Recent Weakening in Interannual Variability of Mean Tropical Cyclogenesis Latitude over the Western North Pacific during Boreal Summer

The western North Pacific (WNP) is the most active and productive area of tropical cyclone (TC) genesis, where the largest number of TCs in the world originate. The TCs formed over the WNP can cause severe natural disasters and affect billions of people in coastal countries (e.g., Zhang et al., 2009), and thus have intrigued a strong research interest in investigations of TC activity variations in the context of global climate change (Knutson et al., 2010; Mendelsohn et al., 2012; Walsh et al., 2015; Knutson et al., 2019). A better understanding of the variability and mechanisms of TC activity over the WNP is not only of great scientific significance, but also of great practical significance in disaster prevention and mitigation.

The mean TC genesis latitude, a relatively new climatic TC activity indicator, has received increasing attention in recent studies. Kossin et al. (2014) showed that the global average TC latitude where TCs achieve their maximum intensity has been migrating poleward systematically over the past 30 years, which is proposed to be plausibly associated with the expansion of the tropical belt under anthropogenic influences. Kossin et al. (2016) further illustrated that a similar poleward migration of the TC genesis latitude in the WNP can be identified in the Coupled Model Intercomparison Project Phase 5 (CMIP5) models and the poleward migration tends to continue into the future based on the projections under the representative concentration pathway 8.5 (RCP8.5) emissions scenario. Daloz and Camargo (2017) also revealed significant poleward shift of the mean TC genesis latitude over the Pacific during 1980–2013. Wang et al. (2016) found that the mean TC genesis latitudes in North Atlantic and eastern North Pacific tend to vary out-of-phase on both the interannual and multidecadal timescales, suggesting that the multidecadal-scale southward shift of TC genesis latitude is modulated by the Atlantic multidecadal oscillation (AMO), which might offset or mitigate the poleward migration trend of TCs in the North Atlantic basin in recent decades. Studholme and Gulev (2018) demonstrated coherent variations in the Hadley circulation and TC genesis latitude. Sharmila and Walsh (2018) also linked the recent poleward shift of TC genesis latitude to the Hadley cell expansion, which may potentially increase TC-associated hazards and destructiveness in regions of higher latitudes. Shen et al. (2018) conducted idealized numerical sensitivity experiments and demonstrated that the climatological mean TC tracks over the WNP tend to shift poleward systematically under the warmer sea surface temperature (SST) conditions.

There is a growing awareness that the WNP TC activity has experienced a significant decadal variation in the late 1990s, including a reduction in TC number (Liu and Chan, 2013; Hsu et al., 2014; Huangfu et al., 2017a; Zhao J. W. et al., 2018; Kim et al., 2020), shifts in TC genesis location and dominant TC track (He et al., 2015; Wu et al., 2015; Hu et al., 2018; Zhao et al., 2019a), changes in TC intensity (Tu et al., 2011; Zhao H. K. et al., 2018), enhanced simultaneous relationships with El Niño–Southern Oscillation (ENSO; Zhao and Wang, 2019) and the WNP monsoon (Zhao et al., 2019b), breakdown of the negative correlation between the pre-winter ENSO index and the WNP TC genesis frequency in the subsequent early TC season (Wang et al., 2019), and a decrease in TC genesis productivity of tropical cloud clusters over the WNP (Zhao et al., 2019c). These significant decadal changes in the WNP TC characteristics that occurred in the late 1990s are considered to be closely associated with the shift of tropical climate regime and shifting ENSO conditions (e.g., He et al., 2015; Hu et al., 2018; Kim et al., 2020).

In addition, as manifested by the weakening of the ENSO amplitude and fluctuation, intensity of the interannual variability (IIV) of ENSO in the tropical Pacific Ocean has weakened significantly since 2000 (Hu et al., 2013, 2020). Considering the impact of ENSO on the WNP TC activity (e.g., Lander, 1994; Chia and Rope-lewski, 2002; Wang and Chan, 2002; Chand et al., 2017), it is possible that IIV of the WNP TC activity might also be weakening during the recent decades in response to the weakening IIV of ENSO. However, possible decadal changes in the IIV of the WNP TC during the recent decades have not been extensively studied yet.

In this paper, we demonstrate that IIV of the seasonal mean WNP TC genesis latitude during boreal summer (June–July–August; JJA) has weakened significantly since the late 1990s, which has not been examined in previous studies. The possible reason and mechanism behind this weakening is discussed. The results presented here are dedicated to filling the knowledge gap in this area. The reminder of this paper is organized as follows. Section 2 describes the data and methods. The observed weakening in the IIV of the mean WNP TC genesis latitude since the late 1990s is presented in Section 3. The relationship between the strength of the north–south TC see-saw and the IIV of the mean TC genesis latitude is described in Section 4. Possible mechanisms for these observed changes are provided in Section 5. Summary and discussion are presented in Section 6.

