Response of Future Summer Marine Heatwaves in the South China Sea to Enhanced Western Pacific Subtropical High

Marine heatwaves (MHWs) are prevalent in the South China Sea (SCS) and are primarily caused by the anomalous western Pacific subtropical high (WPSH) which suppresses the SCS summer monsoon. Our study investigates future summer MHWs in the SCS in response to the enhanced WPSH. The results show that the WPSH will be strengthened in most models under the highest emission scenario. Total days of summer SCS MHWs will significantly increase with the enhancement of the WPSH and peak at the end of the 21st century. The high‐value area of total days of summer MHWs (>50 days) appears in the south‐central SCS. The intensity of MHWs exhibits little response to the strengthening WPSH. In addition, the weakening of the SCS cold filament due to the anomalous WPSH also favors summer MHWs. Furthermore, increased future severe summer SCS MHWs will create a higher bleaching risk for coral reefs.

• Increasing variability in the western Pacific subtropical high (WPSH) leads to more frequent, stronger WPSH events in future projections • The total days and intensity of summer marine heatwaves (MHWs) in the South China Sea (SCS) show different responses to the enhanced WPSH • Total days of summer MHWs in the SCS increase with the strengthened WPSH, resulting in a high coral bleaching risk

Supporting Information:
Supporting Information may be found in the online version of this article. effect provided by this upwelling and cold filament will result in robust basin-wide surface warming, which is conducive to the occurrence of MHWs (Xie et al., 2003(Xie et al., , 2007Yao & Wang, 2021).
Recent studies have revealed that the WPSH is significantly enhanced under different future projections (X. Chen et al., 2020;Cherchi et al., 2018;Preethi et al., 2017). Compared to the previous century, both the variability and the frequency of strong WPSH events will increase in the 21st century under the highest emission scenario (Yang et al., 2022). However, how summer MHWs in the SCS will respond to the future strengthened WPSH has not been fully studied. In this paper, we explore the response of future summer MHWs in the SCS to the variability in the WPSH under the highest emission scenario. This work provides insight into future changes in summer MHWs in the SCS and their impacts on marine ecosystems.

Data Sources
The daily SST data used in this study are obtained from the National Oceanic and Atmospheric Administration (NOAA) Optimum Interpolation SST data set version 2.1 (OISST v2.1) (Huang et al., 2021). The product has a daily 0.25° × 0.25° horizontal resolution and the data set covers the period of 1982-2020.
Monthly atmospheric reanalysis data are from the fifth generation of the European Center for Medium-Range Weather Forecasts atmospheric reanalysis data set (ERA5) with a horizontal resolution of 1° × 1° (Hersbach et al., 2019). Monthly atmospheric variables, including 10 m surface wind, pressure levels geopotential height and horizontal wind for the period of 1982-2020, are used.
The future projections are based on the available global climate model data collected from the Coupled Model Intercomparison Project Phase 6 (CMIP6) multi-model database. The models are selected from historical simulations and future greenhouse gases under the Shared Socioeconomic Pathways 5-8.5 (SSP5-8.5) emission scenario. The selected variables include daily sea surface temperature ("tos"), monthly pressure levels horizontal wind ("ua" and "va") and monthly 10 m surface wind ("uas" and "vas"). The period of "tos" for the historical simulations is 1985-2014 and for the SSP5-8.5 is 2015-2100. The period of historical simulations for monthly wind data are from 1941 to 2014, and the SSP5-8.5 data are from 2015 to 2100. All oceanic model data are regridded to 0.25° × 0.25°, and all atmospheric model data are regridded to 1° × 1°. The first ensemble member (r1i1p1f1) is chosen. In particular, we use 19 CMIP6 models with horizontal resolutions of daily SST ≤ 1° (Table S1 in Supporting Information S1).
All reanalysis datasets and CMIP6 model datasets have been detrended to remove the effects of global warming. Boreal summer in this study includes June, July, and August.

MHW Calculation and Model Selection
A MHW event is defined as SST warmer than the 90th percentile based on the baseline period of 1985-2014 for at least five consecutive days (Hobday et al., 2016). Continuous events less than 2 days apart from other events are classified as the same event. The climate threshold is calculated from data within an 11-day window centered on each calendar day and then smoothed with a 31-day moving average. Due to the vast number of coral reefs in the SCS, it is more reasonable to utilize a fixed baseline definition for SCS MHW in order to better understand its impact on these reefs in the future (Amaya et al., 2023;Oliver et al., 2021;Smith et al., 2023;. Two metrics are used: total days of MHWs per summer (HWT) and average MHW intensity per summer (HWI) (Yao & Wang, 2021). This study only focuses on summer MHWs. The categorization of MHWs is referenced by Hobday et al. (2018) and Kajtar et al. (2021). An MHW category can be defined based on a multiple of the local difference between the fixed climatological mean and the climatological 90th percentile. Based on the HWI, the magnitude of scale descriptors can be assigned at each point in time and space for an MHW event, which are defined as moderate (1-2X), strong (2-3X), severe (3-4X), and extreme (>4X).
We screen CMIP6 models by comparing the historical simulations of summer SST, HWT, and HWI from 1985 to 2014 with OISST data set. A model will be retained if the difference between the 30-year average of the model's simulation for the three indices and the observational data is less than twice the standard deviation.

