The influence of large-scale climate modes on tropical cyclone tracks in the southwest Pacific

Tropical cyclones (TCs) impact the economy, properties, lives and infrastructure of island nations and territories of the southwest Pacific (SWP), accounting for three in four regional disasters each year. To increase the resilience of the SWP to the destructive impacts of TCs, improved TC track forecasts are needed since a high degree of uncertainty exists around the likely path a TC will take in this region post-formation. This requires better comprehension of the factors contributing to TC track variability occurring at different timescales. Therefore, we examine the modulating impact of key Indo-Pacific climate drivers: the El Niño-Southern Oscillation (ENSO), Interdecadal Pacific Oscillation (IPO), Southern Annular Mode (SAM) and the Indian Ocean Dipole (IOD), on SWP TC track variability. We present new insights into the spatial (i.e. prevailing trajectories) and temporal (i.e. track length, average speed and duration) components of TC tracks, being modulated by both individual and combined climate modes. Overall, TC tracks tend to shift northeast during El Niño, IPO positive, IOD east positive and/or SAM negative phases (with a southwest shift observed during the opposite climate phases). Further, we show that when two of these climate modes are in their positive phase (e.g. El Niño with the positive phases of IPO or eastern pole of IOD and SAM), TC track length and average speed are enhanced. However, for cases where either one (e.g. El Niño/negative phase of IPO and IOD east) or two (La Niña/negative phase of IPO, IOD east and SAM) climate modes were in the negative phase, an increase in TC track duration was observed. The findings of this study may be used to improve TC forecasting and better quantify TC-related risks.


