In the shallow water offshore Guangdong Province (Figure 1), approximately 3100 km acoustic backscatter data were collected in July 2020 using an Innomar SES2000 Light parametric sub-bottom profiler, with primary frequency (approximately 100 kHz) and ping rate (up to 50 pings/s). The draft (source depth) of the transducer was 2 m. In this study, we focused on the four easternmost lines with a NW-SE direction (L1–L4; Figures 1, 2) to study the spatiotemporal evolution of the ISWs on the continental shelf. These lines captured a soliton train three times in 20 h repeatedly (Figure 2).

Innomar ISE software (Innomar Technologies GmbH) was used to process the acoustic backscatter data. The acoustic image processing flow was (1) envelope algorithm imaging; (2) noise attenuation by median filter; and (3) time-depth conversion assuming an averaged sound speed of 1500 m/s.

Optical satellite images, such as Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) images, are widely used in ISW studies (Li et al., 2013; Tang et al., 2014, 2015, 2018). True color satellite images with a spatial resolution of 250 m are available from the website.¹ In this study, the data from MODIS and VIIRS sensors on NASA spacecraft (Terra, Aqua, NOAA-20, and Suomi) were used to image the sea surface signature induced by ISWs (Figures 3, 4). Benefitting from the good weather conditions on July 18 and 19, 2020, the ISWs in the shallow water region were clearly imaged by three satellite datasets. The mean wave propagation speeds and directions can be measured from the spatiotemporal variations in the satellite images and acoustic backscatter data (Figures 3, 4).

During the cruise, two XBTs were deployed simultaneously along the survey lines on July 9th at 13:24 and July 10th at 20:25 UTC (Figures 1, 5). The in situ XBT data, less than 50 km away from the four acoustic backscatter survey lines (L1, L2, L3, and L4), could be used as the background temperature structure in the study area in early summer (Figure 5).

Depth profiles of temperature and salinity from three CTD stations were collected from July 27th to August 16th, 2009 (Figures 6A,B; Chen et al., 2017). These historical temperature and salinity data reveal that the water structure can be simplified as a two-layer structure according to the depth of the mixed layer in the study region (Figure 6). Therefore, two-layer structures are finally composited for theoretical KdV and eKdV model calculations and result validation for corresponding ISWs.

Sea surface temperature (SST) images used in this study were derived from the Group for High Resolution Sea Surface Temperature (GHRSST) data, which are available from the NOAA website.² The daily GHRSST data are satellite-composite products by the Jet Propulsion Laboratory (JPL) with a spatial resolution of 1 km derived from the Advanced Microwave Scanning Radiometer (AMSRE), MODIS, WindSat, Advanced Very High Resolution Radiometer (AVHRR), and in situ observation data.

In the continental shelf of this study region (shallower than 300 m), the water column stratification can often be simplified as a two-layer structure according to the density structure (Figure 6). Therefore, we can use a two-layer model to calculate the non-linear phase velocity and the characteristic half-width of ISWs in the study area (Ostrovsky and Stepanyants, 1989). An internal soliton propagation process in continuous water stratification can be described by the KdV equation (Djordjevic and Redekopp, 1978):

and eKdV equation:

where c, β, α, and α₁ are the linear phase velocity, dispersive coefficient, quadratic, and cubic non-linear parameters, respectively. For the simplified two-layer water stratification model, these parameters can be calculated from the following relationships (Ostrovsky and Stepanyants, 1989):

where h₁, h₂, ρ₁, and ρ₂ are the thicknesses and the densities for the upper and lower layers, respectively.

For finite-amplitude waves, the theoretical ISW solutions of Equations 1, 2 have the following form (Ostrovsky and Stepanyants, 1989):

where Δ, D, V_KdV, and V_eKdV are the characteristic widths and the phase velocities from the KdV and eKdV models, respectively.

Therefore, the vertical velocities of particle motions can be computed from the analytical solutions (7) and (8) via partial derivative W(x) = ∂⁡η(x,t)/∂⁡t|_{t = 0} (Trevorrow, 1998; Teague et al., 2011):

and

for the KdV and eKdV models, respectively.

