Trends and Periodicities of Tropical Cyclone Frequencies and the Correlations with Ocean Drivers

Abstract

This study presents a comprehensive analysis on the variations in the tropical cyclone (TC) frequencies during 1980–2021, including the linear trends, periodicities, and their variabilities on both global and basin-wise scales. An increasing trend in the annual number of global TCs is identified, with a significant rising trend in the numbers of tropical storms (maximum sustained wind $35 \mathrm{k}\mathrm{t}\mathrm{s}\le U_{max}<64 \mathrm{k}\mathrm{t}\mathrm{s}$) and intense typhoons ($U_{max}\ge 96 \mathrm{k}\mathrm{t}\mathrm{s}$) and a deceasing trend for weak typhoons ($64 \mathrm{k}\mathrm{t}\mathrm{s}\le U_{max}<96 \mathrm{k}\mathrm{t}\mathrm{s}$). There is no statistically significant trend shown in the global Accumulated Cyclone Energy (ACE). On a regional scale, the Western North Pacific (WNP) and Eastern North Pacific (ENP) are the regions of the first- and second-largest numbers of TCs, respectively, while the increased TC activity in the North Atlantic (NA) contributes the most to the global increase in TCs. It is revealed in the wavelet transformation for periodicity analysis that the variations in the annual number of TCs with different intensities mostly show an inter-annual period of 3–7 years and an inter-decadal one of 10–13 years. The inter-annual and inter-decadal periods are consistent with those in the ENSO-related ocean drivers (via the Niño 3.4 index), Southern Oscillation Index (SOI), and Inter-decadal Pacific Oscillation (IPO) index. The inter-decadal variation in 10–13 years is also observed in the North Atlantic Oscillation (NAO) index. The Tropical North Atlantic (TNA) index and Atlantic Multi-decadal Oscillation (AMO) index, on the other hand, present the same inter-annual period of 7–10 years as that in the frequencies of all the named TCs in the NA. Further, the correlations between TC frequencies and ocean drivers are also quantified using the Pearson correlation coefficient. These findings contribute to an enhanced understanding of TC activity, thereby facilitating efforts to predict particular TC activity and mitigate the inflicted damage.

1. Introduction

Tropical cyclones (TCs) pose great risks to coastal as well as inland communities all around the world [1,2,3,4]. Fully understanding the features of TC activity is crucial for effective risk assessment and mitigation [5,6].

The trend in the frequency of TC genesis is a hot topic in the related fields. It was reported that, globally, the annual number of all the named TCs from 1990 to 2021 varied little [7], but the frequency of TCs in Categories 4–5 increased considerably during 1980–2006 [8,9]. Specifically, on the ocean-basin scale, the annual numbers of all the named TCs occurring within the South Indian Ocean (SIO), South Pacific (SP), Western North Pacific (WNP), and Eastern North Pacific (ENP) showed a declining trend since 1980 [10]. On the contrary, those generated in the Central Pacific (CP) and North Atlantic (NA) increased. Moreover, regarding TCs with high intensity, it was reported that the frequencies of intense TCs (above Category 3 on the Saffir–Simpson scale) in both the WNP and SP presented an increasing trend during recent years [11]. Various composite metrics have been developed to quantify TC activity considering not only the numbers but also the intensities or energies of the cyclones [9]. The Accumulated Cyclone Energy (ACE) [12] is one of the composite metrics and is usually adopted. It was shown in the results of [13] that the global ACE varied little between 1970 and 2011. However, [7] reported a remarkable downward trend in the global ACE from 1990 to 2021, primarily due to those cyclones generated in the WNP and SIO. This clearly shows that the trends in TC activity have not been fully understood regarding the uncertainties in the data and analysis methods as well as the effects of ongoing climate change.

Correlations between TC activity and ocean drivers have been investigated to examine the reasons for the changes in TC activity [6,14,15,16,17]. For example, the El Niño–Southern Oscillation (ENSO) has been shown to be closely related to the TC activity levels in various ocean basins [18,19,20]. During the warm phase of the ENSO, the number of TCs generated in the NA was below average [21,22], while that in the WNP might be above average [23]. The Inter-decadal Pacific Oscillation (IPO) phenomenon, characterized by the difference in the sea surface temperature (SST) of the Tropical and Northern Pacific oceans, also played an important role in affecting the TC activity [24,25]. It was found that the positive phase of the IPO, in which the SST in the Tropical Pacific is higher than that in the Northern Pacific, led to more TCs forming within both the ENP and WNP but less in the NA [26,27,28]. Another mode of oceanic driver is the Indian Ocean Dipole (IOD), which describes the temperature oscillation in the Western and Eastern equatorial Indian Ocean [29,30]. The positive phase of the IOD, in which the SST in the Western Indian Ocean is higher than that in the Eastern Indian Ocean, was correlated with a reduction in TC frequency in the Northern Indian Ocean (NIO) due to the weakened convection and generated anticyclonic circulation [31,32,33]. Moreover, the Tropical North Atlantic SST anomaly (TNA) has also attracted a great deal of attention as there seemed to be a significant negative correlation between the TC frequencies in the WNP and TNA [34,35,36,37,38]. A strong correlation between the landing number of intense TCs in mainland China during autumn and the TNA SST anomaly on an inter-annual time-scale was reported [39]. Furthermore, the North Atlantic Oscillation (NAO), which controlled the multi-decadal changes in the TNA–ENSO connection [40], influenced the TC frequencies in the WNP [41] and NA [42]. In the NA, one more mode of ocean driver is the Atlantic Multi-decadal Oscillation (AMO) [43]. In the positive period of the AMO, when the SST in the NA is warmer than its long-term average, more storms formed in the Caribbean Sea [44]. However, the effects of ocean drivers on changes in TC genesis frequency are far from clearly and fully understood. The related physical processes are too complex, and, more importantly, the processes in different ocean basins may be interrelated. The previous studies on the correlations between the TC frequency changes and ocean drivers in specific ocean basins were not sufficient, and a correlation analysis from a global perspective is necessary.

