Quasi-biennial modulation of rapidly intensifying tropical cyclones by the western North Pacific monsoon

Previous publications have highlighted the relationship between several climate modes such as El Niño–Southern Oscillation and tropical cyclones (TCs) experiencing rapid intensification (RI) over the western North Pacific (WNP), particularly on a 3–7-yr timescale. This study investigates the modulation of TCs experiencing RI by the WNP monsoon on biennial timescales. There is a significant positive relationship between rapidly intensifying TC (RITC) frequency over the WNP during July–November from 1980 to 2021 and the simultaneous WNP monsoon index. After classifying different WNP monsoon years on biennial timescales, we find significantly more TCs forming over the key region from 5°–25°N, 140°–160°E during strong WNP monsoon years. Some of these TCs then move westward into the portion of the WNP that climatologically has the most RI-favorable environmental conditions. Alternatively, other TCs forming in the key region move northward and undergo RI quickly after genesis, subsequently leading to an increase in rapidly intensifying WNP TC frequency. The WNP monsoon influences rapidly intensifying TC activity predominantly through modulation of large-scale atmospheric conditions. During strong WNP monsoon years, increased low-to-mid-level humidity, low-level vorticity and upper-level divergence and decreased vertical wind shear all favor TC genesis and RI development over the key region. A strong WNP monsoon is also associated with an anomalous 850-hPa cyclone, an anomalous 200-hPa anticyclone, increased 600-hPa moisture convergence and a decrease in the magnitude of 200-hPa winds over the key region. Our study highlights that the WNP monsoon significantly modulates TC and RITC activity at distinct timescales.


Introduction
Tropical cyclone (TC) rapid intensification (RI) is commonly defined as a pronounced intensity increase over a short duration. Forecasting RI has posed a significant challenge in improving TC intensity prediction, mainly due to an incomplete understanding of physical mechanisms responsible for RI (DeMaria et al 2021). TCs that undergo RI during their lifetimes (RITCs) typically have an increased chance to attain a greater lifetime maximum intensity (LMI) than other TCs. Lee et al (2016) showed that RITCs account for a majority (79%) of global TCs reaching major TC intensity (LMI 96 kt; Category 3 or higher on the Saffir-Simpson Hurricane Wind Scale). Owing to their generally higher intensities, RITCs can generate devastating losses for coastal regions. For instance, Super Typhoon Haiyan (2013), an RITC over the western North Pacific (WNP), caused 6300+ fatalities and ∼$2 billion US dollars in damage in the Philippines (Galvin 2014). Further investigation of RITCs is thus needed for reduction of TC-related disasters. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Given the increased focus of climate change's impacts on TCs, several studies (e.g., Wang and Zhou 2008, Fudeyasu et al 2018, Gao et al 2020, Shi et al 2020, Cai et al 2022 have investigated interannual changes in WNP RITCs. The WNP is the most TC-active basin around the globe on an annually-averaged basis (Chan 2005). It is now well accepted that El Niño-Southern Oscillation (ENSO) is the primary factor influencing the interannual variability of WNP TCs (Emanuel 2018, Lin et al 2020, Gao et al 2022. Wang and Zhou (2008) and Fudeyasu et al (2018) found that both the occurrence number and ratio of WNP RITCs were on average higher in El Niño years than in La Niña years. During El Niño, increased RITC activity has been linked to more TCs forming over the southeastern quadrant of the WNP. These storms have a greater opportunity to undergo RI since they often pass through the region that climatologically has the most RI-favorable conditions (Fudeyasu et al 2018). Shi et al (2020) reported that the occurrence proportion of WNP RITCs exhibited distinct features between different El Niño flavors. Higher RITC ratios were observed during July-October (November-December) for eastern (central) Pacific El Niño years.
