Seasonality of floc strength in the southern North Sea

Authors


Abstract

The suspended particulate matter (SPM) concentration in the high turbidity zones of the southern North Sea is inversely correlated with chlorophyll (Chl) concentration. During winter, SPM concentration is high and Chl concentration is low and vice versa during summer. This seasonality has often been associated with the seasonal pattern in wind forcing. However, the decrease in SPM concentration corresponds well with the spring algal bloom. Does the decrease of SPM concentration caused by changing wind conditions cause the start of algae bloom, or does the algae bloom decrease SPM concentrations through enhanced flocculation and deposition? To answer the question, measurements from 2011 of particle size distribution (PSD), SPM, and Chl concentrations from the southern North Sea have been analyzed. The results indicate that the frequency of occurrence of macroflocs has a seasonal signal, while seasonality has little impact upon floc size. The data from a highly turbid coastal zone suggest that the maximum size of the macroflocs is controlled by turbulence and the available flocculation time during a tidal cycle, but the strength of the macroflocs is controlled by the availability of sticky organic substances associated with enhanced primary production during spring and summer. The results highlight the shift from mainly microflocs and flocculi in winter toward more muddy marine snow with larger amounts of macroflocs in spring and summer. The macroflocs will reduce the SPM concentrations in the turbidity maximum area as they settle faster. Consequently, the SPM concentration decreases and the light condition increases in the surface layer enhancing algae growth further.

1. Introduction

Seasonal variations are characteristic for biogeochemical processes on tide-dominated midlatitude continental shelves. They are primarily caused by the seasonality of solar forcing that drives physical (e.g., weather conditions, thermal stratification, light) and biological (e.g., primary production) processes. Suspended Particulate Matter (SPM) concentration in the North Sea has a typical seasonal variation with high values in winter and low values in summer [e.g., Howarth et al., 1993]. Very often the seasonal pattern in wind and waves, with more storms in winter than summer, is put forward to explain the seasonality [Eleveld et al., 2008; Dobrynin et al., 2010]. The spring and early summer phytoplankton blooms have been associated with the seasonality of SPM concentration through the formation of larger flocs, and increasing the settling of SPM toward the seafloor [Jago et al., 2007; Borsje et al., 2008; Van Beusekom et al., 2012] and reducing the erodibility of bed sediments [Black et al., 2002].

Flocculation is the combined process of particle size growth and decay through aggregation and breakage in a turbulent flow field, thereby determining the size and settling velocity of SPM. A low turbulent flow enhances particle aggregation and increases the size and settling velocity of particles, but a high turbulent flow enhances floc breakage and decreases the size and settling velocity [Winterwerp, 2002]. Flocculation also combines biomass and minerals particles together into larger aggregates with often multimodal floc size distributions and different floc strength [Verney et al., 2009; Lee et al., 2012]. Flocculation depends on the attractive forces acting between the suspended particles that are caused by the surface properties of the particles. These properties are controlled by cohesive forces of clay minerals and by the microbial products consisting of sticky gel-like particles, called TEP (Transparent Extracellular Polymer) that interact with mineral particles and alter the properties of the SPM [Passow, 2002]. In coastal zones, the SPM is composed mainly of mineral particles and these organic products are acting as an additional binding agent [Hamm, 2002; Fettweis et al., 2006; Maggi, 2009; Bainbridge et al., 2012].

Despite the improved understanding of flocculation dynamics and their interaction with turbulence and biomineralogical composition, our knowledge is still insufficient to describe the impact of high primary production in spring and summer on floc sizes that induce changes in settling, formation of high-concentration mud suspensions, and resuspension of fine-grained sediments. Does a decrease of SPM concentration caused by changing wind and wave conditions trigger the start of the phytoplankton bloom in a high turbidity zone, or does the bloom decrease SPM concentration through enhanced flocculation and deposition? Both processes have a seasonal signal (Figure 1). Since the Belgian nearshore area, located in the southern North Sea, is very turbid and characterized by intense algae blooms, it is a relevant site to investigate links between biomass, SPM concentration and seasonal hydrometeorological.

Figure 1.

Surface SPM and Chl concentration from MERIS for the period 2003–2011 (thick black line) and for 2011 (gray line). Below are the daily averaged wind velocity and significant wave heights for the period 2001–2012 (thick black line) and for 2011. All data are for the station MOW1 (see Figure 2).

