UNIVERSITY of READING Department of Meteorology Dynamically Simulated Tropical Storms: Their Natural Variability and Response To

UNIVERSITY OF READING

Department of Meteorology

Dynamically Simulated Tropical Storms: Their Natural Variability and Response to Climate Change

RAY BELL

A thesis submitted for the degree of Doctor of Philosophy

November 2013 DECLARATION

I conﬁrm that this is my own work and the use of all material from other sources has been properly and fully acknowledged.

Ray Bell

Page i ABSTRACT

Tropical cyclones can cause substantial loss of life and an improved understanding of storm variability and their response to climate change can help inform preparation and future adaptation. The influence of the El Nino˜ Southern Oscillation (ENSO) on global tropical cyclone activity is investigated in a high-resolution coupled climate model (HiGEM) compared to an atmosphere-only simulation using the atmospheric component of HiGEM (HiGAM). HiGEM is able to capture the shift in tropical cyclone locations to ENSO in the Pacific and Indian Oceans but not in the North Atlantic. The vertical wind shear response over the Caribbean is not captured in HiGEM compared to HiGAM and ERA-Interim. Pre- cipitation biases in HiGEM remain in HiGAM over the Western North Pacific, however HiGAM simulates a more accurate representation of the ENSO-tropical cyclone teleconnection. Model experiments are subsequently undertaken to investigate a contemporary issue on how different types of El Nino˜ influence tropical cyclone activity in the Western North Pacific. The HiGEM-HiGAM central Pacific El Nino˜ experiment simulates an increase of tropical cyclones that move towards South East Asia. This response is attributed to a large-scale anti-cyclonic anomaly over east China. In contrast, the low SST in the West Pacific during east Pacific El Nino˜ reduces tropical cyclone activity. How tropical cyclone activity might change due to the influence of increased atmospheric carbon dioxide concentrations using HiGEM is finally investigated. Tropical cyclones are shown to decrease in frequency globally by 9 % in the 2×CO2 (2CO2) simulation and 26 % in the 4×CO2 (4CO2) simulation. Tropical cyclones only become more intense in the 4CO2. A decrease in mean ascent at 500 hPa contributes to the reduction of tropical cyclones in the 2CO2 in most basins. The larger reduction of tropical cyclones in the 4CO2 arises from further reduction of mean ascent at 500 hPa and a large enhancement of vertical wind shear.

Page ii ACKNOWLEDGEMENTS

I would like to thank my supervisors: Pier Luigi Vidale, for allowing me to undertake such an interesting PhD topic and providing me with the HiGEM data; Kevin Hodges, for an incredible amount of technical help and keeping me on my toes with papers; Jane Strachan, for her patience and guidance in helping me to pursue aspects outside of my PhD. I would also like to thank members of the tropical group for fruitful discussions, Andy Heaps for a godsend of IDL codes and many PhD students for not just technical help; but creating such an enjoyable working environment. A big thanks to Eric Guilyardi for sharing ideas on the different types of El Nino,˜ especially when over a beer. Thank you Reinhard Schiemann and Marie-Estelle Demory for helping me setup my UM experiments. Malcolm Roberts, the unsung hero behind high-resolution climate modelling at the Met Ofﬁce. I would also like to give a special thanks to my monitoring commit- tee: Sue Gray and Hilary Weller for encouraging comments and guidance throughout my PhD. To Angela, for coming into my life right when I needed her. To everyone who I have trained with over the last 3 years, for giving me the best possible break from my work. Finally, I would like to thank my family for their continual support throughout my PhD.

Page iii Contents

Contents

1 Introduction 1 1.1 Motivation and research questions ...... 1 1.1.1 Tropical cyclones and their impact on society ...... 3 1.1.2 Tropical cyclones and their role in climate ...... 4 1.1.3 Tropical cyclones and their impact on biology and chemistry . . . 5 1.2 An introduction to tropical cyclones ...... 6 1.2.1 How do tropical cyclones form? ...... 7 1.2.2 The theory of tropical cyclones ...... 13 1.3 Climate change in the tropics ...... 15 1.3.1 Tropical North Atlantic warming ...... 16 1.3.2 Tropospheric water vapour ...... 17 1.3.3 Walkercirculation ...... 20 1.3.4 Hadleycirculation ...... 22

2 Data, Method and Tools 24 2.1 Introduction...... 24 2.2 Generalcirculationmodels ...... 24 2.2.1 Horizontal resolution ...... 27 2.2.2 Modelphysics ...... 28 2.3 Themodelsusedinthisstudy...... 29 2.3.1 HiGEM...... 29 2.3.1.1 Model description ...... 30 2.3.1.2 Climate change experiments ...... 31 2.3.2 HiGAMAMIP ...... 32 2.3.3 HiGEM-HiGAM time slice experiments ...... 32 2.4 Observationaldata ...... 32 2.4.1 Tropical cyclones: IBTrACS ...... 33 2.4.2 SeaSurfacetemperature:HadISST ...... 33 2.4.3 Precipitation: GPCP ...... 34

Page iv Contents

2.5 Reanalysesdata...... 34 2.5.1 ERA-Interim ...... 35 2.6 Tropical cyclone tracking algorithms ...... 35 2.6.1 TRACK...... 37 2.7 HiGEM:Modelperformance ...... 39 2.7.1 Tropical cyclones in the present-day climate of HiGEM ..... 39 2.7.2 Large-scale environmental climatology in HiGEM ...... 43 2.7.2.1 Sea surface temperature ...... 43 2.7.2.2 Precipitation ...... 44 2.7.2.3 Relative humidity ...... 45 2.7.2.4 Walker circulation ...... 46 2.7.2.5 Vertical wind shear ...... 48 2.7.2.6 Upper-level circulation ...... 49

3 The ENSO-Tropical Cyclone Teleconnection 51 3.1 Introduction...... 51 3.1.1 El Nino˜ Southern Oscillation ...... 52 3.1.1.1 Measurements of ENSO ...... 55 3.1.1.2 Construction of ENSO composites ...... 57 3.2 The ENSO-tropical cyclone teleconnection: A review ...... 57 3.3 ENSOsimulationinHiGEM ...... 65 3.4 The ENSO-tropical cyclone teleconnection in HiGEM ...... 66 3.4.1 ENSO and tropical cyclone location ...... 67 3.4.2 ENSO and tropical cyclone frequency ...... 71 3.5 ENSO and large-scale environmental conditions ...... 72 3.5.1 Seasurfacetemperature ...... 74 3.5.2 Precipitation ...... 75 3.5.3 Walkercirculation ...... 77 3.5.4 Verticalwindshear ...... 79 3.5.5 Low-level vorticity ...... 81 3.5.6 Upper-level circulation ...... 83 3.6 Thermodynamic vs. dynamic inﬂuences ...... 86 3.7 Discussion...... 90 3.8 Conclusion ...... 91

Page v Contents

3.9 Futurework...... 93

4 The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity 94 4.1 Introduction...... 94 4.1.1 Different types of El Nino˜ ...... 95 4.1.2 Different types of El Nino˜ and tropical cyclone activity: A review 98 4.2 Simulation of different types of El NinoinHiGEM˜ ...... 104 4.3 Modelexperiments ...... 107 4.4 Tropicalcyclonechanges ...... 110 4.5 Large-scale environmental conditions ...... 115 4.5.1 Precipitation ...... 116 4.5.2 Walker circulation ...... 118 4.5.3 Verticalwindshear ...... 121 4.5.4 Upper-level circulation ...... 122 4.5.5 Steeringﬂow ...... 124 4.5.6 Relative humidity ...... 125 4.6 Discussion...... 127 4.7 Conclusion ...... 129 4.8 Futurework...... 130

5 Tropical Cyclones and Climate Change 132 5.1 Introduction...... 132 5.1.1 Observed trends in tropical cyclone activity ...... 132 5.1.2 Tropical cyclones and climate change: A review ...... 135 5.2 Tropical cyclones and climate change simulations ...... 140 5.2.1 Tropical cyclone location changes ...... 141 5.2.2 Tropical cyclone frequency changes ...... 142 5.2.3 Tropical cyclone intensity changes ...... 143 5.2.4 Tropical cyclone duration changes ...... 146 5.2.4.1 Methodology ...... 146 5.2.4.2 Results...... 147 5.2.5 Tropical cyclone structure changes ...... 148 5.2.5.1 Methodology ...... 148 5.2.5.2 Results...... 151 5.3 Changes in large-scale environmental conditions ...... 156

Page vi Contents

5.3.1 Sea surface temperature change ...... 157 5.3.2 Precipiationchange...... 159 5.3.3 Mid-level relative humidity change ...... 160 5.3.4 Circulation change ...... 161 5.3.5 Verticalwindshearchange ...... 166 5.3.6 Thermodynamic vs. dynamic inﬂuences ...... 169 5.4 Summaryandconclusions ...... 175 5.5 Futurework...... 177

6 Conclusions 179 6.1 Introduction...... 179 6.2 Synthesisofresults ...... 180 6.2.1 How does El Nino˜ Southern Oscillation (ENSO) influence global tropical cyclone activity? ...... 181 6.2.2 What is the role of the eastern tropical Pacific on tropical cyclone activity associated with different types of El Nino?˜ ...... 182 6.2.3 What is the response of tropical cyclones to climate change? . . . 183 6.3 Suggestions for further investigation ...... 184 6.3.1 How does El Nino˜ Southern Oscillation (ENSO) influence global tropical cyclone activity? ...... 184 6.3.2 What is the role of the eastern tropical Pacific on tropical cyclone activity associated with different types of El Nino?˜ ...... 185 6.3.3 What is the response of tropical cyclones to climate change? . . . 186

Bibliography 188

Page vii Chapter 1: Introduction

Chapter 1: Introduction

1.1 Motivation and research questions

Tropical cyclones can cause substantial loss of life and an improved understanding of storm variability could help inform preparation and response to landfalling events. Due to recent advances in available computing resources, General Circulation Models (GCMs) can now be run with a high enough resolution to simulate different aspects of tropical cyclone activity (e.g. Zhao et al., 2009; Smith et al., 2010; Murakami et al., 2012b; Manganello et al., 2012; Strachan et al., 2013). The response of tropical cyclones on climate time scales can also be investigated using GCMs, which provide a platform to examine the physical and thermodynamical mechanisms. The ability of a GCM to simulate tropical cyclone activity is a demanding test of a model’s capability. The following research questions are addressed:

1. How does the El Nino˜ Southern Oscillation (ENSO) inﬂuence global tropical cyclone activity?

2. What is the role of the eastern tropical Paciﬁc on tropical cyclone activity associated with different types of El Nino.˜

3. What is the response of tropical cyclones to climate change?

This thesis is structured as follows. The ﬁrst chapter contains a general introduction to tropical cyclone theory in terms of how the large-scale environmental conditions favour tropical cyclone development and also the mechanisms for tropical cyclogenesis.

Page 1 Chapter 1: Introduction

A literature review is also provided on climate change in the tropics and its implications on future tropical cyclone activity. I have additionally included a separate, more specific literature review to the start of each work chapter. Chapter 2 will introduce the General Circulation Models (GCMs) used in this study as well as discussing how well the GCMs are able to capture the mean state of global tropical cyclone activity. In addition, the simulation of large-scale environmental conditions is also investigated. Literature is also presented on aspects of GCMs which are important for the simulation of tropical cyclones. Observational and reanalyses datasets that are used for verification of the models are also introduced. A literature review on different types of tropical cyclone tracking algorithms is provided along with a description of the tropical cyclone tracking algorithm which is used in this study. Chapter 3 will address the ability of the GCMs to capture the expected response of tropical cyclones to the phase of ENSO. An examination of the teleconnections associated with ENSO on the large-scale environmental conditions is used to explain both the advantages and shortcomings of the simulated tropical cyclone response. Novel research on different types of El Ninos˜ and the tropical cyclone response in the Western North Pacific is investigated in chapter 4. Idealised experiments with SST forcing in the eastern tropical Pacific compared to global forcing is used to examine the role of the eastern tropical Pacific. In chapter 5 how the GCM simulates future changes in tropical cyclone activity is addressed. This chapter looks at changes of tropical cyclone metrics in idealised stabilised

2×CO2 and 4×CO2 experiments. An understanding of the mechanisms responsible for the simulated change is also presented. The thesis will end with a conclusions chapter which brings together results from the previous chapters and attempts to address the original research questions presented in this introduction. Separate conclusions will be given at the end of each work chapter. A series of recommendations of the areas that need to be addressed in order to improve our understanding of tropical cyclones in terms of natural variability and climate change with the use of a GCM is ﬁnally presented.

Page 2 Chapter 1: Introduction

1.1.1 Tropical cyclones and their impact on society

Tropical cyclones can broadly be described as cyclonic disturbances that originate over the tropical or occasionally subtropical regions of the world’s oceans (see section 1.2 for a detailed introduction of tropical cyclones). They are amongst the most devastating natural phenomena and can cause substantial loss of life and damage to infrastructure. This has recently been highlighted by hurricane Katrina, which in August 2005 killed at least 1,833 people; displaced hundreds of thousands more and damaged major oil refineries, causing gasoline prices in the USA to soar (Sampe and Xie, 2007). The top 10 costliest hurricanes have caused in excess of $650 bn worth of damage (2013 $’s) and are shown in figure 1.1 (ICAT, 2013). Cyclone Nargis (2008) made landfall in Burma and caused at least 138,000 fatalities (Swiss Reinsurance Company, 2009). However, the largest number of deaths resulting from a tropical cyclone is associated with a cyclone that hit Bangladesh in 1970 (Bhola). It is believed at least 300,000 people died from the associated storm surge in the low-lying deltas (Holland, 1993). The National Hurricane Centre provides a useful resource of tropical cyclone records: http://www.aoml.noaa.gov/hrd/tcfaq/tcfaqE.html. A better understanding of tropical cyclones and their social impacts has the potential to contribute to efforts to reduce vulnerability to hurricane related events that threaten society (Pielke and Pielke, 1997). Pielke and Pielke (1997) also offer solutions to reduce hurricane damage, such as improved housing designs and land use. A change in tropical cyclone activity due to anthropogenic climate change is a cause for concern for the insurance industry when assessing future climate risk (Dailey et al., 2009) as well as for society in general. A scientific approach to studying tropical cyclones in the context of climate variability and change is needed to provide guidance for adaptation and mitigation strategies over the coming decades. It is also needed in plan- ning the support and funding of public programmes for disaster management, including preparedness and recovery (Pielke and Pielke, 1997).

Page 3 Chapter 1: Introduction

Figure 1.1 Tracks of the top 10 costliest North Atlantic hurricanes. Taken from http://www. icatdamageestimator.com/toptendamages.

1.1.2 Tropical cyclones and their role in climate

Tropical cyclones are an important aspect of the climate system. Tropical cyclones can remove heat and moisture from the ocean, which affects the formation of deep water that drives ocean circulation (Bourassa, 2009). Tropical cyclones can affect the vertical heat profile in the ocean, help drive the poleward flux of heat and thereby the thermohaline circulation and ultimately global climate circulation (Sriver, 2010). In addition, the large amount of precipitation associated with tropical cyclones can lower the local ocean surface density through addition of fresh water (Shi and Wang, 2011). Scoccimarro et al. (2011) found that tropical cyclones can influence the trade winds in the tropics on short time scales. They found that occurrences of tropical cyclones would weaken the trade winds around 5-18oN and strengthen the surface winds away from the equator at 18- 30oN. They attribute more than 10 % of ocean heat transport out of the tropics to tropical cyclones. Recently, studies have emerged which investigate the feedback effects of tropical cyclones onto the climate system (Emanuel, 2008). Figure 1.2 shows a schematic of how

Page 4 Chapter 1: Introduction tropical cyclones can influence the large scale environment. Schenkel (2012) has investigated research on the passage of supertyphoon Tip in the North West Pacific (which occurred in 1979). Compensating for the extremely low minimum surface pressure of Typhoon Tip (870 hPa) there was a broad increase in surface pressure across the tropical Pacific days later. This led to subsequent suppression of tropical cyclone activity in the Pacific and this mechanism is believed to explain the typical spacing between multiple tropical cyclones.

Figure 1.2 Conceptual model depicting the impacts of tropical cyclones upon the large scale environment (Schenkel, 2012).

1.1.3 Tropical cyclones and their impact on biology and chemistry

Tropical cyclones have a noticeable inﬂuence on the levels of ocean productivity by upwelling nutrients from depth, as a response to surface wind divergence associated with the passage of tropical cyclones. This can have a large local impact on the oceanic carbon cycle (Shi and Wang, 2011). High levels of nutrients are contained in deeper cooler water due to decaying biota and faecal pellets. After the passage of a tropical cyclone, phytoplankton are able to utilise the upwelled nutrients and bloom. The phytoplankton

Page 5 Chapter 1: Introduction subsequently show periodic blooming associated with oscillation of the oceanic thermocline back to its resting state (Shay, 2010). Research has shown the ocean colour (related to sediment and biota) has an influence on tropical cyclone steering by affecting the local SST (Gnanadesikan et al., 2010). Tropical cyclones can also damage coral reefs by mechanical action from the production of large waves and also sediment dispersion, preventing the polyps from photosynthesising efficiently (Hongo et al., 2012). In the atmosphere, tropical cyclones are known to have aided speciation of Pacific Islands, in terms of birds, seeds and insects. Post-hurricane Katia, which affected Scotland in September 2011, brought birds often found in North America to Cornwall (BBC, 2011). Tropical cyclones are important for air-sea exchange of gases including carbon dioxide (CO2). Tropical cyclones can cause large outgassing of CO2 from the ocean following their passage, which can come from a depth of up to 500 m and hundreds of kilometres either side of the tropical cyclone (Bond et al., 2011). Nemoto et al. (2009) indicated that

40-60 % of the summertime sea-to-air CO2 ﬂux could be accounted for by the passage of only a few tropical cyclones.

1.2 An introduction to tropical cyclones

Tropical cyclones can broadly be described as cyclonic disturbances that originate over the tropical or occasionally subtropical regions of the world’s oceans. The word cyclone can be deﬁned as a weather system in which winds move in a circular direction around a centre of low barometric pressure. Tropical cyclones have a warm centre whereas baroclinic cyclones are mostly cold core (Longshore, 2008). Tropical cyclones can be cate- gorised into three main classes based on their intensity: a tropical depression is a tropical storm which has 1-minute sustained / 10-minute sustained lower level wind speeds less than 17 m s−1; a tropical storm has sustained winds of 17-33 m s−1; and hurricanes, typhoons and tropical cyclones have wind speed larger than 33 m s−1. The different terms used for tropical cyclones are given depending on which basin they form in. They are

Page 6 Chapter 1: Introduction known as hurricanes in the Atlantic and North East Pacific, typhoons in the North West Pacific and tropical cyclones in other basins. Hurricanes can be classified further using the Saffir-Simpson scale (Simpson and Riehl, 1981) which ranges from category 1: the least intense, to category 5: the most intense, in terms of wind speed. The previous definition included the amount of damage caused. These are defined in table 1.1.

Category Central pressure (mb) wind speed (m s−1) (mph) Surge (m) Damage

1 ≥ 980 33-42 74-95 1-1.5 Minimal 2 965-979 43-49 96-110 2-2.5 Moderate 3 945-964 50-58 111-130 3-3.5 Extensive 4 920-944 59-69 131-155 4-5.5 Extreme 5 < 920 > 69 > 155 > 5.5 Catastrophic

Table 1.1 Safﬁr-Simpson Hurricane scale (Pielke and Pielke, 1997)

1.2.1 How do tropical cyclones form?

Tropical cyclones emerge from initial disturbances. This can be in form of: an African easterly wave (Rielh, 1948; Hopsch et al., 2009); broad scale vorticity convergence (Yanai, 1964); upper tropospheric troughs (Ramage, 1950) and the Inter Tropical Con- vergence Zone (ITCZ; Tanabe, 1963). Tropical cyclones can also form from ordinary cold fronts, which can penetrate into the tropics. Thunderstorm complexes that form over land occasionally drift out over the ocean and can also become tropical cyclones (Emanuel, 2005b). In the Western North Paciﬁc, initiation of tropical cyclones can arise from monsoon troughs (Ritchie and Holland, 1999). In the Atlantic, a large number of tropical cyclones occur from a baroclinic-barotropic instability of the African easterly jet. 85 % of the intense hurricanes formed in the Atlantic are believed to have their ori- gins as easterly waves (Landsea, 1993). In late summer and early autumn the convection associated with the waves increases and winds near the surface evolve from a typical

Page 7 Chapter 1: Introduction wavy pattern into a closed circulation, which can create a tropical depression (Emanuel, 2005b). As the wave approaches a location, the surface winds shift from north-easterly to south-easterly and heavy downpours commence on and off for a day or so. Alvia and Pasch (1995) suggest these easterly waves cause nearly all of the tropical cyclones in the Eastern Pacific Ocean. There has been a great deal of research investigating why some initial disturbances transform into tropical cyclones and why others do not. Dunkerton et al (2009) describes the process of tropical cyclogenisis as the ‘marsupial paradigm’ for African easterly waves. Tropical cyclogenesis takes place within an African easterly wave via a hybrid wave-vortex structure. The flow streamlines associated with the African easterly wave form a closed circulation in the lower troposphere. The large-scale wave is able to provide: containment of moisture within the developing gyre; confinement of mesoscale vortex aggregation and maintenance or enhancement of the parent wave until the vortex becomes a self-sustaining entity and emerges from the wave as a tropical depression. This is known as the marsupial paradigm akin to development of a marsupial infant in its mother’s pouch. The intense surface-concentrated vortex core that defines a tropical cyclone is con- structed when the thermodynamic environment favours the outbreak of widespread and persistent deep convection. The upward driven air in the low-to-mid troposphere enhances convergence and vorticity at this level (Tory and Frank, 2010). A tropical cyclone becomes self-sustaining when the energy derived locally from surface sensible heat and moisture fluxes are sufficient to maintain the circulation against friction to allow intensification of the storm (Rotunno and Emanuel, 1987). Early work on the understanding of large-scale environmental factors important for the formation of tropical cyclones was undertaken by William Gray at Colorado State University. Gray (1968) was the first to devise a global climatology of tropical cyclones, which led to an understanding of the processes which dominate in the regions where tropical cyclones form (known as the main development regions). Since Gray’s research much work has been done to investigate the relationship between tropical cyclone gene-

Page 8 Chapter 1: Introduction sis and the large-scale environmental conditions (Tory and Frank, 2010; Camargo et al., 2010). Tropical cyclones form in the tropics in the vicinity of warm sea surface temperature (SST). Warm SST are needed to provide the storms with a source of energy. Warm SST leads to warm air just above the surface via evaporation, which has a capacity to hold more moisture, given by the Clausius-Clapeyron relationship:

1 dqs 1 des L ≈ = 2 (1.1) qs dT es dT RvT

where qs is the saturation specific humidity, es is the saturation water vapour pressure, T is air temperature, L is the latent heat of vapourisation, Rv is the gas constant for water vapour, and the first term approximates to the second term (see Ambaum, 2010 and Al- lan, 2011). The source of moisture provides energy to the tropical cyclone via latent heat release (see section 1.3.2). In general, SST must be greater than 26 oC to support tropical cyclone development (Palmen, 1948). However, Cione (2012) notes that above this SST, when the air temperature at 10 m is less than the SST, tropical cyclone intensification occurs. Conversely, if the air temperature at 10 m is greater than the SST a shutdown of enthalpy fluxes prevents tropical cyclone intensification. The depth integrated ocean temperature - potential energy - is important for tropical cyclone genesis. Warm SST often triggers convection; however, warm waters below the surface will ensure there is a continuation of energy available for convection. The warm water in the upper levels of the ocean indicate a deep thermocline. Warm core ocean eddies are one example of where the thermocline can deepen locally. Goni et al. (2003) showed that if the tropical cyclone heat potential - the integrated water column heat content down to the 26 oC isotherm - is 30 kJ cm−2 or greater there is a more likely chance a tropical cyclone will intensify. Emanuel (2005a) note that a drop in 2.5 oC in ocean temperature under the eyewall will generally halt energy input into the storm. A relatively deep oceanic mixed layer depth of 50 m is often quoted as being able to support tropical cyclone genesis (Briegel and Frank, 1997). Vertical wind shear - defined as the magnitude of the vector difference between winds

Page 9 Chapter 1: Introduction at 850 hPa and 200 hPa - has long been known to influence tropical cyclogenesis. Ver- tical wind shear can affect the thermodynamics of a tropical cyclone by ventilation of low-entropy air into the core. It can also influence the dynamics of a tropical cyclone by creating convective asymmetries, which distorts the flow around a tropical cyclone. A review on the influence of vertical wind shear on tropical cyclone activity can be found in Tang and Emanuel (2010). A schematic on the influence of vertical wind shear on a tropical cyclone is shown in figure 1.3. Upper level troughs in the atmosphere often provide a source for strong vertical wind shear which prevents tropical cyclogenesis. Al- though vertical wind shear usually has a detrimental effect on tropical cyclones, a small amount of vertical wind shear often aids tropical cyclone development, similar to how it aids continual development of convective systems (Paterson et al., 2005; Tao, 2012).

Figure 1.3 Schematic on the inﬂuence of vertical wind shear on a tropical cyclone (Heymsﬁeld et al., 2006).

Tropical cyclones also require a humid environment to form. Along with warm SST, an increase in humidity promotes the development of convective activity. Humidity,

Page 10 Chapter 1: Introduction which condenses with height in the atmosphere, releases latent heat. Moisture has a strong inﬂuence on the likelihood of instabilities occurring, which trigger convection, and if the conditions are right, tropical cyclones. When the atmosphere is supersaturated with moisture - the change in equivalent potential temperature with height is positive - instabilities occur. Dry air in the mid-to-low troposphere can act as a barrier for tropical cyclone development. One example of this: the formation of tropical cyclones in the North Atlantic can be suppressed by the advection of Saharan air over the ocean. This dry, dusty air between 800-500 hPa overlays a marked temperature inversion and can advect over the ocean with the easterly jet (Dunion and Veldon, 2004). Emanuel et al. (2008) also note that moisture is important in regulating how long it takes an initial disturbance to form, via a χm parameter - normalised difference of entropy in the boundary layer to that of the mid-troposphere:

χ sb − sm m ≡ ∗ (1.2) so − sb

∗ where sm,sb and so are the entropies of the middle troposphere and boundary layer, and the saturation entropy of the sea surface, respectively. The moist entropy is defined in Emanuel (1994) as: L q s ≡ c lnT − R ln p + v − R qlnH (1.3) p d T v here cp is the heat capacity at constant pressure of air, Lv is the latent heat of vapourisation, q is the specific humidity, Rd is the gas constant for dry air, Rv is the gas constant for water vapour, and H is the relative humidity. Tropical cyclones require ‘spin’ to maintain their cyclonic motion. It is physically impossible for a tropical cyclone to form on the equator where the planetary vorticity is 0. A minimum distance of 500 km from the equator can provide enough planetary vorticity to support the development of a tropical cyclone given a precursor. However, one tropical cyclone, Varmei (2003) was able to form within 1.5o of the equator in the South China Sea. Interaction between air flow over topography and local meteorological conditions were favourable for tropical cyclone development. The probability of a similar

Page 11 Chapter 1: Introduction equatorial development is estimated to be once every 100-400 years (Chang et al., 2003). Tropical cyclones are unlikely to form poleward of the tropics due to cooler extratropical SST and stronger vertical wind shear. Recent research has tried to combine these large-scale environmental factors to give a quantitative relationship to tropical cyclogenesis. The development of a Genesis Poten- tial Index (GPI) by Emanuel and Nolan (2004) and later Camargo et al. (2007c) is given as: 5 3 H 3 Vpot 3 2 GPI = |10 η| 2 ( ) ( ) (1 + 0.1V )− (1.4) 50 70 shear where η is the absolute vorticity at 850 hPa (s−1), H is the relative humidity at 700 −1 hPa (%), and Vshear is the wind shear (m s ). Vpot the potential intensity in terms of maximum velocity (m s−1) is given as:

2 Ts Ck ∗ b Vpot = (CAPE −CAPE ) (1.5) T0 Cd

where Ts is the SST, and T0 is the mean outflow temperature (at the level of neutral buoyancy), Ck is the exchange coefficient of enthalpy and CD is the drag coefficient. The Convective Available Potential Energy (CAPE) is the vertical integral of parcel buoyancy as a function of parcel temperature, pressure, specific humidity, and vertical profiles of virtual temperature. CAPE∗ is the value of CAPE for an air parcel which has been lifted from saturation level at the sea surface temperature and pressure, while CAPEb is the value of CAPE for boundary layer air. Both are evaluated at the radius of maximum winds. The potential intensity parameter is discussed further in section 1.2.2. A set of constants are assigned to the variables of absolute vorticity, relative humidity, potential intensity and vertical wind shear, to represent the relative importance and variability of each parameter related to tropical cyclogenesis. However, the constants are somewhat subjective although they are based on multiple regression analysis. It is worth noting that the GPI assumes the relationship of tropical cyclogenesis with the large-scale environmental conditions is stationary with time, limiting its application for climate change studies. Other GPIs are currently being developed to investigate the relationship

Page 12 Chapter 1: Introduction in speciﬁc basins, for example Bruyere` et al. (2012) for the North Atlantic, as well as an improved statistical relation of the variables to tropical cyclone genesis (Tippett et al., 2010; McGauley and Nolan, 2011).

1.2.2 The theory of tropical cyclones

What distinguishes a tropical cyclone from an ordinary convective system is its ability to become self-sustaining. Tropical cyclones can be thought of as a Carnot heat engine (Emanuel, 2006), which is shown in figure 1.4. Starting from point A: a low pressure system is formed and it draws surrounding air inward isothermally and acquires entropy by creation of water vapour over the warm ocean. As air converges the conservation of angular momentum causes it to rotate faster around the storms axis. At point B: the air flows upward nearly adiabatically in a rotating pattern and expands. The tropical cyclone then radiates some of its energy as Infra-Red radiation out to space. Air travels along the outflow of the tropical cyclone - at height, where it has neutral buoyancy, until all precipitation has been lost, from point C to D. Finally, the air undergoes adiabatic compression as it flows back to the surface (from point D to A) and gains water vapour from the ocean surface when travelling towards the storm. This cycle of evaporation and condensation is what brings in the energy from ocean heat and powers the storm. This has recently been clarified in Bister et al. (2011). Whilst the Carnot heat engine is a good conceptual model of how a tropical cyclone works it has some limitations as it assumes the tropical cyclone does no work on the environment, has an idealised steady-state and an axisymmetric flow. An understanding of these processes has led to the development of the Emanuel (1988) maximum potential intensity theory of a tropical cyclone. Whilst making the assumptions above, the rate of input of available energy, into the tropical cyclone from the sea surface is given by: ε ρ ∗ G = Ck Vs(Ko − Ka) (1.6)

Page 13 Chapter 1: Introduction

Figure 1.4 A tropical cyclone as a Carnot heat engine (Emanuel, 2006). where G stands for ‘generation’, Ck is a dimensionless coefficient called the enthalpy ρ 0 transfer coefficient, is density of air, Vs is the surface wind speed, and K∗ and K0 are enthalpies of the ocean surface and the atmosphere near the surface, respectively. ε is the thermodynamic efficiency, and is defined by: T − T ε = s 0 (1.7) Ts where Ts is the temperature of the ocean surface, T0 is the temperature at which heat is exported from the surface. The taller the tropical cyclone is, the lower the temperature at its top and thus, the greater the thermodynamic efficiency. Almost all of the energy generated is dissipated by friction acting between the winds and the sea surface. The rate of mechanical dissipation in the system, is given by:

ρ 3 D = Cd Vs (1.8)

where D stands for ‘dissipation’, Cd is the drag coefﬁcient and other symbols are the

Page 14 Chapter 1: Introduction same as in (1.6). By equating these two expressions an equation for a tropical cyclone’s maximum potential intensity is given in equation (1.5). Holland (1997) has a slightly different theory for maximum potential intensity which differs from the Emanuel (1988) theory in its treatment of relative humidity. It is also worth noting that the Emanuel (1988) theory has been challenged recently by Smith et al. (2008). The theory for tropical cyclone intensity has recently been updated to include the thermocline depth by Lin et al. (2013). This includes cooling of the upper ocean by tropical cyclone induced mixing. It was found substituting the pre-cyclone depth-average ocean temperature instead of SST gave more accurate estimates of potential intensity estimates compared to observations.

1.3 Climate change in the tropics

There is substantial evidence that the large scale environment from which tropical cyclones form and evolve is changing as a result of anthropogenic emissions of greenhouse gases. The Intergovernmental Panel on Climate Change IPCC (2013) summarise ‘it is extremely likely (95-100 % chance) that more than half of the observed increase in global average surface temperature from 1951 to 2010 was caused by anthropogenic increase in greenhouse gas concentrations and other anthropogenic forcings together’. The global mean temperature is still largely inﬂuenced by multi-year natural variability, which includes: solar variability, volcanic activity, aerosol variability and internal climate variability (Stott et al., 2000; Delworth and Knutson, 2000; Meehl et al., 2004; Knutson et al., 2006). All of these factors have been shown to impact regional tropical cyclone activity: Hodges and Elsner (2011) for solar variability; Evan (2012) for volcanic activity and Dunstone et al. (2013) for aerosols.

Page 15 Chapter 1: Introduction

1.3.1 Tropical North Atlantic warming

There is strong evidence that SST is increasing in the tropical cyclone basins (Emanuel, 2005b; Webster et al., 2005), partly because of a long-term increase in greenhouse gas emissions (Santer et al., 2006; Knutson et al., 2006; Knutson et al., 2010b). The North Atlantic basin in particular has received a great deal of attention due to the availability of long-standing observations. The North Atlantic also displays a pronounced mode of variability on multi-decadal time scales, which has a large impact on hurricane activity. Goldenberg et al. (2001) proposed that the Atlantic multi-decadal oscillation (AMO) had caused multi-decadal variations in hurricane activity since the 1940s, using a de- trended SST AMO index and vertical wind shear in the main development region (MDR; 10oN-20oN and 20oW-85oW). This pattern of SST warming and an associated time series is shown in ﬁgure 1.5. However, Mann and Emanuel (2006) suggest an alternative explanation. Using a statistical regressions approach, they showed that the evolutions of summertime tropical Atlantic SST can be reproduced through the 20th century as a combination of global mean surface temperature and sulphate aerosol forcing. This hypothesis has been further developed by Booth et al. (2012) who highlight the importance of aerosols as a driver for North Atlantic climate variability. However, this paper has recently been challenged by Zhang et al. (2013). The AMO has important impacts on vertical wind shear in the Atlantic MDR (Golden- berg et al., 2001; Vitart and Anderson, 2001; Zhang and Delworth, 2006) which affects tropical cyclone activity. Klotzbach and Gray (2010) believe a stronger than average thermohaline circulation is related to a positive AMO which reduces the strength of the sub-tropical Atlantic oceanic gyre. Cooler water is advected into the east Atlantic as the gyre slows, which alters local thermodynamic conditions in the east North Atlantic. The warmer SST can lead to an increase in tropical cyclone activity.

Page 16 Chapter 1: Introduction

Figure 1.5 First empirical orthogonal function of SST in the North Atlantic, representing the AMO. (Left panel) Spatial pattern and (right panel) associated time series (Goldenberg et al., 2001).

1.3.2 Tropospheric water vapour

Water vapour is the Earth’s dominant greenhouse gas. Through the release of latent heat when it condenses, it also plays an active role in dynamic processes that shape the global circulation of the atmosphere and hence the climate (Schneider et al., 2010). As discussed in section 1.2.1 water vapour is directly important for tropical cyclogenesis as relative humidity. The role which water vapour plays in determining the tropical circulation also has a strong influence on regions which are conducive for tropical cyclone development. Warm temperatures in the tropics allow the near surface air to become saturated with water vapour. As a moist parcel of air ascends, it cools as it expands and does work against the rest of the atmosphere. The water vapour in a parcel of air condenses with height and releases its latent heat (e.g. Trenberth and Stepaniak, 2003). Air in the tropics cools with height at a rate approximately 4.5 K km−1, known as the moist adiabatic lapse rate. A typical vertical profile of temperature in the tropics is given by the gray line in the top left panel of figure 1.6. A warming of tropical SST with climate change will also lead to a warming of near

Page 17 Chapter 1: Introduction

Figure 1.6 High resolution model of horizontally homogeneous radiative-convective equilibrium.

Top left: equilibrium temperature profiles for 3 values of CO2 compared to an observed tropical profile. Top right: the temperature differences compared to the response to doubling CO2 in an ensemble of CMIP3 global climate models. Bottom left: relative humidity profiles for 3 values of

CO2 compared to an observed tropical proﬁle. Bottom right: buoyancy of a lifted parcel through the mean sounding from 100 m (Romps, 2010). surface air. According to the Clausius-Clapeyron relationship (equation 1.1) warmer temperatures will result in a greater concentration of water vapour. Therefore, lower tropospheric water vapour is predicted to increase at a rate of 6-7 % K−1. However, precipitation is predicted to increase at a much slower rate, 2-3 % K−1 (Allen and Ingram, 2002). Precipitation is constrained not only by surface temperature; but the

Page 18 Chapter 1: Introduction earth’s energy budget. An increase in surface tropospheric water vapour will condense more water vapour in the upper atmosphere as the parcel of air rises. The increase in latent heat released will warm the upper atmosphere. That is, the moist adiabatic lapse rate decreases with warming as the upper troposphere warms more than the lower troposphere. This is shown in the top right panel of ﬁgure 1.6. As a parcel of air rises the warmer temperatures aloft will reduce its ability to cool radiatively. This slowing of outgoing longwave radiation has to balance the precipitation, explaining why precipitation increases less than tropospheric water vapour. Figure 1.7 shows this response is also robust in GCMs (Held and Soden, 2006).

Figure 1.7 Scatterplot of the percentage change in global-mean column-integrated (a) water vapour and (b) precipitation vs the global-mean change in surface air temperature (Held and Soden, 2006).

Relative humidity is projected to stay constant with climate change, which is highlighted in the bottom right panel of ﬁgure 1.6. If relative humidity would increase with climate change more cloud and rainfall would be produced. This has the effect of being a negative feedback on surface temperature. A ﬁxed relative humidity is also simulated by GCMs (Sherwood et al., 2010). A small increase in surface humidity may contribute to the muted precipitation response, which can be explained by the relationship of evap-

Page 19 Chapter 1: Introduction

∗ oration to speciﬁc humidity q near the surface and saturated speciﬁc humidity qs at the surface using the bulk aerodynamic formula given in Schneider et al. (2010) as:

ρ ∗ E ≈ Cw||~v||(1 − H)qs (1.9)

Here E stands for evaporation; ρ is density of near-surface air; ~v is the near-surface wind;

Cw is a bulk transfer coefficient; with H the near-surface relative humidity. An argument ∗ for the fractional change in E with qs is given in equation (3) of Schneider et al. (2010). The different warming rates in the tropics of the upper and lower troposphere has a profound affect for the large-scale tropical circulation and leads to an increase in atmospheric stability (Knutson and Manabe, 1995; Held and Soden, 2006; Vecchi et al., 2006; Bengtsson et al., 2007a; Richter and Xie, 2008; Allan, 2011). The physical basis for decreased tropical circulation can be understood in terms of the atmospheric mass flux (M) relating precipitation (P) with low-level specific humidity (q) given in equation 1.10.

