Combined effects of climatic factors on extreme sea level changes in the Northwest Pacific Ocean

Extreme sea level events (ESLs) act as a comprehensive consequence of climate change, and a better understanding of the impact processes of climatic factors on ESLs variability is essential for obtaining a coastal sustainable development strategy. In this work, hourly sea-level data from 699 worldwide tide gauges between 1960 and 2013 were used to analyze the ESLs variability, and we selected 72 representative tide gauges in the Northwest Pacific, which stood out in terms of ESLs intensity, duration and occurrence. Under the climate change framework, The Northwest Pacific is a key area that subjected to stronger, longer, and more frequent ESLs after 1990, and reef coast ESLs gradually shifted their sensitivity from intensity to occurrence. TC-summer and winter are the peak seasons of ESLs during the year. The effects of water level discrepancy (WLD) on ESLs are increasing, and the growing impacts even exceeded the impact of sea-level at a few Indonesian tide gauges. Low frequency variations of sea-level are provided as the background of ESLs variabilities on a seasonal time scale and secular change, with more than 80% of tide gauges experiencing accelerated sea-level rise. In terms of magnitude, the high tide level is the major climatic factor to ESLs along the Northwest Pacific coast. The nested Gumbel–Hougaard copula function is built to evaluate the ESLs partial correlation-joint return period, which introduces the combined effect of WLD, sea-level variability, high tide level, and wind. Sea-level variability, WLD, and wind all shorten the ESLs joint return periods at more than 70% tide gauges, particularly, wind exhibits as a significant shortening climatic factor throughout the Northwest Pacific. Meanwhile, the high tide level shows significant and complicated contribution to the ESLs, which fluctuated according to the complex features of the tidal wave system, and this complexity makes it possible to increase the return period. An individual climatic factor influenced weakly on ESLs joint return period when a second climatic factor is introduced considering the interactions between different climatic factors. More tide gauges will likely exhibit complex features of the ESLs joint return period, and some tide gauges that have a shortened ESLs joint return period may even show an increase. Considering the coastal flooding risk framework, coastal flooding risk is growing in the decadal fluctuation, so the enhanced forecasting and prevention strategy against future ESLs will provide an effective blueprint for adaption and mitigation.

1 3 2021; IPCC 2019). ESLs have always induced heavy coastal flooding along the adaptive coastal zone. Climate models have indicated that about half of the world's land area, population, and assets will be at risk of coastal flooding by 2100 in the RCP8.5 scenario (Kirezci et al. 2020). Therefore, the demand for the studies of ESLs and coastal flooding risk assessments (Kopp et al. 2019;Dawson et al. 2009;Bates et al. 2005;Nicholls et al. 1995) has increased concerns in scientific research and coastal plans. Climate change must be observed as a long-term problem that is superimposed on a number of more immediate issues (Nicholls et al. 1995). The report, titled the Directive of the European Parliament, emphasized that EU member states must consider the potential impacts of climate change when conducting coastal flooding risk assessments.
Former studies have demonstrated that ESLs trend have obvious spatial heterogeneity on a global scale (Menéndez and Woodworth 2010;Merrifield et al. 2013;Kirezci et al. 2020). ESLs height were relatively high (Merrifield et al. 2013;Kirezci et al. 2020) and trended positively in the North Sea, the east coast of America, and the west and central Pacific, while a negative trend was detected in the high latitudes of North America and Europe (Menéndez and Woodworth 2010). In the future, ESLs in northwest Europe, the Bay of Bengal, and southeast and east Asia will grow relatively significantly (Kirezci et al. 2020). Meanwhile, ESLs changes have shown to be regionally consistent (Marcos et al. 2015) due to large-scale weather forces. ESLs intensity and occurrence have shown to be regionally consistent on decadal scales (Marcos et al. 2015). Moreover, ESLs have been characterized by multi-timescale variabilities, apart from seasonal variations, with their residual long-term changes not simply linear in growth, but varying in fluctuations. Rashid and Wahl (2020) pointed out that ESLs consist of fluctuations from decadal to multidecadal. Under these multitimescale changes, most coastlines around the world may be exposed to today's 100-year ESLs every year by 2100 (Vousdoukas et al. 2018b, even under RCP2.6 scenario (Frederikse et al. 2020).
Extreme sea level, as an atypical low-likelihood, highimpact event, has intricate variability influenced by a variety of natural and socioeconomic factors. Under the IPCC framework, the climatic factors contributing to ESLs variability along the coast such as storm surges, sea-level variability, and tides will exhibit significant fluctuations from seasonal to decadal (IPCC 2013). Calafat et al. (2022) proposed a view where trends in surge extremes and sealevel rise both made comparable contributions to ESLs overall change in Europe since 1960. Storm surges, as an important influence of ESLs, were found to be significantly influenced by extreme climate events on short time scales. The impact of storm surges on coastal climatology flooding was found to be even larger than sea-level variability for 96% of America's coastline due to increased rainfall by tropical cyclones (Gori et al. 2022). The Northwest Pacific Ocean, as an important region for tropical cyclone activity, was found to be relatively more vulnerable to ESLs. A comprehensive understanding of extreme storm surges will be vital for coastal flooding risk management and hazard mitigation (Zhang and Wang 2021). Significant warming over the twenty-first century has intensified tropical cyclone activity, leading to an increase in high-intensity ESLs occurrence (Grinsted et al. 2013;Lee et al. 2017). The location of storm surges in the Northwest Pacific shifted poleward after the 1980s (Oey and Chou 2016), leading to an expansion in tropical cyclone (TC) influenced ESLs at higher latitudes. When we focused on local water level evolution and tide fluctuations, the water level discrepancy (WLD) between extreme water and high tide level had to be considered. Skew surge has typically been used to capture changes in storm surge activity (Williams et al. 2016), with its strength in risk assessment (Feng et al. 2021). Most ESLs are driven by moderate skew surges combined with high spring tides, and this was verified around the coastline of New Zealand by Stephens et al. (2020). Extreme storm surges often occurring around mid-tide or low-tide (Horsburgh and Wilson 2007).