Recently, several publications have noted other factors controlling WNP RITC variability on interannual timescales. Gao et al (2020) found a significant positive correlation between tropical Indian Ocean sea surface temperature (SST) anomalies and the occurrence ratio of WNP RITCs. During years with positive tropical Indian Ocean SST anomalies, both a warmer ocean and a moister mid-troposphere over the WNP enhance RITC activity, while changes in dynamical conditions played a minor role. In addition, Cai et al (2022) showed a lag relationship between January-March eastern Tibetan Plateau snow depth (TPSD) and WNP RITC frequency during the following July-November. Increased TPSD induced a low-level anomalous anticyclone over the subtropical eastern North Pacific through modulation of the prevailing trans-Pacific westerly jet, subsequently resulting in a negative Pacific meridional mode-like SST feature. This SST pattern further triggered a low-level anomalous anticyclonic circulation over the WNP by a Gill-type Rossby response, dynamically suppressing RITC activity.
Note that all of the aforementioned factors influencing WNP RITCs typically have a timescale of longer than two years. Various ENSO indices exhibit an oscillation with a 3-7-yr preferred timescale (Wang et al 2016), while the basinwide Indian Ocean SST index shows a preferred periodicity of ∼4-yr (Taschetto and Ambrizzi 2012). On the other hand, there is quasi-biennial variability in the tropical troposphere, which is not fully associated with interannual changes in ENSO (Meehl 1987, Meehl and Arblaster 2002, Meehl et al 2003. This quasibiennial signal has been observed over the Asia-Pacific monsoon region and is typically represented by a relatively strong monsoon followed by a relatively weak monsoon and vice versa (Meehl andArblaster 2002, Meehl et al 2003).
Part of the Asia-Pacific monsoon system is the WNP monsoon over the WNP TC-active region. The WNP monsoon is characterized by a northwest-southeast oriented monsoon trough with intense precipitation, lowlevel southwesterlies and upper-tropospheric easterlies (Tao andChen 1987, Murakami andMatsumoto 1994). Previous studies have found that more than 70% of WNP TC genesis events were linked to the monsoon trough, which favors TC development through providing an environment of large low-level cyclonic vorticity (Briegel and Frank 1997, Ritchie and Holland 1999, Chen et al 2004. On interannual timescales, Wu et al (2012) found a strong linkage between WNP TC activity and monsoon trough location. More (fewer) TCs formed over the southeastern quadrant of the WNP during years when the monsoon trough extended eastward (retreated westward). Moreover, Choi et al (2016) showed that on average, WNP TCs attained higher (lower) intensities and tended to have recurving (straight-moving) tracks in strong (weak) monsoon years.
As summarized above, the WNP monsoon has been extensively studied for its impact on TC genesis. However, few studies have investigated the influence of the WNP monsoon on interannual changes related to WNP RITCs. Furthermore, WNP monsoon variability displays two preferred timescales: one on the lowfrequency ENSO timescale (∼4 years) and the other on a quasi-biennial timescale (Li and Wang 2005). Here we investigate the potential timescale(s) linking WNP monsoon variability and WNP RITCs.
The remainder of the manuscript is organized as follows. Section 2 introduces the TC and environmental datasets and the analysis methods. Sections 3 and 4 discuss the temporal and spatial relationship between the WNP monsoon and RITC activity, respectively. Section 5 describes how the WNP monsoon modulates environmental changes. Section 6 concludes and summarizes.

