Impact of Pacific–Japan pattern on temperature and heatwave events in summer over Taiwan

The prediction of heatwave events is vital to provide information on possible hazards caused by extremely high temperature. The Pacific–Japan (PJ) pattern is a significant climate variability that controls the summer climate over East Asia. Previous studies have reported that the meridional propagating wave structure associated with the PJ pattern can modulate summer temperature and heatwave characteristics over northeast Asian countries, such as Korea and Japan. Nevertheless, its impacts on the subtropical regions, including Taiwan, a mountainous subtropical island with an approximate population of 24 million, is not yet fully understood. In this study, we investigate the impacts of the PJ pattern on the summer temperature and heatwave characteristics in Taiwan on the interannual timescale using long‐term station‐based temperature and rainfall dataset. We found an island‐wide increase in temperature during the positive phase of the PJ pattern, which is characterized by an anticyclonic anomalous circulation over Taiwan. Meanwhile, the averaged effective area and frequency of heatwave extremes over Taiwan have increased significantly, indicating an increase of exposure to heatwaves. Our examinations of large‐scale environment suggest that the adiabatic warming due to subsidence anomaly associated with the anticyclone over Taiwan is primary contributor of surface warming, which is balanced by diabatic cooling and meridional advective cooling on a seasonal timescale. Our study consolidated the links between the PJ pattern as a climate driver, hot summers and increased summer heatwave events over Taiwan. By combining with the seasonal prediction systems and early warning systems, such understanding can help sectors that are vulnerable to heat stress to prepare for potential hazards.


| INTRODUCTION
The summer climate of East Asia is heavily influenced by changes in large-scale circulations, such as the position and strength of the subtropical high (Choi & Kim, 2019;Wang et al., 2015Wang et al., , 2019;;Wang & Wang, 2018) and the climate variability across a wide range of time scales, from El Niño-Southern Oscillation (Chen et al., 2020) to boreal summer intraseasonal oscillation (Hsu et al., 2017).However, one of the most dominant interannual climate variability patterns in this region is the Pacific-Japan (PJ) pattern, which is characterized by a meridionally propagating Rossby wave train that travels from the Tropics towards the midlatitude regions of Northeast Asia, affecting the East Asian summer climate through meridional-aligned anomalous wave circulations (Kosaka & Nakamura, 2006;Kubota et al., 2016;Nitta, 1987).This pattern is identified as an intrinsic mode generated by internal atmospheric processes (Hirota & Takahashi, 2012).Hirota and Takahashi (2012) also demonstrate that moist processes intensify PJ circulation anomalies in the lower tropospheric circulation, facilitating the coupling between the northern and southern cells, which is pivotal for shaping the latitudinal alignment of the tri-cell structure.Spatially, the positive (negative) phase of the pattern is characterized by more (less) rainfall in central eastern China, Japan and South Korea, and less (more) rainfall in northern and southern China.Hsu and Lin (2007) reported an asymmetry of positive and negative phases of PJ pattern, with the positive phase having a stronger tropical connection, while the negative phase has a stronger extratropical connection.So the local impacts of positive and negative phases can be very distinct due to the asymmetry of circulation anomaly and needs to be further investigated.
Compared with rainfall, much less attention has been paid to the PJ's impacts on extreme heatwave events.The unusually persistent heatwave event occurred in early August 2015 is attributed to teleconnection of PJ-like circulation anomaly that is excited by tropical cyclone activities over the Northwest Pacific (Takahashi et al., 2016).In 2018, the anomalous circulation pattern similar to the positive phase of the PJ pattern induces widespread heat wave events over Northeast Asia, with an unusually northward shift in the subtropical high and an enhanced monsoon trough in the western North Pacific (Hong et al., 2021;Tseng et al., 2020;Wang et al., 2019).Recent study by Noh et al. (2021) have found that the extreme hot events in Korea and Japan has a 90% increase in numbers of extremely hot days (T max > 35 C) during the negative phase of the PJ pattern events, when the two countries are located in the subsidence cell.Heatwave can cause significant impacts on human health, ecological systems and infrastructure, as evidenced by previous studies (Easterling et al., 2000;Parmesan et al., 2000;Perkins, 2015).Therefore, it is crucial to understand the underlying mechanisms behind the PJ pattern and its impacts on regional climate, especially in densely populated region such as East Asia.
Located in the subtropical region of East Asia and on the edge of the western Pacific, Taiwan's the summer climate has distinct characteristics from midlatitude countries such as Japan and Korea.Specifically, the PJ pattern's tri-cell structure of the Rossby waves is not latitudinally symmetric.As opposed to the midlatitude cell centre over Japan and Korea, the amplitude of the near-tropical cell centre is weaker and more connected to the Tropics.Thus, the impacts of PJ circulations anomalies may have distinct different impacts on Taiwan under a different seasonal climate state and interactions with Asia summer monsoon circulations, compared with midlatitude region.As such, while previous studies have focused on Korea and Japan, their results may not be applicable to Taiwan.This study aims to fill this knowledge gap by examining the impacts of PJ pattern on surface temperature and heatwave characteristics in Taiwan, with a particular focus on the flat plain region below 1000 m above mean sea level (MSL).We have chosen this domain for two reasons.First, more than 95% of Taiwan's population live in such areas, especially over the west plain between the coastal line and central mountains (Figure 1).Second, owing to the complex orography of the central mountains, there are much less observation sites above 1000 m MSL, leading to larger observational uncertainty over the mountainous region (Weng & Yang, 2018).In addition to quantifying the impacts of the PJ pattern on summer climate and heatwaves over Taiwan, we also investigate the driving mechanisms underlying its impacts.Our study fills a knowledge gap in understanding the impacts of the PJ pattern on heatwaves in East Asia, particularly in the subtropical region.The results of this study will provide valuable insights for more accurate climate information to assess the risks of hot summer temperatures and heatwaves, and to develop effective adaptation strategies to mitigate the potential impacts of future climate variations and changes.
The remainder of this paper is organized as follows.In section 2, we describe the data, model and methodology employed in this study.In section 3, we present the impacts of the PJ pattern on mean rainfall, mean temperature and heatwave characteristics during summer.Section 4 offers an in-depth discussion over the physical mechanisms underlying the impacts of the PJ pattern observed in this study.This includes an analysis of largescale anomalous circulations and budget analysis of nearsurface temperature.Finally, in section 5, we summarize our results and provide a comprehensive discussion of their implications.

