-
No significant correlation is found between the time series of the annual TC numbers during 1975–2020 in these two ocean basins (correlation coefficient = 0.16). On decadal time scales, the WNP TC numbers have two low-frequency (1975–88 and 1998–2020) and one high-frequency periods (1989–97) (Fig. 1a), while Atlantic TC numbers show a significant increase since 1995 (Goldenberg et al., 2001), giving a low-frequency period (1975–94) and a high-frequency period (1995–2020), with the mean number increasing from 9.2 to 15.5 (Fig. 1b). It should be noted that although the Atlantic TC number does not follow the Gaussian distribution, the Mann–Whitney U test shows that the difference of the regimes is also statistically significant at the 99% confidence level, suggesting that the regime shift is robust. The climatological mean difference in TC numbers between the two basins (hereafter referred to as W–A) is 13.0, with a standard deviation (σ) of 7.5. The values of W–A vary significantly, ranging from –7 in 2020 to 29 in 1994 (Fig. 1c). The W–A time series shows a significant downward trend (confidence level of 99%) and a significant interdecadal variation, with a decrease since 1998. Using 0.5 σ from the mean as the threshold, the value of W–A is considered to be small if W–A is ≤ 9 and large if the value is ≥ 17. During 1975–97, the mean W–A is 17.4, with the value of W–A being large in 12 years, normal in 10, and small in only one (1995). However, the mean W–A decreases to 8.5 during 1998–2020, suggesting that the difference in TC activity between the two basins has decreased since 1998. The values of W–A are mostly small or normal (9 and 13 years, respectively) and are large in only one year (2018).
Figure 1. Time series (black lines with dots) of the (a) WNP TC number, (b) Atlantic TC number, and (c) W–A from 1975 to 2020. (d) Time series of the 9-point Gaussian filtered AMO index (blue), TPI index (red), and W–A (green) from 1945 to 2020. Red dashed lines indicate the regime shifts. Vertical lines in (d) divide the study period into the periods with AMO+/IPO− and AMO−/IPO+.
Of particular importance is that the TC number in the Atlantic exceeds that in the WNP (i.e., W–A < 0) in four years (2005, 2010, 2011, and 2020), which never happened in the period 1975–97. The Atlantic TC seasons in 2005 (Trenberth and Shea, 2006) and 2020 are extremely active, with 29 and 30 TCs, respectively, much higher than the climatological mean (12.8) for the Atlantic basin and even higher than that of the WNP basin (25.8). The TC numbers in the WNP are 24 in 2005 and 23 in 2020, leading to negative values of W–A (–5 and –7, respectively). In 2010 and 2011, Atlantic TC numbers are also above-normal (both having 19 TCs), but the WNP has below-normal TC activity (15 and 18 TCs, respectively), thus resulting in negative values of W–A.
The decadal shift in W–A is related to the phases of the TC activity in the two basins. Both the Atlantic and WNP are in their low-frequency periods during 1975–88 (see Figs. 1a and 1b). However, the reduction of TC number in the WNP is smaller than that in the Atlantic, and so, the mean W–A (15.5) is slightly higher than the normal value (13.0). During 1989–94, WNP TC activity becomes more active, but the Atlantic is still in its low-frequency period, and so, the mean W–A increases to 22.3. The Atlantic then switches to its active phase in 1995 while the WNP moves into its inactive phase from 1998 (Liu and Chan, 2013), leading to a dramatic decrease in W–A during 1998–2020. Thus, the recent decrease in W–A is partly related to the co-occurrence of the high-frequency period (1995–2020) in the Atlantic and the low-frequency period (1998–2020) in the WNP.
Such a decrease in W–A as a result of the co-occurrence of an active Atlantic and a relatively quiet WNP has occurred only once during the post-satellite era. To determine whether this decrease is actually a trend or just a part of an interdecadal oscillation, a longer data period is required. The relatively reliable record in the Atlantic begins in 1944, when aircraft reconnaissance became routine (Vecchi and Knutson, 2008). The extended time series (1944–2020) of the Atlantic TC number (not shown) shows a high-frequency period between 1944 and 1971, with a mean of 11.6, which is lower than that in the recent high-frequency period 1995–2020 (15.5). The highest number is only 18 (in 1969), and no extremely active TC season like 2005 and 2020 appears in this earlier period. Although the TC number in this period may be underestimated due to the relatively poor observational coverage (Landsea, 2007; Vecchi and Knutson, 2008), the enhancement of Atlantic TC activity in this period is most likely not comparable to the period 1995–2020. Over the WNP, another low-frequency period is found during 1945–59 (not shown). Note again that the TC activity may also be underestimated. Nevertheless, a period with the co-occurrence of an active Atlantic and a quiet WNP likely exists during 1945–59 (Fig. 1d). The mean TC numbers in the WNP and Atlantic are 21.9 and 12.1, respectively, yielding a difference of 9.8, which is higher than that during 1998–2020 (8.5), and there is no TC season in which the Atlantic TC number exceeds that in the WNP. Thus, the decrease in W–A in the recent period (1998–2020) is unprecedented since 1945, although a similar period may have occurred during 1945–59. In the rest of this paper, all analyses will be for the period 1975–2020 unless otherwise mentioned.