1.1. Region of Interest

The Belgian nearshore area is situated in the southern North Sea and is characterized by high SPM concentrations (Figure 2). SPM concentration ranges between 20 and 100 mg/L at the surface and between 100 and more than 3000 mg/L near the bed; lower values (<100 mg/L) occur offshore [Fettweis et al., 2010; Baeye et al., 2011]. In situ measurements are available at MOW1 (51°21.63′N, 3°7.41′E) located in the turbidity maximum zone (water depth about 10 m Mean Lower Low Water Spring (MLLWS)). Holocene medium-consolidated mud characterizes the seabed at the MOW1 site, albeit covered with an ephemeral fluffy mud layer (fluid mud) or muddy fine sand layer with a median grain size of about 170 μm [Fettweis and Van den Eynde, 2003; Verfaillie et al., 2006]. The suspended matter forms flocs that are built of clay and silt-sized particles, CaCO3, and organic carbon with a median particle size of <2 μm [Fettweis, 2008]. The tidal regime is semidiurnal, and the mean tidal range near MOW1 is 4.3 and 2.8 m at spring and neap tide, respectively. The tidal current ellipses are elongated in the nearshore area and become gradually more semicircular toward the offshore. The current velocities at the measuring location MOW1 vary from 0.2 to 1.5 m/s during spring tide and 0.2 to 0.6 m/s during neap tide [Fettweis and Van den Eynde, 2003]. The strong tidal currents and the low freshwater discharge of the Scheldt (yearly average is 100 m3/s) result in a well-mixed water column with almost no salinity and temperature stratification throughout the water column [Lacroix et al., 2004]. South-westerly winds dominate the overall wind climate, followed by winds from the NE sector (Figure 3). Maximum wind speeds coincide with the south-westerly winds; nevertheless, the highest waves are generated under north-westerly winds. The phytoplankton bloom starts in early spring with a diatom bloom and shifts toward a phaeocystis bloom in April and May [Lancelot et al., 1987]. Diatoms and phaeocystis concentrations decrease during June due to a shortage in nutrients and an increase in predation pressure by heterotrophic plankton species [Rousseau et al., 2002].

Figure 2.

The 2011 (left) mean surface SPM and (right) Chl concentration during winter and summer in the Belgian coastal area (southern North Sea). Data are from MERIS satellite. The cross indicates the in situ measuring station MOW1.

Figure 3.

Wind rose diagrams showing (a) the 2011 winter wind data and (b) the 2011 summer data. Legend values are in m s−1.

2. Material and Methods

2.1. In Situ Measurements

Data were collected with a tripod to measure currents, salinity, temperature, turbulence, SPM concentration, and Particle Size Distribution (PSD). The instrumentation suite consisted of three D&A optical backscatter sensors (OBSs), a SonTek 5 MHz Acoustic Doppler Velocimeter (ADV) Ocean, and a Sequoia Scientific LISST (laser in situ scattering and transmissometry) 100 X type C. The ADV was used to measure the three velocity components at 25 Hz at 18 cm above the bed and to estimate the turbulent kinetic energy from the turbulent fluctuations (see below). The LISST-100C measures PSDs in 32 logarithmically spaced size groups over the range of 2.5–500 μm [Agrawal and Pottsmith, 2000]. The volume concentration of each size group is estimated with an empirical volume calibration constant, which is obtained under a presumed sphericity of particles. Uncertainties of the LISST-100C detectors may arise from various causes [Mikkelsen et al., 2007; Andrews et al., 2010; Davies et al., 2012, Graham et al., 2012]. In the study area, these uncertainties are most probably caused by nonspherical particles, particles exceeding the instrument size range, or a too high SPM concentration.

All data (except LISST) were stored in two SonTek Hydra data logging systems. The LISST was mounted at 2 m above the bed (hereafter referred to as mab) and the OBSs at 0.2, 1, and 2 mab. The OBS signal was used to estimate SPM concentration. OBS voltage readings were converted into SPM concentration by calibration against filtered water samples collected during four tidal cycles every year, see Fettweis [2008] for a description of the method. Data gaps in the PSD time series occurred due to biofouling (mainly summer), too low transmission, too short battery life time, or instrument failure. The OBSs used were formatted to measure concentration of up to 3 g/L. During high energy conditions, SPM concentration was regularly higher than 3 g/L. Under these circumstances the OBS will saturate and underestimate the actual SPM concentration.

The tripod was moored at the location between 3 and 6 weeks and was then recovered and replaced with a similar tripod system. Ten deployments were carried out between 15 December 2010 and 18 January 2012. The long deployment ensured accurate assessments of conditions over neap and spring tides, and included a variety of meteorological events. From these, 208 days of good LISST data in 2011 remained after quality check, with about 2/3 recorded during winter (January-March and October-December) and 1/3 during summer (April-September). Good quality data have an optical transmission between 15% and 98%; show no gradual or sudden decrease (increase) in transmission (volume concentration) during the measurements and have a smooth PSD. A gradual decrease is often the result of biofouling and occurs mainly in spring and summer. A sudden decrease in transmission is generally caused by a physical obstruction (e.g., cord entangled in optical path). A misaligned laser beam may cause high peaks in a few size classes making the PSD not smooth; these peaks remain during the whole measurements. The LISST 100 is a delicate instrument, misalignment of the laser beam may occur during deployment or other physical disturbances (collision with fishing gear).