Indices of the WPSH
Previous studies have shown that the 850-hPa stream function can characterize the variability in the WPSH (Hong et al., 2014). Here, we apply an empirical orthogonal function (EOF) to summer 850-hPa stream function anomalies over the northwest Pacific domain (10°-40°N, 110°−180°E) (Yang et al., 2022). The normalized first principal component time series (PC1) is used to present the variability in the WPSH.

Wind Stress and Wind Stress Curl
The wind stress is derived from 10 m surface wind data and the formula are as follows: where ρ air is the air density that is 1.33 kg/m 3 , C d is the drag coefficient (Wu, 1982), and U 10 is the 10 m wind speed. Here are the formulas for defining wind stress curl: In these equations, u and v are the 10 m eastward and northward wind components, respectively. τ x and τ y represent eastward and northward wind stress components.

Severe Summer MHWs in the SCS Related to the Strong WPSH
We apply EOF analysis to stream function anomalies in the observations during 1982-2020 over the northwest Pacific domain to characterize the variability in the WPSH. The first EOF pattern can explain 77.5% of the variance, while the second only explains 9.2% of the variance ( Figure S1 in Supporting Information S1). The standardized PC1 describes the variability in the WPSH. To further verify its accuracy, we also calculate the regional average of 850-hPa stream function anomalies over 15°-25°N, 120°−150°E as well as 850-hPa eddy geopotential height anomalies over the same domain (B. Wang et al., 2013;Xie et al., 2009) (Figure 1a). PC1 is highly correlated with the regional averages of stream function anomalies and eddy geopotential height anomalies. The correlation of the former can reach 0.95, and the latter is close to 0.73. Therefore, PC1 can effectively characterize the variability in the WPSH.
The strong WPSH events is defined as PC1 with a standard deviation larger than 1.5. PC1 can identify the strong WPSH years that occur in 1995, 1998, 2010, and 2020 ( Figure 1a). The high-value years of the regional average of HWT in the SCS correspond well to the strong WPSH years identified by PC1 ( Figure 1b). The correlation between PC1 and HWT reaches 0.485 (p < 0.01). Figures 1c and 1d show the composition of HWT and HWI in the SCS in strong WPSH years. The maximum value of HWT can exceed half a summer (>50 days), and the high-value area is mainly located in the south-central SCS (e.g., Nansha Islands). The HWI in most of the SCS exceeds 1°C and reaches a maximum of 1.8°C. High-value areas of HWI occur off Vietnam's east coast as well as the eastern SCS. Thus, strong WPSH will lead to severe MHW in the SCS. Furthermore, the HWT is strongly correlated with the WPSH. The question is how the WPSH will change in the next 80 years.

