1 INTRODUCTION

The Bohai and Yellow Seas (BYS), connected with the Bohai Strait, comprise semi-closed continental shelf seas bounded by the Chinese mainland and the Korean peninsula. The BYS region is dominated by the East Asian monsoon system. In addition, northerly winds engulf the BYS in winter, and its circulation pattern can be characterized by downwind coastal currents and upwind warm currents in the middle, known as the Yellow Sea Warm Current (YSWC). The YSWC, as the only channel to the open ocean, can reach far north to the Bohai Strait in its peak. It is especially important for water exchange between the BYS.

Despite the presence of rigorous studies conducted in the area, water exchange between the BYS continues to be a hot research topic, especially considering the frequent outbreak of biochemical hazards as well as rapid deterioration of coastal waters, which may be attributed to the significantly compromised capacity of water exchange given its semi-closed geographic nature (Fang et al., 2002; Wu et al., 2004a, b), and insufficient ecological environmental protection (Yu et al., 2000; Wei et al., 2003a). A general exchange pattern of water, i.e. inflow from the northern side and outflow from the south of Bohai Strait, was first proposed by hydrographic investigations (Guan and Chen, 1964; Guan, 1994) and supported by some pieces of sediment evidence (Cheng et al., 2004). Moreover, Laotieshan channel could be responsible for almost 90% of its volume, although the total volume, as well as its exchange rate could be moderate (Yuan et al., 1997). Furthermore, some studies reveal the presence of a significant seasonal cycle in the flow structure of Bohai Strait. Despite a consensus on its wintertime pattern, i.e. inflow from the north and outflow from the south (Wei et al., 2001; Lin et al., 2002; Zhang et al., 2010), its summertime pattern is still debated. During summertime, the pattern could be similar (Wei et al., 2003b) or reverse (Wei et al., 2001) as compared to its wintertime structure, while it can also show outflow from the middle with inflow from both sides (Lin et al., 2002). There is even an opinion that such a regular pattern does not exist (Zhang et al., 2010).

Winds have a unique role in driving water circulation between the BYS, especially the frequent outbreaks of winter storm events (typically on a synoptic scales of 1-2 days), which could force massive water exchanges into the open ocean (Xu et al., 2008), while significantly contributing to the hydrographic aspects of the BYS. Wan (2014) and Wan et al. (2015a) suggest intensive compensatory flow as well as massive exchanges of water in response to winter storms, accompanied with subtidal oscillations of sea level and currents by joint effects of upper Ekman dynamics and Kelvin waves transmitted along both sides of the BYS. Qu et al. (2014) also emphasize the roles of Kelvin waves in BYS water exchange, while Wan et al. (2015b) suggest a possible relationship between the frequency and intensity of winter storm events and patterns of wintertime circulation as well as its long-term variation by modulating water exchanges. Another proof comes from Song et al. (2017), where a positive correlation of winter storm events with water exchanges of the BYS is proposed based on a spectramixing model.

The capacity of water exchanges could be changed, together with an increase in both frequency and duration of winter storm events in the BYS region (Wang, 2013; Zhang et al., 2013). Intensive studies are needed to better understand the associated dynamics of synoptic storm events in modulating water exchanges in the BYS.

2 DATA AND MODELING 2.1 Data collection

The data used in this study are acquired from a special project entitled China's Offshore Marine Integrated Investigation and Evaluation. In the study, two moored Acoustic Doppler Currents Profilers (ADCP) were deployed in the North Yellow Sea (NYS) (Site: NYS, 123°30′E, 38°N) and Bohai Strait (Site: BS, 120°40′E, 38°N), respectively (Fig. 1), while a Sontek 250k Acoustic Doppler Profiler (ADP) was configured in a bin size of 2 m and total bin numbers of 30, providing a valid range of 3-61 m for the NYS site of 70 m. In addition, a total number of 27 910 measurements were made between December 31, 2006 and February 8, 2007, with an ensemble average of 5 min, accompanied with synchronous sea level data by Alec pressure gauge at an interval of 2 min. As for the Bohai Strait site, a TRDI 300k Workhorse ADCP was adopted with a bin size of 1 m and total bin number of 29, providing a valid range of 3-31 m for the Bohai Strait site of 33 m. The measurements were made from December 25, 2006 to January 23, 2007, with an ensemble average of 5 min. Necessary quality controls were performed before the analysis and tidal signals were removed using lowpass filter.