In the present study, we use the TC dataset during 1970–2016 from the Joint Typhoon Warning Center (JTWC), archived in the International Best Track Archive for Climate Stewardship (IBTrACS) version v03r10 (Knapp et al., 2010), to depict TC activity over the WNP. The TC best-track data from the China Meteorological Administration (CMA) and Japan Meteorology Agency (JMA) are also applied to verify our results, which show similar features. The NCEP/NCAR monthly reanalysis (Kalnay et al., 1996) is used to examine large-scale environmental variables associated with TC gene-sis, including relative vorticity, vertical wind shear, and relative humidity. The monthly extended reconstructed SST (ERSST) version 3b (Smith et al., 2008) is applied to investigate SST anomalies related to TC variations.

In order to minimize the uncertainty in identifying weak systems, only those TCs that reach the intensity of at least tropical storm are included in the analysis. TC genesis latitude is defined as the first recorded latitude when the TC reached the tropical storm intensity with the maximum sustained wind speed exceeding 17.2 m s⁻¹. The seasonal mean TC genesis latitude index is calculated based on the average latitude of all TCs formed over the WNP during JJA. The results reported in this study are not sensitive to changes in the selected range of specific months, and very similar results can also be obtained for TCs during the extended summer (June–September or July–September).

The Niño3.4 index is defined as SST anomalies averaged over 5°S–5°N, 170°–120°W, which is used to represent ENSO intensity. Following Xie et al. (2016), a North Indian Ocean SST index is defined as SST anomalies averaged over the tropical North Indian Ocean (5°–25°N, 40°–100°E), which is used to depict the Indian Ocean SST warming or cooling. According to Hong et al. (2014), a tropical Atlantic SST index is defined as SST anomalies averaged over the tropical North Atlantic Ocean (0°–20°N, 25°–80°W), which is used to quantify the tropical Atlantic Ocean SST warming or cooling.

The IIV during a specific period is defined by the standard deviation of the interannual time series during that period. The interannual time series is obtained by applying a nine-yr high-pass filter to the anomalies, which are obtained by removing the long-term trend and seasonal cycle from the original time series.

The analysis methods used in this study mainly include correlation analysis, composite analysis, and regression analysis. We determine the statistical significance level based on the two-tailed P value using the Student’s t test. In the composite analysis, a high (low) TC genesis latitude year is determined if the mean TC genesis latitude anomaly in that year exceeds one positive (negative) standard deviation. Totally, eight high TC genesis latitude years (1970, 1975, 1981, 1985, 1988, 1996, 1999, and 2016) and eight low TC genesis latitude years (1972, 1976, 1979, 1982, 1987, 1993, 1997, and 2014) are identified.

Geographic distribution of JJA TC genesis locations over the WNP during 1970–2016 is presented in Fig. 1a. The mean TC genesis location during this period is 18.04°N, 135.09°E. The TC genesis latitudes range from 5° to 35°N. The latitude where TCs are generated the most is 15°N, followed by the latitude of 18°N with the second maximum TC genesis number (Fig. 1b). The TC genesis number gradually decreases to the north of 18°N and to the south of 15°N. The TC genesis longitudes range from 107.3° to 176.3°E. The maximum TC genesis number is found at 132°E, and the second maximum TC genesis number is found at 126°E (figure omitted).

Figure 1.  (a) Geographic distribution of TC genesis locations (asterisks) over the WNP during boreal summer (JJA) and (b) latitudinal distribution of TC numbers in the WNP during 1970–2016. A total of 499 TCs formed over the WNP during JJA of that period.

We calculate the time series of JJA mean WNP TC genesis latitude on the interannual timescale during 1970–2016 (Fig. 2a). The interannual variation of the time series shows an obvious weakening trend after the late 1990s, which also implies that an interdecadal change in the variability of the mean TC genesis latitude occurred around the late 1990s. This is confirmed by the 21-yr running standard deviation (Fig. 2b), which demonstrates a remarkable weakening in the IIV of the TC genesis latitude since the late 1990s. The same phenomenon can be observed in the analysis results based on TC data from other institutions including CMA and JMA. Two sub-periods are then chosen to represent the periods before and after the late 1990s: 1979–1997 (P1) and 1998–2016 (P2). The two selected sub-periods both have a duration of 19 yr. During P1, the standard deviation of the mean TC genesis latitude was 2.34°; whereas, during P2, the standard deviation of the mean TC genesis latitude reduced to 1.37°. The significance test shows that the differences between the two periods are statistically significant at the 95% confidence level. Compared with the value during P1, the standard deviation of the mean TC genesis latitude during P2 decreased by 41%, which is larger than the decline rate for the variance of Niño3.4 index (decreased by 28%) as shown in Hu et al. (2013).