3 Introduction
Tropical cyclones (TCs) are the most damaging natural hazard (Stephens and Ramsay 2014) that occur across seven basins with a global average of 86 TCs per year (Murakami et al. 2020). Of those, approximately 12-13 TCs occur across the southwest Pacific (SWP; 0°-35°S, 135°E-120°W) (Fig. 1, panel a) region every season (November-April) (Diamond et al. 2013), causing significant impacts on properties, lives and infrastructure. Home to 15 diverse island nations and territories, the population of the SWP are highly vulnerable to the impact of TCs, which have accounted for approximately 1350 deaths since 1950 (EM-DAT 2020). The region's vulnerability to TC impacts (as well as other forms of natural hazards) can be attributed to the geographical location and varying topographies of SWP nations and territories. In addition, most of the population is located no farther than 10 km from the coast (Andrew et al. 2019) for reasons including accessibility to subsistence resources (i.e. dependence upon the ocean for food supply), trade and transport and economic development (Small and Nicholls 2003;Neumann et al. 2015). As such, this means that many vulnerable communities reside in the low-elevation coastal zones, which are highly vulnerable to both short-term (e.g. TCs, storm surges, tsunamis) and long-term (e.g. a rise in sea level) natural disasters (Naylor 2015).
The impact of TCs on the SWP island nations and territories depends on their trajectory (Saverimuttu and Varua 2017), as they often deviate from their modelled/predicted paths and sometimes display erratic behaviour. While most TCs in the SWP region move towards the east/southeast, with few moving towards the west (Holland and Gray 1983), their orientation can be anywhere from the northwest to the southeast (Terry 2007). TCs are also known to meander or oscillate around this mean path (Holland and Lander 1993), with tracks often characterised in terms of duration (long-/short-lived), speed (slow-moving/fast-moving), direction (differing from the expected trend) and shape (from straight to erratic/looping), or a combination of all the above. Erratic behaviour displayed by TC tracks is a distinct feature of the SWP basin compared to other TC basins across the globe (Australian Bureau of Meteorology 2020). As identified by Sharma et al. (2021), TCs take different paths than expected, representing different risks to island nations. Erratic TC tracks are more common within the Coral Sea region; however, the long-term trend also indicates potential changes in the morphology of tracks (as more convoluted contrary to being mostly straight type) within the eastern part of SWP (Sharma et al. 2021). Further, research has highlighted recent variations in TC tracks in the form of poleward shifts, i.e. TCs transition into mid-latitudes or extratropical cyclones (Sinclair 2002;Christensen et al. 2013;Kossin et al. 2014;Sharmila and Walsh 2018). This changing vulnerability for certain island nations to TC track exposure requires further investigation to understand how, where, when and why TCs track across the SWP (post-formation) and impact island nations (either passing by an island nation or making landfall).
The SWP is exceptionally dynamic, and TC activity within the region is modulated by a number of large-scale modes of climate variability operating within and outside of the region (Grant and Walsh 2001;Camargo et al. 2010;Chand et al. 2013;Diamond et al. 2013;Diamond and Renwick 2015;Magee et al. 2017;Magee and Verdon-Kidd 2018). These modes operate over seasonal through to multi-year cycles and include the El Niño-Southern Oscillation (ENSO), Interdecadal Pacific Oscillation (IPO), Indian Ocean Dipole (IOD), Southern Annular Mode (SAM) and Madden Julian Oscillation (MJO). While many previous studies have focused on TC genesis, few studies have shown the potential impact of these large-scale drivers on TC track morphology, except for the well-known ENSO, where it has been shown that TC tracks follow its wellestablished influence, that is, east-northeast modulations during El Niño (EN) and westsouthwest during La Niña (LN) (e.g. Sinclair 2002;Chand and Walsh 2009;Malsale 2011;Ramsay et al. 2012).
Similar to ENSO but operating on decadal and multidecadal timescales, the IPO is also known to modulate TC activity within the SWP region (Grant and Walsh 2001;Magee et al. 2017). However, studies linking the impact of IPO on SWP TCs are limited-mainly due to the question of the reliability of TC data from the pre-satellite era (before 1970) (Holland 1981). Since IPO is a multidecadal phenomenon, long-term data sets are imperative for analysing relationships. Initial investigations by Grant and Walsh (2001) found that IPO negative conditions resulted in more landfalling TCs across eastern Australia, driven by more favourable wind shear conditions towards the continent; however, their analysis was temporally limited to 1996. Magee et al. (2017) investigated the impact of the IPO on TC genesis and found that IPO positive (negative) conditions result in a northeast (southwest) migration of TC activity, regardless of the ENSO phase. However, to date, no studies have investigated the impact of IPO variability and how the co-occurrence of particular ENSO/IPO phases can influence SWP TC tracks.
Other than Pacific drivers (ENSO and IPO), the other modes of climate variability known to influence SWP (or wider SH) TCs include the SAM and Indian Ocean sea surface temperature (SST) variability (Kidson and Sinclair 1995;Ashok et al. 2007;Renwick and Thompson 2006;Christensen et al. 2013;Saha and Wasimi 2013;Diamond and Renwick 2015;Ramsay et al. 2017;Magee and Verdon-Kidd 2018). Kidson and Sinclair (1995) noted that variation in the SAM resulted in the substantial north-south migration of storm tracks in the Southern Hemisphere (SH). Diamond and Renwick (2015) reported that SAM has no significant impact on SWP TC activities, i.e. TC genesis, tracks and intensity, but influences TC transition into extratropical cyclones near New Zealand.
Indeed, previous studies (e.g. Yin 2005; Kidston and Gerber 2010;Christensen et al. 2013) have emphasised on the projected poleward migration of storm tracks by the high index state of the SAM, augmented by the increased surface temperature gradient, poleward shifts in the wind stress and precipitation. Thus, further work is needed to understand the influence of the SAM on these factors. While the influence of Indian Ocean SST variability on SH TC activity has been widely investigated, there are fewer studies accounting for their influence on SWP TC tracks. Saha and Wasimi (2013) highlighted the link between Indian Ocean SST variability and Australian/SWP TC activity. Magee and Verdon-Kidd (2018) revealed a new relationship between the Indian Ocean SST variability and SWP TC genesis. Anomalously warm (cool) SSTs in the Indian Ocean resulted in a northeasterly (southwesterly) shift in TC genesis, even during ENSO neutral conditions. They also hinted that their findings might have significant implications for TC tracks in the SWP region.
The climate modes discussed here do not operate in isolation, and as such, the prevailing environmental conditions across the SWP represent the combined influence of these individual modes of variability. Magee et al. (2017) have shown that when the suites of climate drivers are considered in co-occurrence, the location of TC genesis is significantly shifted, raising the question of how this may impact the entire TC track. Therefore, this paper aims to investigate the spatio-temporal influence of climate drivers (both individually and coupled) on the SWP TC tracks and their morphology. The following objectives will be addressed to achieve this aim: (1) Investigate the individual influence of Pacific (ENSO and IPO) and remote climate drivers (Indian Ocean SST variability and SAM) on SWP TC track activity.
(2) Explore the combined influence of Pacific and remote drivers on the spatio-temporal variability of SWP TC tracks. (3) Assess how climate mode modulation of TC tracks alters the likelihood of TCs passing over or by each island nation in the SWP.
Identifying how climate drivers interact at varying timescales (interannual and interdecadal), and contribute to spatio-temporal variability of SWP TC tracks, may enable meteorologists and forecasting agencies to produce more reliable TC forecasts and seasonal outlooks and allow better decision-making when planning for future TC events. The findings could complement existing TC outlooks with more information on probable track trajectories to increase societies' resilience to the destructive impacts of TCs by estimating the location of landfall/impact once a TC forms.

