Characteristics of Marine Heatwaves in the Japan/East Sea

: Marine heatwaves (MHWs) are prolonged discrete events of anomalously warm seawater observed around the world. The occurrence of MHWs in the Japan/East Sea (JES), a middle latitude marginal sea with abundant aquatic resources, has severe impacts on marine ecosystems and human society. We used satellite remote sensing products spanning 1982–2020 to investigate the mean states and variations of JES MHWs. The results show that the JES MHWs occurred twice per year. More than 40% of the MHWs were shorter than the mean duration of 12.6 days and weaker than the mean intensity of 2.4 ◦ C. Frequent MHWs were observed in August, while high mean intensities were observed in winter and spring in the western JES. In recent years, the area’s averaged yearly sum of MHW intensity in the JES has increased dramatically, with a positive trend of 29.62 ◦ C days per decade—over twice the average global trend (12.37 ◦ C days per decade), which is further conﬁrmed by its ﬁrst empirical orthogonal function mode. The strengthening trend in MHWs may be explained by oceanic dynamics rather than atmospheric forcing, and may result in increased surface latent heat ﬂux from the ocean to the atmosphere.


Introduction
A marine heatwave (MHW) is defined as a prolonged anomalously warm seawater event [1]. Sea surface temperatures in the northwestern Mediterranean Sea were more than 3 • C above the long-term average temperature during the late boreal summer of 2003, which was one of the first MHWs documented in the literature [2][3][4]. MHWs can last from several days to months and cover an area thousands of kilometers wide [1,5,6]. While numerical models are sometimes used to give a future projection of MHWs, satellite observation data is most commonly used in MHW research because of its advantages of wide space, quasi-real-time and long-term results, continuity, and high temporal and spatial resolution.

Data Description
Daily SST data from 1982 to 2020 at a 0.05° × 0.05° resolution were obtained from the Copernicus Marine Environmental Monitoring Service (CMEMS) (the data can be downloaded from https://resources.marine.copernicus.eu/product-detail/SST_GLO_SST_L4_REP_OBSERV ATIONS_010_011, accessed on 10 February 2022). The SST data were generated from satellite data and in situ observations by the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) [34].
The 10-m wind speed, surface net solar radiation (SSR), surface net thermal radiation (STR), surface sensible heat flux (SSHF), surface latent heat flux (SLHF), evaporation (E), total precipitation (P), significant height of wind waves (SWH), and significant height of total swell (SHTS) were investigated with respect to their contributions to the JES MHW trends. These data were provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 high-resolution reanalysis project at a horizontal resolution of 0.25° × 0.25° (the data can be downloaded from https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=for m, accessed on 10 February 2022) [35]. The monthly Niño, Arctic Oscillation (AO), and Pacific Decadal Oscillation (PDO) index data were collected from the Koninklijk Nederlands Meteorologisch Instituut (KNMI) climate explorer at http://climexp.knmi.nl/selectindex.cgi?id=someone@somewhere, accessed on 10 February 2022 [36]. In this study, all data described above cover the period from January 1982 to December 2020.

Data Description
Daily SST data from 1982 to 2020 at a 0.05 • × 0.05 • resolution were obtained from the Copernicus Marine Environmental Monitoring Service (CMEMS) (the data can be downloaded from https://resources.marine.copernicus.eu/product-detail/SST_GLO_SST_L4 _REP_OBSERVATIONS_010_011, accessed on 10 January 2022). The SST data were generated from satellite data and in situ observations by the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) [34].
The 10-m wind speed, surface net solar radiation (SSR), surface net thermal radiation (STR), surface sensible heat flux (SSHF), surface latent heat flux (SLHF), evaporation (E), total precipitation (P), significant height of wind waves (SWH), and significant height of total swell (SHTS) were investigated with respect to their contributions to the JES MHW trends. These data were provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 high-resolution reanalysis project at a horizontal resolution of 0.25 • × 0.25 • (the data can be downloaded from https://cds. climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=form, accessed on 10 January 2022) [35]. The monthly Niño, Arctic Oscillation (AO), and Pacific Decadal Oscillation (PDO) index data were collected from the Koninklijk Nederlands Meteorologisch Instituut (KNMI) climate explorer at http://climexp.knmi.nl/selectindex.cgi?id= someone@somewhere, accessed on 10 January 2022 [36]. In this study, all data described above cover the period from January 1982 to December 2020.

