Modulation of East Asian atmospheric rivers by the Pacific-Japan teleconnection pattern

Despite growing recognition that atmospheric rivers (ARs) play an important role in summer monsoon rainfall in East Asia, the AR variability related to low-frequency (LF) climate modes remains largely unknown. This study reports the significant control of the Pacific-Japan (PJ) teleconnection pattern on East Asian summer ARs. While Korea and central to eastern Japan experience more frequent, intense, and persistent ARs during the positive PJ, ARs in eastern China become more active during the negative PJ. Such AR activity changes are closely related to the PJ modulation of the western North Pacific subtropical high along which anomalous moisture transport organizes on the LF time scale. This finding suggests that the PJ pattern is an important source of East Asian summer AR variability and needs to be considered in medium-range forecasts of AR-related hydrological extremes.


Introduction
East Asia is increasingly noticed as one of the regions most susceptible to AR impacts.East Asian ARs are observed throughout the year (Kamae et al 2017a, 2017b), but they are most pronounced in the summer monsoon period (Mundhenk et al 2016, Pan and Lu 2020, Liang and Yong 2021), explaining about 35%-70% of East Asian summer monsoon rainfall (Kim et al 2021, Park et al 2021, Kwon et al 2022).The seasonal variation of East Asian ARs is closely related to the East Asian summer monsoon, which is accompanied by the northwestward expansion of the western North Pacific subtropical high (WNPSH).When the WNPSH extends to the Asian continent (figures 1(a), (d)), AR formation is facilitated by through enhanced monsoon southwesterlies in summer (Pan andLu 2020, Park et al 2021).East Asian ARs are particularly active in the early monsoon period (June-July) when the WNPSH extends westward to result in an enhanced pressure gradient along its northern boundary (Park et al 2021).Although the frequency is much reduced, ARs are still frequently observed in the late monsoon period (August-September) along the northern flank of the WNPSH.
East Asian ARs also exhibit significant lowfrequency (LF) variabilities in response to the climate variability such as the El Niño Southern Oscillation.This study examines the PJ control of East Asian summer ARs using the long-term AR catalog and PJ index (section 2).In section 3, the frequency of ARs and AR-related precipitation are contrasted between the positive and negative PJ phases.The hydrological impacts of ARs in the two PJ phases are then quantified by considering the hazardous and beneficial ARs.
The key process behind the modulation of East Asian ARs by the PJ is analyzed by decomposing the moisture transport into high-frequency (HF) and LF components.The results are summarized with a discussion in section 4.

Dataset
The fifth-generation European Centre for Mediumrange Weather Forecasts atmospheric reanalysis (ERA5; Hersbach et al 2020) data is used.The 6hourly data are used in 37 pressure levels with a horizontal resolution of 1.5 • × 1.5 • for the period of 1979-2020.The daily anomaly is then defined by subtracting the calendar-day climatology from the calendar-day values.The daily precipitation data are obtained from the gauge-based precipitation analysis of the Climate Prediction Center (CPC; Xie et al 2007) at a 0.5 • × 0.5 • resolution.As the PJ pattern is clearly defined only in the summer (Nitta 1987, Kosaka andNakamura 2006), all analyses are focused on the summer months of June-August (JJA).
For AR analyses, we use the 42 year AR catalog produced by Park et al (2023) who applied the modified AR detection algorithm of Guan and Waliser (2015) to ERA5 data.The algorithm uses the integrated water vapor transport (IVT) anomaly (IVTA).Here, the IVT is computed as where i and j are the zonal and meridional unit vectors; u and v are the zonal and meridional winds; q is the specific humidity; g is the gravitational acceleration.The algorithm searches for an object as a collection of grid points with IVTA > 150 kg m −1 s −1 .The geometric conditions (length > 2000 km and lengthto-width ratio > 2) are then considered to filter out non-filamentary objects.An object trapped in the deep tropics without a considerable amount of poleward moisture transport is also discarded by examining the location of the AR centroid and the objectmean IVT.