TC track data
This study uses the South Pacific Enhanced Archive of Tropical Cyclones (SPEArTC) database for best track TC data, spanning between 1948 and 2021, with track locations available at 6-h intervals for the entire TC lifecycle (Diamond et al. 2012). The first point of each unique track in the SPEArTC data set is considered as the TC genesis point (TC formation), and the last point as the TC decay point (i.e. where TC ended/terminated). This data set was chosen because it is considered the most complete repository of TC data for the SWP region (Magee et al. 2016;Sharma et al. 2020).

Climate indices
The climate modes investigated in this study include ENSO, Indian Ocean SST variability, IPO and SAM-a diagrammatic summary is illustrated in Fig. 1 (panel a) along with their timeseries (panels c-f). Further details on how each index was derived are described below: • ENSO index: monthly Oceanic Niño Index (ONI) utilised in this study was obtained from the National Oceanic and Atmospheric Administration (NOAA), Climate Prediction Centre (Kousky and Higgins 2007). ONI is based on the 3-month running mean of Extended Reconstructed Sea Surface Temperature Version 5 (ERSSTv5) (Huang et al. 2017a;2017b)  (5°N-5°S, 120°-170°W). The anomaly calculation of ONI is based on the multiplecentred 30-year base period, which is updated every five years. This approach ensures that the classification of EN and LN events is defined by their contemporary climatology and remains fixed over most of the historical period. A threshold of ± 0.5 °C is used to define the EN and LN phases when the periods of below-/above-average SSTs occur for at least five consecutive overlapping seasons, while SSTs in-between are classified as ENSO neutral events. • IPO Index (Tripole Index [TPI]): calculated as the difference between the average SST anomalies over the central equatorial Pacific (10°S-10°N, 170°E-90°W) and the average of the SST anomalies in the Northwest Pacific (25°N-45°N, 140°E-145°W) and Southwest Pacific (50°S-15°S, 150°E-60°W). The TPI is calculated using ERSSTv5 SST anomalies derived from NOAA and is available in filtered and unfiltered versions. For the present study, the filtered IPO was used that is obtained by applying a 13-year Chebyshev low-pass filter (Henley et al. 2015), covering the period from 1948 to 2015. • IOD E (eastern Indian Ocean dipole): monthly SST anomalies within the south-eastern Indian Ocean (0°-10°S, 90°E-110°E), whereby positive (negative) IOD E phase indicates warmer (cooler) SSTs (Saji et al. 1999). We also investigated the Dipole Mode Index; however, we found that the eastern pole of the Indian Ocean Dipole has the greatest impact, which is consistent with studies on TC genesis (Magee and Verdon-Kidd 2018). • SAM: is the monthly differences in the mean sea level pressure observed from six insitu stations close to 40°S and 65°S (Marshall 2003;Ho et al. 2012;Diamond and Renwick 2015) and is calculated using the following equation after Gong and Wang (1999): Based on the recommendations of Ho et al. (2012) and its application on SWP TCs (Diamond and Renwick 2015), the Marshall Index was selected, which is available since 1957. To extend the SAM index to match the start date of the present study (1948), 20th Century Reanalysis Version 2c (20CRv2c) sea level pressure data, available from 1851 to 2011, were used (Gong and Wang 1999). Prior to merging, the zonal mean of the Marshall Index was standardised using a similar climatological baseline as 20CRv2c . SAM was then calculated using Eq. 1.
The monthly anomaly values for all climate indices were calculated using the 1981-2010 climatological baseline, except for TPI (which uses the 1971-2000 baseline) (World Meteorological Organisation 2017). Each climate mode's positive, negative and neutral phases were defined using a threshold of ± 0.5 standard deviations from the mean.

Spatio-temporal analysis techniques
The modulation of TC tracks by the climate drivers was assessed in two ways. Firstly, the individual influence of climate drivers on TC tracks was examined spatially through density plots. TC tracks (vector data) were superimposed over 3 × 3° grid cells to transform into a raster. The raster data were normalised according to the length of each climate driver's phase months (seasonal) over 74 years (i.e. from 1948-2021). The difference in TC (1) SAM = P * 40 • S − P * track density was then computed between the positive-negative, positive-neutral and negative-neutral phases of each climate driver. The combined influence of climate drivers on SWP TC tracks was examined by pairing ENSO phases with the positive, negative and neutral phases of other climate modes (IPO and SAM) and indices (IOD E). For example, EN was paired with the positive, negative and neutral phases of IOD E, and TC track trajectories were stratified and analysed for each of these pairings, respectively. A total of 27 pairings were considered in this paper when analysing spatio-temporal modulations in TC trajectories. The temporal characteristics of TC tracks, including duration, track length and average moving speed, were also evaluated for individual climate drivers and the multimode pairings. TC speed in this study is defined as the average moving speed of a TC track over its lifetime and was calculated using the TC track length (km) over its entire duration (in hours), with the assumption that almost all timestamps in SPEArTC are 6-hourly. An unpaired two-tail Student's t-test was performed to determine whether the observed differences in TC characteristics between the phases of each climate driver were statistically significant.
Finally, to identify and evaluate how landfall changes during each climate driver's phases (individually and combined) on individual SWP island and atoll nations and territories, we calculated the mean number of TCs per season observed within each island nation's Exclusive Economic Zone, EEZ ( Fig. 1, panel b) (Flanders Marine Institute 2018). The EEZ is an area beyond and adjacent to the territorial sea with an outer limit not exceeding 200 nautical miles (Flanders Marine Institute 2018). Note: based on the study region (Fig. 1, panel a), the entire Australian region EEZ is not considered for analysis, but the region straddling the 135°E boundary.