The present study investigates the characteristics of the TC frequency from both global and basin perspectives. The trends in the annual numbers of cyclones with different intensities across the whole globe, the two hemispheres, and the six ocean basins are obtained using linear regression. The characteristic periods of variation in the TC frequencies are identified by wavelet transformation, as presented in Section 3.1. The correlations between the TC frequency changes and ocean drivers are examined regarding their characteristic periods and Pearson correlation coefficients.

2. Materials and Methods

2.1. Data

The track and intensity of all named TCs since 1980 are obtained from the International Best Track Archive for Climate Stewardship (IBTrACS), version 4 [45]. Specifically, the data from USA agencies are adopted. Only TCs with their 1 min-averaged maximum sustained wind speed $U_{max}$ exceeding 35 kts (i.e., 65 km/h) are involved in this study. These TCs are classified into tropical storm (TS), weak typhoon (WTY) and intense typhoon (ITY) according to $U_{max}$, as shown in

Table 1. Classifications of tropical cyclones.

$\mathbf{Range} \mathbf{of} {\mathit{U}}_{\mathit{m}\mathit{a}\mathit{x}}$	Category	Classification
35 kts $\le U_{max}<$ 64 kts	TS	Tropical Storm (TS)
64 kts $\le U_{max}<$ 96 kts	1, 2	Weak Typhoon (WTY)
$U_{max}\ge$ 96 kts	3, 4, 5	Intense Typhoon (ITY)

.

In this paper, the study period is the calendar years from 1980 to 2021, considering that reliable TC observations by satellite are available only after 1980 [46]. For spatial distribution of TCs, the whole globe is divided into six basins as shown in Figure 1.

The ACE is taken as a composite metric to quantify TC activity. The value of ACE for a TC equals to the summation of $U_{max}^2$ at 6 h intervals during the whole TC lifetime [12], shown as

$$\mathrm{A}\mathrm{C}\mathrm{E}=\sum U_{max}^2$$

(1)

It is noted that the lifetime of a TC refers to the time length during which $U_{max}$ is larger or equal to 35 kts. The accumulated ACE of a specific ocean basin in a period is obtained by summing up the ACE of all TCs existing in the basin during this period.

ENSO is quantified by Niño 3.4 index [47] and Southern Oscillation Index (SOI) [48]. Niño 3.4 index is calculated regarding the average SST anomaly in the region of 5° S–5° N and 170° W–120° W, while SOI is defined as the normalized sea-level pressure difference at Tahiti and Darwin. IPO index is the difference between the SST anomalies averaged over the Central equatorial Pacific and the Northwest and Southwest Pacific [49]. The strength of IOD is quantified by anomalous SST gradient between two regions, i.e., 10° S–10° N, 50° E–70° E and 10° S–0° S, 90° E–110° E. [28]. Moreover, TNA index is obtained based on the averaged SST anomaly in the tropical NA (5° N–30° N, 80° W–20° W) following [38]. NAO index is defined as the normalized pressure difference between Gibraltar and SW Iceland (Reykjavik) and calculated following [50]. AMO index is the area-weighted average SST over NA (0° N–70° N) and obtained from the Kaplan SST dataset [51].

2.2. Wavelet Transformation

The periodic feature of a time series of any TC metric can be effectively captured via wavelet transformation [17,52,53]. By decomposing a time series into a time-frequency domain, one is able to determine both the dominant modes of variability and how those modes vary in time [54]. The wavelet coefficient $WT$ is generated by the transform function,

$$WT\left(a,b\right)={\displaystyle \frac{1}{\sqrt{a}}}{\int }_{-\infty }^{+\infty }f(t)\psi ({\displaystyle \frac{t-b}{a}})dt,$$

(2)

where $a$ is the scale and $b$ is the translation parameters. $f\left(t\right)$ is the time series and $\psi$ represents a basic wavelet function. The complex Morlet function is selected as the basic wavelet function $\psi$ in this study, which has been frequently adopted in previous meteorological research [55,56] and is expressed as

$$\psi \left(t\right)={\pi }^{-{\displaystyle \frac{1}{4}}}\exp (-{\displaystyle \frac{t^2}{2}})\exp \left(i{\omega }_{0}t\right),$$

(3)

where ${\omega }_0$ is the nondimensional frequency and set to be 6.2 in the present study following [57]. Subsequently, the scale set utilized in wavelet transform is determined as fractional power of 2.