| Data
Atmospheric variables, such as wind velocity, precipitation and surface 2-m temperature, were obtained from the fifth generation of the European Centre for Medium-Range Weather Forecasts (ERA5; Hersbach et al., 2020), which was on a 0.25 × 0.25 grid from the period between 1950 and 2020.Sea surface temperatures were obtained from extended reconstructed SST (ERSST) version 5 with 2 × 2 horizontal grid (Huang et al., 2017).
The 5 km × 5 km high-resolution horizontal grid data of Taiwan were provided by Taiwan Climate Change Projection Information and Adaptation Knowledge Platform (TCCIP) (2021).This dataset was derived from various data from the Central Weather Bureau, Water Resources Agency, Taiwan Power Company, and Irrigation Association (Weng & Yang, 2018).This dataset has been widely used in understanding the rainfall characteristics over Taiwan (Henny et al., 2021;Huang et al., 2022).The variables used here were surface daily averaged temperature (T avg ), surface daily maximum temperature (T max ) and rainfall.

| Definition of PJ pattern index
We follow the PJ pattern index developed by Hsu and Lin (2007), which represents the rainfall tripole pattern closely linked to the PJ pattern), in our study.First, we calculate the average summer rainfall (i.e., June, July and August) from 1960 to 2019.We removed the linear trend at each grid point based on linear regression.We then removed the 5-year running means from the detrended data to eliminate the decadal-interdecadal signals and retain the interannual variability.The empirical orthogonal analysis (EOF) was applied to the data to retrieve the precipitation pattern variability in East Asia (20 -45 N, 105 -145 E).The first EOF (Figure 2a), which explains 16.5% of the total variance, displays a meridionally oriented tripole structure with two negative polarities around southern China-Taiwan and northern China, and one positive polarity extending from central eastern China to Japan.The tripolar rainfall anomalies are elongated along the east-west direction and are relatively narrow in the meridional direction.This pattern is similar to those derived with indices based on sea-level pressure over Japan and Taiwan in previous studies (Kubota et al., 2016).We define the positive and negative phase based on the first principal component (PC1; bars in Figure 2b).By comparing PC1 with interannual variation of the summer mean temperature over Taiwan (black line in Figure 2b), we found the correlation coefficient of 0.44 (0.01 significant level) between the two.This implies a robust association between the mean field of Taiwan summer temperature and the PJ pattern when considering the interannual timescale.We then created composite by defining positive PJ years as those with PC1 larger than 0.8 standard deviation and negative PJ years as those with PC1 smaller than 0.8 standard deviation.The sensitive test suggested while the intensity of anomalies during PJ years amplifies as the threshold increases, the spatial structure remains relatively stable (Table S1 and Figure S1).We have chosen a threshold of 0.8 standard deviations to optimize the degree of freedom in our analysis.Thus, we select 11 positive years (i.e., 1962, 1965, 1977, 1980, 1983, 1993, 1998, 1999, 2003, 2014, 2015) and 13 negative years (i.e., 1961, 1967, 1973, 1978, 1981, 1984, 1985, 1990, 1994, 1997, 2001, 2013, 2018) based on this definition.