-
The factors that cause the recent unprecedented decrease in W–A must be related to those that control the activities in the two basins. In the Atlantic, the AMO has a significant impact on TC activity (Klotzbach and Gray, 2008). Some recent studies also show a relationship between tropical Atlantic sea surface temperature (SST) and WNP TC activity (Huo et al., 2015; Zhang et al., 2018). Thus, the AMO should be related to the W–A variations. The correlation coefficients between the AMO index and the TC numbers in the Atlantic and WNP are 0.63 and –0.34, respectively, both being significant at the 95% confidence level (Table 1). Therefore, a high correlation (r = –0.64) also exists between the AMO index and W–A. Because the correlation with the Atlantic TC activity is higher, the impact of the AMO on W–A is mainly through the alteration of the Atlantic activity, and its influence on the WNP TC activity appears to play only a secondary role. On an interdecadal time scale, the AMO changes phase from negative to positive in 1995 (Fig. 1d), which coincides with the switch to the active Atlantic TC season, suggesting that the interdecadal variation of the Atlantic TC activity is largely related to the AMO (Goldenberg et al., 2001; Klotzbach and Gray, 2008). However, the impact of the AMO on interdecadal WNP TC activity appears to be less prominent. The mean numbers during the negative and positive AMO phases are 26.8 and 25.0, respectively, and the difference is not statistically significant. Nevertheless, the mean W–A during the positive AMO phase (9.5) is significantly lower than that during the negative AMO phase (17.6), with a confidence level of 99%.
WNP AT W–A Contribution to R2 AMO –0.34 0.63 –0.64 50.7% TPI 0.36 –0.59 0.64 30.2% Niño-3.4 0.38 –0.40 0.50 19.1% Table 1. Correlations of the TC numbers in the WNP and Atlantic (AT) with different indices (AMO, TPI, and ENSO). Percentages in the last column are the contributions to the R-squared values of the multiple linear regression model for W–A.
Besides the AMO, the IPO is another type of presumably natural climatic oscillation affecting the TC numbers in both the WNP and Atlantic. The abrupt decrease in TC number in the WNP since 1998 is related to the IPO, with a switch to its negative phase (Fig. 1d) (Zhao et al., 2018b). Li et al. (2015) showed that the IPO modulates the TC activity in the Atlantic, with more (less) TCs in the negative (positive) IPO phase. Thus, the IPO may have a significant impact on W–A. Indeed, the TPI is positively correlated with the WNP TC number (r = 0.36) but negatively correlated with the Atlantic TC number (r = –0.59), resulting in a significant correlation (r = 0.64) between TPI and W–A (Table 1).
The above results show that W–A is related to the co-variability of the AMO and IPO. Murakami et al. (2020) also pointed out that the trend in global TC number since 1980 is related to the multidecadal internal variability associated with the AMO and IPO. Because the effect of the AMO on W–A at a given phase is similar to that of the IPO with an opposite phase, their effects on W–A may be enhanced when they are out of phase. During the post-satellite era, there is one period with a negative AMO phase and positive IPO phase (1976–94) and one period with the opposite sign of phases (1998–2020) (Fig. 1d). Comparing the two periods, the TC frequency is higher in the Atlantic (15.9 vs. 9.2) but lower in the WNP (24.4 vs. 27.1) during the latter, giving a lower value of W–A (8.5 vs. 17.9), and all the differences are statistically significant at the 95% confidence level or above. Thus, the recent decrease in W–A is partly related to the co-occurrence of the positive phase of the AMO and negative phase of the IPO during 1998–2020.