PJ index
The PJ pattern is extracted based on the point-based PJ index (Kubota et al 2016).The index is defined as the difference of the normalized 850 hPa geopotential height (GPH) anomaly (GPHA) between Taiwan (22 • N, 120 • E) and Japan (35 where the asterisk indicates the normalization by the one standard deviation at each location.To eliminate short-term fluctuations, a 10 day Lanczos lowpass filter is applied to the index with 101 weights (Duchon 1979).The positive and negative PJ phases are defined as a period of successive PJ index greater than its ±1.0 standard deviation in magnitude.To ensure the independence between each PJ phase, each PJ phase (either positive or negative) is separated by at least one day.
The positive PJ pattern is characterized by a southwest-northeast dipole of anomalous cyclone over the Philippine Sea and anomalous anticyclone over Japan (figure 1(e)).The negative PJ pattern exhibits the mirrored anticyclone-cyclone dipole (figure 1(f)).On average, the positive and negative PJ patterns last for 5 and 4.1 d, respectively (figure S1), although their time scales vary widely from a few days to over two weeks (figure S2).The two phases are evenly distributed during JJA (figure S3).

Results
The AR frequency difference for the positive and negative PJ phases is first compared with the climatology (figures 1(d)-(f)).Note that the term 'difference' indicates the deviation from the JJA mean by distinguishing it from the daily anomaly.The AR frequency is defined as the percentage ratio of the number of AR-existing time steps to the number of all time steps at each grid point.The southwesterly IVT is concentrated along the boundary of the WNPSH (figure 1(a); see bold contour for the WNPSH boundary).In this region of strong IVT, the AR activity is pronounced with a maximum AR frequency of ∼ 20% (figure 1(d)).Such spatial coherence of AR activity and monsoon southwesterly is well documented in the literature (Pan and Lu 2020, Park et al 2021).
The AR frequency changes substantially when the PJ pattern develops (figures 1(b), (e) and 1(c), (f)).During the positive PJ phase, the northern boundary of the WNPSH shifts northward with a well-defined ridge around the Korean Peninsula and a monsoon trough over the Philippine Sea (figure 1(b)).It directly results from the cyclone-anticyclone anomaly dipole in the southwest-northeast direction (figure 1(e)).In accordance to this circulation change, AR activities are enhanced over the Korea, eastern Japan and the Philippine Sea, but are suppressed in southeastern China and the western Pacific south of Japan.During the negative PJ phase, the WNPSH retreats southward but extends further west into southern China (figure 1(c)), as the circulation anomalies are reversed from those of the positive PJ phase (figure 1(f)).The AR activities also exhibit a mirrored change from that in the positive PJ phase.
Figure 2 presents the AR precipitation and its ratio to the total precipitation.Here, AR precipitation is defined as the daily precipitation at a grid point where AR is identified for at least 12 h for the day (i.e. two time steps; not necessarily consecutive).To consider the spatial coincidence, the AR frequency data are bilinearly interpolated onto the observational (CPC) grids.Climatologically, the AR precipitation is widely distributed along the Meiyu-Changma-Baiu region (figure 2(a)) with a local maximum (∼8 mm day −1 ) in western Japan.ARs account for 30%-50% of summer precipitation in eastern China and over 60% in Korea and western Japan (figure 2(d)), consistent with the previous quantification (Park et al 2021).
During the positive PJ phase, AR precipitation increases in Korea by up to 3 mm day −1 but decreases over inland eastern China and Japanese main islands with the largest decrease of 1-2 mm day −1 over the central Yangtze River basin (figure 2(b)).The marginal increase in precipitation is also observed in the southeast coast of China, and southwest island of Japan.The mirrored AR precipitation change is evident during the negative PJ phase (figure 2(c)).AR precipitation increases in eastern China, and western Japan, but decreases in Korea.The variation of the AR precipitation according to the PJ phase closely resembles that of AR frequency shown in figures 1(e) and (f).When analyzing the difference in AR precipitation ratio from its climatology, it is found that the AR precipitation ratio increases in the regions where AR precipitation increases while it decreases in the regions where AR precipitation decreases (figures 2(e) and (f)).This implies that the PJ pattern alters the relative importance of ARs in the regional hydroclimate over East Asia.One exception is the Japanese main islands.During the positive PJ phase, while the total AR precipitation decreases slightly in the main islands of Japan, the AR precipitation ratio increases significantly.Conversely, during the negative PJ phase, the AR precipitation does not change much, but its ratio decreases.This is because the total precipitation in the Japanese main islands varies during each PJ phase with minor changes in the AR precipitation.