To identify characteristic periodicities of the time series of TC metrics more straightforwardly, the wavelet power is generated based on wavelet coefficient scalogram. Within a specific scale $a$, the wavelet power $E\left(a\right)$ is calculated as

$$E\left(a\right)={\displaystyle \frac{1}{N}}{\sum }_{b}W{T}^2(a,b),$$

(4)

where $N$ is the length of the time series. In the wavelet power spectrum, a characteristic period of variation in $f\left(t\right)$ is identified as the period with a local maximum value of E. The periods shorter than 10 years, between 10–20 years, and larger than 20 years are named as inter-annual, inter-decadal, and multi-decadal periods, respectively. The time-series data $f\left(t\right)$ may have multiple characteristic periods, and we name the one with the maximum E the dominant period. In this study, only inter-annual and inter-decadal periods are obtained due to the limited time length of data.

2.3. Statistical Analysis

In this study, the linear least-squares regression is applied to obtain the trend of data series at a confidence level of 95%, and the F-test method is adopted to evaluate the statistical significance of obtained trends. A confidence level of 90% is also utilized in some cases. Unless stated, the statistically significant trends mentioned in the text are at the confidence level of 95%.

3. Results

3.1. Temporal Changes in TC Frequencies

3.1.1. Linear Trend of Global TC Frequencies

Figure 2 presents the annual variations and linear trends in the numbers of all the named TCs ($U_{max}\ge 35 \mathrm{k}\mathrm{t}\mathrm{s}$) and specifically the TSs ($35\le U_{max}<64 \mathrm{k}\mathrm{t}\mathrm{s}$), WTYs ($64\le U_{max}<96 \mathrm{k}\mathrm{t}\mathrm{s}$), and ITYs ($U_{max}\ge 96 \mathrm{k}\mathrm{t}\mathrm{s}$). The variations in the proportion of each TC classification are also shown. The 42-year-averaged values and the linear trends in the time series during 1980–2021 are listed in Table S1. Globally, there were about 86 TCs generated every year, and the TC annual number increased by 0.19 per year at a statistical significance level of 10%. Among these TCs, annually, there were about 39 TSs generated, 45.32% of all the named TCs, with a statistically significant (confidence level of 95%) increasing trend of 0.22 per year. The proportions of the WTYs and ITYs were close, i.e., 26.96% and 27.72%, respectively. However, the number of WTYs showed a statistically significant decrease at a rate of −0.18/year, while that of the ITYs increased at a mean rate of 0.15/year. The conclusion of more ITYs generated is consistent with those in previous studies [7,8,9,58] but in a more quantitative way. Moreover, the results show that there was a shift in the TS–WTY–ITY distribution pattern of all the named TCs, i.e., more TSs and ITYs but fewer WTYs. The intensification of tropical depressions might lead to an increase in TSs and the intensification of WTYs into ITYs, raising the number of the latter [7]. However, it seems that the intensification of the TSs was far from enough to keep the number of WTYs stable. The reason is still not clear in the present investigation.

In Table S1 and Figure 3, the averaged annual ACE across the whole globe was 777.14 × 10⁴ kt², and its yearly variation presented no significant change, implying the stability in the kinetic energy of the global TCs during 1980–2021. It is noted that the presently obtained trend of the ACE during 1980–2021 was different from the significantly decreased conclusion during 1990–2021 estimated by [7]. The difference was due to the change in the ACE during 1980–1990.

3.1.2. Linear Trend of TC Frequencies in Different Ocean Basins

Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8 show the linear trends of the annual numbers of all the named TCs, TSs, WTYs, and ITYs as well as ACE in the two hemispheres and the six ocean basins. The specific values are listed in Table S2. The 42-year-averaged proportions of the frequencies in the hemispheres and ocean basins to the corresponding global values during 1980–2021 and their linear trends are also illustrated in the figures. The specific values are listed in Table S3. For all the named TCs and the three classifications, the orders of their proportions in the six ocean basins were the same, as shown in Figure 4b, Figure 5b, Figure 6b and Figure 7b. The WNP and ENP are the regions with the first- and the second-largest proportions of the TCs, TSs, WTYs, and ITYs, respectively. The SIO is the third. However, the statistically significant (confidence level of 95%) growing trend in the annual number of all the named TCs in the Northern Hemisphere (NH) as well as that in the global TCs was mainly due to TCs forming in the NA. The annual number of TCs in the NA increased at a mean rate of 0.26 per year, not high but statistically significant. The increasing rates of the TS and ITY frequencies in the NA were statistically significant and the highest among the six basins. Moreover, the proportions of TSs, WTYs, and ITYs forming in the NA all increased statistically significantly and at the highest rate among the basins. The increase in the TC frequencies appeared mainly in the Gulf of Mexico, the Caribbean Sea, and a banded zone in the tropical NA, as shown in Figure S1. On the other hand, the changes in the TC numbers of both the WNP and ENP were insignificant. Moreover, the numbers of all the named TCs in the Southern Hemisphere (SH) and the two specific basins, the SIO and SP, were insignificantly decreasing at very low rates.