2.2. Remote Sensing Measurements

The satellite-based imagery selected for this study was provided by the Medium Resolution Imaging Spectrometer (MERIS, https://earth.esa.int/web/guest/missions/esa-operational-eo-missions/envisat/instruments/meris). This multispectral sensor was on board of ENVISAT, a polar orbiting satellite which was launched in 2002 and provided data until April 2012. MERIS images were available with a daily temporal frequency and spatial resolution of 2 × 2 km2 and provided water leaving reflectance information for 15 bands across 390–1040 nm. Oceanographic parameters related to ocean color, such as the chlorophyll-a (Chl) and SPM concentration were derived from the water leaving reflectance in specific spectral bands. Chl concentration was estimated using the MERIS case 2 algorithm (version MEGS 7.5) as described by Doerffer and Schiller [2006]. A quality control has been applied according to the standard MERIS product confidence flags. Remotely sensed SPM concentration is estimated from water leaving reflectance at 667 nm using the generic multisensor algorithm of Nechad et al. [2010]. In case of quality issues due to atmospheric correction error, stray light or sun-glint, the pixel was masked as unreliable and rejected. Surface Chl and SPM concentration time series were extracted from the imagery data using a 5 × 5 kernel for the location of the MOW1 station for 2011 (Figure 1). The Chl and SPM concentration climatology for the period 2003–2011 was generated by linearly interpolating the available data at a yearly basis after which a interannual mean was calculated per day. Additionally, for both Chl and SPM concentration, multitemporal composite maps were generated for the winter (January-March, October-December) and summer (April-September) season of 2011 (Figure 2).

Satellites cover large-scale scenes, but at a low time resolution, limited to surface data and with gaps in data often occurring during stormy weather conditions, though missing the high ranges of SPM concentrations. Satellites can be seen as random samplers biased toward good weather conditions as they represent only the cloud-free data, but also to nonsatellite-saturating data which occur at high SPM concentration levels. Fettweis and Nechad [2011] have shown that 60 satellite images per year are representative of the mean SPM concentration during good weather. In 2011, 67 good satellite images are available at MOW1 for SPM concentration (46% in winter, 54% in summer) and 37 for Chl concentration (38% in winter, 62% in summer).

2.3. Kolmogorov Scale of Turbulence From ADV

Turbulence in coastal areas controls the flocculation of fine-grained material and impacts the vertical and horizontal flux of SPM. The length scale of the smallest dissipating eddies (Kolmogorov scale of turbulence, λk) generally limits the size of the flocs [van Leussen, 1999; Fettweis et al., 2006; Cross et al., 2013]. Assuming that turbulent kinetic energy production is equal to dissipation, this scale can be calculated as λk = (ν3/ε)¼, where ν is the kinematic viscosity (10−6 m2 s−1) and ε is the turbulent energy dissipation (m2 s3). The turbulence dissipation can be derived from τ = ρ (ε κ z)2/3, where τ is the shear stress, z the elevation above the bed, κ the von Karman constant, and ρ the water density. The turbulent kinetic energy (TKE) and the shear stress can be calculated using the variance of velocity fluctuation from the high-frequent ADV measurements [Stapleton and Huntley, 1995; Thompson et al., 2003]. MOW1 is situated in shallow waters where wave effects are important; therefore, the shear stress was corrected for the advection by waves following the approach of Trowbridge and Elgar [2001], Sherwood et al. [2006], and Fettweis et al. [2010]. With the turbulence dissipation known, the Kolmogorov length scale can be calculated. The length scale was low-pass filtered using the PL64 filter described in Flagg et al. [1976] with a 33 h half-amplitude cutoff to remove tidal and higher-frequency signals.

2.4. Statistical Methods to Analyze PSD

A curve-fitting technique and a statistical method (entropy analysis) were used to analyze the large PSD data set in order to identify seasonality in an objective way. The curve-fitting software (DistFit™, Chimera Technologies Inc., USA) was used to decompose the multimodal PSD into four subordinate lognormal PSDs and to quantify the geometric mean diameter, standard deviation, and volume fraction of them. The lognormality describes a more or less skewed distribution toward small size and the multimodality describes a distribution consisting of multiple modal peaks. The multimodal lognormal distribution function can be written as an integrated distribution function of four lognormal distribution functions [Whitby, 1978; Jonasz and Fournier, 1996]:

display math

where D is the particle diameter, W the volume concentration, inline image the geometric mean diameter, σi the multiplicative standard deviation, and inline image the volumetric fraction of an ith unimodal PSD. The choice of four lognormal functions is based on the fact that flocculation of fine-grained material in coastal areas develops a four-level structure consisting of primary particles, flocculi, microflocs, and macroflocs [van Leussen, 1999; Lee et al., 2012], see Figure 4. The 32 size classes measured by the LISST can thus be reduced to four groups that represent physical concepts rather than numbers. Primary particles consist of various organic and mineral particles (clay and other minerals, calcareous particles, picophytoplankton, bacteria). Flocculi are breakage-resistant aggregates of mainly clay minerals. Microflocs are the medium size aggregates and macroflocs are the very large aggregates that can reach hundreds to thousands of micrometers in organic-rich low-energy conditions, but only up to a few hundred micrometers in organic-limited high energy conditions [Fettweis et al., 2006]. The DistFit software (Chimera Technologies) was applied to the PSDs averaged over 10 min to generate the bestfits, defined as the minimum errors between fitted and measured PSDs [Whitby, 1978]. For two modal peaks, fixed sizes of 3 μm (lowest size class of the LISST) and 15 μm were chosen; the modal peaks of the bigger fractions were variable and chosen in order to represent the larger size classes of the LISST instrument (15–200 and 150–500 μm). The standard deviations varied between 1 and 2.5. The choice of parameters is based on assumptions and experiences [Lee et al., 2012].

Figure 4.

A multimodal PSD and its subordinate lognormal PSDs of primary particles, flocculi, microflocs, and macroflocs (adapted after Lee et al. [2012]).