Figure 2.  (a) Time series of JJA mean TC genesis latitude over the WNP during 1970–2016 on the interannual timescale and (b) 21-yr running standard deviation of the mean TC genesis latitude.

We also compare the standard deviations of the JJA mean TC number and genesis longitude between the two sub-periods. It is interesting to note that the standard deviations of the TC genesis number and longitude did not decrease significantly but increased during the recent period, although the variance of the Niño3.4 index weakened. During P1, the standard deviation of the TC genesis longitude was 3.29°, whereas the value increased to 4.45° (increased by 35%) during P2. The difference between the two periods is statistically significant at the 95% confidence level. Meanwhile, the standard deviation of the TC genesis number was 2.12 during P1, whereas the value slightly increased to 2.36 during P2. However, the difference in the standard deviation of TC genesis number between the two periods is not significant. These results imply that in the context of the weakening ENSO variability since the late 1990s, not all characteristic parameters of the TC have displayed a weakened variability. Compared with the TC genesis number and longitude, TC genesis latitude is the characteristic TC parameter that demonstrates the most obvious and significant IIV weakening since the late 1990s. Therefore, we will focus on the weakening in the IIV of mean TC genesis latitude over the WNP after the late 1990s in the following analysis.

In the following section, we examine the relationship between the strength of the north–south TC see-saw structure and the IIV of the mean TC genesis latitude.

We first investigate the difference in the latitudinal distribution of TC numbers between the high TC genesis latitude years and low TC genesis latitude years. The result is displayed in Fig. 3a, which shows an obvious north–south dipole structure. This result indicates that the variation in TC genesis latitude is accompanied by a see-saw of TC genesis frequency to the north and south of 15°N. The 15°N behaves as the dividing line that separates the northern and southern WNP regions, which was also applied in Cao et al. (2018) and considered to be crucial for understanding the interdecadal changes in the relationship between ENSO Modoki and the WNP TC numbers. The increase in the mean TC genesis latitude over the WNP is accompanied by an increase in the TC number to the north of 15°N (Fig. 3b) and a decrease in the TC number to the south of 15°N (Fig. 3c). The opposite is true in years with decreased mean TC genesis latitude. It is very interesting to note that an extremely similar north–south dipole structure of TC number difference can also be observed in the eastern North Pacific, which is accompanied with the TC genesis latitude variability there, and the zonal dividing line is also around 15°N (Wang et al., 2016).

Figure 3.  (a) Differences in the latitudinal distribution of TC number between the high TC genesis latitude years and low TC genesis latitude years. Scatterplots of the mean JJA TC genesis latitude and TC genesis number to (b) the north of 15°N and (c) the south of 15°N. The asterisk in the upper right corner of the correlation coefficient in (b) and (c) indicates that the correlation is significant at the 95% confidence level.

The negative correlation supports the out-of-phase relationship between the TC genesis numbers to the north of 15°N and south of 15°N (Fig. 4). Furthermore, observations show that the out-of-phase relationship between the TC genesis numbers to the north and south of 15°N has weakened significantly after the late 1990s (Fig. 4b). The correlation coefficient between the TC numbers to the north and south of 15°N is −0.74 during P1, but the value dropped to −0.22 during P2. The decrease in the correlation coefficient indicates a weakened north–south TC see-saw and dipole structure after the late 1990s, which may contribute to the observed weakening in the IIV of the mean TC genesis latitude shown in Fig. 2b.

Figure 4.  (a) Time series of JJA TC genesis number anomalies to the north (black line) and south (red line) of 15°N during 1970–2016 on the interannual timescale. (b) 21-yr sliding correlation coefficients between the JJA TC numbers to the north and south of 15°N. The red dots in (b) indicate the correlation is significant at the 95% confidence level.

Direct comparisons reveal the close relationship between the standard deviation of the mean TC genesis latitude and the correlation coefficient between the mean TC numbers to the north and south of 15°N (Fig. 5). The result indicates that a larger (smaller) standard deviation of the mean TC genesis latitude tends to occur when the TC numbers to the north and south of 15°N are more (less) negatively correlated. That is, a stronger (weaker) north–south TC see-saw would support a larger (smaller) IIV of the mean TC genesis latitude. The above results confirm that the IIV of the mean TC genesis latitude largely depends on the strength of the out-of-phase relationship between TC genesis numbers over the northern part (north of 15°N) and southern part (south of 15°N) of the WNP.

Figure 5.  Scatterplot of the 21-yr running standard deviations (Std) of the JJA mean TC genesis latitude and 21-yr sliding correlation coefficients between the JJA mean TC numbers to the north and south of 15°N. The asterisk in the upper right corner of the correlation coefficient indicates that the correlation is significant at the 95% confidence level.