It is important to note that not all ARs bring hydrological extremes.Their impacts can be either beneficial or hazardous depending on their intensity and local persistency (e.g.Ralph et al 2013).The potential hazard of ARs can be estimated by the AR scale defined with the maximum IVT and duration of an AR event (Ralph et al 2019).The AR scale has five categories from a weak and short-living AR (Category 1: Cat1) to a strong and persistent AR (Category 5: Cat5) (see the diagram in figure 3).This scaling was originally designed for landfalling ARs in the US west coast, but has been also applied to western Europe ARs and global ARs with adjustments of their impacts (Eiras-Barca et al 2021, Guan et al 2023).Kwon and Son (2024) recently showed that the AR scale can be applicable to East Asian ARs to better understand ARrelated monsoon precipitation.
Using the AR scale, we examine if the potential hazard of AR events vary according to the PJ phase.
Here, an AR event is locally defined at each grid point as consecutive time steps during which an AR is observed.A 6 h break is allowed in counting AR events to consider the flickering cases, i.e.ARs temporarily drop below the detecting threshold for a short period (e.g. a few time steps) in the gridded AR data (Reid et al 2022).During East Asian summer, the Cat5 events are concentrated along the WNPSH boundary, while the no-rank events are mostly observed inland (figure S4).In this study, we pay attention to region A, which includes the Korean peninsula and central to eastern Japan, and region B, which includes eastern China, as these are the regions with pronounced responses of AR frequency and precipitation to PJ patterns (see figures 1 and 2).
In the region A, the fraction of the high-rank (Cat3-5) AR events are larger for the positive PJ phase than that for the negative PJ phase (cf red and blue lines in figure 3(a)).Compared to the JJA climatology, the fraction of the Cat5 AR events exhibits a statistically significant increase during the positive PJ phase and decrease during the negative PJ phase, with a more pronounced change for the negative PJ phase (cf red and blue bars in figure 3(a)).Note that the Cat5 events account for over 15% of total AR events for the positive PJ phase but only ∼7% in the negative PJ phase.This is because both the maximum IVT and AR duration slightly increase in the positive PJ phase but substantially decrease in the negative PJ phase (figure 3(c)).
For region B, ARs are more hazardous during the negative PJ phase than the positive PJ phase, but their difference from the climatology is quite subtle (figure 3(b)).During the negative PJ phase, the Cat4 fraction shows a statistically significant increase from the climatology, but the change is insignificant during the positive PJ phase.This result can be explained by the compensating change of the maximum IVT and AR duration (figure 3(d)).During the positive PJ phase, the maximum IVT decreases while the duration increases.During the negative PJ phase, they exhibit opposite changes with much smaller magnitudes.
These results indicate that ARs in region A become more hazardous during the positive PJ phase, whereas AR-related hydrological extremes slightly increase in region B during the negative PJ phase.Combining the changes in AR frequency (figures 1(e), (f)) and scale (figure 3), the PJ pattern modulates both the AR frequency and its hazard in East Asia.
Although the AR response to the PJ pattern is well synchronized with the WNPSH variability (see figure 1), it does not fully explain the underlying mechanisms because East Asian summer ARs are not only driven by quasi-stationary monsoon southwesterly along the WNPSH boundary but also influenced by migrating extratropical cyclones (Park et al 2021).Indeed, Park et al (2023) have recently shown that ARs in the western North Pacific are highly multiscale in boreal summer.The composite analyses reveal that the ARs in region A and region B are organized by both HF IVTA (IVTA HF ) controlled by zonallyaligned transient disturbances and LF IVTA (IVTA LF ) developing between the quasi-stationary cyclone and the anomalous WNPSH (figure S5).This suggests that the PJ pattern may modulate the AR activity not only by affecting the strength and spatial extent of the WNPSH but also by controlling the migrant synoptic disturbances.
To identify the relative role of the LF and HF processes, IVTA HF and IVTA LF are compared between the two PJ phases (figures 4(a)-(d)).The IVTA HF change is not conspicuous (see shading in figures 4(a) and (b), possibly due to the self-smoothing of its In contrast, IVTA LF is noticeably positive in region A during the positive PJ phase and in region B during the negative PJ phase (see shading in figures 4(c), (d)).This is consistent with the AR frequency change following the PJ phase (figures 1(e), (f)).The frequency of IVTA LF > 150 kg m −1 s −1 also exhibits coherent changes (see contours in figures 4(c), (d)).This indicates that AR frequency change is primarily attributed to the variation of WNPSH in direct response to the PJ pattern.