A rising tail in the lowest size classes of the LISST is frequently observed in data and is caused by the presence of particles up to 10 times smaller than the smallest size bin of the instrument (i.e., 0.25–2.5 μm). Andrews et al. [2010] reported that fine out of range particles affect the entire PSD, with a significant increase in the volume concentration of the first two size classes of the LISST, a decrease in the next size classes and, surprisingly, an increase in the largest size classes. Similar remarks have been formulated by Graham et al. [2012], who observed an overestimation of 1 or 2 orders of magnitude in the number of fine particles measured by the LISST. They argue that due to flocculation small particles (such as individual grains, phytoplankton cells, bacteria, and viruses) probably do not exist as isolated individuals in large numbers in coastal waters. The 3 μm mode, therefore, most likely is an overestimation of fine particles due to inaccuracy of the LISST instrument when particles become too small for its range. The volume fraction calculated for this size distribution is interpreted as an indication of the presence of very fine particles rather than providing a correct number. This bias may further enhance the separation between the two peaks of primary particles and flocculi and develop a small peak of macroflocs during the peak flows. In case of low turbulent condition, it may also indicate the presence of very large particles [Andrews et al., 2010]. Particles exceeding the LISST size range of 500 μm also contaminate the PSD. Davies et al. [2012] reported that large out of range particles increase the volume concentration of particles in multiple size classes in the range between 250 and 400 μm and in the smaller size classes and recommended to interpret the PSD with care in case particles outside the size range may potentially occur. The importance of these spurious results depends on the number of large particles in the distribution [Davies et al., 2012]. No particle size data obtained with other methods (video system, holography) are available at the MOW1 site. Macrofloc sizes recorded by a video system at an estuarine sites with similar tidal dynamics were generally smaller than 580 μm [Winterwerp et al., 2006], which indicates that most of the larger flocs are not exceeding the size limit of the LISST.

The structure of the PSD time series is further investigated using the concepts of entropy and complexity. Entropy analysis has been successfully applied to time series of LISST particle size distributions of suspended matter [Mikkelsen et al., 2007; Fettweis et al., 2012]. Entropy-based algorithms quantify the regularity of a time series. Entropy increases with the degree of disorder and is maximal for completely random systems. Applied to PSDs, entropy analysis allows grouping the size spectra without assumptions about the shape of the spectra. It is therefore suited to analyze unimodal and bimodal as well as multimodal distributions. The PSD time series has first been low-passed filtered using a filter of 33 h to remove the tidal signal, before the entropy classification with four PSD groups was carried out using the FORTRAN routine of Johnston and Semple [1983]. The entropy group PSD time series were then evaluated to assess the effects of the neap-spring tidal signal, meteorological effects, and seasons.

3. Results

The time-varying PSD, SPM concentration, Chl concentration, wind, wave, and hydrodynamics parameters constitutes a scientific record of flocculation and transport of the constituent particles and aggregates throughout 1 year that allows understanding of the possible causes of change in PSD and SPM concentration. The temporal distribution of the four constituents of the PSDs (primary particles, flocculi, microflocs, and macroflocs) are shown in Figure 5 with the entropy analysis of the low-passed filtered PSD data, the low-passed filtered SPM volume and mass concentration, the actual and the low-passed filtered Kolmogorov length scale, the significant wave height, tidal elevation, and the wind direction and strength. The curve-fitting analysis of the PSD has been low-pass filtered in the figure for clearness. The linear correlation between the low-pass filtered Kolmogorov scale and the logarithm of the significant wave height is significant (R2 = −0.77, N = 36499). The low-pass filtered Kolmogorov length scale can thus be used as a proxy of the nontidal (waves and wind) turbulence intensity in shallow waters. The data are divided in three groups according to the turbulent intensity as follows: λk < 0.28 mm, 0.28 mm < λk < 0.58 mm, and λk > 0.58 mm. These groups correspond roughly with periods where the significant high wave heights are greater than 1.50 m, between 0.75 and 1.50 m, and lower than 0.75 m. The λk = 0.58 and 0.28 mm are the 15th and 85th percentiles of the low-pass filtered Kolmogorov-scale data, respectively.

Figure 5.

The 2011 time series (in days of the year) of (a) wind velocity; (b) wind direction; (c) significant wave height; (d) tidal elevation; (e) Kolmogorov scale (gray) and low-pass filtered signal (black); (f) low-pass filtered median floc size and entropy groups (group 1: blue, group 2: green, group 3: red, group 4: yellow, see Figure 6); (g) frequency of the four type of SPM constituents obtained by curve fitting: primary particles (dark blue), flocculi (light blue), microflocs (yellow), and macroflocs (brown); (h) SPM volume concentration at 2 mab; and (i) SPM mass concentration at 0.2 mab (gray) and low-pass filtered signal (black).