Detection of Marine Heatwaves
Hobday et al. [1] defined an MHW as a prolonged discrete anomalously warm event that lasts for at least 5 days and throughout which the SST is higher than the 90th percentile ( Figure 2). For each grid point (Figure 3), the climatological mean and threshold were calculated from the daily SST data within an 11-day window centered on each calendar day (to ensure an adequate sample size) and then smoothed using a 31-day moving average [18]. Consecutive events that were less than 2 days apart were classified as the same event.

Detection of Marine Heatwaves
Hobday et al. [1] defined an MHW as a prolonged discrete anomalously warm ev that lasts for at least 5 days and throughout which the SST is higher than the 9 percentile ( Figure 2). For each grid point (Figure 3), the climatological mean threshold were calculated from the daily SST data within an 11-day window centered each calendar day (to ensure an adequate sample size) and then smoothed usin 31-day moving average [18]. Consecutive events that were less than 2 days apart w classified as the same event.   The climatological mean SST at any point grid, T clm , is calculated as follows: where T is the daily SST on day d of year y (y s y y e ), j is the day of the year, y s and y e are the start and end dates of the climatological base period, respectively. The seasonally varying threshold SST at each grid, T 90 , is calculated as follows: where P 90 is the 90th percentile and X(j) = T(y, d) y s y y e , j − 5 d j + 5 is the SST set. The T m and T s are the 31-day smoothed T clm and T 90 , respectively. In this study, y s and y e are 1982 and 2020.

Marine Heatwave Indices
Once an MHW event at any grid point is detected, a set of metrics can be calculated to represent MHW characteristics, including duration and intensity. The duration of the MHW (MHWD) is the time period between the start (t s ) and end (t e ) date of the MHW. The intensity is the SST anomaly relative to the 31-day smoothed climatological mean (T m ). The I max and I mean are the maximum and mean, respectively, of daily intensities during the MHW [1].   An MHW event in one year can last into the following year ( Figure 3d). Therefore, the number of MHWs (MHWN) in this study is the event counts from 1982 to 2020. The MHWT, MHWI, and MHWIM are the sum of MHW-day (t MHW , the day defined as an MHW), the sum of MHW-intensity (I MHW , the SST anomaly at t MHW ), and the mean MHW-intensity from the start (D s ) to the end (D e ) date of the targeted period, respectively. In the following analysis, the MHWT, MHWI, and MHWIM over a certain month are denoted as MHWT m , MHWI m , and MHWIM m . Meanwhile, the MHWT, MHWI, and MHWIM over a certain year are denoted as MHWT y , MHWI y , and MHWIM y . The MHWIM reflects the mean intensity of MHWs of each day over the targeted period and actually determines the MHWI with consideration of the MHWT. All of the above MHW indices are described in detail in Table 1.

Ekman Pumping
The Ekman pumping vertical velocity is calculated as follows: where τ is the wind stress calculated from the 10-m wind data of ECMWF ERA5, ρ 0 is the mean sea water density, which was taken to be 1.025 × 10 3 kg/m 3 , and f is the Coriolis parameter.

Net Surface Heat and Freshwater Flux
The net surface heat flux (SHF) is an important variable for understanding air-sea interactions. It is expressed as the following equation: The freshwater flux (SWF), which is the difference between the E and P, is computed as follows: The SSR, STR, SSHF, SLHF, E, and P data were all obtained from the ECMWF ERA5, with the downward direction set to be positive. That is, a positive (negative) SHF represents ocean gain (loss) of heat, and a positive (negative) SWF represents precipitation (evaporation).

