Introduction

Turbulence in tidal currents affects on the movement of animals and fishing gear operations. Recently, the turbulence flow as a three-dimensional (3-D) component of velocity was measured in coastal tides and analyzed using acoustic Doppler current profilers (ADCP) or acoustic Doppler velocimeters (ADV) (Thorpe, 2007).

The turbulence flow, which can be defined by the velocity field, is not repeatable in either the whole or part of the flow domain referred to as laminar flow (Bernard and Wallace, 2002). In nature, water is generally in a state of non-uniform and variable motion that is referred to as “turbulence”, although there exists no straightforward, unambiguous definition of the term (Thrope, 2007). Turbulence is generally accepted to be an energetic, rotational and eddying state of motion that results in the dispersion of material and the transfer of momentum, heat and solutes at rates far higher than those of molecular processes alone.

Turbulence is often dominated by coherent structure activities and turbulent events which may be defined as a series of turbulent fluctuations that contain more kinetic energy than the average turbulent fluctuations within the relevant section. Therefore, studies in tidal areas of 3-D turbulence structures have been hampered by the limited size of the observed area due to equipment limitations; e.g., Laser Doppler Velocimetry in tank experiments (Pichot et al., 2009).

The characteristics of turbulence in tidal flow have been investigated in various contexts, such as sediment suspension (Ham et al., 2001; Chanson, 2010; Chanson et al., 2011; Yuan et al., 2009 ), tidal straining (Simpson et al., 2005; Thorpe et al., 2008), and tidal energy turbines (Rippeth et al., 2003; Thomson et al., 2012).

Turbulent flow can affect swimming fish (Shadwick and Lauder, 2006; Liao, 2007; Webb and Cotel, 2010; Tritico and Cotel, 2010) and fish escape from codends (Kim, 2012; 2013a). Flow inside codends was analyzed by measuring the turbulence intensity and periodicity for shrimp beam trawls and bottom trawls (Kim, 2012; 2013a). Flow inside codend towed fishing gear may be affected primarily by tidal flow, followed by wakes inside fishing gear (Bouhoubeiny et al., 2011) and ship motion including the wave effect (O’Neill et al., 2003). Turbulence measurements close to Korea were carried out in the western Yellow Sea off the coast of China in both relatively shallow (Liu et al., 2009a; 2009b; Yuan et al., 2009) and deeper (Lozovastsky et al., 2012) waters.

Thus, the 3-D components of tidal turbulence can affect animal kinetics such as swimming performance, or the drag or wake of fishing gear. However, tidal turbulence at the sea bottom has, to our knowledge, not been analyzed in Korean waters as precise high-frequency 3-D flow. Hence, the purpose of this preliminary study was to determine the physical elements of tidal turbulence, such as turbulence intensity and dominant period focused on fish movement and fishing gear dynamics. Two locations at a trench between Manjido and Bujido offshore of Tongyoung, Korea were selected for their different geological conditions and as locations used as the fishing grounds for shrimp beam trawls, two-boat seine nets, etc. 3-D flow measurement data were collected using ADV and the kinetic energy, turbulence intensity, and period in relation to fishing gear dynamics and movements of swimming fish during fishing operations were analyzed.

set in an upright position using a circular frame of 10-mm iron rods and 20-kg lead weights to anchor it firmly on the seabed (Fig. 1). The velocimeter measured water velocity based on a volume of water of 1.5-cm diameter 16 cm from the acoustic probe with an accuracy of ±0.5% according to the manufacturer’s specifications. The overall length of the velocimeter was 82.5 cm; its diameter was 7.5 cm; and its weight in water was 1.5 kg. The instrument also included sensors for measuring temperature (accuracy, 0.1°C), direction (compass accuracy, 2°), tilt (pitch and roll accuracy, 0.2°), and depth pressure (accuracy, 0.25%). A holding frame that formed a quadratic pyramid (diameter 82 cm, height 63 cm) was used to maintain the device in the orthogonal position, maintaining its sampling position at 1 m above the seabed. The equipment was deployed with 8-mm PP rope of 1.5 times length of depth in connection with a 2.5 kg weight and a spherical buoy (diameter, 30 cm) to a upright rod with a flag on the surface.