| Definition of CTX90pct index
There are various definitions for heatwaves, which emphasize different aspects of unusually hot days.We used the heatwave index (referred to as CTX90pct) developed by Perkins and Alexander (2013), which defines the threshold of a heatwave for each calendar date as the 90th percentile of the daily maximum temperature (T max ).To calculate the threshold for a given calendar day, T max values over the 15-day window period centred at the chosen calendar date for all available years were pooled together and ranked to find the 90th percentile value.This 15-day window covers the 7 days before and after the chosen calendar date.The 15-day window period is chosen because it offers a substantial sample size for the computation of a representative percentile value.This length of time window is considered long enough to accurately reflect heatwave conditions while ensuring a reasonable number of events for analysis (Perkins & Alexander, 2013).The threshold is then referred to as CTX90pct.Subsequently, a T max threshold was constructed for each 365 calendar dates, and the T max values were compared with the threshold.A day with T max higher than CTX90pct is defined as a heatwave day.Thus, CTX90pct can remove the impacts of seasonal variation and serve as an index in relative manner of all seasons.This definition was found to be suitable to a subtropical region like Taiwan for heatwave studies (Lee & Hsu, 2017).
However, while the CTX90pct index efficiently captures the localized impacts of heatwaves on each specific location, which is essential for understanding local climate variations, it has limitations in revealing the unique spatiotemporal evolution of heatwave events, which is closely tied to the underlying physical drivers.To address this limitation and complement the CTX90pct index, we have also incorporated an event-based heatwave index called the heatwave magnitude scale (HWMS) method in our study, which is described in the following section.The HWMS method tracks heatwave events across their spatiotemporal extents, allowing us to better understand the temporal evolution and spatial characteristics of heatwaves.

| Heatwave magnitude scale
Apart from the CTX90pct index, we investigated the changes in heatwave events using a three-dimensional perspective heatwave index, heatwave magnitude scale (HWMS; Lo et al., 2021).By tracking the system from an event perspective, this method can document the temporal and spatial scales of a heatwave and identify intensity, affected area and temporal evolution during the evolution to an extreme event.This approach provides a different perspective of heatwave, compared to the traditional grid point based method.
The HWMS includes three steps for heatwave event tracking.The first step is to identify the heatwave events based on the Heatwave Magnitude Index Daily (HWMId) method introduced by Rosso et al. (2015).This method utilizes a threshold derived from the daily maximum temperature (T max ), set above the 90th percentile of climatological T max over a 31-day window during the reference period 1960-2018.That is, for the day d, the threshold is defined as 90th percentile of A d , where S as the union of a dataset, T y,i as the daily T max of the day i in year y.A heatwave event is found when the T max at day d (T d ), passes the threshold for consecutive 3 days.
At the second step, the daily heatwave magnitude (M d ) is calculated for each heatwave grid by normalizing the grid-point maximum temperature by the 25th percentile (T 25p ) and 75th percentile (T 75p ) of the 30-year annual maximum temperatures.If the maximum temperature of the grid for the given day (T d ) does not exceed T 25p , the day at the given grid is not counted as part of the heatwave occurrence.The resultant M d is a standardized temperature intensity for that grid on the target day d and is computed as follows: 8 < : : In the last step, the identified heatwave grids with M d (i.e., where M d >0) on the time axis are grouped as a three-dimensional volume pixel, known as a voxel, which includes longitude, latitude, and time attributes.A heatwave event is then defined by identifying connected voxels and viewing them as one event based on a depth-first search algorithm (Lo et al., 2021).
The properties of each heatwave event can then be computed by summing properties over all voxels.For example, the heatwave magnitude scale (HWMS) is defined as follows: where d 1 represents the first day of the heatwave event, and d n represents the last day.x e and x w represent the starting and ending longitudes of the heatwave event from east to west, while y s and y n represent the starting and ending latitudes of the heatwave event from south to north.Δx and Δy are the cosine of latitude weights.An event mask δ is then used to remove voxels with cooler daily T max as below The details of heatwave properties, including heatwave event volume (HWV), mean magnitude of a heatwave event (HWMM), mean area of a heatwave event (HWMA) and mean duration of a heatwave event (HWMD), can be found in Data S1, Supporting Information and Lo et al. (2021).