Many ISW packets were captured during the survey cruise using 100-kHz acoustic backscatter data in early summer 2020 (Figure 2). Due to continuous and repeated observations of the ISWs, some soliton trains were captured more than once, as shown with similar features and reasonable locations on different lines. In this study, four representative wave packets (ISW1, ISW2, ISW3, and ISW4) at water depths from 24.5 to 40 m were selected for further analysis (Figure 2). Figure 7 shows the detailed backscatter feature and the strategy of deriving the waveform parameters for ISW1, ISW2, ISW3, and ISW4. The captured locations for each leading soliton are plotted in Figure 3, while the observed parameters of positions, times, upper layer thicknesses, water depths, amplitudes, and full widths of these ISWs are listed in Table 1.

ISW1, ISW2, and ISW3 are from the same soliton train captured three times within 20 h from west to east on the repeating acoustic sections of L2–L4 (Figures 2, 7). ISWs propagated from deeper water (32 m) to the shoaling water (24.5 m) from approximately south to north (Figures 2, 7A–F and Table 1). The wave packet ISW1 was first captured on the L2 section, as it had no appearance on the preceding section L1 (Figure 2). The packet consists of at least four solitons with positive polarity and representative waveforms within the first 0.5 km from the leading soliton (Figure 7A). The solitons’ amplitudes are not ordered from the leading soliton to the trailing ones. The third soliton has the largest amplitude of 9 m, while the leading wave amplitude is approximately 7 m. Approximately 0.5 km behind the leading soliton, wave fields become chaos, and solitons are not easy to identify. The passing packet deepens the mixed layer, forming a bore-like structure of ∼2 km. At the trough of the bore, the thickened mixed layer induced by the passing ISWs reverses the non-linear parameter α, and thus, elevated waves are formed (Figure 2B).

In contrast, packets ISW2 and ISW3 only have one leading soliton individually (Figures 2C,D, 7C–F). Similar to ISW1, the amplitudes of the leading waves of ISW2 and ISW3 are ∼7 m, but their following wave amplitudes are only 1–2 m and cannot be recognized as solitons. The mixed layer is suppressed to specific depths after passing the solitons, forming hydraulic jumps. Therefore, the solitons only have leading edges, and rear edges are not well developed, indicating a significant energy loss, and the stratifications are failed to be restored to the starting depths within ∼10 km (Figures 2C,D).

On line L4 in the deeper water region, the large soliton train (ISW4) with more than 10 following well-developed solitons was captured 2 h beyond ISW3 and ∼20 km away from ISW3 (Figures 2D, 7G). The soliton amplitudes of ISW4 are sequentially ordered from the leading wave (∼13.3 m) to the trailing waves (∼5 m). In addition to the deepened mixed layer, the sediments were suspended several meters above the seafloor after passing the solitons (Figure 7G). Meanwhile, because of its larger spatial scale, ISW4 can be observed from the satellite image (Figures 3B,C).

From two satellite images acquired by NASA MODIS Terra and NOAA-20 VIIRS satellites on July 19th, 2020 (1 day after acoustic backscatter survey), many alternating dark and light curving signatures were caused by ISWs from the Dongsha Plateau to the Guangdong coastal water (Figure 4). The propagation speeds and directions vary significantly from the deep-water region to the shallow water region according to the position differences of the ISW crests with a 2-h lag (Figure 4). The propagation speeds at locations #1–#3 gradually decrease from ∼1.6 to 0.45 m/s, and the propagation directions turn from ∼270° to 340°, corresponding to the shoaling water depth from the continental shelf to the coastal area. Therefore, it is expected that the ISWs observed from the acoustic images in the study region should have propagation parameters similar to those of the satellite-observed results at location #3 (Figure 4).