In the condition with a stronger (weaker) north–south TC see-saw structure, the TC genesis number increases (decreases) in the north, but decreases (increases) in the south at the same time. When the TC number increases to the north of 15°N and decreases to the south of 15°N simultaneously, the average TC genesis latitude is naturally higher. Conversely, when the TC number deceases in the north and increases in the south at the same time, the average TC genesis latitude is naturally lower. If the north–south TC see-saw is stronger, the mean TC genesis locations would be more concentrated in the north or the south (farther away from 15°N). Thus, the mean TC genesis latitude would show a wider range of north–south swing on the interannual timescale. This is why a larger IIV (with a larger fluctuation range) in the mean TC genesis latitude would be observed.

On the contrary, when the north–south TC see-saw becomes weaker after the late 1990s, the mean TC genesis latitude shows a north–south variation within a narrower range on the interannual timescale. Compared with the situation before the 1990s, the distribution of TC genesis is more uniform in the north–south direction after the 1990s, causing the mean TC genesis latitude to be closer to 15°N. Thus, a smaller IIV (with a smaller fluctuation range) in the mean TC genesis latitude can be observed after the late 1990s.

In summary, the mean TC genesis latitude varied within a large range (i.e., larger IIV) when the north–south TC see-saw was strong before the late 1990s. On the contrary, the range within which the mean TC genesis latitude varies has become small (i.e., smaller IIV) in response to the weakening TC see-saw that occurred after the late 1990s.

Why has the out-of-phase relationship between the TC genesis numbers to the north and south of 15°N been weakening after the late 1990s? The significant decadal changes in the influences of tropical SST anomalies may be the main driver. Different tropical SST anomalies may trigger different environmental anomaly fields over the WNP, and thus result in different TC formation activity characteristics in the recent period.

In the next, we first compare the regressed SST anomalies with respect to the mean TC numbers to the north and south of 15°N respectively during the above two periods (Fig. 6). As shown in Fig. 3, the variability of TC genesis latitude is closely associated with changes in TC numbers to the north and south of 15°N. During P1, significant cold (warm) SST anomalies in the tropical central-to-eastern Pacific (CP–EP) can be observed in the regressions with respect to the TC number to the north of 15°N (south of 15°N; Figs. 6a, b). These tropical Pacific SST anomalies mainly reflect the effects of ENSO. In addition, cold SST anomalies in the tropical North Indian Ocean and warm SST anomalies in the tropical western Pacific are found in the regression to TC number to the north of 15°N (Fig. 6a). Significant SST anomalies with opposite signs in the tropical North Indian Ocean and western Pacific are found in the regression to TC number to the south of 15°N (Fig. 6b). The La Niña events, accompanied by cold tropical Indian Ocean SST anomalies, tend to increase TC number to the north of 15°N, but decrease TC number to the south of 15°N. On the contrary, the El Niño events, accompanied by warm tropical Indian Ocean SST anomalies, tend to suppress TC number to the north of 15°N, but enhance TC number to the south of 15°N. Thus, the joint effects of SST anomalies in the tropical Pacific and North Indian Ocean that occurred before the late 1990s (Figs. 6a, b) both contribute to the out-of-phase variations of TC numbers to the north and south of 15°N as shown in Fig. 4.

Figure 6.  Regressions of JJA SST anomalies (shadings; °C) and 850-hPa wind anomalies (vectors; m s⁻¹) with respect to (a, c) mean TC number to the north of 15°N and (b, d) TC number to the south of 15°N during (a, b) P1 and (c, d) P2. Only values at or above the 90% confidence level are shown.

Anticyclonic circulation anomalies occurred to the south of 15°N (Fig. 6a), resulting in the reduction of TC number in the south. These anticyclonic circulation anomalies can be regarded as the Gill-type atmospheric response (Gill, 1980) to cold SST anomalies in the tropical central Pacific (CP). The warm SST anomalies in the tropical CP behave like a tropical heating source, which drives a Gill-type response to the diabatic heating. The westward-propagating disturbances via excitation of Rossby waves associated with the Gill-Matsuno mechanism can generate an anomalous cyclonic circulation over the WNP. Similar conditions but with opposite signs occur when cold SST anomalies exist in the tropical CP, which behaves as a tropical heating sink or cooling source. At the same time, cyclonic circulation anomalies can be found to the north of 15°N (Fig. 6a), promoting more TC genesis there. These cyclonic circulation anomalies to the north of 15°N might be explained by the atmospheric responses to cold SST anomalies in the tropical North Indian Ocean. The cold (warm) SST anomalies in the tropical North Indian Ocean can generate a cyclonic (anticyclonic) circulation anomaly to the north of 15°N through atmospheric Kelvin waves (Xie et al., 2009; Du et al., 2011) or atmospheric zonal overturning circulation (He and Wu, 2014; Cao et al., 2018). Very similar dipole structures of circulation anomalies with opposite signs can be found in the regressions to TC number south of 15°N (Fig. 6b), leading to an increase (a decrease) of TC number to the south of 15°N (north of 15°N). Therefore, these consistent SST configurations in the Indo–Pacific region during P1 (Figs. 6a, b) can at least partly explain the strong negative correlation between TC genesis numbers to the north and south of 15°N as shown in Fig. 4.