Regional sea-level rise also plays an increasingly important role on the coastal flooding in the future (Kirezci et al. 2020). Sea-level rise has accelerated the increase in ESLs in most regions of the world. For example, ESLs and sea-level variability have shown significant positive correlations at most tide gauges along the Chinese coast (Feng et al. 2019). Global average sea-levels rose 0.2 m from 1901 to 2018, and the rate of sea-level rise has continued to increase (IPCC 2021). Global sea-level rise is projected to reach 0.43 m (or more) by 2100, even under RCP2.6 (IPCC 2019). In addition, an evident regional difference has been observed in ESLs sensitivity to sea-level variability. In particular, ESLs occurrence will increase by a factor of 1.5-8 in areas with higher ice-melt sensitivity or where the effects of other climatic factors can cause significant regional sea-level rise (Goodwin et al. 2017). Moreover, sea-level variability interacts with the 18.61-year tidal cycle in tide amplitudes, producing decadal to multidecadal fluctuations in ESLs (Talke et al. 2018;Rashid and Wahl 2020). With rising sea levels, increasing moderate and more common high tides will reach flood thresholds, resulting in a rapid increase in high-tide flooding occurrence (Thompson et al. 2021). The 18.61year tidal cycle entering the positive phase may drive flood heights above these thresholds sooner than sea-level rise would alone (Baranes et al. 2020;Thompson et al. 2021).
Extreme weather processes are important factors in triggering ESLs, and accompanying wind changes play a considerable role in ESLs. In some regions, the impact of coastal flooding due to extreme weather may be more pronounced than sea-level rise, such as the Baltic Sea coast (Vousdoukas et al. 2018a). Both wind speed and wind direction can affect ESLs. Ideal model experiments by Andrée et al. (2022) have shown that ESLs will increase linearly to quadratically with increasing wind speed, with the effect of wind direction on coastal ESLs dependent on the time scale of wind action on the coast. Planetary intensification of surface winds has been observed since the early 1990s, displaying a significant sustained increase trend in the future under RCP8.5 (Hu et al. 2020). In addition to understanding the individual effects of each climatic factor on ESLs, interactions between different climatic factors are extremely important in the prediction of ESLs. The interactions can be considered negligible in deep water; however, they will become increasingly important on the continental shelf and in regions of shallow water (Pugh 1987). Obvious regional differences have been found in the interactions between climatic factors (Gori et al. 2022). Arns et al. (2020) determined that the largest tide and surge interactions were found for the US East Coast and the Gulf of Mexico, the UK North Sea coastline, and parts of the southern Japanese coast. If the interactions are not considered, ESLs will be up to 30% higher at most tide gauges (Arns et al. 2020). Tide-skew surge interactions will cause extreme skew surges to occur more often during smaller high-tide levels in some tide gauges with a mixed semidiurnal regime, especially those on the shallow continental shelf (Santamaria-Aguilar and Vafeidis 2018). Moreover, the interactions between sealevel variability and surge in the German Bight may lead to extreme sea-level elevation (Arns et al. 2015). As a result, the interactions between different climatic factors are not considered negligible for coastal flooding.
We need to further understand the joint effects of climatic factors on ESLs, especially for the purpose of disaster risk assessment, long-term risk predictions, and early warning systems (Berkhout et al. 2006). Therefore, joint return period analysis has been used to examine the compounding effects of climatic factors (Liu et al. 2015;Gori et al. 2022). Zhang and Singh (2007) initially used the copula method to calculate the joint return period for river flood occurrence. Subsequently, the nested copula method was used to investigate the hierarchical effects of multiple climatic factors (Hofert 2012;Lo et al. 2020). In this study, we explore the independent and joint effects of climatic factors on ESLs joint return periods which are established on the nested copula approach and partial correlations, details are shown in Sect. 2.2.
Previous studies on ESLs variations, as well as the effects of WLD, wind, sea level, and high tide level on ESLs have shown to be sufficiently extensive, but work on the relative importance of the above climatic factors on ESLs still needs to be further developed. Under the climate change framework, we focused on the variabilities of ESLs affected by various climatic factors under different shoreline conditions. The rest of this paper is organized as follows: Sect. 2 introduces the datasets and the partial correlation-joint return period methods used in this work. Section 3 presents the global ESLs characteristics and focuses on the Northwest Pacific Ocean, where high-intensity ESLs are growing in occurrence. Section 4 discusses how the climactic factors, such as WLD, high tide level, sea-level, and wind, influence ESLs features and explores the combined effects on ESLs by the partial correlation-joint return periods analysis. Finally, the performance of coastal flooding risks along different shorelines is presented in Sect. 5.

Sea level height
Analysis of ESLs, including variabilities and its climatic factors, was conducted based on the global extreme sea level dataset derived from the Global Extreme Sea Level Analysis (GESLA) project (Woodworth et al. 2017). The GESLA project assembled higher-frequency (i.e., hourly or more frequently) sea-level records from 30 resources (as mentioned by IPCC, 2021) into a common format with good consistency. The datasets (GESLA version 2) contained 39,151 years of higher-frequency measurements from 1355 tide gauges, and hourly sea-level data were checked for datum and time shifts, continuity, and suspicious outliers. In this study, we evaluated more than 1000 worldwide tide gauge records based on their continuity and time span. Considering ESLs long-term variability, we resampled tide gauge records with a time span greater than 20 years and data completeness greater than 90%, and obtained 699 worldwide suitable gauges. Although some tide gauges were very close to each other, we attempted to take all into account to investigate ESLs statistics through the performance of different regions' gauges and time spans.

Wind and sea level pressure
The wind and sea level pressure field data were derived from the NCEP/NCAR reanalysis dataset (Kalnay et al. 1996), which was jointly produced by the National Centers for Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR). The wind and sea level pressure field data were available with four times a day, daily and monthly formats (available from 1948 to the present). In this work, we used the 2.5° by 2.5° global daily gridded wind and sea level pressure field data from 1960 to 2013 to study the effects of large-scale wind on ESLs. In addition, changes in the sea level pressure field helped us better determine the impact of tropical cyclones on coastal zones.