Data on winds were extracted from the Reanalysis Dataset for Sea Surface Wind Fields in China's Offshore and Adjacent Ocean Areas from the server of Ocean & Atmosphere Date Center, Ocean University of China (http://coadc.ouc.edu.cn/index.php/index/dataset/id/23), which serves as a 50-year hind cast result of nested MM5 model. It features NCEP-FNL, NASA-SST, NEAR-GOOS-SST, GTS Data, and 132 radiosonde data of 40 years from around the Chinese seas. It also takes into account typhoon corrections from UNIsys Western Pacific Dataset and Japan Meteorology Agency (JMA) based on ECMWF reanalysis. At the same time, it also takes into account improvements made in the NCARAFWA Tropical Cyclone Bogussing Scheme in MM5 model by introducing a more realistic and physically consistent initial cyclonic structure. The winds dataset has a temporal interval of 3 h, from 1960 to 2007, and a spatial resolution of 0.1°.

In addition, ETOPO5 bathymetry data (https://www.ngdc.noaa.gov/mgg/global/Etopo5.html), SODA T/S data (https://climatedataguide.ucar.edu/climate-data/soda-simple-ocean-data-assimilation), NCEP surface heat flux data (https://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html), and TPXO tidal datasets (http://volkov.oce.orst.edu/tides/tpxo8_atlas.html) were also collected for our modeling purpose.

2.2 Modeling

A high-resolution numerical investigation was implemented for the East China Sea (ECS) region based on Regional Ocean Modeling Systems (ROMS). The model domain covers the ECS and Kuroshio adjacent region, within 24°N-41°N, 117°E-130°E (Fig. 1) with a horizontal resolution of 5′×5′. Terrain following S coordinates of 16 vertical layers was adopted, with an additional refinement for the upper layer. The model was integrated from January 1, 2000 to December 31, 2007, with a time step of 30 s for external mode and 600 s for internal mode.

Moreover, the bottom topography was obtained by combining water depth data from the ETOP5 (1/12°) dataset and water depths digitized from the highresolution navigation charts published by the China Maritime Safety Administration. Meanwhile, sea level and currents files from equilibrium state of longterm simulation, together with T/S interpolated from SODA (0.5°) dataset servers as the initial field, NCEP (~1.875°) solar radiation, and downward longwave radiation, provided the upper forcing while the open boundary conditions come from the T/S and currents of SODA dataset. Tidal forcing was included along the open boundaries with four astronomical constituents (M2, S2, O1, and K1) of TPXO8 (1/30°) dataset. Buoyancy fluxes from River discharges were also considered by adding the Yellow River and Changjiang (Yangtze) River based on in-situ observations at Lijin and Datong sites. The resolution of wind forcing, both spatial and temporal, was critical for the purpose of this study. Hence, data on three-hourly winds (0.1°) from the Reanalysis Dataset for Sea Surface Wind Fields in China's Offshore and Adjacent Ocean Areas were adopted.

3 RESULT AND ANALYSIS

According to the analysis of wind field's characteristics, winter is the season with the most frequent high winds in the BYS. Therefore, the study focuses on the impact of winter winds on the sea level and flow field in this region. Moreover, the data shows that extremely high-wind conditions existed between January 5 and 7, 2007, with a maximum wind speed of 17 m/s. In addition, slightly weaker high-wind conditions persisted between January 26 and 27. Hereinafter, the two extreme events shall be taken as examples to analyze the role of the high wind process.

3.1 Circulation and sea level fields 3.1.1 Analysis of observations

To extract the residual flow, the currents were averaged over a 12.42-h period, as M2 tide is the most dominant tidal component in the BYS. Figure 2a demonstrates the time-varying velocity in January 2007 after low-pass filtering of the ADCP observed data from the anchored station in the NYS. From the figure, we can gauge the following characteristics:

1) From the perspective of spatial structure, the residual current field in this station is mainly shown as barotropic mode without obvious vertical change. It reflects that the residual current in winter is the combined result of pressure gradient, the Coriolis force, and surface/bottom stress.

2) From the perspective of flow direction, the residual current of the station is mainly in the meridional direction, while the zonal component is very small. The configuration of flow direction corresponds to its geographical position, which is located at the center of the NYS trough. Thus, the flow is along the isobaths, subject to terrain constraint.

3) In terms of time variation, the residual current shows obvious paroxysmal features, and its vectors exhibit strong time-varying characteristics, especially the speed, which varies from 1-2 cm/s to 14 cm/s. There are also many reverse moments for the current, which indicates that this flow does not represent stable circulation.

