In this study, we used typhoon information obtained from the Tropical Cyclone Data Center of the China Meteorological Administration. As shown in Figure 1B , Typhoon Damrey (No. 1210) originated in the northwestern Pacific Ocean, took a southeast-northwest course, reached the East China Sea on 1 August, and then entered the Yellow Sea on 2 August in 2012. After making landfall in Jiangsu Province at 18:00 on 2 August, it turned northward, entered the Bohai Sea and dissipated at 00:00 on 4 August. Figure 2 displays the 12-hourly sea surface winds over the Yellow Sea, Bohai Sea, and adjacent areas during the passage of Damrey.

Daily SST obtained from the Remote Sensing Systems (REMSS; Gentemann et al., 2010) was adopted to validate the model simulation and examine temperature variations during the typhoons’ passages. The REMSS SST data set combines the through-cloud capability of microwave data with the high spatial resolution and near-coastal capability of infrared data. Merged satellite observations were from both microwave sensors, including the Global Precipitation Measurement Microwave Imager, the Tropical Rainfall Measuring Mission Microwave Imager, the National Aeronautics and Space Administration (NASA) Advanced Microwave Scanning Radiometer for Earth Observing System, the Advanced Microwave Scanning Radiometer 2, the WindSat Radiometer, and infrared sensors, such as the Moderate Resolution Imaging Spectroradiometer on the NASA Aqua and Terra platform and the Visible Infrared Imaging Radiometer Suite on board the Suomi National Polar-orbiting Partnership satellite. The REMSS SST has a spatial resolution of 0.09°, approximately 7.7–8.7 km in the Yellow Sea, and spans from June 2002 to the present.

Daily sea surface height (SSH) from the global eddy-resolving physical ocean and sea ice reanalysis at 1/12° resolution (GLORYS12; Jean-Michel et al., 2021) was adapted to validate the model simulation. The GLORYS12 product covers the altimetry period from 1993 to the present. It was produced by the Copernicus Marine Environment Monitoring Service (CMEMS), utilizing the Nucleus for European Modelling of the Ocean model, which was driven at the surface by atmospheric reanalysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF). A reduced-order Kalman filter was used to assimilate available observations, including satellite-observed SST, SSH, and sea ice concentration, as well as in situ profiles of temperature and salinity.

In addition, Yang et al. (2017) conducted ship observations of temperature and salinity profiles in the southern coastal waters of the Shandong Peninsula on 1 August and 5 August before and after the passage of Damrey over the study area, respectively. Ship observation sites were marked as FS, B, C, D, E and F in Figure 1A . Note that ship observations were conducted at sites FS−E on the two days, but were also conducted at site F on 5 August. The observation sites were very close to each other and the minimum distance between them was much smaller than the model resolution as described in subsection 2.4. There were roughly four model grids that were close to the ship observation sites of Yang et al. (2017). Therefore, as shown in Figure 1A , the ship observations at observation sites FS, C, E and F were compared with the CTR simulation results at model grids FS′, C′, E′ and F′, respectively.

We used variables derived from the fifth-generation atmospheric reanalysis of the global climate generated by ECMWF (ERA5; Hersbach et al., 2020; Wu et al., 2020) as atmospheric forcing at the sea surface for the model simulations. The ERA5 dataset includes hourly estimates of the atmospheric, land, and oceanic climate variables from January 1940 to the present. It has a spatial resolution of 0.25°, approximately 31 km in the study area. Previous observations proved that the ERA5 dataset effectively captured the large-scale spatiotemporal variations in sea surface atmospheric forcing in the study area (Wu et al., 2020).

Additionally, the Hybrid Coordinate Ocean Model (HYCOM; Cummings and Smedstad, 2013) reanalysis data were used to provide initial and open boundary conditions for the model simulations. The HYCOM reanalysis was produced with HYCOM as the dynamical model and Navy Coupled Ocean Data Assimilation for data assimilation (Cummings, 2005). The HYCOM variables are available from January 1994 to present and cover the global ocean with a horizontal resolution of 1/12°.