Observations were performed at two sites at a deep trench between Manjido and Bujido south-west of Tongyoung, as shown in Fig. 2. The trench was oriented north-west to south-east at depths ranging between 50 and 70 m, while the depth of the surrounding area ranged between 30 and 40 m. These sites are generally used as the primary fishing ground of coastal fishing boats, including those using shrimp beam trawls and two-boat seine nets. Measurements were carried out five times (at sites S1, 2, 4, 5, 6) west of Manjido and once (site S3) north of Bujido. The sampling times and tidal conditions are shown in Table 1.

Tidal flow velocity was defined as three components: east (positive)-west (negative) velocity (V_x), north (positive)-south (negative) velocity (V_y), automatically designated by internal compass, and depth (V_z) directions. Vector measurements of tidal current were set at 2 m/s maximum velocity as a 3-D axis with east, north and depth directions with a 16-Hz sampling rate mostly at 34 practical salinity units (psu). The acoustic speed was also calibrated automatically using data from the temperature and pressure sensors, and all sampling data were stored in internal memory.

Following data collection at sea, the landing state of the data was checked using the tilt, noise and correlation data. When the standard deviation of the pitch or roll data per second was higher than the tilt accuracy; i.e., 0.2°, the vector could be shaking due to slack landing. Therefore, measurements taken at Stations S4 and S5 with a tilt of pitch and roll >1° were not included in the analysis (Parra et al., 2014). Additionally, for the 3-D flow velocity data, average signal to noise ratios <15dB and average correlation values <70% were eliminated (McLelland and Nicholas, 2000).

Based on the definition of turbulence flow in Bernard and Wallace (2002), the tidal turbulent flow was analyzed as turbulence intensity, kinetic energy and oscillation period for analysis of the fisheries variables. The three components (V_x, V_y, V_z) of velocity data for the entire measurement period for each site were assessed every 1 min (data n=480 = 8 Hz × 60 s or 960 = 16 Hz × 60 s) to determine turbulence intensity and kinetic energy, as shown in Table 2. The mean resultant velocity U_m represents an estimate for consecutive 1-min samples as follows:

Accordingly S_x, S_y, and S_z represent the standard deviation of flow velocities per minutefor V_x, V_y, and V_z, respectively. Then, turbulence kinetic energy (K_e, m²/s²)can be represented as:

Next , the turbulence intensity (Tr, %) as a ratio can be defined as the square root of K_e divided by the mean flow velocity U_m as:

For periodic analysis, each sampling dataset recorded at 8 and 16 Hz was converted for each resultant velocity (Vr) and horizontal flow direction (Ab) using the V_xand V_y components for strong tides of six flood durations and two ebb durations during each measurement (Table 2). Using these converted data, the specific period was estimated by fast Fourier transform (FFT), global wavelet method, and Morlet wavelet method.

MATLAB (MathWorks) was used to analyze the periodicity of the tidal flow velocity and direction for shorter periods <30 s using FFT for selected strong flood or ebb data from each measurement. The Morlet wavelet method in MATLAB was well localized in both time and space for periodicity analysis, as applied in many studies (Yuan et al., 2009; Seena and Sung, 2011). The Morlet continuous wavelet method (CWT) was also viewed for 2000 burst sampling data for shorter periods <10s, which was more relevant for the analysis of swimming fish (Kim et al., 2008) or fishing gear movements (O’Neill et al, 2003; Kim, 2012; 2013a). In addition, the global and Morlet wavelet spectra for velocity and direction were calculated to determine the moderate dominant period <30 s using software from Torrence and Compo (1998) and Zhang et al. (2010).

