In situ observations were conducted on a sheltered mudflat of Hailing Island, which is located on the west bank of the Pearl River Estuary, China (Figure 1A). According to records from the tide gauge on the west side of the island, the local tides are mixed semi-diurnal with an average of 2.5 m in tidal range. The spring and neap tidal ranges are >3.0 m and 0.5 m, respectively. Monsoon-driven winds are predominantly north-easterly in autumn and winter, south-easterly in spring, and southerly in summer. The multi-year average wind speed is 4.2 m/s. The province, where the study area is located, is hit by 3.6 tropical cyclones per year on average, which is the highest in China. Waves in this area are dominated by wind waves with a frequency of approximately 85%, and the average wave height is 0.2 m with the maximum storm surge reaching over 2 m.

The study site (111°58′42″E, 21°39′2″N) is located on the middle flat (at the mean water level) of the mangrove wetland on the northeast side of Hailing Island (Figure 1B). The south side of the island is open sea, while the north side is connected with an intertidal area. The tidal flat has a long-shore length of 2.4 km and a cross-shore length of 1.6 km. The mudflat faces to north-east-north (25°), with a mean slope of 2.5‰ (Hu et al., 2020). The upper flat is covered by dense mangrove forest, with a height of 2–3 m. As the study area is located in a bay on the back side of the open sea and a barrier island exists at the seaside of the bay, the study area is relatively sheltered. Therefore, the mangrove wetland provides an ideal habitat for benthos and birds, thereby sustaining high biodiversity in the region (Zhu et al., 2008; Hu et al., 2020).

An ADV (6.0 MHz vector current meter, Nortek AS, Norway) attached to an H-shaped sedimentary measurement frame was used to measure the three-dimensional flow velocities at a high sampling rate (Figure 1C) from March 26 to 26 April 2021 (Period 1), and from July 8 to 8 August 2021 (Period 2). The ADV recorded velocities and pressure with a burst interval of 5 min for a period of 300 s at a frequency of 16 Hz. The water pressure, measured by a silicone piezoresistive pressure sensor (Nortek, 2005), was used to analyze wave characteristics (Zhu et al., 2016). Based on the records of the distance between the transmitter of the ADV and the bed sediment surface from the extra acoustic beams of each burst, relative bed levels at high temporal resolution were also measured using ultrasonic echo-ranging with an accuracy of 1 mm (Andersen et al., 2007; Zhu et al., 2017). The accuracy is determined by the sound speed, which is affected by water temperature and salinity, as well as the boundary condition. The acoustic sensor and the pressure sensor were 0.3 m and 0.54 m above the sediment surface, respectively.

Measured wind data of the nearest site (111°58′12″E, 21°52′12″N) was downloaded from Institute for the Environment, the Hong Kong University of Science and Technology (http://envf.ust.hk). The interval of wind speed and wind direction is 3 h.

Data filtration was conducted prior to further analyses. The burst datasets with backscatter acoustic amplitude less than 100 counts and correlation less than 50% were removed.

The wave-induced bed shear stress, τ w , was obtained by analyzing the surface-elevation monitoring data. Significant wave height, H s , and significant wave period, T s , were derived from high-frequency pressure data via linear wave theory (Tucker and Pitt, 2001) (see http://neumeier.perso.ch/matlab/waves for the routines). The near-bed peak orbital excursion ( A ^ δ ) and peak orbital velocity ( U ^ δ ) can be expressed as (van Rijn, 1993): A ^ δ = H 2 ⁡ sinh ( k h ) (1) U ^ δ = ω A ^ δ = π H T ⁡ sinh ( k h ) (2) where H is the wave height (m), k=(2 π /L) is the wave number ( m − 1 ), with wave length L=(g T 2 /2 π )tanh (kh). Where h is the water depth (m), ω is the angular velocity ( s − 1 ), and T is the wave period (s). In practice, the significant wave height H S and significant wave period T S are used for H and T in the formulae. The time-averaged (over half a wave cycle) bed shear stress caused by waves, τw (Pa), can be expressed as (van Rijn, 1993) τ w = 1 4 ρ w f w U ^ δ 2 (3) where ρ w is the water density (kg/ m 3 ), and fw is the friction coefficient (-), which is determined by the hydraulic regime (van Rijn, 1993): f w = { 2 R e w − 0.5   , R e w ≤ 10 5 ( l a m i n a r ) 0.0521 R e w − 0.187   , R e w > 10 5 ( s m o o t h   t u r b u l e n t ) 0.237 r − 0.52   , ( r o u g h   t u r b u l e n t ) (4) where R e w = U ^ δ A ^ δ V and r = A ^ δ K S are the wave Reynolds number (-) and relative roughness (-), respectively. Parameter ks is the Nikuradse roughness value given as k s = 2.5 d 50 , where d50 is the median grain size of the bed sediment, and ν is the kinematic viscosity of sea water ( m 2 / s ).

