Interdecadal variation of tropical cyclone genesis longitudes over the western North Paci�c

The genesis longitudes of the tropical cyclone (TC) for the months of June–October exhibit the gradual westward shift since 1998 and the present study investigates the possible mechanisms. The period of 1998–2020 is characterized by the large anomalous anticyclonic circulation over the western North Paci�c (WNP) and an anomalous cyclonic circulation in East Asia. Owing to this huge anomalous anticyclonic circulation, TCs could not develop over the eastern part of the WNP. In addition, the equatorial western Paci�c shows a warm anomaly, while the equatorial eastern Paci�c presents a cold anomaly; this negative phase of the Paci�c Decadal Oscillation (PDO) demonstrates the strengthening of the Walker circulation, which causes greater prevalence of TCs to the western seas of the WNP.


Introduction
The western North Paci c (WNP) is the water body where tropical cyclone (TC) activity is the most active on Earth.The TCs that occur in these waters migrate west to affect southern China or Vietnam or north to affect the mid-latitude region of East Asia.The regions affected by TCs are mostly determined by the location of TC genesis (Chia and Ropelewski, 2002; Wang and Chan, 2002).The TCs that occurred in the southeastern waters of the WNP migrate northwest and recurve in the East China Sea and eventually affect East Asia.As these TCs travel long distances over the ocean, they gain su cient energy from the ocean to generate higher intensity and affect the mid-latitude region of East Asia.Meanwhile, TCs that occur in the northwestern waters of the WNP generally migrate short distances to the west to affect southern China and Vietnam.These TCs generally have low intensity because of the topographical effect of the Philippines.
Studies have focused on the tendency of TCs to occur over the western part of the WNP and have revealed that the tropical upper-tropospheric trough (TUTT), monsoon trough, and El Niño-Southern Oscillation (ENSO) affected WNP TC genesis locations (Huang et  ).ENSO is the largescale SST uctuation that occurs in the tropical WNP and the resultant changes in air-ocean circulation affect TC genesis in various regions.Gray (1984) showed that ENSO affected hurricane genesis in the North Atlantic region and Hastings (1990)  In Section 2, the data and methodology are shown, and a time series of TC genesis longitude locations is analyzed in Section 3. In Section 4, the recent westward shift of the WNP TC genesis location and atmospheric circulations is analyzed.In Section 5, the effect of SST variation on the recent westward shift of the WNP TC genesis location is analyzed.The results of this study are summarized in Section 6.

Data
The best-track data of the Regional Specialized Meteorological Center-Tokyo Typhoon Center were used as the TC data.In this study, Reanalysis-2 (R-2) monthly average data of the National Center for Environmental Prediction-Department of Energy were used (Kanamitsu 2002).These data consist of 2.5°× 2.5° latitude-longitude grid spacing with a total of 17 vertical layers from 1979 to 2020.Extended Reconstructed Sea Surface Temperature (ERSST) V3b (Smith et al., 2008) data were used as the SST data.ERSST data are the monthly averages from 1854 to the present and consist of 2° × 2° grid spacing.

Methodology
The vertical wind shear (VWS) was calculated as the method of Wingo and Cecil (2010).The genesis potential index (GPI) is de ned from the equation of Camargo et al. (2007).To determine the signi cance of the results, the two-tailed Student's t test was used (Wilks 1995).

Time Series Analysis
Figure 1 shows the time series of TC genesis longitude averaged during June-October for 1979-2020 (red line with dots).This time series has distinct interannual and interdecadal variations.The time series shows a decreasing trend until recently (signi cant at the 90% con dence level), indicating a strong tendency in the occurrence of TCs.Therefore, a statistical change-point analysis was applied to this time series (blue line) using a 9-year sliding window.The largest absolute value of the t-value appears in 1998, indicating a shift in the climate regime around this year.The average longitude of TC genesis was 136.9°E during 1979-1997 and 132.3°E during 1998-2020, with a difference of 3.6° between the two periods.This difference was signi cant at the 95% con dence level.This implies that TCs occurred farther west by approximately 300-400 km during 1998-2020 than during 1979-1997.In the next section, the average difference between the periods 1998-2020 and 1979-1997 is analyzed to determine why more recent TCs have occurred farther west of the WNP than before.