The Regional Ocean Modeling System (ROMS; Shchepetkin and McWilliams, 2005) was used to understand the spatiotemporal variations in temperature and circulation during the passages of typhoons and the underlying dynamics. As shown in Figure 1A , the model domain included the entire East China Sea, Yellow Sea, and Bohai Sea, extending zonally from 117.4390°E to 133.8120°E and meridionally from 24.0919°N to 41.0031°N. Horizontally, ROMS employs an Arakawa C grid, allowing for local grid refinement or nesting. Vertically, ROMS utilizes a terrain-following coordinate system that stretches with changes in topography, enabling refinement at specific layers. In this study, the model simulation had a horizontal resolution of approximately 7.3 km and 20 vertical layers, with finer resolution near the sea surface. Vertical grid intervals basically increased from coastal regions to offshore regions. In the coastal regions, the smallest vertical grid interval was approximately 0.3 m. In the offshore regions where the YSCWM existed, the largest vertical grid interval near the sea surface was approximately 1.6 m and that near the sea bottom was approximately 10 m. The model topography was derived from 1 Arc-Minute Global Relief Model (ETOPO1; Amante and Eakins, 2009). Following Liu et al. (2022a), the generic length scale scheme (Umlauf and Burchard, 2003; Warner et al., 2005) was employed in the model to calculate vertical diffusion coefficients.

Initial values of temperature, salinity and current vector in the model simulation were obtained from the climatological mean variables of the HYCOM data during 1994−2023. There were three open boundaries (non-land boundaries) around the model domain ( Figure 1A ). Values of temperature, salinity, and subtidal current vector at the three open boundaries were derived from daily variables of the HYCOM data. Schemes proposed by Flather (1987) and Chapman (1985) were used for barotropic momentum and free surface at the three open boundaries, respectively. A mixed radiation-nudging scheme (Marchesiello et al., 2001) was adopted for the open boundary conditions of three-dimensional momentum, temperature, and salinity. Eight major tidal constituents (M₂, S₂, K₁, O₁, N₂, K₂, P₁, Q₁) extracted from the global ocean tidal model TPXO8 (Egbert and Erofeeva, 2002) were used as tidal forcing. TPXO8 is a global model of ocean barotropic tides, which best-fits, in a least-squares sense, the Laplace Tidal Equations and assimilates observations. The atmospheric forcing at the sea surface was derived from the ERA5 data Atmospheric forcing variables used in ROMS mainly included 10-m wind vectors, shortwave radiation, longwave radiation, 2-m air temperature, precipitation rate and humidity. The simulation period was from 1 January to 1 September in 2012. The above numerical simulation was defined as control (CTR) simulation.

For two time series or spatial maps of X and Y, the correlation coefficient (R) is calculated as R = 〈 X ′ Y ′ 〉 / 〈 X ′ 2 〉 〈 Y ′ 2 〉 , the mean absolute bias (MAB) is calculated as MAB = 〈 | X − Y | 〉 , the root mean square error (RMSE) is calculated as RMSE = 〈 ( X − Y ) 2 〉 , where X ' and Y ' are temporal or spatial anomalies, and 〈   〉 denotes the time mean or space average. As different products had different spatial resolutions, the fine resolution variable was mapped onto the horizontal grid of the coarse resolution variable before the spatial correlation coefficient was calculated. In this study, the linear correlation coefficients are all above the 95% confidence level unless otherwise specified.

To examine the changes in density stratification, we calculated squared buoyancy frequency (N ²) as N 2 = − g ∂ ρ / ρ ∂ z , where ρ was water potential density, g =9.8 m/s² was the gravity acceleration.