There was a shift in the TS–WTY–ITY distribution pattern from WTYs towards TSs and ITYs over the 42 years not only across the globe but also within the NH. This was the result of TC frequency changes in various ocean basins. WTYs had a statistically decreasing trend with a rate of −0.10/year in the WNP while showing insignificant variations in the other ocean basins of the NH. The annual number of TSs in all the ocean basins except the NA varied insignificantly but showed a significant upward trend in the NA by 0.18 per year, as illustrated in Figure 5a. For ITYs, the annual number rose statistically significantly in the NA and NIO, specifically at rates of 0.06/year and 0.04/year, respectively, and varied little in the other basins. In the SH, only WTYs had a slight decreasing trend. The decreasing rates of the WTY numbers in the SIO and SP were, respectively, −0.03/year (significance level of 10%) and −0.05/year (significance level of 5%). The TSs and ITYs varied little in the SH and the two southern ocean basins.

In Figure 8a, the ACE demonstrated a statistically significant increase only in the NA and NIO, respectively, by 1.91 × 10⁴/year and 0.54 × 10⁴/year, while it varied insignificantly in the other ocean basins as well as on the hemispherical and global scales.

3.1.3. Periodic Variations in TC Metrics

Figure 9 shows the scalograms of the wavelet coefficient magnitude and the wavelet power spectrums of the annual TC numbers across the globe, the two hemispheres, and the six ocean basins. The corresponding results of the TSs, WTYs, ITYs, and ACE are displayed in Figures S2–S5, respectively. It is shown in Figure 9 that the variation in the annual TCs across the whole globe presented four characteristic periods in the wavelet power spectrum, i.e., 2–3, $~4$, 6–7, and 11–13 years. Consistent with the proportions of TCs forming in the ocean basins shown in Figure 4, these four characteristic periods in the global TC frequency also existed regarding the TC frequencies of the WNP, ENP, and SIO. Specifically, the characteristic periods in the variations in the TC frequency in the WNP were $~4$ and 11–13 years. Those in the ENP were 2–3 and 7–9 years, while, in the SIO, they were 2–3 and 5–7 years. This indicates that the variation with a period of 2–3 years in the global TC frequency was mainly due to the TC frequency changes in both the ENP and SIO. Moreover, the 4-year-period variation in the global TC frequency from the mid-1990s to mid-2010s primarily resulted from the changes in the TC frequency in the WNP. The inter-decadal (11–13 years) variation in the global TC frequency was also principally a result of the change in the WNP, but the effect of the inter-decadal change in the NA after the mid-2010s was notable. It is consistent with the finding in the above subsection that the TCs in the NA contributed most to the increase in the global TCs.

In the NIO, the characteristic periods in the variation in the TC frequency, i.e., 2–3 and 4–6 years, were close to those in the SIO. In the SP, the dominant period was 4–5 years. It is noted that the inter-decadal (11–13 years) variations in the TC frequency were not significant in the two major basins of the SH, while, in terms of the hemisphere, the dominant period of the TC frequency variation in the SH was about 11–13 years.

In Figure 9 and Figures S2–S5, most of the frequencies of the TCs, TSs, WTYs, and ITYs, as well as the ACE across the whole globe, the two hemispheres, and the six ocean basins, had a characteristic variation period in the range of 3–7 years, especially in the NH. For example, in the WNP, the characteristic periods of the variations in the frequencies of TS and WTY were 4–5 and 3–5 years, respectively. Even though the dominant period of ITY variation was 10–13 years, there was a significant characteristic period of 3–4 years after the mid-2010s, as shown in Figure S4. Moreover, in the NA, the characteristic periods of variation were 7–10 and 3–7 years in terms of the TS frequency (Figure S2), 2–3 and 4–6 years in terms of the WTY frequency (Figure S3), and 3–7 years in terms of the ITY frequency (Figure S4). These characteristic periods of 3–7 years showed a possible correlation between the TC genesis in the NH and ENSO.

For the ACE, the dominant periods of the variations in the NH and SH were quite different. In Figure S5, the dominant periods of ACE variation in the NH basins other than the NIO were all larger than 7 years, while those in the SH basins (i.e., the SIO and SP) were about 4–7 years. In the NIO, the dominant period of ACE variation was also 4–7 years.

3.2. Correlations between TC Frequencies and Ocean Drivers

3.2.1. Regarding Temporal Evolution Characteristics

The annual variations in the Niño 3.4 index, SOI, IPO index, IOD index, TNA index, NAO index, and AMO index as well as their linear trends are shown in Figure 10. Over the 42 years from 1980 to 2021, there was no statistically significant trend in the annual variations in the Niño 3.4, IOD, and NAO indices. However, the SOI, the other metric for the ENSO, had a statistically significant increasing trend (at a significance level of 90%) with a mean rate of 0.02/year. The mean value of the SOI seemed to change from negative to positive in favor of the La Niña phenomenon. In Figure 10c,e, the decreasing trends in the IPO and TNA indices were both statistically significant, with the same mean rate of −0.02 °C/year. In Figure 10g, the AMO index clearly presented a significant increasing trend at a mean rate of 0.1 °C/ year and a transition from the cool phase (negative values) to the warm phase (positive values) around 2000, consistent with the finding in [59]. Hence, according to the conclusion of [43] that the TC genesis in the Atlantic Ocean was heightened in the positive phase of the AMO, the annual number of TCs forming in the NA increased significantly, as shown in Figure 4a.