Future Summer SCS MHWs Respond to Enhanced WPSH
To investigate the future changes in the WPSH variability, we also apply EOF analysis to the 850-hPa stream function anomalies of 19 CMIP6 models from 1941 to 2100 and obtain a time series for WPSH variability for 160 years. The time series of PC1 is divided into two periods (1941-2020 and 2021-2100), and its variability is compared for 80 years before and after ( Figure S2 in Supporting Information S1). There are 18 models with an increased standard deviation of the WPSH for the next 80 years, implying increased WPSH variability in the future. In addition, the number of PC1 above 1.5 standard deviations increased in 12 models in the latter 80 years. Thus, 12 model show an increase in both variability and number of strong WPSH (Figure 2 and marked * models in Table S1 of Supporting Information S1). Furthermore, for these 12 models ensemble mean, there is a substantial increase in decadal variability of the WPSH over the next 80 years. Additionally, the WPSH after the 2060s is significantly stronger than that of the preceding 40 years ( Figure S3 in Supporting Information S1). It can thus be concluded by climate models that the WPSH will be stronger in the next 80 years. How will future summer MHWs in the SCS respond to the increased variability in the WPSH?
We next select climate models that can more accurately simulate future SCS MHWs from 12 WPSH-enhanced models by comparing models' SST, HWT, and HWI with the OISST data set. We retain the models with all of the differences less than 2 times standard deviation during 1985-2014. The results show that seven climate models (AWI-CM-1-1-MR, EC-Earth3, EC-Earth3-CC, EC-Earth3-Veg, GFDL-ESM4, KIOST-ESM, and MPI-ESM1-2-HR; Marked * models in Table S2 of Supporting Information S1) can simulate the trend and spatial pattern of summer MHWs in the SCS during the historical period, especially the multi-model ensemble mean results ( Figures S4 and S5 in Supporting Information S1). Therefore, these seven models possess both the characteristics of enhanced WPSH in the future and the capability to simulate the trends and spatial patterns of SCS MHW accurately. Our analyses of future summer MHWs in the SCS are based on these seven models with the multi-model ensemble mean results.
Due to the intensification of the WPSH, high-value areas of HWT in the future strong WPSH years are much larger than historical strong WPSH years, and nearly the entire southern SCS lasts for more than half a summer (>50 days; Figure 3a). High-value areas of HWI mainly appear in the central SCS, Hainan Island, and Taiwan Strait. The maximum HWI is close to 1.8°C, which is equivalent to the mean of past strong WPSH years ( Figure 3b). However, if the WPSH is normal, the corresponding HWT of the SCS will decrease, but the HWI Figure 1. (a) Normalized PC1 of 850-hPa stream function anomalies (PC1, black curve) over 10°-40°N, 110°−180°E and normalized regional averages of 850-hPa stream function anomalies (Regional SF, orange curve) and 850-hPa geopotential height anomalies (eddy GpH, cyan curve) over 15°-25°N, 120°−150°E during 1982-2020 in observed datasets. The green dashed line shows the 1.5 standard deviation. Years with PC1 larger than 1.5 standard deviation are identified as 1995, 1998, 2010, and 2020. (b) Annual variation in the regional average HWT during 1982-2020 in the SCS based on observed datasets. The correlation coefficient between HWT and PC1 is 0.485. The light blue bars are strong WPSH years (1995, 1998, 2010, and 2020). Average (c) HWT and (d) HWI in strong WPSH years during 1982-2020 in the SCS for observations. will not change significantly. This indicates that the WPSH will still affect the HWT primarily in the future ( Figure S6 in Supporting Information S1).
Over the next 80 years, the WPSH is expected to strengthen, leading to a rise in the regional average HWT of summer MHWs in the SCS. This trend is likely to peak toward the end of the 21st century ( Figure 3c). Specifically, the average trend of the SCS HWT increases by 2 days every decade (p < 0.01), increasing from less than 10 days per summer between 2020 and 2030 to approximately 30 days per summer at the end of the 21st century. However, the SCS HWI is expected to remain nearly unchanged in the future (Figure 3d). The HWI ranged from about 1.2 to 1.5°C with no significant trend and very little interannual variation. This indicates that there is also no a significant correlation between HWI and the enhanced WPSH in the future.
To explore the reasons for the future increase in HWT, we also calculate the interannual variation in the average duration of summer MHWs for the next 80 years based on the best performing models. The results show that almost all models can simulate the increasing trend in the average summer MHWs duration in the future as the WPSH is enhanced, such as AWI-CM-1-1-MR, EC-Earth3, EC-Earth3-CC, and EC-Earth3-Veg (Figure 4 left column). The increasing trend in the average summer MHWs duration is consistent with the HWT. The long duration MHW events increase after the 2060s. Furthermore, we select the case with the highest HWT in each model and find that all of them exhibited long-lasting MHW events (Figure 4 right column). We infer that the increase in summer HWT in the SCS is mainly caused by the prolonged MHW duration (Costa & Rodrigues, 2021;Oliver et al., 2019;.

Potential Mechanism of Future Summer MHWs in the SCS
Previous studies have shown basin-wide mid-summer cooling in the SCS (Xie et al., 2003(Xie et al., , 2007. By inducing offshore upwelling on its northern flank, the southwesterly monsoon jet maintains a cold filament. As the southwesterly winds of the SCS weaken due to the anomalous WPSH, the upwelling and cold filaments become weaker or even disappear. Without the mid-summer cooling effect, the SST in the central SCS will rapidly warm and result in severe MHWs (Yao & Wang, 2021). To investigate the influence of the enhanced WPSH on the southwesterly monsoon, we derive the 10 m wind stress in strong WPSH years by using the observed data during 1982-2020 and seven CMIP6 models during 2021-2100 with multi-model ensemble mean method, respectively (Figures 5a and 5b). With the enhanced WPSH in the future, the southwesterly monsoon wind stress in strong WPSH years from 2021 to 2100 will decrease compared with the observations from 1982 to 2020 ( Figure S8a in Supporting Information S1). The observed and models' wind stress can differ in their northward components, which may exceed a maximum of 0.03 N/m 2 . We also investigate the impact of weakened SCS wind stress curl on the cold filament and future MHWs. The wind stress curl in the SCS have a robust correlation with the cold filament both in the historical period (r = −0.727, p < 0.01) and the future projection (r = −0.809, p < 0.01) (Figures 5c and 5d). The high correlations indicate that weakened wind stress curl will suppress the cold filament both in historical period and future projection. Without the cooling effect, the SST in the middle SCS will warm rapidly ( Figure S8b in Supporting Information S1). The correlations between the SCS filament and regional averages of HWT in the SCS are 0.823 (p < 0.01) in the historical period and 0.877 (p < 0.01) in the future projection (Figures 5c and 5d). Therefore, the summer MHWs in the SCS are mainly affected by the ocean upwelling and cold filament response to the increased WPSH variability, which is the result of the combined effect of atmospheric forcing and corresponding ocean dynamical processes.