Typhoons
Typhoons are important weather processes that trigger ESLs (Gori et al. 2020;Zhang and Wang 2021. The typhoon data was obtained from the Typhoon Best Track dataset provided by the U.S. Navy's Joint Typhoon Warning Center and included information on all typhoons from 1945 to 2019, including the typhoon's center location, maximum wind speed, minimum pressure, and the typhoon intensity level at a given time. We determined whether ESLs were likely to be influenced by typhoons by combining the location relationships between the typhoon paths, impact ranges, and coastal tide gauge locations in the Northwest Pacific.

ESLs sampling
Based on the concept of extreme sea level from the IPCC (IPCC 2021), exceptionally high sea surface heights were examined in this work. The threshold of extreme sea level had to be high enough to ensure that the exceedances described real extreme events (Menéndez and Woodworth 2010). To highlight the ESLs variabilities and avoid aliasing induced by the large seasonal variations in sea-level amplitude and tides, our ESLs threshold was the top 3‰, approaching most previous studies (Menéndez and Woodworth 2010;Arns et al. 2013;Sweet et al. 2014;Williams et al. 2016;Wahl et al. 2017;Baranes et al. 2020;Johnson and Lumpkin 2021;IPCC 2021). In this work, higher ESLs threshold like the extra-ESLs threshold were conducted by utilizing the daily maximum water levels from the GESLA dataset. An average of one or two ESLs per year was found in this manner for the majority of tide gauges, highlighting the low-probability but high-impact potential. When the water level evolution continuously exceeded the threshold by 2 h and the interval period of the two consecutive ESLs was larger than 3 days, it was defined as ESLs. We characterized the ESLs variabilities by the integrated intensity (the integral area of the water level pulled up above the given threshold during every event), duration, and occurrence of every event.

Partial correlation-joint return period
In this work, we used joint probability analysis (Masina et al. 2015;Salecker et al. 2011;Wahl et al. 2012;Zhang et al. 2019) to compare different climatic factors on the ESLs return period. The copula method was used to calculate the ESLs joint return period to portray the joint effect of multiple climatic factors. Gumbel-Hougaard copula (Zhang and Singh 2007;Wong et al. 2010) and Frank copula (Liu et al. 2015) have been commonly used for hydrological applications to calculate the joint return period, they can describe the interactions among various variables by establishing flexible multivariate distributions. The results of two copula functions are compared by examining the cumulative probabilities separately, and the Gumbel-Hougaard copula showed more reasonable cumulative probability, therefore we chose the Gumbel-Hougaard copula function to construct the partial correlation-joint return period analysis. Using the nested method, the bivariate joint return period was extended to the trivariate joint return period.
Based on the relative importance of each climatic factor on the ESLs in the Northwest Pacific, we used partial correlation to portray the relative importance of each climatic factor and constructed the ESLs joint return period considering the joint influence of multiple climatic factors. This was titled the partial correlation-joint return period (p-JRP) in this work: where the trivariate joint exceedance probability (Pe) was calculated by (Liu et al. 2015): where C 2 is the bivariate Gumbel-Hougaard copula function (Wong et al. 2010;Nelsen 2006), C 3 is the nested trivariate Gumbel-Hougaard copula function, is the water level, and are the climatic factors, t , a t , and t are their threshold (the top 3‰), d , d and d are the marginal distribution functions of variables, and cor is the correlation coefficient of the two climatic factors. When studying the combined influence of the two climatic factors on ESLs, partial correlations between the water level and two climatic factors ( cor and cor ) were considered. For the sake of calculation, cor is the mean value of partial correlation coefficients between water level and two climatic factors.

Global ESLs characteristics
To determine the global ESLs long-term changes associated with climate change from 1960 to 2013, we analyzed 699 representative tide gauges worldwide and divided them into 11 geographical regions. Among these, European, Northwest Pacific, and North American coasts are crucial regions with relatively denser tide gauge distributions. Global coastlines are under constant threat from ESLs, with a global average of 1.36 ESLs annually at each gauge; however, obvious spatial heterogeneity is observed in ESLs due to climatic conditions and geography. The regions represented by Europe and the east coast of North America are significantly affected by high-impact ESLs with longer duration and higher intensity, while other regions are found to be more sensitive to ESLs occurrence. The global ESLs characteristics are shown in Fig. 1a. Affected by storm surges, sea-level rise, and high tides (Vousdoukas et al. 2017), Europe experiences more frequent ESLs with a much higher intensity and longer duration than most other regions ( Fig. 1b-d). The ESLs intensity in Europe is almost nine times stronger than the global average, mainly due to longer ESLs duration along the North Sea coast (around 6 h for each event). Therefore, this result in inevitable consequences for European flooding risk demands (Vousdoukas et al. 2018a;Calafat et al. 2022). Moreover, it is worth noting that the western boundary current coastlines also suffer from frequent high-intensity ESLs. Sea-level rise combined with climate extreme events such as tropical cyclones, extratropical cyclones, and monsoon (Feng et al. 2019;Lai et al. 2021), has made western boundary current coasts more vulnerable to ESLs. ESLs intensity and duration along the Northwest Pacific coast are 1.5-2 times compared with the eastern coast. Many low-lying islands and reefs distributed in the mid-Pacific Ocean are found to be more sensitive to ESLs occurrence and suffer from ESLs more than once. However, the neritic islands and reefs are more susceptible to high-intensity ESLs. The tide gauges along the east coast of South American are statistically exposed to extremely frequent ESLs (Fig. 1c). Since 1990, sea-level rise has accelerated (IPCC 2021; Hay et al. 2015) at a rate of 3.1 mm/yr, and global coastal tide gauges have also experienced the annual sea-level rise of 2.1 mm from 1991 to 2013. Meanwhile, ESLs intensity, occurrence, and duration have also increased significantly compared with the steady changes observed before 1990 ( Fig. 2b; Table 1). During the period of 1991-2013, ESLs frequently affect coastal zones at an increased rate of 0.06 occurrences per year at each gauge. There have been 885 ESLs per year globally since 1990, which is more than two times from 1960 to 1990. Even more alarming is that ESLs intensity even increases by a factor of 11.8 between 1991 and 2013 compared with between 1960 and 1990, and this is enhanced by a rate of 0.17 m·h/yr. ESLs duration increased 0.04 h every year since 1990, and this is 1.4 times longer than during the period between 1960 and 1990. Despite the slight increase in ESLs duration around the globe, some tide gauges along the North Sea coast are found to be vulnerable to long-lasting ESLs. Globally, coastal zones have suffered from more frequent and more damaging ESLs since 1990, with about 10% of tide gauges worldwide experiencing ESLs in more than 5 times occurrence and intensity for 1991 to 2013 than before. In addition, the increase is most pronounced in the subtropical oceanic west coasts and islands in the Pacific and Europe (Fig. 2a). Among these, more than half of the tide gauges in the Northwest Pacific experience more frequent high-intensity and long-duration ESLs, which is significantly higher than the global average. Considering the high sensitivity of the Northwest Pacific to ESLs and rapid sea-level rise, as well as the complex types of coastlines including islands, reefs, and continental coasts, this work focused on the Northwest Pacific to study ESLs long-term variability and its climatic factors, such as WLD, high tide level, sea-level variability, and wind.

