, G., Camargo, S. J., Chan, J. C. L., Lea, A., Saunders, M., Vitart, F. (2011). Seasonal Forecasting of Tropical Cyclones. Ed. Guard C. Global Guide to Tropical Cyclone Forecasting, 2nd edition. World Meteorological Organization.

Knaff, J. A. (1997). Implications of summertime sea level pressure anomalies in the tropical Atlantic region. Journal of Climate, 10(4), 789-804.

Kossin, J. P., Camargo, S. J., & Sitkowski, M. (2010). Climate modulation of North Atlantic hurricane tracks. Journal of Climate, 23(11), 3057-3076.

Kug, J. S., & Kang, I. S. (2006). Interactive feedback between ENSO and the Indian Ocean. Journal of Climate, 19(9), 1784-1801.

Kunreuther, H. (2006). Disaster mitigation and insurance: Learning from Katrina. The Annals of the American Academy of Political and Social Science, 604(1), 208-227.

Landsea, C. W. (1993). A climatology of intense (or major) Atlantic hurricanes. Monthly Weather Review, 121(6), 1703-1713.

Landsea, C. W., Gray, W. M., Mielke, P. W., & Berry, K. J. (1994). Seasonal forecasting of Atlantic hurricane activity. Weather, 49(8), 273-284.

Landsea, C. W., Pielke Jr, R. A., Mestas-Nunez, A. M., & Knaff, J. A. (1999). Atlantic basin hurricanes: Indices of climatic changes. Climatic Change, 42(1), 89-129.

Latif, M., Dommenget, D., Dima, M., & Grötzner, A. (1999). The role of Indian Ocean sea surface temperature in forcing east African rainfall anomalies during December-January 1997/98. Journal of Climate, 12(12), 3497-3504.

Loschnigg, J., & Webster, P. J. (2000). A Coupled Ocean-Atmosphere System of SST Modulation for the Indian Ocean*. Journal of Climate, 13(19), 3342-3360.

Loschnigg, J., Meehl, G. A., Webster, P. J., Arblaster, J. M., & Compo, G. P. (2003). The Asian Monsoon, the Tropospheric Biennial Oscillation, and the Indian Ocean Zonal Mode in the NCAR CSM*. Journal of Climate, 16(11), 1617-1642.

Mack, J. (2013). The Cryosphere and North Atlantic Tropical Cyclone Activity: Statistical Forecasting and Physical Mechanisms. Southern Illinois University at Carbondale.

Maue, R. N. (2011). Recent historically low global tropical cyclone activity. Geophysical Research Letters, 38(14).

Meyers, G. (1996). Variation of Indonesian throughflow and the El Niño‐Southern Oscillation. Journal of Geophysical Research: Oceans (1978–2012), 101(C5), 12255-12263.

Meyers, G., McIntosh, P., Pigot, L., & Pook, M. (2007). The years of El Niño, La Niña, and interactions with the tropical Indian Ocean. Journal of Climate, 20(13), 2872-2880.

Minobe, S. (2000). Spatio-temporal structure of the pentadecadal variability over the North Pacific. Progress in Oceanography, 47(2), 381-408.

NHC. (2015a). Saffir-Simpson Hurricane Wind Scale. Retrieved from

NHC. (2015b). Tropical Cyclone Climatology. Retrieved from

Nicholls, N. (1978). Air-sea interaction and the quasi-biennial oscillation. Monthly Weather Review, 106(10), 1505-1508.

Nicholls, N. (1979). A possible method for predicting seasonal tropical cyclone activity in the Australian region. Monthly Weather Review, 107(9), 1221-1224.

Nicholls, N. (1989). Sea surface temperatures and Australian winter rainfall. Journal of Climate, 2(9), 965-973.