Conclusions and Discussion
Using multi-model ensembles of CMIP6 models, the future variability in the WPSH and the response of MHWs in the SCS under the SSP5-8.5 scenario are revealed. The results show that the WPSH variability will increase in the next 80 years in most CMIP6 models. In response to the increased WPSH variability, the future summer SCS HWT will increase significantly and peak at the end of the 21st century. However, the HWI is negligibly affected by the enhanced WPSH. The increased HWT is mainly due to the extended duration of MHW events. Under the impact of the enhanced WPSH, the upwelling and cold filament off southern Vietnam will be weak or even disappear. Without the mid-summer cooling effect, the SST in the SCS will warm rapidly and cause severe MHWs. The combined effect of atmospheric forcing and ocean dynamical processes is the main drivers of future summer MHWs in the SCS.
The HWT and HWI in the SCS show different response to the strengthening of the WPSH. Specifically, the HWT responds strongly to the enhanced WPSH and increases with the WPSH strength. However, unlike the significant response of HWT, the HWI exhibits little response to the strengthening or variability of the WPSH, in both the historical and future periods. In previous studies, the HWI in other tropical ocean regions also had little variation and typically did not exceed 2°C (Holbrook et al., 2019;Oliver et al., 2018;Qi et al., 2022;. This is due to the SST-clouds feedback in tropical regions. When MHWs occur, the high SSTs in the SCS will enhance surface evaporation, increase latent heat loss, and strengthen convection, resulting in the formation of clouds, which in turn reduces solar radiation ( Figure S9 in Supporting Information S1) (Park et al., 2022;Stephens, 2005;Ying & Huang, 2016;Zhang & Li, 2014). This negative feedback mechanism of SST-clouds will limit the HWI within a relatively normal range and contribute to the weak response of HWI in the SCS to the strengthening of WPSH.
There has been an ongoing debate about the choice of baseline period when defining MHWs. Previous research emphasized that a fixed-baseline period is particularly useful when studying marine ecosystems, such as coral reefs, with a slow (or no) capacity to adapt to ocean warming (Amaya et al., 2023;Chiswell, 2022;Oliver et al., 2021;. In this study, we used a fixed 30-year baseline period and detrended all data to better comprehend the impact of MHWs on coral reefs in the SCS. However, using a fixed baseline period can magnify the impact of global warming on MHWs, resulting in year-round (the whole summer) MHWs (Costa & Rodrigues, 2021;Frölicher et al., 2018;Oliver et al., 2019Oliver et al., , 2021Plecha & Soares, 2020). This study's focus is on the response of MHWs in the SCS to the enhanced WPSH, rather than discussing extensively the impact of global warming on MHWs. In addition, the enhancement of the WPSH also leads to frequent severe MHW events after the 2060s, in which the HWT reaches a magnitude similar to that in 1998 and 2020 that occur massive coral bleaching in the SCS (Figures 5c and 5d) (Y. Chen et al., 2022;Li et al., 2008;Lyu et al., 2022). The HWT is strongly and significantly correlated with increased coral bleaching, decreased seagrass density and is a key metric for assessing ecological impacts (Hayashida et al., 2020;Smale et al., 2019). The SCS is a vital part of the world's coral reef triangle, and to prevent the tipping point for coral reefs, we must keep global warming within 1.5°C as quickly as possible (Armstrong McKay et al., 2022).  (1995,1998,2010,2020), (b) during 2021-2100 for multi-model ensemble mean. HWT in the SCS (red curve), the SCS cold filament indices (green curve), and SCS wind stress curl indices × 10 7 (blue bars) (c) during 1982-2020 based on observed datasets and (d) during 2021-2100 for multi-model ensemble mean. *** represent the correlation significant at the 99% confidence level. The index of wind stress curl is the regional average of wind stress curl in the bright green box in Figure S8a of Supporting Information S1 (9.5°-15°N, 109°−118°E) (Yao & Wang, 2021). The index of cold filament is the regional average of SST anomalies in the green box in Figure S8b of Supporting Information S1 (10°-13°N, 109°−117°E) (Xie et al., 2003).