Wave-orbital-motion-induced velocities were removed from the total burst velocities by the moving-average method with a 1-s smooth window (Williams et al., 2003). Subsequently, the turbulent kinetic energy method with vertical fluctuations (TKEw) was used to calculate the bottom bed shear stress τ _c (Pa). The formula is as follows (Kim et al., 2000; Zhu et al., 2016): τ c = C ρ w w t 2 ¯ (5) In which ρ _w is seawater density, constant coefficient C=0.9, and w t 2 ¯ is the change rate of near-bed vertical turbulent velocity.

Each measurement period lasted for 1 month or 62 tidal cycles. However, because the ADV on the middle flat was equipped with probes at 0.3 or 0.5 m above the seabed, low tides were not measured. During some periods after April 18, the velocity and wave data were missed due to battery failure (Figure 2). Overall, data from 40 tidal cycles were obtained in Period 1, and data from 45 tidal cycles were obtained in Period 2.

The average wind speed during both periods was 3.0 m/s. We defined the tidal cycle during a wind event as one where the intratidal maximum wind speed exceeded 6.0 m/s. An obvious wind event occurred during each period: around April 9 during Period 1 (Figure 2A) and around July 20 during Period 2 (Figure 3A). The maximum wind speeds and predominant directions were 8.0 m/s and north-east-north in Period 1, and 17.0 m/s and north and east in Period 2, respectively (Table 1).

Measured maximum water depth, h, varied from 0.5 m to 1.7 m during Period 1 (Figure 2A), and from 0.5 m to 2.0 m during Period 2 (Figure 3B). Tides in Period 1 were irregular, i.e., the neap-spring tidal cycle was not 15 days. Significant wave height, H _s , was ordered by centimeters with mean values of 0.08 m and 0.06 m in each measurement period. The mean H _s during wind events was 2.6 times the value during calm conditions (Table 1). Correspondingly, the wave-induced bed shear stress, τ _w , was approximately three times larger during wind events than during calm conditions (Table 1).

The current velocity, U, showed the tide-driven spring-neap cyclicity. However, during the wind event in Period 2, U was greater than normal values (Figure 5E). Thus, the current-induced bed shear stress, τ _c , was four times larger during wind events than during calm conditions.

The bed level, Z, was relative stable without obvious spring-neap tidal cyclicity (Figure 3G and Figure 5G). The bed level statistics showed that the standard deviation was 3.0 mm. This indicates that the inter-tidal bed level change was generally at the millimeter level. During wind events, the bed level variation reached 26–27 mm.

During a typical calm tidal cycle, the bed level remained nearly constant (Figure 4D1 and Figure 5D1). Several tidal cycles show erosion-deposition-erosion cyclicity corresponding to the peak-slack-peak process (Figure 6). In contrast, the intratidal bed levels during wind events showed different patterns. During the wind events of Period 1, there were three patterns of bed-level changes:

(1) Type I: cyclicity motion (Figure 4D2 and Figure 5D3): the April 9 tidal cycle showed clear cyclicity in a period of approximately 2 h. The impact of waves (τ _w ) was stronger than that of currents (τ _c ) in this case.

The maximum variation in the bed level during each tidal cycle, ΔZ, was used to quantify the intratidal bed-level dynamics. ΔZ was less than 10 mm for 90% of the measurement period and during relatively calm conditions (wind speed <6.0 m/s) (Figure 7A). As the wind speed increased, ΔZ increased correspondingly. The linear regression analysis showed that ΔZ has a stronger relationship with bottom orbital velocity (U _δ ) and τ _w than current velocity (U) and τ _c (Figure 6). This indicates that bed-level changes are more sensitive to waves.