In contrast, the configurations of SST anomalies during P2 were found to be inconsistent, which is clear when we compare the SST anomalies regressed onto the TC number to the north of 15°N (Fig. 6c) with that regressed onto the TC number to the south of 15°N (Fig. 6d). During P2, negative SST anomalies in the tropical North Indian Ocean can still be found in the regression with respect to TC number to the north of 15°N (Fig. 6c). However, SST anomalies in the tropical Pacific that existed in the previous period of P1 (Fig. 6a) almost completely disappeared in the recent period of P2 (Fig. 6c). Only cyclonic circulation anomalies are found (Fig. 6c) to the north of 15°N. However, there were no anticyclonic circulation anomalies located to the south of 15°N, which can be found during P1 (Fig. 6a). The disappearance of the southern anticyclonic circulation anomaly is considered as the result of the absence of the tropical CP SST anomalies during P2 (Fig. 6c). By comparison, significant warm SST anomalies in the tropical CP were observed in the regressions with respect to TC number to the south of 15°N (Fig. 6d). The SST anomalies in the tropical eastern Pacific (EP) were relatively weak (Fig. 6d), indicating possibly more influences from the CP-type ENSO rather than that from the EP-type ENSO. Previous studies have also revealed an increasing frequency in the occurrence of these CP-type ENSO events after the 1990s (e.g., McPhaden et al., 2011; McPhaden, 2012; Xiang et al., 2013; Lübbecke and McPhaden, 2014), which may result in more influences from the CP-type ENSO during P2. These warm SST anomalies in the tropical CP can produce atmospheric cyclonic circulation anomalies to the west via the Gill-type atmospheric responses (Gill, 1980). The westward-propagating disturbance via the excitation of Rossby waves triggered by the tropical heating source associated with warm SST anomalies can generate an anomalous cyclonic circulation over the WNP via the Gill-Matsuno mechanism. Compared with Fig. 6b, SST anomalies in the tropical North Indian Ocean have weakened distinctly (Fig. 6d). However, the negative SST anomalies in the tropical North Atlantic Ocean tended to be enhanced during P2 (Fig. 6d). The tropical Atlantic SST anomalies can then convey their impacts on the WNP atmospheric circulation anomalies via trans-basin atmospheric zonal and vertical circulation anomalies (e.g., Hong et al., 2014; Jin and Huo, 2018; Wang and Yu, 2018; Wu and Wang, 2019), and may further influence TC genesis and activity over the WNP. The cold SST anomalies in the tropical Atlantic may trigger descending anomalies over the tropical Atlantic and ascending anomalies over the equatorial CP. This anomalous ascending motion over the equatorial CP can further induce a low-level cyclonic anomaly to the west, which can then constructively reinforce the cyclonic anomalies triggered by the warm SST anomalies in the tropical CP. The dipole structure of circulation anomalies during P1 (Fig. 6b) tends to be weakened substantially during P2 (Fig. 6d). During P2, the TC number to the north of 15°N (south of 15°N) tends to be associated with the SST anomalies in the tropical North Indian Ocean (the tropical Pacific Ocean and the North Atlantic Ocean) as suggested by Fig. 6c (Fig. 6d). These inconsistent SST configurations during P2 (Figs. 6c, d) may be responsible for the significant weakening of the relationship between TC genesis number to the north and south of 15°N after the late 1990s as shown in Fig. 4.

Very similar remarkable differences can also be observed in the regressed SST anomalies with respect to the mean TC genesis latitude between P1 and P2 (Fig. 7). During P1, obvious cold SST anomalies can be found in the tropical CP–EP (Fig. 7a), supporting the argument that La Niña events can help increase the TC genesis latitude. Along with cold SST anomalies in the tropical CP–EP, cold SST anomalies appeared in the tropical North Indian Ocean and warm SST anomalies appeared in the tropical western Pacific during P1. These SST configurations could induce anticyclonic circulation anomalies (favorable for suppressed TC numbers) to the south of 15°N and cyclonic circulation anomalies (favorable for enhanced TC numbers) to the north of 15°N over the WNP (Fig. 7a). As a consequence, the increased TC number to the north of 15°N and decreased TC number to the south of 15°N resulted in a higher mean TC genesis latitude. In contrast, different SST configurations during P2 were observed (Fig. 7b). During P2, cold SST anomalies were concentrated in the tropical CP but not in the tropical EP, suggesting that the variation of mean TC genesis latitude during P2 may be more affected by the CP-type ENSO rather than the EP-type ENSO. These conclusions are consistent with the results shown in Fig. 6. In addition, cold SST anomalies in the tropical North Indian Ocean and warm SST anomalies in the tropical western Pacific, which were clearly identified during P1 (Fig. 7a), tended to decay greatly during P2 (Fig. 7b). It is interesting to note that significant warm SST anomalies occurred in the tropical North Atlantic during P2 (Fig. 7b) but they did not occur in the previous period of P1 (Fig. 7a). These results indicate that the variation of the TC genesis latitude is mainly modulated by the summertime SST anomalies over the tropical CP–EP and the tropical North Indian Ocean during P1, but it is more controlled by the summertime SST anomalies of the tropical CP and the tropical North Atlantic Ocean during P2 (Fig. 8).