The characteristic periods of variation in the seven ocean drivers were identified through wavelet analysis, as shown in Figure 11. It is shown that the Niño 3.4 index, SOI, and IPO index all had two notable characteristic periods of 3–7 and 10–13 years, one inter-annual and the other inter-decadal. The inter-annual period of 3–7 years is a typical feature of the ENSO, which has been widely reported [16,17]. Comparatively, the inter-decadal period of 10–13 years was reported much less frequently. However, this inter-decadal variation period should not be ignored as it existed in almost the entire period from 1980 to 2021, as shown in Figure 11. The similarity in the characteristic periods of the Niño 3.4 index, SOI, and IPO index implies that the inter-decadal Pacific oscillation phenomenon is highly related to the ENSO. Moreover, the variation in the NAO index also presented a characteristic period of 10–13 years after 2000, when the transition of the AMO from the cool phase to the warm phase occurred. This indicates a possible connection between the NAO and ENSO on an inter-decadal scale, not only on the multi-decadal scale already reported in [40]. Another fresh point is displayed in the variation spectrums of the TNA and AMO indices. The dominant characteristic periods in the TNA and AMO variations were both 7–10 years, implying an inter-annual connection between the TNA and AMO.

First, the correlations between the TC frequencies and ocean drivers were investigated by comparing their characteristic periods of annual variation. As discussed in Section 3.1.3, most of the TC frequencies shown in Figure 9 and Figures S2–S5 had a characteristic variation period of 3–7 years. This period was consistent with the characteristic ones of the ENSO and IPO, as widely reported in the literature. Furthermore, the present results show another inter-decadal connection between the TC frequencies and the ENSO, which has not been widely reported. The inter-decadal characteristic periods (11–13 years) in the variations in the TC and ITY numbers in the WNP (Figure 9 and Figure S4, respectively), the TS and WTY numbers in the SP (Figures S2 and S3, respectively), and the WTY number in the ENP (Figure S3) were in good agreement with those of the Niño 3.4 index, SOI, IPO, as well as NAO indices. These show an inter-decadal correlation between the TC frequencies and the ENSO, IPO, and NAO, which have not been reported in the literature yet.

Attention should be paid to a fresh connection between the TC frequencies and the TNA as well as the NAO on an inter-annual scale of 2–3 years. In Figure 9 and Figures S2–S5, the frequency variation in the TCs with different intensities across the globe, hemispheres, and six ocean basins mostly presented a characteristic period of 2–3 years. However, in Figure 11, only in the results of the TNA and NAO indices, the corresponding characteristic period could be found. This indicates that the short-term variation in the TNA and NAO might lead to notable fluctuations in the TC frequencies across the globe. Further, in the NA, the frequencies of all the named TCs (Figure 9) and specifically TSs (Figure S2) might be related to the TNA and AMO on the inter-annual scale. They all presented an inter-annual variation with a characteristic period of 7–10 years.

3.2.2. Regarding Pearson Correlation Analysis

Furthermore, the correlations between the TC frequencies and ocean drivers were examined using the Pearson correlation technique. The Pearson correlation coefficients between the annual numbers of all the named TCs, TSs, WTYs, and ITYs as well as the values of ACE across the whole globe, the two hemispheres, and the six ocean basins and the seven ocean drivers are listed in

Table 2. Pearson correlation coefficients between frequencies of all named TCs in various regions and the ocean variabilities.

Region	Niño 3.4 Index	SOI	IPO Index	IOD Index	TNA Index	NAO Index	AMO Index
Globe	0.1043	−0.1014	−0.0233	0.0891	0.0691	0.3170 **	−0.0422
NH	0.0846	−0.067	−0.1154	0.0932	−0.1369	0.3404 **	0.1031
SH	0.0621	−0.084	0.1521	0.0006	0.3604 **	0.0400	−0.2524
WNP	0.1532	−0.2780 *	0.1803	0.1147	0.4732 **	0.3883 **	−0.4611 **
NIO	0.0973	−0.0183	0.0047	−0.1338	−0.171	−0.0638	0.1941
ENP	0.4323 **	−0.4487 **	0.3821 **	0.2002	0.1925	0.5642 **	−0.2879 *
NA	−0.3949 **	0.5016 **	−0.6220 **	−0.0857	−0.6546 **	−0.2747 *	0.6718 **
SIO	−0.1987	0.1252	−0.1569	−0.0111	0.3387 **	−0.0598	−0.1933
SP	0.2819 *	−0.2551	0.3721 **	0.0162	0.2018	0.1320	−0.1829

Note: The minus sign indicates the negative correlation between two variables; i.e., one variable increases versus the decrease in the other. The numbers in bold and with an asterisk in superscript are statistically significant with a significance level of 0.10 and those with two asterisks are at a significance level of 0.05. Other numbers are statistically insignificant. The same symbols are also used in following Table 3, Table 4, Table 5 and Table 6.

,

Table 3. Pearson correlation coefficients between the frequencies of TSs in various regions and the ocean variabilities.