Pacific climate mode influences on TC tracks (ENSO and IPO)
The differenced density plots of SWP TC tracks for EN, LN and ENSO neutral phases are illustrated in Fig. 2 (panels a-c), with the accompanying boxplots showing TC duration, track length and average moving speed (hereafter referred to as average speed) laterally organised in panels d-f. The results of the Student's t-test (p-values) for all Pacific climate modes are shown in Table 1. Figure 2 (panels a-c) displays the typical east-northeast and west-southwest modulation of SWP TC activity during EN and LN conditions. New insights are also presented regarding the impact of ENSO on the temporal characteristics of SWP TC tracks (panels d-f). For instance, the duration plots (panel d) reveal that during EN, TC tracks have a shorter lifetime (mean duration of 9.69 days) compared to LN (mean duration of 10.32 days). Conversely, the TC tracks during LN (mean of 3392 km) exhibit longer trajectories than during EN (mean of 3160 km in panel e). Consequently, these attributes result in TCs that are faster-moving (slower-moving) during LN (EN), with a mean speed of 15.32 km/hr (14.27 km/hr).   The spatio-temporal variability during IPO (Fig. 3) indicates increased TC track activity in the central and eastern Pacific during IPO positive (compared to IPO negative), evident in panels a and b of Fig. 3. This differs from the northeast/southwest shift in trajectory observed during ENSO events (Fig. 2). Further, panels d-f and Table 1 confirm that IPO positive (negative) tracks tend to be significantly fast-moving (slow-moving) with an average speed of 15.85 km/hr (13.40 km/hr), but with a shorter (longer) mean lifetime of 9.38 days (10.87 days) and lengthier (shorter) tracks of 3457 km (2951 km) on average.
Indeed, such spatio-temporal modulations also increase the likelihood of risk impact over the SWP island nations, depicted in panel g of Fig. 3. It is observed that during IPO negative periods, higher number of TCs are observed, regardless of the nation's geographical location and/or spatial modulation during the negative phase. The nations with the highest risk during IPO negative phase are located in the western SWP, including Australia (3.16 TCs/season), New Caledonia (2.54 TCs/season), Norfolk Island (0.71 TCs/season), PNG (0.99 TCs/season), Solomon Islands (1.78 TCs/season) and Vanuatu (1.66 TCs/season). Interestingly, the TC risk imposed during IPO negative extends to island nations located east of the dateline-e.g. American Samoa, Cook Islands, Fiji, French Polynesia, Niue, Samoa, Tokelau, Tonga and Wallis and Futuna. Of all three phases, IPO neutral represents the lowest risk of TCs impacting individual EEZs.

Other climate mode influences on TC tracks (IOD E and SAM)
The difference density plots of SWP TC tracks categorised as per IOD E phases are illustrated in Fig. 4 (panels a-c). A clear eastward propagation of TC tracks is observed when SSTs in the northeast Indian Ocean are warmer than average (IOD E positive), while TC tracks are more confined to the west of the study region when IOD E is negative (cool) (panels a-c). In particular, IOD E negative results in a clear elevation of TC tracks frequency around the northeastern coast of the Australia/Coral Sea region (panels a and c)increasing TC-related hazard (2.12 TCs/season) demonstrated by the density plot in panel g. As evident in the boxplots (panel e) and Table 2, TC tracks associated with IOD E positive are significantly lengthier (3563 km) when compared with IOD E negative (2797 km). These longer trajectories have shorter durations (mean duration of 9.38 days) compared to smaller tracks of TCs that occur during the IOD E negative phase (mean duration of  For TCs that occur during negative SAM phases, tracks are shown to migrate further east/northeast when compared to the positive phase, as shown in Fig. 5 (panels a and c). Additionally, the typical influence of SAM positive is evident in panel a, whereby most of the tracks are modulated poleward (i.e. longer tracks of an average of 3585 km; panel e), significantly in contrast to SAM negative (Table 2), which has shorter tracks (average of 2907 km; panel e). In particular, panels a and b demonstrate the influence of SAM positive on the extension of TCs south of ~ 40°S, while during SAM negative, this tends to occur north of ~ 40°S (panel c). Specifically, these longer tracks are characteristic of mid-latitude cyclones (or extratropical cyclones), also reported in Sharma et al. (2021), that tend to continue beyond the tropical latitudes, with fast-moving characteristics during SAM positive. This is important as the extension of TC tracks beyond tropical latitudes (transition into mid-latitudes) tends to impact New Zealand (0.66 TCs/season, panel g). During SAM positive, the extension of TC tracks is coupled with increased speed (mean speed of 15.80 km/ Fig. 5 As in Fig. 2, but for SAM. SAM positive (Pos), SAM negative (Neg) and SAM neutral (Neut) hr) and a mean lifetime of 9.89 days. Severe TC Gita (2018) is one such example, where SAM was in a neutral-to-weak positive phase (Australian Bureau of Meteorology 2018), that transitioned into an extratropical cyclone impacting New Zealand. Conversely, TC tracks influenced by SAM negative tend to be slower-moving (mean speed of 13.33 km/hr) but with an extended duration (mean of 10.88 days).
Almost all SWP island nations are susceptible to a higher number of TCs within their EEZ when the SAM is positive, except for New Zealand, which is a result of the poleward shift during negative SAM. Panel g demonstrates that New Zealand is more vulnerable during SAM negative (0.89 TCs/season) and SAM neutral (0.95 TCs/season).