| Heat budget
To investigate the contribution of physical processes to temperature variation, we used the heat budget analysis of potential temperature θ defined by Yanai et al. (1973), which is presented in the following equation: where v represents horizontal winds, r represents the horizontal advection operator, ω represents vertical pressure velocity, P is the pressure and P 0 is for reference pressure (i.e., 1000 hPa).In the equation κ =R=C P , R and C P are the gas constant and the specific heat at constant pressure of dry air, respectively.Q1 denotes the apparent heat source, which represents the heat source from unresolved processes, including radiation, latent heat release and turbulent processes.From left to right, the terms of thermodynamic budget are temperature tendency, horizontal advection term and adiabatic term.To emphasize the effect of each heating term on temperature tendency, the equation is written as following: where u as the zonal wind component, v as the meridional wind component, and

| CHARACTERISTICS OF TEMPERATURE AND HEATWAVE IN TAIWAN
3.1 | Impacts on summer-mean surface temperature and rainfall of Taiwan We used the high-resolution gridded data from TCCIP to address the impacts of the PJ pattern on temperature and rainfall.Figure 3 shows the summer-averaged temperature and rainfall composite of the two phases of the PJ pattern.The positive phase is associated with island wide warmer temperature and reduced rainfall appeared (Figure 3a,c).However, the magnitude of changes varies across Taiwan, with a significant decrease in rainfall observed in the northwest and southeast region (Figure 3a).The increase in temperature is widespread across Northern and Central Taiwan and is mostly statistically significant (Figure 3c).In this and subsequent analyses, the t test is employed to verify the significance of anomalies and mean states.We have selected a significance level of 0.1 to determine the threshold for significance.In contrast, the negative phase is characterized by cooler temperature and increased rainfall, especially over the southern and eastern Taiwan, but with fewer and weaker anomalies that lack statistical significance (Figure 3b,d).This suggests that the impacts of negative PJ pattern may not be as robust over time.Importantly, our findings reveal that one standard deviation change of the PJ index can cause a local increase in summer temperature of up 0.4 C in Northern Taiwan, which is a much more significant increase than warming trend observed during 1969-2019.As illustrated in Figure S2, the warming trend for summer's maximum temperature in Taiwan is 0.07 CÁdecade −1 over past 60-year period (1960-2019) and 0.03 CÁdecade −1 over past 30-year period (1990-2019).Therefore, even if we project the trend of the past 60 years forward, the increase in averaged temperature would be 0.05 C. When compared to the 0.4 C increase in maximum temperature during a positive PJ phase over Taiwan when PJ index increases for 1 standard deviation, the PJ pattern could exert a more dramatic influence, especially in certain local areas.This result suggests that temperature variation caused by the PJ pattern could have a more significant impact on summer temperatures in Taiwan in the coming decades than the warming trend caused by global warming.Consequently, incorporating strategies for adapting to interannual climate variation such as the PJ pattern should be a crucial component of future adaptation plans for Taiwan.However, it is also important to acknowledge that future potential impacts of the PJ pattern on Taiwan's temperature should be interpreted cautiously, considering strong natural climate variability over East Asia.Continued research and future refined model study will be crucial for providing more robust conclusions regarding the relative influences of the PJ pattern and global warming on Taiwan's temperature.