Figure 7.  Regressions of JJA SST anomalies (shading, °C per latitudinal degree) and 850-hPa wind anomalies (vector, m s^-1 per latitudinal degree) with respect to mean TC genesis latitude during (a) P1 and (b) P2. Only values at or above the 90% confidence level are shown.

Figure 8.  Comparisons of correlation coefficients between JJA SST anomalies in key regions and the WNP TC characteristics during two periods of P1 (blue) and P2 (red): (a) mean TC genesis latitude, (b) mean TC number to the north of 15°N, and (c) mean TC number to the south of 15°N. SST anomalies in three key regions, i.e., Niño3.4, tropical North Atlantic (TA), and tropical North Indian Ocean (TIO), are displayed. The dashed lines represent the 95% confidence level.

These observed decadal changes in the configuration of SST anomalies are considered to be closely associated with the ENSO conditions shifting from the EP type to CP type during recent decades. Previous studies suggested that the more frequent occurrences of the CP ENSO during recent decades may possibly be associated with the increased influences of the Atlantic Ocean SST on the Pacific (e.g., Ham et al., 2013; Yu et al., 2015; Wang et al., 2017). Recent studies also support the rise of the tropical Atlantic influence since the late 1990s (e.g., Cai et al., 2019). Thus, it is natural that enhanced signals from SST anomalies over the tropical Atlantic can be observed during P2 (Figs. 6d, 7b), accompanied by the dominant occurrence of CP-type ENSO. In contrast, the tropical Indian Ocean SST anomalies tend to have a stronger association with the EP ENSO than with the CP ENSO (Yuan et al., 2012). Therefore, more significant signals in the Indian Ocean could be observed during P1 (Fig. 7a), when the EP ENSO dominated. However, the tropical Indian Ocean SST anomalies tended to decay greatly during P2 (Fig. 7b), when the CP ENSO became more dominant.

In summary, the significant decadal changes in influences from the tropical SST anomalies during JJA are considered as the main driver for the recent weakening in the north−south TC see-saw structure and IIV of mean TC genesis latitude. Before the late 1990s, the summertime SST anomalies in the tropical CP−EP associated with ENSO tended to be accompanied by summertime SST anomalies in the tropical North Indian Ocean. The contemporaneous summertime SST anomalies in the tropical Pacific−North Indian Oceans can modulate the TC genesis number to the south and north of 15°N (with opposite effects). Therefore, the north−south TC see-saw and dipole structure were strong during P1, which supported a larger IIV of mean TC genesis latitude. After the late 1990s (during P2), the summertime SST anomalies in the tropical Pacific associated with ENSO tended to be accompanied by summertime SST anomalies in the tropical North Atlantic Ocean. Thus, the dominant effect of the summertime SST anomalies shifted to a combined effect from SST anomalies over the tropical CP and the tropical North Atlantic Ocean. The concurrent summertime SST anomalies in the tropical Pacific−North Atlantic Oceans can only modulate the TC genesis number to the south of 15°N, while they have little impact on areas to the north of 15°N. Therefore, the north−south TC see-saw and dipole structure became weak during P2, which caused a smaller IIV of mean TC genesis latitude.

In the next, we used the Genesis Potential (GP) index to identify how different environmental factors can contribute to the observed north−south TC see-saw and dipole structure. The above observed decadal shift of SST may influence the large-scale background environment factors (e.g., factors in the GP index) over the WNP, and thus modulate the decadal shift of the WNP TC activities and characteristics.

Four environmental factors (low-level atmospheric relative vorticity, vertical wind shear, relative humidity, and potential intensity) are included in the GP index to assess the impact of large-scale environmental factors on tropical cyclogenesis (Emanuel and Nolan, 2004; Camargo et al., 2007). These large-scale background environmental factors are considered to influence the TC genesis greatly (e.g., Gray, 1979; Camargo et al., 2007). Low-level atmospheric relative vorticity and vertical wind shear are dynamic factors, and relative humidity and potential intensity are thermodynamic factors.