Multimodal modulation of SWP TC tracks
The preceding section demonstrated that both Pacific and remote large-scale climate modes modulate TC track variability in terms of position, duration, speed and length. However, none of these modes operates in isolation and may act to enhance or suppress the impact of each other. As such, the following sections explore the coupled influence of IPO, IOD E and SAM with the canonical ENSO on TC tracks in the SWP.

Co-influence of ENSO and IPO on TC tracks
To evaluate the co-influence of ENSO and IPO phases on SWP TC track variation, each phase of ENSO (i.e. EN, LN and ENSO neutral) was paired with the positive, negative and neutral phase of IPO, as shown in Fig. 6 (panels a-c). The temporal characteristics, such as the duration (panels d-f), average speed (panels g-i) and track lengths (panel j-l) and their statistical difference (shown as p-values) in the last row, were also analysed. Figure 6 highlights a higher proportion of TCs in IPO negative for all ENSO phases in panels a-c. Panel b illustrates how TC tracks shift westward during LN/IPO negative, whereas the pairing of EN/IPO positive distinctly shows enhanced eastward modulation (east of 150°W) of TC tracks (panel a), consistent with observed relationships with TC genesis by Magee et al. (2017). The EN/IPO positive pairing is particularly important for nations towards the eastern SWP, including French Polynesia, the Pitcairn Islands and the Line Islands of Kiribati, where TCs are relatively infrequent compared to the western SWP Sharma et al. 2021). Indeed, the heatmap depicting the average number of TCs per season for this multimodal pairing further enhances our understanding of the potential risk of impact to each SWP island nation, atolls and territories (Fig. 7). For instance, the above-mentioned island nations are shown to be impacted mostly during EN/ IPO positive than other phase pairings-e.g. French Polynesia (0.19 TCs/season). This is consistent and further complements the increased risk impact demonstrated individually by EN (Fig. 2) and IPO positive (Fig. 3). Figure 6 further demonstrates modulation in the TC track attributes, which either has been enhanced or suppressed when compared to the typical influence of ENSO only in Fig. 2. For example, the interaction between IPO positive and EN tracks exhibits an average track length of 3577 km and average speed of 16.26 km/hr (panels d and g)-a substantial increase of 13% and 14% when compared to EN TCs in Fig. 2. A negligible difference in duration was observed between IPO positive and EN tracks (suppressed by 3%) (panel j). Conversely, the combined influence of LN/IPO negative influenced the mean duration of 10.76 days (a 4% enhancement) while suppressing track length (3336 km) and average speed (14.90 km/hr) by 2 and 3%, respectively, when compared to the individual influence of LN on TC tracks in Fig. 2 (panels d-f). Furthermore, the dominant influence of IPO negative is exhibited through the westward modulation of EN tracks in panel a, with a reduction in length (18%) and average speed (13%), and the difference is statistically significant (Fig. 6).
The dominant influence of IPO negative over ENSO impacts on TC tracks is also illustrated through the heatmap (Fig. 7). It is shown that most SWP island nations, atolls and territories are impacted when IPO negative is coupled with all three ENSO phases, i.e. EN, LN or ENSO neutral. Importantly, eastern Australia's highest risk occurs during the LN/ IPO negative combination (1.34 TCs/season in Fig. 7, central column), given TC tracks are modulated further west (Fig. 6, panel b). Additionally, Fig. 7 potentially provides additional insight into the cause of the observed elevated risk over PNG and Solomon Islands (western SWP nations, Fig. 2, panel g) that appeared to be unusual based on the location of these islands.