Summary and discussion
This study examines how East Asian ARs are modulated by the PJ pattern, one of the dominant teleconnection patterns in East Asia in boreal summer.The positive PJ phase is characterized by a southwestnortheast dipole of anomalous cyclone over the Philippine Sea and anomalous anticyclone over Japan, resulting in the northward shift of the northern boundary of the WNPSH.Corresponding to this circulation change, the AR frequency is enhanced over Korea, eastern Japan and the Philippine Sea but suppressed in southeastern China and the Pacific open ocean.During the negative PJ phase, the WNPSH retreats southward but extends westward, and the AR frequency shows opposite changes from the positive PJ phase.The AR precipitation also changes consistently with the AR frequency change.In particular, the AR precipitation ratio increases in the regions where AR precipitation increases while it decreases in the regions where AR precipitation decreases.This suggests that the PJ pattern strongly modulates the East Asian summer ARs and the related precipitation.The overall results are summarized in table 1.
The AR scale analysis is applied to region A, which includes the Korean peninsula and central to eastern Japan, and region B, which includes eastern China, as the AR activity in these regions varies strongly with the PJ pattern.The results reveal that the potential hazards of individual ARs are also altered by the PJ pattern (table 1).During the positive PJ phase, the ARs in region A become more hazardous with increasing the highest-rank (i.e.Cat5) AR events.For the negative PJ phase, both the AR intensity and persistency decrease greatly, resulting in reduced highrank (Cat3-5) AR events.In region B, ARs become more hazardous during the negative PJ phase, however, the difference from the JJA climatology is much ) and low-frequency (IVTA LF ) components, it is further found that the LF processes play a key role in shaping the effects of the PJ pattern on the AR activity.The IVTA LF is significantly positive in region A during the positive PJ phase and in region B during the negative PJ phase, consistently with the AR frequency change.This implies that the LF modulation, in direct response to the PJ pattern, which is closely related with the WNPSH variability, primarily explains the PJ modulation of the East Asian summer ARs.
This study adopts an AR detection algorithm with an absolute threshold of IVTA = 150 kg m −1 s −1 .To test the robustness of the result, a more restrictive absolute threshold of IVT = 500 kg m −1 s −1 (Reid et al 2020), and a relative threshold of the 85thpercentile IVT (Guan and Waliser 2015) are further considered.Both algorithms capture the distribution of AR frequency and its response to the PJ pattern (figure S8).Moreover, AR precipitation changes to the PJ pattern are well reproduced, although the precipitation values are different in magnitude (figure S9).Essentially the same results are also found when using the EOF-based PJ index is used instead of the pointbased PJ index (figure S10).These results indicate that the East Asian summer AR responses to the PJ pattern are robust regardless of the AR detection algorithms and PJ indices used.
Several studies have recently evaluated the prediction skill of ARs on various timescales (Wick et al 2013, Nayak et al 2014, DeFlorio et al 2018, 2019).For instance, it is shown that the certain combination of the Madden-Julian oscillation and the quasibiennial oscillation phases can provide the empirical prediction of AR in subseasonal time scale (Baggett et al 2017, Mundhenk et al 2018).In this regard, the PJ pattern could be the potential source of AR predictability.Further study is desired to evaluate the prediction skill of the East Asian summer ARs and associated precipitation under the varying PJ phase.
Atmospheric rivers (ARs) are filamentary objects of intense water vapor transport which play a critical role in global hydroclimate and local precipitation extremes.Early AR studies are mostly focused on the cold-season ARs affecting the western coasts of North America (Bao et al 2006, Ralph et al 2006, Neiman et al 2008, Dettinger et al 2011, Ralph and Dettinger 2012).However, recent studies show that ARs are ubiquitous across the globe including the Asian monsoon regions (Dhana Laskhmi and Satyanarayana 2020, Liang and Yong 2021, Park et al 2021), Southern Hemisphere (Viale et al 2018, Ye et al 2020), and even in high latitudes (Nash et al 2018, Wille et al 2019) and continental interiors (Nayak and Villarini 2017, Akbary et al 2019).