Data and methods
This study examines TCs forming over the WNP during 1980-2021 that reached an LMI of tropical storm intensity or higher (1-min maximum sustained wind speed 34 kt), as derived from the Joint Typhoon Warning Center best track data as compiled in the International Best Track Archive for Climate Stewardship (IBTrACS) (v04r00, Knapp et al 2010). TC genesis is identified as the first instance for each TC where the intensity is at least 34 kt. TCs that form over the eastern North Pacific and then enter the WNP are excluded, consistent with Fudeyasu et al (2018). TC best track data from 1980-2021 are usually considered to be of higher quality given the near-global coverage of geostationary satellites (Daloz and Camargo 2018). Similar to previous RI-related publications (e.g., Kaplan and DeMaria 2003, Kaplan et al 2010, Shu et al 2012, Knaff et al 2018, an RI event is defined as a 24-h intensity increase of at least 30 kt over water, while an RITC is a TC experiencing at least one RI event during its lifetime. A higher threshold (e.g. 40 kt or 45 kt) was applied by Li et al (2022) to define RI over different basins. This definition provided a more physically-based threshold and described a more reliable representation of extreme intensification events (Li et al 2022). The results that we display are almost unchanged if a 45-kt threshold is used to identify WNP RI events (see supplementary figures 1 and 2 for further information). We investigate RITC activity during July-November. These five months include a majority (84%) of WNP RI events occurring over the entire year ( Following Wang and Fan (1999), the annual WNP monsoon index (WNPMI) is calculated as: where U850 denotes the ERA5 areal-averaged zonal wind during July-November at 850 hPa. Choi et al (2016) defined the ten years with the highest (lowest) values as the positive (negative) WNPMI phase and then discussed the WNP monsoon's effect on interannual changes in WNP TC activity. Their study did not focus on a specific period but on a relatively wider range of periods. To highlight the monsoon-RITC relationship on biennial timescales, we employ the classification method proposed by Meehl and Arblaster (2002) and Meehl et al (2003). This classification method has been applied to analyze the influence of the Tropospheric Biennial Oscillation (TBO). A strong monsoon during year i is defined as: while a weak monsoon year is defined as: Table 1 displays the 13 strong monsoon years and the 14 weak monsoon years identified by equations (2) and (3).
The significance levels (p) of correlation coefficients (r) between the two series and the differences in means of the samples between strong and weak monsoon years are both estimated using a two-tailed Student's t-test. In evaluating statistical significance, the effective sample size proposed by Trenberth (1984) is used to minimize the influence of autocorrelation. Although TC and RITC frequencies are highly correlated (r = 0.54, p < 0.01), their spectra peak at different periods ( figure 1(b), (c)). A period of 3-6 years dominate changes in TC frequency ( figure 1(b)), which corresponds to the preferred timescale of ENSO. By comparison, RITC frequency exhibits a significant periodicity from 2-3 years ( figure 1(c)), which is relatively shorter than the dominant period for TC frequency. Figure 1(a) also displays the annual change of the WNPMI, which significantly correlates with both TC frequency (r = 0.32, p = 0.04) and RITC frequency (r = 0.42, p < 0.01). Choi et al (2016) reported a much higher correlation between the WNPMI and TC frequency (r = 0.62, p < 0.01), which possibly is a result of the different season analyzed (June-August). As shown in figure 1(d), similar to Li and Wang (2005), the WNPMI shows two preferred periods of 5-6 years and 2-3 years, which roughly correspond to the dominant periods of TC frequency and RITC frequency, respectively. Furthermore, TC frequency and the WNPMI show significant coherence only on a timescale of 5-8 years ( figure 1(e)). Comparatively, significant coherence between RITC frequency and the WNPMI is observed on a shorter timescale of 2-3 years. This means that although the WNP monsoon notably influences annual changes in both TC and RITC frequencies, its modulation peaks on distinct timescales. When focusing on biennial timescales, as displayed in table 1, the difference in the average TC frequency between strong and weak WNP monsoon years (2.6 TCs) is not statistically significant (p = 0.08). Accordingly, we suspect that the significant WNPMI-TC relationship found in Choi et al (2016) is primarily due to its relationship on longer than biennial timescales. By contrast, the average RITC frequency in strong WNP monsoon years is 2.7 TCs greater than that in weak WNP monsoon years. This difference is statistically significant (p < 0.01), confirming a notable WNP monsoon-RITC frequency linkage on the biennial timescale. Note that the TC frequency difference is comparable to the RITC frequency difference, while the differences in significance are mainly the result of the standard deviation of TC frequency (3.9 TCs) being much larger than that of RITC frequency (2.4 TCs).

Spatial relationship
Figure 2 displays differences in several TC-related metrics between the 13 strong WNP monsoon years and 14 weak WNP monsoon years. Choi et al (2016) reported that more TCs tended to form over the southeast (northwest) quadrant of the WNP during June-August during the positive (negative) phase of the WNPMI. By comparison, during July-November, the TC genesis difference between strong and weak WNP monsoon years exhibits an east-west dipolar pattern, with a boundary at around 135°E (figure 2(a)). During strong WNP monsoon years, unlike the dipolar pattern of June-August TC passage frequency differences shown in Choi et al (2016), there are higher TC track densities over almost the entire WNP during July-November ( figure 2(b)). Because of enhanced TC genesis far away from the East Asian continent, more TCs experience recurving or northward-moving tracks. By contrast, there are only weak changes in TC track density east of 130°E, likely attributed to the offsetting of the following two factors. One is decreased TC genesis over the western WNP, which directly leads to suppressed TC activity over this region. The other is more TCs forming over the eastern WNP. Some of these TCs move straight westward or northwestward and subsequently enter the western WNP.