| Impacts on seasonal statistics of heatwave events in summer
As the impacts of the positive phase of PJ pattern on temperature and rainfall are more robust, our analysis has focused primarily on the impacts of positive phase and its difference from the negative phases in this study.We used three indices in Table 1 to characterize the seasonal features of heatwave events over Taiwan from both a grid-based and event-based perspective.Figure 4 shows the composite of daily maximum temperature (T max ) averaged over summer for the positive years and its difference from the negative years.In most parts of the island, the T max is warmer by approximately 0.2-0.6C (Figure 4a), and the difference between positive phase and negative phase can range from 0.4 to 0.8 C (Figure 4b).Over the northeastern part of Taiwan, T max increased by up to 0.5 C for the positive years (Figure 4a), which is similar to the T avg increase pattern (Figure 3a).Notably, T max also increased over the central and southern parts of the west plains and its piedmonts with statistical significance, while T avg did not show significant changes.Moreover, during the positive phase years, a notable increase of downward solar radiation is observed, coinciding with an augmentation in suppressed rainfall over the southwestern part of Taiwan (Figure S3).This alignment suggests a concurrent reduction in cloud cover, leading to the amplification of T max there, which is not reflected in the T avg .When stratifying the days of positive and negative years of the PJ patterns according to T max in summer, we found a clear shift in distribution towards the higher temperature end during positive years compared with the negative years (Figure 4c).During the positive phases, 26.4% of the days had T max higher than 33 C, and 2.3% of days were warmer than 35 C. In contrast, the corresponding values were 18.2% and 0.8%, respectively.In other words, during the 90 days in summer, the year in the positive phase can have up to 23.76 (2) days warmer than 33 C (35 C), while the negative year only have 16.4 (0.72) such days.
Second, we analyse the frequency and intensity of heatwaves days exceeding CTX90pct (Figure 5).Overall, the frequency exceeded 20 days per summer in the positive years (Figure 5a), which is approximately 10 more days than in the negative phase (Figure 5b).Interestingly, the heatwave intensity was the highest in southern Taiwan (Figure 5c), while the change was not very clear in the northern part of Taiwan.The difference between positive and negative phases can reach 0.3 C over the southern Taiwan (Figure 5d).The discernible regional discrepancy in heatwave intensity can be attributed to diverse mechanisms of surface warming prompted by the Pacific-Japan (PJ) circulation anomaly over Taiwan.Supported by Wu et al.'s study in 2020, it is evident that a PJlike circulation can engender intensified temperature advection over northern Taiwan, facilitated by the southwesterly anomaly of PJ circulations, which brings warm air temperatures to the northern region.Conversely, in southern Taiwan, a distinct driver is at play.The conspicuous augmentation in downward solar radiation over this region (Figure S3) contributes significantly to surface warming.However, unravelling the comprehensive underlying causes of these regional disparities needs a more extensive study, employing higher-resolution regional datasets and modelling to understand local response to the PJ pattern.
Moreover, to further understand the association of CTXpct90 heatwave properties and T max , we performed a correlation analysis to examine the potential of changes in the heatwave properties as a dependable predictor for T max .As illustrated in Figure S4, it becomes evident that the T max anomalies exhibit a strong linear correlation with changes in heatwave frequency, as identified by CTXpct90 (Figure S4a), while such a correlation is not observed with intensity changes (Figure S4b).
By plotting the probability density function of temperature when the temperature exceeds the CTX90pct thresholds during the three phases, Figure 5c illustrates that the number of heatwave extreme days have increased during the positive phase of PJ pattern compared with negative and neutral phases.Specifically, the percentage of the days with an intensity of CTX90pct exceeding 33 C was 70.2% in the positive phase, 65.8% in the neutral phase, and 63.0% in the negative phase, respectively.For the 35 C threshold, the corresponding values were 14.4%, 11.3%, and 9.2% in the positive, neutral and negative phases, respectively.
Finally, the HWMS approach is used to analyse the seasonal statistics of heatwave events, as shown in Figure 6.This method identifies the heatwave events by connectivity in both space and time, enabling the analysis of event duration, affected area and volume as the multiplication of duration and area.By examining temporalspatial characteristics from a three-dimensional perspective, we can better characterize the changes in heatwave events in temporal and spatial dimensions.Figure 6 summarizes the heatwave characteristics of HWMS in the neutral years, positive phase years and negative phase years of the PJ pattern.Generally, during the positive phase years, a marked increase is observed in the volume characteristics of an averaged event compared with neutral and negative years (i.e., effective volume = area × duration = HWMA × HWMD; Figure 6a).Upon decomposition, we found this increase mainly comes from the increased maximum effective area of identified heatwave events T A B L E 1 Three heatwave indices used in this study

Heatwave indices References
Daily maximum temperature (T max ) Heatwave defined by exceeding the 90th percentile of the daily maximum temperature in climatology (CTX90pct) Perkins and Alexander (2013) Heatwave magnitude scale (HWMS) Lo et al. (2021) (HWMA; Figure 6b), as the average duration of the events remains the same in both phases and neutral years (HWMD; Figure 6c).Meanwhile, the averaged intensity of the identified events, HWMM, remains similar intensity between the two phases and neutral years (Figure 6d).However, the most notable changes between the three phases are the increase of event occurrence, which can be up to 2 times in the positive years than in the negative years, and 1.2 times than the neutral years (frequency; Figure 6f).This frequency increases results in 2-time increase in the accumulated intensity of total events, integrated magnitude (MWMM_ACC; Figure 6e).In summary, the large-scale environment associated with the positive PJ pattern results in an increase of effective area of the individual events, as well as more frequent occurrence of heatwave events.For the intensity (HWMM) and duration (HWMD) of individual heatwave events, there is little difference between the three phases of the PJ pattern.