Region	Niño 3.4 Index	SOI	IPO Index	IOD Index	TNA Index	NAO Index	AMO Index
Globe	−0.1559	0.2019	−0.3083 **	0.0491	−0.1528	0.0456	0.1681
NH	−0.1793	0.2302	−0.3772 **	0.1743	−0.2815 *	0.0448	0.2798 *
SH	0.0247	−0.0244	0.0836	−0.2297	0.2077	0.0119	−0.1744
WNP	−0.3889 **	0.3409 **	−0.4046 **	0.1233	0.1067	0.1229	−0.0282
NIO	0.1784	−0.1847	0.2229	−0.2172	0.0549	−0.0573	−0.0741
ENP	0.1778	−0.1753	0.0647	0.3067 *	−0.0351	0.1295	0.0016
NA	−0.2562	0.3901 **	−0.5308 **	0.0604	−0.6014 **	−0.1016	0.5728 **
SIO	−0.0021	−0.028	0.0656	−0.2678 *	0.1491	−0.0161	−0.1323
SP	0.0743	−0.0553	0.0984	−0.0728	0.2133	0.0801	−0.1627

Note: The minus sign indicates the negative correlation between two variables; i.e., one variable increases versus the decrease in the other. The numbers in bold and with an asterisk in superscript are statistically significant with a significance level of 0.10 and those with two asterisks are at a significance level of 0.05. Other numbers are statistically insignificant.

,

Table 4. Pearson correlation coefficients between the frequencies of WTYs in various regions and the ocean variabilities.

Region	Niño 3.4 Index	SOI	IPO Index	IOD Index	TNA Index	NAO Index	AMO Index
Globe	−0.0973	−0.03	0.0634	−0.1671	0.3840 **	0.1807	−0.3491 **
NH	−0.1427	0.0083	−0.0421	−0.2723 *	0.2836 *	0.2191	−0.2719 *
SH	0.0385	−0.0801	0.2131	0.1052	0.3405 **	0.0091	−0.2848 *
WNP	0.0166	−0.1841	0.1852	−0.1108	0.5556 **	0.1390	−0.5700 **
NIO	−0.1079	0.1717	−0.1349	−0.1903	−0.0974	0.0274	0.1605
ENP	0.0262	−0.0842	0.0557	−0.1578	0.1565	0.4279 **	−0.2102
NA	−0.2834 *	0.2534	−0.3139 **	−0.1384	−0.2566	−0.1984	0.3218 **
SIO	−0.3026 *	0.2071	−0.1725	0.0977	0.2665 *	−0.1772	−0.2170
SP	0.2731 *	−0.2509	0.3931 **	0.0622	0.2251	0.1479	−0.1952

Note: The minus sign indicates the negative correlation between two variables; i.e., one variable increases versus the decrease in the other. The numbers in bold and with an asterisk in superscript are statistically significant with a significance level of 0.10 and those with two asterisks are at a significance level of 0.05. Other numbers are statistically insignificant.

,

Table 5. Pearson correlation coefficients between the frequencies of ITYs in various regions and the ocean variabilities.

Region	Niño 3.4 Index	SOI	IPO Index	IOD Index	TNA Index	NAO Index	AMO index
Globe	0.4629 **	−0.3956 **	0.2909 *	0.2501	−0.0566	0.2978 *	0.0489
NH	0.5805 **	−0.4903 **	0.4017 **	0.1849	−0.1141	0.3611 **	0.0404
SH	0.0431	−0.0457	−0.0297	0.2132	0.0612	0.0468	0.0368
WNP	0.6153 **	−0.5969 **	0.5124 **	0.1544	0.0832	0.3337 **	−0.1315
NIO	−0.0275	0.1127	−0.2481	0.3639 **	−0.3048 **	−0.0427	0.3120 **
ENP	0.5013 **	−0.4888 **	0.5136 **	0.1266	0.2352	0.4744 **	−0.3167 **
NA	−0.3886 **	0.4897 **	−0.5253 **	−0.2101	−0.5538 **	−0.4031 **	0.5861 **
SIO	−0.0781	0.0702	−0.1709	0.1952	0.1266	0.0432	0.0150
SP	0.1878	−0.1801	0.2102	0.0484	−0.0924	0.0104	0.0368

Note: The minus sign indicates the negative correlation between two variables; i.e., one variable increases versus the decrease in the other. The numbers in bold and with an asterisk in superscript are statistically significant with a significance level of 0.10 and those with two asterisks are at a significance level of 0.05. Other numbers are statistically insignificant.

and

Table 6. Pearson correlation coefficients between the annual values of ACE in various regions and the ocean variabilities.

Region	Niño 3.4 Index	SOI	IPO Index	IOD Index	TNA Index	NAO Index	AMO Index
Globe	0.4777 **	−0.4590 **	0.4140 **	0.082	0.2056	0.3841 **	−0.1544
NH	0.5632 **	−0.5253 **	0.4594 **	0.1151	0.1208	0.4447 **	−0.1377
SH	0.1093	−0.1382	0.1546	−0.0217	0.3016 *	0.1047	−0.1280
WNP	0.6236 **	−0.6220 **	0.5914 **	0.111	0.3396 **	0.3527 **	−0.3713 **
NIO	0.0211	0.0536	−0.1828	0.226	−0.2704 *	0.0189	0.2511
ENP	0.5230 **	−0.5243 **	0.5225 **	0.1775	0.2922 *	0.5744 **	−0.3567 **
NA	−0.3886 **	0.4469 **	−0.5028 **	−0.1899	−0.5307 **	−0.2901 *	0.6228 **
SIO	−0.1519	0.0893	−0.1402	0.0533	0.3351 **	−0.0159	−0.1084
SP	0.3699 **	−0.3323 **	0.4253 **	−0.1031	0.0232	0.1863	−0.0571

Note: The minus sign indicates the negative correlation between two variables; i.e., one variable increases versus the decrease in the other. The numbers in bold and with an asterisk in superscript are statistically significant with a significance level of 0.10 and those with two asterisks are at a significance level of 0.05. Other numbers are statistically insignificant.