Mudflat bed is composed of fine sediments, which are sensitive to hydrodynamic disturbances. Sheltered mudflats are less dynamic than open mudflats (Hu et al., 2017). For instance, monthly variations in the bed levels of open mudflats in the meso-to macro-tidal range reached 10–30 cm (Miao et al., 2016; Zhu et al., 2017). Even the bed variation can be measured in centimeters under normal weather conditions (Zhu et al., 2014). In brief, if a mudflat is open and exposed to strong hydrodynamic forcing, daily and monthly bed variations are in the order of centimeters and tens of centimeters, respectively. In contrast, sheltered mudflats were quite stable. Our study shows that the monthly bed variation was only approximately 1 cm. These sheltered mudflats can sometimes be stable over a longer time scale. Studies of sheltered mudflats in the Western Scheldt Estuary, the Netherlands, showed that the bathymetry profile was stable, with the variation of bed level less than 10 cm over a period of 25 years (de Vet et al., 2017; Maan et al., 2018). A stable bed environment is a potentially advantageous for maintaining the biodiversity and biomass of wetland plants and benthic animals on sheltered mudflats (Herman et al., 2001; Chargulaf and Tibbetts, 2015).

Bed-level changes between tidal submergence, which are measured by the double/triple-rod method, buried-plate measurements or optical detections, are useful for understanding the daily evolution of mudflats (Yang et al., 2003; Hu et al., 2015; Zhu et al., 2017). However, acoustic ranging detection shows that the maximum intratidal bed level variation is much greater than the net intertidal bed level changes. This phenomenon is notable in open mudflat systems (Zhu et al., 2014; Zhu et al., 2017). The tidal cycle on the morning of April 9 demonstrates the same phenomenon on sheltered flat. The intratidal maximum variation was 20 mm, whereas the net intertidal bed-level change was 9 mm (Figure 7E2). To date, the acoustic measurement of bed surface dynamics at high temporal resolution is the only approach for determining intratidal bed level changes (Thorne et al., 2018; Zhu et al., 2019). By applying advanced acoustic bed detection, our study further indicates that the sheltered mudflat is more variable as we expected.

The typical intratidal bed-level changes in normal weather are shown in Figure 6E1 and Figure 7E1. There may be a slight trend, and the bed level was nearly constant in general. This agrees with the measurements from other sheltered mudflats (Zhu et al., 2019). Some tidal cycle still shows the cyclicity of erosion during flood and peak and deposition during slack water (Figure 6). Notably, the minimum ΔZ was larger than 1 mm, which is the accuracy of bed level detection. So far, the accuracy of acoustic bed detection is ± 1–2 mm, depending on individual instrument (Gallagher et al., 1996; Thornton et al., 1998; Christie et al., 1999; Andersen et al., 2006; Turner et al., 2008). In calm weather, both wave-and current-induced bed shear stresses were less than 0.1 Pa (Figure 6C1 and Figure 7C1), which is the typical value of critical bed shear stress for mud erosion (Winterwerp et al., 2012). No erosion and little deposition lead to a stable bed level with a slight accumulation of sediment. In our study, 73.7% of the measured ΔZ was <5 mm, indicating the stability of the sheltered mudflat. When the tidal-averaged wind speed was >3.0 m/s, more drastic intratidal bed-level changes were observed. The mechanism of local bed-level change differed from that of hydrodynamic conditions.

The observed fluctuation in the bed level was partially caused by measurement errors and accuracy issues, as mentioned above. By incorporating a moving average and despike, such noise signals can be minimized. The smoothed bed-level series exhibited a periodic signal (Figure 4D2). We defined this as the Type I bed-level variation pattern, which was closely related to the formation and migration of bed ripples. The magnitude and period of the bed-level fluctuation, referring to ripple height and ripple migration periods, were 1–2 cm and 1–2 h, respectively. This matches the typical ripple migration parameters of mudflats (Baas et al., 2013; Lin and Venditti, 2013; Zhu, 2017). Under dynamic environments with changing currents and waves, reworking and mobilization of bed sediments leads to formation of bed ripples in coastal systems (Catano-Lopera and Garcia, 2006; Thorne et al., 2018; Guerrero and Guillen, 2020; Guerrero et al., 2021; Stella, 2021). Bed ripples preferably occur in the presence of waves (Chakraborty, 2001; Lorenz and Valdez, 2011; Guerrero and Guillen, 2020; Jin et al., 2020). For the tidal cycle on the morning of April 9, τ _c was as low as 0.1 Pa due to low tide, meaning that flow could hardly stir up surface bed sediment, whereas τ _w was 2.5 times higher (Figure 4D2). Such wave-dominated conditions further prove that the Type I bed-level variation is caused by ripple migration.