The temperature budget equation is expressed as

with the surface and bottom boundary conditions being

In the above equations, T is water temperature, A k t is vertical diffusivity coefficient, and u , v and w are zonal, meridional and vertical components of the three-dimensional current vector. x is the zonal coordinate being positive eastward, y is the meridional coordinate being positive northward, and z   is the vertical coordinate being positive upward. Q n e t is the net heat flux at sea surface, C p =3985 J (kg °C)⁻¹ is the specific heat capacity of seawater at constant pressure, ρ 0 = 1025 kg m⁻³ is the reference density of seawater, and h is the water depth. The abbreviation R A T T denoted the total change rate of temperature, A D V T denoted the change rate of temperature caused by three-dimensional current transports, and D I F F T denoted the change rate of temperature caused by vertical turbulent diffusion. D T was the change rate of temperature caused by horizontal diffusion and was neglected in the analysis due to its small value resulting from the weak horizontal temperature gradient. All of the above temperature budget terms were directly output from the numerical model.

Figures 3A–J show daily SSTs in the satellite observation and CTR simulation during Damrey’s passage over the study area. A visual comparison indicated that satellite-observed SSTs agreed with CTR-simulated SSTs, showing significant cooling on the right side of the typhoon track. The spatial correlation coefficients between satellite-observed and CTR-simulated SSTs were 0.82, 0.73, 0.69, 0.76, and 0.76 in the five days, respectively. In addition, MAB and RMSE were also relatively low, ranging from 0.77°C to 0.93°C and from 1.07°C to 1.32°C, respectively.

Meanwhile, daily SSHs in the satellite observation also agreed with those in the CTR simulation ( Figures 3K–T ). The spatial correlation coefficients between the satellite-observed and CTR-simulated SSHs were 0.96, 0.96, 0.97, 0.95, and 0.91 in the five days, respectively. MAB and RMSE between the satellite observation and CTR simulation were also relatively low, ranging from 0.14 m to 0.16 m and from 0.17 m to 0.18 m, respectively. SSH was relatively low in the entire Yellow Sea before the passage of Damrey. On 1 August, however, SSH began to rise in the South Yellow Sea, while remained almost unchanged in the North Yellow Sea. On 2−3 August, high SSH appeared in the Bohai Sea and South Yellow Sea, especially in the northwestward coastal areas of the Bohai Sea and southern coastal areas of the Shandong Peninsula. On 4 August, SSH decreased to low values after the typhoon dissipated.

The ship observations demonstrated the commonly-existed vertical structure of temperature profiles in the Yellow Sea, which was composed of an upper mixed layer, thermocline and a deep mixed layer, consistent with previous studies ( Figures 4A–C ; Zhai et al., 2021; Li et al., 2024). Damrey induced surface water cooling of approximately 2°C, subsurface water warming up to 10°C, bottom water warming of approximately 4°C, and deepening of both the mixed layer and thermocline by roughly 10 meters ( Figures 4D ). Figures 4E–H show the CTR-simulated temperature profiles in days of ship observations of Yang et al. (2017). Overall, the CTR simulation well reproduced the ship-observed vertical structures of temperature profiles. The deep layer temperature was warmer in the CTR simulation than in ship observations, and hence the thermal stratification was slightly weaker in the CTR simulation than in ship observations, possibly because that tidal mixing in the model simulation was weaker than that in reality. Though with smaller magnitudes, the CTR simulation well reproduced dramatic changes of temperature profiles during Damrey’s passage, which consisted of surface layer cooling, stronger subsurface layer warming and weaker deep layer warming.

We also compared the ship-observed and CTR-simulated salinity profiles before and after Damrey’s passage in Figures 4I–P . The salinity profiles were composed of an upper mixed layer, halocline and a deep mixed layer, consistent with temperature profiles. Damrey induced rapidly decreasing of salinity throughout the whole water column, especially at nearshore sites, accompanying the deepening of upper mixed layer ( Figures 4I–L ). The largest salinity decrease was approximately −0.45 and occurred at about 9.0 m in the lower halocline. The CTR simulation well reproduced the ship-observed vertical structures of salinity profiles. The CTR-simulated salinity was larger than ship-observed salinity throughout the whole water column, possibly because that we had no observations of runoffs of all rivers from the Shandong Peninsula into the Yellow Sea and excluded them in our model simulation. Though with smaller magnitudes, the CTR simulation well reproduced the dramatic salinity decrease in the subsurface and deep layers during Damrey’s passage. The largest salinity decrease in the CTR simulation was approximately −0.15 and occurred at about 9.4 m in the lower halocline. With no river runoffs from the Shandong Peninsula, however, the CTR-simulated salinity increased during Damrey’s passage due to the strong wind-enhanced turbulent mixing.