ESLs variations in the Northwest Pacific
When we focused on the Northwest Pacific Ocean, we excluded tide gauges within the bay, which were subjected to complex processes on account of the terrain effect, such as seiche (Saharia et al. 2021). In addition, we gave preference to tide gauges with better quality among the neighboring tide gauges, to reduce double counting of ESLs at different tide gauges caused by the same weather process. Finally, we employed 72 (16 tide gauges along the Chinese coast, 24 tide gauges along the Japanese coast, and 32 tide gauges in Indonesia) representative tide gauges to explore the ESLs variabilities in the Northwest Pacific, a key area where high-intensity ESLs occur frequently. Nearly 80% of the tide gauges along the coasts of China and Japan experience ESLs at least once a year, which occur on average once every 2 years in the Indonesian islands and reefs. ESLs duration is relatively short along the Chinese coast, lasting about 3 h on average. Although not sensitive to the ESLs duration, the tide gauges along the Chinese coast are also threatened by high-intensity ESLs under the influence of a higher sea surface height. For example, Hong Kong, Lianyungang, Lvsi, and Shijiusuo are affected by dramatic ESLs, but with a much lower duration than the median. Meanwhile, along the Japanese coast, especially along the northern coast, ESLs average duration can reach 6 h or even longer. Many low-lying islands and reefs distributed in the tropical Pacific have shown to be highly vulnerable to ESLs, experiencing long-duration ESLs that can even last 7 h. The Japanese coast and some islands and reefs have suffered more severe ESLs as a result of long ESLs duration (Fig. 3b). In addition, nearly a third of the gauges experience low-probability but high-impact ESLs, with alternatively half of the gauges threatened by highprobability but low-impact ESLs (Fig. 3a). The ESLs characteristics along the northern and southern coasts of Japan also differ significantly due to different influencing processes. ESLs in the northern tide gauges are stronger, longer, and more frequent, which is possibly related to the influence of strong extratropical cyclones, cold waves, and some crossisland tropical cyclone activities along the northern coast.
With climate change, sea levels are rising, and extreme weather processes are simultaneously showing different characteristics than before. More slow-moving and longlasting tropical cyclones have shifted northward (Oey and Chou 2016), and the characteristics of ESLs have shown obvious changes. Overall, ESLs intensity in the Northwest Pacific has been relatively stable, remaining at a relatively high level of about 0.2 m·h since the 1980s (Fig. 4a). Among these, ESLs intensity in Indonesia increases rapidly after 1985, and the coasts of China and Japan show  a relatively stable increasing trend. ESLs duration continues to rise in the Northwest Pacific, with the duration increasing from 2.4 to 4.3 h. Although there is a slight decrease in the 1990s, the impact time of ESLs on coastal zones begin to increase slowly after the twenty-first century (Fig. 4b). ESLs duration increases most significantly along the Japanese coast, and Indonesia also shows an overall increasing trend, but a slight decrease after the 1990s. ESLs occurrence has also been constantly increasing over time, from 0.26 events annually at each gauge in the 1960s to more than three events in the twenty-first century (Fig. 4c). A total of 729 ESLs have occurred in the Northwest Pacific since 2000. In particular, ESLs occurrence along the Indonesian and Japanese coasts increases significantly in the twenty-first century, with up to 5-6 events annually at each gauge along the Japanese coast, posing a significant threat to the safety of the coastal zone. Indonesia also grows to more than twice annually at each gauge, and China remains at an average level of once a year. Sea-level variability, which can be closely related to the ESLs, has steadily increased continuously since the 1960s. And the rate of sea-level rise has increased significantly since the 1990s (Fig. 4d), resulting in a noticeable impact on low-lying islands and reefs. High tide level ESLs show significant seasonal variability that is particularly prominent along the coasts of China and Japan. ESLs seasonal variability closely follows the seasonal pattern of sea-level anomaly (Stephens et al. 2020) and climate extreme events. High sea levels (Fig. 4h) and frequent climate extreme events in the summer and winter in the Northwest Pacific result in a double peak in ESLs occurrence during the year (Table 2). Considering the influence of tropical cyclone activity on ESLs, we analyzed the tropical cyclone data and delineated two critical areas in the Northwest Pacific based on the tropical cyclone movement paths. Only when a tropical cyclone pass through the critical area is it considered likely to affect Northwest Pacific coastal zones. Combined with the changes in wind field near the tide gauges, if ESLs occur at the tide gauges within 10 times the maximum wind radius (Kawai et al. 2005) of tropical cyclones, we recognized them as TC-involved ESLs. Based on this, we distinguished the two peaks as summer when tropical cyclones are prevalent (TC-summer, JASO) and winter (NDJ). Along the coasts of China and Japan, tropical cyclones are the main triggers of ESLs in TC-summer with nearly 70% of ESLs caused by tropical cyclones. The high sea levels superimposed on extreme weather-induced WLD make TC-summer a high period for ESLs, with intensity and duration that are relatively high (Fig. 4e-g). In winter, sea levels are still relatively high even though they have been falling since October, and the extratropical cyclones and other strong wind processes affect onshore winds, leading to extreme WLD. Winter and spring are the seasons with the highest intensity and occurrence of extratropical cyclones in the Northwest Pacific (Lee et al. 2020), and some low-latitude tide gauges are affected by tropical cyclones, resulting in ESLs. Indonesia lies in the low latitude zone where semi-annual variations have been more prominent. The area is exposed to ESLs throughout the year and is relatively more susceptible to frequent high-intensity and long-duration ESLs in the winter. Low-probability but high-intensity ESLs also threaten Indonesia in the summer. ESLs long-term and seasonal variations are regulated by various climatic factors, and only by fully understanding their influence processes can we better predict ESLs.