| DRIVING MECHANISMS OF PACIFIC-JAPAN PATTERN'S IMPACTS ON TAIWAN
Figures 7-9 provide a physical picture of how the impacts of the tri-cellar circulations of PJ pattern affecting rainfall and surface warming over Taiwan.Figure 7 showcases the composite of anomalous rainfall, surface temperature and low-level stream functions during positive and negative phases.
The dry-wet-dry (wet-dry-wet) rainfall anomalies in the positive (negative) phase along East Asia, with a Rossby wave train structure spreading from the Tropics to Northeast Asia around Korea and Japan, are consistent with previous studies (Figure 7a,b).During the positive phase, an anticyclonic anomaly at 850 hPa is evident in the Northwest Pacific between 20 N and 35 N (Figure 7a).Conversely, the low-level circulation during the negative phase is of opposite polarity, with centres slightly shifted southward (Figure 7b).Taiwan's summer precipitation is influenced by the western Pacific-East Asia monsoon system, where southwesterly winds transport moisture from the South China Sea and Southeast Asia (Figure S5).The local rainfall variations in Taiwan are thus the outcome of the interplay between PJ-related circulation anomalies, the mean monsoonal flow and the local topography.During the positive PJ phase, the anticyclonic circulation diminishes moisture advection from the South Asian monsoon region, reducing southwesterly moisture transport.This anticyclonic anomaly also brings subsidence anomalies that stabilize the atmosphere, suppress local convection and encourage clear skies (Figure 3a).In contrast, the negative PJ phase amplifies westerly anomalies over Indo-China, enhancing moisture advection to Taiwan's Southern and Eastern regions, intensifying rainfall over the southern Taiwan (Figure 3b).
The surface temperature structure of the two phases also displays a tri-cell structure, which is similar to the rainfall anomaly, revealing higher surface temperatures collocated with dry anomaly and vice versa for colder surface temperatures (Figure 7c,d).However, surface temperature variability is largest at 40 N, while rainfall signals have the largest variability at 20 N. Furthermore, the tri-cell structure is better illustrated by the low-level streamfunction and rotational wind (Figure 7e,f).Consistent with the anomalous pattern of 2-m temperature at low-level, an anticyclonic (cyclonic) anomaly at 20 N covers Taiwan, and a cyclonic (anticyclonic) anomaly at 40 N in the positive (negative) phase (Figure 7e,f).These findings showcase the linkage between the anticyclonic (cyclonic) anomalies of the PJ pattern and surface temperature anomaly during the two phases.
Figure 8 presents a latitudinal-height cross section averaged over 115 -125 E, showing a vertical profile of the tri-cellar circulations of the PJ pattern during its two phases.During the positive phase, the anticyclonic anomaly lying at 20 N causes a strong subsidence anomaly over the latitudinal band of 10 -25 N (Figure 8a).To the north of the anticyclone, an ascending region located at 30 N is associated with the cyclonic circulations.On the other hand, during the negative phase, a cyclonic circulation lies around 20 N, extending to the south end of Taiwan.Notably, as part of the anticyclonic circulation, the subsidence anomaly of PJ positive years reaches the ground (Figure 8a), leading to stronger lowlevel southwesterly winds near Taiwan (Figure 7e). Figure 8 also presents the vertical profiles of heat due to vertical advection motion along the longitudinal cross section.Aligned with the vertical motion patterns (indicated by vectors in Figure 8), substantial warming coincides with the subsidence anomaly over Taiwan during the positive phase (Figure 8a).Conversely, cooling effects align with downward motion during the negative phase (Figure 8b).
To look for the driving mechanisms that responsible for an increase of surface temperature over Taiwan during the PJ positive phase, Figure 9 2)), the temperature change is governed by the interplay of horizontal advection terms, vertical advection terms, and the residual term Q1.Horizontal advection delineates heat transfer driven by horizontal wind dynamics, while vertical advection terms encapsulate temperature changes as air parcels experience ascent (resulting in cooling) or descent (leading to warming) adiabatically devoid of heat exchange with the environment.The diabatic term Q1 includes a spectrum of processes, including latent heat release/absorption during precipitation and evaporation processes, radiative processes and turbulence diffusion.
While the average seasonal surface temperature remains poised in a state of equilibrium, the other terms within the thermodynamic equation exhibit a tendency to mutually offset one another, as depicted in Figure 9.During the positive PJ phase years, the adiabatic warming term emerges as the most substantial contributor to the surface potential temperature budget.This warming effect is balanced by diabatic cooling Q1 and negative zonal temperature advection.The presence of southwesterly flow over Taiwan during the positive phase leads to a negligibly positive meridional advection term (Figure 7a,b), in line with our findings.As an increased occurrence of clear skies and a reduced rainfall are observed over regions marked by subsidence (Figure 8), the Q1 cooling evident is primarily attributed to radiative processes involving atmospheric gases.
In contrast, during the negative PJ phase years, a contrasting response is observed, with all terms flipped in sign (Figure 9b).The prevailing east-northeasterly winds over Taiwan now lead to a balance between vertical adiabatic cooling and meridional cold advection, which are offset by diabatic Q1 warming and zonal warm advection.However, compared to the positive phase, the magnitude of the adiabatic cooling term is reduced by approximately half, and other terms also weaken.For the negative phase, the augmented rainfall in Taiwan implies heightened convective activity, suggesting that Q1 heating likely stems from convective processes.
Notably, the asymmetry in the surface temperature budget is consistent with the asymmetry observed in both low-level and upper-level circulations between the two PJ phases mentioned in Hsu et al. (2017) and in Figure 7a,b.During the positive phases, a more pronounced meridional wave train in the lower troposphere, attributed to its tropical origin, exerts stronger effects on surface temperature.As a result, the contribution from vertical adiabatic terms is notably stronger during positive phase years compared to negative phase years.This asymmetry in the surface temperature budget is further corroborated by the composite analysis of surface temperature and heatwave statistics presented in previous sections.The findings suggest that the negative PJ pattern plays a comparatively less dominant role in influencing nearsurface temperature in the Taiwan region.