Empirical Orthogonal Function Analysis
The Empirical Orthogonal Function (EOF) analysis is commonly used for time series analysis to reveal possible spatial patterns of climate variability and how they change with time, and was conducted using the Climate Data Toolbox for MATLAB developed by Greene et al. [37]. The EOF method was used to reveal the spatiotemporal variability in the JES MHWs and explain the MHW variability in the South China Sea [18] and off of eastern Tasmania [20]. In this study, MATLAB software was used for calculation and analysis.

Statistical Characteristics
On average, there were about two MHWs in the JES during the period from 1982 to 2020. A high MHWN occurred in the southwestern JES and a low MHWN occurred in the Japan Basin (Figure 4a). The average MHWT y peaked at more than 26 days in the western JES at approximately 40 • N (Figure 4b). Except for the Tsushima Strait, the mean MHWD was more than 10 days over the entire JES and was more than 15 days in the central JES and along the North Korea coast (Figure 4c). The longest MHW event occurred off the east coast of North Korea, lasting more than 200 days from August 2019 to the spring of 2020. The spatial patterns of the mean I max and I mean are similar to that of the average MHWT (Figure 4d   Lifetime and intensity are important parameters for describing the characteristics of MHWs. The mean MHWD and mean I mean of all MHWs in the JES were 12.6 days and 2.4 • C, respectively, which are represented by the dot-dash line and the dashed line, respectively, in Figure 5. The probability density function (PDF) distribution is divided into four regions by these two lines. Beginning at the top right quadrant and moving counterclockwise, the values account for cumulative PDFs of 17.8%, 12.9%, 41.8%, and 27.5%. These statistical results show that MHWs of short duration and weak intensity prevail in this region, where the most frequently occurring MHWs have a duration of 5 days and a mean intensity of 2.1 • C.

Seasonal Variability
The average JES MHWT m varied from 0 to more than 6 days, showing strong seasonal variability ( Figure 6). MHW events occurred in all seasons in the JES, with a high MHWT m in summer (from June to August), and a low MHWT m in spring (from February to May). In January and February, the spatial distribution of the MHWT m was similar, with low values in the Tartar Strait and the southeastern coastal region (Figure 6a,b). In April, the MHWT m in the Japan Basin decreased significantly ( Figure 6d). The MHWT m was at a low level from April to May, with high values (more than 2 days) in the North Korea coastal area (Figure 6d