4) The wind in the same period of the station is mainly northwestwards, displaying high speeds, with the average and maximum speeds of 7 and 17 m/s, respectively. However, the mean flow is mainly in the north or north by west direction, which means that the main direction of the residual current is opposite to that of the wind stress. This indicates that the residual current structure of the station is characterized by upwind flow, and this windward flow can sometimes even occupy the area close to the surface. Correlation analysis between wind field and residual current shows a significant negative correlation between the two variables, with the correlation coefficient reaching -53.5%.

5) It can be seen that strong northwards flow in the residual current sequence is accompanied by a high wind process. It indicates that the northerly wind plays an important role in inducing the strong upwind flow. However, this study has found that not all high wind processes can lead to strong windward flow. For example, from January 21 to 23, there was a strong northwesterly wind, but the corresponding upwind flow was weak, indicating that the wind field, at least the local wind field, is not the only determinant of the northward upwind flow. In addition, other factors may also influence windward flow.

6) Another remarkable feature of the residual current structure is that there was a very strong upwind flow from January 7 to 8, with depth-averaged velocity of close to 12 cm/s. The windward flow distributed throughout the water column, even at the surface that is not affected by wind stress. Compared with other moments, both the intensity of the upwind flow and the depth reached were far greater than the average level. However, the difference between the wind speed at the corresponding moment and the other high wind processes was not large. In sharp contrast with it on January 5 to 7, the wind field was dominated by westerly winds almost without north component, and the corresponding mean flow appeared as the south by east component of all depths.

7) Obvious diversities have been found for wind stress in the upper Ekman layer, especially above the depth of 17 m. Sometimes there were strong winddriven currents, such as during January 2 to 4. However, more often, the role of wind stress in the upper Ekman layer is very small, and even disappears at certain times. For example, during the high wind from January 21 to 23, there was almost no flow in the upper Ekman layer. Nevertheless, it is still evident that the effect of wind stress can work in the surface Ekman layer above the depth of 17 m.

Figure 3c shows the results after the same low-pass filtering of the sea level data. It can be seen that there is obvious abnormal increase and decrease of water in the time series of the residual water level of the station, with the maximum drop of sea level exceeding 1 m. This abnormal rise or drop of the sea level of the station is inevitably accompanied by the water exchange in the NYS with the adjacent ocean. Therefore, it should be reflected in the residual current structure. Comparing this abnormal change in the sea level with the residual current structure, we can see the obvious correlation between them. The abnormal drop in the sea level near the station is often accompanied by a relatively strong northwards upwind flow. The intensity of the windward flow is proportional to the magnitude of the abnormal increase or decrease in water. For example, the sea level abnormally decreased by more than 1 m during January 6 to 7, and the velocity of corresponding northwards current exceeded 10 cm/s. This relationship between sea level change and residual current intensity seems to indicate that the upwind flow is closely related to the water exchange between the entire NYS and the adjacent sea.

The aforementioned analysis shows the presence of a significant upwind flow in the middle of the NYS trough, i.e. the YSWC, which shows obvious barotropic features with vertical uniform structure in winter. In addition, the upwind flow is not a stable mean flow, and its vector has a drastic change with time, which is sometimes even reversed. Its speed also varies from 1-2 to 14 cm/s. In addition, the occurrence of strong upwind flow has an obvious relationship with the wind speed. However, the observation results also show that the local wind field is not the only determinant and other important factors also affect the residual current structure in the central region of the NYS. The relationship between the abnormal change in sea level and the intensity of upwind flow seems to indicate that the residual structure of the station is closely related to the exchange process of the entire NYS and the adjacent sea.

3.1.2 Simulation of high wind process

In order to further explore the circulation and sea level changes of the BYS under high winds, this study simulated the aforementioned phenomenon using the ROMS model.

First, the model was driven by the monthly average wind field. The results of flow field in January show that (figure omitted), there is a stable northwards warm current in the Central Yellow Sea, which can extend northwards along the Yellow Sea Trough to the Bohai Strait. At the same time, there is a southward coastal current along the east coast of the Chinese mainland and the west coast of the Korean Peninsula. The entire flow field did not change significantly over time.