P ≈ Mq (1.10)

As low-level speciﬁc humidity increases faster than precipitation with warming the mass ﬂux must decrease to compensate. This is discussed further in Allan (2011) and refer- ences therein. The implications of this on future tropical cyclone activity is discussed further in section 5.1.3.

1.3.3 Walker circulation

An important aspect of the tropical overturning circulation is the Walker circulation. The Walker circulation acts as the main pathway to inﬂuence regional tropical cyclone activity on a year-to-year time scale. The Walker circulation was discovered by Sir Gilbert Walker (Walker, 1923; see section 3.1.1) and is driven by the east-west temperature gradient across the tropical Paciﬁc. A schematic showing how the Walker circulation is likely to

Page 20 Chapter 1: Introduction change with climate change is shown in ﬁgure 1.8. As the tropical overturning circulation slows down: precipitation increases slower than surface moisture at point C. The upward motion has to slow down to transport the same amount of moisture. At point A the surface moisture increases which favours convection and the subducting motion in this region starts to slow. To compensate for these slowing branches the winds across the surface of the Paciﬁc are also likely to slow (Vecchi, 2006). Vecchi et al. (2006) report evidence for a weakening trend in the Walker circulation

Figure 1.8 A schematic of weakening of the Walker circulation (Vecchi, 2006). during the 20th century, which is consistent with hindcast predictions by a number of climate models. The weakening of the Walker circulation allows the associated upper branch to extend further into the North Atlantic, similar to what occurs during an El Nino˜ event (see section 3.7.4.) and hinder future tropical cyclone development over the Caribbean Sea. Wang and Lee (2008) report a statistical association between increasing vertical wind shear in the tropical Atlantic since 1949 and a global warming SST mode which is physically related to the weakening of the Walker circulation. Vecchi and Soden

Page 21 Chapter 1: Introduction

(2007b) also ﬁnd this relationship is due to global warming.

1.3.4 Hadley circulation

The Hadley cell represents a region of strong ascent off the equator, which follows the maximum seasonal heating and creates the inter-tropical convergence zone (ITCZ). The ITCZ can act as an initial disturbance for tropical cyclones and it also preconditions the atmosphere for tropical cyclone development. Convection causes air to rise to the upper atmosphere until it cools to become neutrally buoyant. Air in the upper branch conserves angular momentum and moves towards the pole. At some latitude the air starts to descend. The latitudinal extent of the Hadley cell is believed to primarily be related to the equator-pole temperature difference. Held and Hou (1980) deﬁne the latitudinal extent of the Hadley cell theoretically as: ∆ φ 5 gHt h 1/2 HH ≈ ( 2 2 ) (1.11) 3 Ω a T0

where Ω is the planetary vorticity, ∆h is the (vertically-averaged) pole-equator temperature contrast in radiative equilibrium, and T0 is a reference temperature. Ht is the height of the tropopause, given in Schneider et al. (2010) as:

Ts Ht ≈ (1 − c) γ + cHe (1.12)

where c is a constant, the emission height He increases as the concentration of greenhouse gases (or the optical thickness of the longwave absorber) increases, as well as the tropical surface temperature Ts. The tropical lapse rate γ also decreases with increased greenhouse gases. All three factors contribute to the increase in tropopause height. The Held-Hou model assumes the poleward ﬂow in the upper branches is approximately angular momentum-conserving and the circulation is energetically closed, so that diabatic heating in ascent regions is balanced by radiative cooling in descent regions. Similar to the Walker circulation, the Hadley cell is predicted to weaken with climate

Page 22 Chapter 1: Introduction change (Schneider et al., 2010). In addition to weakening, the Hadley cell is predicted to migrate poleward with climate change (Lu et al., 2007). The latitude of the Hadley cell terminus against global-mean surface temperature in idealised GCM simulations show the poleward shift of the Hadley cell in figure 1.9. A weaker pole-to-equator temperature gradient can be attributed to this shift, however, there are other factors which control the latitudinal extent of the Hadley cell. Walker and Schneider (2006); Schneider et al. (2010) found the Hadley circulation widens much more slowly with increasing radiative equilibrium than Held and Hou (1980); it also widens with increasing low-latitude static stability, whereas Held and Hou (1980) would imply that it is independent of static stability. Therefore, it is plausible to attribute the widening of the Hadley circulation with increasing low-latitude static stability to a poleward displacement of deep baroclinic eddy fluxes (Schneider et al., 2010). An increase in static stability implies the baroclinic eddy fluxes first become deep enough to reach the upper troposphere move poleward. Obser- vations have also shown the Hadley cell to widen in recent decades (Johanson and Fu, 2009). There is still a lack of research on the seasonality of the Hadley cell and therefore the relationship to Northern hemisphere and Southern hemisphere tropical cyclone activity.

Figure 1.9 Hadley circulation width versus global-mean surface temperature in idealized GCM simulations. Shown is the latitude of the subtropical terminus of the Hadley circulation, deﬁned as the latitude at which the mass ﬂux stream function at approximately 725 hPa is zero. The termination latitudes in both hemispheres are averaged (Schneider et al., 2010).

Page 23 Chapter 2: Data, Method and Tools

Chapter 2: Data, Method and Tools

2.1 Introduction

In this chapter the tools required to answer the research questions outlined in section 1.1. are introduced. The chapter provides an introduction to global climate modelling and describes the climate models used in this study. The reanalyses data, as well as observational products are also introduced. A review of tropical cyclone tracking algorithms is presented and the objective feature tracking algorithm which is used in this study is described. The mean state climatology of HiGEM along with the tropical cyclone climatology is presented with comparisons to reanalyses data and observations.

2.2 General circulation models

To understand the mechanisms behind global atmospheric and oceanic processes General Circulation Models (GCMs) can be used as a framework to provide additional information outside of traditional uses of observations and theory. The global climate system is incredibly complex, incorporating interactions between processes which occur at different spatial and temporal scales. GCMs represent the earth system on a three dimensional grid on which equations are solved which govern large-scale ﬂuid dynamics and thermodynamics. Reviews on the use of GCMs are given in Kalnay (2003); Held (2005); and Slingo et al. (2009). GCMs use time and space discretisation to solve the basic dynamical and thermody-

Page 24 Chapter 2: Data, Method and Tools namical equations representing the climate system. Either grid-point or spectral formu- lations can be used (Durran, 1999). For the grid-point model, a field is represented by its value at discrete grid points. On the other hand, a field is expressed using a discrete set of coefficients of known functions for the spectral model (Doron et al., 1974). GCMs have to take into account many different physical processes, some of which occur on smaller spatial scales than GCMs are able to explicitly resolve. Parameterisations aim to represent these sub-grid processes as accurately as possible in terms of the properties of the grid-box. One example of parameterisation that is important for simulated tropical cyclone activity is the cumulus convection scheme, which influences tropical precipitation (Murakami et al., 2012a). There are many different types of GCM which are used depending on the problem at hand. Atmosphere-only GCM (AGCM) studies typically use the ‘time-slice’ method (Bengtsson et al., 1995) to allow the use of higher resolution in the atmosphere than would otherwise be possible. The time slice approach utilises sea surface temperature (SST) and sea ice distributions taken from relatively low resolution coupled Atmosphere- Ocean GCMs (AOGCMs) experiments or observations as boundary conditions. The At- mospheric Model Intercomparison Project (AMIPII; Taylor et al., 2000) was set up to allow for comparison between AGCMs and to gain an understanding of limitations of the atmosphere only models between different modelling centres. Time-slice experiments which are used to study tropical cyclones typically have short integration lengths, of around ∼ 20 to 30 years. This makes it difficult to address whether the changes in tropical cyclones seen in these simulations are likely to be outside the range of natural variability, which can occur on timescales of several decades, and would therefore not be captured by such a short integration period. However, the integration lengths of AGCMs are becoming longer to take this into account or with the use of ensembles. Additionally, AGCMs do not allow tropical cyclone activity to feedback onto the SST and ocean heat content (Scoccimarro et al., 2011). However, the costs are greatly reduced compared to running a fully coupled Atmosphere-Ocean GCM (AOGCM). AOGCMs are coupled climate models which couple an ocean model with an atmo-

Page 25 Chapter 2: Data, Method and Tools sphere model and usually encompass other models such sea ice and land. The coupling allows feedback to take place between the two systems. However, the computational costs are much greater due to the larger number of equations which have to be solved. As a compromise, the resolution of AOGCMs are often quite low, for example the models used in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) have resolutions between 1.5o and 3o in the atmosphere and 1o in the ocean (IPCC, 2007). The models used in the more recent Fifth Assessment Report (AR5) (IPCC, 2013) have a slightly improved resolution, with a small number of models having resolution almost as high as the AOGCM used in this study (Tory et al., 2013; Camargo, 2013). Models of this resolution limits some applications to tropical cyclone studies, such as a change in tropical cyclone intensity with climate change. Earth system models (ESMs) incorporate increasingly complex components of the climate system, such as interactive biogeochemistry, including the carbon cycle. The latest models in the IPCC fifth assessment report (AR5) include earth system models. These models are of increasing interest to understand future tropical cyclone activity, for example Dunstone et al. (2013) found the mitigation pathway can influence future North Atlantic hurricane activity via aerosol levels. In addition, Evan et al. (2011) argue that changes in aerosols over the Indian Ocean have altered the atmospheric circulation in the pre-monsoon season over the Arabian Sea, leading to decreased vertical wind shear and more intense tropical cyclones. Regional Climate Models (RCMs) can be run at much higher resolutions than their global counterpart as they cover a limited geographical area (Christensen et al., 2007). RCMs are therefore better at resolving smaller scale features than a global model, such as tropical cyclone intensity and structure. However, the use of RCMs are limited due to coarse resolution boundary forcing from the GCMs which are used to drive the model. In addition, RCMs are only as good as the forcing model and for large domains, boundary conditions are not sufficient to constrain the inner domain unless spectral nudging is used. RCMs are not able to capture the interaction outside of the domain and therefore limit studies such as the El Nino˜ Southern Oscillation which have an impact on global

Page 26 Chapter 2: Data, Method and Tools tropical cyclone activity. Computational limitations place constraints on GCM experiments and a comprise must be made between grid-cell resolution, spatial domain, length of simulation, size of ensemble (a single GCM simulation is just one realisation of the climate, to sample the climate uncertainty associated with the non-linearity of the governing equations ensemble methods are used) and physical complexity, which are shown in ﬁgure 2.1.

Figure 2.1 Computational limits of simulating GCMs (Slingo, 2007).

2.2.1 Horizontal resolution

Horizontal resolution is arguably the most important aspect of GCMs in terms of their ability to realistically simulate tropical cyclone activity. Strachan et al. (2013) used a hierarchy of Hadley Center AGCMs with varying resolutions and found horizontal resolutions of 135 km or higher are able to capture the geographical location of tropical cyclones. Strachan et al. (2013) found that resolutions higher than 100 km are needed to simulate interannual variability; however much higher resolution models are needed to simulate tropical cyclone intensity and structure than used in this study. Bengtsson

Page 27 Chapter 2: Data, Method and Tools et al. (2007a) also notes that higher resolution models are needed to identify changes in tropical cyclone activity with increasing CO2. Manganello et al. (2012) investigated simulated tropical cyclone activity in a range of resolutions with the European Centre for Medium-Range Weather Forecasts Integrated Forecast System (ECMWF IFS) down to 10 km and noticed a marked improvement of intensity and structure of the most intense tropical cyclones. The role of horizontal resolution on the simulation of tropical cyclone activity has also been investigated using idealised studies. To assess the role of initial conditions and horizontal resolution on tropical cyclone intensification, Reed and Jablonowski (2010) investigate solutions of a balanced warm-core vortex in an aquaplanet configuration, using the finite-volume dynamical core of the the Community Atmosphere Model version 3.1 (CAM 3.1) developed at the National Center for Atmospheric Research. Reed and Jablonowski (2010) find thresholds for initial wind speed and the radius of maximum winds are needed to support intensification, at horizontal grid spacings less than 55 km. This idealised tool has led to a better understanding of the simulation and representation of tropical cyclone-like vorticies in global atmospheric models. Reed and Jablonowski (2012) have since investigated a simpler case by assessing the role of the dynamic core on the simulation of tropical cyclone intensity and structure.

2.2.2 Model physics

The physics package of a GCM has been shown to influence the simulation of tropical cyclone activity. Murakami et al. (2012a) have run an AGCM at 60 km (TL319) to investigate multi-physics and multi-sea surface temperature experiments on future tropical cyclone changes. Murakami et al. (2012a) showed variations in physics package - mainly the cumulus convection scheme - can influence the simulation of future tropical cyclone activity. Changes in the horizontal cumulus mixing rate and divergence damping coefficient has also been shown to affect simulated tropical cyclones (Zhao et al., 2012). Research of tropical cyclones in numerical weather prediction models can help explain

Page 28 Chapter 2: Data, Method and Tools deficiencies in climate models. For example, surface friction and micro-physics scheme influence tropical cyclone intensity in the Unified Model (Chamberlain, pers. comm.).

2.3 The models used in this study

The main aim of this thesis is to investigate natural variability, in terms of the El Nino˜ Southern Oscillation (ENSO), and climate change on the simulation of tropical cyclones in a high-resolution coupled climate model. In order to do this two versions of a high- resolution climate model have be used, the U.K. High-Resolution Global Environment Model (HiGEM) and the atmosphere-only model HiGAM. The section below gives a brief summary on the main components of HiGEM which are relevant for this study and an evaluation of the model performance with respect to tropical cyclone simulation. The simulations with the atmospheric model of HiGEM (HiGAM) which has been forced with observed SST and sea ice is used to compare to HiGEM. Further experiments using HiGEM SSTs to force HiGAM are explained which are used to address the issue of ocean coupling on the simulation of tropical cyclone activity. Finally, the model experiments used to investigate the research question on the role of the eastern tropical Paciﬁc on typhoon activity associated with different types of El Nino.˜

2.3.1 HiGEM

The UK HiGEM project developed from HadGEM1 (Johns et al., 2006; Ringer et al., 2006) as an exercise to investigate the role of resolution on GCM processes. The development of HiGEM is described in Shaffrey et al. (2009) and Catto (2009) provides a full list of the dynamics and physics used in HiGEM. In addition, Demory (2012); De- mory et al. (2013) provides useful technical information about HiGEM and the choice of resolution.

Page 29 Chapter 2: Data, Method and Tools

2.3.1.1 Model description

Resolution - HiGEM has 38 terrain-following, height-based vertical atmospheric layers extending to over 39 km in height, which are given in table 2 of (Ringer et al., 2006). The atmospheric horizontal resolution is N144 (1.25o latitude × 0.83o longitude) which has grid spacing or approximately 90 km at 50oN. The ocean component has horizontal resolution of 1/3o latitude × 1/3o longitude (approximately 30 km). The ocean component has 40 vertical levels with the majority near the surface to give an improved representation of the mixed layer and atmosphere-ocean coupling.

Initial conditions - The models are conﬁgured for current climate conditions, with present-day atmospheric loadings for aerosols, ozone and greenhouse gases for 1990 forcing. The exact values of present day greenhouse gases can be found in table 2.1. of Catto (2009).

Atmosphere-ocean coupling - HiGAM is coupled to the ocean model (higher resolution version of HadGOM1 in Johns et al., 2006) by passing mean daily fluxes from the atmosphere to the ocean and mean daily ocean surface boundary conditions to the atmosphere. As the land sea masks differ in the atmospheric and oceanic models a coastal tiling method is applied within the MOSES-II surface exchange scheme (Essery et al., 2003). Fluxes are computed separately for land and ocean. The atmosphere thus expe- riences fluxes of heat, water and momentum originating from a mixture of land and sea surface types. Fluxes are transferred to the ocean grid via horizontal bilinear interpolation with a correction for discontinuities in interpolating across the land-sea boundary. A local weight- ing given in Roberts et al. (2004) is then applied to the interpolated fields to ensure that fluxes are globally conserved.

Page 30 Chapter 2: Data, Method and Tools

2.3.1.2 Climate change experiments

To address the research question on tropical cyclones and climate change, idealised climate change scenarios are used with HiGEM and are shown in ﬁgure 2.2. The HiGEM control simulation was completed using present-day radiative forcing for 150 years. From the HiGEM control simulation, a transient climate change integration was performed with

CO2 levels increasing by 2 % per year from 1990 levels. The CO2 levels were then stabilised at 2×CO2 levels and integrated for a further 30 years. The transient integration was continued and the CO2 levels stabilised at 4×CO2. Again this was integrated for another 30 years. The two runs with the stabilised CO2 levels will be referred to as the 2CO2 and 4CO2. Tropical cyclone activity is investigated in both experiments and assessed to ﬁnd out whether any changes are outside the range of natural variability given by the 5×30-year control simulation. These experiments were run on the Earth super- compter in Japan in 2006. The data from these model experiments were provided to me.

Figure 2.2 Schematic showing the idealised climate simulations (Catto et al., 2011).

Page 31 Chapter 2: Data, Method and Tools

2.3.2 HiGAM AMIP

HiGAM is the atmospheric component of HiGEM. The model is forced with prescribed monthly SST and Sea Ice Concentration boundary conditions from the Atmospheric Model Intercomparison Project (AMIPII; Taylor et al., 2000), integrated for the period 1979 until 2002. This model is used to investigate the global ENSO-tropical cyclone teleconnection and is compared to the results obtained in HiGEM. The AMIPII SST data are available monthly but are linearly interpolated to daily values and are given on a 1o × 1o grid. The resolution was increased to force the higher resolution atmospheric component using the Met Ofﬁce ancillary programme (see Demory, 2012). The ability of the HiGAM model to simulate tropical cyclone activity is investigated fully in Strachan et al. (2013) and the ﬁndings are discussed in section 2.2.1.

2.3.3 HiGEM-HiGAM time slice experiments

The time slice experiments reveal differences in coupled and uncoupled simulations, which are investigated for projections of tropical cyclone intensity in section 5.2.3. The SSTs from the HiGEM simulations are used to provide the boundary conditions for HiGAM as time slice experiments. The present-day and 2CO2 experiments were integrated for 30 years, whereas only 25 years were available for the 4CO2.

2.4 Observational data

Climate models need to be evaluated to understand biases. These biases are likely to have large implications on the simulation of tropical cyclones in the climate models. This section includes a description of the observations used for validation of the models and analysis, which the results are shown in section 2.10. The observational products are also evaluated during El Nino˜ and La Nina˜ years in comparion to the models in chapter

Page 32 Chapter 2: Data, Method and Tools

2.4.1 Tropical cyclones: IBTrACS

Global observed tropical cyclone data from the International Best Track Archive for Cli- mate Stewardship (IBTrACS) (Knapp et al., 2010) are used to validate the model in terms of its ability to simulate the present-day tropical cyclone climatology. IBTrACS provides the best observations of tropical cyclones from multiple regional observational centres, which are merged into one product and quality controlled. However, there are known limitations of IBTrACS, such as global inhomogeneity (Landsea, 2007) and changes in the observational techniques (Landsea et al., 2006). IBTrACS brings the national records together which use identification based on wind thresholds to define when a storm is a tropical cyclone and therefore do not record the very early stage and later extra-tropical transition stage. Recent research has identified that prior to the satellite era (pre 1970s) many tropical cyclones over the ocean would have gone unnoticed (Vecchi and Knutson, 2010). With regard to global inhomogeneity the national tropical cyclone centres study activity in their region of interest and subsequently areas such in the Central Pacific and South West Pacific going largely unobserved.

2.4.2 Sea Surface temperature: HadISST

Observed SST are used from the Met Ofﬁce Hadley Centre’s sea ice and SST dataset, HadISST for the 1979-2010 period (Rayner et al., 2003) to evaluate the mean-state SST simulation of HiGEM and its ability to capture spatial SST changes with ENSO. The dataset is comprised of in situ sea surface observations and satellite derived estimates at the sea surface. The dataset extends back to 1850 and it uses Empirical Orthogonal Functions (EOFs) to recreate past global SST coverage from a few points using present- day SST patterns. The period of 1979 onwards uses homogeneous data which has global

Page 33 Chapter 2: Data, Method and Tools coverage.

2.4.3 Precipitation: GPCP

Observed precipitation is obtained from the Global Precipitation Climatology Project (GPCP) dataset in comparison to the models, which is a 2.5o × 2.5o gridded dataset that combines satellite estimates and rain gauge data (Adler et al., 2003). The data are used to evaluate the model performance in terms of its mean-state and ENSO variability. Precipi- tation rates in GPCP are produced through combining empirical infra-red estimates from geostationary satellites with empirical microwave estimates from polar orbiting satellites. The estimates are adjusted where gauge data is available over land, however, there are known limitations over the Ocean (Smith et al., 2006).

2.5 Reanalyses data

Reanalyses are based on the combination of a short range forecast made with a forecast model and a diverse set of observations using data assimilation. This is often based on a frozen operational system. The forecast is a cyclical process with the background forecast being made from an earlier assimilation. Reanalyses provide our best 4D view of the atmosphere constrained by observations and are homogeneous compared to the inho- mogeneous observations. Reanalysis data depend on the observation type and distribution, the data assimilation method and forecast model. Changes in the observing system can introduce spurious trends. The tracking algorithm (described below) is applied to reanalyses data using the same procedure as for HiGEM, providing a more consistent comparison between tracked model data and observational data (Strachan et al., 2013).

Page 34 Chapter 2: Data, Method and Tools

2.5.1 ERA-Interim

The ERA-Interim reanalysis is based on an operational forecast model constrained by the assimilation of data. ERA-Interim is the latest re-analysis product provided by the European Centre of Medium range Weather Forecasts (ECMWF). ERA-Interim differs from its predecessor, ERA-40, by using an improved Integrated Forecasting System (IFS) which uses complete four-dimensional variational data assimilation spectral model. The ERA-Interim period is from 1979 onwards to include the vast amounts of improved observations which started in 1979 such as: cloud motion winds obtained from geostationary satellites as well as surface pressure, temperature and winds from buoys. The overwhelming majority of data originates from satellites. This includes clear-sky radi- ance measurements from polar-orbiting satellites, scatterometer wind data, and ozone retrievals from various satellite-borne sensors. ERA-Interim has an improved moisture analysis leading to better representation of tropical precipitation, and variational bias correction which has improved the homogeneity in time of the reanalysis. The resolution of ERA-Interim is ∼ 80 km (T255) with 60 model layers up to 0.1 hPa (Dee et al., 2011). The tracking algorithm (see section 2.9.1) is applied to ERA-Interim using the same procedure as for HiGEM, providing a more consistent comparison between tracked model data and observational data (Strachan et al., 2013).

2.6 Tropical cyclone tracking algorithms

Feature tracking algorithms of tropical cyclones tend to be unique to each study, each basin and can be resolution dependent, which makes it difﬁcult to undertake systematic comparisons between basins and models. An evaluation of tracking algorithms is currently being undertaken in The Tropical Cyclone Climate Model Intercomparison Project (TCMIP; Walsh et al., 2010), as well as in the CLIVAR hurricane working group. The different tropical cyclone tracking algorithms that are used are described below.

Page 35 Chapter 2: Data, Method and Tools

Broccoli and Manabe (1990) developed the ﬁrst tracking algorithm which used a low level wind speed threshold in a GCM with 4.5o×7.5o resolution. The addition of a warm core temperature threshold was added later by Haarsma et al. (1993), to distinguish tropical cyclones from extratropical cyclones. Bengtsson et al. (1995) included a further criterion of the lower-level wind speed to be greater than the upper-level wind speed, a similar approach to many tracking algorithms currently use with vorticity as it implies a warm core. Camargo and Zebiak (2002) apply basin-dependent threshold criteria to three variables: low-level vorticity, surface wind speed, and vertically integrated temperature anomaly. Although Camargo and Zebiak (2002) argue this approach is taken to lengthen the tracks compared to those in Bengtsson et al. (1995). This approach is not entirely objective, as the tracking algorithm is tuned to match observations in each basin. Hart (2003) uses mean sea level pressure criteria to identify tropical cyclones. The cyclone phase space is them applied to identify tropical cyclones according to their struc- tural evolution (Hart, 2003). It is also used to identify the point at which a tropical cyclone undergoes extratropical transition. The cyclone phase space is described using parameters of asymmetry (symmetric versus asymmetric) and horizontal height gradient (warm core versus cold core). Walsh et al. (2007) comment on the limitations of tracking tropical cyclones in models and reanalyses of different resolutions when making comparisons with observed tropical cyclones. To overcome this, Walsh et al. (2007) apply a resolution dependent threshold for lower level wind speed. However, the thresholds are somewhat subjective and created so that the number of tropical cyclones tracked are close to that observed. Murakami et al. (2012a) use a tracking algorithm which is optimised for a given model conﬁguration to ensure that the present-day global annual mean tropical cyclone number matches that observed (84 year−1 1979-2003). They use a number of parameters including: maximum 850 hPa relative vorticity; maximum 850 hPa wind speed; a temperature threshold at 700, 500 and 300 hPa for evidence of a warm core aloft. They also use a radius of maximum wind threshold to remove tropical depressions in the North

Page 36 Chapter 2: Data, Method and Tools

Indian Ocean. Each model used in the study therefore varies its location of tropical cyclones, although most simulate far too many in the North Indian Ocean. The majority of these tracking algorithms apply the identiﬁcation criteria during the tracking which restricts the tracks to just the tropical cyclone deﬁne part. Therefore earlier and later parts of the tracks can be missing. The tropical cyclone tracking algorithm used in this study tracks all centres of high vorticity and applies additional criteria post tracking which is explained below.

2.6.1 TRACK

The method used in this study is an objective feature tracking methodology designed to be of general application and has been used to track extra-tropical cyclones, easterly waves, polar lows as well as tropical cyclones. The method is fully described in Hodges, 1995; Hodges, 1996; Hodges, 1999; Bengtsson et al., 2007b and Strachan et al., 2013. For initial identification and tracking, 850 hPa relative vorticity is computed at a spectral resolution of T42. The initial identification is made for vorticity maxima with intensities greater than 0.5 × 10−5 s−1 in the Northern Hemisphere (minima less than −0.5 × 10−5 s−1 in the Southern Hemisphere). The tropical cyclone features are identified using intensity thresholds and evidence of a warm core using the following criteria:

i. T63 relative vorticity at 850 hPa is larger than 6 × 10−5 s−1.

ii. A positive T63 vorticity centre must exist at 850, 500 and 200 hPa.

iii. There must be a minimum reduction in vorticity at T63 from 850 hPa to 200 hPa of 6 × 10−5 s−1, to provide evidence of a warm core.

iv. There must be a reduction in T63 vorticity with height between pressure levels of 850 hPa and 500 hPa, as well as, 500 hPa and 200 hPa.

v. Criteria i. to iv. must be attained for a minimum of 4×6 hourly time steps (one day).

Page 37 Chapter 2: Data, Method and Tools

The same criterion above also apply for tracking systems in the Southern Hemisphere, but with the values multiplied by -1. The choice of these criteria is discussed in Bengts- son et al. (2007b). The triplet 6,6,4 (maximum T63 intensity of relative vorticity at 850 hPa during the lifetime, a T63 vorticity maxima at each level up to 250 hPa and the difference in vorticity between 850 hPa and 250 hPa (850-250), for the number of 6-hourly time steps) used in the ECMWF operational analysis gave the best calibration to observed tropical cyclones. The feature points (i.e. vorticity maxima (minima) in the Northern Hemisphere (Southern Hemisphere) of the cyclone are identified and are initialised into tracks using a nearest neighbour approach to constrain the maximum distance between the feature points at consecutive timesteps. These are them improved by minimising a cost function for the ensemble track smoothness to obtain the minimal set of smoothest tracks. The minimisation is performed subject to adaptive constraints on the track smoothness and displacement distance (Hodges, 1999). This is done directly on the sphere to avoid the need to use projections which can introduce biases (Hodges, 1995). Systems with a lifetime of over two days are retained for further analysis. As discussed in Strachan et al. (2013), relative vorticity allows identification of smaller spatial scales than is possible with mean sea level pressure and hence earlier storm identification. The tracking is applied in the region 60o S to 60o N and uses 6- hourly temporal frequency. For the Northern (Southern) Hemisphere tropical cyclone season months May to November (October to May) are assessed. For initial identification and tracking, 850hPa relative vorticity is used on a standard spectral resolution of T42. The use of vorticity at T42 and T63 reduces noise and ensures that the identification and tracking methodology is as resolution independent as possible. The refined warm core criteria, which has been included in addition to the criteria used by Bengtsson et al. (2007b), helps exclude non-tropical-cyclone like features such as monsoon-like features that may be cold core in the lower troposphere and warm core aloft. Full resolution information, such as relative vorticity and 850 hPa wind speeds, are added back onto the remaining tracks for analysis.

Page 38 Chapter 2: Data, Method and Tools

Track density statistics are used to investigate the spatial distribution of tropical cyclones via the tracking algorithm. They are computed using spherical nonparametric kernel estimators which are described in Hodges (1996). Tropical cyclones from the analysis period are composited and monthly mean storm transits per unit area (equivalent to a 5o spherical cap, or approximately 106 km2) are calculated. The tracking of HiGEM and ERA-Interim identiﬁcation is based on the T42 vorticity. The tracking algorithm follows the life-cycle from ‘genesis’, when the initial T42 vorticity maxima threshold is achieved, through to ‘lysis’, when the T42 vorticity maxima threshold is no longer achieved (Strachan et al., 2013). Restricting the tracks to only those parts where the criteria apply to IBTrACS would remove valuable information, for example Easterly Wave precursors (Hopsch et al., 2009) and the extra-tropical transitions (Evans and Hart, 2003), often associated with tropical cyclones.

2.7 HiGEM: Model performance

2.7.1 Tropical cyclones in the present-day climate of HiGEM

It is important to evaluate the ability of HiGEM to simulate the present-day tropical cyclone climatology to provide a degree of confidence for the climate change study. A comparison of track densities from IBTrACS and reanalysis with HiGAM results, which are based on observed SSTs, reduces the complexity of understanding systematic errors in HiGEM due to inaccurate SST. Previous work has focussed on the ability of HiGAM to simulate present-day tropical cyclone climatology (Strachan et al., 2013). Strachan et al. (2013) found that HiGAM simulates slightly reduced tropical cyclone activity in the North Atlantic and North East Pacific basins compared to observations and tropical cyclones identified in several reanalyses. HiGAM also simulates a more even split in tropical cyclone numbers between hemispheres than observed, with for example almost twice as many tropical cyclones

Page 39 Chapter 2: Data, Method and Tools identified in the South Pacific. Figure 5.6 (a) shows the track density for the HiGEM present-day simulation, as well as the mean number of tropical cyclones simulated in each basin per season. HiGEM simulates slightly fewer tropical cyclones in the North Atlantic compared to HiGAM and observations (figure 2.3). It is noted that the model tropical cyclone track length is longer than typical observed track length, which is explained further in section 2.9.1. A lack of recurving tropical cyclones in the North West Pacific can also be seen in comparison to those simulated in HiGAM and identified in ERA-Interim (figure 2.3). Cold SST biases in HiGEM (which can be seen in figure 3 (a) of Shaffrey et al., 2009) may contribute to both of these systematic errors. HiGEM simulates the same number of tropical cyclones in the South Pacific as HiGAM, although their genesis is simulated more eastward. In addition, HiGEM simulates more tropical cyclones in the South Indian basin than HiGAM. A positive precipitation bias (which can be seen in figure 6 of Shaffrey et al., 2009) is partly explained by higher tropical cyclone counts in both basins compared to observations. Despite these differences, the general spatial distribution of tropical cyclones in HiGEM is in good agreement with the observed tropical cyclone distribution, especially in the Northern Hemisphere. Both HiGEM and ERA-Interim capture tropical cyclone- like features off the coast of South America, where only one storm has been recorded, Catarina in 2004. This is likely to be due to the tracking criteria which identifies warm- core storms. Strachan et al. (2013) found that relative humidity is the dominant factor in explaining interannual variations of identified storms in the South Atlantic. It can be seen that atmosphere-ocean coupling has little impact on the number of tropical cyclones simulated in each basin as the coupled HiGEM results look similar to the HiGAM-HiGEM time slice experiment (figure 2.3 (d) and figure 5.6 (a)).

Page 40 Chapter 2: Data, Method and Tools

Figure 2.3 Tropical cyclone track density (storm transits/month/106 km2 or equivalent to a 5o radius) during May to November in the northern hemisphere and October to May in the southern hemisphere for (a) IBTrACS, (b) ERA-Interim reanalysis, (c) HiGAM AMIPII simulation (Strachan et al., 2013) and (d) HiGAM forced with HiGEM SST. The numbers shown in each subdomain are the climatology annual count of tropical cyclones. Note: IBTrACS removes extratropical position.

The 150-year present-day simulation shows multi-decadal variability of tropical cyclones. Figure 2.4 shows that HiGEM is able to capture the interannual and inter-decadal variability associated with North Atlantic tropical storms, even though the mean frequency is underestimated. HiGEM simulates slightly less intense tropical cyclones than those simulated by HiGAM shown in ﬁgure 2.5 and ﬁgure 5.8, as HiGEM allows negative feedback mechanisms between the atmosphere and ocean to be simulated, for example those associated with wind-stress induced cold water upwelling (Scoccimarro et al., 2011). The differences between HiGEM and HiGAM could also be due to SST biases.

Page 41 Chapter 2: Data, Method and Tools

Figure 2.4 North Atlantic tropical cyclone count from BestTrack, left panel and the 150-year present-day HiGEM simulation. Thanks to Jane Strachan.

Figure 2.5 Normalized distributions of Northern Hemisphere extracted storm maximum intensities in terms of maximum 850-hPa relative vorticity from the hierarchy of GCM resolutions and reanalyses (Strachan et al., 2013).

Page 42 Chapter 2: Data, Method and Tools

2.7.2 Large-scale environmental climatology in HiGEM

This section utilises new observational products during a more recent time period and further evaluates the mean-state biases of HiGEM which were discussed in Roberts et al. (2009); Shaffrey et al. (2009) and Catto (2009). Deﬁciencies in the mean-state of HiGEM impact the simulation of the tropical cyclone climatology and the ENSO associated teleconnections. The biases also provide knowledge of associated uncertainties in the climate change experiments. The large-scale environmental parameters investigated are important for tropical cyclone activity as stated in section 1.2.1.

2.7.2.1 Sea surface temperature

Figure 2.6 shows the difference in HiGEM simulated SST during July-October (JASO) to HadISST and AMIP SST. The HiGEM SSTs are shown to be too cool by 1-2 oC throughout the tropics. The small area of positive SST bias in the upwelling region off the coast of Peru is partly due to errors in the low-lying stratocumulus cloud decks which are poorly captured in HiGEM (Shaffrey et al., 2009). This has a large implication on the simulation of ENSO and associated tropical cyclone teleconnections which is discussed in section 3.5.1. It is also known that HiGEM simulates too strong wind stress, in particular the equatorial Pacific and equatorial Atlantic (Shaffrey et al., 2009) which increases ocean mixing (Lu and Zhao, 2012). The cool SSTs in the tropical Atlantic relate to the lower number of tropical cyclones simulated in this basin compared to observations. The lack of recurvature of tropical cyclones in the North West Pacific (shown in figure 2.3) is attributed to the cold SST bias.

Page 43 Chapter 2: Data, Method and Tools

Figure 2.6 Sea surface temperature (oC), July-October for (a) HadISST 1979-2010 climatology, (b) AMIP 1979-2002 climatology, (c) HiGEM 150-year present-day climatology minus HadISST and (d) HiGEM minus AMIP.

2.7.2.2 Precipitation

The difference in simulated precipitation between HiGEM, HiGAM and GPCP during JASO is shown in figure 2.7. HiGEM shows there are large-scale errors in precipitation in the tropical cyclone basins. HiGEM simulates too much precipitation in the Western North Pacific compared to GPCP relating to a greater number of tropical cyclones simulated in this region compared to in IBTrACS (figure 2.3). A substantial dry bias can be seen over the North Indian Ocean which highlights the poor simulation of the monsoon (Shaffrey et al., 2009). The main development region in the tropical North Atlantic shows reduced precipitation to the west of Africa along with reduced tropical cyclone activity. HiGAM shows similar precipitation biases to HiGEM in the West Pacific. HiGAM has a positive precipitation biases in the Caribbean Sea, unlike HiGEM, which relates to the slightly enhanced tropical cyclone climatology. In the coupled model precipitation errors arise from absolute SST errors as well SST gradient errors. As for the atmosphere-only precipitation errors arise due to the imperfect atmospheric representation of convection. Precipitation is related to the product of upward velocity and low-level humidity which

Page 44 Chapter 2: Data, Method and Tools are discussed below.

Figure 2.7 Precipitation (mm day−1), July-October for (a) GPCP 1979-2010 climatology, (b) HiGAM 1979-2002 minus GPCP and (c) HiGEM 150-year present-day climatology minus GPCP.

2.7.2.3 Relative humidity

Figure 2.8 shows the difference in HiGEM and HiGAM simulated relative humidity at 700 hPa SST during July-October (JASO) to ERA-Interim. The biases are very similar in HiGEM and HiGAM, and similar to the biases in precipiation. HiGEM and HiGAM have a positive relative humidity bias to the west of Hawaii which favour tropical cyclone activity. Negative relative humidity biases can be seen in both HiGAM and HiGEM in the eastern tropical North Atlantic, however HiGAM shows a more moist environment in the Caribbean Sea. Both HiGEM and HiGAM simulate the Indian monsoon associated relative humidity poorly. The cool SST bias in the tropical North Atlantic is likely to explain the dry bias over the region along with the large vertical wind shear. It is unclear what causes the errors in the atmospheric model, HiGAM. The errors in relative humidity

Page 45 Chapter 2: Data, Method and Tools are in phase with the errors of precipitation in the tropical Paciﬁc suggesting they are related. However relative humidity and precipitation biases are of the opposite sign in the Gulf of Mexico in HiGAM which states the precipitation bias in this area is not soley caused by relative humidity.

Figure 2.8 Relative humidity at 700 hPa (%) July-October for (a) ERA-Interim 1979-2010, (b) HiGAM 1979-2002 minus ERA-Interim and (c) HiGEM 150-year present-day climatology minus GPCP.

2.7.2.4 Walker circulation

The biases in the simulation of the Walker circulation is shown in ﬁgure 2.9. HiGEM shows a stronger Walker circulation compared to HiGAM and ERA-Interim. The circulation is stronger in magnitude and the region of ascent over the warm pool is much nar- rower compared to in ERA-Interim, around 150oE. Both HiGAM and HiGEM simulate

Page 46 Chapter 2: Data, Method and Tools upper-level flow over the North Atlantic (300oE) too strong compared to ERA-Interim which relates to the increase in vertical wind shear discussed below. The stronger Walker circulation in HiGEM arises due do a larger tropical SST gradient compared to observations which was shown in figure 2.6. The cool bias in the east Pacific stregthens the east-west gradient across the tropical Pacific which is important for the Walker circulation.

Figure 2.9 Height-longitude cross section of Walker circulation, 0-10oN July-October for (a) ERA-Interim 1979-2010 climatology, (b) HiGAM 1979-2002 minus ERA-Interim and (c) HiGEM 150-year present-day climatology minus ERA-Interim. The colours show mean ascent (-ω) and the vectors are mean ascent and a change in the velocity potential with respect to latitude.