Seasonal Variability
The average JES MHWTm varied from 0 to more than 6 days, showing str seasonal variability ( Figure 6). MHW events occurred in all seasons in the JES, wi high MHWTm in summer (from June to August), and a low MHWTm in spring (f February to May). In January and February, the spatial distribution of the MHWTm similar, with low values in the Tartar Strait and the southeastern coastal region (Fig  6a,b). In April, the MHWTm in the Japan Basin decreased significantly ( Figure 6d). MHWTm was at a low level from April to May, with high values (more than 2 days the North Korea coastal area (Figure 6d,e). Spatially, the MHWTm in June showe meridional distribution pattern of being high in the south and low in the north (Fig  6f). The MHWTm increased in July, except in the southeastern JES ( Figure 6g). frequency and spatial extent of MHWs both reached their annual maximums in Aug (Figure 6h). The MHWTm in June showed a zonal spatial distribution pattern of be high in the east and low in the west (Figure 6i). From October to December, the level spatial distribution of the MHWTm were roughly similar to those from January to Ma (Figure 6j-l). In contrast to the MHWT m and MHWI m , the seasonal variability of the average MHWIM m was reflected mainly in its spatial distribution (Figure 7). In winter and spring, the MHWIM m was spatially distributed more inhomogeneously than in summer and autumn. The MHWIM m varied from 0.01 • C to more than 5 • C in winter, with high values in the western JES and low values in the Tartar Strait and southeastern JES. In spring, the inhomogeneity of the spatial distribution decreased, the maximum MHWIM m decreased to 4 • C, and the minimum MHWIM m increased to 1.7 • C. The variation range and regional differences of the MHWIM m were the smallest in summer. The inhomogeneity increased in September, with a high-value area in the eastern JES. In October and November, the distribution of a high MHWIM m was scattered. A low MHWIM was observed mainly in the southeastern JES. Figure 8 shows that the average MHWI m was high in summer and low in spring, which is consistent with the seasonal variation of MHWT m . By comparing the seasonal variation of the spatial distribution of MHWI m and MHWT m , it can be found that the distribution characteristics of the two are relatively similar between June and October. This pattern means that the MHWI m was mainly affected by the MHWT m in this period. In the other months, a high MHWI m (more than 5 • C days) was observed in the subpolar front area around 40 • N, especially in the coastal regions near North Korea.
Although the spatial distributions of the average MHWT m , MHWI m , and MHWIM m are different, they have similar seasonal variations in temporal distribution, with the maximums appearing in summer ( Figure 9). Due to the high frequency of MHWs in August, the MHWT m reached a maximum of 4.6 days. Because of the wide spatial extent of MHWs in August, the MHWI m reached a maximum of 12.6 • C days. This explains why many studies focus only on the summer MHWs [18]. However, the seasonal variation in the MHWIM m is weak, with the maximum of 2.8 • C occurring in July and the minimum of 2.0 • C occurring in February. In contrast to the MHWTm and MHWIm, the seasonal variability of the a MHWIMm was reflected mainly in its spatial distribution (Figure 7). In wint spring, the MHWIMm was spatially distributed more inhomogeneously than in s and autumn. The MHWIMm varied from 0.01 °C to more than 5 °C in winter, wi values in the western JES and low values in the Tartar Strait and southeastern spring, the inhomogeneity of the spatial distribution decreased, the ma MHWIMm decreased to 4 °C, and the minimum MHWIMm increased to 1.7 ° variation range and regional differences of the MHWIMm were the smallest in su The inhomogeneity increased in September, with a high-value area in the eastern October and November, the distribution of a high MHWIMm was scattered. A MHWIM was observed mainly in the southeastern JES.   variation of the spatial distribution of MHWIm and MHWTm, it can be found that distribution characteristics of the two are relatively similar between June and Octo This pattern means that the MHWIm was mainly affected by the MHWTm in this per In the other months, a high MHWIm (more than 5 °C days) was observed in the subp front area around 40°N, especially in the coastal regions near North Korea. August, the MHWTm reached a maximum of 4.6 days. Because of the wide spatial ex of MHWs in August, the MHWIm reached a maximum of 12.6 °C days. This expl why many studies focus only on the summer MHWs [18]. However, the seas variation in the MHWIMm is weak, with the maximum of 2.8 °C occurring in July and minimum of 2.0 °C occurring in February.

Interannual Variability and Long-Term Trends
In addition to seasonal variations, the JES MHWs also showed interann variabilities during the 1982-2020 period ( Figure 10). The frequency of MHWs in the was relatively high, with MHWs occurring in all months except for June 1983, Octo 1984, June 1986, and May 1987. Moreover, the MHW frequency increased significantl the most recent years, which is clearly reflected in the temporal distribution of the h MHWTm and MHWIm. With the exception of 2019 and 2020, the JES MHWs tende