Later on, this study used real-time high wind process to drive the model. The Reanalysis Dataset for Sea Surface Wind Fields in China's Offshore and Adjacent Ocean Areas can provide a good upper boundary for our numerical model, which can be used to study the rapid change process of the YSWC. This study first compared it with the observed wind field and found that both the magnitude and direction were consistent, and several high wind processes were characterized accurately (Fig. 3a, b). We used the wind field to force the model every 3 h and found that several abnormal water increase and decrease processes of the anchored station (123.5°E, 38°N) in the NYS could be reproduced well (Fig. 3c, d): there was obvious abnormal water reduction from January 6 to 7, 2007, with the maximum sea level decline of nearly 1m on January 7, while there was also abnormal water reduction from January 26 to 27, just with a smaller amplitude. The simulated flow velocity changes (Fig. 4) also corresponded to the actual situation, and the flow reversed quickly before and after the strong wind. All of these are consistent with the observations, indicating that the model can simulate the abnormal increase and decrease of sea level and the corresponding velocity in the NYS.

The flow field has also changed significantly. In particular, the YSWC is no longer a steady flow, but a flow reciprocating with time in the Yellow Sea. Analysis of the simulated flow field from January 6 to 9, 2007 (Fig. 5) shows that on January 6, the entire area was dominated by westerly winds, with the corresponding southwards flow in the Yellow Sea Trough. Later on January 7, a northwesterly high wind appeared, which pushed the water in the Yellow Sea towards south rapidly, especially on the west side, resulting in the lowest sea level that spreads from the Bohai Sea to the South Yellow Sea. On January 8, the wind still blew northwest, but in order to supplement the reduced water level of the Yellow Sea, there was a strong northward warm wave in the Yellow Sea. On January 9, the wind direction remained unchanged. With the recovery of the sea level, the YSWC continued to flow northwards, but the intensity weakened. The meridional velocity profile through the 34.5°N section in the Yellow Sea (Fig. 6) also shows that the flow in almost the entire depth moved southwards immediately after the northerly high wind erupted, while the flow in the whole depth turned northwards immediately after the northerly wind weakened. The reciprocating processes of the YSWC that first flows southwards and then flows northwards have been fully performed.

3.2 Water exchange 3.2.1 Water exchange and sea level rise or drop

The YSWC plays an important role in the Bohai Sea water exchange. The exchange volume between the BYS is closely related to the strength of the YSWC. A comparison of the observed current in the Bohai Strait with the wind field shows that the flow between the BYS is consistent with the intensity of wind field. As shown in figure, there was a strong eastward current in the Bohai Strait during high winds of January 6 to 7, which indicates that the Bohai Sea could experience large water reduction (Fig. 2b).

To verify the relationship between the water exchange through the Bohai Strait and the abnormal sea level fluctuation, the flow of three sections in January were calculated, including 34.5°N in the South Yellow Sea, 37.5°N in the NYS, and 121°E in the Bohai Strait (Fig. 1), and their changes were put in the same coordinate (Fig. 7). A comparison with the simulated sea level changes at the observation stations in January showed that the flow changes of the three sections were consistent with the anomalous water increase and decrease in the NYS.

During the process of abnormal sea level fluctuation around January 6, the flow of the three sections changed dramatically. The process of water level change before and after January 26 also corresponds to the larger flow change. From the zonal velocity profile of the Bohai Strait (not shown), we can learn that the velocity before and after high winds rapidly changes from outflow to inflow. Thus, it can be seen that the variation of discharge is closely related to the anomalous increase and decrease of water in the NYS, which also indicates that the water exchange between the BYS is closely linked to the anomalous increase and decrease of water caused by high winds.

Briefly, correlations of the strength of wind field with the YSWC, abnormal sea level fluctuation, and the water exchange through the Bohai Strait are very consistent and clear. In addition, the response of water exchange to wind field is quasi-synchronous.

3.2.2 Estimation of water exchange rate

To facilitate the statistics, the 121°E line in Bohai Strait was used as the boundary between the BYS, and then the Bohai Sea was divided into Bohai Bay, Laizhou Bay, and Liaodong Bay by 118°55′E, 37°45′N, and 39°35′N lines, respectively (Fig. 1). Later on, the volume transports of the Bohai Strait and the bay mouth sections were calculated. The variation with time in January 2007 is shown in Fig. 8. We found that water flux responded rapidly to high winds in each sea area, with the value reaching more than 10 times than usual, and high wind process played an important role in water exchange between the BYS.