| DISCUSSION AND SUMMARY
Our study investigates the impacts of the PJ pattern on the summer temperature and heatwave extremes in Taiwan, as well as the associated driving mechanisms.Particularly, during the positive phase of the PJ pattern, an anomalous anticyclonic circulation develops above Taiwan, leading to an increase of temperature of up to 0.2-0.6C and rainfall deficiencies of up to 80 mmÁmonth −1 in summer.Conversely, during the negative phase of the PJ pattern, relatively colder and wetter conditions prevail in Taiwan due to the presence of lying cyclonic anomalies.However, our analysis suggests that the impact of the PJ pattern on surface temperature is more significant than on rainfall.This is evident from significant test results presented in Figure 3, indicating that the PJ pattern may not be the dominant factor in interannual variation of summer rainfall in Taiwan.Therefore, we focus primarily on the impacts of PJ pattern on temperature, especially during the positive phase.
In warm summers during the positive PJ pattern, we found that heatwave events become more frequent and severe over Taiwan.To quantify the impacts, we used two grid-based heatwave indices, T max and CTX90pct, to identify heatwave days occurring during summer.Both indices showed a significant increase in the occurrence of identified heatwave days.Days with T max above 33 C increased by approximately 8% (15 days) compared with negative years, while CTX90pct showed a difference of around 5-10 daysÁyear −1 difference between positive and negative years.Furthermore, distribution of T max over Taiwan region shows that not only the mean T max but also the skewness of T max shift towards warmer temperatures during the positive PJ phase.In addition to the grid-base indices, we also used an event-based index, HWMS, to analyse heatwave events.Our analysis indicates that the occurrence of heatwave events can increase by more than 2 times during the positive phase compared with the negative phase.Moreover, when characterizing heatwave events with duration and affected area, the temporal-spatial multiplication (i.e., volume) of each event can increase by approximately 1.5 times during the positive phase.These results suggest that the positive PJ pattern can increase the total range of exposure area to the heatwave events, both in time and space.
The increase in summer temperature during the positive phases was attributed to the subsidence anomaly over Taiwan, which is part of the anticyclonic circulation of meridional propagating Rossby waves.A budget analysis conducted over the near-surface temperature over Taiwan revealed that the subsidence associated with anomalous circulations causes significant adiabatic warming.Although meridional advection plays a minor role in warming, it transports warmer air from the South China Sea into Taiwan region.On the seasonal scale, warming tendency from subsidence is balanced by the diabatic cooling and zonal transport of colder air from the west.The dominant diabatic cooling indicates that diabatic physical processes, particularly radiative cooling effect, play a significant role in balancing the adiabatic warming introduced by PJ circulations.A schematic showing how anomalous circulations of subsidence due to anticyclonic circulations impact surface temperature through atmospheric processes is presented in Figure 10.
Our analysis provides insight into the impact of the PJ pattern on summer temperature and heatwave characteristics in Taiwan.We find that a positive PJ pattern can potentially increase the frequency and duration of heatwave events, thereby increasing the climate risk for sectors that are vulnerable to heat stresses.Meanwhile, numerous previous studies have suggested that other climate variability on interannual and interdecadal timescale can modulate the summer climate in Taiwan.For example, Wu et al. (2020) reported an increase in extreme hot days during the summers following El Niño events, and a decrease during La Niña events.Lee and Hsu (2017) shows that the westward extension and intensification of the western North Pacific subtropical high result in anticyclonic and descending anomalies, which cause persistent warm weathers over Taiwan.Furthermore, the Pacific Decadal Oscillation can also affect temperature and rainfall over Taiwan (Huang et al., 2018;Lo & Hsu, 2008).Future studies are needed to explore how these variabilities may affect the heatwave events over Taiwan in conjunction with the PJ pattern.Combining seasonal prediction systems and early warning systems, along with a deeper understanding of driving mechanisms of heatwaves, can help society to prepare for potential climate hazards.