Page 47 Chapter 2: Data, Method and Tools

2.7.2.5 Vertical wind shear

The biases in the simulation of vertical wind shear have a large impact on the simulation of the tropical cyclone climatology. Figure 2.10 shows the difference of vertical wind shear in HiGAM and HiGEM compared to ERA-Interim during JASO. Vertical wind shear is too strong in the tropical North Atlantic by up to 10 m s−1, an increase of 100 % when comapred to ERA-Interim. The strong vertical wind shear in the North Atlantic is reduced in HiGAM explaining why more tropical cyclones are simulated than HiGEM. Shaffrey et al. (2009) noted that HiGEM has a 850 hPa westerly wind bias in the West- ern North Pacific which increases the vertical wind shear around 30oN and also limits tropical cyclone recurvature. Although both HiGEM and HiGAM simulated enhanced relative humidity to the west of Hawaii, the models have different levels of vertical wind shear in this region. HiGAM simulates too much vertical wind shear and HiGEM too little; this relates to the presence of tropical cyclones forming in the Western North Pa- cific much further east in HiGEM than HiGAM. The large SST gradient in the tropical Pacific strengthens the Walker circulation and increases the upper-level westerlies. This mechanism in discussed in section 3.5.4. In HiGAM, the biases in vertical wind shear are likely to be related to biases in precipitation. Vertical wind shear occurs in tropics as the circulation responds to regions of ascending and descending motion. It is also known that vertical wind shear can subsequenty influence vertical motion as the change in convection and circulation is coupled process in the tropics.

Page 48 Chapter 2: Data, Method and Tools

Figure 2.10 Vertical wind shear (m s−1), July-October for (a) ERA-Interim 1979-2010, (b) HiGAM 1979-2002 minus ERA-Interim and (c) HiGEM 150-year present-day climatology minus ERA-Interim.

2.7.2.6 Upper-level circulation

The upper-level circulation response is investigated using the velocity potential - an integrated measurement of the irrotational part of the upper level ﬂow; and the stream function - the rotational part. The stream function emphasises the anomalous wave propagation, whereas the velocity potential shows the forcing of these waves. Figure 2.11 shows the 200 hPa velocity potential and stream function during JASO for ERA-Interim and the differences in HiGEM and HiGAM. HiGEM shows enhanced upper-level convergence in the central Paciﬁc, relating to the increase in local tropical cyclone Activity. The stream function anomaly over the North Atlantic highlights the strong westerly wave

Page 49 Chapter 2: Data, Method and Tools activity from the tropical Paciﬁc acting to increase vertical wind shear. HiGAM shows enhanced upper-level convergence over the Caribbean Sea and anomalous wave forcing to the east. Many of the large-scale environmental conditions are still unfavourable in the North Atlantic in HiGAM to enhance tropical cyclone activity greater than observations.

Figure 2.11 The 200 hPa velocity potential (1×10−6 m2 s−1) in colours and the 200 hPa stream function (1×10−6 m2) in black contours, June-October for July-October for ERA-Interim 1979-2010 (top), HiGAM AMIP 1979-2002 minus ERA-Interim (middle) and HiGEM 150-year present-day simulation minus ERA-Interim (bottom).

Page 50 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.1 Introduction

Tropical cyclones can cause substantial loss of life and an improved understanding of storm variability can help inform preparation and response to landfalling events. Natural climate variability, for instance, the El Nino˜ Southern Oscillation (ENSO) has a large influence on global storm variability. This has been long observed e.g. Gray (1984) for the North Atlantic; Chan (1985) for the Western North Pacific and Nicholls (1979) for the Australian region). Due to recent advances in available computing resources, General Circulation Models (GCMs) can now be run with a high enough resolution to simulate different aspects of tropical cyclone activity (e.g. Zhao et al., 2009; Smith et al., 2010; Murakami et al., 2012b; Manganello et al., 2012; Strachan et al., 2013). Long integrations of these GCMs can be used to understand a robust response of tropical cyclones to the phase of ENSO beyond the traditional use of observations, as well as provide a platform to examine the dynamical and thermodynamical mechanisms. This chapter investigates the simulated response of global tropical cyclone activity to the phase of ENSO. The long integration of 150 years at present-day CO2 levels are used with the aim of obtaining a large sample size of ENSO events to increase the robustness of the simulated ENSO-tropical cyclone teleconnection. The present-day forcing removes further anthropogenic influence by holding greenhouse gases and aerosols constant and therefore a possible influence on a changing ENSO, such as its frequency or

Page 51 Chapter 3: The ENSO-Tropical Cyclone Teleconnection strength. An investigation of the mechanisms simulated in the AOGCM are compared to observations and reanalysis data. The ability of the atmospheric component (HiGAM) in an Atmospheric Model Intercomparison Project (AMIP) simulation to capture the expected ENSO-tropical cyclone teleconnection is also examined. The structure of this chapter will be as follows: the definition of ENSO is first discussed followed by a literature review on the expected ENSO-tropical cyclone teleconnection. The ability of the model to simulate the mean state and variability of ENSO is discussed in context of simulating the ENSO-tropical cyclone teleconnection. The results of changing tropical cyclone location and frequency to the phase of ENSO are shown, along with the changing large-scale environmental conditions. The final section discussed the advantages and shortcomings of the simulated ENSO-tropical cyclone teleconnection in HiGEM and HiGAM. The results are lastly summarised along with concluding remarks. Work in this chapter has been published in Bell et al. (2013b).

3.1.1 El Nino˜ Southern Oscillation

The El Nino˜ Southern Oscillation (ENSO) is the term given to a global weather/climate phenomenon which occurs across the tropical Pacific basin approximately every 4-5 years and usually lasts for 1-2 years (Burgers and Stephenson, 1999). There has been an extensive amount of research into ENSO and in depth review papers are given in Chunzai and Picaut (2004); Guilyardi et al. (2009); Wang et al. (2012), as well as a summary of the strong 1998/1999 El Nino˜ by McPhaden (1999). ENSO is a coupled phenomenon which is associated with the Southern Oscillation Index - the varying difference in atmospheric pressure between Darwin, Australia and Tahiti, Hawaii (Walker, 1924) and tropical Pacific SST (Bjerknes, 1966). A schematic of the coupled phenomenon is shown in figure 3.1. The mean-state (neutral conditions) in the tropical Pacific consists of warm water in the Western Pacific (known as the Western Equatorial Warm Pool) and cooler water in the East Pacific along the Peruvian coast. This cooler water occurs due to northerly wind stress causing coastal upwelling which

Page 52 Chapter 3: The ENSO-Tropical Cyclone Teleconnection tilts the thermocline so it is higher in the east Paciﬁc. The SST gradient is maintained by the Walker Circulation (see section 1.3.3) as warmer SST in the west increases convection and precipitation whereas the cooler water in the east causing the atmosphere to be drier. This drives upper-level westerly winds and surface easterlies which maintains the temperature gradient. El Nino˜ is deﬁned as a weakening of this SST gradient. One mechanism which

Figure 3.1 A schematic showing the atmospheric circulation and tilt of the thermocline during: El Nino˜ (left) and La Nina˜ (right) (Tropical Atmosphere Ocean project, 2000). initiates El Nino˜ are Westerly Wind Bursts which can slacken the easterly trade winds. The upwelling in the east Pacific is reduced and the thermocline becomes more flat. The atmospheric convection adjusts and the Walker circulation weakens further weakening the easterly trade winds (Wyrtki, 1975; Kiladis and Diaz, 1989; Glantz et al., 1991). The schematic in figure 3.2 shows how the pattern of precipiation changes. The counterpart of El Nino˜ is known as La Nina,˜ where anomalous cold waters are found in the eastern tropical Pacific. This is related to a larger tilt of the thermocline. The Walker circulation stregthens in this case which enchances the easterly trade winds and in turn stregthens the SST gradient. ENSO can affect parameters globally by means of atmospheric processes, usually operating temporally on the order of a few months and oceanic processes which can take years to decades to affect other basins (Alexander et al., 2002; Wu and Hsieh, 2004; Liu and Alexander, 2007). These are known as teleconnections.

Page 53 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.2 Schematic of ENSO teleconnections (NOAA, 2012).

Page 54 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.1.1.1 Measurements of ENSO

ENSO is defined mainly by the normalised NINO3.4 SST anomaly (in the area 5oS-5oN, 120oW-170oW, shown in figure 3.3). El Nino˜ is defined when the normalized Nino-˜ 3.4 SST anomalies have an amplitude of greater than 1, a La Nina˜ event when the normalized Nino-3.4˜ SST anomalies have an amplitude of less than -1 for the December- January-February (DJF) period. Observed El Nino˜ and La Nina˜ years using the Ex- tended Reconstructed SST (ERSST.v3b) can found at http://www.cpc.ncep.noaa. gov/products/analysis monitoring/ensostuff/ensoyears.shtml. The El Nino˜ and La Nina˜ events since 1979 - which is the start of the AMIP period - are shown below in table 3.1.

Figure 3.3 A map of NINO regions, which are commonly studied (Golden Gate Weather Ser- vices, 2000).

Page 55 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

El Nino˜ years SSTa La Nina˜ years SSTa

1982/1983 2.2 1986/1987 1.2 1988/1989 -1.7 1991/1992 1.6 1994/1995 1 1997/1998 2.2 1998/1999 -1.5 1999/2000 -1.7 2002/2003 1.1 2007/2008 -1.5 2009/2010 1.6

Table 3.1 Observed El Nino˜ and La Nina˜ years given by normalised NINO3.4 DJF SST anomalies greater than 1 or less than -1 since 1979 (http://www.cpc.ncep.noaa.gov/ products/analysis monitoring/ensostuff/ensoyears.shtml).

Page 56 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.1.1.2 Construction of ENSO composites

ENSO events are defined using the normalized Nino-3.4˜ SST anomalies (in the area 120oW-170oW, 5oS-5oN) which had an amplitude of greater than 1 or less than -1 for the December-January-February (DJF) period. Tropical cyclone seasons in the Northern Hemisphere (May-November) are defined prior to an ENSO event. The SSTs in the peak of the Northern Hemisphere tropical cyclone season, August-September-October (ASO), are usually reflective of the DJF SST anomalies. Southern Hemisphere tropical cyclone seasons (October-May) were selected during the ENSO event. The seven El Nino˜ events of 1982/83, 1986/87, 1991/92, 1994/95, 1997/98, 2002/03 and 2009/10 and the six La Nina˜ events of 1984/85, 1988/89, 1998/99, 1999/00, 2007/08, 2010/11 are composited and compared with the 1979-2010 observed climatology. The ENSO events within the time period 1979-2002 are composited in the AGCM, HiGAM dataset and compared to its climatology. The AOGCM, 150-year integration of HiGEM simulates 31 El Nino˜ and 25 La Nina˜ events, which are compared with the 150-year climatology. Composites are used to increase robustness of the ENSO-tropical cyclone teleconnection.

3.2 The ENSO-tropical cyclone teleconnection: A review

An association between ENSO and the incidence of Atlantic basin tropical storms was ﬁrst observed by Gray (1984). The long standing record of North Atlantic tropical storms allows for reliable statistics of tropical cyclone activity between the two phases of ENSO to be calculated. Although there are known biases in the long-term record in particular before satellites (see section 2.4.1). It is observed that approximately two times more tropical cyclones occur in La Nina˜ years as opposed to El Nino˜ years, as well as three times more hurricanes (Landsea et al., 1999). Tropical cyclones also become more intense during La Nina˜ years and there is an increase in the likelihood of landfall (Pielke and Landsea, 1999; Xie et al., 2002a; Smith et al., 2007; Klotzbach,

Page 57 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

2011). During an El Nino˜ event warm SSTs in the Pacific lead to a warming of the free troposphere. The anomalously warm temperature is transported across the tropics via upper tropospheric equatorial wave dynamics (Tang and Neelin, 2004). In the North Atlantic, in particular, the increase in upper level temperatures can decrease the lapse rate and make the atmosphere more stable. This leads to a reduction in the number of tropical cyclones which can form per season (Knaff, 1997; Vecchi and Soden, 2007a). In addition, a weakening of the Walker circulation leads to an increase in vertical wind shear over the North Atlantic, as the upper branch of the Walker circulation extends into the Caribbean. This both reduces the number of tropical cyclones and influences the location of genesis (e.g. Gray and Sheaffer, 1991; Goldenberg and Shapiro, 1996; Kossin et al., 2010). The presence of El Nino˜ conditions is shown to be associated with a weakening of the North Atlantic subtropical high, an increase in the percentage of recurving ocean tropical cyclones, and a decrease in the percentage of recurving landfalling tropical cyclones (Colbert and Soden, 2011). Shaman et al. (2009) also note the importance of Rossby waves influencing upper-level vorticity over the tropical Atlantic, suppressing tropical cyclogenesis during El Nino˜ years. Camargo et al. (2007c) found relative humidity and vertical wind shear are important for the reduction in genesis seen in the North Atlantic using the genesis potential index (GPI) on the National Centers for Environmental Prediction (NCEP) reanalysis data (Kalnay et al., 1996) from 1950-2004. ENSO can also influence North Atlantic tropical storm activity indirectly. ENSO has a noticeable influence on the African Easterly Wave cycle, as well as the Atlantic Meridional Mode, a mode of variability which represents the meridional temperature gradient and related to North Atlantic tropical cyclone activity (Bell and Chelliah, 2006; Kossin et al., 2010; Maue, 2011). The Western North Pacific is the basin which has the second longest standing records of tropical storm counts, going back to 1945. There is a well known relationship: as the warm tropical Pacific SSTs shift to the east during El Nino˜ years, so does the genesis location. This is thought to be connected with an eastward expansion of the

Page 58 Chapter 3: The ENSO-Tropical Cyclone Teleconnection monsoon trough and westerlies, which increases low level vorticity (Wang and Chan, 2002). However, Camargo et al. (2007c) argue a decrease in mid-level humidity near the Asian continent during El Nino˜ years suppresses activity. There is in general more tropical cyclones during El Nino˜ years than La Nina˜ years (Chan and Liu, 2004; Chan and Liu, 2012), although this is masked by multi-decadal variability (ﬁgure 3.4). The eastward genesis of tropical cyclones with El Nino˜ is also associated with an increase in intensity as the tropical cyclones are usually longer lived and are also kept at warmer lower latitudes for longer (Saunders et al., 2000; Elsner and Liu, 2003; Wu et al., 2004; Chan, 2007). It is also worth mentioning studies which relate the lead-lag relationship of ENSO on Western North Paciﬁc typhoons (Li and Zhou, 2012; Ha et al., 2012).

Figure 3.4 Tropical cyclone numbers in the Western North Paciﬁc. The arrows qualatatively show the trend of tropical cyclone numbers (Chan and Liu, 2012).

Tropical cyclone statistics in the North East Pacific show a see-saw of activity in opposite phase to how ENSO influences tropical cyclone activity in the North Atlantic (Frank and Young, 2007). Tropical cyclone activity is enhanced in this basin during El Nino˜ years and reduced during La Nina˜ years (Gray, 1984). However, other studies show ENSO has no influence on North East Pacific tropical cyclone frequency. Tropical cyclone genesis shifts westward during El Nino,˜ which can be seen in figure 3.5. This is related to a weaker Walker circulation which occurs during El Nino˜ years and increases

Page 59 Chapter 3: The ENSO-Tropical Cyclone Teleconnection the vertical wind shear. The region of enhanced vertical wind shear often extends into the North East Pacific from the Caribbean (Camargo et al., 2007c). Gray and Sheaffer (1991) noticed that tropical cyclones are more intense in El Nino˜ years, likely attributed to warmer local SSTs which increase the potential intensity (Camargo et al., 2007c). The ENSO-tropical cyclone teleconnection has not been well studied in the North In- dian Ocean basin mainly due to poor tropical cyclone observations in this region. Singh et al. (2000) found fewer tropical cyclones in the May-November period in the Bay of Bengal during El Nino˜ years. Chiang and Sobel (2002) noted during El Nino˜ years an increase in static stability over the North Indian Ocean relates to a reduction of tropical cyclone frequency. Similarly, tropical cyclone observational records in the Southern Hemisphere are limited, resulting in a lack of understanding of how ENSO influences tropical cyclone activity. Nicholls (1979) first investigated the ENSO-tropical cyclone teleconnection around the Australian region. The South Pacific shows a pronounced shift of tropical cyclone activity with fewer tropical cyclones between 145oE and 165oE and more from 165oE eastward during El Nino˜ events. There is also a smaller tendency for tropical cyclones to originate closer to the equator. The opposite is true in La Nina˜ (Nicholls, 1992; Kuleshov et al., 2008). There is an increased chance of landfall on the east coast of Australia during La Nina˜ events as tropical cyclones form close to the coast (Evans and Allan, 1992). The number of tropical cyclones has been found to increase with El Nino˜ (Basher, R. Zheng, 1995). Low-level relative vorticity and vertical wind shear also play roles in determining where tropical cyclones form dependent on the phase of ENSO (Ramsay et al., 2008). In addition, Chand et al. (2013) recently found that an eastward extension of low vertical wind shear, increased relative humidity and SST explains the enhancement of tropical cyclone numbers during El Nino˜ years in that region. Recent work by Diamond et al. (2012) categorize the type of ENSO event based on the strength of ocean-atmosphere coupling and its impact on tropical cyclone activity in the Southwest Pacific. Diamond et al. (2012) found that tropical cyclones west of 170oE have a strong relationship to SST and tropical cyclones east of 170oE have a closer connection to atmospheric circulation.

Page 60 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

In the Southern Indian Ocean, Ho et al. (2006) found that tropical cyclone genesis is shifted westward during El Nino˜ events, enhancing cyclogenesis west of 75oE. Ho et al. (2006) explain this shift due to changes in the Walker circulation, which leads to anomalous anticyclonic low-level circulation in the eastern part of the South Indian Ocean during El Nino˜ years, reducing activity. Kuleshov et al. (2008) found that there is a decrease in activity south east of Madagascar and a moderate increase in activity in the central subtropical South Indian Ocean during El Nino˜ events. Kuleshov (2003) noticed how ENSO influences the intra-seasonal cycle of tropical cyclone activity in the South Indian basin. During El Nino˜ years the most intense tropical cyclone activity is extended to the end of February and early March, whereas in La Nina˜ years the peak of activity is found in January. Tropical cyclone tracks tend to be more zonal and frequent in La Nina˜ years, attributed to an exceptional number of tropical cyclones making landfall in Mozambique in 2000 (Vitart et al., 2003). Vitart et al. (2003) showed a high resolution AOGCM was able to simulate this relationship. Xie et al. (2002b) found an association between local SST influenced by ENSO which affects tropical cyclone activity. Tropical cyclone landfalls in Indonesia are more likely in El Nino˜ years, even though large-scale precipitation is suppressed during El Nino˜ events (Camargo et al., 2010), but the landfalling events are still rare. A summary of how ENSO influences tropical cyclone activity in cyclone basins is given in table 3.2.

Page 61 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Cyclone basin El Nino˜ La Nina˜ Frequency Intensity Frequency Intensity North Atlantic large decrease small decrease small increase small increase North West Pacific small increase slight increase small decrease slight decrease North East Pacific small increase increase slight decrease no change North Indian small decrease ? small increase ? South Pacific slight increase ? decrease ? South Indian small decrease ? small increase ?

Table 3.2 Summary of the ENSO-tropical cyclone teleconnection adapted from http: //cawcr.gov.au/bmrc/pubs/tcguide/ch5/ch5 2.htm. A ? denotes the relationship has not been studied in greatdepth. A summary of location changes can be found in ﬁgure 3.5.

Whilst the ENSO-tropical cyclone relationship is reasonably well known within each basin, there has only been one study that investigates the global ENSO-tropical cyclone teleconnection. Camargo et al. (2007c) examined how different environmental factors contribute to the ENSO-tropical cyclone teleconnection using the GPI developed by Emanuel and Nolan (2004) on the National Centers for Environmental Prediction (NCEP) reanalysis data (Kalnay et al., 1996) from 1950-2004 (shown in ﬁgure 3.5). Camargo et al. (2007c) found in El Nino˜ years, relative humidity and vertical wind shear are important for the reduction in genesis seen in the Atlantic basin, and relative humidity and vorticity are important for the eastwards shift in the mean genesis location in the Western North Paciﬁc.

Page 62 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.5 GPI anomalies between El Nino˜ years minus climatology (left) and La Nina˜ years minus climatology (right). The Northern Hemisphere is August-September-October and Southern Hemisphere is January-February-March. Adapted from Camargo et al. (2007c).

The ENSO-tropical cyclone teleconnection has also received little attention in GCM studies compared to research on tropical cyclones and climate change. Wu and Lau (1992) were the first to investigate the global ENSO-tropical cyclone teleconnection using a very coarse Atmosphere-only GCM (AGCM) with resolution R15 (7.5o × 4.5o). Vitart and Anderson (2001) used a 10 member ensemble AGCM at T42 (2.8o × 2.8o) to investigate the ENSO-tropical cyclone teleconnection in the North Atlantic. Vitart and Anderson (2001) were able to simulate the expected tropical cyclone response with the phase of ENSO due to simulated changes in vertical wind shear. Whilst the ensemble approach is useful, Vitart and Anderson (2001) only investigated one El Nino˜ event and one La Nina˜ event. More recently, Murakami and Wang (2010) showed a 20-km resolution AGCM was able to capture the broad tropical cyclone response to ENSO in the North Atlantic although they did not comment on which simulated parameters were important. AGCM experiments are limited by short integration lengths which makes it difficult to assess the robustness of the simulated ENSO-tropical cyclone teleconnection. In addition, AGCMs are forced with observed SSTs which have signatures of time-varying radiative forcing including that from aerosols and greenhouse gases. This makes it difficult to isolate the simulated ENSO-tropical cyclone teleconnection from these experiments. In-

Page 63 Chapter 3: The ENSO-Tropical Cyclone Teleconnection vestigating the ENSO-tropical cyclone teleconnection using AOGCMs has received little attention due to the larger computational costs involved. In addition, the complexity of understanding tropical cyclone genesis and moving processes has limited progress on understanding the ENSO-tropical cyclone teleconnection. Shaman and Maloney (2011) investigated the ability of the Coupled Model Inter- comparison Project 3 (CMIP3) Atmosphere-Ocean GCMs (AOGCMs) to simulate the expected large-scale environmental conditions associated with ENSO, which are important for tropical cyclones, over the North Atlantic. Shaman and Maloney (2011) found the impacts of ENSO on Caribbean vertical wind shear were the most poorly simulated. However, the GCMs used in Shaman and Maloney (2011) had coarse horizontal resolution (2o × 2o). These coarse resolution CMIP3 models have an inaccurate representation of ENSO itself and ENSO-associated teleconnections (Guilyardi et al., 2009). In addition, coarse resolution models simulate a poor tropical cyclone climatology (Murakami and Sugi, 2010). Whereas, Higher-resolution GCMs have shown to greatly improve the simulation of ENSO variability (e.g. Shaffrey et al., 2009; Delworth et al., 2012), tropical cyclone climatology (e.g. Strachan et al., 2013) and ENSO-associated teleconnections (e.g. Dawson et al., 2013). Constant present-day forcing experiments with an AOGCM can be used to assess how well a model is able to simulate modes of natural variability and associated teleconnections. Iizuka and Matsuura (2008) use a high resolution AOGCM with present- day forcing integrated for 100 years to investigate the simulated ENSO-tropical cyclone teleconnection in the Western North Paciﬁc; Iizuka and Matsuura (2009) for the North Atlantic. Both studies show the AOGCM is able to capture the expected tropical cyclone response to ENSO. However Iizuka and Matsuura (2008); Iizuka and Matsuura (2009) fail to discuss if the AOGCM is capturing the tropical cyclone response due to the expected mechanisms. The results in the North Atlantic by Iizuka and Matsuura (2009) show a spurious trend of increasing tropical cyclone frequency throughout the simulation reducing the credibility of their ﬁndings. Kim et al. (2013) show the simulation of the global ENSO-tropical cyclone teleconnection in a present-day experiment using the

Page 64 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Geophysical Fluid Dynamics Laboratory Coupled Model 2.5 (GFDL CM2.5) however do not comment on the simulated mechanisms as their paper focusses on climate change results. It is important to investigate this phenomenon in more than one GCM to improve understanding of the simulated ENSO-tropical cyclone teleconnection.

3.3 ENSO simulation in HiGEM

The ability of HiGEM to simulate a realistic ENSO in terms of its amplitude, variability, spatial structure and associated teleconnections is discussed in Shaffrey et al. (2009). HiGEM simulates Nino-3˜ (90oW-150oW, 5oS-5oN) SST anomalies of a similar variability to those observed in HadISST, with a standard deviation of 0.89 K compared to a standard deviation of 0.84 K in HadISST. Figure 3.6 shows a composite of the eight largest El Ninos˜ during December-January-February (DJF). The ENSO-SST simulated in HiGEM has a spatial pattern and amplitude that is much closer to observed than in previous lower resolution versions of the model. However, the El Nino˜ SSTs still extend too far into the western tropical Pacific, which is a common failing of most climate models (Guilyardi et al., 2009). The precipitation response over the west Pacific does not move as far eastward into the central Pacific as it does in the observations. HiGEM also simulates a more meridional shift in precipitation in the central Pacific, whereas observations show a more zonal shift. HiGEM has some skill in replicating the observed transitions of the Walker Circulation, for example capturing the precipitation response over the Indian Ocean. HiGEM successfully captures the deepening of the Aleutian low and the response over the Eurasian sector during El Nino˜ events. However, the deepening of the Aleutian low in HiGEM occurs to the west of that in observations, which may be related to the rainfall not moving as far eastward into the central tropical Pacific (Spencer and Slingo, 2003; Shaffrey et al., 2009). More recently, Dawson et al. (2013) investigated the ability of HiGEM to capture the expected extratropical teleconnection associated with ENSO. The upper level vorticity response is captured by HiGEM in the

Page 65 Chapter 3: The ENSO-Tropical Cyclone Teleconnection extratropical Paciﬁc, although is slightly shifted westward compared to observations.

Figure 3.6 El Nino˜ DJF composite anomalies for SST (K) and precipitation (mm day−1) from (a) the HadISST SST dataset and (b) the CMAP precipitation dataset and from (c), (d) HiGEM1.2 and (e), (f) HadGEM1.2 (Shaffrey et al., 2009).

3.4 The ENSO-tropical cyclone teleconnection in HiGEM

The results in this section focus on the simulated response of global tropical cyclone activity during El Nino˜ and La Nina˜ years. The differences in tropical cyclone location and frequency are shown for both HiGAM and HiGEM, with comparisons to tropical cyclones identiﬁed in ERA-Interim and those observed in IBTrACS.

Page 66 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.4.1 ENSO and tropical cyclone location

Tropical cyclone track densities of El Nino˜ and La Nina˜ years minus the climatology are shown in figure 3.7. The tropical cyclone location changes in IBTrACS show an increase in tropical cyclones towards the dateline in the Western North Pacific as found in Wang and Chan (2002); Camargo et al. (2007c). Tropical cyclones are suppressed in the North Atlantic during El Nino˜ years and enhanced during La Nina˜ years, a well known response (Gray and Sheaffer, 1991; Goldenberg and Shapiro, 1996; Kossin et al., 2010). The increase in tropical cyclone activity in the Bay of Bengal during La Nina˜ year has also been found by Felton et al. (2013). In the Southern Hemisphere there is a reduction of tropical cyclones to the west of Australia during El Nino˜ events. The opposite is true for La Nina˜ events. Tropical cyclones form closer to the east coast of Australia in the South Pacific during El Nino˜ years and further offshore during La Nina˜ years which is discussed in Kuleshov et al. (2008). There are some differences of tropical cyclone location changes in IBTrACS to those tracked in ERA-Interim. This is in part due to the tracking algorithm applied to ERA- Interim as we track storms from their genesis to lysis via T42 vorticity, whereas IBTrACS is based on observations of near-surface sustained wind speed (see Strachan et al., 2013). However, when focusing on the tropics it can be seen that that the tropical cyclones tracked in ERA-Interim match very closely to those in IBTrACS. This provides increased confidence in the tracking algorithm used. The tropical cyclone location changes in ERA- Interim obtained by using an explicit tracking algorithm are similar to the GPI changes in Camargo et al. (2007c). However, the large reduction of tropical cyclone activity in the North Atlantic during El Nino˜ years, which is similar to observations, does not correspond with the change in GPI shown in Camargo et al. (2007c). Similarly, Murakami and Wang (2010) found simulated tropical cyclone changes with ENSO did not correspond with GPI changes in this basin. HiGAM captures the response of tropical cyclone location to ENSO in the Pacific and Indian Oceans. In the North Atlantic, variability is confined to the Caribbean due

Page 67 Chapter 3: The ENSO-Tropical Cyclone Teleconnection to biases in the tropical cyclone climatology (Strachan et al., 2013). The variability is over pronounced in the Western North Pacific in HiGAM compared with observations and ERA-Interim. HiGEM is able to capture broadly the shift in tropical cyclones in the Pacific and Indian Oceans. However, the tropical cyclones show a shift in location which is more meridional as opposed to the zonal shift in observations in the South Pacific. This is likely to be related to the meridional shift of ENSO associated precipitation in HiGEM (see figure 20. of Shaffrey et al., 2009). HiGEM is unable to capture the expected response of tropical cyclone location changes in the North Atlantic, with a small tendency to simulate more tropical cyclones during El Nino˜ years, the opposite to observed. There are known biases of tropical cyclones in the mean-state in this basin (see chapter 2) which likely limit the expected ENSO variability being captured and are discussed later. However, this limitation is not present in all AOGCMs. The Geophysical Fluid Dynam- ics Laboratory Coupled Model 2.5 (GFDL CM2.5; Delworth et al., 2012), which has an atmospheric resolution of 50 km, captures the expected sign of the tropical cyclone response to ENSO in the North Atlantic, even though the model similarly simulates a smaller tropical cyclone climatology than observed (Kim et al., 2013). The higher atmospheric resolution is likely to be key in capturing this, just as Strachan et al. (2013) found atmospheric resolution to be important for a better representation of interannual variability in the North Atlantic which mainly came from the improvement in interannual variability of vertical wind shear. In addition, Murakami and Wang (2010) was able to capture the ENSO-tropical cyclone teleconnection in the North Atlantic using an AGCM with 20 km resolution. The statistical significance is shown on figure 3.7 using a student’s t-test. Tropical cyclone changes in the Pacific are shown to significant in all datasets. Tropical cyclones changes in the North Atlantic with ENSO are only significant in ERA-Interim. Work using the non-parametric Monte-Carlo method, which requires no assumption about the underlying data, developed by Hodges (2008) proved unsuccessful in this case. This method works by re-sampling without replacement the track datasets. The process ran-

Page 68 Chapter 3: The ENSO-Tropical Cyclone Teleconnection domly samples the individual cyclone tracks to generate 2000 new pairs of track datasets. The statistics are then calculated for each of these new sets and the differences between the statistics of each new pair is then found. The issue may have been related to compar- ing a small sample size (e.g. El Nino˜ years) to a large sample size (climatology).

Page 69 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.7 Tropical cyclone track density (storm transits/month/106 km2 or equivalent to a 5o radius) during May to November in the northern hemisphere and October to May in the southern hemisphere for IBTrACS: (a) El Nino˜ years minus 1979-2010 climatology; (b) La Nina˜ years minus 1979-2010 climatology; ERA-Interim: (c) El Nino˜ years minus 1979-2010 climatology; (d) La Nina˜ years minus 1979-2010 climatology; HiGAM: (e) El Nino˜ years minus 1979-2002 climatology; (f) La Nina˜ years minus 1979-2002 climatology; HiGEM: (g) El Nino˜ years minus 150-year climatology; (h) La Nina˜ years minus 150-year climatology. Stippling shows where changes have a p-value < 0.05 using a student’s t-test. Note: Stippling is not shown for IBTrACS.

Page 70 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.4.2 ENSO and tropical cyclone frequency

The effect of ENSO on tropical cyclone counts in each basin is shown in figure 3.8. The average number of tropical cyclones which form in each basin is given in the text below the percentage change. The North Atlantic shows the greatest response of tropical cyclone frequency with ENSO. Tropical cyclones are less frequent in El Nino˜ years by 35 % and more frequent in La Nina˜ years by 18 %. Klotzbach (2010) found a more symetri- cal respone of tropical cyclone numbers in El Nino˜ and La Nina˜ years using a longer time period of 1900-2009. However, there are uncertainties in the historical tropical cyclone record (Landsea and Franklin, 2013) and historical tropical Pacific SSTs (Solomon and Newman, 2012) especially in the early twentieth-century. Slightly more tropical cyclones form in El Nino˜ years than La Nina˜ years in the Western North Pacific (Chan and Liu, 2004). There is no change in the observed number of tropical cyclones which form each year in the North Indian basin with ENSO. There are some notable differences between tropical cyclone frequency changes in ERA-Interim and IBTrACS. The variability of tropical cyclones in response to ENSO are much more pronounced in ERA-Interim compared with IBTrACS in the North Indian Ocean. Tropical cyclones are 18 % less frequent during La Nina˜ years in the North In- dian Ocean in ERA-Interim, whereas in IBTrACS there is no change in the number of tropical cyclones in El Nino˜ or La Nina˜ years. However, the observations in the North Indian Ocean have a large uncertainty shown by the confidence intervals. It should be noted that the tracking algorithm may pick up monsoon depressions in this region similar to other tracking algorithms. ERA-Interim shows greater percentage variability of tropical cyclones in response to ENSO in the North Atlantic and Western North Pacific compared with IBTrACS. HiGAM simulates a much stronger response of tropical cyclone frequency to ENSO in the North Indian Ocean than observed. However, the response is associated with large uncertainty due to the limited sample size of ENSO events. HiGAM is also able to capture the expected response in the North Atlantic although the interannual variability is

Page 71 Chapter 3: The ENSO-Tropical Cyclone Teleconnection large shown by the confidence intervals. HiGAM does not simulate the observed magnitude of change of tropical cyclone frequency in the North Atlantic with a 10 % reduction in the number of tropical cyclones that form compared to a 35 % reduction shown by IBTrACS during El Nino˜ years. The tropical cyclone frequency response in the South Indian Ocean is smaller in HiGEM than in HiGAM and therefore simulates a response similar to that observed. Although HiGEM captures the shift of tropical cyclones in the Pacific region to the phase of ENSO, the amplitude of the tropical cyclone frequency changes is much less than observed. One outstanding difference between HiGEM and HiGAM is that HiGEM simulates the response of tropical cyclone frequency change in the North Atlantic with ENSO of the opposite sign to HiGAM and that expected, although HiGAM simulates large variability in the number of tropical cyclones per season during El Nino˜ and La Nina˜ years shown by the large confidence intervals.

3.5 ENSO and large-scale environmental conditions

The number of tropical cyclones which form each year and in each basin depends largely on the large-scale environment (Camargo et al., 2007b; Camargo et al., 2007c). The results in this section investigate the changing large-scale environmental conditions associated with ENSO. Variables considered include SST, precipitation, vertical wind shear and tropical circulation.

Page 72 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.8 Percentage change of tropical cyclone counts in El Nino˜ years and La Nina˜ years compared to climatology: (a) IBTrACS (1979-2010), (b) ERA-Interim (1979-2010), (c) HiGAM (1979-2002) and (d) HiGEM. The climatology is shown at the bottom of the x-axis label. Error bars denote the 90% conﬁdence interval.

Page 73 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.5.1 Sea surface temperature

Figure 3.9 shows the change in SST during July-October (JASO), the peak of the North- ern Hemisphere tropical cyclone season, for El Nino˜ and La Nina˜ years compared to the climatology of HadISST, AMIPII SST and HiGEM. The warmest SST associated with ENSO during JASO can be seen to extend too far into the western tropical Pacific in HiGEM. In addition, the meridional extent of the ENSO associated SST is wider in HiGEM than in HadISST which is associated with biases in the simulated atmospheric response. This may be responsible for the large response of tropical cyclones in the central Pacific to ENSO which is not present in observations or ERA-Interim seen in figure 1.1. However, the results from HiGAM indicate than poor simulated spatial patterns of SST variability in HiGEM do not explain all of HiGEM’s deficiencies at simulating the expected ENSO-tropical cyclone teleconnection.

Page 74 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.9 Sea surface temperature (oC), July-October for (a) HadISST El Nino˜ years minus 1979-2010 climatology, (b) HadISST La Nina˜ years minus 1979-2010 climatology, (c) AMIP SST El Nino˜ years minus 1979-2002 climatology, (d) AMIP La Nina˜ years minus 1979-2002 climatology (e) HiGEM El Nino˜ years minus 150-year climatology and (f) HiGEM La Nina˜ years minus 150-year climatology. The climatology is shown in black contours for HadISST in (a) and (b), AMIP SST in (c) and (d) and in HiGEM for (e) and (f). Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

3.5.2 Precipitation

The precipitation response to ENSO is a good indicator of the atmospheric teleconnections in the tropics (Alexander et al., 2002). Examining the accuracy of simulated ENSO associated precipitation provides a test bed for comparison of modeled to observed precipitation (Langenbrunner and Neelin, 2013). Figure 3.10 shows JASO precipitation changes in El Nino˜ and La Nina˜ years compared to the climatology for GPCP, HiGAM and HiGEM. The climatology for each dataset is shown in black contors. Tropical cy-

Page 75 Chapter 3: The ENSO-Tropical Cyclone Teleconnection clone activity follows large-scale changes in precipitation with ENSO (shown in figure 1.1). HiGAM is able to simulate the expected precipitation response across the tropics; however, the magnitude of the response is larger in HiGAM than seen in GPCP. HiGEM captures the shift of precipitation into the central Pacific and the suppression of precipitation around the Maritime continent during El Nino˜ years. However, the ENSO- precipitation response is too strong in HiGEM over the Indian Ocean, as noted by Shaf- frey et al. (2009). During El Nino˜ years GPCP shows that precipitation is suppressed over the Caribbean and enhanced during La Nina˜ years. While HiGAM captures this variability, HiGEM simulates no variability. The lack of precipitation in the mean-state of HiGEM in the Caribbean also contributes to the poor simulation of the teleconnection (see figure 6; Shaffrey et al., 2009).

Page 76 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.10 Precipitation (mm day−1), July-October for (a) GPCP El Nino˜ years minus 1979- 2010 climatology, (b) GPCP La Nina˜ years minus 1979-2010 climatology, (c) HiGEM El Nino˜ years minus 1979-2002 climatology and (d) HiGAM La Nina˜ years minus 1979-2002 climatology, (e) HiGEM El Nino˜ years minus 150-year climatology and (f) HiGEM La Nina˜ years minus 150-year climatology. The climatology is shown in black contours for GPCP: (a) and (b); HiGAM: (c) and (d) and HiGEM: (e) and (f). Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

3.5.3 Walker circulation

The Walker circulation response to ENSO acts as an ‘atmospheric bridge’ (Alexander et al., 2002) for tropical cyclone activity to respond in basins away from the Paciﬁc. Figure 3.11 shows the Walker circulation for El Nino˜ and La Nina˜ years compared to climatology in ERA-Interim, HiGAM and HiGEM. The Walker circulation inﬂuences regions of large-scale motion as well as shifts the location of vertical wind shear patterns which modulate tropical activity (Kossin et al., 2010). The anomalously warm SSTs

Page 77 Chapter 3: The ENSO-Tropical Cyclone Teleconnection in the central Pacific during El Nino˜ years favors convection, which drives the large- scale tropical atmospheric circulation. The descending motion over the Indian Ocean, at around 90oE, is too strong in both HiGAM and HiGEM, leading to greater ENSO-tropical cyclone variability. The region of maximum ascent during El Nino˜ years in the central Pacific is constrained to around 160oE in HiGEM, whereas in ERA-Interim the region of maximum ascent occurs over a broad region from 160oE to 190oE. Similarly, during the La Nina˜ phase, HiGEM shows enhanced subsidence throughout the central Pacific east of 180oE which was not found in ERA-Interim. HiGEM does not show the change in vertical motion over the North Atlantic seen in HiGAM and ERA-Interim, explaining the erroneous response of tropical cyclones in this region. A change in the upper-level circulation can be seen in ERA-Interim around 280oE, which may influence the mean vertical motion over the North Atlantic. This driving mechanism is not simulated and as a result HiGEM does not capture the expected ENSO-tropical cyclone variability in the North Atlantic. The change in both vertical motion and upper-level circulation in the North Atlantic is somewhat better simulated in HiGEM during La Nina˜ years compared to El Nino˜ years however the observed response is smaller in La Nina˜ years than El Nino˜ years. The lack of ENSO-SST variability in the tropical east Pacific in HiGEM may impact the Hadley cell (Wang, 2002) which may also relate to the poor simulation of ENSO-associated vertical wind shear over the North East Pacific and North Atlantic. The localized ascending motion around 160oE during an El Nino˜ event in HiGEM may prevent the upper-tropospheric circulation response over the North Atlantic.