NOAA. (2015). Weather Impacts of ENSO: The Jetstream. Retrieved from

Nyberg, J., Malmgren, B. A., Winter, A., Jury, M. R., Kilbourne, K. H., & Quinn, T. M. (2007). Low Atlantic hurricane activity in the 1970s and 1980s compared to the past 270 years. Nature, 447(7145), 698-701.

Okumura, Y. M., & Deser, C. (2010). Asymmetry in the duration of El Niño and La Niña. Journal of Climate, 23(21), 5826-5843.

Pielke Jr, R. A., & Landsea, C. W. (1998). Normalized hurricane damages in the United States: 1925-95. Weather and Forecasting, 13(3), 621-631.

Pielke Jr, R. A., Rubiera, J., Landsea, C., Fernández, M. L., & Klein, R. (2003). Hurricane vulnerability in Latin America and the Caribbean: Normalized damage and loss potentials. Natural Hazards Review.

Pielke Jr, R. A., Landsea, C., Mayfield, M., Laver, J., & Pasch, R. (2005). Hurricanes and global warming. Bulletin of the American Meteorological Society, 86(11), 1571-1575.

Powell, M. D., & Reinhold, T. A. (2007). Tropical cyclone destructive potential by integrated kinetic energy. Bulletin of the American Meteorological Society, 88(4), 513-526.

Prospero, J. M., & Lamb, P. J. (2003). African droughts and dust transport to the Caribbean: Climate change implications. Science, 302(5647), 1024-1027.

Rogers, J. C. (1984). The association between the North Atlantic Oscillation and the Southern Oscillation in the northern hemisphere. Monthly Weather Review, 112(10), 1999-2015.

Ropelewski, C. F., & Halpert, M. S. (1987). Global and regional scale precipitation patterns associated with the El Niño/Southern Oscillation. Monthly Weather Review, 115(8), 1606-1626.

Saji, N. H., Goswami, B. N., Vinayachandran, P. N., & Yamagata, T. (1999). A dipole mode in the tropical Indian Ocean. Nature, 401(6751), 360-363.

Saji, N. H., & Yamagata, T. (2003). Possible impacts of Indian Ocean dipole mode events on global climate. Climate Research, 25(2), 151-169.

Saunders, M. A., & Lea, A. S. (2005). Seasonal prediction of hurricane activity reaching the coast of the United States. Nature, 434(7036), 1005-1008.

Saunders, M. A., & Lea, A. S. (2008). Large contribution of sea surface warming to recent increase in Atlantic hurricane activity. Nature, 451(7178), 557-560.

Saunders, M. A., & Lea, A. S. (2014). June Forecast Update for Atlantic Hurricane Activity in 2014. University College London, Department of Space and Climate Physics.

Schlesinger, M. E., & Ramankutty, N. (1994). An oscillation in the global climate system of period 65-70 years. Nature, 367(6465), 723-726.

Shapiro, L. J. (1987). Month-to-month variability of the Atlantic tropical circulation and its relationship to tropical storm formation. Monthly Weather Review, 115(11), 2598-2614.

Shapiro, L. J. (1989). The relationship of the quasi-biennial oscillation to Atlantic tropical storm activity. Monthly Weather Review, 117(7), 1545-1552.

Shapiro, L. J., & Goldenberg, S. B. (1998). Atlantic sea surface temperatures and tropical cyclone formation. Journal of Climate, 11(4), 578-590.

Simpson, R. H., & Saffir, H. (1974). The hurricane disaster potential scale. Weatherwise, 27(8), 169, 186.

Smith, I. (1994). Indian ocean sea‐surface temperature patterns and australian winter rainfall. International Journal of Climatology, 14(3), 287-305.

Solow, A. R., & Moore, L. J. (2002). Testing for trend in North Atlantic hurricane activity, 1900- 98. Journal of Climate, 15(21), 3111-3114.

Terray, P., & Dominiak, S. (2005). Indian Ocean sea surface temperature and El Nino-Southern Oscillation: A new perspective. Journal of Climate, 18(9), 1351-1368.