. For the frequency of all the named TCs, most of their correlation coefficients with the ocean drivers are in the range of $-0.4~0.4$, indicating a statistically insignificant or very weak correlation with the ocean drivers. In the NA, the annual number of all the TCs over the 42 years demonstrated a strong negative correlation with the IPO and TNA but a strong positive one with the AMO index, with an absolute coefficient larger than 0.6. The TC number in the NA also had an intermediate positive correlation with the SOI, with an absolute coefficient in the range of $0.4~0.6$. This is consistent with the correlation analysis based on the characteristic periods. The annual number of TCs in the ENP had an intermediate positive correlation with the Niño 3.4 and NAO indices but an intermediate negative one with the SOI, which implies that the El Niño and NAO favored the occurrence of tropical cyclones in the ENP. Specifically, the TC frequency in the ENP had a stronger correlation with the NAO than that with the ENSO. Moreover, the TC frequency in the WNP had an intermediate positive correlation with the TNA index but a nonequivalent negative correlation with the AMO. All these show a teleconnection effect of the NA ocean variabilities on the TC genesis in the Pacific.

For the annual number of TSs, most of the correlation coefficients shown in Table 3 are statistically insignificant or smaller than 0.4. Only in the NA, the correlation between the TS frequency and the TNA index was strong, with the absolute value of the coefficient larger than 0.6. The correlation between the TS frequency in the NA and the AMO index was also considerable. The correlation coefficient between the TS frequency in the NA and the IPO index was −0.5308, lower in absolute value than that between the TC frequency in the NA and the IPO index but still in the intermediate range. The positive correlation between the TS number in the NA and the SOI was also statistically significant at a confidence level of 95%.

In Table 4, the correlations of the WTY frequencies with the ocean drivers were extremely weak. Only the correlation coefficients of the WTY frequency in the WNP with the TNA and AMO indices were higher than 0.4. Similar to results of all the named TCs shown in Table 2, the WTY frequency in the WNP had an intermediate positive correlation with the TNA index but a nonequivalent negative correlation with the AMO. However, the correlations between the WTY frequency and the two ocean drivers were stronger than those of the TC frequency.

It is noted that the ITY frequencies regarding the whole globe, NH, WNP, ENP, and NA all had an intermediate or strong correlation with the ENSO-related and IPO indices. Specifically, in Table 5, the ITY frequencies regarding the globe, NH, WNP, and ENP were all positively correlated with the Niño 3.4 index and IPO index while being negatively correlated with the SOI. However, the ITY frequency in the NA had a totally converse result, being negatively correlated with the Niño 3.4 and IPO indices and positively correlated with the SOI. It also had an intermediate negative correlation with the NAO, a notable negative correlation with the TNA, and a strong positive correlation with the AMO, indicating the critical effect of the three ocean variabilities in the Atlantic Ocean on the genesis of the ITYs in the NA. Further, the correlations between the ITY numbers in the NIO and IOD and TNA indices were also statistically significant, albeit lower than 0.4.

Similar to the ITY frequency, the annual values of the ACE regarding the whole globe, NH, WNP, and ENP were all positively correlated with the Niño 3.4 index and IPO index while being negatively correlated with the SOI to an intermediate or greater extent, as indicated in Table 6. Also, the ACE value in the NA had a totally converse result, significantly negatively correlated with the Niño 3.4 and IPO indices while being positively correlated with the SOI. The positive correlation between the ACE and AMO index was strong. A difference is that the ACE value in the SP also had a statistically significant positive correlation with the Niño 3.4 and IPO indices as well as a negative correlation with the SOI.

4. Discussion

In the present study, a comprehensive investigation was conducted on the temporal trends and periodicities of the tropical cyclone frequencies on both the global and basin-wise scales. Even though this critical topic has been widely investigated in the literature [6,7], some findings in the present study are more quantitative in terms of details or have rarely been reported previously:

(1) There was a shift in the TS–WTY–ITY distribution pattern of all the named TCs during 1980–2021, i.e., more TSs and ITYs but fewer WTYs, across the globe and within the NH. The intensification of tropical depressions into TSs and WTYs into ITYs might explain the increase in the TSs and ITYs. However, the intensification of the TSs was not enough to stabilize the number of WTYs, and the reason needs to be investigated in the future.

(2) On the basin-wise scale, the shift in the TS–WTY–ITY distribution pattern was the result of TC frequency changes in various ocean basins. The increase in the TSs was statistically significant only in the NA, and the rise in the ITY frequencies appeared in the NA and NIO. The decrease in the WTY frequency occurred mainly in the WNP and secondarily in the SP.