Overall, the above comparisons indicated that the model simulation well reproduced the dramatic changes in hydrodynamic conditions in the study area during the passage of Damrey.

Note that Damrey passed over the Yellow Sea mainly on 2 August ( Figure 2 ). To gain more insights into the underlying dynamics, Figure 8 gives daily mean temperature budget terms in the surface, subsurface, and bottom layers on 2 August. The spatial distributions of temperature change rates ( R A T T ) were consistent with those of temperature changes. In the surface layer, R A T T showed spatially coherent negative values in most regions of the South Yellow Sea, and the largest temperature cooling rates occurred in a southeast-northwest band on the right side of the typhoon track. The negative   R A T T was primarily caused by D I F F T in most regions of the South Yellow Sea and by A D V T in sporadic coastal regions of the Korean Peninsula.

In the subsurface layer at 15 m, positive R A T T values dominated most of the South Yellow Sea, forming a southeast-northwest band along the right side of Damrey’s track. The positive R A T T was primarily attributed to D I F F T . This finding indicated that surface temperature cooling and subsurface temperature warming in the majority of the Yellow Sea were caused by the strengthened vertical mixing generated by the typhoon’s strong sea surface winds (Liu et al., 2022a, b). On the contrary, the positive R A T T in the southern coastal areas of the Shandong Peninsula was mainly caused by A D V T . The negative R A T T above the southern bottom slope of the South Yellow Sea just off the Subei Shoal was also caused by A D V T .

In the bottom layer, however, the R A T T was caused by different processes in different regions. The significant band-shaped negative R A T T along the bottom slope off the Subei Shoal and in the western coastal areas of the Korean Peninsula in the South Yellow Sea was dominated by A D V T . The positive R A T T in the southern coastal areas of the Shandong Peninsula, however, was caused by the combined effects of D I F F T and A D V T .

From the above analyses, we could conclude that surface temperature cooling and subsurface temperature warming were primarily caused by strengthened vertical mixing in the majority of the Yellow Sea. The bottom water warming off the southern coast of the Shandong Peninsula was a joint effect of current transport and vertical mixing. However, the cooling of the entire water column just off the Subei Shoal was mainly contributed by the current transports across the temperature front and thermocline around the YSCWM.

Temperature budget analyses suggested that three-dimensional current transport made significant contributions to temperature changes during Damrey’s passage, especially in the coastal areas and above the bottom slope. To fully understand the underlying processes, it was necessary to examine the three-dimensional circulation changes during Damrey’s passage over the Yellow Sea. For that purpose, Figures 9A–F display the spatiotemporal variations in daily ocean currents in both the sea surface and bottom layers from 31 July to 2 August. In each day, we subtracted tidal current vectors from hourly ocean current vectors and then calculated the average of residual current vectors to obtain daily ocean current vectors. By comparison, ocean currents on 31 July before Damrey’s passage were much weaker and thus not discussed further in this analysis.

Damrey’s powerful winds began to trigger strong sea surface currents in the southeastern Yellow Sea after 12:00 on 1 August. Sea surface currents in the southeastern Yellow Sea flowed northwestward, with a maximum speed of approximately 0.6 m/s appearing to the southwest of the Cheju Island. On 2 August, as Damrey’s center moved primarily through the Yellow Sea, sea surface currents greatly strengthened across the entire basin. Sea surface currents were characterized by a strong northwestward current band just to the right side of the typhoon track and shifted to flow northward in other areas of the Yellow Sea. The current speeds in the strong northwestward current band usually exceeded 0.4 m/s, reaching up to 0.6 m/s. In contrast, sea surface current speed on the left side of the typhoon track was significantly weaker. In comparison with Figure 2 , one could know that the strong sea surface currents were wind-forced Ekman currents in nature, which were deflected to the right with respect to sea surface winds, consistent with the situations during the passages of Lekima and Bavi (Liu et al., 2022a, b).