Page 78 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.11 Height-longitude cross section of Walker Circulation 0-10oN, July-October for (a) ERA-Interim El Nino˜ years minus 1979-2010 climatology, (b) ERA-Interim La Nina˜ years minus 1979-2010 climatology, (c) HiGEM El Nino˜ years minus 1979-2002 climatology and (d) HiGAM La Nina˜ years minus 1979-2002 climatology, (e) HiGEM El Nino˜ years minus 150-year climatology and (f) HiGEM La Nina˜ years minus 150-year climatology. The colors show mean ascent (-ω) and the vectors are mean ascent and a change in the velocity potential with respect to longitude. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

3.5.4 Vertical wind shear

Vertical wind shear is deﬁned as the magnitude of the vector difference between winds at 850 and 200 hPa. The response of vertical wind shear to ENSO during JASO is shown in ﬁgure 3.12. HiGAM is able to capture the shift in vertical wind shear in El Nino˜ and La Nina˜ years when compared to ERA-Interim, although the simulated response is stronger in magnitude. Vertical wind shear is also much stronger in HiGAM than in HiGEM. Figure 3.9 shows AMIPII associated ENSO SSTs are constrained along the

Page 79 Chapter 3: The ENSO-Tropical Cyclone Teleconnection equator more-so than in HadISST and HiGEM. The larger meridional SST gradient will result in a stronger upper-level westerlies via thermal wind balance which will increase vertical wind shear. In addition, as the atmosphere in HiGAM has a larger source of heat without the presence of air-sea coupled feedbacks, the large-scale ENSO associated tropical convection is more intense and therefore drives stronger winds. The responses observed in the West Pacific and Indian Oceans are captured well by HiGEM, although with slightly reduced magnitude than that seen in ERA-Interim. The smaller magnitude response over the Western North Pacific can help to explain why tropical cyclones show less variability in HiGEM than observed. ERA-Interim reveals a dipole of vertical wind shear over the North East Pacific and North Atlantic due to the response of ENSO, which was also noted by Aiyyer and Thorncroft (2006). This pattern explains the dipole of tropical cyclone activity between the North Atlantic and North East Pacific (Frank and Young, 2007; Maue, 2009). Camargo et al. (2007c) also notes that vertical wind shear is an important parameter for explaining the ENSO-tropical cyclone teleconnection in these basins. HiGEM does not capture the expected vertical wind shear shift in this region.

Page 80 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.12 Vertical wind shear (m s−1), July-October for (a) ERA-Interim El Nio years minus 1979-2010 climatology, (b) ERA-Interim La Nia years minus 1979-2010 climatology, (c) HiGEM El Nino˜ years minus 1979-2002 climatology and (d) HiGAM La Nina˜ years minus 1979-2002 climatology, (e) HiGEM El Nino˜ years minus 150-year climatology and (f) HiGEM La Nina˜ years minus 150-year climatology. The climatology is shown in black contours for ERA-Interim: (a) and (b); HiGAM: (c) and (d) and HiGEM: (e) and (f). Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

3.5.5 Low-level vorticity

Low-level vorticity represents the ‘spin’ of the atmosphere that is required to form cy- clonically rotating storms. Camargo et al. (2007c) note that low-level vorticity is important for tropical cyclogenesis with ENSO in the Western North Paciﬁc. Figure 3.13 shows the 850 hPa JASO relative vorticity for El Nino˜ and La Nina˜ years compared to climatology in ERA-Interim, HiGAM and HiGEM. The observed vorticity changes in the Western North Paciﬁc match those of Wang and Chan (2002) and Mori et al. (2013).

Page 81 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

HiGAM simulates an increase in vorticity north of the equator in the Pacific with a similar magnitude to that observed in ERA-Interim during El Nino˜ years. There is a large change in vorticity in the Western North Pacific around 150oE, 20oN in HiGAM during El Nino˜ and La Nina˜ events with is stronger than observed. HiGAM is able to simulate the large-scale observed response of vorticity over the tropical North East Pacific and North Atlantic. HiGEM broadly captures the increase in vorticity in the Pacific region during El Nino˜ years albeit weaker than observed. In addition, the decrease in vorticity in the main development in the North Indian Ocean during El Nino˜ years is simulated in HiGEM. The vorticity response during La Nina˜ years is poorly simulated to the west of Hawaii. Iizuka and Matsuura (2008) similarly found that vorticity does not decrease as much as in observations during La Nina˜ years in the eastern parts of the Western North Pacific. HiGEM simulates no change in relative vorticity in the tropical North East Pacific and North Atlantic.

Page 82 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.13 he 850 hPa relative vorticity (1×10−5 s−1), July-October for (a) ERA-Interim El Nino˜ years minus 1979-2010 climatology, (b) ERA-Interim La Nina˜ years minus 1979-2010 climatology, (c) HiGEM El Nino˜ years minus 1979-2002 climatology and (d) HiGAM La Nina˜ years minus 1979-2002 climatology, (e) HiGEM El Nino˜ years minus 150-year climatology and (f) HiGEM La Nina˜ years minus 150-year climatology. Stippling shows where changes have a p-value < 0.0001 using a student’s t-test..

3.5.6 Upper-level circulation

The upper-level circulation response to ENSO is investigated further using the velocity potential - an integrated measurement of the irrotational part of the upper level ﬂow; and the stream function - the rotational part. The stream function highlights the anomalous wave propagation, whereas the velocity potential highlights the forcing of these waves. Figure 3.14 shows the velocity potential and stream function at 200 hPa during JASO for ERA-Interim, HiGAM and HiGEM. ERA-Interim shows that in El Nino˜ years twin anti-cyclones straddle the equator as a response to increased convection in the central

Page 83 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Pacific (Spencer and Slingo, 2003), similar to a Gill-type response (Gill, 1980). In ERA- Interim and HiGAM during El Nino˜ years the upper-level divergent circulation is constrained to the central Pacific, which was also found in the ECHAM5 model by Bengts- son et al. (2007b). HiGEM simulates enhanced upper-level convergence in the Indian Ocean, which also explains the greater tropical cyclone variability related to anomalous large-scale ascent. The spatial extent of the upper-level divergent circulation in HiGEM reaches too far into the North Atlantic from the central Pacific. As a result of the spatial errors in the maximum velocity potential in HiGEM, the upper-level wave propagation is not captured over the North Atlantic, which has previously been related to tropical cyclogenesis (Shaman et al., 2009).

Page 84 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.14 The 200 hPa velocity potential (1×10−6 m2 s−1) in colors and the 200 hPa stream function (1×10−6 m2) in black contours, July-October for (a) ERA-Interim El Nino˜ years minus 1979-2010 climatology, (b) ERA-Interim La Nina˜ years minus 1979-2010 climatology, (c) HiGAM El Nino˜ years minus 1979-2002 climatology (d) HiGAM La Nina˜ years minus 1979- 2002 climatology, (e) HiGEM El Nino˜ years minus 150-year climatology and (f) HiGEM La Nina˜ years minus 150-year climatology. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

Page 85 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.6 Thermodynamic vs. dynamic inﬂuences

Figures 3.15 and 3.16 show the relationships between large-scale environmental conditions that are important for tropical cyclone activity over the main development regions and tropical cyclone counts for the North Atlantic and Western North Pacific, respectively. The area-average used for the North Atlantic is the same as in chapter 5: 275oE- 340oE, 10o-20oN. The region with the largest change in tropical cyclone activity is used for the Western North Pacific: 130oE-170oE, 5o-20oN. This analysis includes: thermodynamic variables of SST and relative humidity at 700 hPa; dynamic variables of relative vorticity at 850 hPa, mean ascent at 500 hPa (-ω500) and vertical wind shear. The figures highlight biases in the mean state of the models as well as limitations in capturing the magnitude of ENSO associated large-scale environmental changes. Both figures show Nino-3.4˜ SSTs in panel (a). It can be seen that HiGEM has a cool SST bias in the tropical Pacific although it captures the expected magnitude of ENSO SST anomalies. In the North Atlantic, it can be seen that vertical wind shear is an important parameter for tropical cyclone activity in ERA-Interim, which is also discussed in Camargo et al. (2007c). The mean vertical wind shear is too strong in HiGAM and more so in HiGEM, which explains the reduced mean number of tropical cyclones compared to observations. This was also found in other GCMs (Shaman and Maloney, 2011). The associated change of vertical wind shear with ENSO in HiGEM is opposite to that observed, with stronger vertical wind shear during El Nino˜ years and slightly more tropical cyclones. The coupled model used in Iizuka and Matsuura (2009) was able to simulate the vertical wind shear response although the simulated amplitude of the Nino-3.4˜ SST anomalies are much larger than observed. Tropical cyclones are not suppressed as strongly during El Nino˜ in HiGAM compared to ERA-Interim, as the change in large- scale deep ascent is not captured. Although vertical wind shear shows a slight increase in La Nina˜ years in HiGAM the number of tropical cyclones increase possibly related to an increase in mid-level relative humidity and mean ascent at 500 hPa. The thermodynamic parameters are of secondary importance to dynamic parameters in explaining the tropical

Page 86 Chapter 3: The ENSO-Tropical Cyclone Teleconnection cyclone response in the North Atlantic in HiGAM. HiGEM simulates both a poor mean- state and little variability of mid-level relative humidity and mean ascent at 500 hPa and therefore does not capture the expected ENSO-tropical cyclone teleconnection. In the Western North Paciﬁc, HiGAM and HiGEM simulate too many tropical cyclones, which is discussed further in chapter 5, even though vertical wind shear is stronger in HiGAM and HiGEM than in ERA-Interim. In addition, vertical wind shear increases during El Nino˜ years along with an increase in tropical cyclones, indicating that vertical wind shear is not an important driver of tropical cyclone activity over the main development region (Zhao and Held, 2011; Strachan et al., 2013). The variability of tropical cyclones with ENSO in this basin is related to the variability in vorticity (Camargo et al., 2007c). HiGAM simulates a large variability of relative vorticity which causes large variability in the number of tropical cyclones. HiGEM is also able to simulate the relationship of vorticity with ENSO although the tropical cyclone counts do not respond during El Nino˜ years. Research using track density, as a measurement of spatial tropical cyclone activity, in the region of interest shows an increase of tropical cyclones in HiGEM along with the increase of relative vorticity during El Nino˜ years (not shown). The relative humidity at 700 hPa is slightly larger in the mean-state of both HiGAM and HiGEM which increases the number of tropical cyclones which form each year compared to that observed. The relationship of relative humidity at 700 hPa with ENSO in HiGAM and HiGEM show distinct differences, however the changes are small and are associated with larger interannual variability. The simulated tropical cyclones in this region are sensitive to a change in mean ascent at 500 hPa. Wang et al. (2013) discussed how environmental factors impact tropical cyclone frequency differently on the inter-basin scale in the Western North Paciﬁc.

Page 87 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.15 Scatter plot of tropical cyclone counts versus large-scale environmental parameters for the North Atlantic, averaged over the region 275-340oE, 10-20oN during July-October. Red symbols represent the climatology of El Nino˜ years, blue symbols represent the climatology of La Nina˜ years and black symbols represent the climatology of all years for (a) NINO3.4 SST anomaly, (b) vertical wind shear, (c) relative vorticity, (d) mean ascent at 500 hPa and (e) relative humidity at 700 hPa. Observations are the cross, ERA-Interim the diamond, HiGAM the circle and HiGEM the triangle. The line styles distinguish between model and observations. The error bars denote the 90 % conﬁdence interval.

Page 88 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Figure 3.16 Scatter plot of tropical cyclone counts versus large-scale environmental parameters for the Western North Pacific, averaged over the region 130-170oE, 5-20oN during July-October. Red symbols represent the climatology of El Nino˜ years, blue symbols represent the climatology of La Nina˜ years and black symbols represent the climatology of all years for (a) NINO3.4 SST anomaly, (b) vertical wind shear, (c) relative vorticity, (d) mean ascent at 500 hPa and (e) relative humidity at 700 hPa. The error bars denote the 90 % confidence interval. Symbols and line styles are the same as in figure 3.15.

Page 89 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.7 Discussion

Simulating the correct spatial SST pattern of ENSO is a necessary, but not sufficient requirement for an accurate representation of the global ENSO teleconnections in an AOGCM. Dawson et al. (2013) found an increase in the oceanic resolution produced more accurate ENSO teleconnections compared to an increase in only the atmospheric resolution due to an improvement in the mean-state and ENSO variability. Errors in the atmospheric teleconnections simulated in HiGEM stem from mean-state SST biases and errors in the spatial pattern of ENSO associated SST. Zhu et al. (2011) found that the simulated mean-state of vertical wind shear over the North Atlantic was shown to play a critical role in how the remote influence of ENSO modulates the vertical wind shear. HiGEM has a large mean-state bias of vertical wind shear over the North Atlantic and a underestimation of climatological frequency of tropical cyclones compared to observed. In addition, HiGEM does not capture the expected tropical cyclone variability associated with El Nino˜ and La Nina˜ events in the North Atlantic. Errors in the variability of ENSO- vertical wind shear teleconnection over the North Atlantic in HiGEM are caused by errors in the spatial pattern of ENSO associated SST. As the maximum SST warming occurs too far west compared to observations, the eastwards atmospheric branch of the Walker circulation does not reach into the North Atlantic. As a result the coupled model does not capture the dipole of vertical wind shear over the North East Pacific and North Atlantic. However, the westward extension of ENSO associated SSTs simulates a realistic ENSO- tropical cyclone teleconnection in the Indian Oceans. The mean-state of vertical wind shear over the tropical North Atlantic is reduced in the uncoupled model, HiGAM. There are still errors in the vertical wind shear variability, with slightly more vertical wind shear in La Nina˜ years than the climatology. When the vertical shear is low enough so that it does not dominate the reduction in tropical cyclone signal, the other governing environmental parameters may have a role in influencing tropical cyclone variability, such as the mean ascent at 500 hPa and relative vorticity, which are better captured in HiGAM than in HiGEM.

Page 90 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

Whilst the HiGAM simulation shows a marked improvement of the ENSO-tropical cyclone teleconnection in the Western North Pacific there are still some limitations due to inaccurate representation of atmospheric teleconnection processes. As discussed in Shaffrey et al. (2009) HiGEM has a large mean-state precipitation bias over the Western North Pacific. The mean-state precipitation bias in HiGEM remains in HiGAM over the Western North Pacific. The precipitation is in response to stronger ascent which likely results in too many tropical cyclones forming. The variability in terms of tropical cyclone location and the number of tropical cyclones per season is simulated more accurately in HiGAM then HiGEM. This is attributed here to the greater variability of mean ascent at

500 hPa (-ω500). The atmosphere in HiGAM has a larger source of heat than HiGEM without the presence of air-sea coupled feedbacks. During El Nino˜ years the warmer SST leads to more intense tropical convection. Although this is not entirely clear in the variability of relative humidity shown in figure ?? this parameter is associated with large errors bars and is sensitive to the area-average used. A basin-wide area average shows a similar mid-level relative humidity pattern in HiGAM to that observed (not shown). It may be suggested that the greater variability of tropical cyclone activity in the Western North Pacific in HiGAM is influenced by the differences in simulation lengths of HiGEM and HiGAM. However, A 23-year period with a similar number of ENSO events as that in HiGAM in HiGEM show the results remain largely unchanged (not shown).

3.8 Conclusion

It is important to evaluate the ability of GCMs to simulate realistic ENSO associated tropical cyclone teleconnections for seasonal forecasting and before predictions are made for tropical cyclones and climate change using GCMs (Shaman and Maloney, 2011). The global ENSO-tropical cyclone teleconnection has been investigated with the use of a 150-year high-resolution AOGCM. The 150-year simulation gives robust statistics of the ENSO-tropical cyclone teleconnection. The relationship was also investigated further

Page 91 Chapter 3: The ENSO-Tropical Cyclone Teleconnection using an atmosphere-only GCM forced with observed SSTs. The conclusions of this study are as follows:

• The coupled model, HiGEM, is able to capture the observed shift in tropical cyclone location with ENSO in the Paciﬁc and Indian Oceans.

• HiGEM is not able to capture the expected tropical cyclone response in the North Atlantic and simulates a response opposite to observed, with slightly more tropical cyclones during El Nino˜ years.

• The precipitation biases and lack of mid-level relative humidity variability in the Western North Paciﬁc are found to be the limiting factors in capturing the expected magnitude of the ENSO-tropical cyclone teleconnection in HiGEM.

• The large-scale environment in the North Atlantic is not inﬂuenced by ENSO variability in HiGEM. In particular, the vertical wind shear response over the Caribbean is not captured.

• HiGAM simulates the dipole of vertical wind shear in the North East Paciﬁc and Caribbean Sea which is not found in HiGEM and therefore the tropical cyclone teleconnection is not simulated as expected.

• Although the large-scale precipitation bias remains in HiGAM over the Western North Paciﬁc the tropical cyclone variability is much better simulated possibly related to a more accurate representation of mid-level relative humidity with ENSO.

The tropical cyclone changes with climate change found in chapter 5 are somewhat similar to the El Nino˜ response found in this study. How ENSO may change in the future (Collins et al., 2010) will have a large inﬂuence on future tropical cyclone activity. In the near-term future, ENSO and other modes of natural variability will dominate variability of tropical cyclone activity. It is therefore important that a similar emphasis is placed on understanding the global ENSO-tropical cyclone teleconnection as with tropical cyclones and climate change using GCMs.

Page 92 Chapter 3: The ENSO-Tropical Cyclone Teleconnection

3.9 Future work

HiGEM and HiGAM are compared to ERA-Interim reanalysis only. Future work should include comparisons to other high-resolution reanalysis such as The Modern Era Retrospective-Analysis for Research and Applications (MERRA; Rienecker et al., 2011), Japanese reanalysis (JRA; Onogi et al., 2007) or National Centers for Environmental Pre- diction Climate Forecast System reanalysis (NCEP CFSR; Saha et al., 2010). Research by Jane Strachan has showed that tropical cyclones tracked in these different reanalyses give slightly different results. An investigation into the change of tropical cyclone intensity in each basin with ENSO would be of use to understand dynamical seasonal forecasts, however this should be left to a higher resolution atmosphere-only GCM due to limitations of capturing observed tropical cyclone intensities at this resolution (Strachan et al., 2013). This study only uses one climate model. As tropical cyclones associated with ENSO events are rare occurrences a multi-model study will increase the sample size of tropical cyclone events and allow for a comparison to other coupled models. Pre- vious research shows a rich diversity of ENSO with considerable natural variability at interdecadal-centennial timescales (Li et al., 2013). To capture the representation of this variability model integrations are needed of multi-centennial time scale. The role of ocean coupling and its importance on the ENSO-tropical cyclone teleconnections can be investigated further using the HiGEM-HiGAM timeslice experiment. However, the integration length for this experiment is only 30 years and ideally would have to be integrated for the same length as HiGEM.

Page 93 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.1 Introduction

A recent development in ENSO research has been an interest in El Nino˜ event diversity. Ashok et al. (2007) suggested the existence of another type of El Nino˜ with the centre of action in the central Pacific different to typical canonical types which have their centre of action in the eastern Pacific. There is an ongoing debate in the ENSO community on the existence of two distinct modes, however their impacts are believed to be different. The previous section showed that HiGEM was unable to capture the ENSO-tropical cyclone teleconnection in the North Atlantic. Therefore this chapter investigates how different types of El Nino˜ affect tropical cyclone activity in the Western North Pacific only. This basin has also been widely studied and allows for comparison to observations. Idealised atmosphere-only model simulations are used to attribute the response when specifying SST associated with the different types of El Nino˜ in the eastern tropical Pacific only. SST Composites of different types of El Nino˜ are taken from HiGEM and used to force HiGAM. Tropical eastern Pacific SST experiments are compared to global SST experiments to isolate the information coming from the El Nino˜ region alone. These experiments can help identify the regions that are important to improving skill in seasonal forecasting (Spencer et al., 2004; Smith et al., 2010). This chapter is structured as follows: The different types of El Nino˜ are first defined followed by a literature review describing their impact on tropical cyclone activity. The

Page 94 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity simulation of different types of El Nino˜ in HiGEM is shown in relation to observations and other models. The results of changing tropical cyclone location and frequency in the Western North Paciﬁc associated with the different El Nino˜ experiments are presented followed by an investigation into the change in large-scale environmental conditions. Finally, the ﬁndings are discussed in comparison to other idealised modelling studies followed by concluding remarks.

4.1.1 Different types of El Nino˜

The study of Ashok et al. (2007), using an empirical orthogonal function (EOF) analysis of monthly tropical Pacific SST anomaly, showed that central Pacific El Nino˜ events (CP-EN) are represented by the second mode that explains 12 % of the variance. Since this work there have been a vast number of observational and modelling studies to better define CP-EN, which have led to various definitions (e.g. Kug et al., 2009; Kao and Yu, 2009; Ren and Jin, 2011; Ham and Kug, 2011). Figure 4.1 is taken from Wang et al. (2013) and shows the different types of El Nino˜ using the definition of: eastern Pacific El Nino˜ (EP-EN) if the 5-month running means of SST anomalies in the Nino-˜ 3 region exceed 0.5 oC for more than 6 months and the maximum SST anomalies are located in the eastern Pacific (i.e., the Nino-3˜ SST anomalies are larger than the Nino-˜ 4 SST anomalies). CP-EN events are associated with the maximum SST anomalies in the central Pacific and requires that the 5-month running means of SST anomalies in the Nino-4˜ region exceed 0.5 oC for at least 6 months. The EP-EN composites the years 1951, 1957, 1965, 1972, 1976, 1982, 1987 and 1997. The CP-EN composites the years 1969, 1991, 1994, 2002 and 2004. It is not clear how rare these two types of events are and therefore the difficulty in obtaining robust statistical relationships using observations. Whether there really are distinctly different types of ENSO, or whether there is one type of ENSO with variability in its location is still a topic of debate (CLIVAR, 2013).

Page 95 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.1 Composite SST anomalies (oC) during the mature phase of ENSO (November- January) for (top) EP-EN and (bottom) CP-EN adapted from (Wang et al., 2013).

Yeh et al. (2009) further suggest that the CP-EN has become more frequent in recent years, along with the studies by Latif et al. (1997); Lee and McPhaden (2010); and McPhaden (2012). DiNezio et al. (2009) argue that during transient climate change the upper tropical Pacific Ocean continually warms. El Nino˜ events on top of this background surface warming are more influenced by surface forcing which favours CP-EN. DiNezio et al. (2009) believe EP-EN is mainly forced by changes in the thermocline which will be prominent again as the deep ocean warms. However, Nicholls (2008); and McPhaden et al. (2011) found no trend in the increase of CP-EN events using other statistical approaches. The CP-EN has been found to give remarkably different atmospheric teleconnections than the EP-EN. Figure 4.2 shows global precipitation and surface temperature teleconnections associated with the different types of El Nino.˜ Ashok et al. (2007) found the teleconnections over Japan and New Zealand were of the opposite sign to that expected from the typical El Nino.˜ The increase in convection in the central Pacific leads to a surplus of rainfall flanked by negative rainfall anomalies in the western equatorial Pacific and eastern equatorial Pacific. The deficit of rainfall in the western Pacific region is seen

Page 96 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity to extend southward to southeastern Australia, influencing a significant part of eastern Australia. The negative rainfall anomalies over the equatorial eastern Pacific also extend over the western coast of North America. This was studied further by Hill et al. (2011), who investigated the sensitivity of South American summer rainfall to regional tropical Pacific Ocean SST anomalies using idealised modelling experiments. The change in position of deep tropical convection drives changes in the large-scale circulation which are different from EP-EN. Chen and Tam (2010) used a baroclinic AGCM to understand the role of different heating sources associated with different types of El Nino.˜ Chen and Tam (2010) found during CP-EN a large cyclonic anomaly forms over the Western North Pacific. In contrast, during EP-EN years, a zonally-elongated heating source and sink exhibit a meridional dipole pattern, which induces an anticyclonic anomaly in the subtropics and a cyclonic anomaly near the equatorial central Pacific. These are discussed further in the section below in relation to tropical cyclone activity.

Figure 4.2 Teleconnections associated with different types of El Nino.˜ Modoki is another term for CP-EN. taken from http://www.jamstec.go.jp/frcgc/research/d1/iod/enmodoki home s.html.en.

Page 97 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.1.2 Different types of El Nino˜ and tropical cyclone activity: A review

As discussed in the previous section the different types of El Nino˜ can lead to different atmospheric teleconnections. Chapter 3 in this thesis showed how important ENSO associated teleconnections are for global tropical cyclone activity. This section presents contemporary research on how the different types of El Nino˜ may affect tropical cyclone activity. The study of Kim et al. (2009) investigated the impact of different types of El Nino˜ on North Atlantic tropical cyclones. Kim et al. (2009) found an increased occurrence of landfalling storms in the Southwest United States during CP-EN events, greater than that during La Nina˜ events. The study of Kim et al. (2009), along with other observational studies investigating the impact of different types of El Nino˜ on tropical cyclone activity, is limited as it uses a small sample size of CP-EN events (5) for which the tropical cyclone record is reliable. This is especially problematic for understanding the tropical cyclone response in the North Atlantic as the number of tropical cyclones which form each year is less than for the Western North Pacific, for example. Lee et al. (2010) dis- agree with Kim et al. (2009) by stating that only two of the chosen CP-EN events had enhanced tropical cyclone activity in the North Atlantic. Lee et al. (2010) undertook idealised model experiments using the CP-EN years defined in Kim et al. (2009). They found that local forcing from the Atlantic warm pool had a greater influence on tropical cyclone activity during some of the CP-EN events used in Kim et al. (2009). The majority of research on tropical cyclone activity associated with different types of El Nino˜ focusses on the Western North Pacific. This region is subject to dynamical and thermodynamical changes associated with El Nino˜ in the equatorial tropical Pacific. There is also an interest in the threat of typhoon landfall in East Asia (Zhang et al., 2012). In the summer of 2004 during a CP-EN year a record-breaking ten typhoons made landfall in Japan (Kim et al., 2005). Hong et al. (2011) found tropical cyclones track further westward and recurve later

Page 98 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity leading to more landfall over Taiwan and South China during CP-EN events. They attribute this to changes in the local Hadley cell and the location of the subtropical high affecting the steering flow. Kim et al. (2011) similarly investigate different types of El Nino˜ on the modulation of tropical cyclones in the North Pacific, which are shown in figure 4.3. The westward shift of west Pacific induced heating is shown to move the anomalous westerly wind and monsoon trough to the northwest section of the Western North Pacific and provides a more favourable condition for tropical cyclone landfall during CP- EN. Pradhan et al. (2011) relate the change in low-level cyclonic flow over the northern South China Sea to a quasi-stationary Rossby wave response during CP-EN events. Ha et al. (2012) have further studied how the change in tropical cyclone location with the different types of El Nino˜ relates to intensity. Wang et al. (2013) used the NCEP/NCAR re-analysis and Joint Typhoon Warming Center (JTWC) and China Meteorological Ad- ministration (CMA) typhoon observations, which span the period of 1950-2009. They found during the peak tropical cyclone season EP-EN is associated with steering flows that are unfavourable for tropical cyclones to move northwestward or westward, whereas CP-EN favours the northwestward track and suppresses the straight westward track. This can be seen in figure 4.4.

Page 99 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.3 Composites of (a) Eastern Paciﬁc warming (EPW; EP-EN) and (b) Central Paciﬁc warming (CPW; CP-EN) track density anomaly adapted from (Kim et al., 2011).

Figure 4.4 Track density anomaly (colour; number of tropical cyclones passing through a 5o × 5o grid box) and steering flow anomalies (m s−1) for (left) EP-EN and (right) CP-EN. Yellow contours and stippling indicate statistically significant at the 90 % level for track density and steering flow, respectively. Adapted from Wang et al. (2013).

Model studies can complement observational studies by increasing the sample sizes of CP-EN events and provide a platform to further investigate what mechanisms are

Page 100 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity important. Chen and Tam (2010) use a simple baroclinic AGCM to examine the atmospheric response to different heating sources and delineate circulation anomalies for the two types of El Nino˜ events. The coarse resolution of the model used (T42) was found to broadly replicate large-scale environmental conditions which they relate to observed tropical cyclone frequency. They found the CP-EN induces a large-scale cyclonic anomaly over the Western North Pacific which increases vorticity anomalies and is cor- related with the increase in tropical cyclone numbers. During the EP-EN experiments a negative heating anomaly around 160oE, 20oN exists. The cooling perturbation drives a low-level anticyclonic response which merges with the equatorial westerlies forming a cyclonic shear pattern over near-equatorial Western North Pacific. A similar coarse resolution model study was undertaken by Hong et al. (2011) who used ECHAM5 at T42 resolution. Hong et al. (2011) argue the difference in tropical cyclone tracks between CP-EN and EP-EN is due to local SST warming in the Western North Pacific. During CP-EN a warm SST anomaly in the central Pacific increases convection and strengthens the Hadley cell, which leads to an increase in subsidence in the subtropics. The enhanced subsidence further dries the environment and causes a westward extension of the subtropical high. Numerical experiments of global CP-EN minus EP-EN SST, and Western North Pacific CP-EN minus EP-EN SST show the Western North Pacific experiment captures the Hadley cell response similar to the global experiment. Hong et al. (2011) argue the Western North Pacific experiment captures the expected Hadley cell response and causes the different tropical cyclone tracks (figure 4.5). However, the model is of coarse resolution and they do not identify tropical cyclones in their experiments and only infer relationships with large-scale changes. In addition, the study does not investigate other drivers such as the change in steering flow between the two experiments.

Page 101 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.5 (b) Global EP-EN minus CP-EN El Nino˜ experiment and (a) the local Hadley circulation averaged over 150oE-160oE, apparent heat source (Q1 from Yanai et al., 1973) and vectors are vertical velocity and meridional wind. (d) Western North Paciﬁc experiment and (c) Hadley circulation response (Hong et al., 2011).

In contrast, the study of Jin et al. (2012) find remote SSTs are more important in driving tropical cyclone location during CP-EN events. Jin et al. (2012) use a Weather Research and Forecast (WRF) regional climate model with a global 150 km resolution forcing and a 50 km horizontal resolution over the Western North Pacific. They per- form a CP-EN minus climatology SST forcing experiment, where the SSTs are taken from a region on the equator around 180oE (see red box in figure 4.6 (e)). An additional experiment is undertaken which increases the region of SST forcing poleward to include the positive SST anomaly off the equator in the Northern Hemisphere. The off- equatorial CP-EN SST anomaly was found to expand the anomalous cyclonic response in the Philippine Sea farther northward. The location of the cyclonic anomaly produces a tunnel effect in the East China Sea as more tropical cyclones move towards east Asia, shown in figure (figure 4.6 (b) and (f)). However, the results of the experiments are not

Page 102 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity substantially different. Further sensitivity studies of similar regions would have helped identify now robust the response is. In addition no comparison to EP-EN events were made.

Figure 4.6 Region 1 SST forcing (left hand side) uses CP-EN SST in the red shaded box in (e) and climatological SSTs elsewhere. Region 1 and 2 SST forcing (right hand side) uses CP-EN SST in the red shaded box in (f) and climatological SSTs elsewhere. Composite anomalies of (a),(b) track density; (c),(d) genesis density; and (e),(f) 500 hPa geopotential height (m; contour), steering flows (m s−1; vector), and OLR (W m−2; shading) during CP-EN. The 5880-500 hPa geopotential height lines (contour) for climatology (dashed) and CP-EN (solid) are also plotted in (c),(d). Dots, crosshatching and thick vectors denote regions where the anomalies are significant at the 90 % confidence level based on a student’s t test (Jin et al., 2012).

Page 103 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.2 Simulation of different types of El Nino˜ in HiGEM

The two types of El Nino˜ are defined using the method of Ham and Kug (2011). Ham and Kug (2011) found most climate models have difficulty in simulating the two types of El Nino˜ and La Nina˜ events. The ability of the CMIP3 models along with Extended Reconstructed Sea Surface Temperature Version 2 (ERSST V2) observational data to simulate the two types of El Nino˜ is shown in figure 4.7.

Page 104 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.7 The scatter plots between normalized Nino-3˜ and Nino-4˜ indices during the ENSO in Boreal winter season (Dec-Feb): (a) observations, (b) modified Nino˜ indices in the observation, and (c-v) the CMIP3 models. The definition of modified Nino-3˜ (Nino-4)˜ is seasonal-mean SST anomalies averaged over 170oW-110oW, 5oS-5oN (140oE-170oW, 5oS-5oN). Note that the blue (red) dots denote the cases for (Warm pool (WP): CP-EN) and (Cold tongue (CT): EP-EN), respectively. (Ham and Kug, 2011).

HiGEM simulates two types of El Nino˜ similar to a number of other climate models, such as the Geophysical Fluid Dynamics Laboratory Coupled Model 2.0; Institute for Numerical Mathematics Coupled Model Version 3.0; Hadley Centre Coupled Model version 3 and MIROC3.2 (hires) (ﬁgure 4.8). The results highlight the asymmetric nature

Page 105 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity of the different types of El Nino,˜ with EP-EN showing Nino-3˜ normalised SST anomalies of up to 3.2 and CP-EN showing Nino-4˜ normalised SST anomalies of up to 2. The amplitude of the different types of El Nino˜ is comparable to observations along with the relative frequency of each event. The spatial pattern of the different types of El Nino˜ is given in figure 4.10. As discussed in section 3.5.1, HiGEM El Nino˜ SSTs extends too far westward and do not warm near the coast of Peru. This can also be seen in the EP-EN events, which is very similar to a typical El Nino˜ event. Hence the location of maximum warming is shifted westward compared to Kim et al. (2009). The magnitude of the CP- EN is greater than that found in observations by Kim et al. (2009); however, the HiGEM SST composite uses the five strongest El Nino˜ events. Kim et al. (2009) also show the SST anomalies associated with CP-EN are small in the east, whereas in HiGEM they are slightly stronger. Again this is attributed to the composite of strong El Nino˜ events in HiGEM than in the observations used by Kim et al. (2009). Both types of El Nino˜ give different SST patterns in the Western North Pacific. The CP-EN is much weaker in magnitude compared to EP-EN and shows no change of SST in the West Pacific. Whereas, the EP-EN is associated with a cooling in the West Pacific.

Figure 4.8 Magnitude of different types of El Nino˜ from the HiGEM 150-year present-day simulation. See ﬁgure 4.7.

Page 106 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.3 Model experiments

This section describes the El Nino˜ experiments which are used to address a contemporary topic on the teleconnections associated with different types of El Nino.˜ The scientific motivation of this research is described in section 4.1. SSTs are taken from the 150-year present-day simulation of HiGEM. The reason HiGEM SSTs are used is because it is in- line with this thesis. The HiGEM model has been analysed and the biases are understood as discussed in chapter 2 and 3. Using HiGEM also has an advantage over using observations. Reliable observed SSTs in the tropical Pacific extend back to approximately 1950 (Solomon and Newman, 2012) therefore there have been few observed different types of El Nino˜ which limit the robustness of observed tropical cyclone response studies. The 150-year present-day simulation of HiGEM simulates a larger sample of different types of El Nino.˜ The five strongest El Nino˜ events during August-September-October (ASO) are composited as the focus is on the Northern Hemisphere circulation and the tropical cyclone response. The two types of El Nino˜ are defined using the methodology of Ham and Kug (2011):

• East Paciﬁc El Nino˜ (EP-EN) is deﬁned as normalised Nino-3˜ (150oW-90oW, 5oS- 5oN) SST anomaly > 1 and Nino-3˜ SST anomaly > Nino-4˜ (160oE-150oW, 5oS- 5oN) SST anomaly.

• Central Paciﬁc El Nino˜ (CP-EN) is deﬁned as Nino-4˜ SST anomaly > 1 and Nino-4˜ SST anomaly > Nino-3˜ SST anomaly.

Additional remote SST experiments are also used to investigate the role of the eastern tropical Pacific. These experiments use SSTs from the tropical eastern Pacific (160oE- East Pacific coast, 10oS-20oN) whilst applying a Tukey filter (tapered cosine; Bloomfield, 2000) out to 10o either side. SSTs are imposed in this region and ENSO neutral SSTs are imposed elsewhere. The Tukey filter can be regarded as a cosine lobe convolved with a rectangular window (see figure 4.9) and is given mathematically as:

Page 107 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.9 Tukey window used to smooth SST in the eastern tropical Paciﬁc El Nino˜ experiments, α = 0.5.

α 1 1 + cos π 2n − 1 0 6 n 6 (N−1)  2 h α(N−1) i 2 α w(n) =  (N−1) 6 6 α (4.1)  1 2 n (N − 1)(1 − 2 ) α 1 1 + cos π 2n − 2 + 1 (N − 1)(1 − ) 6 n 6 (N − 1)  2 h α(N−1) α i 2  Where n is the longitude or latitude, α is 0.5 and N the number of longitude or latitude points. The application of the Tukey filter to the region of study is shown in figure 4.10. Figure 4.10 shows the different SST experiments. During CP-EN there is an increase in SST in the tropical North Atlantic. The remote (eastern tropical Pacific) SST experiments can be used to attribute tropical cyclone activity that occurs outside of the eastern tropical Pacific due to the teleconnections coming from the forcing region alone, similar to the experiments in Spencer et al. (2004). However, the model has a poor mean-state of tropical cyclones in the North Atlantic (see section 2.7.1), therefore the focus is on the Western North Pacific as with many previous studies. The SSTs are used to force the atmospheric component of HiGEM, HiGAM, with 10 ensemble members. Each ensemble member has slightly different initial conditions created by perturbing the prognostic potential temperature field at the bit level (Demory, 2012). Each experiment is integrated from March until December.

Page 108 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.10 (a) Composite of ENSO neutral SST during July-August-September-October (JASO) taken from the 150-year HiGEM present-day simulation, (b) composite of the five strongest Cen- tral Pacific El Ninos˜ minus climatology, (c) composite of the five strongest Eastern Pacific El Ninos˜ minus climatology, (d) same as for (b) with tropical eastern Pacific SSTs only and ENSO neutral SSTs elsewhere, (e) same as for (c) with tropical Eastern Pacific SSTs only and ENSO neutral SSTs elsewhere. The solid lines is the SST forcing region. The dotted lines shows the Nino˜ boxes (Nino-4˜ to the left and Nino-3˜ to the right). The dot-dash line shows the SST smooth- ing region.

Table 4.1 below describes the experiments and deﬁnes acronyms that will be used throughout the rest of the chapter. The regions used in the experiment are shown in ﬁgure 4.10.