Large-scale Environment And Atmospheric Circulation
First, we analyzed the difference in TC genesis frequency between the two periods (Fig. 2a).TCs during 1998-2020 had a strong tendency to occur mainly in the sea near the Philippines and in the South China Sea.However, during 1979-1997, TCs had a higher preference in the eastern waters of the WNP.That is, recently, TCs occurred more often to the western side of the WNP.The TC passage frequency indicates that the frequency of TC migration along the East Asian coast is higher (Fig. 2b); therefore, East Asian coastal countries have the increased TC-related risk.
The reason why recent TCs have occurred west of the WNP can be found from GPI eld.A positive anomaly was located in the northern region as well as in the western region of the WNP (Fig. 3a).
Conversely, a negative anomaly was located south of 20°N and east of 150°E.As a proxy for deep convection, OLR is shown in Fig. 3b where a negative OLR anomaly is distributed in the western region of the WNP (relative to 140°E), and a positive anomaly is in the eastern part of the WNP.This means that during 1998-2020, convection was more active in the western WNP than during 1979-1997.The precipitable water shows the similar results as those of the GPI (Fig. 3c).Except for the eastern seas of the WNP, most of the analyzed areas showed a positive anomaly.Rainfall eld indicates a spatial distribution opposite to that of the OLR (Fig. 3d); the region west of 140°E shows the positive anomaly, while the region east of 140°E shows the negative anomaly.
The occurrence of TCs in more or less farther western area of the WNP during 1998-2020 can also be attributed to environmental factors affecting the occurrence of TCs.For example, the 600-hPa relative humidity (RH600) shows that the region west of 150°E has a positive anomaly, while the region east of 150°E has a negative anomaly (Fig. 3e).The VWS eld shows that the negative anomaly south of 10°N is located farther west than the positive anomaly at 10°-20°N (Fig. 3f). Figure 3g shows a positive low-level relative vorticity anomaly located slightly farther northwest than the negative anomaly in the WNP.The 850-hPa horizontal divergence shows anomalous convergence over the western region of the WNP, while anomalous divergence over the eastern region of the WNP (Fig. 3h).All these indicate that convection during 1998-2020 is more active in the west of the WNP than during 1979-1997 To examine whether the climate regime shift occurred in 1998 in all of the above variables, statistical change-point analysis for GPI, OLR, and RH600 was applied to the area-averaged time series in the western (0°-25°N, 100°-140°E) and eastern regions (0°-25°N, 140°-180°E) of the WNP (Fig. 4), where 140° E is the longitude of dividing the opposite anomalies in the OLR eld (Fig. 3b).Trends for GPI and RH600 averaged in the western WNP shows an increasing trend, but OLR shows a decreasing trend (these trends were signi cant at the 95% con dence level), indicating that the large-scale environment favorable for the occurrence of TCs in the western region of the WNP has formed in recent years.The statistical change-point applied to the time series reveals that the largest absolute value of the t-value existed for 1998 for all three variables.Thus, the supportive environments for the formation of TCs have been present in the western WNP since 1998.Meanwhile, environments unfavorable for the occurrence of TCs have been formed in the eastern WNP (the area-averaged GPI and RH600 decreased, while OLR increased).This region also shows the largest absolute value of the t-value occurring in 1998, indicating that fewer TCs have occurred in the eastern region of the WNP since 1998.
The changes in the large-scale environments, shown above, between 1998-2020 and 1979-1997 can be identi ed in the difference in the 850-hPa stream ow between the two periods (Fig. 5a).Whereas the large anomalous anticyclonic circulation exists in the WNP region, an anomalous cyclonic circulation is located in East Asia.This is the west-low-east-high spatial pressure pattern.Owing to the huge anomalous anticyclonic circulation located in the WNP during 1998-2020, TCs could not develop in the eastern WNP but rather occurred in the western WNP.In addition, from this anomalous pressure pattern, the eastern coast of China and the Korean peninsula are affected by anomalous southerlies; these steering ows affect the TC track shown in Fig. 2b, in such a way that during 1998-2020, TCs could mainly affect the Korean Peninsula from the Philippines region through the east coast of China.The tendency for TCs to occur to the west during 1998-2020 was identi ed in the longitudinal position of monsoon troughs in the two periods (represented by red lines in Figs.5b and 5c).The monsoon trough appeared only up to 125°E during 1998-2020 (Fig. 5b) but expanded eastward to approximately 140°E during 1979-1997 (Fig. 5c); the latter situation supports the development of TC to more eastern side of the WNP.
Consistent with the OLR eld shown in Fig. 3b, for the latitudinal range of 100°-140°E the meridional circulation pattern (left panel of Fig. 6a) shows the anomalous upward motions over the region where TCs mainly occur (0°-28°N), whereas the anomalous downward motions are formed in the mid-latitude zone of East Asia (30°-40°N); in contrast, for the domain of 140°-180°E, the anomalous downward motions develop strongly at 5°-20°N, while anomalous upward motions are formed at 20°-40°N (right panel of Fig. 6a).This property is also manifested in the longitude-height plot (Fig. 6c).The relative humidity anomalies (Fig. 6b) shows roughly the consistent result with other variables.These atmospheric circulations and moisture conditions are directed toward the formation of more TCs in the western region of the WNP during 1998-2020.
The TUTT is an elongated trough stretching from the mid-latitude to the tropics (red lines in Figs.7a and  b) and since it induces a large amount of VWS, tropical disturbances can be subdued.However, in the vicinity of TC center (usually to the northeast of it), TUTT can help ventilate the upper-level air so that TC can develop and/or intensify (Barry and Carlton 2001).Figures 7a and 7b show evidently that the TUTT developed westward to 155°E during 1998-2020 but contracted eastward to 165°E during 1979-1997, indicating the favorable environment for more TCs to develop over western WNP.The longitudinal location of the TUTT averaged in June-October was analyzed each year (Fig. 7c).The longitudinal location of the TUTT shows a decreasing trend, which is signi cant at the 95% con dence level.There is a clear in-phase relationship between the locations of the TC genesis and TUTT longitudes; a positive correlation of 0.62 (0.59 if the linear trends are removed) was found between the two variables.The statistical change-point analysis (Fig. 7d) shows that the largest absolute value of the t-value occurred in 1998, indicating the interdecadal variation of the TUTT longitudes.