The seasonality of SPM and Chl concentration is obvious from satellite images and in situ data at MOW1 (Figures 1 and 5 and Table 1). The geometric mean near-bed SPM concentrations are 28% higher during calm weather in winter than in summer. With increasing wave influence, these differences decrease to 17% (0.28 mm < λk < 0.58 mm) and 7% (λk < 0.28 mm) at 0.2 mab. Higher in the water column (1 mab) similar trends are observed, except that in summer the geometric mean SPM concentration during higher waves (λk < 0.28 mm, 119 mg/L) is almost the same as during lower wave conditions (0.28 mm < λk < 0.58 mm, 118 mg/L). The latter is probably caused by the limited amount of data fitting this criterion in summer. Surface SPM concentrations from satellite images are about 100% higher in winter than summer. The seasonal variation in Chl concentration is shown in Figure 1b, the mean Chl concentration during summer 2011 was 11 μg/L (maximum is 29 μg/L) and during winter 3 μg/L (maximum is 14 μg/L).

Table 1. Geometric Mean and Standard Deviation of SPM Mass Concentration (mg/L) From the OBS at 0.2 and 1 m Above Bed (mab) and the SPM Volume Concentration (µL/L) From the LISST at 2 mab (mg/L and µL/L)a
 0.2 mab (mg/L)1 mab (mg/L)2 mab (µL/L)
  1. a

    The data are calculated according to season and low-pass filtered Kolmogorov scale (λk); λk < 0.3 and λk > 0.6 mm are the 15th and 85th percentiles, corresponding to periods with significant high wave heights greater than about 1.50 m and lower than about 0.75 m.

Winter 2011
all431 */2.1145 */2.6674 */1.9
λk > 0.6 mm420 */1.8136 */1.8545 */2.0
0.3 < λk < 0.6 mm448 */2.0146 */2.5718 */1.9
λk < 0.3 mm484 */2.0205 */2.3664 */1.9
Summer 2011
all357 */2.5117 */2.2449 */2.2
λk > 0.6 mm304 */2.6110 */2.2338 */2.2
0.3 < λk < 0.6 mm374 */2.4118 */2.2495 */2.2
λk < 0.3 mm449 */2.2119 */1.8423 */2.0

The PSD of the four entropy groups can be found in Figure 6 and Table 2. The groups are ordered in increasing size of the median particle size. The entropy analysis was based on the low-passed filtered data and the tidal variability of the PSDs was thus removed prior to analysis. The resulting classification into the four groups reflects the neap-spring and meteorological variations and not the tidal variations. In Figures 7 and 8 and Table 1, seasonality is investigated by grouping the data into a summer (April to September) and winter (October to March) period. These months have been chosen based on Figure 1 and correspond with a biological (Chl concentration) and physical (Wind, waves) influenced season. In winter, groups 1 and 2 occur during spring tides and storm periods. The differences between groups 1 and 2 are not very large, but are significant. A shift from group 2 in January toward group 1 in February and March is visible in Figure 5. This difference is correlated with changes in wind direction from S-SW toward an eastern direction (SE-E-NE). Group 3 is typically associated with neap tides and low waves. Nevertheless in May, it was dominant during spring tide conditions, indicating that the turbulence was not strong enough to break the flocs into smaller constituents and as a result microfloc and macrofloc were more abundant. Group 4 occurs mainly during summer and is characterized by a shift of the PSD toward larger size classes. Macroflocs are most abundant in this group, and represent about 40% of the total volume concentration. The frequency of the four groups is shown in Figure 7 as a function of season and turbulence intensity (low-pass filtered Kolmogorov scale). During storm periods (λk < 0.28 mm, Hs > 1.50 m), groups 1 and 2 are occurring more frequently, whereas the frequency of occurrence of group 3 is lower and of group 4 absent. Seasonal variations in the distribution of the groups can be distinguished. Group 2 is most frequent during winter (57%) followed by groups 1 (24%), 3 (19%), and 4 (1%). Group 3 is associated with neap and group 2 with spring tidal conditions during calm weather. During summer, group 3 emerges as most frequent (34%) followed by groups 2 (30%), 1 (21%), and 4 (15%). During this season, group 4 is typically associated with neap and group 3 with spring tidal conditions during calm weather.

Table 2. Frequency (%) of Primary Particles (PP), Flocculi, Microflocs, and Macroflocs of the Four Entropy Groups (Figure 6)a
 PPFlocculiMicroflocsMacroflocs
  1. a

    The geometric mean floc size is shown between brackets for the microfloc and macrofloc. Primary particles and flocculi have constant size of 3 and 15 µm, respectively.

Group 15.629.554.9 (52 µm)10.1 (250 µm)
Group 210.713.954.1 (52 µm)21.4 (250 µm)
Group 32.314.166.0 (66 µm)17.4 (249 µm)
Group 41.711.747.7 (70 µm)38.8 (249 µm)
Figure 6.

The PSD of the four entropy groups of the low-pass filtered data. Also shown are the subordinate lognormal PSDs of primary particles, flocculi, microflocs, and macroflocs obtained by curve fitting (see also Table 2).

Figure 7.

Frequency (%) of the four entropy groups for the summer and winter season and according to the low-pass filtered Kolmogorov scale (λk); λk < 0.28 and λk > 0.58 mm are the 15th and 85th percentiles, corresponding to periods with significant high wave heights greater than about 1.50 m and lower than about 0.75 m.

Figure 8.