The Bohai Sea covers an area of 7.7×10⁴ km² with an average depth of 18 m, and the total volume of the Bohai Sea is about 1.39×10¹² m³. The exchange rate of the BYS can be estimated by analyzing the proportion of discharge in the Bohai Strait. Table 1 shows the exchange time and volume of each sea area during the two high-wind incidents in January 2007. In the early period of the high wind process in the beginning of January, the Bohai Sea outflow lasted for 54 h, with the volume of water reduction totaling about 1.13×10¹¹ m³, accounting for 8.1% of the total volume of the Bohai Sea. In the later period of the high wind process, the inflow lasted for 24 h, with the volume of water increase totaling about 1.08×10¹¹ m³, accounting for 7.8% of the total volume of the Bohai Sea. In the early stage of the high wind process at the end of January, the Bohai Sea outflow lasted for 27 h, with the volume of water reduction totaling about 3.45×10¹⁰ m³, accounting for 2.5% of the total volume of the Bohai Sea. In the later stage of the high wind process, the inflow lasted for 21 h, with the volume of water increase totaling about 2.50×10¹⁰ m³, accounting for 1.8% of the total volume of the Bohai Sea.

In the early period of the high wind process in the beginning of January, the outflow of Bohai Bay lasted for 27 h, with the volume of water reduction totaling about 1.19×10¹⁰ m³. In addition, the outflow of Laizhou Bay lasted for 27 h, with the volume of water reduction totaling about 0.79×10¹⁰ m³. Moreover, the outflow of Liaodong Bay lasted for 51 h, with the volume of water reduction totaling about 3.77×10¹⁰ m³. Furthermore, in the later stage of the high wind process, the inflow of Bohai Bay lasted for 21 h, with the volume of water increase totaling about 1.80×10¹⁰ m³. Meanwhile, the inflow of Laizhou Bay lasted for 24 h, with the volume of water increase totaling about 0.85×10¹⁰ m³. At the same time, the inflow of Liaodong Bay lasted for 24 h, with the volume of water increase totaling about 3.64×10¹⁰ m³. In addition, the reduction time of Bohai Bay and Laizhou Bay lagged behind the total reduction time of Bohai and Liaodong Bay by about 30 h.

In the early period of high wind process at the end of January, the outflow of Bohai Bay lasted for 30 h, with the volume of water reduction totaling about 0.48×10¹⁰ m³. The outflow of Laizhou Bay lasted for 30 h, with the volume of water reduction totaling about 0.29×10¹⁰ m³. In addition, the outflow of Liaodong Bay lasted for 18 h, with the volume of water reduction totaling about 1.11×10¹⁰ m³. Moreover, in the later stage of the high wind process, the inflow of Bohai Bay lasted for 18 h, with the volume of water increase being about 0.37×10¹⁰ m³. The inflow of Laizhou Bay lasted for 18 h, with the volume of water increase totaling about 0.19×10¹⁰ m³. Meanwhile, the inflow of Liaodong Bay lasted for 24 h, with the volume of water increase totaling about 0.95×10¹⁰ m³. The progress of water increase and decrease in the Bohai Sea and other sea areas has been reported to be the same.

It can be inferred that the impact of a high wind process is enormous. It can make the flux through the Bohai Strait, as well as that through the mouth of each constituent bay far greater than usual, resulting in a significant increase in the water exchange rate. Therefore, we believe that the exchange between the BYS is completed by only a few occasional high wind processes in winter. However, individual results are not enough to fully prove this conjecture, and further studies are required.

4 CONCLUSION

Based on the observed data from the special project titled China's Offshore Marine Integrated Investigation and Evaluation, this study uses diagnostic analysis and numerical simulation to reveal the influence of high wind processes on the circulation and water exchange between BYS in winter. This study has reached the following results:

(1) In winter, the vertical structure of the YSWC is relatively uniform under the action of high winds, with obvious barotropic features. However, this flow is not a stable mean flow, showing strong paroxysmal and reciprocating characteristics. A comparison of the changes in sea level suggests that the intensity of the northwards upwind flow is consistent with the abnormal fluctuations in the sea level. It indicates that the upwind flow is closely related to the water exchange between the BYS.

(2) The impact of the high wind process on the water exchange between the BYS is enormous. It can make the flux through the Bohai Strait, as well as that through the mouth of each constituent bay far greater than usual, resulting in a significant increase in the water exchange rate. The exchange capacity, which is about 8% of the total volume of the Bohai Sea, can be completed in a few days. Therefore, it is concluded that the water exchange of the Bohai Sea may be completed by only a few occasional high wind processes in winter.

5 DATA AVAILABILITY STATEMENT

The observation data that support the findings of this study are available from National Marine Data and Information Service (NMDIS) conditionally, and certain restrictions may apply to the purpose of the data usage under special license, such that are not open for public access. Therefore, data are available from the authors upon reasonable request and with permission of NMDIS.