Page 109 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Experiment Name SST

CP-EN Central Pacific El Nino˜ NINO4 SSTa > 1 and NINO4 SSTa > NINO3 SSTa EP-EN East Pacific El Nino˜ NINO3 SSTa > 1 and NINO3 SSTa > NINO4 SSTa rCP-EN Remote (tropical cen- Same for CP-EN but SSTs are extracted tral Pacific) CP-EN from (160oE-East Pacific coast, 10oS-20oN) and neutral SSTs elsewhere rEP-EN Remote (tropical east- Same for EP-EN but SSTs are extracted from ern Pacific) EP-EN (160oE-East Pacific coast, 10oS-20oN) and neutral SSTs elsewhere

Table 4.1 Deﬁnitions of the El Nino˜ experiments, see ﬁgure 4.10.

4.4 Tropical cyclone changes

The tropical cyclone changes in the Western North Paciﬁc in the El Nino˜ experiments are shown in ﬁgure 4.11. During CP-EN there is an increase in tropical cyclone occurrence within the region 120oE-160oE, 15oN. This represents the zonal westward track of tropical cyclones moving towards South East Asia, which is in line with previous observational studies (Kim et al., 2011; Chen and Tam, 2010; Wang et al., 2013). There is also a small recurving branch of tropical cyclones to the south-east of Japan. During CP-EN the number of tropical cyclones slightly increases to 30 per season compared to 28 per season during ENSO neutral. Chen and Tam (2010) also found an increase in tropical cyclone frequency during CP-EN, however they focus on regional activity and not the total count. Tropical cyclones remain offshore during EP-EN similar to the results discussed in section 3.4.1. The results are found to be similar to Chen and Tam (2010) for EP-EN as less tropical cyclones occur in the northern part of the basin and more tropical cyclones form in the southeastern part of the basin. The number of tropical cyclones which form each year is less than CP-EN events at 26 per season.

Page 110 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

In the eastern tropical Pacific SST CP-EN experiment (rCP-EN), tropical cyclones still move westward, however only as far as 130oE, east of the Philippines. There is a small increase in the number of tropical cyclones in the South China Sea, however less than the increase in the CP-EN experiment. Tropical cyclones become more frequent in the rCP-EN experiment with 33 storms per season forming. The results are similar to Jin et al. (2012) who undertook a similar modelling study using SST from a region in the central equatorial Pacific (see figure 4.6). The anomalous decrease of tropical cyclones to the west of Japan is also simulated in HiGAM, similar to the findings of Jin et al. (2012). However, Jin et al. (2012) found more of an increase in the region 120oE-140oE, 20oN-40oN than this study. All in all, the tropical cyclone changes in the rCP-EN experiment are not greatly different from the CP-EN experiment. In contrast, the eastern tropical Pacific EP-EN SST experiment (rEP-EN) shows a different pattern of tropical cyclone activity compared to the EP-EN experiment. It can be seen that tropical cyclones are more prominent around 140oE, 15oN, whereas during EP-EN tropical cyclones were reduced in this location. More than 6 tropical cyclones (an increase of 24 %) occur during the rEP-EN experiment compared to the EP-EN experiment. The largest changes are shown to be significantly significant at the 95 % level using a student’s t-test. The p-value of 0.05 was chosen as the 0.01 value, which is used for large-scale parameters, did not show any significance. The shift of tropical cyclones to the central Pacific is shown to be significant for the EP-EN experiment which was discussed in chapter 3. In the CP-EN there are only a few regions which show statisitcally significant changes in the east part of the basin. For both rCP-EN and rEP-EN there is a significant increase in tropical cyclone activity around 140oE-20oN. A larger sample size would help understand if the track density changes are not significant at the 99 % level due to a sampling error or if the track density changes are not physically different. Fur- ther work using the non-parametric Monte-Carlo method, which requires no assumption about the underlying data, developed by Hodges (2008) is needed.

Page 111 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.11 Tropical cyclone track density (storm transits/month/106 km2 or equivalent to a 5o radius) for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Pa- ciﬁc SST EP-EN minus ENSO neutral. The average number of tropical cyclone simulation for each experiment is shown in the upper-left hand corner. Stippling shows where changes have a p-value < 0.05 using a student’s t-test.

Figure 4.12 shows the tropical cyclone response to the different types El Nino˜ in the coupled HiGEM model. In comparison with the HiGEM-HiGAM experiment above HiGEM does not clearly simulate an increase in westward propagating zonal tropical cyclones during CP-EN as found in observations. The tropical cyclones in EP-EN mainly shift to the east, however there is a large increase in tropical cyclones forming near the Asian coastline, which is not observed. Figure 4.13 shows the differences of the HiGAM, HiGEM and HiGEM-HiGAM compared to GPCP. Whilst HiGEM is wrong for the right

Page 112 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity reasons the presence of air-sea coupling leads to large precipitation biases in the western North Paciﬁc. By switching off the air-sea coupling in the HiGEM-HiGAM simulation the large precipitaion bias is reduced. The atmosphere-only simulation uses a composite SST of the different types of El Nino,˜ therefore giving a more robust tropical cyclone response. The underlying SST in a AGCM has a large impact on the simulation of tropical cyclones (Roberts et al., 2013). The smooth SST composite forcing in these experiments are used to give more control over the simulation of tropical cyclones. Using the same SSTs, multiple ensembles members can be used and provide a larger sample than the number of different types of El Nino˜ events in HiGEM. The larger sample is necessary to identify robust changes. In addition, the atmosphere-only model is less expensive to use than the coupled model to diagnose the response associated with different regional forcing. The reduced cost is therefore utilised to allow the integration of many ensembles.

Figure 4.12 Tropical cyclone response in the HiGEM coupled model which uses the individual yearly SSTs which are used are a composite in ﬁgure 4.11. See ﬁgure 4.11.

Page 113 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.13 Precipitation (mm day−1), July-October for (a) 10 GPCP neutral years (1981-2001 exluding El Nino˜ and La Nina˜ years, see table 3.1), (b) 10 HiGAM neutral years (1981-2001) minus GPCP, (c) 10 HiGEM neutral years minus GPCP and (d) 10 HiGEM-HiGAM neutral years minus GPCP.

The spatial pattern of tropical cyclogenesis in the El Nino˜ experiments is shown in figure 4.14. Tropical cyclogenesis is related to tropical cyclone path as previous research has shown the genesis location influences the track due to large-scale flow patterns at the point of genesis in the Western North Pacific (Camargo et al., 2007a). The genesis pattern somewhat matches the tropical cyclone track density results above, although there are some differences. The majority of tropical cyclones in the CP-EN experiment form in a region around 150oE, 15oN. In the rCP-EN experiment genesis increases westward at 15oN compared to CP-EN. The tropical cyclones forming at this latitude are related to the increase in track density at 15oN in the rCP-EN experiment. During EP-EN the genesis is shifted eastward around the dateline. In the rEP-EN experiment, a large increase in genesis occurs to the east of the Philippines which increases the track density in this region. The regions with the largest change are significant at the 95 % level.

Page 114 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.14 Tropical cyclone genesis density (storm/month/106 km2 or equivalent to a 5o radius) for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.05 using a student’s t-test. 4.5 Large-scale environmental conditions

The results in this section investigate the driving mechanisms behind the change in tropical cyclone activity in the El Nino˜ experiments. The variables considered are the same as in section 3.5: precipitation (latent heat release), vertical wind shear, tropical circulation and relative humidity. The variables of geopotential height and steering ﬂow are also included to explain tropical cyclone movement.

Page 115 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.5.1 Precipitation

Section 3.5.2 describes the important precipitation as a teleconnection response. Figure 4.15 shows the July-October (JASO) precipitation changes in the El Nino˜ experiments. The precipitation pattern matches the changes in tropical cyclone activity. All experiments show a similar large-scale El Nino˜ pattern with enhanced precipitation east of 160oE on the equator and a decrease over the Maritime Continent. The main differences in the El Nino˜ experiments occur to the east of the Philippines. During CP-EN an increase in precipitation occurs to the east of the Philippines and in the South China Sea. The positive precipitation anomaly in this region has also been found in Kao and Yu (2009); Kug et al. (2009); Wang and Wang (2012). The precipitation pattern slightly increases in magnitude in the rCP-EN experiment with a larger region showing statistical signiﬁcance. Precipitation can be seen to decrease by 5 mm day−1 in the central Western North Paciﬁc in the EP-EN. In contrast, the rEP-EN experiment simulates a comparable precipitation response to the CP-EN experiment with an increase in precipiation to the east of the Philippines.

Page 116 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.15 Precipitation (mm day−1), July-October for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

The outgoing longwave radiation (OLR) changes are more smooth than precipitation changes as OLR also takes into account clouds which do not precipitate. Figure 4.16 shows that most features in the OLR can be explained by the precipitation response, however there are some differences. For example, the rCP-EN experiment shows a decrease in OLR to the north east of the Philippines, whereas the precipitation pattern is mixed. During EP-EN a narrow tongue of increased OLR spans from the Maritime Continent into the Western North Paciﬁc, which separates regions of decreased OLR in the equatorial central Paciﬁc and along the coast of Asia. The results for the EP-EN and CP-EN

Page 117 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity match well with Chen and Tam (2010) who performed idealised modelling studies using a baroclinic AGCM.

Figure 4.16 Outgoing longwave radiation (W m−2), July-October for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

4.5.2 Walker circulation

The change in the Walker circulation in El Nino˜ and La Nina˜ years is discussed in section 3.5.3. Figure 4.17 shows the response of the Walker circulation in the different El Nino˜ experiments. During CP-EN, when the maximum SST anomaly in located to the west, there is a small increase in vertical motion around the date-line. The largest changes

Page 118 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity occur around 100oE-150oE with an increase in subsidence, which was also found by Kim et al. (2009). The Walker circulation in the EP-EN is similar to that discussed in section 3.5.3. although this experiment shows a stronger response due to the composite of fewer and stronger El Nino˜ events.

Figure 4.17 Height-longitude cross section of Walker circulation 0-10oN, July-October for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP- EN minus ENSO neutral. The colors show mean ascent (-ω) and the vectors are mean ascent and a change in the velocity potential with respect to longitude. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

The change in vertical motion is investigated further in ﬁgure 4.18. The response looks very similar to the precipitation response shown in ﬁgure 4.15. A shift in the region of maximum ascent westward in the rCP-EN experiment is attributed to the increase

Page 119 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity in zonal tropical cyclone activity around 130oE-150oE, 20oN. The increase in tropical cyclone activity east of the Philippines in the rEP-EN experiment is due to reduced low- level subsidence around 120oE.

Figure 4.18 Vertical velocity at 500 hPa, July-October for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. The colors show mean ascent (-ω) and the vectors are mean ascent and a change in the velocity potential with respect to longitude. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

Page 120 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.5.3 Vertical wind shear

The change in vertical wind shear to the different types of El Nino˜ during JASO is shown in figure 4.19. The CP-EN shows increased vertical wind shear to the east of Japan and a decrease over the South China Sea. However, the connection to the tropical cyclone activity is not clear. Section 3.6 shows that vertical wind shear has a small inverse relationship to tropical cyclones over the entire Western North Pacific and is not an important parameter. In the rCP-EN experiment the increase of tropical cyclone density and genesis occurs across a narrow band at 20oN, where there is no change in vertical wind shear or a slight reduction in the west. In the EP-EN, the weak vertical wind shear anomaly located to the east of the date line favours tropical cyclones entering the central Pacific, which was also noted by Kim et al. (2009). The vertical wind shear pattern in the rEP- EN experiment does not simulate a decrease in vertical wind shear over the Philippines compared to EP-EN, however this parameter does not match the corresponding tropical cyclone activity suggesting other parameters are more important.

Page 121 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.19 Vertical wind shear (m s−1), July-October for (a) ENSO neutral simulation, (b) CP- EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

4.5.4 Upper-level circulation

The upper-level circulation response associated with the different types of El Nino˜ using the velocity potential and the stream function is shown in figure 4.20. Panel (b) highlights that during CP-EN, upper-level convergence is not as large over the Maritime Continent as in EP-EN, therefore the environment is less conducive to tropical cyclone activity. The strong anti-cyclonic anomaly in the central Pacific in EP-EN is related to a surface cyclonic anomaly and enhances the likelihood of tropical cyclones entering the central Pacific (see figure 4.11). The tropical eastern Pacific experiments both show stronger

Page 122 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity upper-level convergence in the Indian Ocean compared to their respective global SST experiments. The main difference between the rCP-EN experiment and rEP-EN experiment is the upper-level divergence in the Eastern Paciﬁc. The rEP-EN experiment simulates enhanced upper-level divergence eastwards than the rCP-EN experiment. As a result the stream function shows a stronger anti-cyclonic circulation eastward in the rEP-EN experiment than the rCP-EN experiment.

Figure 4.20 The 200 hPa velocity potential (1×10−6 m2 s−1) in colors and the 200 hPa stream function (1×10−6 m2) in black contours, July-October for (a) ENSO neutral simulation, (b) CP- EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

Page 123 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

4.5.5 Steering ﬂow

The steering flow is calculated as the vertically-integrated wind from 850 to 300 hPa and highlights the large-scale circulation which influences the movement of tropical cyclones. The geopotential height at 500 hPa, which represents the thickness of the atmosphere, highlights the large-scale seasonal anti-cyclones and cyclones. Figure 4.21 shows an increase of geopotential height over east China simulated in the CP-EN experiment associated with an anti-cyclonic anomaly. The anti-cyclonic anomoly inhibits the potential for tropical cyclogenesis and maintains tropical cyclone tracks at lower latitudes. An anti-cyclonic anomaly to the south east of Japan can be seen in the steering flow and explains the small recurvature of tropical cyclone tracks. During the rCP-EN experiment the anti-cyclonic anomaly strengthens and moves eastward. A cyclonic anomaly is also simulated to the east of the Philippines giving a meridional dipole circulation anomaly. The cyclonic anomaly not only acts as a source of vorticity, which increases tropical cyclogenesis (Chen and Tam, 2010), but to tunnel the tropical cyclones westward with the anti-cyclonic anomaly, towards South East Asia (Kim et al., 2011; Chen and Tam, 2010; Wang et al., 2013). The circulation pattern simulated in rCP-EN experiment matches that simulated in Jin et al. (2012). The lower SSTs in the West Pacific during EP-EN leads to an increase in geopotential height at 500 hPa. A small anti-cyclonic anomaly can be seen in the steering flow at 145oE, 25oN, which was found in Chen and Tam (2010), however it is much weaker. In the West Pacific, the rEP-EN experiment has neutral SSTs whereas the EP-EN experiment has decreased SSTs. A small cyclonic anomaly to the east of the Philippines is simulated in the rEP-EN experiment similar to the rCP-EN experiment. The circulation anomaly is associated with an increase in tropical cyclogenesis around 130oE-140oE, 15oN even though geopotential height is still anomalously positive. There is no obvious large-scale steering flow change in this simulation compared to CP-EN, therefore suggesting dynamic parameters are not important in controlling typhoon activity associated with EP-EN.

Page 124 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.21 Geopotential height at 500 hPa (contour) and steering flow calculated as the vertically-integrated wind from 850 to 300 hPa for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Pacific SST CP-EN minus ENSO neutral and (e) eastern tropical Pacific SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

4.5.6 Relative humidity

Section 3.6 in this thesis discusses the importance of relative humidity in the Western North Pacific to tropical cyclone activity. This section presents the change in relative humidity at 700 hPa simulated in the El Nino˜ experiments. A small increase in relative humidity occurs to the east of China in the CP-EN experiment, although this finding is not statistically significant (figure 4.22). A decrease of relative humidity in the region 150oE, 25oN is attributed to the decrease in tropical cyclone activity in that particular

Page 125 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity location. In the rCP-EN experiment the region of increased relative humidity to the east of China is greatly enhanced. There is no change in relative humidity in the South China sea which explains why tropical cyclones increase to the east of Philippines but do not move into the South China Sea compared to the CP-EN experiment. In the EP-EN experiment it can be seen that relative humidity changes closely resem- ble that of the tropical cyclone changes, with a decrease in the centre of the basin. A large increase in relative humidity east of the date-line is associated with the enhancement in tropical cyclone activity, which was also found in Kim et al. (2011). The region of decreased relative humidity over the Maritime Continent which extends to the centre of the basin is reduced in the rEP-EN experiment and as a result more tropical cyclones form in this region.

Page 126 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

Figure 4.22 Relative humidity at 700 hPa (%), July-October for (a) ENSO neutral simulation, (b) CP-EN minus ENSO neutral, (c) EP-EN minus ENSO neutral, (d) eastern tropical Paciﬁc SST CP-EN minus ENSO neutral and (e) eastern tropical Paciﬁc SST EP-EN minus ENSO neutral. Stippling shows where changes have a p-value < 0.01 using a student’s t-test.

4.6 Discussion

Different El Nino˜ warming patterns in the tropical Pacific can have impacts on tropical cyclone location in the Western North Pacific. The remote SST experiments can be attributed as the cause of tropical cyclone activity if the response is similar to the global SST experiments. The SSTs in this region influence tropical cyclone activity through changes in the circulation over the Western North Pacific (Jin et al., 2012). However, if the remote SST experiment is different to the global SST it is likely that local SSTs play

Page 127 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity an important role (Hong et al., 2011). For the CP-EN, both the global and remote SST experiment showed enhanced tropical cyclone activity to the east of the Philippines. Therefore, the increase in tropical cyclone activity in this region during CP-EN is largely influenced by the remote SSTs in the eastern tropical Pacific. The SSTs in this region affect the Western North Pacific through changes in the Walker circulation, which in turn influences regions of subsidence. The change in geopotential height and steering flow to the east of China is simulated in the tropical eastern Pacific experiment with slightly stronger magnitude. This is in contrast to Hong et al. (2011) who argued that local SSTs are more important. However, Hong et al. (2011) undertook EP-EN minus CP-EN experiments and therefore did not distinguish between local and remote influences for the different types of El Nino.˜ During CP-EN the SSTs do not show large regions of cooling in the West Pacific and actually show some regions of warming. Therefore the large SST anomaly to the east has a stronger influence on the atmospheric circulation. The local SST in this case acts to modulate the response from the eastern tropical Pacific. The influence of the tropical eastern Pacific SST on tropical cyclone activity associated with EP-EN is different from that associated with the global SST experiment. In contrast to CP-EN, during EP-EN the SST show strong cooling in the Western Pacific related to the stronger warming in the east. The lower SSTs in the West Pacific have a strong local thermodynamic influence which was shown as a relative humidity signature. Without the cool SSTs, relative humidity increases associated with an increase in tropical cyclone activity in the rEP-EN experiment compared to the EP-EN experiment. It is interesting to note the tropical cyclone response in the rEP-EN experiment is similar that in the rCP-EN experiment. Changing the zonal position of the SST anomalies, and hence changing the anomalous diabatic heating from convection, would alter the wave-train response to ENSO (Alexander et al., 2002). Both remote SST experiments simulate a cyclonic anomaly to the east of the Philippines. However, the location is shifted westward in the CP-EN case, in which case the SSTs anomalies are in the Nino-4˜ region. The location of the circulation anomaly results in westward zonal tropical

Page 128 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity cyclone propagation. The shift of the cyclonic anomaly eastward during EP-EN is also forced by the small cold SST anomalies in the west of the forcing region.

4.7 Conclusion

Landfalling tropical cyclones in South East Asia can have devastating impacts. While many observational studies have found more tropical cyclones move towards South East Asia during CP-EN, they are based upon a limited number of years. It is therefore difﬁcult to attribute the change in tropical cyclone activity with such a small sample size. High- resolution GCMs can now credibly simulate different aspects of tropical cyclone activity and can be used to investigate a change in tropical cyclone location due to different types of El Nino.˜ The conclusions of this study are as follows:

• The tropical cyclones changes are shown to be statistically signiﬁcant at the 95% level from climatology. However, the changes are not as signiﬁcant as previous observational studies.

• The HiGEM-HiGAM CP-EN experiment simulates an increase of tropical cyclones that move towards South East Asia, similar to previous observational studies.

• Tropical cyclone activity is shown to be dominated by the steering ﬂow during CP- EN. A large-scale anti-cyclonic anomaly over east China inﬂuences low-latitude westward tropical cyclone movement.

• The tropical eastern Paciﬁc SSTs have a large inﬂuence on the CP-EN response. The rCP-EN SST experiment simulates an enhancement of the circulation pattern associated with the CP-EN experiment, especially through the cyclonic anomaly to the east of the Philippines which, along with the anti-cyclonic anomaly to the north, acts to tunnel tropical cyclones to South East Asia.

Page 129 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity

• For the EP-EN, the associated SSTs are lower in the Western Pacific compared to the CP-EN. The low SSTs influence local thermodynamic parameters which play an important role in influencing tropical cyclone activity.

• Changes in relative humidity dominate the tropical cyclone response in the EP- EN experiments. A change in sign of relative humidity in the rEP-EN experiment explains the change in sign of tropical cyclone activity.

This study investigates the tropical cyclone response in the Western North Paciﬁc by utilising the coupled model SSTs (HiGEM) which are used to force the atmospheric component (HiGAM). The context of this assessment ﬁts in with the thesis in the analysis of HiGEM and HiGAM, however the results cannot be directly compared to observations. HiGEM has large SST biases (see section 2.7.2.1) and these are likely to have an impact on the tropical cyclone response associated with the different types of El Nino˜ compared with observations. If this work was repeated I would use observational SST such as HadISST to force the HiGAM model. This observational SST experiment would therfore be directly comparable to the real world. The remote SST experiments would be repeated using HadISST and comapred to the HiGEM study. The ultimate goal of this study would be to identify regions of importance for landfalling tropical cyclones in the South East Asia associated with different types of El Nino.˜ These regions would ideally be improved in seasonal forecast models perhaps by improving observations in that region.

4.8 Future work

A total of 10 ensembles members were run for each experiment. The precipiation response is still noisy with 10 ensemble members therefore an increase of up to 50 ensembles would provide a much larger sample size to be able to investigate the robustness of the tropical cyclone changes. Additional experiments would provide key information to address the role of local vs. remote SSTs on tropical cyclone activity associated with

Page 130 Chapter 4: The Impact of Different Types of El Nino˜ on Tropical Cyclone Activity different types of El Nino.˜ The same methodology used in this study can be applied to investigate the response when the forcing is just in the West Paciﬁc region, similar to the study by Hong et al. (2011). The local SST experiments should be compared to the eastern tropical Paciﬁc SST experiments to investigate how well each experiment is able to replicate the response simulated in the global SST forcing experiments. These experiments should also be repeated in other climate models. An observational study using this methodology would provide information on how well HiGAM is able to simulate the response associated with the different types of El Nino˜ without taking into account coupled model SST biases. With HadISST data going back to 1851 a sample size of EP-EN and CP-EN events would be comparable to the HiGEM 150 year present-day sample size. The HadISST data can also be used to further compare how well HiGEM is able to simulate the different types of El Nino.˜ Spencer et al. (2004) undertook 24 month El Nino˜ and La Nina˜ composite studies to investigate the lead and lag of atmospheric responses to remote ocean SSTs. This study could be furthered by using the methodology of Spencer et al. (2004) to investigate seasonal predictability of teleconnections associated with the different types of El Nino.˜

Page 131 Chapter 5: Tropical Cyclones and Climate Change

Chapter 5: Tropical Cyclones and Climate Change

5.1 Introduction

This chapter investigates the response of tropical cyclones to climate change in HiGEM. A literature review of tropical cyclones and climate change is ﬁrst presented. A comparison of tropical cyclone activity in idealised climate change experiments stabilised at

2×CO2 (2CO2) and 4×CO2 (4CO2) with the long integration of 150 years at present- day CO2 concentration is investigated. Changes of tropical cyclone metrics including: location; frequency; intensity; duration and structure are all discussed. The change in large-scale environmental conditions are subsequently examined in relation to the change in tropical cyclone activity. Work in this chapter has been published in Bell et al. (2013a).

5.1.1 Observed trends in tropical cyclone activity

Observational records of tropical storm and hurricanes are essential in order to discern how climatic changes have inﬂuenced tropical storms and hurricanes, and to build predictive understanding of the inﬂuence of climate on hurricanes. The study of tropical cyclones and climate change using observations is hampered by limited global data of tropical cyclones in the pre-satellite era (before 1970s; which is discussed in section 2.4.1). The short length of the observational record prevents the determination of accurate trends. However, this section discusses published current observed trends in tropical cyclone activity using the observational records available. The recent upswing of tropical cyclone activity in the North Atlantic - from 1995

Page 132 Chapter 5: Tropical Cyclones and Climate Change onwards - has received a great deal of attention. The science article by Webster et al. (2005) was published in September 2005 and concluded: tropical cyclones are becoming more intense globally. Coincidentally, a few weeks after the publication hurricane Katrina made landfall, which resulted in the media seeking to attribute global warming associated with Katrina. Many papers followed, debating the role of natural variability compared to anthropogenic warming in causing the recent increase of intense tropical cyclones (e.g. Emanuel, 2005b; Pielke et al., 2005; Trenberth and Shea, 2006). However, the lack of reliable observations of global intense tropical cyclones back to 1970 is still an issue (e.g. Lea, 2012). Additionally, tropical cyclone activity is subject to large multi-decadal variability (Klotzbach and Gray, 2008; Chan and Liu, 2012) which also hampers the detection of an anthropogenic influence in the period of good global observations. Ryan Maue from Florida State University keeps an updated web-page of current global tropical cyclone activity: http://policlimate.com/tropical/. Maue (2011) noticed a lull in global tropical cyclone activity in recent years (see figure 5.1 and 5.2). This study found the frequency and intensity of landfalling tropical cyclones shows no trend since 1970 (Weinkle et al., 2012), although landfalling storms only represent a small proportion of the total number of tropical cyclones reducing the statistical significance of the study’s findings.

Page 133 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.1 Tropical storm and hurricane frequency of the last 4-decades from http:// policlimate.com/tropical/frequency 12months.png.

Figure 5.2 Global and Northern Hemisphere Accumulated Cyclone Energy: 24 month running sums from http://policlimate.com/tropical/global running ace.png.

There have been a lack of studies investigating trends, or even understanding, of changing tropical cyclone duration and structure under climate change. Holland (2012) recently presented work using the observational record and found that tropical cyclones are now intensifying faster.

Page 134 Chapter 5: Tropical Cyclones and Climate Change

5.1.2 Tropical cyclones and climate change: A review

A recent review paper on tropical cyclones and climate change by Knutson et al. (2010a) concluded future projections, based on theory and high-resolution dynamical models, consistently indicate that greenhouse warming will cause the globally averaged intensity of tropical cyclones to shift towards stronger storms, with intensity increases of 2-11 % by 2100. Existing modelling studies also consistently project decreases in the globally averaged frequency of tropical cyclones, by 6-34 %. Balanced against this, higher resolution modelling studies typically project substantial increases in the frequency of the most intense cyclones, and increases of the order of 20 % in the precipitation rate within 100 km of the storm centre (Chauvin et al., 2006; Knutson et al., 2010a). Due to recent advances in available computing resources, General Circulation Models (GCMs) can now be run with a high enough resolution to simulate different aspects of tropical cyclone activity (e.g. Zhao et al., 2009; Smith et al., 2010; Murakami and Sugi, 2010; Manganello et al., 2012; Strachan et al., 2013). The response of tropical cyclones to changes in the large-scale environment associated with climate change can also be investigated by using high-resolution GCMs, which provide a platform to examine the mechanisms. There have been a number of previous studies investigating tropical cyclones and climate change using Atmosphere-only General Circulation Models (AGCMs) (e.g. Mc- Donald et al., 2005; Bengtsson et al., 2007a; Held and Zhao, 2011). A consistent result from these studies is that there is a general decrease in the global number of tropical cyclones the extent of which depends on the scenario (see Supplementary Material Table S1 in Knutson et al., 2010a for a summary). However, there is still a large spread in regional changes of tropical cyclone frequency between models, with some basins showing a range of ± 50 % change (Zhao et al., 2009; Sugi et al., 2009). The North Atlantic is the basin which gives the largest uncertainty in terms of tropical cyclone frequency projections although it is subject to large natural variability (see Supplementary Material Table S1 in Knutson et al., 2010a).

Page 135 Chapter 5: Tropical Cyclones and Climate Change

The models also predict an increase in tropical cyclone intensity (see Table S3 in Knutson et al., 2010a) in line with theory presented by Emanuel (1987) (see section 1.3.2.) and also those in idealized experiments by Shen et al. (2000). Tropical cyclones are predicted to become more intense with climate change due to an increase in the temperature difference between the Ocean and typical tropical cyclone outflow temperature (e.g. Vecchi et al., 2013). However, the ability of models to simulate a change in tropical cyclone intensity is largely resolution dependent (Bengtsson et al., 2007a; Murakami and Sugi, 2010). The simulated decrease in global tropical cyclone frequency has previously been related to changes in dynamical parameters, such as an increase in the vertical wind shear over the main development regions (Tsutsui, 2002; Garner et al., 2009) and also a decrease in the convective mass flux (Sugi et al., 2002; Yoshimura, 2005; Oouchi et al., 2006; Bengtsson et al., 2007a; Held and Zhao, 2011). Held and Zhao (2011) give a scaling argument stating that an increase in static stability with climate change will slow down convective mass flux from the planetary boundary layer to the upper atmosphere. Conservation of mass leads to a reduction of detrainment from moist convective systems into the surrounding environemnt. Subsequently the upper troposphere becomes drier and therefore the intrusion of drier air with vertical shear is able to suppress tropical cyclogenesis more effectively. In the study of Murakami et al. (2012a) an AGCM was integrated at 60 km (TL319) to investigate the influence of different physical Parame- terisations and sea surface temperatures (SST) on future tropical cyclone changes. They showed that variations in the projected SST pattern affected the dynamical parameters of vertical motion at 500 hPa (ω500) and relative vorticity at 850 hPa, which caused larger deviations in global tropical cyclone activity than changes in the model physics (see figure 5.3). Other studies point to changes in thermodynamics, such as the mid-troposphere saturation deficit being more important (Emanuel et al., 2008). A change in tropical cyclone genesis with climate change has recently been investigated by Nolan and Rappin (2008). Nolan and Rappin (2008) use a radiative-convective simulation on an f-plane, sugesting that vertical wind shear becomes more effective at preventing genesis as temperatures increase. Nolan and Rappin (2008) attribute this to

Page 136 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.3 Ensemble mean future changes in (a) tropical cyclone frequency and (b) tropical cyclone genesis frequency. Cross marks indicate that the difference are statistically different at the 90 % confidence level or above according to the two-sided student’s t-test and that more than 10 experiments (80 % of experiments) project mean changes of the same sign (Murakami et al., 2012a). the higher altitude of the developing mid-level vortex and increased static stability with warmer SST. This relationship is also investigated further in Held and Zhao (2011). Fur- ther research with these simulations by Rappin et al. (2010) find increased mid-level saturation deficits are the primary reason for slowing or preventing genesis. The AGCM studies typically use the ‘time slice’ method (Bengtsson et al., 1995) to allow the use of higher resolution in the atmosphere than would otherwise be possible. The time slice approach utilises SST and sea ice distributions taken from relatively low resolution coupled Atmosphere-Ocean GCMs (AOGCMs) experiments as boundary conditions. Time slice experiments typically have short integration lengths, of around ∼ 20 to 30 years. This makes it difficult to address whether the changes in tropical cyclones

Page 137 Chapter 5: Tropical Cyclones and Climate Change seen in these simulations are likely to be outside the range of natural variability. Addi- tionally, AGCMs do not allow tropical cyclones to feedback onto the SST and ocean heat content (Scoccimarro et al., 2011), which may inﬂuence future tropical cyclone activity (Emanuel, 2008). Investigating tropical cyclones and climate change using AOGCMs has received less attention due to the larger computational costs involved. The study of Tsutsui (2002) used the National Centre for Atmospheric Research (NCAR) Community Climate Model version 2 (CCM2), which had a resolution of T42, and Bengtsson et al. (2007a) used the Max Planck Institute (MPI) ECHAM5 at T63. Gualdi et al. (2008) assessed 30-years at o o present day, 2×CO2 and 4×CO2 using SINTEX-G, which had a 2 × 2 resolution ocean model and a T106 resolution atmospheric model. However, Gualdi et al. (2008) only focussed on averages of tropical cyclones across different zonal bands, which does not show how tropical cyclones will shift location within a basin. More recently Scoccimarro et al. (2011) use a higher resolution T159 atmospheric model and 2o × 2o ocean model in CMCC-MED; however, they only compare two 20-year periods of a transient experiment, 1950-1970 to 2049-2069 under the A1B IPCC scenario. Other studies have used output from low-resolution AOGCMs to allow a focus on inferred changes of tropical cyclone activity, e.g. changing genesis parameters (Kim et al., 2010). It should be noted that the AOGCM simulations give similar results of tropical cyclone frequency changes compared to the AGCM experiments, but simulate a smaller increase in intensity, related to negative feedback associated with wind-stress induced cold water upwelling (Scocci- marro et al., 2011). Future tropical cyclone intensity changes in an 20-km AGCM are shown in ﬁgure 5.4. Murakami et al. (2012b) found a large decrease in the number of weak tropical cyclones and a small increase in the most intense tropical cyclones, especially in the Northern Hemisphere.

Page 138 Chapter 5: Tropical Cyclones and Climate Change

There are few studies which investigate a change in tropical cyclone duration and structure with climate change. Figure 5.5 shows the change in power dissipation, frequency, intensity and duration from many GCMs (Emanuel et al., 2008). The figure shows an overall increase in tropical cyclone intensity and a decrease in tropical cyclone frequency, which is more variable between basins. However, there is no obvious change in tropical cyclone duration. McDonald et al. (2005) found a reduction in short lived tropical cyclones less than 5 days simulated by HadAM3. Unlike Emanuel et al. (2008), Mc- Donald et al. (2005) found longer lived storms were simulated in the North Atlantic and North East Pacific. Kanada et al. (2013) investigated tropical cyclone structure changes in a 2-km non-hydrostatic model simulation down-scaled from a 20-km AGCM simulation. They find a significant reduction in mean sea level pressure of intense tropical cyclones by 23 % and a decrease in the azimuthally averaged radius of maximum wind of 30 %, although they only used a small sample size of tropical cyclones. An increase in SST and a thinner planetary boundary layer are attributed to the increase in tropical cyclone intensity.

Figure 5.4 Annual mean tropical cyclone frequency for all years, observations (1979-2003, green lines), present-day simulations (1973-2003, blue lines) and future projections (2075-2099, red lines) for (g) global, (h) northern hemisphere, (i) southern hemisphere. Red (blue) circles indicate that the binned values for the future projection is signiﬁcantly higher (lower) than that for the present-day simulation at the 95 % conﬁdence level (two-sided student’s t-test). Bin width is 5 m s−1. Pink lines show the difference between the future projection and the present-day simulation. Adapted from (Murakami et al., 2012b).

Page 139 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.5 Change in basin-wide tropical cyclone (a) power dissipation, (b) frequency, (c) intensity, and (d) duration from the last 20 yr of the twentieth century to the last 20 yr of the twenty-second century, as predicted by running 2,000 synthetic events in each basin in each period of 20 yr. The different colours correspond to the different global climate models, as given in the legends. See Emanuel et al. (2008) for deﬁnitions of intensity and duration. The change here is given as 100 multiplied by the logarithm of the ratio of twenty-second- and twentieth-century quantities. Note that (a) is the sum of (b)-(d). The values of the changes averaged across all models are given by the numbers in black (Emanuel et al., 2008).

5.2 Tropical cyclones and climate change simulations

This section presents results from the climate change simulations with HiGEM which were introduced in section 2.7.1. It is important to characterize changes in tropical cyclone activity and to determine those which are outside the range of natural variability

Page 140 Chapter 5: Tropical Cyclones and Climate Change given by the 5×30-year present-day simulations. The results in this section focus on the difference between the 2CO2 and 4CO2 for tropical cyclone metrics of location, frequency and intensity, structure and duration.

5.2.1 Tropical cyclone location changes

The track density difference between the 2CO2, 4CO2 and control experiment are shown in figure 5.6. There is a reduction of tropical cyclone track density in the Southern Hemi- sphere and in the North West Pacific in both the 2CO2 and 4CO2, which was also found by Sugi et al. (2002); Oouchi et al. (2006) and Zhao et al. (2009). There is an increase in tropical cyclone track density in the Central North Pacific region, which agrees well with previous studies (e.g. McDonald et al., 2005; Li et al., 2010; Zhao and Held, 2011). In the North East Pacific, tropical cyclones shift towards the south west of the basin in the 2CO2. However, in the 4CO2 the tropical cyclones are greatly reduced throughout the entire basin. The change in location in the North Atlantic is shown in more detail in figure 5.6 (d and f): tropical cyclones migrate poleward in the 2CO2, whereas there is a reduction throughout the basin in the 4CO2. Tropical cyclone changes in the South Atlantic are shown to shift poleward in the 2CO2 and moreso in the 4CO2. In general, the tropical cyclones show a slight poleward shift which was also found by Zhao and Held (2011), mainly in the Northern Hemisphere, similar to the study by Emanuel et al. (2010). The stippled regions show changes which are outside the range of natural variability as represented by 5×30-year present-day simulations. Both the North West Pacific and some regions in the Southern Hemisphere show a robust reduction of tropical cyclones.

Page 141 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.6 Tropical cyclone track density, same as ﬁgure 2.6, for (a) HiGEM present-day simulation. The numbers shown in each subdomain are the climatology annual count of tropical cyclones. (b) The same as for (a) but North Atlantic (c) 2CO2 minus present-day simulation (d) North Atlantic 2CO2 minus present-day simulation (e) 4CO2 minus present-day simulation and (f) North Atlantic 4CO2 minus present-day simulation. Stippling shows where changes are outside the range of 5×30-year present-day simulations.

5.2.2 Tropical cyclone frequency changes

The global, hemispheric and regional changes in the mean annual number of tropical cyclones are shown in ﬁgure 5.7, which shows tropical cyclones decrease in frequency globally by 9 % in the 2CO2 and 26 % in the 4CO2, and also show error bars that are a measure of current climate variability. The tropical cyclone frequency changes are similar to previous studies which give a range of 6 % to 34 % (Knutson et al., 2010a). There is a larger reduction in the number of tropical cyclones in the southern hemisphere:

Page 142 Chapter 5: Tropical Cyclones and Climate Change

12 % in the 2CO2 (30 % in the 4CO2) compared to 6 % in the northern hemisphere in the 2CO2 (22 % in the 4CO2). The North Atlantic basin shows the greatest percentage reduction in tropical cyclones; however, in the control simulation this is the basin with the smallest number of storms. The mean frequency in the North East Paciﬁc in the 2CO2 is similar to the control simulation (18 year−1), although the tropical cyclones shift their location (ﬁgure 5.7). However, in the 4CO2 the tropical cyclones decrease by 40 % (11 year−1). The reduction is greater than the error bars associated with the 2CO2 and are now outside the range of 5×30-year natural variability.

Figure 5.7 Percentage change of annual tropical cyclone counts. The error bars denote the maximum and minimum 5×30-year present-day simulations. The present-day climatology is shown at the bottom of the x-axis label.