Type II intratidal bed-level variations show an unclear fluctuation pattern. A typical example is the tidal cycle observed on July 20 morning (Figure 5E2). Both τ _c and τ _w were greater than 0.1 Pa, and they are comparable. This means that both the current and waves played a role in bed sediment movement. Under normal weather conditions, the intratidal bed level shows a pattern of degradation during tidal flood and ebb peaks, and accretion during high slack water on open flats (Andersen et al., 2007; Zhu et al., 2014). However, a continuous accretion trend was observed in the tidal cycle of July 20. During wind events, wind-induced flow (Figure 3D) and wind-induced turbulence result in that τ _c does not increase with tides (Figures 3, 4F) (Banerjee et al., 2015; Su et al., 2015). Meanwhile, the advection of sediment flow was strong during wind events. These two factors may break the peak-slack water cyclicity, but accretion or degradation trends are determined by the timing of wind event with regard to tidal cycle, sediment supply, residual flow, and settling time (Ding et al., 2003; Freeman et al., 2015; Xu et al., 2015; Zhu et al., 2017). The bed-level change pattern of the middle tidal flat was the most dynamic and difficult to predict (Fan et al., 2006). Additionally, there was no clear ripple migration signal in the bed-level series, as sheet flow during wind events tends to wash out ripples (Li and Amos, 1999; Thorne et al., 2018; Guerrero et al., 2021).

However, the low tide tidal cycle on July 20 (Figure 5D3) showed a different bed level pattern than that of April 9 (Figure 4D2). This cycle marked the transition between Type I and Type II variation, that is, the intratidal bed level change consisted of a cyclic motion and continuous accretion trend. The tidal conditions of the two tidal cycles were similar, but the current flow was quite different. Wind direction was offshore during Period 1 and almost onshore during Period 2 (Figure 2A and Figure 3A). Correspondingly, tidal flow was less influenced by winds during Period 1 (Figure 4B2), or by the northern component of the flow velocity (Figure 5B3). However, the eastern component was considerably influenced (Figure 5B3). There was an eastward flow during the majority of the tidal cycle, which was likely driven by wind induced flow pattern. Both τ _c and τ _w were much higher than those of other tidal cycles, reaching up to 0.7 Pa.

In summary, wave and tide conditions determine the intratidal bed level changes: 1) low τ _c and low τ _w lead to a stable bed; 2) low τ _c and high τ _w promote the generation and migration of bed ripples (Type I); 3) comparable τ _c and τ _w , with medium-to-high values lead to non-cyclicity bed-level change patterns (Type II); and 4) high τ _c and high τ _w result in a combination of Type I and Type II variations.

Sheltered mudflats is considered stable, with the bed-level-change is in the order of millimeters in the time scale of tidal-cycle to even month, and is less affected by winds (Andersen et al., 2006; Hu et al., 2017). However, sheltered mudflats may be more dynamic than previously believed. Our study shows that the critical wind speed triggering distinct changes in the local sediment bed is approximately 3.0 m/s, a light-to-gentle breeze according to the Beaufort wind scale. If the wind is onshore, the bed-level-change can immediately respond to the wind. Otherwise, the bed need about one tidal cycle, i.e., half a day, to respond to wind. Therefore, more significant morphological changes would be expected during gale or storm events, wherein wind speed exceeds 15 m/s.

Although the sheltered mudflat is easily disturbed, it also recovers rapidly. The monthly bed-level measurements showed that the bed level fluctuated at a nearly constant level (Figures 2, 3G). These sheltered mudflats are stable and can recover quickly even after disturbance (Maan et al., 2018). For tidal flats, rapid recovery of disturbed sediment beds is enabled by the redistribution of sediments from high-energy to low-energy areas (Yang et al., 2003; Friedrichs, 2011). For low dynamic tidal flats, such as a sheltered flat, the distance of sediment movement may be smaller than in open tidal flats, and are likely to be transported back to maintain an equilibrium state after wind events (Zhu et al., 2017). The sediment transport patterns of sheltered mudflat systems must be further studied on a horizontal spatial scale.