The regional mean values of the GP index over the north and south of 15°N tend to vary out-of-phase (Fig. 9a). The negative correlation coefficient confirms the out-of-phase relationship between the GP indexes over the northern and southern regions (Fig. 9b), which supports the formation of the north−south TC see-saw and dipole structure as shown in Figs. 3, 4. The out-of-phase relationship between the northern and southern GP indexes tends to be weakening during recent decades (Fig. 9b), which is consistent with the weakening of the out-of-phase relationship between the TC genesis numbers to the north and south of 15°N (Fig. 4b). Therefore, the changes in the GP index can also explain the recent changes in the north−south TC see-saw and dipole structure.

Figure 9.  (a) Time series of regional average GP index over the regions to the north (black line) and south (red line) of 15°N during 1970−2016 on the interannual timescale. (b) 21-yr sliding correlation coefficients between the mean GP index anomalies to the north and south of 15°N, respectively. The red dots in (b) indicate the correlation is significant at the 95% confidence level.

We further examine the four environment factors in the GP index to identify their relative contributions to changes in TC genesis and activity. These four environmental factors are respectively regressed onto the mean TC genesis latitude and results are compared (Fig. 10). An obvious dipole structure can be observed in the regression of the 850-hPa relative vorticity (Figs. 10a, e) with significant positive (negative) anomalies to the north (south) of 15°N. These signals in the 850-hPa relative vorticity are consistent with atmospheric circulation anomalies (opposite signs in cyclonic/anticyclonic circulation anomalies to the north and south of 15°N) as shown in Fig. 7. For other three environmental factors, no obvious and continuous north−south dipole structure can be found. These results suggest that the low-level atmospheric relative vorticity has made the largest contribution to the observed see-saw of TC genesis number and the variation of mean TC genesis latitude.

Figure 10.  Regressions of environmental factors associated with TC genesis with respect to mean TC genesis latitude during P1 (left) and P2 (right). (a, e) 850-hPa relative vorticity (10⁻⁶ s⁻¹ per latitudinal degree), (b, f) vertical wind shear (m s⁻¹ per latitudinal degree), (c, g) 500-hPa relative humidity (% per latitudinal degree), and (d, h) potential intensity (m s⁻¹ per latitudinal degree). The dots denote the regression exceeding the 90% confidence level.

We further calculate the regional mean time series of 850-hPa relative vorticity to the north and south of 15°N, respectively (Fig. 11a). Negative correlations can be observed (Fig. 11b), consistent with the north−south dipole structure shown in Figs. 10a, e. The negative correlation for the 850-hPa relative vorticity tended to weaken substantially after the late 1990s, which is consistent with the observed weakening of the negative correlations between TC numbers to the north and south of 15°N during the recent decades (Fig. 4b). The correlation coefficient between the 850-hPa relative vorticity to the north and south of 15°N was −0.48 during P1, but the value dropped to be −0.27 during P2.

Figure 11.  (a) Time series of regional average 850-hPa relative vorticity to the north (black line) and south (red line) of 15°N during 1970−2016 on the interannual timescale. (b) 21-yr sliding correlation coefficients between the mean 850-hPa relative vorticity anomalies to the north and south of 15°N. The red dots in (b) represent the correlation is significant at the 95% confidence level.

By examining four environmental factors in the GP index, the low-level atmospheric relative vorticity is found to be the one that makes the largest contribution. Before the late 1990s, the joint effects of the tropical CP−EP and the tropical North Indian Ocean SST anomalies dominated, rendering the north−south dipole structure of low-level atmospheric circulation anomalies (opposite signs in cyclonic/anticyclonic circulation anomalies to the north and south of 15°N). An obvious dipole structure can be observed for the 850-hPa relative vorticity, with opposite-signed anomalies to the north and south of 15°N. Correspondingly, a stronger north−south TC see-saw and a larger IIV of the mean TC genesis latitude are found before the late 1990s. After the late 1990s, the dominant effect of summertime SST anomalies has shifted to the combined effects from the tropical CP and the tropical North Atlantic Ocean SST anomalies, which may weaken the north−south dipole structure of low-level atmospheric circulation anomalies. The dipole structure for the 850-hPa relative vorticity has also weakened greatly. Correspondingly, significantly weak-ened north−south TC see-saw and reduced IIV of mean TC genesis latitude can be observed after the late 1990s.

In this study, the IIV of the mean TC genesis latitude over the WNP is found to be weakening greatly since the late 1990s, which represents a significant decadal variation in the IIV of the mean TC genesis latitude occurred in the late 1990s that has not been well examined in previous studies. The results of the present study suggest that this observed significant decadal change in the IIV of the mean TC genesis latitude over the WNP is closely associated with the shifting ENSO conditions that occurred in the late 1990s.