Attribution of ESLs
In this work, WLD, high tide level, and sea-level variability together form the sea surface height. The time series data of sea-level variability was obtained by removing the mean sea level for the entire period at each tide gauge, and the tidal contribution to ESLs was estimated by the MATLAB-based harmonic analysis toolbox T_Tide (Pawlowicza et al. 2002). Considering nodal modulation correction and the Rayleigh criterion, tidal harmonic analysis using 369 days of data can yield distinguish the diurnal and semidiurnal tidal constituents more accurately. On this basis, the WLD was characterized by the difference between the maximum observed water level around the astronomical high tide (within 6 h) and the predicted high tide level, regardless of timing (Arns et al. 2020;Stephens et al. 2020;Feng et al. 2021). The occurrence time of the maximum observed water level and high tide level had important implications when we focused on the impact of WLD on ESLs. Therefore, we separated WLD into skew (the maximum observed water level did not coincide with high tide level) and surge (the maximum observed water level occurred simultaneously with high tide level). To understand the contribution of different climatic factors to ESLs, we conducted an attribution analysis of ESLs. Sea-level variability is an important driver of ESLs at most tide gauges (IPCC 2013), providing an important context for ESLs variations on a seasonal time scale and secular change (Fig. 4d, h). In some regions, storm surges (mentioned here as WLD) or tide contributions dominate (IPCC 2021). WLD, sea-level variability, and high tide level are the most significant contributors to ESLs, and we focus on the impact of these three on ESLs. ESLs attribution analysis (Fig. 5) indicates that the direct contribution of sea-level variability to ESLs is not obvious in terms of magnitude; however, low-frequency sea-level variabilities provide a background for ESLs variations overall Northwest Pacific. In the Northwest Pacific, more than 95% of tide gauges have experience sea-level rise since the 1960s, and the rate of rise has accelerated, especially after 1990, exposing coastal zones to stronger ESLs more frequently. In addition to rising sea levels, WLD and high tide level have regulated the occurrence of ESLs. There is a clear decadal variation in ESLs in the Northwest Pacific, mainly along the Chinese and Indonesian coasts (Fig. 5), which is mainly due to decadal variations in the high tide level. High tide level plays a dominant role in the Northwest Pacific in terms of magnitude (Fig. 5), and high tide level decadal variations play a moderating role in the contribution of sea-level variabilities to ESLs (Thompson et al. 2021). However, for tide gauges with smaller tidal ranges, especially those along the Japanese northern coast and Indonesian reefs, WLD is a more dominant climatic factor that affect ESLs. Along the northern coast of Japan, the correlation coefficient between ESLs height and WLD exceeds 0.9 and even 0.95 for ESLs height and skew. In addition, most ESLs are driven by high tide level and an appropriately WLD (Stephens et al. 2020). WLD caused by tropical cyclones and other strong wind Japan, and d Indonesia. WLD included the annual mean skew and surge, which were superimposed on the sum of sea-level anomaly and high tide level, respectively. Data went missing in 11 Chinese tide gauges after 1998, resulting in their unexpectedly decreased ESLs height processes may generate severe flooding over low-lying coastal zones, particularly during high tides. High tide level and WLD modulate each other on a short time scale, where the phase difference between them can also regulate the feature of the ESLs. ESLs in the Northwest Pacific are more correlated with changes in skew, with correlation coefficients exceeding 0.6 in all regions except the Chinese coast. With climate change, the phase displacement between WLD and high tide level undergoes constant adjustments and changes, and the influence of skew on ESLs is enhanced and even have a tendency to exceed that of surge (Fig. 5).
Wind-driven piling-up along coastlines is the main driver of extreme water levels (Andrée et al. 2022). By comparing the probability distributions of skew, surge, and wind speed, we find that the distribution of wind speed and WLD, especially skew, is very similar. They all show a significant positive skewed distribution (Fig. 6a). When focusing on flooding in coastal zones, we focus on low-probability and high-impact events in the long-tail distributions. More than half of the ESLs in the Northwest Pacific during 1960-2013 are caused by strong wind processes, and 10% of extremely strong winds cause WLD of more than 1 m. WLD increase linearly with the wind speed (Fig. 6b). The higher the wind speed, the more likely it is to trigger higher WLD, leading to destructive ESLs. There is also 15% of ESLs are caused by smaller winds. In addition to the effects of high tide level, sea-level variability, precipitation, as well as other climatic factors, wind direction regulation is an important climatic factor to ESLs. Between 1960 and 2013, more than half of the ESLs in the Northwest Pacific are caused by onshore wind-driven piling-up, which benefit more water rushing into coastal zones. Notably, 30% of ESLs are caused by offshore winds, which is even comparable to onshore winds along the coast of Japan (Table 3). Seasonal variations in wind direction are obvious in the Northwest Pacific. The Chinese coast is mainly influenced by southerly winds in the summer (Fig. 6c). After September, winter winds gradually affect northern China, and northeasterly winds start to dominate in the winter (Fig. 6d).