By identifying the time and space differences of the wave crests using the backscatter data, the estimated mean propagation velocities were ∼0.44 m/s (between ISW1 and ISW2) to ∼0.41 m/s (between ISW2 and ISW3), assuming that the ISWs were propagating northward (Figure 2). However, the propagation velocities of ISW1, ISW2, and ISW3 cannot be determined by satellites because of their small spatial scales. Therefore, we resort to the theoretical calculation to verify the propagation in the next section. In contrast, ISW4 on L4 was captured by the MODIS image ∼4 h after capture by acoustic backscatter observation (Figure 3C). The estimated mean propagation velocity of ISW4 from the acoustic backscatter to the satellite is ∼0.35 ± 0.04 m/s, which is less than the results from two satellite images in the coastal region (Figure 4).

There are generally two generation sources of ISWs observed on the continental shelf of the northern SCS. One is remotely from the Luzon Strait, and the other is locally from the northern SCS continental shelf (Lien et al., 2005; Cai et al., 2012). According to satellite images, most ISWs in the northern SCS propagate westward from the Luzon Strait and are refracted near the Dongsha Plateau and then torn by the island into two new trains (northern and southern arms) continuing to propagate northwestward to coastal areas (Li et al., 2013; Wang et al., 2013; Ma et al., 2016).

In this study, we suggest that large-scale ISWs were also generated in the Luzon Strait based on the following considerations. Two satellite images within the acoustic backscatter observation period directly show that large-amplitude ISWs are successively distributed from the Dongsha Plateau to the coastal areas (Figure 4). Meanwhile, ISWs at the Dongsha Plateau must be generated in the Luzon Strait, and ISWs are still traceable tens of kilometers south of the study region evolved from the northern arm of the ISW train (Figure 4). Moreover, the surface signature of ISW4 was clear enough and was captured by both satellite and acoustic backscatter data. Therefore, it can be safely inferred that the periodically active ISWs in the coastal region are generated at the Luzon Strait and then propagate into the study region after long-range evolution.

The acoustic backscatter data may have captured some locally generated internal waves with internal tides on the continental shelf. For example, there is an isolated ISW approximately 1 km ahead of ISW4 in Figure 7G. This ISW does not belong to either the trailing soliton of ISW3 or the leading soliton of ISW4, which are successive ISW packets from the Luzon Strait. Therefore, the isolated ISW might be generated locally with steep topography, most likely at the continental shelf break by the internal tide. More integrated observations are necessary to identify the origins of these ISWs.

In this study, all ISWs have large amplitude frontal edges and small amplitude rear edges (Figure 7). The frontal edges are steep and well-developed. In contrast, the rear edges are much weaker and fail to resort to the normal depth. It is suggested that these divergent zones of the rear edges were under collapse by losing potential energy and finally forming asymmetric waveforms. The breaking event of ISWs is mostly caused by Kelvin–Helmholtz instability and convective overturn in the steep rear edge, forming the breaking tail of the ISWs. These features are similar to previous simulated and observed results in shelf-coastal regions (Ostrovsky and Stepanyants, 1989; Novotryasov et al., 2015).

Numerical simulations suggest that large internal waves always dissipate energy by developing a dispersive wave tail on gentle slopes (e.g., Vlasenko and Hutter, 2002). The coastal topography in the study region is gentle (slope angle γ < 0.17°) with small variation, and seawater stratification might be a key factor affecting the propagation and dissipation of ISWs. Following the ISW breaking criterion h_b = h₁ + η₀/(0.8°/γ + 0.4), where h_b is the critical water depth for solitary wave breaking (Vlasenko and Hutter, 2002), and according to the stratification parameters of ISW1–4 (Table 1), the critical water depth h_b is approximately 17.5 m. Such a water depth is typically smaller than the water depth (∼30 m) of the study. Therefore, the prediction suggests that stratification favors passing ISWs through the study area with a large leading wave and dispersive trailing waves, consistent with acoustic observations (Figure 7). The trailing waves weaken as the ISWs propagate toward coastal water.