The spatial features of genesis (track density) differences for RITCs between strong and weak WNP monsoon years are quite similar to that for all TCs, with a pattern correlation coefficient of 0.86 (0.95) (figures 2(c), (d)). Enhanced TC activity is generally associated with enhanced RITC activity, since all TCs and RITCs share several common favorable environmental conditions (Wang and Zhou 2008). During strong WNP monsoon years, there are significantly more RITCs forming over the region of 5°-25°N, 140°-160°E (hereafter the key region), which only covers the eastern edge of the RI main development region (MDR; 10°-20°N, 125°-145°E; Song et al 2021) (figure 2(c)). During strong WNP monsoon years, more RITCs move westward across the MDR (figure 2(d)), where the climatological environment is most favorable for RI. Compared with TCs, RITCs take more westward-moving tracks and fewer northward-moving tracks during strong WNP monsoon years. TCs entering the MDR are more likely to undergo RI, while TCs entering higher-latitude regions are less likely to undergo RI. Moreover, during strong WNP monsoon years, there are notably more RI occurrences and greater RI ratios over the key region (figures 2(e), (f)). Note that the region with significantly enhanced RI occurrence is a little farther north of the region with significantly enhanced RITC genesis. These findings imply that some northward-moving RITCs experience RI quickly after genesis. Figure 2 summarizes how RITC activity is influenced by the WNP monsoon. During strong WNP monsoon years, more TCs tend to form east of 135°E, particularly over the key region. Some of these TCs move westward into the MDR that climatologically has the most RI-favorable environmental conditions, These TCs subsequently have a great chance to experience RI. Other TCs undergo RI over the key region quickly after genesis, while moving northward towards higher latitudes. Both of these types of TCs can lead to an increase in RITC frequency during strong WNP monsoon years. Figure 3 shows differences in environmental conditions during July-November between strong and weak WNP monsoon years. Similar spatial structures are observed in differences of thermodynamic variables (e.g., SST, TCHP and 700-500-hPa relative humidity) (figures 3(a)-(c)). During strong WNP monsoon years, there are warmer waters, larger TCHPs and a moister atmosphere over the tropical central Pacific, while cooler waters, lower TCHPs and a drier atmosphere extend from the South China Sea (SCS) to the subtropical central Pacific. There are almost no significant SST and TCHP changes over the entire WNP, including over the key region (figures 3(a), (b)). However, significantly increased relative humidity is found in the zonal belt of 10°-20°N and east of 145°E ( figure 3(c)).

Environmental changes
Differences in 850-hPa relative vorticity mainly show latitudinal differences ( figure 3(d)). During strong WNP monsoon years, there is significantly increased relative vorticity between 10°-30°N and significantly decreased relative vorticity south of 10°N and north of 30°N. The difference in 200-hPa divergence has a similar spatial pattern to the vorticity difference (figure 3(e)), with a pattern correlation of 0.56 (p < 0.01). Over the region east of 120°E, there is increased (decreased) upper-level divergence south (north) of 25°N, roughly corresponding to increased (decreased) low-level vorticity. In addition, increased 850-200-hPa vertical wind shear is concentrated over the equatorial western WNP, while reduced vertical wind shear extends from the tropical central Pacific to the subtropical western WNP ( figure 3(f)).