Interannual Variability and Long-Term Trends
In addition to seasonal variations, the JES MHWs also showed interannual variabilities during the 1982-2020 period ( Figure 10). The frequency of MHWs in the JES was relatively high, with MHWs occurring in all months except for June 1983, October 1984, June 1986, and May 1987. Moreover, the MHW frequency increased significantly in the most recent years, which is clearly reflected in the temporal distribution of the high MHWT m and MHWI m . With the exception of 2019 and 2020, the JES MHWs tended to occur in summer. Compared to the interannual variation of the MHWT m and MHWI m , the interannual variation of the MHWIM m is weak.
Significant positive trends in the MHWT y occurred almost throughout the JES region (Figure 11a). A rapidly increasing trend of over 15 days/decade was found in the area south of 45 • N in the JES. The annual regionally averaged MHWT y time series shows a rapidly increasing trend at 12.01 ± 3.48 days/decade, passing over a significance level of p < 0.01 (Figure 11b). Positive trends in the MHWI y were also found almost across the entire area of the JES, with high trends of more than 90 • C days/decade appearing in the western corner of the JES (Figure 11c). Additionally, the annual area-averaged MHWI y time series shows a rapid growth trend of 29.62 ± 9.00 • C days/decade (p < 0.01), which is consistent with the annual area-averaged MHWT y (Figure 11d). Although the trends of the MHWIM y are basically positive in the whole JES, the linear trend in the regional averaged time series is only 0.07 ± 0.07 • C/decade, with a relatively weak significance level of p < 0.05 (Figure 11e,f)  Significant positive trends in the MHWTy occurred almost throughout the region (Figure 11a). A rapidly increasing trend of over 15 days/decade was found in area south of 45°N in the JES. The annual regionally averaged MHWTy time series sh a rapidly increasing trend at 12.01 ± 3.48 days/decade, passing over a significance l of p < 0.01 (Figure 11b). Positive trends in the MHWIy were also found almost across entire area of the JES, with high trends of more than 90 °C days/decade appearing in western corner of the JES (Figure 11c). Additionally, the annual area-averaged MH time series shows a rapid growth trend of 29.62 ± 9.00 °C days/decade (p < 0.01), whic consistent with the annual area-averaged MHWTy (Figure 11d). Although the trend the MHWIMy are basically positive in the whole JES, the linear trend in the regio averaged time series is only 0.07 ± 0.07 °C/decade, with a relatively weak significa level of p < 0.05 (Figure 11e,f). The MHWI y , including the properties of frequency, duration, and mean intensities of the MHWs, is a comprehensive index for measuring MHWs [18]. Therefore, EOF analysis was conducted on the MHWI y to reveal the spatiotemporal patterns of JES MHWs. The first three leading modes respectively account for 44.89%, 15.49%, and 6.23% of the total variance. The first EOF mode (EOF1) essentially represents basin-scale trends in the JES MHWI y (Figure 12). The EOF1 pattern is reminiscent of the spatial distribution of trends in the MHWI y , showing positive values throughout the JES, with high values in the western corner and center of the JES (Figures 11c and 12a). The corresponding time coefficients of EOF1 (PC1) show an increasing trend above the 99% significance level (p < 0.01), with a turning point in approximately 2004. The MHWIy, including the properties of frequency, duration, and mean inten of the MHWs, is a comprehensive index for measuring MHWs [18]. Therefore, analysis was conducted on the MHWIy to reveal the spatiotemporal patterns of MHWs. The first three leading modes respectively account for 44.89%, 15.49%, 6.23% of the total variance. The first EOF mode (EOF1) essentially represents basintrends in the JES MHWIy ( Figure 12). The EOF1 pattern is reminiscent of the sp distribution of trends in the MHWIy, showing positive values throughout the JES, high values in the western corner and center of the JES (Figures 11c and 12a). corresponding time coefficients of EOF1 (PC1) show an increasing trend above the significance level (p < 0.01), with a turning point in approximately 2004.

Factors Associated with Long-Term Trends in Marine Heatwaves
Previous investigations have qualitatively discussed the potential var associated with the occurrence of MHWs, including climate modes, air-sea heat fl mixing processes, and Ekman pumping [5,33,[38][39][40][41]. The correlations between the and the PDO index, the AO index, and the Niño index are −0.29, 0.13, and