Fig. 6
ESLs variabilities based on wind and WLD (incl. skew and surge) in the Northwest Pacific. a Probability distributions of the normalized wind speed (where 'Windt/Windm' denote the normalized daily wind speed on the ESLs occurrence day by the exact monthly wind speed climatology), WLD, skew, and surge water levels; b scatter plot of the normalized wind speed with skew and surge water levels, and their linear fits; c ESLs statistics rose (shading) diagram variations with wind directions at the 16 azimuth angles in the Northwest Pacific (China, Japan, and Indonesia) during the TC-summer season (JAS&O), compared with the prevailing winds (colorful arrows) in monthly climatology. The yellow and pink arrows indicate the prevailing wind directions in the JAS and October during the TC-summer seasons, respectively. The radius of the polar coordinate indicated the probability of ESLs heights in this direction with a 10% interval. The wind directions are divided into four segments and are shown as the double-headed arrow ranges, e.g., red indicated that the wind direction range most likely to occur ESLs; d similar to (c), but in winter (NDJ) Although the wind direction varies greatly with the seasons, the ESLs are mainly influenced by northeasterly winds both in the winter and TC-summer at most Chinese tide gauges. In addition, some gauges possibly suffered low-probability, but high-impact ESLs are influenced by winds from other directions. The Japanese coast is mainly affected by tropical cyclones from low latitudes in TC-summer, and winds from the southwest (Fig. 6c) are more likely to cause stronger ESLs, while winter is affected by extratropical cyclones coming from China (Fig. 6d). Because Japan is an island nation with tortuous shore boundary conditions, winds from multiple directions can cause ESLs during cross-island tropical cyclone crossings. And high-impact ESLs outside the influence of the prevailing seasonal winds cannot be ignored. Despite the relatively low wind speeds in Indonesia, islands and reefs are also prone to strong ESLs due to their low topography. The winds over Indonesia vary in different seasons, with southwesterly winds prevailing in summer (Fig. 6c) and northeasterly winds prevalent from October. Nevertheless, easterly winds continue to dominate ESLs in Indonesia. In the reefs, ESLs are mainly influenced by easterly winds (Fig. 6d) from Pacific Ocean. Some islands are more susceptible to stronger ESLs due to southwesterly winds in the summer and northwesterly winds in the winter. Therefore, the wind direction and wind speed jointly regulate ESLs.
*Wind blows parallel to the local coastline of the tide gauge within a 40° angle.
Outside the dominant wind direction, low-probability but high-intensity ESLs also potentially occur due to the complex influence of WLD, sea-level variability, high tide level, wind, and the interactions between them. Therefore, we focus on the combined effect of climatic factors on ESLs. In addition to ESLs with higher intensity and longer duration, which occur more frequently in the Northwest Pacific after 1990, wind speeds in the Northwest Pacific also increased significantly, especially during TC-summer when ESLs are highly prevalent. Considering this regime shift, we compared the changes in the ESLs joint return period under the influence of climatic factors during 1991-2013, compared with 1960-1990. We chose 61 of the 72 tide gauges in the Northwest Pacific because of their sufficient data from tide gauges both before and after 1990. When water level combined with different climatic factors, the changes in ESLs joint return periods for the two periods show that WLD, sea-level variability, high tide level, wind, and the combined effect of different climatic factors have varying degrees of influences on the ESLs joint return period (Fig. 7). Under the climate change framework, continued sea-level rise and dramatic WLD in coastal zones both result in a shortening ESLs joint return period at 70% of the tide gauges along the coasts of China, southern Japan, and Indonesia. By contrast, Japanese northern coast is less affected by sea-level variability and WLD. Wind has a wide range of influence and present consistently across the field, and it also plays an important role in shortening of ESLs joint return periods in the Northwest Pacific. High tide level causes more frequent ESLs along Japanese northern coast and most reefs. However, the effect of high tide level on the ESLs joint return period in most other regions is not obvious and even leads to an increase, which is mainly due to complex variations in the tidal wave system and its stable periodic variation characteristics by celestial gravity. Because the different climatic factors are not completely independent, but have interactions between them, when we combine the water level with two different climatic factors, the effects of the individual climatic factors on ESLs joint return period are weakened. More tide gauges exhibit the complex features of the ESLs joint return periods when we introduce the second climatic factor, and some tide gauges that have a shortened ESLs joint return period may even show an increase. The relative importance of the climatic factors is also reflected in the ESLs joint return period. Nevertheless, under the combined influence of multiple climatic factors, ESLs joint return periods still show a shortening at more than half of the gauges. The interactions between wind and high tide level are the most obvious and can be characterized by skew. The phase differences between the accumulation of seawater caused by onshore wind and high tide levels result in a combined effect of the two, the combined effect potentially lengthen the ESLs joint return period. Under the combined influence of various climatic factors, ESLs joint return period at most tide gauges is significantly shorten during 1991-2013, and ESLs affect the Northwest Pacific frequently.

ESLs involve in coastal flooding risk
With climate change, climate extreme events such as tropical cyclones and extratropical cyclones in the Northwest Pacific with high intensity and low movement speed have become more frequent (Lee et al. 2017). The location of tropical cyclone maximum intensity in the Northwest Pacific shifts poleward after the 1980s (Oey and Chou 2016), implying an increased threat of stronger tropical cyclones for higher latitudes. Extreme WLD accompanied by highly damaging weather processes constantly threaten the safety of coastal zones. For WLD, most studies have focused on the effects on short time scales, whereas we focused on WLD long-term changes in the Northwest Pacific and found its trends made comparable contributions to ESLs with sea-level rise. We compiled long-term trends for WLD and sea level for each station in the Northwest Pacific, and presented representative tide gauges in each region (Fig. 8). Sea-level rise occurred at most tide gauges in the Northwest Pacific (Fig. 8a) and accelerated in the tropical western Pacific during the 1990s and 2000s (IPCC 2021). Among these, the overall sea-level trend is significantly higher for islands than for continental and reefs coasts. Surge and skew show an increase for more than half of the islands, and nearly 83% of the continental and reefs gauges show a rise (Fig. 8b). The rising surge and skew indicate that the contributions of WLD to ESLs grow at most tide gauges in the Northwest Pacific. The rising trends of skew and surge even exceed sea-level variability on a few Indonesia islands and reefs (Fig. 8a). However, a significant declining trend is observed for sea level, surge, and skew water levels in eastern Indonesia, which is possibly related to a reversed situation with sea-level falling in the tropical western Pacific Ocean during the 2010s (IPCC 2021). Sea level, surge, and skew water levels vary widely among the island tide gauges, with a few gauges showing sea level decreasing at a rate of 5 mm/yr, while the sea-level rising rate in some gauges can reach 7 mm/yr. In general, most tide gauges in the Northwest Pacific are experiencing sea-level rise and increasingly strong WLD.