5.2.3 Tropical cyclone intensity changes

To assess the impact of climate change on the intensity of simulated tropical cyclones, the distributions of maximum intensities in terms of 850 hPa wind speed for the global basins are shown in ﬁgure 5.8. Lower level wind speed data are not used as they are signiﬁcantly damped in these simulations compared to those observed. However, the

Page 143 Chapter 5: Tropical Cyclones and Climate Change main concern regarding the use of model output wind speeds to assess the distribution of maximum intensities against observations is that the winds are not directly comparable in terms of vertical level, temporal sampling and resolution (Walsh et al., 2007; Strachan et al., 2013). There is a shift to more intense tropical cyclones along with a reduction of weak storms in the southern hemisphere. However, this is only significant in the 4CO2; the 2CO2 changes are within the 5×30-year range of natural variability (which is indicated by the error bars). The northern hemisphere basins also show a similar pattern. The North Atlantic shows the greatest shift to more intense tropical cyclones in the 4CO2, similar to the results by Oouchi et al. (2006) who used a 20 km AGCM. However, Oouchi et al. (2006) found a decrease in maximum intensity in other basins, although they used short 10-year simulations limiting the significance of this finding. Tropical cyclones become relatively more intense in the North East Pacific in the 4CO2, although there is a reduction in weak storms in both the 2CO2 and 4CO2. The increase in intensity in all basins, which are barely outside the range of natural variability in the 4CO2, is lower than in most previous studies used in Knutson et al. (2010a). The majority of studies in Knutson et al. (2010a) used AGCMs with higher resolution than HiGEM and focus on the late st 21 century period, with carbon dioxide levels of around 2×CO2. The small increase in simulated tropical cyclone intensity is in part an issue of model resolution, but also linked to atmosphere-ocean coupling. Bengtsson et al. (2007a) argued that higher resolution models are needed to simulate more realistic intensities and therefore required to assess the change in intensities with climate change however relative intensity changes in lower resolution models can still be assessed and add value to the research. Gualdi et al. (2008) found no increase in tropical cyclone intensity as measured by mean sea level pressure for their 4×CO2 simulation, albeit using a model with a coarser atmospheric component. However, assessing changes in mean sea level pressure is not a good measure compared to wind speed or vorticity as it is less sensitive, especially at low resolutions.

Page 144 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.8 Normalised distributions of storm maximum intensities in terms of 850 hPa wind speed from HiGEM for (a) North Atlantic (b) North West Pacific (c) North East Pacific (d) North Indian (e) South Pacific and (f) South Indian. The error bars denote the maximum and minimum 5×30-year present-day simulations. Note the difference in the scaling of the y-axis for (c) and (d). Bin widths are 5 m s−1.

The role of atmosphere-ocean coupling on future tropical cyclone intensity is investigated further using uncoupled-coupled (HiGAM-HiGEM) time slice experiments (see section 2.3.3), shown in ﬁgure 5.9. These results indicate an increase in tropical cyclone intensity more so in the 2CO2 experiments compared to those in the coupled simulations, although there are some differences on a regional scale. The North West Paciﬁc shows a robust response of tropical cyclones becoming more intense in the 2CO2 in the uncoupled simulations. Negative feedback associated with cold water upwelling can not occur in the uncoupled simulations. As a result tropical cyclones become more intense with increasing CO2 in the uncoupled simulations. There is also a clear shift of more intense tropical cyclones in the southern hemisphere in the 2CO2. The response in the North

Page 145 Chapter 5: Tropical Cyclones and Climate Change

East Paciﬁc is very similar in the coupled and uncoupled simulations.

Figure 5.9 Normalised distributions of storm maximum intensities in terms of 850 hPa wind speed from HiGAM-HiGEM time slice experiments for (a) North Atlantic (b) North West Pacific (c) North East Pacific (d) North Indian (e) South Pacific and (f) South Indian. The error bars denote the maximum and minimum 5×30-year present-day simulations. Note the difference in the scaling of the y-axis for (c) and (d). Bin widths are 5 m s−1.

5.2.4 Tropical cyclone duration changes

5.2.4.1 Methodology

Tropical cyclone duration is deﬁned by the sum of the number of 6-hourly time steps at which the tropical cyclone is identiﬁed as having a warm core. However, it should be noted that the warm core criteria may still pick up Easterly waves and other precursors which also have a warm core.

Page 146 Chapter 5: Tropical Cyclones and Climate Change

5.2.4.2 Results

The simulated relationship between tropical cyclone duration and intensity shows a great deal of spread in the present-day experiment as tropical cyclones intensify at different rates, shown in ﬁgure 5.10. The general relationship between duration and intensity is well simulated by HiGEM as tropical cyclones take time to intensify. For example, it is known the most intense hurricanes in the North Atlantic are usually those which travel across the entire basin, after forming near the Cape Verde Islands. However, simulated tropical cyclones do take longer to intensify than observed, which was shown by Manganello et al. (2012) using life-cycle composites, even at a model resolution of 10 km. It is not clear why this is the case.

Figure 5.10 Scatter graph of all northern hemisphere tropical cyclone data (150 year present-day simulation) for duration (number of 6 hour time steps as a warm core storm) vs. maximum 850 hPa relative vorticity. The colour denotes the number of storms at each value.

Page 147 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.11 shows a change in normalised tropical cyclone duration in the Northern Hemisphere basins in the 2CO2 and 4CO2 compared to the 5×30-year present-day simulations. HiGEM simulates no change in the longest lived storms similar to Emanuel et al. (2008) who noted there were no clear changes in simulated tropical cyclone velocity weighted duration between several models across all basins. The North Atlantic basin shows an increase in the number of short lived storms which have a lifetime of less than 4 days. The North West Pacific shows little change in tropical cyclone duration. There is an increase in short lived storms in the North East Pacific in the 2CO2 and 4CO2 which comes from a reduction in the number of storms lasting 8-10 days. The North Indian Ocean shows a reduction in short lived storms in the 2CO2, however, this basin has the smallest sample size of tropical cyclones. The reduction of tropical cyclone frequency (section 5.2.2.) was shown to come from weaker shorter lived tropical cyclones, similar to the results found in McDonald et al. (2005) using non-normalised data (not shown). However, in contrast to McDonald et al. (2005) there are no simulated longer storms in the North Atlantic and North East Pacific. The results from this study show that tropical cyclones become more intense but do not become more longer lived. This response led to a hypothesis that ‘tropical cyclones intensify more rapidly with climate change’, which was discussed in Holland (2012). This hypothesis is taken into context when discussing a change in tropical cyclone structure below.

5.2.5 Tropical cyclone structure changes

5.2.5.1 Methodology

The method used to investigate the structure of tracked features is given fully in Catto, 2009; Catto et al., 2010 and Manganello et al., 2012. Bengtsson et al. (2007a) investigated tropical cyclone structure in models with different resolution using this compositing methodology and gives detail in their Appendix A, but the process will be

Page 148 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.11 Normalised distributions of storm lifetimes in terms of days from HiGEM for (a) North Atlantic (b) Western North Pacific (c) North East Pacific (d) North Indian. The error bars denote the maximum and minimum 5×30-year present-day simulations. Note the difference in the scaling of the y-axis. Bin widths are 2 days. summarised here. The 100 most intense tropical cyclones in the Northern Hemisphere are selected based on the 850 hPa wind speed. The structure is evaluated at the point of maximum intensity along the tropical cyclone track. Fields are sampled on a 10o radial spherical cap centered on the storm location. In addition, each storm is rotated so the direction of travel is the same for all storms. This is shown in figure 5.12, taken from Catto (2009). Although the schematic shows compositing of extratropical cyclones, the methodology is exactly the same for tropical cyclones though the direction is altered westward. As discussed in section 5.2.5, to remove extra-tropical influences the maximum intensity is restricted to south of 35oN.

Page 149 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.12 Schematic of steps to the compositing. 1: Identify and select the tracks to be used. 2: Find the position of maximum intensity along the track and move the spherical polar cap grid to be centred on this position. Rotate preferred direction of spherical cap to direction of storm propagation and extract the region for averaging (Catto, 2009).

Here the focus is on the strongest tropical cyclones in the Northern Hemisphere. The limited region of analysis was to reduce CPU. As well, the Northern Hemisphere has more intense tropical cyclones than in the southern hemisphere, with a large proportion occurring in the North West Pacific (Manganello et al., 2012; also see figure 5.14). The region of analysis is restricted to those which reach maximum intensity below 30oN to avoid any influence of extratropical transition, which is discussed further below.

Page 150 Chapter 5: Tropical Cyclones and Climate Change

5.2.5.2 Results

The location of the 100 most intense tropical cyclones at maximum intensity in the 150- year present-day simulation, 2CO2 and 4CO2 are shown in figure 5.13. The locations at which storms reach maximum intensity are found more northerly in the present-day simulation, which is a 150-year integration, compared to both climate change experiments, which are 30-year integrations. The longer present-day integration increases the chance of simulating a more intense tropical cyclone, which is associated with a more poleward location of maximum intensity. It is noticeable that the locations are found well into the extra-tropics, most likely as the tropical cyclone undergoes an extra-tropical transition. As a tropical cyclones undergoes an extra-tropical transition it becomes larger and more asymmetric in response to its interaction with the baroclinic midlatitude environment and will develop a cold-core structure (Evans and Hart, 2003). The cyclone also reintensifies (Hart and Evans, 2001). In observations tropical cyclones reach their most intense point in the tropics, while they are classified as warm core storms. For example there are no intense tropial cyclones around the Philippines. It is thought that the slow intensification of tropical cyclones in GCMs is an issue of resolution (Manganello et al., 2012).

Page 151 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.13 Location of the Northern Hemisphere 100 most intense tropical cyclones in (a) the 150-year present-day simulation, (b) 2CO2 and (c) 4CO2. Deﬁned as the point of maximum 850 hPa wind speed.

This issue of tropical cyclones becoming most intense at higher latitudes is not only limited to HiGEM. The same analysis has been applied to tropical cyclones tracked in ERA-Interim and is shown in ﬁgure 5.14. Again the tropical cyclones reach their most intense in the extra-tropics, especially in the North Atlantic. It should be noted that HiGEM does simulate the observed proportion of intense storms in each basin in comparison to the frequency of storms detected by our algorithm when tracking ERA-Interim.

Figure 5.14 Location of the Northern Hemisphere 100 most intense tropical cyclones in ERA- Interim (1979-2010) at the point of maximum 850 hPa wind speed.

Page 152 Chapter 5: Tropical Cyclones and Climate Change

One way to address the issue of extra-tropical influences is to restrict the location of maximum intensity of tropical cyclones south of 35oN, similar to the methodology used by Manganello et al. (2012). The tropical cyclones which fit this criterium are now shown in figure 5.15. 30 years of the present-day simulation are used to compare with the 30-year climate change experiments. Intense tropical cyclones in the North West Pacific show a noticeable shift to the north-east, which is also present in tropical cyclones of all intensities (shown in figure 5.6). This north-eastwards shift was also found by Haarsma et al. (2013) although for the North Atlantic. HiGEM has a large negative bias for At- lantic tropical storms but if the same mechanisms are the same the results in Haarsma et al. (2013) may help explain the results in the North West Pacific here. Haarsma et al. (2013) attribute the north-eastwards shift to an expansion of warm SSTs and no change in vertical wind shear in their model for the North Atlantic. The proportion of intense tropical cyclones decreases in the North Atlantic, most likely as a result of the increasing intensity of tropical cyclones in the North West Pacific. Although some studies have looked at changing spatial patterns of tropical cyclone related precipitation (e.g. Chauvin et al., 2006), there has been less research on changing spatial patterns of tropical cyclone associated winds with climate change. The methodology to investigate a change in lower level winds is given in section 2.6.1.2. Figure 5.16 shows the system relative winds and earth relative winds (which incorporates the translational speed). Both increase in the 2CO2 and 4CO2 compared to the 5×30-year present-day simulations. The winds speeds are almost always stronger in the front-right quadrant as the forward motion is added to the near axisymmetric system-relative winds. Although the spatial pattern of wind speed associated with tropical cyclones in HiGEM are good, the wind speeds are much lower than observed. Manganello et al. (2012) investigated simulated tropical cyclone structures in a 10 km resolution AGCM with comparisons to lower resolution versions of the AGCM. The highest resolution AGCM was able to simulate 10 m tropical cyclone wind speeds of 60 m s−1. However, this was only for a composite of the 5 most intense cyclones, unlike the 100 most intense investigated in this study. HiGEM also simulates the region of maximum winds over a much broader area,

Page 153 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.15 Location of the Northern Hemisphere 100 most intense tropical cyclones in (a) a 30-year present-day simulation, (b) 2CO2 and (c) 4CO2. Deﬁned as the point of maximum 850 hPa wind speed, restricted to maximum below 35oN. with the maximum winds occurring 2.5o (250 km) away from the storm centre compared with observations. Whereas, in observations the radius of tropical cyclone maximum winds is much more narrow, on the order of 10 km (Chavas and Emanuel, 2010).

Page 154 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.16 850 hPa wind speed for the 100 most intense tropical cyclones at the peak of their intensity. (a) System-relative 5×30-year present-day simulations winds; (b) Earth-relative 5×30- year present-day simulations winds; (c) 2CO2 minus 5×30-year present-day simulations system- relative winds; (d) 2CO2 minus 5×30-year present-day simulations earth-relative winds; (e) 4CO2 minus 5×30-year present-day simulations system-relative winds; (f) 2CO2 minus 5×30- year present-day simulations earth-relative winds. Restricted to tropical cyclones that reached maximum intensity south of 35oN.

Page 155 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.16 shows that the wind speed becomes stronger in the climate change experiments, similar to the results found in Kanada et al. (2013). The maximum wind speeds increase by approximately 2 m s−1 and 5 m s−1 in the 2CO2 and 4CO2, respectively. However, in contrast to Kanada et al. (2013) the radius of maximum winds are shown to expand in size. Kanada et al. (2013) used a model with much higher resolution (2 km), which showed a decrease in the azimuthally averaged radius of maximum winds. This is a more likely projection of future tropical cyclone associated wind change. In observations as a tropical cyclone intensifies the radius of maximum winds often tighten. The increase in size of the radius of maximum winds associated with tropical cyclones in HiGEM is most likely due to the resolution. The increase in earth-relative winds compared to system-relative winds, especially in the 4CO2, indicates that the most intense tropical cyclones move faster in the climate change experiments. This is something which has not yet been previously studied. The increase of tropical cyclone translational speed can also help explain the increase in tropical cyclone intensity, given in section 5.2.3. The translational speed is an important component for tropical cyclone intensification as faster moving tropical cyclones are less likely to be influenced by the lower SSTs caused by the negative feedback associated with shear induced cold water upwelling and therefore can become more intense (Zeng et al., 2007; Mei et al., 2012). This mechanism is proposed to help explain how tropical cyclones intensify faster with climate change.

5.3 Changes in large-scale environmental conditions

The number of tropical cyclones that form each year and their maximum intensities are largely dependent on the large-scale environmental conditions. An understanding of these relationships has allowed for statistical seasonal forecasts to be developed (Ca- margo et al., 2007b). The discussion in this section focusses on the differences between the 2CO2 and 4CO2 to the 5×30-year present-day simulations, with respect to aspects of

Page 156 Chapter 5: Tropical Cyclones and Climate Change the large-scale environment known to modulate tropical cyclone activity. These include SST, vertical wind shear, tropical circulation, and the relationship between changing thermodynamic and dynamic parameters.

5.3.1 Sea surface temperature change

The change in SST during July to October (JASO), shown in figure 5.17, indicates a warming everywhere in the tropics in line with previous studies (e.g. Bengtsson et al., 2007a; Zhao et al., 2009; Xie et al., 2010; Zhao and Held, 2011). One striking feature, which occurs in both the 2CO2 and 4CO2, is a tongue of relatively cooler water in the tropical North Atlantic, as shown in figure 5.17 (b) and (d). Research by Vecchi and Soden (2007a) and Murakami et al. (2012a) suggests that for tropical cyclones, the spatial structure of future SST are more important than the absolute SST changes. This reduced SST warming in the North Atlantic, which is less than the tropical average SST warming, has a strong impact on the number of tropical cyclones that can form in the region (Lee et al., 2011). The increase in SST in the North East Pacific is related to a weakening of the Walker Circulation as noted by DiNezio et al. (2009) and Catto et al. (2011).

Page 157 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.17 Sea surface temperature (oC), June-October for (a) 2CO2 minus present-day simulation, (b) 2CO2 tropical (30oS-30oN) mean sea surface temperature change, the number at the top right shows the tropical mean anomaly (TM). (c) 4CO2 minus present-day simulation, (d) 4CO2 tropical mean sea surface temperature change. The present-day simulation climatology is shown in black contours in (a). Signiﬁcance was found everywhere in (a) and (c) and stippling is therefore not shown.

In addition, SSTs are shown to increase more in the Northern Hemisphere than the Southern Hemisphere during their peak tropical cyclone seasons, as was also found by Vecchi and Soden (2007b). Figure 5.18 shows that in the Southern Hemisphere SST increases on average by approximately 1 oC in the 2CO2 during the peak tropical cyclone season. In the 4CO2 the region of maximum warming occurs in the Central Paciﬁc. The reduced warming in the Southern Hemisphere compared to the Northern Hemisphere relate to a greater reduction of tropical cyclone frequency with climate change (see section 5.2.2). The reduced warming leads to a reduction of relative SST warming compared to the rest of the tropics. Similar to the results for the North Atlantic it is likely a reduction in relative SST warming will increase the atmospheric lapse rate and therefore make the environment less conducive for tropical cyclones to form.

Page 158 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.18 Sea surface temperature (oC), December-March for (a) 2CO2 minus present-day simulation, and (b) 4CO2 minus present-day simulation. The present-day simulation climatology is shown in black contours in (a). Signiﬁcance was found everywhere and stippling therefore not shown.

5.3.2 Precipiation change

The simulated change in northern hemisphere JASO precipitation in the 2CO2 and 4CO2 is given in figure 5.19. The patterns looks very similar to precipitation change simulated in the CMIP3 models (e.g. DiNezio et al., 2009; Xie et al., 2010). There is an increase in precipitation across most of the Pacific, excluding the North East Pacific. The precipitation in the Bay of Bengal also increases in the 2CO2 and more so in the 4CO2. The inter-tropical convergence zone (ITCZ) is shown to strengthen in the tropical North East Pacific as precipitation increases locally by 7 mm day−1 in the 4CO2. The large increase in precipitation in the equatorial Pacific was also noted by Xie et al. (2010). Similarly, DiNezio et al. (2009) found most models simulate an increase in equatorial Pacific precipitation, as well as in the North East sub-tropical Pacific. The North Atlantic shows a decrease in precipitation of approximately 1-2 mm day−1 in the 2CO2.

Page 159 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.19 Precipitation change (mm day−1), July-October for (a) 2CO2 minus present-day simulation, and (b) 4CO2 minus present-day simulation. The present-day simulation climatology is shown in black contours in (a). Stippling shows where changes are outside the range of 5×30- year present-day simulations.

5.3.3 Mid-level relative humidity change

Mid-level humidity is important for tropical cyclone genesis as stated in section 1.3.1. Figure 5.20 shows the change in relative humidity at 700 hPa during JASO. The mid- level humidity change shows large-scale changes compared to precipitation change. The region of increased relative humidity to the east of Hawaii is associated with the increase in tropical cyclone activity (shown in figure 5.6). This enhancement of relative humidity in the North East sub-tropical Pacific is also simulated by other CMIP3 models (Vecchi and Soden, 2007b). The change in mid-level relative humidity in the North West Pacific is small, even though tropical cyclones are significantly reduced in this basin.

Page 160 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.20 Relative humidity change at 700 hPa (%), July-October for (a) 2CO2 minus present- day simulation, and (b) 4CO2 minus present-day simulation. The present-day simulation climatology is shown in black contours in (a). Stippling shows where changes are outside the range of 5×30-year present-day simulations.

5.3.4 Circulation change

Catto et al. (2011) showed that HiGEM simulates greater warming in the upper tropical atmosphere compared to the lower tropopause, similar to other GCM studies (Held and Soden, 2006). The upward mass flux decreases because the top of atmosphere radiation cannot compensate for the increase in lower-tropospheric specific humidity and associated latent heating with a fixed mass flux (Held and Soden, 2006). This has previously been discussed as the cause of the simulated reduction in global tropical cyclone frequency (e.g. Sugi et al., 2002; Yoshimura, 2005; Oouchi et al., 2006; Held and Zhao, 2011; Sugi et al., 2012). Held and Zhao (2011) argue that a reduction of mean ascent at 500 hPa makes the mid-to-upper level environmental air drier, as there is less detrainment of moist air from convective systems. This reduces the relative humidity at 700 hPa and

Page 161 Chapter 5: Tropical Cyclones and Climate Change the drier air is able to suppress tropical cyclone genesis more effectively via entrainment and downdrafts. The velocity potential and stream function are useful parameters which give an integrated view of the large-scale circulation. The stream function emphasises the anomalous wave propagation, whereas the velocity potential shows the forcing of these waves. The change in velocity potential and stream function at 200 hPa for JASO are given in figure 5.21. The changing velocity potential highlights the significance of large-scale convergence in the Central Pacific, especially in the 4CO2. The change in stream function shows that the warming in the Central Pacific sets up a Gill type response (Gill, 1980) as two large-scale anticyclones straddle the equator to the west of maximum large-scale convergence associated with Rossby waves. This relationship was also noted by Li et al. (2010) and used to explain the increase in occurrence of tropical cyclones in the Central Pacific and the reduction in the North West Pacific. Additionally, the pattern in the 4CO2 looks like an eastward shifted El Nino˜ response (section section 3.5.5).

Page 162 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.21 Velocity potential (m2 s−1) and stream function (m2) change at 200 hPa, July- October for (a) 2CO2 minus present-day simulation, and (b) 4CO2 minus present-day simulation. Stippling shows where changes are outside the range of 5×30-year present-day simulations.

When assessing changes in the mean zonal circulation, figure 5.22 shows that there is a weakening of the rising branch of the Walker circulation in the North West Pacific, particularly in the 4CO2. The weaker mean ascent reduces the convective mass flux, which makes the environment less favorable to tropical cyclone development. In contrast, the descending branch in the Central Pacific shows anomalous ascent, which favours the enhanced development of tropical cyclones in this region (see also Li et al., 2010).

Page 163 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.22 Height-longitude cross section of Walker circulation, 0-10oN June-October for (a) present-day simulation; (b) 2CO2 minus present-day; (c) 4CO2 minus present day. The colors show mean ascent (-ω) and the vectors are mean ascent and a change in the velocity potential with respect to latitude. Stippling shows where changes are outside the range of 5×30-year present-day simulations.

Whilst previous studies have focussed on possible changes in the Walker circulation, a changing Hadley circulation has received less attention. However, this is perhaps even more relevant as tropical cyclones tend to form north of 10oN. To also address a change in the vertical mass flux in the meridional direction, an investigation of the change in the Hadley cell circulation is shown in figure 5.23. Regions of ascent only are used in the zonal mean to reflect regions that are conducive to tropical cyclone development. HiGEM simulates a weakening of the Intertropical Convergence Zone (ITCZ) with an increase in ascent either side of 9oN, related to a weakening of the Hadley circulation similar to the results found by Lu et al. (2007) and Kang and Lu (2012). The weakening suppresses tropical cyclone activity in low latitudes around 10oN, a feature which is more pronounced in the 4CO2. Although the magnitude of the change is much less than the weakening of the Walker circulation, the weakening will also aid the overall reduction

Page 164 Chapter 5: Tropical Cyclones and Climate Change of tropical cyclones. The anomalous ascent north of 16oN, may be responsible for the poleward migration of tropical cyclones. A widening of the Hadley cell is related to a rise in the tropopause under global warming and a poleward shift of mid-latitude eddies (Lu et al., 2007; Schneider et al., 2010).

Figure 5.23 Height-latitude cross section of Hadley circulation, 0-360oE, June-October for: (a) present-day simulation; (b) 2CO2 minus present-day; (c) 4CO2 minus present day. The zonal average includes regions of updrafts only using the 500 hPa level. The colors show mean ascent (- ω) and the vectors are mean ascent and a change in the velocity potential with respect to longitude. Stippling shows where changes are outside the range of 5×30-year present-day simulations.

Page 165 Chapter 5: Tropical Cyclones and Climate Change

5.3.5 Vertical wind shear change

Vertical wind shear is defined as the magnitude of the vector difference between winds at 850 and 200 hPa. The weakening of the Walker circulation has previously been related to the increase in vertical wind shear over the main development region (MDR) of the North Atlantic (Vecchi and Soden, 2007b). This response is similar to that which occurs during an El Nino˜ event. The increase in vertical wind shear over the Caribbean is also likely to be associated with the tongue of relatively cooler water in the tropical North Atlantic compared to the warmer SST to the north (Zhang and Delworth, 2006). This SST pattern creates a larger meridional temperature gradient, which in turn leads to stronger vertical wind shear over the region via thermal wind balance. There is a small decrease of vertical wind shear to the east of North America which would help explain the increase of tropical cyclone activity in the 2CO2 in this region. Vertical wind shear change in the North East Pacific in the 2CO2 is strongly related to the change in tropical cyclone location (shown in figure 5.24 and figure 5.6), as the tropical cyclones move to the south west of the basin. In the 4CO2, however, the spatial extent of increased vertical wind shear expands across the entire basin and is likely to be responsible for the reduced tropical cyclone frequency. The large increase of vertical wind shear is also likely responsible for the overall reduction of tropical cyclone activity in the North Atlantic. There is a reduction in vertical wind shear in the Central North Pacific, which would favour tropical cyclone development, both in the 2CO2 and 4CO2.

Page 166 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.24 Vertical wind shear m s−1, July-October for (a) 2CO2 minus present-day simulation, and (b) 4CO2 minus present-day simulation. The present-day simulation climatology is shown in black contours in (a). Stippling shows where changes are outside the range of 5×30-year present-day simulations.

The change of vertical wind shear during December-March is shown in figure 5.25. There is no obvious change of vertical wind shear in the 2CO2 compared to the 5×30- year present-day simulations; however, tropical cyclones are still less frequent (shown in figure 5.6). It is therefore unlikely that the change of vertical wind shear has a large influence in future southern hemisphere tropical cyclone activity in the 2CO2. In the 4CO2 the vertical wind shear pattern remains the same. Vertical wind shear is increased yet again around 30oS. There is a large reduction of vertical wind shear in the tropical Pa- cific. This arises due to a weakening of the Walker circulation which is discussed above. HiGEM has a bias to generate too many storms in the South Pacific Convergence Zone. The increase in vertical wind shear in this region and to the east is likely to be responsible causing the reduction of tropical cyclone frequency. Vecchi and Soden (2007b) notes that vertical wind shear increases in the Southern Hemisphere more so than in the Northern Hemisphere and is responsible for the greater reduction of tropical cyclone frequency (see figure 5.7).

Page 167 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.25 Vertical wind shear m s−1, December-March for (a) 2CO2 minus present-day simulation, and (b) 4CO2 minus present-day simulation. The present-day simulation climatology is shown in black contours in (a). Stippling shows where changes are outside the range of 5×30-year present-day simulations.

Page 168 Chapter 5: Tropical Cyclones and Climate Change

5.3.6 Thermodynamic vs. dynamic inﬂuences

Figure 5.26 shows the percentage change of other large-scale environmental conditions that are important for simulated changes in tropical cyclone activity over the MDRs. This includes, using the terminology of Held and Zhao (2011), thermodynamic variables of SST, tropical relative SST, relative humidity at 700 hPa, and precipitation. In addition, dynamic parameters including mean ascent at 500 hPa (-ω500; a proxy for deep convection) and vertical wind shear. The North Atlantic shows a linear reduction in tropical cyclone frequency between the 2CO2 and 4CO2. The mechanism that reduces tropical cyclone frequency is similar to that which occurs during an El Nino˜ event. Warming in the tropical Pacific, compared to the North Atlantic, shown by the tropical relative SST, increases the upper tropospheric temperature across the tropics, and decreases the lapse rate over the North Atlantic, increasing the vertical stability. This makes the environment less favorable for genesis (Tang and Neelin, 2004). The reduction in -ω500 explains the reduction in tropical cyclone frequency, which is discussed further in Held and Zhao (2011). HiGEM also simulates an increase in vertical shear in the Caribbean, which expands in size and magnitude in the 4CO2. Garner et al. (2009) find vertical wind shear to be more important than large-scale subsidence, in contrast to the results reported here, although they use a regional model as opposed to the global approach used in this study. It should be noted the North Atlantic basin has a poor mean-state bias of vertical wind shear and tropical cyclone activity shown in figure 5.27. The large percentage reduction in tropical cyclone activity occurs due to the small number of tropcial cyclones which form in the present- day experiment. The North West Pacific shows a robust reduction in tropical cyclones for both the

2CO2 and 4CO2, even though vertical wind shear decreases. The small decrease in -ω500 of 3 % (12 %) in the 2CO2 (4CO2) is of the same sign as the tropical cyclone changes, unlike the other variables. This is shown further in the changing Walker circulation (ﬁg- ure 5.22). Figure 5.28 shows the nonlinear relationship between large-scale parameter

Page 169 Chapter 5: Tropical Cyclones and Climate Change changes and tropical cyclones changes. A change in vertical wind shear is unlikely to have an affect on tropical cyclone frequency as it was discuss in chapter 3 that vertical wind shear is not an important parameter in this basin. The change in -ω500 clearly shows a reduction in tropical cyclone frequency. This is plausible as panel d) shows the other datasets have a realtionship that the number of tropical cyclones which form per season is related to the season acerage of -ω500. There is a significant nonlinear change in tropical cyclone frequency and track density between the 2CO2 and 4CO2 in the North East Pacific. Most large-scale parameters remain relatively unchanged in the 2CO2 and the tropical cyclones show a slight reduction in frequency, although not outside the range of 5×30-year control variability (18 year−1). The magnitude of vertical wind shear increases by 46 % over the MDR in the 4CO2 and −1 tropical cyclones are greatly reduced to 11 year . The change in -ω500 is small in both experiments, therefore it is likely that the tropical cyclones decline in number as a result of an increase in vertical wind shear in this basin. In the North Indian Ocean tropical cyclones initially increase in frequency in the 2CO2 but do not change in the 4CO2. However, track density increases in the MDR in both experiments, as tropical cyclones tend to form more in the Bay of Bengal. There is an increase in -ω500 and also relative humidity at 700 hPa, relating to the tropical cyclone changes. Both the South Pacific and South Indian basins have a similar tropical cyclone response. The reduction of tropical cyclones in the South Pacific in the 2CO2 is related to a small decrease in -ω500 and a slight increase in vertical wind shear. In the South Indian basin in the 2CO2 there is a small decrease in mean ascent at 500 hPa. A further reduction of tropical cyclones in the 4CO2 in both basins is related to less -ω500 and an increase in vertical wind shear.

The dynamical parameters of -ω500 and vertical wind shear show large percentage changes in the climate change experiments compared to the present-day simulations. The changes are of the same sign as the tropical cyclones changes in most basins. On the other hand, local SST and precipitation changes can be of the opposite sign to the

Page 170 Chapter 5: Tropical Cyclones and Climate Change tropical cyclone changes. It is known that a small change in the tropical relative SST pattern drives larger changes in dynamical parameters, which ultimately inﬂuences tropical cyclone activity (Zhao and Held, 2011).

Page 171 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.26 Percentage change of ocean-only large-scale monthly-mean parameters. The area deﬁned is shown above the plot. (a) - (d) for June-October, (e) and (f) are for December-March. The error bars show the maximum and minimum variability of the 5×30-year present-day simulations. The climatologies in the control simulation are shown at the bottom. SST (oC), tropical relative SST (30oS-30oN; oC), relative humidity at 700 hPa (RH, %), precipitation (ppt, mm

−1 −1 −1 day ), mean ascent at 500 hPa (-ω500,Pas ), vertical wind shear (vws, m s ), tropical cyclone track density (TCden, same as ﬁgure 1), tropical cyclone frequency (TCfreq, same as ﬁgure 3).

The percentage change of -ω500 in the NATL is shown as minus because the region has subsiding air in the present-day simulation. Note the difference in the scaling of the y-axis.

Page 172 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.27 Scatter plot of tropical cyclone counts versus large-scale environmental parameters for the North Atlantic, averaged over the region 275-340oE, 5-20oN during July-October. Red symbols represent the climatology of the 2CO2 experiment, green symbols represent the climatology of the 4CO2 experiment black symbols represent the climatology of the present-day experiment for (a) SST, (b) vertical wind shear, (c) relative vorticity, (d) mean ascent at 500 hPa and (e) relative humidity at 700 hPa. Observations are the cross, ERA-Interim the diamond, HiGAM the circle and HiGEM the triangle. The line styles distinguish between the CO2 experiments. The error bars denote the 90 % conﬁdence interval.

Page 173 Chapter 5: Tropical Cyclones and Climate Change

Figure 5.28 Scatter plot of tropical cyclone counts versus large-scale environmental parameters for the Western North Paciﬁc, averaged over the region 110-170oE, 5-20oN during July-October. Red symbols represent the climatology of the 2CO2 experiment, green symbols represent the climatology of the 4CO2 experiment black symbols represent the climatology of the present-day experiment for (a) SST, (b) vertical wind shear, (c) relative vorticity, (d) mean ascent at 500 hPa and (e) relative humidity at 700 hPa. Observations are the cross, ERA-Interim the diamond, HiGAM the circle and HiGEM the triangle. The line styles distinguish between the CO2 experiments. The error bars denote the 90 % conﬁdence interval.

Page 174 Chapter 5: Tropical Cyclones and Climate Change

5.4 Summary and conclusions

Tropical cyclones simulated in a high-resolution AOGCM with a 150-year present-day control simulation have been compared against idealized climate change simulations of stabilized 2CO2 and 4CO2. Changes in tropical cyclone location, frequency, intensity and lower-level wind structure are all considered. The conclusions on the changes of tropical cyclone activity are as follows:

• Tropical cyclones show a poleward shift in the Northern Hemisphere in the 2CO2 and more so in the 4CO2, associated with a poleward shift of the rising Hadley cell branch.

• There is an increase of tropical cyclones in the North Central Paciﬁc in the both the 2CO2 and 4CO2 which is related to a weakening of the Walker circulation (consistent with the results of Li et al., 2010). This may increase the future risk of tropical cyclones making landfall in Hawaii.

• Tropical cyclones decrease in frequency globally by 9 % and 26 %, in the 2CO2 and 4CO2, respectively.

• The North Atlantic is the basin which shows the largest decrease in tropical cyclone frequency, although it shows the greatest shift to more intense tropical cyclones. This basin is also the location of maximum current underestimation.

• The increase in tropical cyclone intensity only becomes noticeable in the 4CO2 compared to the 5×30-year present-day simulations.

• The shift to more intense tropical cyclones in the 4CO2 is robust across all basins and is consistent with a shift to fewer weaker tropical cyclones.

• The increase in intensity was found to be less than found in previous studies (Knut- son et al., 2010a), due to the atmosphere-ocean coupling in HiGEM. This was

Page 175 Chapter 5: Tropical Cyclones and Climate Change

further highlighted by uncoupled simulations which showed tropical cyclones becoming more intense in the 2CO2.

• Tropical cyclones in the Northern Hemisphere do not become longer lived in a warmer world; the reduction of tropical cyclones mainly comes from the reduction of short lived storms.

• An increase of earth-relative winds to system-relative winds, indicates that tropical cyclones move faster when simulated in the climate change experiments. This has not been previously studied.

The reduction arises in the 2CO2 due to a decrease in mean ascent at 500 hPa, similar to what is found in Held and Zhao (2011). The further reduction of tropical cyclone frequency in the 4CO2 is thought to be connected to a large increase in vertical wind shear, as well as further reduction of mean ascent at 500 hPa, especially in the southern hemisphere. A weaker Walker circulation drives changes in tropical cyclone location: an increase in the North Central Pacific and eastward shift in the North West Pacific. Tropi- cal cyclones respond to an increase in vertical wind shear in the North East Pacific, with a shift to the south west, which was also found by Vecchi and Soden (2007b). A poleward shift of ascent associated with the northern hemisphere Hadley cell, during peak tropical cyclone season, helps to explain the poleward migration of tropical cyclones. It has been shown that future tropical cyclone activity does not follow local SST and precipitation changes. The threshold of tropical convection (often quoted at 26oC; see section 1.3.1) has recently been found to be dependent on global mean temperature (Johnson and Xie, 2010). An enhanced warming of the upper troposphere with climate change may increase the threshold over which tropical cyclogenesis can occur. The upper troposphere is well mixed and influenced by tropical average SST (Vecchi and Soden, 2007a). Local SSTs, which warm less than the tropical average, were shown to have the largest reduction in future tropical cyclone activity, as found by Lee et al. (2011). How- ever, an increase in local SST in the North East Pacific is not related to an increase in tropical cyclone activity. Some basins show an increase in precipitation but do not show

Page 176 Chapter 5: Tropical Cyclones and Climate Change an increase in tropical cyclone activity. This study does not address other changing environmental parameters, such as the mid-troposphere saturation deﬁcit (Emanuel et al., 2008) due to the limited number of required parameters that were stored as pressure level diagnostics, although other studies show this changes homogeneously and cannot explain regional changes in future tropical cyclone activity (Sugi et al., 2012). Previous limitations of computer resources have not allowed for climate change experiments to be compared to these types of integrations. The increase in atmospheric

CO2 simulations reveals robust changes of tropical cyclone activity.

5.5 Future work

Whilst the tropical cyclone metrics of location, frequency and intensity simulated in HiGEM have been well validated against observations in Strachan et al. (2013), the metrics of tropical duration and lower-level wind structure require further validation against observations and tropical cyclones tracked in reanalysis data. The focus of tropical cyclone structure changes can also include vertical changes. Unfortunately, HiGEM data was only available at 6-hourly temporal frequency for the pressure levels of 850, 500 and 200 hPa. Data is required on more levels to attain a good representation of simulated tropical cyclone structure and how it might change in a warmer world. An objective threshold for the location of maximum intensity in the tracked cyclones needs to be implemented which is based on physics. This could be done using the method of Hart (2003). It would also be important to investigate future precipitation associated with tropical cyclones, however the data was not available at 6-hourly temporal frequency. Moreover, further work is needed to validate tropical cyclone structure in HiGEM to other resolution models and observations (e.g. Chavas and Emanuel, 2010; Reed and Jablonowski, 2010; Reed and Jablonowski, 2012; Manganello et al., 2012; Kanada et al., 2013). An objective threshold for the maximum

Page 177 Chapter 5: Tropical Cyclones and Climate Change

However, there is still a lack of good observations of tropical cyclone structure due to the difﬁculties of obtaining measurements. Further work with composite life-cycles could be used to investigate the rate at which tropical cyclones intensify and whether this will increase in the climate change experiments. Longer integrations of the 2CO2 and 4CO2 of 150 years each - to match the present- day simulation - would be required to further investigate robust tropical cyclones changes with climate change. Tropical cyclones show large multidecadal variability in observations as well as in HiGEM, and there is a possibility that within a 30-year simulation the Atlantic Meridional Overturning Circulation (AMOC) was in a particular phase; or there was an usually large number of El Nino˜ events. The results in the 2CO2 and 4CO2 may therefore slightly change if the simulations were run for many more years. The experiments investigated here assess the tropical cyclone response to an increase in CO2, and do not assess the response to interactive aerosol forcing. Recent work by Booth et al. (2012) highlight the importance of aerosols as a driver for tropical North At- lantic climate variability, and therefore tropical cyclone variability. Future research using interactive aerosol forcing in HiGEM, could be exploited to investigate future tropical cyclone activity.