Co-influence of ENSO and IOD E on TC tracks
Complimentary to their individual influences (in Figs. 2 and 4), the combination of positive (negative) phases of ENSO and IOD E also results in east-northeast (west-southwest) modulation of TC tracks (Fig. 8, panels a and b). EN is more likely to co-occur with IOD E positive (159 TCs) and LN tends to occur with IODE negative (108 TCs), consistent with Liu and Chan (2012) and Magee and Verdon-Kidd (2018). Furthermore, positive phases of IOD E contribute significantly to increased TC track lengths during EN phases, while the converse occurs for IOD E negative LN combinations. Compared to ENSO alone (Fig. 2), which generally results in LN associated tracks being slightly faster-moving than EN (15.32 vs 14.27 km/hr), the co-influence of IOD E positive phase results in TCs during EN being faster-moving (average speed of 15.53 km/hr; Fig. 8, top row) than during LN/ IOD E negative (average speed of 14.41 km/hr; Fig. 8, panel h). The heatmap generated for ENSO and IOD E combinations in Fig. 9 illustrates the risk of TCs impacting SWP island nations is substantially different for EN/IOD E positive and LN/IOD E negative. EN/IOD E positive tends to impact almost all SWP island nations in contrast to the other two IOD E phases with EN (left column). The LN/IOD E pairings (central column) is shown to mainly impact eastern Australia (0.84 TCs/season), Fiji (0.43 TCs/season), New Caledonia (0.61 TCs/season), New Zealand (0.36 TCs/season), Vanuatu (0.35 TCs/season) and Solomon Islands (0.30 TCs/season)-whereby majority of these nations are on the western part of SWP (or west of the dateline).
Interestingly, under ENSO neutral/IOD E negative conditions, there is an observed increase in TC occurrence (163 TCs) in panel c (Fig. 8), indicating that IOD E negative can influence SWP TCs independent of ENSO influence (since ENSO is neutral in these cases). La Niña Neutral Fig. 9 As in Fig. 7 but for ENSO and IOD E phases from 1948-2021 A similar finding was reported by Magee and Verdon-Kidd (2018), where it was shown that regardless of ENSO (i.e. during ENSO neutral conditions), Indian Ocean SST variability also spatially modulates SWP TC genesis. Nevertheless, compared to ENSO neutral, TC tracks during ENSO neutral/IOD E negative display a reduction in length (2480 km compared to 3392 km) with an extended lifetime (11.28 days compared to 10.32 days), indicative of a slow-moving nature (average speed of 11.82 km/hr compared to 15.32 km/ hr) (Fig. 8, panels f, i and j) when compared with the track attributes during ENSO neutral in Fig. 2. TCs of this nature tend to be impactful for island nations similar to those mentioned above for LN/IOD E, but with an enhanced number of TCs per season (Fig. 9, right column)-e.g. Australia (1.11 TCs/season), New Caledonia (0.97 TCs/season), Fiji (0.62 TCs/season), Vanuatu (0.61 TCs/season), Solomon Islands (0.57 TCs/season) and New Zealand (0.47 TCs/season).

Co-influence of ENSO and SAM on TC tracks
The SAM modulates the impact of ENSO on TC tracks by suppressing or magnifying the typical ENSO track characteristics (Fig. 10). Spatially, the TC tracks during SAM negative (in all three ENSO phases) are extended further east than SAM positive tracks that are more confined westward (panels a-c). Temporally, the negative and neutral phases of SAM result in enhanced TC occurrences during EN (112 and 101 TCs), LN (91 and 88 TCs) and ENSO neutral (136 and 141 TCs) as compared to SAM positive (which suppresses ENSObased TC frequency). Indeed, enhanced TC frequency indicates greater risk impacts for SWP island nations, atolls and territories. Figure 11 demonstrates (through larger TCs/season values) that many island nations tend to get impacted mostly when SAM negative is associated with all three phases of ENSO. Most of these include eastern Australia, Fiji, New Caledonia, Solomon Islands, Vanuatu, New Zealand and Tonga, with average TCs ranging from 0.27 to 0.78 per season. The p-values in the last row (of Fig. 10) demonstrate significant temporal modulations mostly during LN and SAM positive phase pairings-with TCs being lengthier (4024 km) yet faster-moving (17.32 km/hr). Indeed, this is somewhat similar to the individual influence of ENSO and SAM portrayed in Figs. 2 and 5, respectively, where LN TCs were found to be fast-moving, and tracks were also lengthier during SAM positive. Additionally, TCs during LN/SAM positive are shown to typically impact island nations central-to-western SWP-eastern Australia (0.41 TCs/season), New Caledonia (0.32 TCs/season) and Fiji (0.35 TCs/season) and, to some extent, New Zealand (0.28 TCs/season) and Vanuatu (0.24 TCs/season), as shown in Fig. 11 (central column).