Frequency of primary particles, flocculi, microflocs, and macroflocs for the summer and winter season and according to Kolmogorov scale (λk); λk < 0.25 and λk > 0.65 mm are the 15th and 85th percentiles. The geometric mean size of the microflocs is 66 ± 24 µm (winter) and 73 ± 21 µm (summer), and for the macroflocs 222 ± 44 µm (winter) and 220 ± 44 µm (summer). Primary particles and flocculi have constant size of 3 and 15 µm, respectively.

In contrast with the entropy grouping (Figures 5f and 7), the curve-fitting grouping (Figures 5g and 8) is based on physical and not mathematical characteristics of the PSD time series. The results indicate that microflocs are most abundant in terms of volume concentration, followed by flocculi, macroflocs, and primary particles. Yearly average percentages of these four constituents are 3% (primary particles), 20% flocculi, 66% microflocs, and 11% macroflocs (Figure 8). The data are divided in three groups in a similar way as the entropy groups, but now using the (not low-pass filtered) Kolmogorov-scale data. The 15th and the 85th percentile are λk = 0.65 mm and λk = 0.25 mm, respectively. Primary particles and flocculi are abundant during breakup periods as well as microflocs. During periods with lower turbulence, i.e., large Kolmogorov-scale numbers, macroflocs are more abundant. The geometric mean size of the microflocs is 69 ± 23 μm (all year) with slightly lower values during winter (66 ± 24 μm) and slightly higher ones during summer (73 ± 21 μm). The geometric mean size of the macroflocs is 221 ± 44 μm (all year) with almost no variation between seasons (winter: 222 ± 44 μm, summer: 220 ± 44 μm). In contrast with microflocs that do not show a strong seasonal signal, macroflocs are found to be more frequent in summer than in winter (14.2% versus 8.9%, see Figure 8). The frequency of macroflocs in the SPM during periods with high turbulence (λk < 0.25 mm), is on average 9.5% during summer and 6.7% during winter. During calm periods, macroflocs are almost 2 times more abundant in summer than during similar periods in winter (20.2% versus 11.0%). The higher/lower frequency of macroflocs in summer/winter is compensated by lower/higher frequencies of primary particles and flocculi.

4. Discussion

SPM dynamics are controlled by flocculation, which influences the size and deposition rate of the SPM [Winterwerp, 2002]. Flocculation depends on the turbulent intensity (tides, wind, waves) and on the surface properties of the suspended particles, which are of electrochemical or microbial origin [Mietta et al., 2009]. Microbial products, such as TEPs, are released by algae and bacteria and influence aggregation [Logan et al., 1995; Engel, 2000]. Chl concentration, wind velocity, and wave height all have a seasonal signal. The seasonality of the Chl concentration signal is, however, more pronounced (Figure 1). Below, we will discuss to what extent the seasonality of floc size is controlled by these physical and biological effects.

4.1. Physical Controls

The hydrodynamics (i.e., flow velocity and turbulent intensity) vary with meteorological conditions, neap-spring and tidal periods. As tidal forcing is approximately equal during both seasons, we will focus on meteorological conditions. The wind climate in the study area is characterized by mainly SW and NE winds (Figure 3), which affect the direction and strength of alongshore water mass transport. During summer, the main wind sectors are WSW and NNE and during winter S to SW and NE. These small shifts in the main wind sectors do not affect the direction of the residual alongshore transport [Baeye et al., 2011]. A modification of the residual transport influences the position of the coastal turbidity maximum and thus the SPM concentration at the measuring site [Baeye et al., 2011]. The influence of this alongshore advection is reflected in the shift from entropy groups 2 and 3 toward groups 1 and 2 during the first 3 months of the year (Figure 5). This shift is caused by changes in wind direction from S-SW toward more easterly directions (SE toward NE) and also reflects a change in source of the SPM. During SW winds the suspended matter is transported toward the NE and the PSD correspond with group 2 (spring tide) and group 3 (neap tide), whereas during E wind direction the SPM is advected out of the Scheldt estuary and transported toward the SW. During these conditions, spring tides and storms have PSD according to group 1 and neap tides according to group 2 or 3. The frequency of both dominant wind directions does not vary significantly between seasons and therefore, they cannot explain seasonal variations in SPM concentration (Figure 3).

The data indicate that the differences in wind direction and strength between the seasons are small. The wind speeds smaller than 8 m/s are more frequent in summer (55% versus 46%), whereas wind speeds between 8 and 16 m/s are more frequent during winter (49% versus 42%). Storms (>16 m/s) are more frequent (5%) in winter than summer (3%). Similar results have been obtained for the waves. The mean significant wave height (Hs) during winter 2011 was 0.55 m and during summer 0.50 m. The mean Hs during the LISST measurements at MOW1 were 0.56 m during summer and 063 m during winter. The mean during the measurements was thus slightly higher than the mean over the whole year. Significant wave heights greater than 1.5 m during the LISST measurements were less frequently in summer (2% versus 7%), whereas periods with low wave heights (<0.75 m) were less frequently in winter (60% versus 65%). A seasonal signal exists for wave heights greater than 1.5 m for 2011. The 75% of the higher wave heights occurred during winter and the mean of these waves was higher during winter (1.89 m) than during summer (1.60 m). The summer of 2011 was, compared with the mean occurrence of significant wave heights greater 1.5 m during the period 2001–2012, less stormy. Periods with significant wave heights greater than 1.5 m occurred during 4.1 days versus 7.8 days in the summers of 2001–2012. The frequency of significant wave heights greater than 1.5 m during the winter of 2011 was nearly equal to the mean value (12.5 versus 12.3 days).