Page 178 Chapter 6: Conclusions

Chapter 6: Conclusions

6.1 Introduction

Tropical cyclones can cause substantial loss of life and an improved understanding of storm variability and their response to climate change can help inform preparation and future adaptation. A recent example, hurricane Sandy, was the deadliest and most de- structive hurricane of the 2012 North Atlantic hurricane season. At least 286 people were killed as a result of hurricane Sandy and it was the second costliest hurricane to strike the US with an estimated US$65 billion worth of damage (ICAT, 2013). Tropical cyclones are also important elements of the climate system. Tropical cyclones remove heat and moisture from the ocean, which affects the formation of deep water that drives ocean circulation (Sriver, 2010). It is therefore vital that climate models are able to ad- equately simulate tropical cyclone activity to increase conﬁdence in projections of how their activity will change from interannual to inter-decadal time scales. The evaluation of climate models in their ability to simulate tropical cyclone activity will contribute to further improvement of climate models. The study of dynamically simulated tropical storms: their natural variability and response to climate change, has been investigated with the aim of answering the research questions posed in chapter 1:

1. How does the El Nino˜ Southern Oscillation (ENSO) inﬂuence global tropical cyclone activity?

2. What is the role of the eastern tropical Paciﬁc on tropical cyclone activity associ-

Page 179 Chapter 6: Conclusions

ated with different types of El Nino.˜

3. What is the response of tropical cyclones to climate change?

Climate model resolutions are typically of the order of 1o-3o in the atmosphere. How- ever, due to recent advances in available computer resources climate model resolution has increased and climate models are becoming widely used to study tropical cyclone activity. GCMs can be used to investigate how large-scale climate variability can influence tropical cyclone activity and support quantification of predictability. To accurately study changes in tropical cyclones a GCM has to run with as high a resolution as possible for multi-centennial integrations (Strachan et al., 2013). The model used in this study is HiGEM, which has a resolution of N144 (1.25o latitude × 0.83o longitude; grid spacing of approximately 90 km at 50oN). The HiGEM control simulation was completed using present-day radiative forcing for an integration length of 150 years. This study has identified opportunities and limitations in simulating global tropical cyclone activity in GCMs. Firstly, the mean state biases are taken into considerations to investigate the simulation of the global ENSO-tropical cyclone teleconnection in a high- resolution Atmosphere-Ocean General Circulation Model (AOGCM) with comparisons to an Atmosphere-only General Circulation Model (AGCM), reanalysis data and observations. Secondly, a modelling framework has been used to investigate how different types of El Nino˜ affect tropical cyclone activity in the Western North Pacific. The final chapter investigates how tropical cyclones may change due to the influence of increased atmospheric CO2.

6.2 Synthesis of results

In this section the three questions from this thesis are addressed using the results from chapters 3 to 5.

Page 180 Chapter 6: Conclusions

6.2.1 How does El Nino˜ Southern Oscillation (ENSO) inﬂuence global tropical cyclone activity?

It has been long observed that the ENSO has a large influence on global tropical cyclone variability. It is important to evaluate the ability of GCMs to simulate realistic ENSO associated tropical cyclone teleconnections for seasonal forecasting and before predictions are made for tropical cyclones and climate change using GCMs (Shaman and Maloney, 2011). The ability of the models used in this study to represent the global ENSO-tropical cyclone teleconnection was assessed in chapter 3. The model results were systematically compared to observations and reanalysis. The coupled model (HiGEM) accurately simulates ENSO-tropical cyclone teleconnections in the Pacific and Indian Oceans. During an El Nino˜ year tropical cyclones shift toward the Central Pacific in the Western North Pa- cific and South Pacific. In the Indian Oceans tropical cyclones increase in number during La Nina˜ years. However, HiGEM does not simulate the expected ENSO-tropical cyclone teleconnection in the North Atlantic leading to a decrease in tropical cyclones during La Nina˜ years and an increase during El Nino˜ years. HiGEM, along with a large majority of coupled climate models, has large SST biases in the eastern equatorial Pacific. This SST bias leads to ENSO associated SSTs extending further westwards compared to observations as does the Walker circulation response, which leads to a simulation of enhanced tropical cyclone variability in the Central Pacific. Additionally, the upper-level branch does not extend eastwards enough to influence vertical wind shear over the Caribbean Sea and therefore HiGEM does not simulate tropical cyclone variability in the North At- lantic associated with this. The uncoupled model (HiGAM), forced by observed SSTs, simulates the global ENSO-tropical cyclone teleconnection more accurately than HiGEM compared to observations. However, HiGAM over pronounces the variability of tropical cyclone activity between El Nino˜ and La Nina˜ years. One of the likely causes is the fact that the lack of atmosphere-ocean feedback causes an effective infinite source of heat for the atmosphere, which results in greater precipitation variability. Mid-level relative humidity was

Page 181 Chapter 6: Conclusions found to be better simulated over the Western North Pacific in HiGAM than HiGEM and explains the improvement of the ENSO-tropical cyclone teleconnection in this region. HiGAM was also shown to simulate the dipole of vertical wind shear in the North East Pacific and the Caribbean Sea with ENSO as observed in ERA-Interim. The dipole of vertical wind shear in the Caribbean Sea and North East Pacific explains why tropical cyclone activity shows a dipole pattern in the North Atlantic and North East Pacific with ENSO.

6.2.2 What is the role of the eastern tropical Paciﬁc on tropical cyclone activity associated with different types of El Nino?˜

The previous work chapter showed that HiGEM was unable to capture the ENSO-tropical cyclone teleconnection in the North Atlantic. Therefore this chapter investigates how different types of El Nino˜ affect tropical cyclone activity in the Western North Pacific only. In addition, HiGAM simulates the ENSO-tropical cyclone teleconnection much better in the Western North Pacific than HiGEM, which is why experiments were undertaken using HiGAM forced by HiGEM SSTs. In this study idealised atmosphere-only model simulations are used to attribute the response associated when specifying SST of the different types of El Nino˜ in the eastern tropical Pacific only. The experiments are: a composite of central Pacific El Nino˜ (CP-EN) SSTs; a composite of eastern Pacific El Nino˜ (EP-EN); remote central Pacific El Nino˜ (rCP-EN); same for CP-EN but SSTs are extracted from 160oE-East Pacific coast, 10oS-20oN and neutral SSTs elsewhere; remote eastern Pacific El Nino˜ (rEP-EN); same for EP-EN but SSTs are extracted from 160oE- East Pacific coast, 10oS-20oN and neutral SSTs elsewhere. The HiGEM-HiGAM CP-EN experiment simulates an increase of tropical cyclones that move towards South East Asia, similar to previous studies. This response is explained by the anti-cyclonic anomaly which forms over east China and influences low- latitude westward tropical cyclone movement. The tropical cyclone response simulated in the EP-EN experiments match those found in the previous work chapter. The tropi-

Page 182 Chapter 6: Conclusions cal cyclone response in the CP-EN experiments is largely replicated by prescribing the SSTs in the tropical eastern Pacific only. The rCP-EN simulate an enhancement of the circulation pattern associated with the CP-EN experiment, especially through the cyclonic anomaly to the east of the Philippines. The cyclonic anomaly, along with the anti-cyclonic anomaly to the north, acts to tunnel tropical cyclones to South East Asia. For the EP-EN, the associated SSTs are lower in the Western Pacific compared to the CP- EN. The low SSTs influence mid-level relative humidity which play an important role in influencing tropical cyclone activity. So from these results the role of the eastern tropical Pacific on tropical cyclone activity associated with the different types of El Nino˜ is to influence circulation patterns over the Western North Pacific.

6.2.3 What is the response of tropical cyclones to climate change?

The ability of HiGEM and HiGAM to simulate the aspects of natural variability of tropical cyclone formation in various regions of the globe indicates that they are capturing some of the essential physical relationships governing the links between the global climate system and tropical cyclones. A comparison of tropical cyclone activity in idealised climate change experiments stabilised at 2×CO2 (2CO2) and 4×CO2 (4CO2) was compared to the present-day integration which was investigated in chapter 3. More tropical cyclones were found to form in the North Central Paciﬁc, in both the 2CO2 and 4CO2, this is related to a weakening of the Walker circulation, a result that agrees with previous studies (Vecchi et al., 2006; Li et al., 2010). Tropical cyclones were shown to decrease in frequency globally by 9 % and 26 %, in the 2CO2 and 4CO2, respectively. The increase in tropical cyclone intensity only becomes noticeable in the 4CO2 compared to the 5×30-year present-day simulations. The small increase in intensity was partly related to atmosphere-ocean coupling. This was further highlighted by uncoupled simulations which showed tropical cyclones becoming more intense in the 2CO2. It has been shown that future tropical cyclone activity does not follow local SST and precipitation changes. The reduction arises in the 2CO2 due to a decrease in mean ascent

Page 183 Chapter 6: Conclusions at 500 hPa. The further reduction of tropical cyclone frequency in the 4CO2 is thought to be connected to a large increase in vertical wind shear, as well as further reduction of mean ascent at 500 hPa, similar to what is found in Bengtsson et al. (2007a); Held and Zhao (2011) and Sugi et al. (2012). Tropical cyclones respond to an increase in vertical wind shear in the North East Paciﬁc, with a shift to the south west, which was also found by Vecchi and Soden (2007b).

6.3 Suggestions for further investigation

In this section further work is suggested to improve our ability to answer the research questions posed in this thesis.

6.3.1 How does El Nino˜ Southern Oscillation (ENSO) inﬂuence global tropical cyclone activity?

Reanalysis datasets can be used to complement model and observation comparisons. Identifying tropical cyclones in multiple reanalyses leads to a greater understanding of the observational uncertainty. The tracking algorithm could be applied to other reanalysis datasets beyond ERA-Interim, for example as in Schenkel and Hart (2011). The diagnosis of tropical cyclones and large-scale environmental conditions undertaken in this study should be expanded to other GCMs. The capability of models to distinguish the response of tropical cyclones to ENSO is important for seasonal prediction systems. The Met Ofﬁce have a suite of uncoupled models with different resolutions using HadGEM3 ranging from N96 (130 km) to N512 (25 km) (Mizielinski et al., 2013; Vi- dale et al., 2013 (in prep)). It would be interesting to repeat this analysis for the different model resolutions. Exactly how important resolution is for capturing the ENSO-tropical cyclone teleconnection in the North Atlantic is of great interest. However, this should

Page 184 Chapter 6: Conclusions be expanded to a global analysis using coupled models as well. The multi-model study would provide a much larger ENSO sample size and therefore a more robust tropical cyclone response in the multi-model mean. Malcolm Roberts at the Met Ofﬁce has recently being running a 12 km AGCM without convective parameterisation. Along with other very high resolutions models, these studies can be used to assess the role of physics in the simulation of tropical cyclone activity, including its impact on variability associated with ENSO. The multi-parameter analysis in chapter 3 can also be applied to other basins. Whilst outside of the scope of this study, an agreed upon tracking algorithm needs to emerge to allow for better comparisons of simulated tropical cyclones in GCMs. In the meantime multiple tracking algorithms should be used for a single study analogous to using multiple reanalysis datasets.

6.3.2 What is the role of the eastern tropical Paciﬁc on tropical cyclone activity associated with different types of El Nino?˜

Research on the different types of El Nino˜ and their associated teleconnections is a contemporary topic. A unified classification of the different types of El Nino˜ in observations and models needs to be agreed upon for continual assessment on how well climate models simulate the different types of El Nino.˜ A counter-part model experiment can be conducted by forcing the atmosphere with local SSTs only. This would be the first task to improve the understanding on what regions have the largest influence on tropical cyclone movement in the Western North Pacific. The experimental design could be used to analyse tropical cyclone activity in other basins. The local and remote SST experiments should also be undertaken with an observed SST dataset similar to Hong et al. (2011) and Jin et al. (2012). New SST products have lead to a number of studies which show that the simulation of

Page 185 Chapter 6: Conclusions tropical cyclones is sensitive to the use of different SST datasets (LaRow, 2013). This is of interest to AGCM simulated tropical cyclone activity which includes tropical cyclone activity in the Western North Paciﬁc associated with different types of El Nino.˜

6.3.3 What is the response of tropical cyclones to climate change?

Other aspects of tropical cyclone metrics and how they may change with climate change should be studied, such as storm associated precipitation and track length (Kim et al., 2013). Kim et al. (2013) found the maximum precipitation associated with tropical cyclones increases by 25 %. A change in tropical cyclone associated precipitation with climate change will have large social and econmics impacts due to associated ﬂooding hazard. An intercomparison model study of how tropical cyclones may change with climate change is currently being undertaken within the Tropical Cyclone Model Intercompari- son Project (TCMIP) with the use of 2×CO2 and +2K global SST increase experiments, similar to the study by Held and Zhao (2011). However, these are AGCM studies only and an emphasis is needed for high resolution coupled models of 50 km or greater to obtain more accurate projections of tropical cyclones and climate change. The largest uncertainty with tropical cyclones and climate change occurs on a regional scale which is mainly related to simulated differential ocean warming patterns. More effort is needed to reduce SST and precipitation biases in coupled models to reduce the uncertainties in future projections. An analysis similar to Stott et al. (2004), which is a conceptual framework to estimate the contribution of human-induced increases in atmospheric concentrations of greenhouse gases and other pollutants, can be applied to high resolution (< 25 km) GCM simulations. The analysis will quantify the attribution of present-day tropical cyclone changes due to anthropogenic emissions. Due to the large costs associated with running high-resolution coupled models Nick Klingaman is currently undertaking work investigating how air-sea coupling affects pro-

Page 186 Chapter 6: Conclusions jections in regional weather and climate extremes via a high resolution atmosphere model coupled with a K Proﬁle Parameterisation (KPP) boundary-layer ocean model (Klinga- man et al., 2011). Work has also been proposed by Pier Luigi Vidale on this that will use a hiearachy of KPP, NEMIX (mixed layer ocean) and fully coupled atmosphere to NEMO ocean model at 1/12o to investigate air-sea coupling of tropical cyclones without introducing the full cost and large SST biases associated with current coupled models. Investigating how tropical cyclone intensity changes with climate change in this dataset will tackle one of the most important research areas in tropical cyclones and climate change. Higher resolution coupled models than those used in this thesis may lead to interesting studies on how tropical cyclones with climate change can feedback onto the large-scale environment (Emanuel, 2008). Tropical cyclone associated ocean mixing occurs within the most intense tropical cyclones. These would have to be resolved with the use of higher resolution models to capture the magnitude of this mechanism and any potential feedbacks such as ocean heat transport.

Page 187 BIBLIOGRAPHY

Adler, R. F., Huffman, G. J., Chang, A., Ferraro, R., Xie, P.-P., Janowiak, J., Rudolf, B., Schneider, U., Curtis, S., Bolvin, D., Gruber, A., Susskind, J., Arkin, P., and Nelkin, E. (2003). The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979-Present). J. Hydromet., 4(6), 1147–1167. Aiyyer, A. R. and Thorncroft, C. (2006). Climatology of Vertical Wind Shear over the Tropical Atlantic. J. Climate, 19(12), 2969–2983. Alexander, M., Blade,´ I., Newman, M., Lanzante, J. R., Lau, N.-C., and Scott, J. D. (2002). The Atmospheric Bridge: The Inﬂuence of ENSO Teleconnections on Air-Sea Interaction over the Global Oceans. J. Climate, 15(16), 2205–2231. Allan, R. P. (2011). The Role of Water Vapour in Earths Energy Flows. Surveys in Geophysics. Allen, M. R. and Ingram, W. J. (2002). Constraints on future changes in climate and the hydrologic cycle. Nature, 419(6903), 224–232. Alvia, L. A. and Pasch, R. J. (1995). Atlantic tropical systems of 1993. Mon. Weath. Rev., 123, 887–896. Ambaum, M. H. P. (2010). Thermal physics of the atmosphere, page 239. Wiley- Blackwell. Ashok, K., Behera, S. K., Rao, S. A., Weng, H., and Yamagata, T. (2007). El Nino Modoki and its possible teleconnection. J. Geophys. Res., 112(C11), C11007. Basher, R. Zheng, X. (1995). Tropical cyclones in the southwest Paciﬁc: Spatial patterns and relationships to the Southern Oscillation and sea surface temperature. J. Climate, 8, 1249–1260.

188 Bibliography

Bell, G. D. and Chelliah, M. (2006). Leading Tropical Modes Associated with Interan- nual and Multidecadal Fluctuations in North Atlantic Hurricane Activity. J., 19(4), 590–612. Bell, R., Strachan, J., Vidale, P. L., Hodges, K. I., and Roberts, M. (2013a). Response of tropical cyclones to idealized climate change experiments in a global high resolution coupled general circulation model. J. Climate, 26(20), 7966–7980. Bell, R., Strachan, J., Vidale, P. L., Hodges, K. I., and Roberts, M. (2013b). Simulation of the Global ENSO-Tropical Cyclone Teleconnection by a High-Resolution Coupled General Circulation Model. J. Climate (in press). Bengtsson, L., Botzet, M., and Esch, M. (1995). Hurricane-type vortices in a general circulation model. Tellus A, 47, 175–196. Bengtsson, L., Hodges, K. I., Esch, M., Keenlyside, N., Kornblueh, L., Luo, J.-J., and Yamagata, T. (2007a). How may tropical cyclones change in a warmer climate? Tellus A, 59(4), 539–561. Bengtsson, L., Hodges, K. I., and Esch, M. (2007b). Tropical cyclones in a T159 resolution global climate model: comparison with observations and re-analysis. Tellus A, 59(4), 396–416. Bister, M., Renno, N., Pauluis, O., and Emanuel, K. (2011). Comment on Makarieva et al. A critique of some modern applications of the Carnot heat engine concept: the dis- sipative heat engine cannot exist. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Science, 467(2125), 1–6. Bjerknes, J. (1966). A possible response of the atmospheric Hadley circulation to equatorial anomalies of ocean temperature. Tellus, 18(4), 820–829. Bloomﬁeld, P. (2000). Fourier Analysis of Time Series: An Introduction. New York: Wiley-Interscience. Bond, N. A., Cronin, M. F., Sabine, C., Kawai, Y., Ichikawa, H., Freitag, P., and Ronnholm, K. (2011). Upper ocean response to Typhoon Choi-Wan as measured by the Kuroshio Extension Observatory mooring. J. Geophys. Res., 116(C2), C02031. Booth, B. B. B., Dunstone, N. J., Halloran, P. R., Andrews, T., and Bellouin, N. (2012).

Page 189 Bibliography

Aerosols implicated as a prime driver of twentieth-century North Atlantic climate variability. Nature, 484, 228–232. Bourassa, M. A. (2009). Remotely sensed winds and wind stresses for marine forecasting and ocean modelling. Technical report, Ocean Obs ’09. Briegel, L. M. and Frank, W. M. (1997). Large-Scale Influences on Tropical Cyclogenesis in the Western North Pacific. Monthly Weather Review, 125(7), 1397–1413. Broccoli, A. J. and Manabe, S. (1990). Can existing climate models be used to study anthropogenic changes in tropical cyclone climate? Geophys. Res. Lett., 17(11), 1917– 1920. Bruyere,` C. L., Holland, G. J., and Towler, E. (2012). Investigating the Use of a Genesis Potential Index for Tropical Cyclones in the North Atlantic Basin. J. Climate, 25(24), 8611–8626. Burgers, G. and Stephenson, D. B. (1999). The ’normality’ of El Nino. Geophys. Res. Lett., 26(8), 1027–1030. Camargo, S. J. (2013). Global and regional aspects of tropical cyclone activity in the CMIP5 models. J. Climate. Camargo, S. J. and Zebiak, S. E. (2002). Improving the Detection and Tracking of Trop- ical Cyclones in Atmospheric General Circulation Models. Weather and Forecasting, 17(6), 1152–1162. Camargo, S. J., Robertson, A. W., Gaffney, S. J., Smyth, P., and Ghil, M. (2007a). Cluster Analysis of Typhoon Tracks. Part II: Large-Scale Circulation and ENSO. J. Climate, 20(14), 3654–3676. Camargo, S. J., Barnston, A. G., Klotzbach, P. J., and Landsea, C. W. (2007b). Seasonal tropical cyclone forecasts. WMO Bulletin, 56, 297–309. Camargo, S. J., Emanuel, K. A., and Sobel, A. H. (2007c). Use of a Genesis Potential Index to Diagnose ENSO Effects on Tropical Cyclone Genesis. J. Climate, 20(19), 4819–4834. Camargo, S. J., Sobel, A. H., Barnston, A. G., and Klotzbach, P. J. (2010). The Influence of Natural Climate Variability on Tropical Cyclones, and Seasonal Forecasts of Trop-

Page 190 Bibliography

ical Cyclone Activity. In Global Perspectives on Tropical Cyclones: From science to mitigation, pages 325–360. World Scientific Pubslishing Co. Catto, J. L. (2009). Extratropical cyclones in HiGEM: Climatology, Structure and Future Predictions. Ph.D. thesis, University of Reading. Catto, J. L., Shaffrey, L. C., and Hodges, K. I. (2010). Can Climate Models Capture the Structure of Extratropical Cyclones? Journal of Climate, 23(7), 1621–1635. Catto, J. L., Shaffrey, L. C., and Hodges, K. I. (2011). Northern Hemisphere Extratrop- ical Cyclones in a Warming Climate in the HiGEM High Resolution Climate Model. Journal of Climate, 24, 5336–5352. Chan, J. C. (2007). Interannual variations of intense typhoon activity. Tellus A, 59, 455–460. Chan, J. C. and Liu, K. S. (2012). An Inactive Period of Western North Pacific Trop- ical Cyclone Activity: 1998-2011. In 30th Conference on Hurricanes and Tropical Meteorology. Chan, J. C. L. (1985). Tropical Cyclone Activity in the Northwest Pacific in Relation to the El Nino/Southern˜ Oscillation Phenomenon. Mon. Wea. Rev., 113(4), 599–606. Chan, J. C. L. and Liu, K. S. (2004). Global Warming and Western North Pacific Typhoon Activity from an Observational Perspective. J. Climate, 17(23), 4590–4602. Chand, S. S., McBride, J., Tory, K. J., Wheeler, M. C., and Walsh, K. J. E. (2013). Impact of different ENSO regimes on southwest Pacific tropical cyclones. J. Climate, 26(2), 600–608. Chang, C.-P., Liu, C.-H., and Kuo, H.-C. (2003). Typhoon Vamei: An equatorial tropical cyclone formation. Geophys. Res. Lett., 30(3), 1150. Chauvin, F., Royer, J.-F., and Dequ´ e,´ M. (2006). Response of hurricane-type vortices to global warming as simulated by ARPEGE-Climat at high resolution. Climate Dynam- ics, 27(4), 377–399. Chavas, D. R. and Emanuel, K. A. (2010). A QuikSCAT climatology of tropical cyclone size. Geophys. Res. Lett., 37(18), L18816. Chen, G. and Tam, C.-Y. (2010). Different impacts of two kinds of Pacific Ocean warm-

Page 191 Bibliography

ing on tropical cyclone frequency over the western North Pacific. Geophys. Res. Lett., 37(1), L01803. Chiang, J. C. H. and Sobel, A. H. (2002). Tropical tropospheric temperature variations caused by ENSO and their influence on the remote tropical climate. J. Climate, 15, 2616–2631. Christensen, J., Carter, T., Rummukainen, M., and Amanatidis, G. (2007). Evaluating the performance and utility of regional climate models: the PRUDENCE project. Climatic Change, 81(0), 1–6. Chunzai, W. and Picaut, J. (2004). Understanding ENSO physics: A review. Geophysical monograph, 147, 21–48. Cione, J. (2012). Hurricane intensity and inner-core SST: Warmer is not always better. In 30th Conference on Hurricanes and Tropical Meteorology. CLIVAR (2013). ENSO Variations. U.S. CLIVAR, 11(2), p28. Colbert, A. J. and Soden, B. J. (2011). Climatological Variations in North Atlantic Trop- ical Cyclone Tracks. Journal of Climate, 25(2), 657–673. Collins, M., An, S.-I., Cai, W., Ganachaud, A., Guilyardi, E., Jin, F.-F., Jochum, M., Lengaigne, M., Power, S., Timmermann, A., Vecchi, G., and Wittenberg, A. (2010). The impact of global warming on the tropical Pacific Ocean and El Nino.˜ Nature Geoscience, 3(6), 391–397. Dailey, P., Huddleston, M., Brown, S., and Fasking, D. (2009). The financial risks of climate change. Technical report, Associated of British Insurers. Dawson, A., Matthews, A., Stevens, D., Roberts, M., and Vidale, P.-L. (2013). Impor- tance of oceanic resolution and mean state on the extra-tropical response to El Nino˜ in a matrix of coupled models. Climate Dyn., 41(5-6), 1439–1452. Dee, D. P., Uppala, S. M., Simmons, a. J., Berrisford, P., Poli, P., Kobayashi, S., Andrae, U., Balmaseda, M. a., Balsamo, G., Bauer, P., Bechtold, P., Beljaars, a. C. M., van de Berg, L., Bidlot, J., Bormann, N., Delsol, C., Dragani, R., Fuentes, M., Geer, a. J., Haimberger, L., Healy, S. B., Hersbach, H., Holm,´ E. V., Isaksen, L., Ka˚ llberg, P., Kohler,¨ M., Matricardi, M., McNally, a. P., Monge-Sanz, B. M., Morcrette, J.-J., Park,

Page 192 Bibliography

B.-K., Peubey, C., de Rosnay, P., Tavolato, C., Thepaut,´ J.-N., and Vitart, F. (2011). The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137(656), 553–597. Delworth, T. L. and Knutson, T. R. (2000). Simulation of Early 20th Century Global Warming. Science, 287(5461), 2246–2250. Delworth, T. L., Rosati, A., Anderson, W., Adcroft, A. J., Balaji, V., Benson, R., Dixon, K., Griffies, S. M., Lee, H.-C., Pacanowski, R. C., Vecchi, G. A., Wittenberg, A. T., Zeng, F., and Zhang, R. (2012). Simulated Climate and Climate Change in the GFDL CM2.5 High-Resolution Coupled Climate Model. J. Climate, 25(8), 2755–2781. Demory, M.-E. (2012). Sensitivity of the Global Hydrological Cycle to Horizontal Res- olution in an Atmosperhic General Circulation Model. Ph.D. thesis, University of Reading. Demory, M.-E., Vidale, P., Roberts, M., Berrisford, P., Strachan, J., Schiemann, R., and Mizielinski, M. (2013). The role of horizontal resolution in simulating drivers of the global hydrological cycle. Climate Dynamics, pages 1–25. Diamond, H. J., Lorrey, A. M., and Renwick, J. A. (2012). A Southwest Pacific Tropical Cyclone Climatology and Linkages to the El Nino-Southern˜ Oscillation. J. Climate, 26(1), 3–25. DiNezio, P. N., Clement, A. C., Vecchi, G. A., Soden, B. J., Kirtman, B. P., and Lee, S.-K. (2009). Climate Response of the Equatorial Pacific to Global Warming. Journal of Climate, 22(18), 4873–4892. Doron, E., Hollingsworth, A., Hoskins, B. J., and Simmons, A. J. (1974). A comparison of grid-point and spectral methods in a meteorological problem. Quarterly Journal of the Royal Meteorological Society, 100(425), 371–383. Dunion, J. P. and Veldon, C. S. (2004). The impact of the Saharan Air layer of Atlantic tropical cyclone activity. Bull. Amer. Meteorol. Soc., 85, 353–365. Dunstone, N. J., Smith, D. M., Booth, B. B. B., Hermanson, L., and Eade, R. (2013). Anthropogenic aerosol forcing of Atlantic tropical storms. Nature Geosci, 6(7), 534– 539.

Page 193 Bibliography

Durran, D. (1999). Numerical methods for wave equations in geophysical ﬂuid dynamics. Springer. Elsner, J. and Liu, K. (2003). Examining the ENSO-typhoon hypothesis. Climate Re- search, 25, 43–54. Emanuel, K. (1994). Atmospheric convection. Oxford Univeristy Press. Emanuel, K. (2005a). Divine wind: the history and science of hurricanes. Oxford Uni- veristy Press. Emanuel, K. (2005b). Increasing destructiveness of tropical cyclones over the past 30 years. Nature, 436(7051), 686–8. Emanuel, K. (2006). Hurricanes: Tempests in a greenhouse. Physics Today, pages 2–3. Emanuel, K. (2008). The Hurricane-Climate Connection. Bulletin of the American Me- teorological Society, 89(5), ES10–ES20. Emanuel, K., Sundararajan, R., and Williams, J. (2008). Hurricanes and Global Warm- ing: Results from Downscaling IPCC AR4 Simulations. Bulletin of the American Meteorological Society, 89(3), 347–367. Emanuel, K., Oouchi, K., Satoh, M., Tomita, H., and Yamada, Y. (2010). Comparison of Explicitly Simulated and Downscaled Tropical Cyclone Activity in a High-Resolution Global Climate Model. J. Adv. Model. Earth Syst, 2(9), 9. Emanuel, K. A. (1987). The dependence of hurricane intensity on climate. Nature, 326(6112), 483–485. Emanuel, K. A. (1988). The Maximum Intensity of Hurricanes. J. Atmos. Sci., 45(7), 1143–1155. Emanuel, K. A. and Nolan, D. S. (2004). Tropical cyclone activity and global climate. In 26th Conference on Hurricanes and Tropical Meteorology, pages 240–241, Miami, FL. Amer. Meteor. Soc. Essery, R. L. H., Best, M. J., Betts, R. A., Cox, P. M., and Taylor, C. M. (2003). Explicit Representation of Subgrid Heterogeneity in a GCM Land Surface Scheme. Journal of Hydrometeorology, 4(3), 530–543. Evan, A. T. (2012). Atlantic hurricane activity following two major volcanic eruptions.

Page 194 Bibliography

Journal of Geophysical Research: Atmospheres, 117(D6), D06101. Evan, A. T., Kossin, J. P., Chung, C., and Ramanathan, V. (2011). Arabian Sea tropical cyclones intensified by emissions of black carbon and other aerosols. Nature, 479(7371), 94–97. Evans, J. L. and Allan, R. J. (1992). El Nino/Southern Oscillation modification to the structure of the monsoon and tropical cyclone activity in the Australisan region. Inter- national Journal of Climatology, 12, 611–623. Evans, J. L. and Hart, R. E. (2003). Objective Indicators of the Life Cycle Evolution of Extratropical Transition for Atlantic Tropical Cyclones. Monthly Weather Review, 131(5), 909–925. Felton, C. S., Subrahmanyam, B., and Murty, V. S. N. (2013). ENSO modulated cyclogenesis over the Bay of Bengal. J. Climate, 26(24), 9806–9818. Frank, W. M. and Young, G. S. (2007). The interannual variability of tropical cyclones. Mon. Wea. Rev., 135, 3587–3598. Garner, S. T., Held, I. M., Knutson, T., and Sirutis, J. (2009). The Roles of Wind Shear and Thermal Stratification in Past and Projected Changes of Atlantic Tropical Cyclone Activity. Journal of Climate, 22(17), 4723–4734. Gill, A. E. (1980). Some simple solutions for heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc., 106(449), 447–462. Glantz, M. H., Katz, R. W., and Nicholls, N. (1991). Teleconnections Linking Worldwide Climate Anomalies: Scientific Basis and Societal Impact. Cambridge University Press, Cambridge. Gnanadesikan, A., Emanuel, K., Vecchi, G. A., Anderson, W. G., and Hallberg, R. (2010). How ocean color can steer Pacific tropical cyclones. Geophysical Research Letters, 37(18), 2–6. Goldenberg, S. B. and Shapiro, L. J. (1996). Physical Mechanisms for the Association of El Nino˜ and West African Rainfall with Atlantic Major Hurricane Activity. J. Climate, 9(6), 1169–1187. Goldenberg, S. B., Landsea, C. W., Mestas-Nunez, a. M., and Gray, W. M. (2001).

Page 195 Bibliography

The recent increase in Atlantic hurricane activity: causes and implications. Science, 293(5529), 474–9. Goni, G. J., Black, P., and Trinanes, J. (2003). Using satellite altimetry to identify regions of hurricane intesification. Technical report, AVISO. Gray, W. M. (1968). Global view of the origin of tropical disturbances and storms. Mon. Weath. Rev., 96(10), 669–700. Gray, W. M. (1984). Atlantic seasonal hurricane frequency. Part I: El Nino˜ and 30 mb quasi-biennial oscillation influences. Mon. Weath. Rev., 112, 1649–1668. Gray, W. M. and Sheaffer, J. D. (1991). El Nino˜ and QBO influences on tropical cyclone activity. In Teleconnections linking worldwide anomalies, pages 257–284. Cambridge University Press, New York. Gualdi, S., Scoccimarro, E., and Navarra, A. (2008). Changes in Tropical Cyclone Ac- tivity due to Global Warming: Results from a High-Resolution Coupled General Cir- culation Model. Journal of Climate, 21(20), 5204–5228. Guilyardi, E., Wittenberg, A., Fedorov, A., Collins, M., Wang, C., Capotondi, A., van Oldenborgh, G. J., and Stockdale, T. (2009). Understanding El Nino˜ in Ocean- Atmosphere General Circulation Models: Progress and Challenges. Bull. Amer. Me- teor. Soc., 90(3), 325–340. Ha, K.-J., Yoon, S.-J., Yun, K.-S., Kug, J.-S., Jang, Y.-S., and Chan, J. (2012). Depen- dency of typhoon intensity and genesis locations on El Nino˜ phase and SST shift over the western North Pacific. Theoretical and Applied Climatology, 109(3-4), 383–395. Haarsma, R. J., Mitchell, J. F. B., and Senior, C. A. (1993). Tropical disturbances in a GCM. Climate Dynamics, 8(5), 247–257. Haarsma, R. J., Hazeleger, W., Severijns, C., de Vries, H., Sterl, A., Bintanja, R., van Oldenborgh, G. J., and van den Brink, H. W. (2013). More hurricanes to hit western Europe due to global warming. Geophysical Research Letters, 40(9), 1783–1788. Ham, Y.-G. and Kug, J.-S. (2011). How well do current climate models simulate two types of El Nino? Climate Dynamics, pages 1–16. Hart, R. E. (2003). A Cyclone Phase Space Derived from Thermal Wind and Thermal

Page 196 Bibliography

Asymmetry. Monthly Weather Review, 131(4), 585–616. Hart, R. E. and Evans, J. L. (2001). A Climatology of the Extratropical Transition of Atlantic Tropical Cyclones. Journal of Climate, 14(4), 546–564. Held, I. M. (2005). The Gap between Simulation and Understanding in Climate Model- ing. Bulletin of the American Meteorological Society, 86(11), 1609–1614. Held, I. M. and Hou, A. Y. (1980). Nonlinear Axially Symmetric Circulations in a Nearly Inviscid Atmosphere. J. Atmos. Sci., 37, 515–533. Held, I. M. and Soden, B. J. (2006). Robust Responses of the Hydrological Cycle to Global Warming. Journal of Climate, 19(21), 5686–5699. Held, I. M. and Zhao, M. (2011). The response of tropical cyclone statistics to an increase in CO2 with fixed sea surface temperatures. Journal of Climate, 24, 5353–5364. Heymsfield, G. M., Simpson, J., Halverson, J., Tian, L., Ritchie, E., and Molinari, J. (2006). Structure of Highly Sheared Tropical Storm Chantal during CAMEX-4. Jour- nal of the Atmospheric Sciences, 63(1), 268–287. Hill, K. J., Taschetto, A. S., and England, M. H. (2011). Sensitivity of South American summer rainfall to tropical Pacific Ocean SST anomalies. Geophys. Res. Lett., 38(1), L01701. Ho, C.-H., Kim, J.-H., Jeong, J., and Kim, H.-S. (2006). Variations of tropical cyclone activity in the South Indian Ocean: El Nino-Southern Oscillation effects. Journal of Geophysical Research, D22191. Hodges, K. I. (1995). Feature tracking on a unit sphere. Mon. Weath. Rev., 123, 3458– 3465. Hodges, K. I. (1996). Spherical Nonparametric Estimators Applied to the UGAMP Model Integration for AMIP. Monthly Weather Review, 124(12), 2914–2932. Hodges, K. I. (1999). Adaptive Constraints for Feature Tracking. Monthly Weather Review, 127(6), 1362–1373. Hodges, K. I. (2008). Confidence Intervals and Significance Tests for Spherical Data Derived from Feature Tracking. Monthly Weather Review, 136(5), 1758–1777. Hodges, R. E. and Elsner, J. B. (2011). Evidence linking solar variability with US hurri-

Page 197 Bibliography

canes. International Journal of Climatology, 31(13), 1897–1907. Holland, G. (1993). Ready Reckoner. Technical report, World Meteorological Organiza- tion; Geneva, Switzerland. Holland, G. (2012). Global Tropical Cyclones and Climate: Current Signal and Future Change. In 30th Conference on Hurricanes and Tropical Meteorology. Holland, G. J. (1997). the maximum potential intensity of tropical cyclones. J. Atmos. Sci., 54, 2514–2519. Hong, C.-C., Li, Y.-H., Li, T., and Lee, M.-Y. (2011). Impacts of central Pacific and eastern Pacific El Ninos on tropical cyclone tracks over the western North Pacific. Geophys. Res. Lett., 38(16), L16712. Hongo, C., Kawamata, H., and Goto, K. (2012). Catastrophic impact of typhoon waves on coral communities in the Ryukyu Islands under global warming. J. Geophys. Res., 117(G2), G02029. Hopsch, S. B., Thorncroft, C. D., and Tyle, K. R. (2009). Analysis of African East- erly Wave Structures and Their Role in Influencing Tropical Cyclogenesis. Monthly Weather Review, 138(4), 1399–1419. Iizuka, S. and Matsuura, T. (2008). ENSO and Western North Pacific tropical cyclone activity simulated in a CGCM. Climate Dyn., 30(7-8), 815–830. Iizuka, S. and Matsuura, T. (2009). Relationship between ENSO and North Atlantic Tropical Cyclone Frequency Simulated in a Coupled General Circulation Model. In J. B. Elsner and T. H. Jagger, editors, Hurricanes and Climate Change, pages 323–338. Springer US. IPCC (2007). Climate Change 2007: The physical science basis. Contribution of Working Group 1 to the Fourth Assessment Report of the intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge. IPCC (2013). IPCC Fifth Assessment Report, Summary for Policymakers. Jin, C.-S., Ho, C.-H., Kim, J.-H., Lee, D.-K., Cha, D.-H., and Yeh, S.-W. (2012). Critical Role of Northern Off-Equatorial Sea Surface Temperature Forcing Associated with Central Pacific El Nino˜ in More Frequent Tropical Cyclone Movements toward East

Page 198 Bibliography

Asia. Journal of Climate, 26(8), 2534–2545. Johanson, C. M. and Fu, Q. (2009). Hadley Cell Widening: Model Simulations versus Observations. Journal of Climate, 22(10), 2713–2725. Johns, T. C., Durman, C. F., Banks, H. T., Roberts, M. J., McLaren, A. J., Ridley, J. K., Senior, C. A., Williams, K. D., Jones, A., Rickard, G. J., Cusack, S., Ingram, W. J., Crucifix, M., Sexton, D. M. H., Joshi, M. M., Dong, B.-W., Spencer, H., Hill, R. S. R., Gregory, J. M., Keen, A. B., Pardaens, A. K., Lowe, J. A., Bodas-Salcedo, A., Stark, S., and Searl, Y. (2006). The New Hadley Centre Climate Model (HadGEM1): Evaluation of Coupled Simulations. J. Climate, 19(7), 1327–1353. Johnson, N. C. and Xie, S.-P. (2010). Changes in the sea surface temperature threshold for tropical convection. Nature Geoscience, 3(12), 842–845. Kalnay, E. (2003). Atmospheric Modelling, Data Assimilation and Predictability. Cam- bridge University Press, Cambridge. Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Leetmaa, A., Reynolds, R., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K. C., Ropelewski, C., Wang, J., Jenne, R., and Joseph, D. (1996). The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77(3), 437–471. Kanada, S., Wada, A., and Sugi, M. (2013). Future Changes in Structures of Extremely Intense TCs by a 2-km Mesh Nonhydrostatic Model. J. Climate. Kang, S. and Lu, J. (2012). Expansion of the Hadley cell under global warming: winter versus summer. J. Climate, 25, 8387–8393. Kao, H.-Y. and Yu, J.-Y. (2009). Contrasting Eastern-Pacific and Central-Pacific Types of ENSO. Journal of Climate, 22(3), 615–632. Kiladis, G. N. and Diaz, H. F. (1989). Global Climatic Anomalies Associated with Ex- tremes in the Southern Oscillation. Journal of Climate, 2(9), 1069–1090. Kim, H.-M., Webster, P. J., and Curry, J. A. (2009). Impact of Shifting Patterns of Pacific Ocean Warming on North Atlantic Tropical Cyclones. Science, 325(5936), 77–80. Kim, H.-M., Webster, P. J., and Curry, J. A. (2011). Modulation of North Pacific Tropical

Page 199 Bibliography

Cyclone Activity by Three Phases of ENSO. Journal of Climate, 24(6), 1839–1849. Kim, H.-S., Vecchi, G. A., Knutson, T. R., Anderson, W. G., Delworth, T. L., Rosati, A. J., Zeng, F., and Zhao, M. (2013). Tropical Cyclone Simulation and Response to CO2 Doubling in the GFDL CM2.5 High-Resolution Coupled Climate Model. J. Climate (submitted). Kim, J.-H., Ho, C.-H., and Sui, C.-H. (2005). Circulation features associated with the record-breaking typhoon landfall on Japan in 2004. Geophysical Research Letters, 32(14), L14713. Kim, J.-H., Brown, S. J., and McDonald, R. E. (2010). Future changes in tropical cyclone genesis in fully dynamic ocean- and mixed layer ocean-coupled climate models: a low- resolution model study. Climate Dynamics, 37, 737–758. Klingaman, N. P., Woolnough, S. J., Weller, H., and Slingo, J. M. (2011). The Impact of Finer-Resolution Air-Sea Coupling on the Intraseasonal Oscillation of the Indian Monsoon. Journal of Climate, 24(10), 2451–2468. Klotzbach, P. J. (2010). El Nino-Southern˜ Oscillations Impact on Atlantic Basin Hurri- canes and U.S. Landfalls. J. Climate, 24(4), 1252–1263. Klotzbach, P. J. (2011). The Inﬂuence of El Nino-Southern˜ Oscillation and the Atlantic Multidecadal Oscillation on Caribbean Tropical Cyclone Activity. J. Climate, 24(3), 721–731. Klotzbach, P. J. and Gray, W. M. (2008). Multidecadal Variability in North Atlantic Tropical Cyclone Activity. Journal of Climate, 21(15), 3929–3935. Klotzbach, P. J. and Gray, W. M. (2010). Summary of 2010 Atlantic Tropical Cyclone activity and veriﬁcation of Author’s seasonal and two-week forecast. Technical report, Colardo State University. Knaff, J. a. (1997). Implications of Summertime Sea Level Pressure Anomalies in the Tropical Atlantic Region. Journal of Climate, 10(4), 789–804. Knapp, K. R., Kruk, M. C., Levinson, D. H., Diamond, H. J., and Neumann, C. J. (2010). The International Best Track Archive for Climate Stewardship (IBTrACS). Bull. Amer. Meteor. Soc, 91(3), 363–376.