Teleconnection Related To Sst Variation
The difference in SST between the two periods (Fig. 8a) indicates a warm anomaly over the western region of the WNP and a cold one over the eastern region.This spatial distribution of SST anomalies causes TCs during 1998-2020 to occur in the western WNP.In fact, the whole SST eld is typical of the cold PDO pattern and its index shows a decreasing trend until recently (Fig. 8b).This PDO time series has a correlation of 0.52 (0.49 if the linear trends are removed) with that of TC genesis longitudes averaged from June to October.This implies that the lower (higher) the PDO index is, the more TCs is formed in the western (eastern) region of the WNP.The statistical change-point analysis (Fig. 8c) demonstrates the largest absolute value of the t-value in 1998, indicating the interdecadal variation of the PDO index time series.
The difference in 500-hPa omega between the two periods (Fig. 9a) indicates the similar result.Over the latitudinal area of 0°-10°N, anomalous upward motions developed west of 160° E, while anomalous downward motions east of 160°E, implying the environment supportive of the formation of more TCs in the western WNP.The dipole pattern exhibits the strengthening of the Walker circulation after 1998.To examine whether Walker circulation is correlated with the location of TC genesis longitude, the average Walker circulation index during June-October is obtained by using the method of Vecci et al. ( 2006).This index shows the increasing trend until recently, but this trend is not statistically signi cant (Fig. 9b).In addition, an out-of-phase relationship between the two variables was clear (r = −0.68),implying that the stronger (weaker) the Walker circulation is, the higher the possibility is of TCs occurring on the western (eastern) side of the WNP.The Walker circulation index also produces the largest t-value at 1998, as before.