Torrence, C., & Compo, G. P. (1998). A practical guide to wavelet analysis. Bulletin of the American Meteorological Society, 79(1), 61-78.

Torrence, C., & Compo, G. P. (2000). Interactive wavelet plot. Retrieved from

Trenberth, K. E., & Shea, D. J. (2006). Atlantic hurricanes and natural variability in 2005. Geophysical Research Letters, 33(12).

Trenberth, K., Branstator, G., Karoly, D., Kumar, A., Lau, N. C., & Ropelewski, C. (1998). Progress during TOGA in understanding and modeling global teleconnections associated with tropical. Journal of Geophysical Research, 103, 14-291.

TSR. (2014). Tropical Storm Risk (TSR). Retrieved from

Vecchi, G. A., & Knutson, T. R. (2008). On Estimates of Historical North Atlantic Tropical Cyclone Activity*. Journal of Climate, 21(14), 3580-3600.

Vinayachandran, P. N., Iizuka, S., & Yamagata, T. (2002). Indian Ocean dipole mode events in an ocean general circulation model. Deep Sea Research Part II: Topical Studies in Oceanography, 49(7), 1573-1596.

Walker, G. T. (1924). Correlation in seasonal variation of weather IV. Mem. Indian Meteor. Dept, 24, 275-332.

Wang, X., & Wang, C. (2014). Different impacts of various El Niño events on the Indian Ocean Dipole. Climate Dynamics, 42(3-4), 991-1005.

Webster, P. J., Moore, A. M., Loschnigg, J. P., & Leben, R. R. (1999). Coupled ocean– atmosphere dynamics in the Indian Ocean during 1997–98. Nature, 401(6751), 356-360.

Webster, P. J., & Hoyos, C. D. (2010). Beyond the spring barrier?. Nature Geoscience, 3(3), 152-153.

Wong, S., & Dessler, A. E. (2005). Suppression of deep convection over the tropical North Atlantic by the Saharan Air Layer. Geophysical Research Letters, 32(9).

Wu, R., & Kirtman, B. P. (2004). Understanding the impacts of the Indian Ocean on ENSO variability in a coupled GCM. Journal of Climate, 17(20), 4019-4031.

Zhao, Y., & Nigam, S. (2015). The Indian Ocean Dipole: A Monopole in SST. Journal of Climate, 28(1), 3-19.

Zhu, J., Huang, B., Zhang, R. H., Hu, Z. Z., Kumar, A., Balmaseda, M. A., Marx, L., & Kinter III, J. L. (2014). Salinity anomaly as a trigger for ENSO events. Scientific reports, 4.

APPENDIX

ABBREVIATIONS AND ACRONYMS

Table 9

Abbreviations and acronyms.

Abbreviation or Acronym Definition or Organization ACE Accumulated Cyclonic Energy Index AEW African Easterly Wave AMO Atlantic Multidecadal Oscillation AOML Atlantic Oceanographic and Meteorological Laboratory BOM Australian Bureau of Meteorology CLIPER CLImatological-PERsistence ENSO Model CPC Climate Prediction Center CSU Colorado State University DMI Dipole Mode Index ECMWF European Centre for Medium-Range Weather Forecasts EIOD Eastern Pole of Indian Ocean Dipole ENSO El Niño Southern Oscillation ESRL Earth System Research Laboratory HURDAT HURricane DATabases maintained by the NHC IOD Indian Ocean Dipole JISAO Joint Institute for the Study of the Atmosphere and Ocean NAO North Atlantic Oscillation NCEP National Centers for Environmental Prediction NHC National Hurricane Center NOAA National Oceanic and Atmospheric Administration NTC Net Tropical Cyclone Index ONI Oceanic Niño Index QBO Quasi-Biennial Oscillation SST Sea Surface Temperature TSR Tropical Storm Risk WIOD Western Pole of Indian Ocean Dipole