Page 200 Bibliography

Knutson, T. R. and Manabe, S. (1995). Time-Mean Response over the Tropical Pacific to Increased C02 in a Coupled Ocean-Atmosphere Model. Journal of Climate, 8(9), 2181–2199. Knutson, T. R., Delworth, T. L., Dixon, K. W., Held, I. M., Lu, J., Ramaswamy, V., Schwarzkopf, D., Stenchikov, G., and Stouffer, R. J. (2006). Assessment of twentieth- century regional surface temperature trends using the GFDL CM2 coupled models. J. Climate, 19(9), 1624–1651. Knutson, T. R., McBridge, J. L., Chan, J., Emanuel, K., Holland, G., Landsea, C., Held, I., Kossin, J. P., Srivastava, A. K., and Sugi, M. (2010a). Tropical cyclones and climate change. Nat. Geosci., 3, 157–163. Knutson, T. R., Landsea, C., and Emanuel, K. A. (2010b). Tropical Cyclones and Climate Change: A Review. In Global Perspectives on Tropical Cyclones: From science to mitigation, pages p243–284. World Scientific Pubslishing Co. Kossin, J. P., Camargo, S. J., and Sitkowski, M. (2010). Climate Modulation of North Atlantic Hurricane Tracks. J. Climate, 23(11), 3057–3076. Kug, J.-S., Jin, F.-F., and An, S.-I. (2009). Two Types of El Nino˜ Events: Cold Tongue El Nino˜ and Warm Pool El Nino.˜ Journal of Climate, 22(6), 1499–1515. Kuleshov, Y. (2003). Tropical cyclones in the southern hemisphere: Influence of the El Nino-Southern Oscillation phenomenon. In Proc. 7th Int. Conf. Southern Hemisphere Meteorology and Oceanography, Wellington, New Zealand. Kuleshov, Y., Qi, L., Fawcett, R., and Jones, D. (2008). On tropical cyclone activity in the Southern Hemisphere: Trends and the ENSO connection. Geophys. Res. Lett., 35(L14S08). Landsea, C. W. (1993). A climatology of intense (or major) Atlantic hurricanes. Mon. Weath. Rev., 121, 1703–1713. Landsea, C. W. (2007). Counting Atlantic tropical cyclones back to 1900. Eos, Trans. Am. Geophys. Union, 88, 197–208. Landsea, C. W. and Franklin, J. L. (2013). Atlantic Hurricane Database Uncertainty and Presentation of a New Database Format. Mon. Wea. Rev., 141(10), 3576–3592.

Page 201 Bibliography

Landsea, C. W., Pielke Jr, R. A., Mestas-Nunez, A., and Knaff, J. A. (1999). Atlantic Basin Hurricanes: Indices of Climatic Change. Climatic Change, 42, 89–129. Landsea, C. W., Harper, B. A., Hoarau, K., and Knaff, J. A. (2006). Can We Detect Trends in Extreme Tropical Cyclones? Science, 313(5786), 452–454. Langenbrunner, B. and Neelin, J. D. (2013). Analyzing ENSO Teleconnections in CMIP Models as a Measure of Model Fidelity in Simulating Precipitation. J. Climate, 26(13), 4431–4446. LaRow, T. E. (2013). The Impact of SST Bias Correction on North Atlantic Hurricane Retrospective Forecasts. Monthly Weather Review, 141(2), 490–498. Latif, M., Kleeman, R., and Eckert, C. (1997). Greenhouse Warming, Decadal Variabil- ity, or El Nino?˜ An Attempt to Understand the Anomalous 1990s. Journal of Climate, 10(9), 2221–2239. Lea, A. (2012). Improving counts of global intense tropical cyclone events back to 1970. In 30th Conference on Hurricanes and Tropical Meteorology. Lee, S.-K., Wang, C., and Enfield, D. B. (2010). On the impact of central Pacific warming events on Atlantic tropical storm activity. Geophys. Res. Lett., 37(17), L17702. Lee, S.-K., Enfield, D. B., and Wang, C. (2011). Future Impact of Differential Interbasin Ocean Warming on Atlantic Hurricanes. Journal of Climate, 24(4), 1264–1275. Lee, T. and McPhaden, M. J. (2010). Increasing intensity of El Nino in the central- equatorial Pacific. Geophys. Res. Lett., 37(14), L14603. Li, J., Xie, S.-P., Cook, E. R., Morales, M. S., Christie, D. A., Johnson, N. C., Chen, F., D’Arrigo, R., Fowler, A. M., Gou, X., and Fang, K. (2013). El Nino modulations over the past seven centuries. Nature Clim. Change, 3(9), 822–826. Li, R. C. Y. and Zhou, W. (2012). Changes in Western Pacific Tropical Cyclones As- sociated with the El Nino-Southern˜ Oscillation Cycle. Journal of Climate, 25(17), 5864–5878. Li, T., Kwon, M., Zhao, M., Kug, J.-S., Luo, J.-J., and Yu, W. (2010). Global warming shifts Pacific tropical cyclone location. Geophysical Research Letters, 37(21), 1–5. Lin, I.-I., Black, P., Price, J. F., Yang, C.-Y., Chen, S. S., Lien, C.-C., Harr, P., Chi, N.-H.,

Page 202 Bibliography

Wu, C.-C., and D’Asaro, E. A. (2013). An ocean coupling potential intensity index for tropical cyclones. Geophysical Research Letters, 40(9), 1878–1882. Liu, Z. and Alexander, M. (2007). Atmospheric Bridge, Oceanic Tunnel, and Global Climatic Teleconnections. Rev. Geophys., 45, 1–34. Longshore, D. (2008). Encyclopedia of hurricanes, typhoons and cyclones, new edition, chapter A to Z Ent, page 468. Facts On File. Lu, J. and Zhao, B. (2012). The Role of Oceanic Feedback in the Climate Response to Doubling CO2. Journal of Climate, 25(21), 7544–7563. Lu, J., Vecchi, G. A., and Reichler, T. (2007). Expansion of the Hadley cell under global warming. Geophys. Res. Lett., 34(6), L06805. Manganello, J. V., Hodges, K. I., Kinter, J. L., Cash, B. A., Marx, L., Jung, T., Achuthavarier, D., Adams, J. M., Altshuler, E. L., Huang, B., Jin, E. K., Stan, C., Towers, P., and Wedi, N. (2012). Tropical Cyclone Climatology in a 10-km Global Atmospheric GCM: Toward Weather-Resolving Climate Modeling. J. Climate, 25, 3867–3893. Mann, M. E. and Emanuel, K. A. (2006). Atlantic hurricane trends linked to climate change. Eos Trans. AGU, 87(24). Maue, R. N. (2009). Northern Hemisphere tropical cyclone activity. Geophys. Res. Lett., 36(5), L05805. Maue, R. N. (2011). Recent historically low global tropical cyclone activity. Geophysical Research Letters, 38(14), 1–6. McDonald, R. E., Bleaken, D. G., Creswell, D. R., Pope, V. D., and Senior, C. A. (2005). Tropical storms: representation and diagnosis in climate models and the impacts of climate change. J. Climate, 18, 1262–1275. McGauley, M. G. and Nolan, D. S. (2011). Measuring Environmental Favorability for Tropical Cyclogenesis by Statistical Analysis of Threshold Parameters. J. Climate, 24(23), 5968–5997. McPhaden, M. J. (1999). Genesis and Evolution of the 1997-98 El Nino.˜ Science, 283(5404), 950–954.

Page 203 Bibliography

McPhaden, M. J. (2012). A 21st century shift in the relationship between ENSO SST and warm water volume anomalies. Geophys. Res. Lett., 39(9), L09706. McPhaden, M. J., Lee, T., and McClurg, D. (2011). El Nino˜ and its relationship to changing background conditions in the tropical Pacific Ocean. Geophysical Research Letters, 38(15), L15709. Meehl, G. A., Washington, W. M., Amman, C. M., Arblaster, J. M., Wigley, T. M. L., and Tebaldi, C. (2004). Combinations of natural and anthropogenic forcings in twentieth- century climate. J. Climate, 17, 3721–3727. Mei, W., Pasquero, C., and Primeau, F. (2012). The effect of translation speed upon the intensity of tropical cyclones over the tropical ocean. Geophys. Res. Lett., 39(7), L07801. Mizielinski, M., Roberts, M. J., Vidale, P. L., Schiemann, R., Demory, M.-E., Strachan, J., Edwards, T., Stephens, A., Lawrence, B., Pritchard, M., Chiu, P., Iwi, A., Churchill, J., Del Cano Novales, C., Kettleborough, J., Roseblade, W., Selwood, P., Foster, M., Glover, M., and Malcolm, A. (2013). High resolution global climate modelling; the UPSCALE project, a large simulation campaign. Geoscientific Model Development. Mori, M., Kimoto, M., Ishii, M., Yokoi, S., Mochizuki, T., Chikamoto, Y., Watanabe, M., Nozawa, T., Tatebe, H., T. Sakamoto, T., Komuro, Y., Imada, Y., and Koyama, H. (2013). Hindcast Prediction and Near-Future Projection of Tropical Cyclone Activity over the Western North Pacific Using CMIP5 Near-Term Experiments with MIROC. J. Meteor. Soc. Japan, 91(4), 431–452. Murakami, H. and Sugi, M. (2010). Effect of Model Resolution on Tropical Cyclone Climate Projections. SOLA, 6, 73–76. Murakami, H. and Wang, B. (2010). Future Change of North Atlantic Tropical Cyclone Tracks: Projection by a 20-km-Mesh Global Atmospheric Model. J. Climate, 23(10), 2699–2721. Murakami, H., Mizuta, R., and Shindo, E. (2012a). Future changes in tropical cyclone activity projected by multi-physics and multi-SST ensemble experiments using the 60- km-mesh MRI-AGCM. Climate Dynamics, 39, 2569–2584.

Page 204 Bibliography

Murakami, H., Wang, Y., Yoshimura, H., Mizuta, R., Sugi, M., Shindo, E., Adachi, Y., Yukimoto, S., Hosaka, M., Kusunoki, S., Ose, T., and Kitoh, A. (2012b). Future Changes in Tropical Cyclone Activity Projected by the New High-Resolution MRI- AGCM. J. Climate, 25(9), 3237–3260. Nemoto, K., Midorikawa, T., Wada, A., Ogawa, K., Takatani, S., Kimoto, H., Ishii, M., and Inoue, H. Y. (2009). Continuous observations of atmospheric and oceanic CO2 using a moored buoy in the East China Sea: Variations during the passage of typhoons. Deep Sea Research Part II: Topical Studies in Oceanography, 56(810), 542–553. Nicholls, N. (1979). A possible method for predicting seasonal tropical cyclone activity in the Australian region. Mon. Weath. Rev., 107, 1221–1224. Nicholls, N. (1992). Recent performance of a method for forecasting seasonal tropical cyclone activity. Aus. Met. Mag., 40, 105–110. Nicholls, N. (2008). Recent trends in the seasonal and temporal behaviour of the El Nino-Southern˜ Oscillation. Geophysical Research Letters, 35(19), L19703. Nolan, D. S. and Rappin, E. D. (2008). Increased sensitivity of tropical cyclogenesis to wind shear in higher SST environments. Geophys. Res. Lett., 35(14), L14805. Onogi, K., Tsutsui, J., Koide, H., Sakamoto, M., Kobayashi, S., Hatsushika, H., Mat- sumoto, T., Yamazaki, N., Kamahori, H., Takahashi, K., Kadokura, S., Wada, K., Kato, K., Oyama, R., Ose, T., Mannoji, N., and Taira, R. (2007). The JRA-25 Reanalysis. Journal of the Meteorological Society of Japan. Ser. II, 85(3), 369–432. Oouchi, K., Yoshimura, J., Yoshimura, H., Mizuta, R., Kusunoki, S., and Noda, A. (2006). Tropical Cyclone Climatology in a Global-Warming Climate as Simulated in a 20 km-Mesh Global Atmospheric Model: Frequency and Wind Intensity Analyses. Journal of the Meteorological Society of Japan, 84(2), 259–276. Palmen, T. (1948). On the formation and structure of tropical cyclones. Technical report, University of Helsinki. Paterson, L. A., Hanstrum, B. N., Davidson, N. E., and Weber, H. C. (2005). Inﬂuence of Environmental Vertical Wind Shear on the Intensity of Hurricane-Strength Tropical Cyclones in the Australian Region. Monthly Weather Review, 133(12), 3644–3660.

Page 205 Bibliography

Pielke, R. A. J. and Landsea, C. N. (1999). La Nina, El Nino, and Atlantic Hurricane Damages in the United States. Bull. Amer. Meteorol. Soc., 80(10), 2027–2033. Pielke, R. A. J. and Pielke, R. A. S. (1997). Hurricanes their nature and impacts on society, page 279. Wiley. Pielke, R. A. J., Landsea, C., Mayﬁeld, M., Laver, J., and Pasch, R. (2005). Hurricanes and global warming. Bull. Amer. Meteorol. Soc., 86, 1571–1575. Pradhan, P. K., Preethi, B., Ashok, K., Krishnan, R., and Sahai, A. K. (2011). Modoki, Indian Ocean Dipole, and western North Paciﬁc typhoons: Possible implications for extreme events. J. Geophys. Res., 116(D18), D18108. Ramage, C. S. (1950). Hurricane Development. J. Met., 21(9), 917–925. Ramsay, H. A., Leslie, L. M., Lamb, P. J., and Richman, M. B. (2008). Interannual variability of tropical cyclones in the Australian region: Role of large-scale environment. J. Climate, 21, 1083–1103. Rappin, E. D., Nolan, D. S., and Emanuel, K. A. (2010). Thermodynamic control of tropical cyclogenesis in environments of radiative-convective equilibrium with shear. Quarterly Journal of the Royal Meteorological Society, 136(653), 1954–1971. Rayner, N. A., Parker, D. E., Horton, E. B., Folland, C. K., Alexander, L. V., Rowell, D. P., Kent, E. C., and Kaplan, A. (2003). Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108(D14), 4407. Reed, K. A. and Jablonowski, C. (2010). An Analytic Vortex Initialization Technique for Idealized Tropical Cyclone Studies in AGCMs. Monthly Weather Review, 139(2), 689–710. Reed, K. A. and Jablonowski, C. (2012). Idealized tropical cyclone simulations of inter- mediate complexity: A test case for AGCMs. J. Adv. Model. Earth Syst., 4, M04001. Ren, H.-L. and Jin, F.-F. (2011). Nino indices for two types of ENSO. Geophys. Res. Lett., 38(4), L04704. Richter, I. and Xie, S.-P. (2008). Muted precipitation increase in global warming simulations: A surface evaporation perspective. J. Geophys. Res., 113(D24), D24118.

Page 206 Bibliography

Rielh, H. (1948). On the formation of typhoons. J. Met., 5(6), 247–264. Rienecker, M. M., Suarez, M. J., Gelaro, R., Todling, R., Bacmeister, J., Liu, E., Bosilovich, M. G., Schubert, S. D., Takacs, L., Kim, G.-K., Bloom, S., Chen, J., Collins, D., Conaty, A., da Silva, A., Gu, W., Joiner, J., Koster, R. D., Lucchesi, R., Molod, A., Owens, T., Pawson, S., Pegion, P., Redder, C. R., Reichle, R., Robertson, F. R., Ruddick, A. G., Sienkiewicz, M., and Woollen, J. (2011). MERRA: NASAs Modern-Era Retrospective Analysis for Research and Applications. Journal of Cli- mate, 24(14), 3624–3648. Ringer, M. A., Martin, G. M., Greeves, C. Z., Hinton, T. J., James, P. M., Pope, V. D., Scaife, A. A., Stratton, R. A., Inness, P. M., Slingo, J. M., and Yang, G.-Y. (2006). The Physical Properties of the Atmosphere in the New Hadley Centre Global Environ- mental Model (HadGEM1). Part II: Aspects of Variability and Regional Climate. J. Climate, 19(7), 1302–1326. Ritchie, E. A. and Holland, G. J. (1999). Large-scale patterns associated with tropical cyclogenesis in the western Paciﬁc. J. Met., 127, 2027–2043. Roberts, M., Vidale, P. L., Bell, R., and Hodges, K. (2013). Sensitivity of tropical cyclones simulation to SST forcing. U.S. CLIVAR, 11(3), 12–17. Roberts, M. J., Banks, H., Gedney, N., Gregory, J., Hill, R., Mullerworth, S., Pardaens, A., Rickard, G., Thorpe, R., and Wood, R. (2004). Impact of an Eddy-Permitting Ocean Resolution on Control and Climate Change Simulations with a Global Coupled GCM. Journal of Climate, 17(1), 3–20. Roberts, M. J., Clayton, A., Demory, M.-E., Donners, J., Vidale, P. L., Norton, W., Shaf- frey, L., Stevens, D. P., Stevens, I., Wood, R. A., and Slingo, J. (2009). Impact of Resolution on the Tropical Paciﬁc Circulation in a Matrix of Coupled Models. J. Cli- mate, 22(10), 2541–2556. Romps, D. M. (2010). Response of Tropical Precipitation to Global Warming. Journal of the Atmospheric Sciences, 68(1), 123–138. Rotunno, R. and Emanuel, K. A. (1987). An Air-Sea Interaction Theory for Tropical Cy- clones. Part II: Evolutionary Study Using a Nonhydrostatic Axisymmetric Numerical

Page 207 Bibliography

Model. Journal of the Atmospheric Sciences, 44(3), 542–561. Saha, S., Moorthi, S., Pan, H.-L., Wu, X., Wang, J., Nadiga, S., Tripp, P., Kistler, R., Woollen, J., Behringer, D., Liu, H., Stokes, D., Grumbine, R., Gayno, G., Wang, J., Hou, Y.-T., Chuang, H.-Y., Juang, H.-M. H., Sela, J., Iredell, M., Treadon, R., Kleist, D., Van Delst, P., Keyser, D., Derber, J., Ek, M., Meng, J., Wei, H., Yang, R., Lord, S., Van Den Dool, H., Kumar, A., Wang, W., Long, C., Chelliah, M., Xue, Y., Huang, B., Schemm, J.-K., Ebisuzaki, W., Lin, R., Xie, P., Chen, M., Zhou, S., Higgins, W., Zou, C.-Z., Liu, Q., Chen, Y., Han, Y., Cucurull, L., Reynolds, R. W., Rutledge, G., and Goldberg, M. (2010). The NCEP Climate Forecast System Reanalysis. Bulletin of the American Meteorological Society, 91(8), 1015–1057. Sampe, T. and Xie, S.-P. (2007). Mapping high sea winds from space. Bull. Amer. Meteorol. Soc., 88, 1965–1978. Santer, B. D., Wigley, T. M. L., Gleckler, P. J., Bonfils, C., Wehner, M. F., AchutaRao, K., Barnett, T. P., Boyle, J. S., Bruggemann,¨ W., Fiorino, M., Gillett, N., Hansen, J. E., Jones, P. D., Klein, S. A., Meehl, G. A., Raper, S. C. B., Reynolds, R. W., Taylor, K. E., and Washington, W. M. (2006). Forced and unforced ocean temperature changes in Atlantic and Pacific tropical cyclogenesis regions. Proceedings of the National Academy of Sciences, 103(38), 13905–13910. Saunders, M. A., Chandler, R. E., Merchant, C. J., and Robers, F. P. (2000). Atlantic hurricanes and NW Pacific typhoons: ENSO spatial impacts on occurrence and landfall. Geophys. Res. Lett., 27(8), 1147–1150. Schenkel, B. (2012). Assessing the Climate Footprint of Tropical Cyclones: Pertinent Players or Irrelevant Pawns? In 30th Conference on Hurricanes and Tropical Meteo- rology. Schenkel, B. A. and Hart, R. E. (2011). An Examination of Tropical Cyclone Position, Intensity, and Intensity Life Cycle within Atmospheric Reanalysis Datasets. Journal of Climate, 25(10), 3453–3475. Schneider, T., O’Gorman, P. A., and Levine, X. J. (2010). Water Vapor and the Dynamics of Climate Changes. Rev. Geophys., 48(3), RG3001.

Page 208 Bibliography

Scoccimarro, E., Gualdi, S., Bellucci, A., Sanna, A., Giuseppe Fogli, P., Manzini, E., Vichi, M., Oddo, P., and Navarra, A. (2011). Effects of Tropical Cyclones on Ocean Heat Transport in a High-Resolution Coupled General Circulation Model. Journal of Climate, 24(16), 4368–4384. Shaffrey, L. C., Stevens, I., Norton, W. A., Roberts, M. J., Vidale, P. L., Harle, J. D., Jrrar, A., Stevens, D. P., Woodage, M. J., Demory, M. E., Donners, J., Clark, D. B., Clayton, A., Cole, J. W., Wilson, S. S., Connolley, W. M., Davies, T. M., Iwi, A. M., Johns, T. C., King, J. C., New, A. L., Slingo, J. M., Slingo, A., Steenman-Clark, L., and Martin, G. M. (2009). U.K. HiGEM: The New U.K. High-Resolution Global Environment Model-Model Description and Basic Evaluation. J. Climate, 22(8), 1861–1896. Shaman, J. and Maloney, E. D. (2011). Shortcomings in climate model simulations of the ENSO-Atlantic hurricane teleconnection. Climate Dyn., 38, 1973–1988. Shaman, J., Esbensen, S. K., and Maloney, E. D. (2009). The Dynamics of the ENSO- Atlantic Hurricane Teleconnection: ENSO-Related Changes to the North African- Asian Jet Affect Atlantic Basin Tropical Cyclogenesis. J. Climate, 22(9), 2458–2482. Shay, L. (2010). Air-sea interactions in tropical cyclones. In Global Perspectives on Tropical Cyclones: From science to mitigation, pages 93–132. World Scientiﬁc Pub- slishing Co. Shen, W., Tuleya, R. E., and Ginis, I. (2000). A Sensitivity Study of the Thermodynamic Environment on GFDL Model Hurricane Intensity: Implications for Global Warming. Journal of Climate, 13(1), 109–121. Sherwood, S. C., Ingram, W., Tsushima, Y., Satoh, M., Roberts, M., Vidale, P. L., and O’Gorman, P. A. (2010). Relative humidity changes in a warmer climate. J. Geophys. Res., 115(D9), D09104. Shi, W. and Wang, M. (2011). Satellite observations of asymmetrical physical and bio- logical responses to Hurricane Earl. Geophys. Res. Lett., 38(4), L04607. Simpson, R. H. and Riehl, H. (1981). The hurricane and its impact, page 398. Louisianna State University Press. Singh, O. P., Khan, T. M. A., and Rahman, M. S. (2000). Changes in the frequency of

Page 209 Bibliography

tropical cyclones over the North Indian Ocean. Meteorol. Atmos. Phys., 75, 11–20. Slingo, J., Bates, K., Nikiforakis, N., Piggott, M., Roberts, M., Shaffrey, L., Stevens, I., Vidale, P. L., and Weller, H. (2009). Developing the next-generation climate system models: challenges and achievements. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 367(1890), 815–831. Smith, D. M., Eade, R., Dunstone, N. J., Fereday, D., Murphy, J. M., Pohlmann, H., and Scaife, A. (2010). Skilful multi-year predictions of Atlantic hurricane frequency. Nature Geoscience, 3(12), 846–849. Smith, R. K., Montgomery, T., and Vogl, S. (2008). A critique of Emanuels hurricane model and potential intensity theory. Q. J. R. Meteorol. Soc., 134, 551–561. Smith, S. R., Brolley, J., OBrien, J. J., and Tartaglione, C. A. (2007). ENSOs Impact on Regional U.S. Hurricane Activity. Journal of Climate, 20(7), 1404–1414. Smith, T. M., Arkin, P. A., Bates, J. J., and Huffman, G. J. (2006). Estimating Bias of Satellite-Based Precipitation Estimates. J. Hydromet., 7(5), 841–856. Solomon, A. and Newman, M. (2012). Reconciling disparate twentieth-century Indo- Paciﬁc ocean temperature trends in the instrumental record. Nature Clim. Change, 2(9), 691–699. Spencer, H. and Slingo, J. M. (2003). The Simulation of Peak and Delayed ENSO Tele- connections. J. Climate, 16(11), 1757–1774. Spencer, H., Slingo, J. M., and Davey, M. K. (2004). Seasonal predictability of ENSO teleconnections: the role of the remote ocean response. Climate Dynamics, 22(5), 511–526 LA – English. Sriver, R. L. (2010). Tropical cyclones in the mix. Nature, 463(25), 1032–1033. Stott, P. A., Tett, S. F. B., Jones, G. S., Allen, M. R., Mitchell, J. F. B., and Jenkins, G. J. (2000). External Control of 20th Century Temperature by Natural and Anthropogenic Forcings. Science, 290(5499), 2133–2137. Stott, P. A., Stone, D. A., and Allen, M. R. (2004). Human contribution to the European heatwave of 2003. Nature, 432(7017), 610–614. Strachan, J., Vidale, P. L., Hodges, K., Roberts, M., and Demory, M.-E. (2013). Investi-

Page 210 Bibliography

gating global tropical cyclone activity with a hierarchy of AGCMs: the role of model resolution. J. Climate, 26, 133–152. Sugi, M., Noda, A., and N. Sata (2002). Influence of global warming on tropical cyclone climatology: an experiment with the JMA global model. J. Meteor. Soc. Japan, 80, 249–272. Sugi, M., Murakami, H., and Yoshimura, J. (2009). A Reduction in Global Tropical Cyclone Frequency due to Global Warming. SOLA, 5, 164–167. Sugi, M., Murakami, H., and Yoshimura, J. (2012). On the mechanism of tropical cyclone frequency changes due to global warming. J. Meteor. Soc. Japan, 90A, 397–408. Tanabe, S. (1963). Low Latitude Analysis at the Formative Stage of Typhoons in 1962. Kishocho Kenkyu Jiho, 15(6), 405–418. Tang, B. and Emanuel, K. (2010). Midlevel Ventilations Constraint on Tropical Cyclone Intensity. Journal of the Atmospheric Sciences, 67(6), 1817–1830. Tang, B. H. and Neelin, J. D. (2004). ENSO Influence on Atlantic hurricanes via tropospheric warming. Geophysical Research Letters, 31(24), 1–4. Tao, D. (2012). Impact of vertical wind shear on the predictability of tropical cyclones. In 30th Conference on Hurricanes and Tropical Meteorology. Taylor, K., Williamson, D., and Zwiers, F. W. (2000). The sea surface temperature and sea-ice boundary conditions for AMIPII simulations. Technical report, PCMDI. Tippett, M. K., Camargo, S. J., and Sobel, A. H. (2010). A Poisson Regression Index for Tropical Cyclone Genesis and the Role of Large-Scale Vorticity in Genesis. J. Climate, 24(9), 2335–2357. Tory, K. and Frank, W. (2010). Tropical cyclone formation. In Global Perspectives on Tropical Cyclones: From science to mitigation, pages 55–92. World Scientific Pub- slishing Co. Tory, K. J., Chand, S. S., McBride, J. L., Ye, H., and Dare, R. A. (2013). Projected changes in late 21st century tropical cyclone frequency in thirteen coupled climate models from the Coupled Model Intercomparison Project Phase 5. Journal of Climate. Trenberth, K. E. and Shea, D. J. (2006). Atlantic hurricanes and natural variability in

Page 211 Bibliography

2005. Geophys. Res. Lett., 33(12), L12704. Trenberth, K. E. and Stepaniak, D. P. (2003). Covariability of Components of Poleward Atmospheric Energy Transports on Seasonal and Interannual Timescales. Journal of Climate, 16(22), 3691–3705. Tsutsui, J. (2002). Implications of Anthropogenic Climate Change for Tropical Cyclone Activity: A Case Study with the NCAR CCM2. Journal of the Meteorological Society of Japan, 80(1), 45–65. Vecchi, G. A. and Knutson, T. R. (2010). Estimating Annual Numbers of Atlantic Hur- ricanes Missing from the HURDAT Database (1878-1965) Using Ship Track Density. Journal of Climate, 24(6), 1736–1746. Vecchi, G. A. and Soden, B. J. (2007a). Effect of remote sea surface temperature change on tropical cyclone potential intensity. Nature, 450(7172), 1066–70. Vecchi, G. A. and Soden, B. J. (2007b). Increased tropical Atlantic wind shear in model projections of global warming. Geophysical Research Letters, 34(8). Vecchi, G. A., Soden, B. J., Wittenberg, A. T., Held, I. M., Leetmaa, A., and Harrison, M. J. (2006). Weakening of tropical Paciﬁc atmospheric circulation due to anthropogenic forcing. Nature, 441(7089), 73–76. Vecchi, G. A., Fueglistaler, S., Held, I. M., Knutson, T. R., and Zhao, M. (2013). Impacts of Atmospheric Temperature Trends on Tropical Cyclone Activity. J. Climate, 26(11), 3877–3891. Vitart, F. and Anderson, J. L. (2001). Sensitivity of Atlantic tropical storm frequency to ENSO and interdecadal variability of SSTs in an ensemble of AGCM integrations. J. Climate, 14(4), 533–545. Vitart, F., Anderson, D., and Stockdale, T. (2003). Seasonal Forecasting of Tropical Cyclone Landfall over Mozambique. Journal of Climate, 16(23), 3932–3945. Walker, C. C. and Schneider, T. (2006). Eddy Inﬂuences on Hadley Circulations: Simu- lations with an Idealized GCM. Journal of the Atmospheric Sciences, 63(12), 3333– 3350. Walker, G. (1923). Correlation in seasonal variations of weather, VIII. A preliminary

Page 212 Bibliography

study of world weather. Memoirs of the India Meteorological Department, 24(4), 75– 131. Walker, G. (1924). Correlations in seasonal variations of weather, IX. Memoirs of the India Meteorological Department, 24, 333–345. Walsh, K., Lavender, S., Murakami, H., Scoccimarro, E., Caron, L.-P., and Ghantous, M. (2010). The Tropical Cyclone Climate Model Intercomparison Project. Walsh, K. J. E., Fiorino, M., Landsea, C. W., and McInnes, K. L. (2007). Objectively Determined Resolution-Dependent Threshold Criteria for the Detection of Tropical Cyclones in Climate Models and Reanalyses. Journal of Climate, 20(10), 2307–2314. Wang, B. and Chan, J. C. L. (2002). How strong ENSO events affect tropical storm activity over the western North Pacific. J. Climate, 15(13), 1643–1658. Wang, C. (2002). Atlantic Climate Variability and Its Associated Atmospheric Circula- tion Cells. J. Climate, 15(13), 1516–1536. Wang, C. and Lee, S.-K. (2008). Global warming and United States landfalling hurricanes. Geophysical Research Letters, 35(2), 2–5. Wang, C. and Wang, X. (2012). Classifying El Nino˜ Modoki I and II by Different Impacts on Rainfall in Southern China and Typhoon Tracks. Journal of Climate, 26(4), 1322– 1338. Wang, C., Deser, C., Yu, Y.-Y., DiNezio, P., and Clement, A. (2012). El Nino and South- ern Oscillation (ENSO): A review. In Coral Reefs of the Eastern Pacific. Springer. Wang, C., Li, C., Mu, M., and Duan, W. (2013). Seasonal modulations of different impacts of two types of ENSO events on tropical cyclone activity in the western North Pacific. Climate Dyn., 40(11-12), 2887–2902. Webster, P. J., Holland, G. J., Curry, J. A., and Chang, H.-R. (2005). Changes in Tropical Cyclone Number, Duration, and Intensity in a warming environment. Science, 309, 1844–1846. Weinkle, J., Maue, R., and Pielke, R. (2012). Historical Global Tropical Cyclone Land- falls. Journal of Climate, 25(13), 4729–4735. Wu, A. and Hsieh, W. W. (2004). The nonlinear Northern Hemisphere winter atmospheric

Page 213 Bibliography

response to ENSO. Geophys. Res. Lett., 31(2), L02203. Wu, G. and Lau, N.-C. (1992). A GCM simulation of the relationship between tropical- storm formation and ENSO. Mon. Weath. Rev., 120, 958–977. Wu, M. C., Chang, W. L., and Leung, W. M. (2004). Impacts of El Nino-Southern Oscil- lation Events on Tropical Cyclone Landfalling Activity in the Western North Paciﬁc. J. Climate, 17(6), 1419–1428. Wyrtki, K. (1975). El Nino-The˜ Dynamic Response of the Equatorial Paciﬁc Ocean to Atmospheric Forcing. Journal of Physical Oceanography, 5(4), 572–584. Xie, L., Pietrafesa, L. J., and Wu, K. (2002a). Interannual and decadal variability of landfalling tropical cyclones in the southeast coastal states of the United States. Adv. Atmos. Sci., 19, 677–686. Xie, S.-P., Annamalai, H., Schott, F. A., and McCreary, J. P. (2002b). Structure and mechanisms of South Indian Ocean climate variability. J. Climate, 25, 864–878. Xie, S.-P., Deser, C., Vecchi, G. A., Ma, J., Teng, H., and Wittenberg, A. T. (2010). Global Warming Pattern Formation: Sea Surface Temperature and Rainfall. Journal of Climate, 23(4), 966–986. Yanai, M. (1964). Formation of tropical cyclones. Rev. Geophys., 2(2), 367–414. Yanai, M., Esbensen, S., and Chu, J.-H. (1973). Determination of Bulk Properties of Tropical Cloud Clusters from Large-Scale Heat and Moisture Budgets. Journal of the Atmospheric Sciences, 30(4), 611–627. Yeh, S.-W., Kug, J.-S., Dewitte, B., Kwon, M.-H., Kirtman, B. P., and Jin, F.-F. (2009). El Nino˜ in a changing climate. Nature, 461(7263), 511–4. Yoshimura, J. (2005). Tropical Cyclone Climatology in a High-resolution AGCM Im- pacts of SST Warming and CO2 Increase. SOLA, 1, 133–136. Zeng, Z., Wang, Y., and Wu, C.-C. (2007). Environmental Dynamical Control of Tropical Cyclone Intensity-An Observational Study. Monthly Weather Review, 135(1), 38–59. Zhang, R. and Delworth, T. L. (2006). Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geophysical Research Letters, 33(17), 1–5.

Page 214 Bibliography

Zhang, R., Delworth, T. L., Sutton, R., Hodson, D. L. R., Dixon, K. W., Held, I. M., Kushnir, Y., Marshall, J., Ming, Y., Msadek, R., Robson, J., Rosati, A. J., Ting, M., and Vecchi, G. A. (2013). Have Aerosols Caused the Observed Atlantic Multidecadal Variability? Journal of the Atmospheric Sciences, 70(4), 1135–1144. Zhang, W., Graf, H.-F., Leung, Y., and Herzog, M. (2012). Different El Nino˜ Types and Tropical Cyclone Landfall in East Asia. Journal of Climate, 25(19), 6510–6523. Zhao, M. and Held, I. M. (2011). TC-Permitting GCM Simulations of Hurricane Frequency Response to Sea Surface Temperature Anomalies Projected for the Late- Twenty-First Century. J. Climate, 25(8), 2995–3009. Zhao, M., Held, I. M., Lin, S. J., and Vecchi, G. A. (2009). Simulations of Global Hur- ricane Climatology, Interannual Variability, and Response to Global Warming Using a 50-km Resolution GCM. J. Climate, 22(24), 6653–6678. Zhao, M., Held, I. M., and Lin, S.-J. (2012). Some Counterintuitive Dependencies of Tropical Cyclone Frequency on Parameters in a GCM. Journal of the Atmospheric Sciences, 69(7), 2272–2283. Zhu, X., Saravanan, R., and Chang, P. (2011). Inﬂuence of Mean Flow on the ENSO- Vertical Wind Shear Relationship over the Northern Tropical Atlantic. Journal of Cli- mate, 25(3), 858–864.

Page 215