Factors Associated with Long-Term Trends in Marine Heatwaves
Previous investigations have qualitatively discussed the potential variables associated with the occurrence of MHWs, including climate modes, air-sea heat fluxes, mixing processes, and Ekman pumping [5,33,[38][39][40][41]. The correlations between the PC1 and the PDO index, the AO index, and the Niño index are −0.29, 0.13, and −0.01, showing that the PDO, AO, and El Niño-Southern Oscillation (ENSO) had no significant correlation with the JES MHWs. However, the average MHWT y values of the PDO positive phase years were relatively lower than those of the PDO negative phase years. Furthermore, the spatial distribution of the average MHWT y values between positive and negative AO phase years is different. Figure 13c shows a dipole-like pattern, with high values identified in the northwestern JES and low values identified in the southern and southeastern JES; the pattern is shifted in Figure 13f. Compared to the PDO and AO, the ENSO had a relatively weak effect on the JES MHWs (Figure 13d,g). We explored variations in the SHF, SWF, wind speed, and , as shown in F 14. The SHF of almost the entire JES region shows significant negative trends, indic that the ocean was sending heat to the atmosphere, especially in the area of the Tsu Strait (−30 W/m 2 /decade) (Figure 14a). The annual regionally averaged SHF time in the JES shows a decreasing trend of −5.37 ± 2.31 W/m 2 /decade at a significance le p < 0.01 (Figure 14b). Meanwhile, the SWF and 10-m wind speed have negative positive trends in the area of the Tsugaru Strait, with area-averaged trends of 0.11 We explored variations in the SHF, SWF, wind speed, and w E , as shown in Figure 14. The SHF of almost the entire JES region shows significant negative trends, indicating that the ocean was sending heat to the atmosphere, especially in the area of the Tsugaru Strait (−30 W/m 2 /decade) (Figure 14a). The annual regionally averaged SHF time series in the JES shows a decreasing trend of −5.37 ± 2.31 W/m 2 /decade at a significance level of p < 0.01 (Figure 14b). Meanwhile, the SWF and 10-m wind speed have negative and positive trends in the area of the Tsugaru Strait, with area-averaged trends of 0.11 ± 0.09 mm/d/decade and −0.04 ± 0.03 m/s/decade, respectively. Due to the inhomogeneous spatial distribution of the trends in w E , the corresponding area-averaged trend is relatively weak, below the 90% significant level. Since a significant increasing trend of the annual area-averaged time series of the SHF was observed, we studied the variations of each component of the SHF further. The weak increasing trends of the SSR were observed in most of the JES except for the area of Tsugaru Strait, with a regionally averaged linear trend of 0.70 ± 0.72 W/m 2 /decade (Figure 15a,b). The STR shows significant decreasing trends in the northeastern JES region, indicating that the surface net upward longwave flux had an increasing trend (Figure 15c). The annual regionally averaged STR time series in the JES shows a decreasing trend of −0.76 ± 0.42 W/m 2 /decade at a significance level of p < 0.01 (Figure 15d). Meanwhile, the decreasing trends of the SSHF also hold in the eastern JES, The weak increasing trends of the SSR were observed in most of the JES except for the area of Tsugaru Strait, with a regionally averaged linear trend of 0.70 ± 0.72 W/m 2 /decade (Figure 15a,b). The STR shows significant decreasing trends in the northeastern JES region, indicating that the surface net upward longwave flux had an increasing trend (Figure 15c). The annual regionally averaged STR time series in the JES shows a decreasing trend of −0.76 ± 0.42 W/m 2 /decade at a significance level of p < 0.01 (Figure 15d). Meanwhile, the decreasing trends of the SSHF also hold in the eastern JES, with area-averaged trend of −1.27 ± 1.10 W/m 2 /decade above a 95% significant level (Figure 15e,f). Decreasing trends in the SLHF were found across almost the entire area of the JES, indicating that the surface upward latent heat flux of the JES had positive trends from 1982 to 2020 (Figure 15g). A relatively faster trend of −4.05 ± 1.17 W/m 2 /decade was observed in the area-averaged SLHF above a 99% significant level (Figure 15h). In addition, we explored the correlations between the PC1 and the variables and processes associated with air-sea interaction ( Figure 16). As shown in Figure 16a, the SHF in most areas of the JES was negatively correlated with PC1, with a high correlation coefficient of −0.7 (p < 0.01) in the west of the Tsugaru and Soya Straits, indicating that the ocean had a forcing effect on the atmosphere. Compared to the SSR, STR, and SSHF, the SLHF had a relatively stronger correlation with the PC1, showing that the ocean affected the atmosphere mainly by releasing surface latent heat flux upward. In addition, we explored the correlations between the PC1 and the variables and processes associated with air-sea interaction ( Figure 16). As shown in Figure 16a, the SHF in most areas of the JES was negatively correlated with PC1, with a high correlation coefficient of −0.7 (p < 0.01) in the west of the Tsugaru and Soya Straits, indicating that the ocean had a forcing effect on the atmosphere. Compared to the SSR, STR, and SSHF, the SLHF had a relatively stronger correlation with the PC1, showing that the ocean affected the atmosphere mainly by releasing surface latent heat flux upward. Meanwhile, the weak negative correlation between the SWF and the PC1 was mainly due to the increase in E, which is consistent with the variation of the SLHF. However, the mixing processes had no significant impact on the positive trend of the JES MHWI y (Figure 16i-l).