Figure 3 .
Figure 3. (a), (b) Fraction of AR events for five different AR scales (see the right diagram) during positive PJ (red line) and negative PJ (blue line) phases over the grid points in (a) region A, and (b) region B (boxed regions in figures 1(e), (f)).Filled circles indicate statistically significant differences between the two PJ phases based on a two-tailed student's t test at the 95% confidence level.Pink and skyblue bars represent the difference in fraction (%) from JJA climatology for positive PJ phase, and negative PJ phase, respectively.Hatched bars indicate statistically significant differences between the JJA and each PJ phase based on a two-tailed student's t test at the 95% confidence level.(c), (d) Differences in maximum IVT (left y-axis), and AR duration (right y-axis) for AR events from JJA climatology in the (c) region A, and (d) region B for positive PJ (red box), and negative PJ (blue box) phases.The error bars indicate the 95% confidence interval of the difference in maximum IVT or AR duration between each PJ phase.

Figure 4 .
Figure 4. (a), (b) IVTAHF (kg m −1 s −1 , shading) for the (a) positive and (b) negative PJ phases.Difference in the frequency of IVTAHF > 150 kg m −1 s −1 from JJA climatology (%, contours) is overlayed contoured.Only the statistically significant values are shown based on a two-tailed student's t test at the 95% confidence level.(c), (d) Same as in (a), (b), but for IVTALF.
Such results indicate that AR prediction skill can be improved by considering slowly varying climate modes.However, these studies mainly focus on the western United States or the eastern Pacific region.No study has evaluated the prediction skill of East Asian ARs.Since more than 50% of East Asian summer precipitation is associated with ARs (Kim et al 2021, Park et al 2021, Kwon et al 2022), improved prediction of ARs may lead to improved prediction of East Asian summer precipitation which suffers from low prediction skill (Wang et al 2009, Gong et al 2016, Liang and Lin 2018).

Table 1 .
Differences of AR frequency (%), AR precipitation ratio (%), and fraction of AR events for five different AR scales (%) from JJA climatology averaged in the region A, and region B (boxed regions in figures 1(e), (f)) for each PJ phase.The asterisk indicates the statistically significant difference based on a two-tailed student's t test at the 95% confidence level.