In the bottom layer, ocean currents were much smaller than those in the surface layer. The currents remained relatively unchanged on 1 August. However, on 2 August, distinct seaward currents, flowing in the opposite direction to the surface currents, appeared in the southern coastal areas of the Shandong Peninsula. Another hotspot region with strong ocean currents was along the bottom slope just off the Subei Shoal, where water depth increased northwestward rapidly from approximately 20 m to 40−60 m. The strong ocean currents exceeded 0.15 m/s and shifted to northwestward, crossing the isobaths and flowing consistently toward the typhoon track.

Clearly, ocean current changes in the surface and bottom layers in response to Damrey’s passage were opposite to each other in the southern coastal areas of the Shandong Peninsula and along the bottom slope off the Subei Shoal. To find out whether ocean currents in these areas exhibited a two-layer vertical structure, we examined ocean currents along Sec1 and Sec2. As seen in Figure 9G , the most striking feature along Sec1 was that ocean currents were characterized by a two-layer structure with opposite directions in depth layers above and below the upper limit of the subsurface temperature warming layer in the western part along the section. Especially along the southern coastal areas of the Shandong Peninsula, the two-layer currents were connected by coastal downwelling, crossing the tidal front and bringing warm waters downward. Along Sec2, two-layer ocean currents with opposite directions also could be seen along the bottom slope just off the Subei Shoal, and were connected to each other with significant upwelling ( Figure 9H ). The upwelling crossed the tidal front and brought the cold water upward. The above analyses suggested that just off the Subei Shoal, the strong wind-driven Ekman flow in the surface layer had significantly larger speed on the right side of Damrey’s track than on the left side. This phenomenon led to Ekman divergence below sea surface, especially on the right side of the typhoon track, causing Ekman pumping and resulting in upwelling just below and on the slightly right side of the typhoon center. The bottom layer water then flowed towards the center of this upwelling zone driven by horizontal pressure gradient force (Liu et al., 2022b).

To conclude, sea surface currents were primarily wind-forced Ekman currents in the South Yellow Sea during Damrey’s passage. They transported a large amount of water to accumulate along the south coast of the Shandong Peninsula, generating coastal downwelling. The coastal downwelling then transported warm surface water to deeper layers, leading to vertically consistent temperature warming below the surface. In contrast, wind-forced Ekman currents dispersed above the bottom slope just off the Subei Shoal, inducing upwelling. The upwelling transported cold water from the deep layer to sea surface layer, leading to vertically consistent temperature cooling.

As indicated in previous studies (Liu et al., 2022a, b; Yan et al., 2023) and above sections, bottom topography played a quite important role in the three-dimensional responses of temperature and circulation in the shelf seas to typhoons. We conducted a numerical experiment (EXP_TOPO) to give more insights into the exact impacts of bottom topography in the current study. Model configurations of the EXP_TOPO simulation were nearly the same to those of the CTR simulation except that bottom topography in the EXP_TOPO simulation was modified to decrease linearly eastward to exclude the bottom slope just off the Subei Shoal ( Figure 10A ).