Discussion and Conclusions
The MHWs in the JES are samples of enhancing MHWs over the world. As shown in Figure 17a, compared to the rest of the world, the MHW frequency in the JES is at a

Discussion and Conclusions
The MHWs in the JES are samples of enhancing MHWs over the world. As shown in Figure 17a, compared to the rest of the world, the MHW frequency in the JES is at a moderate level, with an area-average MHWN of 1.97 occurrences, which is slightly higher than the global averaged level of 1.82 occurrences (Table 2). Globally, except among the Arctic and Antarctic regions, the average MHWT y ranges around 23.92-30.77 days, similar to the JES. The increasing trend of the area-averaged MHWT y in the JES is about 12.01 days/decade, exceeding that of the global averaged MHWT of 8.54 days/decade. Although the spatial distributions of increasing trends in the MHWT y and the MHWI y are similar in the JES (Figure 17c,d), the increasing trend in the MHWI y is relatively strong, reaching 29.62 • C days/decade, which is slightly slower than that in the Mediterranean Sea and in the western North Pacific Ocean. Taking into consideration the rather high baseline of the MHWI y , which already reached an area-averaged value of 59.65 • C days, this significant increasing trend of the MHWI y highlights the importance of MHWs in the JES and suggests the need for further studies.    Many studies have revealed increasing trends in the frequency, duration, and intensity of MHWs due to global warming [31,32]. Indeed, as the definition of MHWIy index goes, the PC1 of the MHWI y coincided with the area-averaged yearly SST anomalies, with a significant correlation coefficient of 0.77 (p < 0.01). Compared to the global averaged SST trend (0.12 • C/decade at a significance level of p < 0.01), the SST trend in the JES is much stronger, especially in the central JES ( Figure 1). Furthermore, even if the global averaged SST trend is removed from the daily SST for each grid point, the increasing trends in the MHWT y and MHWI y also hold throughout the JES after rerunning the MHW analysis following the same procedures as we did for the original SST data, with area-averaged increasing trends of 6.07 days/decade and 16.69 • C days/decade, respectively.
MHWs show that various scale changing patterns and climate modes are promising candidates to explain the causes of variation in MHWs. Indeed, the results show that the PDO, AO, and the El Niño-Southern Oscillation (ENSO) have no significant correlation with the significant increasing trend in the MHWI y [14]. However, the average MHWT y values in the positive PDO phase years were higher than those in the negative PDO phase years. During the positive phases of the PDO, the JES was generally in a cooler state in the upper layer, thereby suppressing the occurrence of MHWs [42][43][44]. A positive AO phase generally led to a weaker East Asia winter monsoon and a warmer winter over the JES [45][46][47]; as a result, the MHWs in the northern JES were slightly more active than in the southern JES because of the weakened wind speeds and decreased sea ice induced by the positive AO phase.
Besides climate modes, some of the most likely causes of marine heatwaves are ocean currents. Along the Tsushima Strait, there is a perennial circulation from the southwest toward the northeast, referred to as the Tsushima Warm Current (TSWC), which is the only inflow for the JES, a semi-enclosed marginal sea. Additionally, the majority of the TSWC flows out through the Tsugaru Strait into the Pacific Ocean, and through the Soya Strait into the Okhotsk Sea, known as the Tsugaru Warm Current and the Soya Warm Current, respectively [48,49]. Therefore, the water temperature anomalies are closely related to the variations in the TSWC of both the volume transport and the pathways. Under positive PDO conditions, the water transport of the TSWC was decreased, leading to a reduction in the heat supply to the JES, and thereby resulting in relatively fewer MHWs than under the negative phase of the PDO [42,43]. Besides variation and along with climate modes, an increasing trend of 0.1 Sv per decade was observed not only in the ADCP measurements but also in the sea-level differences across the Tsushima Strait [50]. However, due to the scarcity of in situ observations, it is too soon to jump to the conclusion of the relationship between the ocean currents and MHWs, which may need to be further evaluated by high-resolution coupled models.
The strengthening trends of MHWs may have vital importance for human activities. The long-term trend in JES MHWs reflects that the ocean has a forcing effect on the atmosphere, which causes increases in the surface temperatures in the surrounding lands, increasing the occurrence of land heatwaves [51]. Moreover, the high amplitude area in the EOF1 of the MHWI y is located in the western JES around 40 • N, where fishing resources are rich due to the interaction of warm and cold currents. The warmer environment could alter the habitats of fishes, and the likely induced HABs could result in fish deaths. Due to a sudden increase in the water temperature in the upper layer (50-100 m) in 1987-1989, a shift of fish assemblage structure occurred in 1998-1989, such that the common squid became the dominant species of commercial fish catches (over 60%) instead of filefish, anchovy, and sardine in the Tsushima Basin after 1993 [52]. The MHWs' environmental and economic significance may be also subjected to MHWs' influential depth, besidesds their duration and intensity. In addition, strong subsurface MHWs have been observed in the tropical western Pacific Ocean based on buoy data [27]. However, due to the lack of temperature profile data, the depth of the impact of the MHWs is unclear so far and requires further comprehensive field observation.