It is found that the IIV of the mean TC genesis latitude largely depends on the strength of the out-of-phase relationship between TC genesis numbers over the north (north of 15°N) and south (south of 15°N) of the WNP. The interannual variation of the mean TC genesis latitude over the WNP is accompanied by a see-saw of TC occurrence frequency to the north and south of 15°N. The increase in the mean TC genesis latitude is accompanied by an increase in the TC number to the north of 15°N and a decrease in the TC number to the south of 15°N. The opposite is true in years with decreased mean TC genesis latitude. The weaker (stronger) north−south TC see-saw can lead to a smaller (larger) IIV of the mean TC genesis latitude after (before) the late 1990s. The weakening of the north−south TC see-saw and dipole structure after the late 1990s contribute to the observed weakening in the IIV of the mean TC genesis latitude.

Different configurations of summertime SST anomalies are found to be responsible for the decadal change in the north−south TC see-saw and dipole structure (Fig. 12). Different tropical SST anomalies can trigger different environmental anomaly fields over the WNP, and thus result in different characteristics of TC genesis and activity in the prior and recent periods. Before the late 1990s, the joint effects of the tropical CP−EP and the tropical North Indian Ocean SST anomalies dominated, rendering the obvious north−south TC see-saw and larger IIV of the mean TC genesis latitude (Fig. 12a). After the late 1990s, the dominant effect of summertime SST anomalies shifted to the combined effects from the tropical CP and the tropical North Atlantic Ocean SST, which may weaken the north−south TC see-saw and reduce the IIV of the mean TC genesis latitude (Fig. 12b).

Figure 12.  A schematic diagram of the physical processes responsible for the recent weakening in the IIV of the mean TC genesis latitude over the WNP. (a) Before the late 1990s and (b) after the late 1990s.

These observed decadal changes in the configurations of summertime SST anomalies are considered to be closely associated with the ENSO conditions shifting from the EP type to the CP type during recent decades, which may be associated with the increased influences of the Atlantic Ocean SST on the Pacific as suggested in previous studies (e.g., Yu et al., 2015; Wang et al., 2017). The greatly enhanced signals of SST anomalies in the tropical Atlantic can be observed during the recent period, accompanied by the dominant occurrence of CP-type ENSO. In contrast, the associated summertime tropical Indian Ocean SST anomalies tended to decay greatly during the recent period, whereas it was significant during the prior period when the EP-type ENSO prevailed. The results suggest that the increased influences from the tropical Atlantic Ocean have become more important to the WNP TC activity variations during recent decades. The increased influences of the tropical Atlantic Ocean during recent decades may trigger the recent weakening in IIV of mean TC genesis latitude over the WNP.

Four environmental factors in the GP index are examined to identify how different environmental factors contribute to the observed north−south TC see-saw. The results suggest that the low-level atmospheric relative vorticity may make the largest contribution to the observed north−south TC see-saw and the variations of the mean TC genesis latitude.

These results may have important implications for assessing latitudinal distributions of the TC-associated hazards. Along with the poleward migration of the TC latitude recently (e.g., Kossin et al., 2014; Daloz and Camargo, 2017), poleward migrations of the destructive effects of TCs were also documented (e.g., Oey and Chou, 2016; Altman et al., 2018). Increasing hazard exposure and disaster risks from TCs may be experienced in coastal regions with relatively higher latitudes, accompanied with the poleward-shifting of TC locations. Similarly, the weakening in the IIV of mean TC genesis latitude reported in this study may reduce the dispersion of typhoon disasters in the spatial distribution and make the area affected by typhoon disasters more concentrated at some specific latitudes.

The present study mainly focuses on the summer season. Previous studies also have revealed that changes in the TC activity over the WNP may be different during different seasons (Tu et al., 2011; Hsu et al., 2014, 2017; Wang, 2016; Xu and Wang, 2014; Zhao and Wang, 2016; Huangfu et al., 2017b; Fan et al., 2019; Wang et al., 2019; Yao et al., 2020). The possible changes in the IIV of TC characteristics in the early or late season need to be investigated in details in the future study. The results also show that the changes in different TC characteristics (TC genesis number, longitude, and latitude) may be different. In this paper, we mainly focus on interdecadal variation in the IIV of the mean TC genesis latitude since the late 1990s. The possible changes in the IIV of the mean TC genesis longitude and genesis number need to be studied in detail in the future.

Along with the weakening of the variability in the tropical Pacific including ENSO during recent decades (e.g., Hu et al., 2013), a weakened SST variability in the tropical Atlantic Ocean since 2000 was also revealed (Prigent et al., 2020). Wang (2017) revealed the recent weakening of the variability in the coupled tropical Pacific and Atlantic climate system, including the rainfall, winds, and SSTs in both the tropical Pacific and Atlantic regions. The weakened variability in the tropical Atlantic Ocean may also influence the IIV of TC characteristics over the North Atlantic and Northeast Pacific regions. Just like the weakened IIV of the mean TC genesis latitude over the WNP presented in this study, similar interdecadal changes may also exist in the IIV of TC characteristics in other ocean basins (such as the tropical North Atlantic and Northeast Pacific), which deserve further research in the future.

Acknowledgments. We thank the three anonymous reviewers for their comments and suggestions.