Floc size and SPM concentration have been evaluated as a function of sea state characterized by the low-pass filtered Kolmogorov length scale. The geometric mean SPM concentration at MOW1 increased from 357 mg/L (117 mg/L) during summer toward 431 mg/L (145 mg/L) during winter at 0.2 mab (1 mab). The influence of waves is significant (Table 1): the geometric mean SPM concentration was 64 mg/L (145 mg/L) lower during calm conditions than during stormy periods in winter (summer). The volume concentration of the SPM had a slightly different response than the mass SPM concentration. The highest volume concentrations were occurring during intermediate wave conditions. During high wave conditions (λk < 0.3 mm) we observed a decrease of the volume concentration (Table 1). This reflects the fact that the volume concentration depends on effective density and floc size. In case of high waves, large portions of the macroflocs with lower density were broken up into smaller particles with higher densities, resulting thus in a decrease of the volume concentration.

The effects of the storm extend a certain period after the storm. The duration of storm influence depends on wind direction, wind strength and wave height and can last up to a few days after the storm. The influence period is longer when waves are higher as more sediments have been resuspended or fluidized. Influence of storms is mainly detected in the near-bed layer and decreases toward the surface. Storms with wave heights of more than 2 m affect the SPM concentration for a period of about 5 days after the storm [Fettweis et al., 2010]. Storms with significant wave heights above 2 m occurred once during summer and nine times during winter. The total duration of these high wave events was 0.1 days (summer) and 4.3 days (winter). A 2.8 days of winter storms occurred during LISST measurements. Higher SPM concentration due to these meteorological conditions influenced the signal over a period of about 14 days. This represents 10% of the measurements in winter. It does not, however, explain the 20% higher SPM mass concentration and the 50% higher SPM volume concentration near the bed or the 100% higher SPM mass concentration in the surface during winter.

OBSs have primarily been designed to be most sensitive to SPM mass concentration; size effects are an order of magnitude lower than those of concentration, and flocculation effects are even smaller [Downing, 2006]. OBSs are most sensitive to fine-grained sediments. When SPM composition changes from very fine material (clay and silt particles or flocs) to silt-sized or sand-sized grains without changes in concentration, the optical backscatter signal will decrease, resulting in an apparent decrease in SPM mass concentration, without affecting SPM volume concentration. This apparent decrease has been observed at a nearby site during NE storm conditions when sand grains were resuspended [Fettweis et al., 2012]. However, as NE storms are not frequent these inaccurate SPM concentrations will only have a slight effect on the mean values.

4.2. Biological Controls

Flocculation in coastal waters depends on the attractive forces acting between the suspended particles. These forces depend on the surface properties of the particles, which are of physochemical and microbial origin. TEP was first described by Alldredge et al. [1993] and consists of mostly polysaccharides that are negatively charged, very sticky, and frequently colonized by bacteria. They interact with mineral particles and alter the properties of the SPM [Passow et al., 2001]. In general, TEPs are found to be in the same size and abundance range as phytoplankton [Mari and Burd, 1998]. The production of TEPs in the ocean has been connected with algae blooms and bacteria in that the formation of large aggregates following blooms was primarily controlled by TEP concentrations [Logan et al., 1995; Mari and Burd, 1998; Passow et al., 2001]. In shallow turbid systems, the phytoplankton balance is strongly affected by SPM dynamics. Phytoplankton growth in such environments depends on the light adsorption coefficient of the water, which varies according to the tidal and neap-spring variation of SPM concentration [Desmit et al., 2005]. Our data show that SPM concentration is highest during peak velocity during spring tide and lowest during slack water. Desmit et al. [2005] have shown that these short-term, tidally driven SPM concentration variations allow the incident sunlight energy to sustain phytoplankton production in these environments during spring and summer. The increase of Chl concentration during spring (Figure 1) indicates the start of the algae bloom and thus the start of the “biologically active” season. The surface Chl concentration drops after the spring bloom, and is followed by a summer bloom between day 150 and 190. Although the summer bloom was less pronounced as shown by the Chl concentration climatology (Figure 1b), it resulted in sufficiently Chl concentration levels that when combined with the TEP production by heterotrophic bacteria and macrobenthos was able to maintain the higher frequency of macroflocs found in the measurements.