The above methods, however, are unable to calculate the amplitude as the velocity or directional range of a period, and are also unsuitable for periodic data with a wide range. Therefore, the event analysis method, such as the difference between peak and valley values, was adapted using software designed by ourselves based on Narasimha et al. (2007) to analyze the depth change of the shaking codend (Kim, 2013b). When the depth increased until the peak and then decreased in consecutive time series data, the peak value was detected as a positive peak value, and the valley value was detected as a negative valley value. Fig. 3 shows the results using the event analysis method for Station S2b. The minimum time interval between peaks or valleys was limited to 1 s, taking into consideration the sampling rate and oscillation of fishing gear (O’Neill et al., 2003; Kim, 2013a). The initial threshold value (Vi, +:peak, -:valley) between neighboring peaks or neighboring valleys was categorized by the mean value ± 0.5 SD for data at 30-s intervals, respectively. Then, a peak event value (Vp+ for peak, Vp- for valley) can be selected as the highest value among Vi. The mean period was estimated from the average of the total intervals between peaks and between valleys, while the mean amplitude was estimated from the velocity or directional difference between peak events and valley events that occurred only consecutively as a pair.

Results and discussion

Fig. 4 shows the 3-D tidal velocities Vx west, Vy north, and Vz depth directions and their resultant velocity (Um) and tidal direction for data sampled at 16 Hz at flood tide at S2b given in Table 3. The 16-Hz sampling rate exhibited higher variations in flow velocity and flow direction as turbulence of tidal flow, although a higher sampling rate is needed to compare these variations.

Fig 5 shows the resultant mean tidal velocity for 1 min = 60 s resultant velocity (Fig. 4(B)) estimated from the 3-D data of east, north and depth velocity sampled at 16 Hz at Stations S2 and 6. Over the measurement period, the mean resultant velocity changed with tidal range while the flood velocity was greater than the ebb velocity.

The variation in velocity was represented as a kinetic energy as the absolute mean value >1 min using the resultant velocity yielded by Eq. (2), and as a turbulence intensity as the relative rate yielded by Eq. (3). The kinetic energy and turbulence intensity in relation to the mean resultant velocity at Station S3 are shown in Fig. 6. The kinetic energy Ek or turbulence intensity (T_r) can be expressed as a function of the resultant velocity by the following equations:

The turbulence intensity when the mean velocity was >0.2 m/s ranged between 10 and 50%, which was higher than the values of 8–15% recorded 4.7 m above the seabed at 22-m water depth in Puget Sound, USA and 10% recorded at a site off Seattle, USA (Thomson et al., 2012). The turbulence intensity affected the swimming speed of perch in tank experiments (Lupandin, 2005), and was the key variable affecting fish swimming in turbulent flow (Liao, 2007).

The intercept and slope along with correlation coefficients for Eqs. (4) and (5), respectively, are given in Table 3. Ek increased with increasing resultant velocity, while T_r decreased with increasing resultant velocity. Furthermore, the relationship between the resultant velocity and the kinetic energy and the resultant velocity and the turbulence intensity was significant (high correlation coefficient), with the exception of the turbulence intensity at S1, which exhibited a lower correlation coefficient. Thus, the main features of the velocity change in turbulence can be expressed accurately with the kinetic energy as absolute values in relation to the mean velocity rather than the relative values of turbulence intensity.

The turbulence kinetic energy at S3 was similar to the ~0.03 m²/s² recorded in the eastern English Channel at 20 m above the seabed at 60-m water depth at a maximum flow velocity of 1 m/s (Korotenko and Senchev, 2011). In one study, the turbulence kinetic energy was adopted as the main cue in the reaction of blue crab at the post-larval stage in tank experiments (Saiz, 1994; Welch et al., 1999), while in another study, the turbulence kinetic energy was not considered in the effects of turbulence on the movements of swim ming salmon (Enders et al., 2003).

The FFT periodicity spectrum of the resultant velocity and flow direction for all sampling data from S2b are given in Fig. 7. Among the several peaks of period of <30 s, one dominant peak is observed around 10 s as a shorter period caused by oscillation of fishing gear or movements of fish. The shorter periods of velocity and direction produced by FFT ranged between 4 and 15 s, as shown in Table 4. Complex changes in tidal flow velocity with variations in flow direction have been observed to generally occur in near-bottom tidal flow (Roget et al., 2010; Walter et al., 2011). Because a shorter flow period could have greater effects on fish swimming or escape behavior (Kim et al., 2008), in this study we considered shorter only periods of ~10 s.

Fig. 8 shows the estimated period of velocity and direction using the global wavelet spectrum (GWL) method (Torrence and Compo, 1998) for the eight scenarios of strong tide for periods <30 s given in Table 2. The dominant period of velocity and direction ranged between 4 and 17 s, while no peak period was present at S3a, S3b and S6a, as shown in Table 4.