As sea level and WLD in the Northwest Pacific increase significantly, the contributions of climatic factors to ESLs also change with time. Sea-level rise is evident in the Northwest Pacific Ocean. Along the Japanese coast, with an increasing contribution to ESLs from a negative share of sea-level variability before the 1990s, to a share of about 10% in recent years (Fig. 9b). In the context of continuous sea-level rise, WLD and high tide level contribute steadily to ESLs with a constant fluctuating change (Fig. 9). In addition, there is clear decadal variability in the contribution of WLD and high tide level toward ESLs along the Japanese coast. Therefore, it is critical to consider tidal non-stationarity in planning coastal flooding. ESLs along the Northwest Pacific coast are tidally dominated in terms of magnitude. The high tide level contributions to ESLs can reach more than 70% alone the coasts of China, islands, and reefs. By contrast, along the Japanese coast, the contribution of WLD often exceeds that of high tide level due to the lower tidal range, although high tide level still accounts for a major part of ESLs. In some years, reefs are also more significantly affected by WLD and its contribution is even similar to the high tide level. Since 1990, the contribution of high tide level to ESLs along the continental coast has decreased, and ESLs are more likely to occur at lower high tides. This is possibly due to increased intensity tropical cyclones in the Northwest Pacific (Oey and Chou 2016) and more extreme WLD affecting coastal ESLs. The tide gauges along the continental coast are experiencing a positive contribution of skew and surge. The timing of WLD and high tide level is an important indicator that affects ESLs. According to statistics, more than half of ESLs in the Northwest Pacific between 1960 and 2013 occurred at high tide, and the average surge and skew water level associated with ESLs are 0.29 and 0.39 m, respectively. We counted ESLs with annual incidence, where WLD manifests itself as skew. The results show that the association between ESLs annual incidence where WLD manifests itself as skew and annual WLD contribution to ESLs is not significant along the coast of China before 1990, while the correlation coefficient reaches 0.47 after 1990. Along the Japanese (reefs) coast, the two are better related throughout the time period, with the correlation coefficient ranging from 0.59 (0.6) before 1990 to 0.76 (0.69) after 1990, and the significance level exceeding 99%. This is due to stronger seawater accumulation brought by skew that the skew annual incidence affects the contribution of WLD to annual ESLs.
An increasing number of tide gauges are possibly affected by dramatic WLD from extreme weather for longer periods of time, and skew surge (mentioned here as WLD) can be selected for use in risk-based coastal planning frameworks (Feng et al. 2021). We referred to the coast flooding risk index proposed by Little et al. (2015) and adopted the threshold used in this work to characterize the changes in ESLs characteristics under the influence of extreme WLD. The results show that the Chinese coasts have been at high risk from extreme WLD for a long time (Fig. 10), and the Indonesian islands have also been significantly affected by WLD since 1990 (Fig. 10b). The coastal flooding risks under the influence of WLD were very low along the Japanese coast before 2000 (Fig. 10b). Only a few tide gauges along the northern coast of Japan are vulnerable to extreme WLD (Fig. 10a). However, this started to change in the twenty-first Fig. 8 Trends in sea level, surge, and skew from 1960 to 2013. a Trends at the tide gauges and b the regional mean were compared, with downward bars indicating the negative trends. The boxplots were overlaid with the mean (central rod), 17-83% CI (shading box), and range of changes (whiskers) in China (CHN), Japan (JPN), Indonesia islands, and reefs century, with flooding occurring along the Japanese coast, increasing rapidly or even exceeding that of the Chinese and Indonesian coasts. Although the Indonesian reefs are generally less affected by extreme WLD, their interannual variability exposes them to coastal flooding risk in some years as well. In the past, low-probability, but high-impact ESLs have severely affected the reefs. Since the twenty-first century, reefs have been subject to more frequent ESLs with lower intensity (supplementary Fig. 1a) due to the major construction of reef protection facilities by government departments and the strengthening of coastal resilience to disasters. This shows that reef sensitivity gradually shifts from ESLs intensity to occurrence. Compared with reefs, islands are less vulnerable to ESLs, but they still suffer from increasingly frequent ESLs and the intensity also increases slightly. The seasonal differences (supplementary Fig. 1b) are reflected in the fact that high-impact ESLs occur more frequently during the winter on the islands. In the summer, the islands are rarely affected by ESLs, except for some tide gauges in the Philippines, which are affected by tropical cyclones. Reefs are exposed to frequent high-intensity ESLs for longer periods of time throughout the year, especially from August to December. With the rising sea level, serious coral losses have occurred, and many tide gauges have lost natural wave barriers, coupled with tropical cyclones, earthquakes, volcanoes, and other disasters. As a result, ESLs have seriously affected the safety of islands and reefs, and disaster control remains critical. Many low-lying megacities and Fig. 9 Climatic factor contributions to ESLs along the China (a), Japan (b), Indonesia islands (c), and reef (d) coasts. Sea level anomaly (SLA), high tide level (HTL), WLD contributions to ESLs variability are shown with their percentages, along with an annual incidence of ESLs (purple bar), where WLD manifests itself as skew (ESLs _skew , shown in the right y-axis) 1 3 small islands at almost all latitudes are predicted to experience historically rare ESLs annually by 2050 (IPCC 2019). Coastal flooding risk shows a clear decadal feature along the Northwest Pacific coast. It increases along the entire coastal zones, and most notably along the coasts of Japan and Indonesian islands. At the same time, the Chinese coast, which has been affected by WLD for a long time, is still exposed to its high impact.

Discussion and summary
Climate change has emerged as the greatest threat to human survival and the future of the planet. Under the climate change framework, an increasing number of global coasts will be exposed to a risk of ESLs. We used hourly sea-level data from 699 global tide gauges to capture ESLs variability and its climatic factors, such as WLD, high tide level, sealevel variability, and wind. The coastal zones are found to be more frequently exposed to ESLs with stronger intensity and longer duration since 1990. Dramatic ESLs occur in the subtropical western boundary current regions, Pacific islands and reefs, and Western Europe frequently, signifying that these areas are more sensitive to ESLs under the climate change framework. Among these, the Northwest Pacific stands out with stronger ESLs intensity and more occurrences than the global mean. ESLs intensity in the Northwest Pacific is 1.5-2 times that of the east coast, and occurrence increases with 557 ESLs in the 2000s. ESLs duration increases from 2.4 to 4.3 h. In the future, western boundary current regions, South Pacific, and northwestern Europe will still be hotspots vulnerable to ESLs (Vousdoukas et al. 2018a;Kirezci et al. 2020). Seasonal variations are observed for ESLs occurrence, duration, and intensity, which closely follows the seasonal pattern of sea-level variability (Stephens et al. 2020) and climate extreme events. A double peak is observed in ESLs occurrence, duration, and intensity. ESLs occur frequently and with a relatively higher intensity and longer duration in TC-summer and winter. More than 60% of ESLs that occur along the Chinese and Japanese coasts in TC-summer involve tropical cyclones, with some months even reaching 88%. In addition, ESLs intensity interannual variability is quite large (Feng et al. 2021), with decadal variations closely following the decadal sea-level and high tide level patterns. The periodic variations of high tide level play a good moderating role with sea-level variability (Thompson et al. 2021) and WLD at different timescales. ESLs fluctuated from seasonal to decadal, and even multidecadal variations.