Based on the decomposition of the measured PSD into four subordinate lognormal functions it was found, despite the seasonal signal in median floc size (D50-summer: 64 μm, D50-winter: 51 μm), that the sizes of macroflocs only show small variations during seasons (summer: 220 μm, winter: 222 μm). The frequency of macroflocs, however, has a seasonal signal. Macroflocs are more abundant in the SPM in summer than winter regardless of the turbulence intensity (Figure 8). Similar results are found in low-passed filtered entropy data. During neap tides and low wave activity entropy group 4 is dominant in summer and group 3 in winter. Group 4 is characterized by 39% of macroflocs and 48% of microflocs, whereas group 3 consists of 17% macroflocs and 66% microflocs. This observation is somewhat in contrast to other studies, arguing that for a given turbulence level flocs become larger with abundant organic matters during summer [Lunau et al., 2006; Cross et al., 2013]. The rate of break up of large flocs and the equilibrium size of flocs in turbulent flow depend on their strength [Kranenburg, 1999; Winterwerp, 2002]. Our observations of PSDs from a highly turbid coastal zone suggest that the maximum size is mainly controlled by the intensity of turbulence (tidal signal and waves) and the flocculation time. The tidal current ellipses are elongated at the measuring site and time available for floc formation is limited to the short periods of slack water (current velocity below 0.2 m/s at 1.8 m above the bed last on average 45 min at the measuring location), which are not sufficient for the flocs to attain their equilibrium size. Mineralogical composition of the SPM shows only minor changes throughout the year [Zeelmaekers, 2011]. The data suggest that the TEP formed during spring and summer increases the strength of the macroflocs rather than their size. If the abundance of macroflocs as a function of turbulence intensity is a proxy of floc strength then Figures 8 shows that flocs in summer are stronger than in winter. In our data, we can see that group 3, which is typically correlated with neap tides during winter, was dominant in May 2011 during a spring tide (around day 120 in Figure 5). Turbulence was then apparently not strong enough to break up the flocs and to shift the PSD toward group 2. Similar results were observed by Lee et al. [2012] who found that aggregates were armored against breakage during the algae bloom in April 2008. The stronger flocs resist shear-induced breakup and—if we assume that the macroflocs behave similar as in other coastal and estuarine environments [Winterwerp et al., 2006]—the higher proportion of large flocs results in a higher settling rate during summer and thus a lower SPM concentration. Only during storms (wave height > 2 m) in summer did a significant break up of the larger flocs into smaller particles (see around day 150 and 200 in Figure 5) occur.

The higher frequency of macroflocs in summer is compensated by lower frequencies of primary particles and flocculi. The size and frequency of the microfloc population in the PSDs, which is the major part of the SPM, has almost no seasonal signal (Figure 8). The predominance of these intermediate floc sizes can possibly be explained by a low TEP concentration relative to the mineral concentration. The sticky organic matter will decay and be quickly saturated by the mineral particles and will only help a limited part of the SPM to develop into breakage-resistant macroflocs.

Our results suggest that floc size controls settling and deposition and thus sediment dynamics. TEPs and other biostabilizators reduce erosion and resuspension of mud deposits [Droppo et al., 2001; Black et al., 2002; Gerbersdorf et al., 2008; Maerz and Wirtz, 2009]. Therefore, we presume that during summer a larger part of the cohesive sediments is kept in a high-concentration mud suspension (HCMS), fluid mud, or consolidated bed layer. The presence of HCMS or fluid mud results in a reduction of the bottom shear stress and thus a decrease of erosion [Geyer et al., 1996]. In winter, the strength of the deposits decreases, due to lower TEP concentrations flocs getting less strong and therefore are more easily resuspended, resulting in higher SPM concentrations. Evidence of HCMS formation during an algae bloom period was observed at a nearby site in April 2008 in contrast with a winter period [Fettweis et al., 2012]. The much greater decrease in surface SPM concentration compared with near-bed SPM concentration during summer suggests faster settling in the summer.

5. Conclusions

The annual cycle of SPM concentration in the high turbidity area off the Belgian coast is mainly caused by the seasonal biological cycle, rather than wind and waves. Wind strengths and wave heights have a seasonal signal, but these are not sufficient to explain the large differences observed in SPM concentration. The data are in line with the literature that emphasizes the stabilizing effect of biomass on bed erosion and floc strength. In the tidal-dominated southern North Sea, biomass effects increase the strength of macroflocs rather than their size, as was reported from other sites. The results highlight the transformation of mainly microflocs and flocculi in winter toward more muddy marine snow with larger amounts of macroflocs in spring and summer. The larger fraction of macroflocs reduces the SPM concentrations in the turbidity maximum area as they settle faster. The fact that macroflocs are more abundant and that SPM concentration decreases will increase light condition in the surface layer and enhance algae growth. Whence, it is mainly the biological activity in spring and summer that lead to a decrease in SPM concentration in the study area rather than the seasonal pattern in wind conditions. The data are, however, not able to explain the initiation of this transition as PSD and in situ Chl and TEP concentration are not available for the 2011 spring algal bloom period. Nevertheless, our data clearly show the importance of microbial activity on the cohesive sediment dynamics. The proposed mechanisms may differ in other marine habits. To date, very little comparative data exist with which to contrast our results from a eutrophied high turbidity area, with areas that are less affected by excess supply of nutrients.

Acknowledgment

The study was supported by the Maritime Access Division of the Ministry of the Flemish Community (MOMO project). Ship Time RV Belgica was provided by BELSPO and RBINS–Operational Directorate Natural Environment. The wave and wind data are from the Ministry of the Flemish Community (IVA MDK—afdeling Kust—Meetnet Vlaamse Banken). We thank L. Naudts, J. Backers, W. Vanhaverbeke, and K. Hindryckx for all technical aspects of instrumentation and moorings; F. Francken for data processing and archiving; and K. Ruddick, X. Desmit, G. Lacroix, and Q. Vanhellemont for useful discussions. We would also like to acknowledge the two reviewers for useful comments on the first version of the manuscript.