To further validate this hypothesis, it is useful to examine the behavior of W–A in the previous period with a similar condition in AMO and IPO. During the pre-satellite era, one more period with a positive AMO phase and a negative IPO phase (1945–62, namely the first AMO+/IPO− period) is identified (Fig. 1d). The first AMO+/IPO− period generally matches the period with the co-occurrence of an active Atlantic and a quiet WNP TC season (1945–59), with mean TC numbers of 11.6 and 23.1, respectively. The mean W–A (11.5) is lower than the climatological mean during 1975–2020 (13.0), which further confirms the dependence of W–A on the co-variability of the AMO and IPO. It should be noted that no additional AMO−/IPO+ period can be identified during the pre-satellite era.
-
The decadal change in W–A has been shown to be partly related to the natural variability associated with the AMO and IPO. However, the greater decrease in W–A in the second AMO+/IPO− period (1998–2020) as compared with the first AMO+/IPO− period (1945–62) suggests that the variability of W–A may not be completely explained by the AMO and IPO, and other external forcings may make a significant contribution. Because the recent decrease in W–A is mainly due to the increase in Atlantic TC number, the role of the external forcings is most likely on the modification of the Atlantic TC activity. Murakami et al. (2020) also showed that the observed change in TC frequency since 1980, with a decrease in the WNP and an increase in the Atlantic, is unlikely to be explained entirely by multidecadal internal variability, and the external forcings such as greenhouse gases, aerosols, and volcanic eruptions may also play an important role. Some other studies have shown that the long-term change in Atlantic TC frequency is related to the change of tropical Atlantic SST (see also Fig. 2a), which is affected by the warming arising from greenhouse gases (Mann and Emanuel, 2006; Holland and Webster, 2007; Saunders and Lea, 2008) and anthropogenic aerosol cooling (Dunstone et al., 2013; Qin et al., 2020). Thus, external forcing from greenhouse gases and aerosols may also play a role in the variations of W–A.
Figure 2. (a) Correlation map between the SST and W–A during 1975–2020. (b) Spatial distribution of the trends in SST. Contour lines in (a) and (b) represent the correlation coefficients and trends in SST (units: °C yr–1), respectively. Blue and red shadings indicate the negative and positive correlations and trends significant at the 95% confidence level, respectively. The rectangular boxes in the WNP and Atlantic indicate the areas used in the definition of the TPI index and the tropical Atlantic region, respectively.
The upward trend in tropical Atlantic SST is probably partly related to the global warming induced by greenhouse gases, which is likely to have been less significant in the first AMO+/IPO− period while the natural variability associated with the AMO and IPO tended to be dominant. However, because the warming effect becomes more significant in the second AMO+/IPO− period, it superimposes on the effect of the AMO and IPO to cause a substantial increase in Atlantic TC frequency and hence the larger decrease in W–A. On the other hand, the aerosols may affect the TC frequency in the Atlantic through the cooling of SST in the tropical Atlantic (Dunstone et al., 2013).
Indeed, the aerosol cooling from the 1960s to mid-1990s might have largely canceled out the greenhouse warming (Mann and Emanuel, 2006; Qin et al., 2020), but this cooling becomes less significant in response to the decrease in anthropogenic aerosols at the end of the 20th century (Dunstone et al., 2013). Thus, the anthropogenic aerosol forcing over the Atlantic basin likely reduces the TC frequency in the AMO−/IPO+ period (1976–94) but enhances it in the second AMO+/IPO− period (1998–2020). Over the WNP, Takahashi et al. (2017) showed that changes in sulfate aerosol emissions contribute more than half of the observed decreasing trend in TC frequency in the southeastern part of the WNP during 1992–2011. Murakami et al. (2020) found that external forcing (greenhouse gases, aerosols, and volcanic eruptions) plays a role in the observed deceasing trend in TC frequency in the WNP since 1980. Thus, the decrease of WNP TC number in the second AMO+/IPO− period may be partly related to the anthropogenic aerosol forcing over the WNP. The decadal change in W–A is therefore fundamentally related to the co-variability of the AMO and IPO but modified by the external forcings of greenhouse warming and anthropogenic aerosol forcing, which is consistent with the results from Murakami et al. (2020). However, the relative importance of these forcings is still not clear.