The Northwest Pacific has been frequently affected by climate extreme events, and the impact of WLD on ESLs on short time scales has been a matter of concern. In recent years, an increasing number of studies have started to focus on water level discrepancy between the maximum observed water level and the predicted high-tide level (Feng et al. 2021;Stephens et al. 2020;Williams et al. 2016). Not only that, the time difference between the maximum observed water level and the predicted high-tide level is also a considerable climatic factor for ESLs. We divided WLD into skew and surge to comprehensively analyze the influence of WLD on ESLs, where skew represents the interactions of high tide level, WLD, and wind on ESLs. The skew water level associated with ESLs is usually higher than the surge water level. Low-frequency variations in sea level, which occur at most tide gauges, provide the background for ESLs variabilities on a seasonal time scale and secular change. In the Northwest Pacific, sea level has been rising at more than 80% of the tide gauges since the 1960s. There are also a few tide gauges along the Indonesian islands, where a negative trend in sea level is observed. Climate-induced sea-level rise will occur over the next few decades and will remain significant thereafter (Nicholls et al. 2021). In terms of magnitude, the high tide level is the major climatic factor to ESLs along the Northwest Pacific coast, whether it occurs at islands, reefs, or continental coasts. The 18.61-year tidal cycle is mutually regulated with the sealevel low-frequency fluctuation (Thompson et al. 2021), and it affects ESLs along with WLD on a short time scale.
Wind-driven piling-up at the coast is an essential driver for ESLs. About 70% of ESLs along the coasts of China and Japan are associated with tropical cyclones due to frequent tropical cyclone activity in the Northwest Pacific, especially during TC-summer. The variations in the wind field are also a key climatic factor in the study of ESLs. Ideal model experiments by Andrée et al. (2022) have shown that European ESLs increased linearly to quadratically with wind speed, which contained large uncertainty during strong wind conditions. We used the measured data to analyze WLD variability under the effect of the wind field and found a good linear relationship between wind speed and WLD. The relationship between wind direction and shoreline also plays an important role in the ESLs occurrence. Onshore winds are more likely to cause ESLs, but about 30% of ESLs are also caused by offshore winds, which are possibly related to apparent water replenishment. The coast of China is susceptible to southerly winds in the winter and northeasterly winds in the summer. However, despite significant seasonal changes in wind direction, ESLs along the coast of China are significantly influenced by northeasterly winds during both winter and TC-summer seasons. ESLs along the Japanese coast are influenced by northwesterly winds in the winter and more pronounced southwesterly winds in the summer. However, islands and reefs are susceptible to ESLs due to easterly winds, although in the summer when southwesterly winds are prevalent.
WLD, high tide level, sea-level variability, and wind are not stable with climate change, and there are great uncertainties in the interactions of different climatic factors, all of which will lead to the changes in the ESLs joint return period. For example, the TC rainfall-surge joint return periods are expected to decrease significantly along the entire southeastern coastline of the USA at the end of this century (Gori et al. 2022). We focused on comparing the changes in joint effects of climatic factors on ESLs between 1991 and 2013 with those between 1960 and 1990 in the Northwest Pacific. By comparing the ESLs joint return period in the two stages, we found that WLD, sea-level variability, and wind can shorten the ESLs joint return period at more than 70% tide gauges. In particular, the wind field variability shortens the ESLs joint return period across the entire northwestern Pacific Ocean. The ESLs joint return period is stable or even increases at most tide gauges under the influence of high tide level on account of its complex variation; however, high tide causes more frequent ESLs along the northern coasts of Japan and most reefs. ESLs may be underestimated when the interactions are ignored (Arns et al. 2015;Arns et al. 2020). The effect of individual climatic factor on ESLs joint return period could weaken when another climatic factor is introduced due to the interactions between the different climatic factors. And more tide gauges will likely exhibit completely opposite features of the ESLs joint return periods due to the interaction between climatic factors. The relative importance of the climatic factors is also reflected in the ESLs joint return period. The interaction between wind and high tide level is the most obvious, which can be characterized by skew, whose contribution to ESLs has been growing compared with surge, notably in Japan and Indonesia. The interactions among climatic factors must be taken into account.
With climate change, WLD has become increasingly important for ESLs and its long-term changes cannot be ignored in the Northwest Pacific, as also proven around the European coastline by Calafat et al. (2022). The increasing contributions of WLD to ESLs even exceeds sea-level rise at a few Indonesian tide gauges. The Northwest Pacific coast is found to be influenced by extreme WLD over the decadal timescale. After 1990, the annual incidence of ESLs where WLD manifests itself as skew is more closely related to the annual contribution of WLD. Their correlation coefficients exceed 0.5 and even approaching 0.7 (99.9% significance level) in some areas. It is precisely because of stronger seawater accumulation brought by skew that the skew annual incidence affects the contribution of WLD to annual ESLs. As WLD increases significantly, the northwest Pacific coasts have been increasingly exposed to coastal flooding risk caused by extreme WLD, especially along the Japanese coasts. The coastal flooding risk along the coast of Japan, affected by extreme WLD, has increased significantly or even exceeded that of China and Indonesia since the start of the twenty-first century. Coastal flooding risk along the Indonesian islands and reefs has also shown obvious acceleration, while the coast of China has been under constant high-impact coastal flooding risk.
were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.