F
I G U R E 1 (a) Location of Taiwan in the East Asia region.The inset shows the elevation of the study area, and the 1000-m altitude line is marked in black.(b) The population distribution of Taiwan.Source: https://data.gov.tw/[Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 2 (a) Geospatial pattern and (b) associated standardized principal component (bar) of the 1st EOF mode derived from summer-mean rainfall in East Asia.The dash lines indicate 0.8 and −0.8 standard deviation.The black solid line is the summer-mean temperature over Taiwan, with a 5-year running mean removed.Correlation coefficient between the temperature and PC1 is 0.44.[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 3 Composites of (a, c) monthly rainfall anomalies (mmÁmonth −1 ) and (b, d) monthly temperature anomalies ( C) averaged over positive and negative phases of PJ pattern.The dots indicate regions where anomalies are statistically significant at the 0.1 significant level.[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 4 (a) Composites of daily maximum temperature (T max ) in the positive PJ pattern (shading; C).The dots denote the region where the anomalies are statistically significant at the 0.1 significant level.(b) The difference between positive and negative phase of the PJ pattern.(c) Probability density function of the daily maximum temperature regarding the positive (red line), neutral (black line) and negative (blue line) phase.The proportion of days exceeding 33 C and 35 C are shown for positive, neutral and negative phases, respectively, divided by the total number of days.[Colour figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 5 (a, b) Number of days when T max exceeding CTX90pct index each summer (frequency; unit: dayÁsummer −1 ).(c, d) The exceeding temperature during the hot days identified by CTX90pct (intensity; unit: C). (a, c) Composites of positive phase of PJ pattern.The dots denote the region where the anomalies are statistically significant at the 0.1 significant levels.(b, d) The difference between positive and negative phase of the PJ pattern.(e) Probability density function of daily maximum temperature of the hot days identified by the CTX90pct index regarding the positive (red line), neutral (black line) and negative (blue line) phases.The proportion of days exceeding 33 C and 35 C are shown for positive, neutral and negative phases, respectively, divided by the total number of days.[Colour figure can be viewed at wileyonlinelibrary.com] presents the nearsurface temperature budget of the Taiwanese region (115 -125 E, 20 -26 N) based on daily averaged variables from the ERA5 reanalysis.The figure shows the individual terms of the thermodynamic budget of potential temperature, which are, from left to right: temperature change ( ∂θ ∂t ), diabatic heating (Q1), zonal horizontal advection (p Ã u ∂θ ∂x ), meridional horizontal advection (p Ã v ∂θ ∂y ) and adiabatic heating/cooling (p Ã ω ∂θ ∂p ).In the thermodynamic budget equation (Equation (

F
I G U R E 8 Composites of latitude-height cross section averaged over 115 -125 E, with shading representing heating/cooling effect from adiabatic vertical advection (WÁm −2 ) and vectors representing wind circulation (v: mÁs −1 ; omega: PaÁs −1 ) anomalies.(a, b) The positive and negative phases of the PJ pattern.The dots denote the regions where the anomalies are statistically significant at the 0.1 significant levels.The topography of Taiwan is marked as green at 23 N. [Colour figure can be viewed at wileyonlinelibrary.com]F I G U R E 9 Potential temperature budget in (a) positive and (b) negative phase near the surface at 925 hPa calculated using ERA5 reanalysis averaged over the area of 115 -125 E, 20 -26 N. From left to right, the terms of thermodynamic budget are temporal change of potential temperature, diabatic heating term (Q1), zonal horizontal advection term, meridional horizontal advection term and adiabatic term.Unit: 10 −6 KÁs −1 .

F
I G U R E 1 0 Schematic of processes of positive PJ phase contributing to surface warming in summer in Taiwan.[Colour figure can be viewed at wileyonlinelibrary.com]