-
In addition to the AMO and IPO, ENSO has long been known to have significant impacts on annual TC activity in both basins (Gray, 1984; Chan, 1985, 2000; Goldenberg and Shapiro, 1996; Wang and Chan, 2002). The Niño-3.4 index during the period 1975–2020 is positively correlated with the WNP TC number (r = 0.38) but negatively correlated with the Atlantic TC number (r = –0.40), both correlations being significant at the 95% confidence level (Table 1). Thus, a significant correlation (r = 0.50) also exists between W–A and ENSO. During El Niño years, TC activity is generally enhanced in the WNP, with a mean number of 27.0, but suppressed in the Atlantic (the mean number being 8.8), yielding a larger mean value of 18.2. The situation is just the opposite for the La Niña years. The mean TC numbers in the WNP and Atlantic are 22.0 and 16.4, respectively, indicating a significant decrease in W–A (the mean value being 5.6). Note that the differences in TC numbers are all statistically significant at the confidence level of 95% or above. Thus, the recent unprecedented decrease in W–A appears to be mainly attributed to the co-occurrence of a positive AMO phase and a negative IPO phase, the effects of greenhouse gas and aerosol forcing, and the La Niña condition.
It has been shown that W–A is related to the AMO, IPO, and ENSO. If the indices associated with these climatic oscillations are used as the predictors of W–A, the multiple regression model gives a correlation coefficient of 0.79, which is statistically significant at the 99% confidence level. The relative weight analysis shows that the contributions of the AMO, IPO, and ENSO to the R-squared values of the multiple linear regression model are 50.7%, 30.2%, and 19.1%, suggesting that the AMO is the main factor controlling the W–A.
-
W–A is negatively correlated with the SSTs in the tropical North Atlantic (10°–20°N, 80°–20°W) associated with the AMO and other external forcings, with a correlation coefficient of –0.73 (Fig. 2a). The spatial pattern of the correlations over the Pacific reveals the SST pattern associated with the IPO. Thus, the Atlantic TC genesis frequency appears to be more sensitive to the change of local SSTs than WNP TC genesis frequency. Climatologically, the SST in the main development region of the WNP is generally high enough for TC genesis and a further increase in SST may not lead to an additional increase in TC frequency. The variation of the tropical Atlantic SST consists of an upward trend (Fig. 2b) arising from greenhouse warming and aerosol forcing (Santer et al., 2006) and an interdecadal variation associated with the AMO (see Fig. 1d). The fast warming of the North Atlantic in recent decades is, therefore, a result of the co-occurrence of the human-induced temperature increase and reduced aerosol cooling. To remove this effect, W–A and SSTs are detrended, and the correlation coefficient between W–A and tropical Atlantic SSTs decreases from –0.73 to –0.66, which suggests that the downward trend of W–A is partly due to the upward trend of the tropical Atlantic SSTs, and a large part of its interdecadal variation is explained by the AMO and IPO.
The trend is insignificant in the equatorial Central and Eastern Pacific, and upward trends are generally found in the Northwest and Southwest Pacific, which reveals a La Niña-like SST pattern (Fig. 2b). The insignificant trend in the equatorial Eastern Pacific is related to the cooling of the equatorial Pacific associated with the recent global warming hiatus since 1998 (Kosaka and Xie, 2013; Zhao et al., 2018a). Some other studies have also shown that the response of the mean-state Pacific to greenhouse warming is a La Niña-like SST pattern (Kohyama et al., 2017; Lian et al., 2018). This pattern tends to strengthen the Walker circulation in the tropical Pacific, which weakens the monsoon trough and enhances the VWS in the southeastern part of the WNP, and therefore suppresses TC genesis in the WNP (Zhao et al., 2018a). Zhao et al. (2020) also showed that historical greenhouse warming and the negative phase of the IPO reduce TC genesis in the WNP, with the latter having a much greater impact. While historical greenhouse warming may enhance the TC frequency in the Atlantic, that in the WNP is not likely enhanced, as pointed out in previous studies (Chan and Liu, 2004). In fact, some modeling studies actually show a decreasing frequency in response to greenhouse warming (Murakami et al., 2020). The different responses of TC numbers in these two basins to greenhouse warming therefore leads to the larger decrease in W–A in recent years.