The complex periodicity of tidal flow was examined using the continuous wavelet method with Morlet in MATLAB as shown in Fig. 9. The period plumes are clearly visible for shorter periods of 3–6 s, represented in blue, and the curves of the correlation coefficients are shown in the plots at the bottom of the figure. Although, we were unable to plot the entire analyzed period for each scenario, the shorter period of each flow velocity or direction appeared between 3 and 20 s, similar to the results yielded by the global wavelet method shown in Table 4. The periods produced by Morlet wavelet method for data recorded in Jiaozhou Bay, Qingdao, China at a water depth of 7 m and flow velocity of ~0.5 m/s with sampling at 16 Hz were estimated to be 4–64 s (Yuan et al., 2009). However, the periods produced by event detection method in the Eprapah Creek Estuary in eastern Australia at the seabed at a water depth of 2 m and flow velocity of ~0.25m/s with sampling at 50 Hz were estimated to be 0.25–1 s (Yuan et al., 2009). The results from these two studies are in agreement with the shorter periods of turbulent flow observed in our work. However, at a lower sampling rate of 2 Hz, the period at slack tide in Lunenburg Bay, Nova Scotia, Canada at a water depth of ~6 m and flow velocity of ~0.5 m/s ranged between 100 and 300 s (Rennie and Hay, 2008). This longer period was also observed in our study using the global wavelet and FFT methods.

Tables 5 and 6 show the mean period, and the amplitude of velocity and direction, respectively, produced by peak event analysis for the eight scenarios of strong tide in Table 2. The mean periods ranged between 4 and 8 s for velocity and direction with no significant difference between flood and ebb tides. The amplitude of the velocity variation ranged between 0.1 and 0.6 m/s and increased with increasing mean velocity. Similarly, the amplitude of the directional variation ranged between 20 and 90° and increased with increasing mean velocity. The period rate for peak or valley events ranged from at least 80 to 100% as covered time ratio of periodic events, while the pair rate of a consecutive pair of peak and valley events of period <30 s ranged between 30 and 70%. These rates are characteristic of the periodic ratios in highly variable data of turbulence flow similar to the coefficients yielded by the Morlet wavelet method. However, the peak event method can provide an estimate of the absolute range of amplitude for each period for comparison with other period analysis methods such as FFT or the wavelet method. The amplitude obtained by peak event analysis was considered as different terms in relation to the standard deviation, which was calculated for all data without periodic analysis (Bendat and Piersol, 2000).

The period frequencies of tidal velocity and tidal direction given by peak event analysis are shown in Fig. 10. When the minimum period was <1 s the peak periods ranged between 2 and 4 s for the tidal velocity, and between 1 and 3 s for the tidal direction.

The period of the swim speed variation of fish depends on body length and tail beat frequency and has been variously reported at between 0.1 and several seconds (Videler and Wardle, 1991). The dominant period of swimming acceleration in juvenile roundfish near the upper panel of the codend was 2~3 s at a towing speed of 1.5 m/s (Kim et al., 2008). The effect of turbulence on Atlantic salmon was investigated at a dominant flow period of 6 s in tank experiments.

The dominant period of turbulence flow on the codend of the shrimp beam trawl or bottom trawl was estimated at between 3 and 8 s when a shorter period of <60 s was selected (Kim, 2012; 2013a) and its value is possibly related to the period of swimming response in escaped fish from the codend (Kim et al., 2008). Therefore, period analysis of shorter periods of a few seconds is suitable for studying the movements of swimming fish or fishing gear dynamics.

The turbulent water flow inside the codend could be resultant turbulence mixed up by tidal flow, towing motion of a fishing boat, wake of fishing gear, etc. In addition, the main index of turbulence effects on animal movements or swimming fish movements should be considered as kinetic energy, and the dominant period of tidal flow as 3-D and flow direction for analyzing the stability control of fish (Kim and Gordon, 2010; Webb and Cotel, 2010: Tritico and Cotel, 2010). Therefore, in future, turbulent mixing inside the codend should be analyzed and interpreted as complex flow including the tidal turbulence flow in the overlying water.