Discussion
This paper aimed to evaluate the spatio-temporal influence of climate drivers on SWP TC track activity. To achieve this, Pacific (ENSO and IPO) and remote (IOD E and SAM) drivers of climate variability were investigated in terms of their spatio-temporal influence on TC tracks and related attributes, both individually and in co-occurrence. Additionally, the hazard presented by the number of TCs intercepting the EEZ boundary for each island nation was evaluated in response to the modulation of TCs by the ocean-atmospheric interactions, both individually and coupled.

Variability in TC tracks by individual climate drivers
Both Pacific climate drivers (ENSO and IPO) exhibited eastern and western modulation of SWP TC tracks during positive and negative phases, respectively-consistent with previous studies (e.g. Grant andWalsh 2001, Chand andWalsh 2009;Camargo et al. 2010;Chand et al. 2013;Magee et al. 2017). Building upon previous research that were mostly on TC genesis, our findings illuminate additional impacts on track characteristics. It is revealed that TC tracks associated with EN (LN) are slow-moving (fast-moving) with reduced (extended) TC lifetime (Fig. 2, panels d and f)-with the latter attribute (i.e. shorter lifetime) being consistent with Ramsay et al. (2012). EN influenced TCs being slow-moving is in contrast with Sinclair's (2002) findings, which generally suggested otherwise; however, this could be due to the longer dataset used in this study and differences in the approach for calculating TC speed between the studies. Similar to Magee et al. (2017), who reported clustering of TC genesis points around central SWP during IPO positive, our findings also demonstrate comparable variability for TC tracks, i.e. increased TC activity around this region (Fig. 3, panel a). Conversely, the large concentration of IPO negative tracks westward (panel c) aligns with the reported findings of Grant and Walsh (2001). Temporal variability showed that TC tracks during the positive (negative) phase are significantly enhanced (suppressed) in terms of length and average speed yet have a shorter (extended) lifetime (Fig. 3, panels d-f). Interestingly, these temporal variations (track length, speed and duration) are also evident for TCs influenced by IOD E positive and negative phases (Fig. 4, panels d-f). Further, the temporal characteristics of the TC tracks are shown to be opposite with respect to EN and LN conditions, possibly due to differences in the TC seasonal cycle of the Indian Ocean and SWP region (Saji et al. 1999;Magee and Verdon-Kidd 2018).
Unlike the typical eastward and westward modulation of TC tracks observed for Pacific and Indian Ocean modes, positive and negative SAM produced northward and southward shifts in SWP TC activity, respectively (Fig. 5, panels a-c). These variations are attributable to the dynamics of SAM, i.e. lower pressure (over Antarctica) with strengthened westerly winds (around Antarctica) and higher pressure in the mid-latitudes, causing alteration of storm tracks more towards the South Pole during positive SAM (opposite for negative SAM) (Renwick and Thompson 2006;Ho et al. 2012;Diamond and Renwick 2015). It was also revealed that SAM positive/negative conditions resulted in westward/eastward modulation of TC tracks. On a broader scale, variations due to SAM are associated with extratropical cyclones or mid-latitude cyclones. Spatial variation in TC tracks during SAM positive indicates that most southern island nations of SWP are at elevated risk, e.g. New Zealand. However, our findings demonstrate that the risk impact for New Zealand (along with other island nations) is amplified during negative SAM.