Strong dissipation zones could be identified as a series of bright (white) bands in the northern part of the study region (Figures 3, 4). However, these dissipation zones should differ from the ISW-induced dissipation zones, although they highly overlap with the ISW field spatially. First, there are no clear frontal lines relating to the leading waves of the ISWs. Second, the dissipation zones move eastward with a very high speed, contradicting the northwestward propagating ISWs with quite low speeds.

The water masses in the coastal region are complex. They include solar-heated SCS water in the south, eastward drifting river water, along coastal wind-driven currents, and coastal upwelling cold water. On L1 and L2 in the shallow water region, the acoustic backscatter images show the upper layers with substantially weak without a continuous scattering interface are above the lower thick strong scattering layers (Figure 2). We suggest that these phenomena were the river plumes sitting on top of the seawater close to the river mouth side. Moreover, the SST acquired on July 18 and 19, 2020 in the northern SCS reveals high-temperature SCS water in the south, low temperature close to the shore region, and a transition zone between (Figure 10).

The dissipation zones are highly correlated with the temperature transition zone or the frontal zone between the northern ventilated water and southern heated water (Figures 3, 10, 11). On the L3 and L4 acoustic backscatter sections, the scattering feature has been intensified even before the arrival of ISWs (Figure 2). This region is within the dissipation zone on the satellite images and the temperature transition zone on the SST image (Figures 3, 10, 11). Therefore, a schematic model is proposed for internal wave propagation and dissipation in this study region (Figure 12). Far field ISWs generated at the Luzon Strait can penetrate into the wide shelf region because of the strong stratification caused by solar-headed surface water. Distinct dissipation processes at the trailing waves of the ISW packets make the ISWs weaken toward the shoreward region. These ISWs finally disappear before entering the upwelling cold water, as their stratification does not support ISW transmission. The surface dissipation patches are near-surface internal wave breaks induced by the wind force. Their occurrences show a quick response to the wind force along the frontal instability zone.

Our study area is offshore of the Guangdong–Hong Kong–Macao Greater Bay Area. It is one of the most developed areas in China. There are numerous near-shore artificial facilities in the study area, such as wind power facilities, cross-sea bridges, oil platforms, and submarine cables. Large amplitude ISWs can lead to strong horizontal and vertical currents in the ocean. A few studies have been carried out to evaluate the impact induced by strong waves on offshore structures (e.g., Cai et al., 2006). This study shows that the duration of ISWs is a few minutes, and the vertical displacements and vertical velocities are on the order of ∼10 m and 10 cm/s, respectively. For example, ISWs recorded that the vertical displacement was up to 20 m at a period of ∼6–8 h at station ZHJ2 (Figure 1; Lee et al., 2021).

During the cruise, a high-resolution multichannel streamer, which is geophysical equipment used in near-surface engineering surveys, was deployed simultaneously with acoustic backscatter. Due to the lack of depth controlling units, the streamer went through strong vertical movements induced by ISWs, resulting in apparent undulating topography waves of ∼500 m wavelength and ∼2–4 m amplitude. Tang et al. (2015) have also reported that the streamer with depth controlling units was uplifted to the surface by the ISWs for ∼20 min (∼3 km) in NE SCS. In addition, offshore engineering vessel operations, such as dynamic positioning and drilling, are often influenced by ISWs.

Artificial facility safety is sensitive to sediment siltation or erosion on the seafloor. ISWs can give rise to remarkable interactions between the seafloor and ocean currents in a relatively short time (Apel et al., 2007; Huang et al., 2016). The total amount of sediments resuspended by the ISWs is 2.7 times that of the sediments supplied by the river in the northern SCS (Jia et al., 2019). The suspended sediments diffuse along the isopycnals and redeposit in a different area depending on sediment grain sizes and ocean currents (Tian et al., 2019a,b). Acoustic backscatter data have shown that ISWs induce large-scale sediment resuspension offshore of Guangdong. Therefore, ISWs combining coastal currents may have potential harm to offshore engineering structures by inducing large-scale sediment siltation or erosion.