Summary
The statistical change-point analysis was applied to the TC genesis longitude averaged from June to October in each year from 1979-2020.It was found that the longitude of TC genesis has shifted to the west since 1998.The average longitude location during 1979-1997 was 136.9°E, whereas it was 132.3°E during 1998-2020, a difference of 300-400 km in distance between the two periods.The present study investigates the possible mechanisms of this shift.
The recent trend for TCs to occur in the western WNP was con rmed by many thermodynamic and dynamics variables.Here, positive precipitation, GPI, and 600-hPa relative humidity (OLR) anomalies were observed in the western (eastern) WNP.In contrast, negative precipitation, GPI, and 600-hPa relative humidity (OLR) anomalies were located east (west) of 140°E.All these indicate that convection was more active in the western area of the WNP during 1998-2020 than during 1979-1997.The VWS anomalies showed that the negative anomalies below 10°N were located further west than the positive anomalies at 10°-20°N.The 850-hPa relative vorticity anomalies also showed that positive anomalies were located farther northwest than were the negative anomalies in the WNP.
The changes in the large-scale environment between 1998-2020 and 1979-1997 as shown above, could be con rmed by the difference in the 850-hPa stream ow between the two periods.The large anomalous anticyclonic circulation existed in the WNP, while the anomalous cyclonic circulation existed in East Asia.
Owing to the huge anomalous anticyclonic circulation located in the WNP, TCs during 1998-2020 could not occur in the eastern area of the WNP but rather occurred in the western side.This was con rmed by the longitudinal location of development of monsoon troughs during the two periods; the monsoon troughs only developed up to 125°E during 1998-2020 but developed eastward to approximately 140°E during 1979-1997.
TUTTs developed more westward during 1998-2020 than during 1979-1997.The TUTT time series show the climate regime shift in 1998, and since 1998, there has been a strong tendency toward a westward shift of the TUTT.
The differences in SST between the two periods shows the warm (cold) anomaly over the western (eastern) region of the WNP, and, considering the midlatitude anomalies together, this SST anomaly eld is the characteristic pattern of cold PDO phase.The difference in the 500-hPa omega between the two periods demonstrates the strengthening of the Walker circulation, which also showed a climate regime shift in 1998 (Fig. 10).
This study showed that TC genesis longitude was shifted westward in the WNP during the cold PDO phase.The precise physical mechanisms can be veri ed using a series of sensitivity tests using coupled climate model, which will be pursued as a future work.As shown above, more TCs develop over the western area of the WNP in recent years, implying the higher risk of natural hazard along the coastal countries of East Asia, which requires increased precaution in those areas.
al. 2016; Wang and Wu 2016; Huangfu et al. 2017; Deng.et al. 2018).Huang et al. (2017) attributed the westward shift of the TUTT to the recent westward migration of the monsoon trough.Hu et al. (2018) recently reported that more TCs tended to occur in the northwestern part of the WNP and showed that this was related to a shift in convective anomalies for El Niño-Southern Oscillation (ENSO) and a climate regime shift in the Paci c Ocean in 1998.Through observational data analysis and model simulations, Cao et al. (2019) suggested that changes to the La Niña-type Paci c sea surface temperature (SST), which induced the recent global warming hiatus, are responsible for the northwest migration of the location of the WNP TC genesis.Many studies have shown that the WNP TC genesis is affected by ENSO on the interannual timescale (Gray 1984; Chan 2000; Dowdy et al. 2012; Villarini et al. 2014; Zhao and Wang 2019 found that the frequency of tropical disturbances in the southwestern Paci c increased in the eastern region of Australia during anti-El Niño years.Chia and Ropelewski (2002) and Wang and Chan (2002) showed that El Niño years have a strong tendency for WNP TCs to occur in the southeastern waters of the WNP, whereas in La Niña years, they clearly occur in the northwestern part of the WNP.It is also well known that ENSO affects the TC tracks (Kim and Seo 2016).It is well known that SST is an important environmental factor in uencing TC activity.In a study on the relationship between Paci c Decadal Oscillation (PDO) and TC, Lee et al. (2021) indicated that the frequency of TCs passing through Korea and Japan during the negative phase of PDO increased by approximately 50%, compared to those in the positive phase.Wang and Liu (2016) investigated how PDO modulates the effect of ENSO on the rapid intensi cation (RI) of TCs in the WNP.The relationship between ENSO and RI frequency was shown to be statistically signi cant in the warm PDO phase.Scoccimarro et al. (2021) showed that PDO modulates TC days in the North Paci c Ocean.In addition, many other studies have investigated the association between TC activity and PDO (Zhou et al. 2007; Kubota and Chan 2009; Liu and Chan 2008, 2013; Yang et al. 2018).So far, however, no studies have been conducted that the westward shift of the WNP TC genesis location during peak TC season (June to October) is associated with PDO.Therefore, we analyze the relationships between the westward shift of the WNP TC genesis location during peak TC season and the PDO on a decadal timescale.
series of TC genesis longitude averaged for June-October (JJASO) (red line with dot) and the statistical change-point pro le (blue line).

Figure 2 Differences
Figure 2

Figure 4 Time
Figure 4