Summary
In this study, we used newly released high-resolution satellite daily SST data from 1982 to 2020 to investigate the characteristics of JES MHWs, including the mean states, seasonal and interannual viabilities, and the long-term trends. Our results are summarized as follows: (1) JES MHWs occurred about twice per year, with a mean MHWD and a mean I mean of 12.6 days and 2.4 • C, respectively. MHWs with a short duration and weak intensity dominated in this region. A duration of 5 days and a mean intensity of 2.1 • C were particularly common. High MHWT y and high SST anomalies were both observed in the western JES at approximately 40 • N.
(2) Both the spatial and temporal distributions of MHW indices were subject to strong seasonal variability. The maximum values of the regional averaged MHWT m , MHWI m , and MHWIM m (4.6 days, 12.6 • C days, and 2.8 • C, respectively) all occurred in August. However, a high MHWI m and MHWIM m in the western corner of the JES also appeared in winter (from December to February) and spring (from May to March). The MHWs in the eastern JES were also active in September. (3) The MHWT y and MHWI y have increased significantly in recent years over the entire JES, except for the Russian coastal regions and the Tsushima Strait. Significantly positive trends were observed in the MHWT y (12.01 ± 3.48 days/decade) and the MHWI y (29.62 ± 9.00 • C days/decade), which are still significant when the global averaged SST trend is removed. A prolonged MHW event in the western JES lasting more than 200 days (from August 2019 to spring of 2020) is an example of this increasing trend. The increasing trend in the MHWIM y (0.07 ± 0.07 • C/decade) is rather trivial. The EOF1 of the MHWI y accounts for 44.89% of the total variance. The corresponding time coefficient shows a significantly increasing trend. (4) The average JES MHWN was 1.97 occurrences, which is slightly higher than the global average level of 1.82 occurrences. However, regional comparison results indicate that the increasing trend of the MHWI y in the JES is twice that of the global averaged MHWI trend of 12.37 • C days/decade.