Variability in TC tracks by combined climate drivers
Evaluation of climate drivers (i.e. IOD E, IPO and SAM) in combination with the canonical ENSO phases highlighted notable variations in the typical spatio-temporal variability of SWP TCs. Compared with the established eastern and western modulation by the EN and LN phases (Chand and Walsh 2009;Camargo et al. 2010;Chand et al. 2013), it was shown that with some climate mode/index pairings, these modulations could be enhanced or reversed. For instance, EN/IPO negative influenced TC tracks resulted in westward modulation, demonstrating IPO negative's dominant influence over TC tracks during EN. Magee et al. (2017) reported a similar regime for IPO negative over EN and attributed this variation to the dynamic conditions of this pairing (i.e. conducive relative vorticity and vertical wind shear). In this paper, we also find that these phase pairings substantially suppressed track length and average speed, while a negligible difference was observed for TC duration (enhanced by 1%). The leading influence of IPO negative is also exhibited when paired with LN, where TC duration is enhanced while other temporal attributes are suppressed. A possible reason for such variation could be attributed to the characteristics of western SWP TC tracks, i.e. generally intricate, as reported by Sharma et al. (2021).
The results also demonstrated that both positive and negative phases of IOD E produce notable variations on ENSO influenced TC tracks. Magee and Verdon-Kidd (2018) reported that couplings such as EN/IOD E positive and LN/IOD E negative result in significant spatial modulation of TC genesis. Our findings show that these combinations also influence TC tracks, i.e. EN/IOD E positive (LN/IOD E negative) enhances TC track length and average speed (TC track length and duration) by 13% and 10% (4% and 11%), respectively. The modulation observed during these pairings, i.e. EN/ IOD E positive and LN/IOD E negative, are mainly evident on TC tracks within the eastern and western SWP region, respectively, complementing the findings of Sharma et al. (2021). Furthermore, it was shown that ENSO neutral/IOD E negative pairing significantly resulted in the suppression (enhancement) of track length and average speed (duration) by 21% and 16% (6%), respectively. Importantly, this pairing highlights that modulations in SWP TC tracks can occur regardless of ENSO (i.e. during ENSO neutral conditions) through variations in the Indian Ocean SSTs.
A particularly interesting pairing was ENSO and SAM, which provided greater insight into these variations than that provided when these modes were assessed individually. In particular, the lengthening/shortening of TC tracks during LN/EN conditions was shown to be significantly influenced by the positive/negative phases of SAM, respectively, across distinct regions of the SWP. For example, the interaction between EN/SAM negative (both modulating TC activity eastward) resulted in shorter tracks over the eastern part of the SWP, while usually, the tracks are lengthier within this region based on the findings of Sharma et al. (2021). The co-influence of these two modes of climate variabilities (ENSO and SAM) on TC tracks is vital to the broader SWP region and the island nations within the extratropical region (beyond 25°S latitude)-e.g. New Zealand and the southeast coast of Australia. These findings may be used to quantify TC-related hazards better, considering both ENSO and SAM in future TC-related climate attribution studies. For instance, Adam et al. (2021) has shown potential impacts of ENSO and SAM in influencing TCs (later as extratropical cyclones) that results in storm surges. With this information, preparedness and mitigation plans can be enhanced. Furthermore, the westward variation with longer tracks during ENSO neutral and LN conditions (Fig. 2) is generally co-influenced by SAM positive. Complimentary to the findings of Sharma et al. (2021), the effects on the nature of TC tracks (e.g. fast-moving, more prolonged and transitioning into extratropical cyclones) within the western SWP are potentially a result of this co-influential relationship. Indeed, recent events such as TC Gabrielle (2023) that had devastating impact over New Zealand could be attributed to this co-influential relationship, given this event occurred when the state of ENSO was LN (although weakening) and SAM was in positive phase (Australian Bureau of Meteorology 2023; Harrington et al. 2023).

TC risk impact induced by Pacific and remote climate drivers
TC risk impact analyses have provided insights into how this risk varies for SWP island nations, atolls and territories during various phases of Pacific and remote climate drivers. In many cases, the results demonstrated that TC-induced impacts for island nations are amplified because of each climate phase's spatio-temporal modulation. For example, eastern (western) SWP island nations are more at risk during positive (negative) phases of ENSO, IPO and IOD E. However, in some cases, central SWP island nations are equally impacted during both positive and negative phases-e.g. Fiji during EN and LN. Some interesting cases were revealed where island nations were found to be at risk regardless of their geographical location or the typical spatial modulation exhibited by climate drivers. These occurrences were countered through risk impact assessment of the multimodal analysis. For instance, the impact of EN influenced TCs on western island nations was influenced by the negative phase of IPO. This again highlights the importance of considering multimodal analysis influencing TC track variability. It was also found that irrespective of each climate driver's positive and negative phases, island nations are also vulnerable during the neutral phase.

Conclusions
While ENSO is known as the dominant mode of variability within the SWP region, we show that it is not the only factor influencing TC track behaviour. Indeed, IPO, IOD E and SAM also play important roles in the overall prevailing track trajectories. Further, the findings of this study demonstrated that these modes of climate variability do not operate in isolation due to the dynamic nature of the ocean-atmosphere system and hence, modulate the impact of individual drivers. Indeed, the multimodal approach revealed important and distinct trends that were not reflected when climate modes were evaluated individually. It follows that when two climate modes/indices were in phase (e.g. EN with the positive phase of IPO, IOD E and SAM), TC track length and average speed were enhanced. For cases where either one (e.g. EN/negative phase of IPO and IOD E) or both (LN/negative phase of IPO, IOD E and SAM) climate modes/indices were in the negative phase, an increase in TC track duration was observed, implying a dominant impact of negative phases. We also show how variability in TC tracks induced by Indo-Pacific climate drivers imposes a risk impact on SWP island nations. The outcomes of this study, along with statistical or numerical modelling, could be useful in refining forecasts of TC paths once they form and may assist in improved quantification of TC-related risks.