Figures 10B–D show the temperature changes in the EXP_TOPO simulation. Both SST cooling and bottom warming off the Subei Shoal moved closer to the coast in the EXP_TOPO simulation than in the CTR simulation. The vertically consistent temperature cooling moved shoreward from the right side of the typhoon to the left side along Sec2. On the right side of the typhoon track, there was still temperature cooling below the subsurface layer. To understand the underlying dynamics, we investigated ocean currents along Sec2 in the EXP_TOPO simulation on 2 August, because three-dimensional temperature changes dynamically resulted from three-dimensional current transports at an earlier time. Due to the absence of steep bottom slope, Damrey-induced upwelling off the Subei Shoal in the EXP_TOPO was much weaker than that in the CTR simulation ( Figure 10E ). The reduced upward transport of cold water then led to weaker temperature cooling. In addition, Damrey-induced upwelling also moved westward much closer to the coast in the EXP_TOPO simulation than in the CTR simulation. One could see that the bottom cold water in the EXP_TOPO simulation extended much closer to the coast and hence strengthened the thermal stratification in the coastal areas than that in the CTR simulation. This resulted in the Damrey-induced temperature cooling being located much closer to the coast than in the CTR simulation.

Therefore, the steep bottom slope was quite important in the development of the strong upwelling just off the Subei Shoal, which transported upward cold water and resulted in vertically consistent temperature cooling.

For typhoons moving northwestward over the Yellow Sea like Damrey, another important question should be discussed was whether the distance between the typhoon tracks and the bottom slope just off the Subei Shoal could greatly influence the three-dimensional temperature responses or not. To answer that question, we conducted a series of numerical experiments by directly shifting the sea surface winds during Damrey’s passage southward by 0.5°N (Wexp1) and northward by 0.5°N (Wexp2), 1.5°N (Wexp3), and 2.5°N (Wexp4), respectively. As examples, Figures 11A–D show the sea surface winds at 12:00 on 2 August in Wexp1−Wexp4.

Figures 11E–L display the temperature changes in the surface and bottom layers on 3 August since 31 July in the four numerical experiments. In the sea surface layer, the strong southeast-northwest band of temperature cooling in the South Yellow Sea moved northeastward, with its spatial extent gradually expanding and its intensity decreasing, from Wexp1 to Wexp4. In contrast, SST cooling off the northeastern Shandong Peninsula greatly strengthened from Wexp1 to Wexp4. In the bottom layer, temperature warming in the coastal waters and its maximum both moved northward from Wexp1 to Wexp4.

The above results indicated that the spatial patterns of changes in SST and BLT in the numerical experiments were basically same to those in the CTR simulation. Meanwhile, the underlying dynamics responsible for temperature changes in the numerical experiments were also same to those in the CTR simulation (Figure not shown). The largest differences between temperature changes in the CTR simulation and numerical experiments mainly occurred just off the Subei Shoal and in the coastal waters of the Shandong Peninsula. To give more details of these differences and the underlying dynamics, we showed in Figure 12 the temperature changes on 3 August since 31 July and vertical velocity on 2 August along a meridional section extending northward from the Subei Shoal to the Liaodong Peninsula (Sec3, the green line in Figure 11 ). From Wexp1 to Wexp4, vertically consistent temperature cooling just off the Subei Shoal weakened and disappeared, vertically temperature warming below sea surface in the southern coastal waters of the Shandong Peninsula and Liaodong Peninsula strengthened, and vertically consistent temperature cooling in the northeastern coastal waters of the Shandong Peninsula also strengthened. The underlying dynamics could be inferred from the spatiotemporal variations in vertical velocity along Sec3. From Wexp1 to Wexp4, upwelling was gradually replaced with downwelling on the bottom slope just off the Subei Shoal. However, there were strengthening downwelling in the southern coastal waters of the Shandong Peninsula and Liaodong Peninsula but upwelling in the northeastern coastal waters of the Shandong Peninsula.

To summarize, the distance between the typhoon track and Subei Shoal was also a crucial factor governing the three-dimensional temperature responses in the Yellow Sea to northwestward-moving typhoons. With the typhoon track moving northeastward away from the Subei Shoal, vertically consistent temperature cooling just off the Subei Shoal would abate and disappear, but vertically temperature warming below sea surface in the southern coastal waters of the Shandong Peninsula and Liaodong Peninsula and vertically consistent temperature cooling in the northeastern coastal waters of the Shandong Peninsula would strengthen.