Besides SST, the upper oceanic condition, as measured by the TCHP, may also affect TC development and hence W–A. Because TCHP data are only available between 1980 and 2017, the analysis can only be performed for this period. The correlation map between the TCHP and W–A is similar to that for SST, with positive correlations in the equatorial Central Pacific and negative correlations in the tropical Atlantic and WNP (Fig. 3a). A high correlation (r = –0.75) exists between the tropical Atlantic TCHP and W–A, suggesting that the local upper oceanic condition is another important factor controlling Atlantic TC activity and hence W–A. In contrast, W–A is related to the TCHP pattern associated with the IPO rather than the local change of TCHP in the tropical WNP. The trend in TCHP is also similar to that for SST, with positive trends in the Atlantic and WNP but insignificant trends in the equatorial Central and Eastern Pacific (Fig. 3b). While the increasing trend of the TCHP in the tropical Atlantic may be partly responsible for the increase in Atlantic TC number, the increase of the TCHP in the tropical WNP may not lead to an increase in WNP TC number because the TCHP in this area is generally high enough for TC genesis and a further increase in TCHP may not have a significant impact on TC genesis. Therefore, the different responses to TCHP increases may partly contribute to the recent decease in W–A.
Figure 3. (a) Correlation map between the TCHP and W–A during 1980–2017. (b) Spatial distribution of the trends in TCHP. Contour lines in (a) and (b) represent the correlation coefficients and trends in TCHP (units: kJ cm–2 yr–1), respectively. Blue and red shadings indicate the negative and positive correlations and trends significant at the 95% confidence level, respectively. The rectangular boxes in the WNP and Atlantic indicate the areas used in the definition of the TPI index and the tropical Atlantic region, respectively.
In addition to the forcing by SST and TCHP, TC occurrence in both basins has been known to be related to VWS (Gray, 1979; Zhao and Wang, 2019), which should therefore contribute to the variation of W–A. Indeed, W–A is positively correlated with the VWS in the tropical Atlantic (10°–20°N, 80°–30°W) and negatively correlated with the VWS in the southeastern part of the WNP (10°–20°N, 150°–180°E) (Fig. 4a). Both the AMO and IPO have significant connections to the VWS in the tropical Atlantic (Li et al., 2015) and the southeastern part of the WNP (Zhang et al., 2018; Zhao et al., 2018b). The VWS in the tropical Atlantic is negatively correlated with the AMO index (r = –0.53) but positively correlated with the TPI index (r = 0.61), indicating a reduction in VWS during the AMO+/IPO− phase and hence an enhancement of Atlantic TC activity (Figs. 4b and 4c). The VWS in the southeastern part of the WNP is negatively correlated with the TPI index (r = –0.53) but positively correlated with the AMO index (r = 0.28). TC genesis is generally suppressed during the AMO+/IPO− phase as a result of the enhanced VWS. Note that the contribution of the AMO may be smaller, as indicated by the small correlation. Thus, the AMO and IPO affect the value of W–A through the alteration of the VWS in the tropical Atlantic and the southeastern part of the WNP.
Figure 4. Correlation maps (a) between the VWS and W–A, (b) between the VWS and AMO index, and (c) between the VWS and TPI index during 1975–2020. Contour lines represent the correlation coefficients. Blue and red shadings indicate the negative and positive correlations significant at the 95% confidence level, respectively. The rectangular boxes indicate the prescribed areas with significant correlations between the VWS and W–A.
Some previous studies have proposed the physical mechanism for which the local TC activity is influenced by the remote SST change (Li et al., 2015; Zhang et al., 2018). The primary connection of the characteristic SST pattern associated with the AMO and IPO to the anomalous atmospheric circulation over the WNP and Atlantic is through the Gill–Matsuno-type response, in which equatorial Kelvin (Rossby) waves modulate the circulation east (west) of the heating center (Matsuno, 1966; Gill, 1980). During the AMO-/IPO+ period, low-level easterly (westerly) anomalies are found in the tropical Atlantic (WNP) in response to the warming in the equatorial Central Pacific and cooling in the tropical Atlantic and western part of the WNP. At the upper level, the response is the westerly (easterly) anomalies in the tropical Atlantic (WNP). Thus, in the tropical Atlantic, the increased upper-level westerlies enhance the VWS while the increased low-level easterlies lead to anomalous anticyclonic flow, both providing an unfavorable environment for TC genesis. In the tropical WNP, the upper-level easterly anomalies reduce the VWS while the low-level westerly anomalies induce anomalous cyclonic flow, indicating a favorable environment for TC genesis. The changes in atmospheric circulations during the AMO+/IPO- period is opposite to those during the AMO-/IPO+ period.
WNP | AT | W–A | Contribution to R2 | |
AMO | –0.34 | 0.63 | –0.64 | 50.7% |
TPI | 0.36 | –0.59 | 0.64 | 30.2% |
Niño-3.4 | 0.38 | –0.40 | 0.50 | 19.1% |