(3) The present study enriched our understanding of the TC genesis in the NA. It was illustrated that the WNP, ENP, and SIO were the regions with the first-, second-, and third-largest proportions of TCs as well as TSs, WTYs, and ITYs, respectively. However, the NA contributed most to the significant increase in the TCs in the NH and across the globe. The increasing rates of TS and ITY frequencies in the NA were statistically significant and the highest among the six basins. Moreover, the proportions of TSs, WTYs, and ITYs forming in the NA all grew statistically notably and at the highest rate among the basins. Further, through wavelet analysis, the characteristic periods of the TC frequency in the NA were obtained, i.e., 7–10 and 3–7 years regarding the TS frequency, 2–3 and 4–6 years regarding the WTY frequency, and 3–7 years regarding the ITY frequency. The frequencies of all the named TCs and specifically the TSs in the NA had a correlation with the TNA and AMO on the inter-annual scale of 7–10 years. The Pearson correlation coefficients between the TC frequencies in the NA and the seven ocean variability drivers were calculated. Generally, as indicated in previous papers [40,41,42,43,44], the TC frequencies in the NA were strongly correlated with the TNA, NAO, and AMO, while a few of them examined the variation in the TC frequencies in the NA as quantitatively as in the present study.

(4) In the wavelet analysis, there was demonstrated to be an inter-decadal characteristic period of 10–13 years in the ENSO-related indices as well as the IPO index, other than the typical period of 3–7 years. The inter-decadal characteristic period also existed regarding the variation in the NAO, indicating a possible connection between the NAO and ENSO on an inter-decadal scale, not only on the multi-decadal scale already reported in the literature. Further, there was also an inter-decadal characteristic period of 11–13 years in the variation in TC frequencies within the ocean basins, demonstrating fresh inter-decadal correlations between the TC frequencies and the ENSO, IPO, and NAO that have not been reported yet in the previous work [16,17,24,40].

5. Conclusions

The temporal genesis characteristics of tropical cyclones (TCs) during 1980–2021 were investigated. It was shown that the annual number of all the named TCs across the whole globe demonstrated an increasing trend with a rate of 0.19/year over the 42 years. Specifically, the annual numbers of tropical storms (TSs), whose maximum sustained wind speed $U_{max}$ is in the range of $35 \mathrm{k}\mathrm{t}\mathrm{s}\le U_{max}<64 \mathrm{k}\mathrm{t}\mathrm{s}$, and intense typhoons (ITYs), with $U_{max}\ge 96 \mathrm{k}\mathrm{t}\mathrm{s}$, increased, while the amount of weak typhoons (WTYs), with $64 \mathrm{k}\mathrm{t}\mathrm{s}\le U_{max}<96 \mathrm{k}\mathrm{t}\mathrm{s},$ decreased. The Western North Pacific Ocean (WNP), Eastern North Pacific Ocean (ENP), and South Indian Ocean (SIO) are the basins with the first-, second-, and third-largest densities of generated TCs in the world, respectively. However, the statistically significant growing trend in the global TC number was mainly attributed to those TCs generated in the North Atlantic Ocean (NA). The changes in the TC numbers in the WNP, ENP, and SIO were all insignificant. Moreover, the decrease in the WTY frequency mainly occurred in the WNP, the increase in the TS frequency in the NA, and the increase in the ITY frequency in the NA and the North Indian Ocean (NIO). Additionally, the Accumulated Cyclone Energy (ACE) underwent a statistically significant increase only in the NA and NIO and varied insignificantly in the other ocean basins.

Through wavelet transformation, the characteristic periods of TC genesis were identified. There were four characteristic periods in the variation in global TC frequency, i.e., 2–3, ~4, 6–7, and 11–13 years. These periods were mainly attributed to the TCs generated in the WNP, ENP, and SIO, consistent with their proportions of TCs. Specifically, the variation with a period of 2–3 years in the global TC frequency was mainly attributed to the TC frequency changes in both the ENP and SIO. The 4-year-period variation was primarily due to the variation in the TC frequency in the WNP. The inter-decadal (11–13 years) variation was principally due to the changes in the TC genesis in the WNP and NA. Most of the TC metrics (frequencies of TCs, TSs, WTYs, ITYs, and ACE) across the whole globe, the two hemispheres, and the six ocean basins had a characteristic variation period of 3–7 years, which was consistent with the characteristic ones of the ENSO and Inter-decadal Pacific Oscillation (IPO). The correlations between TC genesis and ocean variabilities, i.e., the Niño 3.4 index, Southern Oscillation Index (SOI), IPO index, Indian Ocean Dipole (IOD) index, Tropical North Atlantic SST anomaly (TNA), North Atlantic Oscillation (NAO) index, and Atlantic Multi-decadal Oscillation (AMO) index, are determined first by comparing their characteristic periods of frequency variation and then calculating the Pearson correlation coefficients. The variations in the annual number of TCs, especially ITYs, were correlated with the ENSO, IPO, and AMO phenomena on both an inter-annual scale of 3–7 years and an inter-decadal scale of 10–13 years. The ITY numbers regarding the globe, WNP, and ENP were all positively correlated with the Niño 3.4 and IPO indices while being negatively correlated with the SOI. On the contrary, the ITY number in the NA was negatively correlated with the Niño 3.4 index and IPO index but positively correlated with the SOI. The TC frequencies in the NA generally presented a strong negative correlation with the TNA index but a positive one with the AMO index.