In summary, over the key region, SST and TCHP changes likely play only a minor role in modulating RITC activity between different WNP monsoon years (figures 3(a), (b)). In contrast, changes in atmospheric conditions primarily act to favor TC genesis and RI occurrence during strong WNP monsoon years compared with weak WNP monsoon years. These more favorable conditions during strong WNP monsoon years include increases in low-to-mid-level humidity, low-level vorticity and upper-level divergence and decreases in vertical wind shear (figures 3(c)-(f)). We note that when considering the biennial influence of the WNP monsoon, TC genesis and RI occurrence share similar preferred environments. Figure 4 further shows how the WNP monsoon modulates changes in the aforementioned atmospheric conditions. Similar to Li and Wang (2005), a strong WNP monsoon is characterized by enhanced precipitation over the tropical WNP, with a suppressed convective region extending from Borneo to the North Indian Ocean ( figure 4(a)). The only discrepancy between our study and Li and Wang (2005)'s study is over the SCS, where increased and decreased rainfall was observed in Li and Wang (2005) and our study, respectively. This discrepancy may be due to the different seasons considered. Meanwhile, there is reduced precipitation over the mid-latitude regions of East Asia, indicating an opposite change of the WNP monsoon and the East Asian monsoon on biennial timescales. Consistent with the precipitation difference over the WNP, during strong WNP monsoon years, there is an anomalous low-level cyclonic (anticyclonic) circulation south (north) of 30°N ( figure 4(b)). This anomalous cyclone is centered over the key region, providing positive anomalous vorticity that is favorable for TC genesis and RI development. At 600 hPa, there are two anomalous cyclonic circulations over the tropical WNP, with the eastern part of the key region located between them ( figure 4(c)). Over the eastern key region, there is moisture convergence due to northerlies to the east and southerlies to the west, leading to an increase in low-to-mid-level humidity. The anomalous circulation over the WNP at 200 hPa is nearly opposite to that at 850 hPa, exhibiting an anomalous anticyclone (cyclone) south (north) of 30°N ( figure 4(d)). The key region is located near the center of the anticyclone, with associated upper-level divergence based on quasi-geostrophic theory. In addition, over the key region, anomalous westerlies (easterlies) oppose climatological easterlies (westerlies) at 850 hPa (200 hPa), indicating a slowdown of the environmental wind speed (figures 4(b), (d)). Given that the areal-averaged speed of the anomalous low-level westerlies (1.2 m s −1 ) is much lower than that of the anomalous upper-level easterlies (3.3 m s −1 ), the result is reduced vertical wind shear over the key region.

Conclusion
The modulation of WNP RITC activity by the WNP monsoon on biennial timescales is investigated in this study.
There is a significant positive relationship between WNP RITC frequency during July-November and the simultaneous WNPMI, which primarily results from their correlation on biennial timescales. During strong monsoon years, there is enhanced TC genesis over the key region (5°-25°N, 140°-160°E). Some of these TCs move westward into the MDR which climatologically has the most RI-favorable environmental conditions. These TCs subsequently have a relatively high chance of undergoing RI. Other TCs undergo RI over the key region quickly after their genesis, while moving northward to higher latitudes. These both result in an increase in RITC frequency during strong WNP monsoon years. The influence of the WNP monsoon on RITC activity can be explained by changes in large-scale environmental variables. In general, during strong WNP monsoon years, there is increased low-to-mid-level humidity, low-level vorticity and upper-level divergence and decreased vertical wind shear over the key region, which all favor TC genesis and RI development. Given the small observed changes in SST and TCHP, we find that atmospheric variable changes play a dominant role in influencing RITC activity over the WNP. Further analysis shows that a strong WNP monsoon is characterized by an anomalous 850-hPa cyclone and an anomalous 200-hPa anticyclone over the tropical WNP, providing positive vorticity and divergence at lower and upper levels, respectively. Over the key region, 600-hPa moisture convergence increases low-to-mid-level relative humidity, while 200-hPa wind has a larger decrease in magnitude than 850-hPa wind, reducing vertical wind shear. Li and Wang (2005) concluded that the interannual variation of the WNP monsoon displayed two preferred timescales: one at low-frequency ENSO timescales (∼4 years) and the other at quasi-biennial timescales. Our study highlights that the WNP monsoon significantly modulates TC and RITC frequency at distinct timescales. Our results are mainly obtained by statistical composite analysis. These results should be verified using numerical sensitivity tests. We intend to conduct these simulations in future work.