An 11 hour boat survey (6:00 - 17:00 GMT) was performed on the 17th of September 2014 in the Rhine Region of Freshwater Influence (Rhine-ROFI) about 10 km downstream of the mouth of the Rotterdam Waterway during calm weather conditions. This field survey was part of a large field campaign that was conducted in September-October 2014 in the mid-field region of the Rhine ROFI (Flores et al., 2017), close to the Sand Engine (Stive et al., 2013). A mooring and two frames were deployed at 12 m depth as part of the Stratification Impact on Near-shore Sediment Transport (STRAINS II) experiment (Flores et al., 2017; Horner-Devine et al., 2017; Rijnsburger et al., 2018). In Figure 1 the position of the survey vessel is indicated by the capital letter S and cross. The mooring and two frames location is indicated by the capital letter M and the dot.

The tidal elevation at Hoek van Holland was between 0.85 m and -0.62 m, and the wind speed stayed below 6 m/s coming from the east. In addition, the significant wave height was below 0.5 m, which was retrieved from a wave rider buoy 1 km south from the survey site. The total river discharge from the Rhine and Meuse River was about 1500 m ³/s, for that day. The boat survey started at flood tide, before the flood peak. The water column was weakly stratified with a salinity difference of 2 PSU (Practical Salinity Unit) between surface and bottom. The water column was divided into different salinity layers after a front arrival 2 hours later. The stratification prevailed almost the whole day as it was a neap tide day with low energetic mixing conditions. Note that the salinity close to the bottom remained close to 30 PSU.

The one day field-survey was used to study SPM in the near bed part of the water column. Several in-situ instruments, based on optical techniques, were used simultaneously during this day. Two of these instruments (LISST 100X and LISST-HOLO) were on a frame in the vicinity of the survey vessel whereas the underwater camera was operated from the vessel and the LabSFLOC2 camera set-up was used on board. The samples studied with the LabSFLOC2 camera were collected using a Van Dorn sampler and analysed as soon as the sampler was retrieved.

LISST 100X (Laser In-Situ Scattering and Transmissometry) uses laser diffraction to determine the volume concentrations of particles in 32 logarithmically spaced effective spherical diameter classes. The obtained Particle Size Distribution (PSD) ranges for the 100X type from 2.5 – 500 μm (Traykovski et al., 1999; Agrawal, 2001). The LISST 100X was placed on the bottom frame, 1.65 m above the bed (mab) and measurements were taken every 10 seconds. The data was averaged over 10 minutes bursts.

LISST-HOLO (Submersible Digital Holographic Camera) is an in-situ digital holographic technology that uses solid state diode laser at 658 nm and 4.4 μ m pixel size digital camera (1600 × 1200 pixels). It measures equivalent spherical diameters in the size range of 25-2500 μ m, within 50 log spaced size classes. The LISST- HOLO was placed at 1.40 mab on one of the two frames and took a picture every 5 minutes. The LISST-HOLO was tuned to get the most appropriate values for clean stack and threshold parameters (Davies et al., 2011). Clean Stack 3% and default threshold of 0.15 were found to give the best pictures and were used to process the raw data (LISST HOLO manual). The particle density can be estimated from the solidity calculated by the LISST-HOLO software. The solidity is defined as the solid fraction of a given particle, and can be related to the porosity of that particle (Choi et al., 2018). A fully solid particle will have a solidity of 1, whereas a solidity close to 0 will indicate a very porous particle. The solidity is obtained by dividing the black area of a particle (A _B =N×r ² where r ² is the area of a black pixel and N the number of black pixels within a particle) by the total area (A) of that particle. We assume that the volume of the particle is given by V=A×c , where c is the dimension of the particle perpendicular to the field of view of the camera, and V _B =A _B ×c is the volume of the black area. Assuming that the particles properties do not change along the direction indicated by c, we can express solidity as function of the volume of the particle by:

The density ρ _p of a particle is given by:

where ρ _s is the absolute density of the solid part and ρ _w is the absolute density of water (that fills the voids). It is assumed that no air is trapped in the flocs. From this we can deduce that the effective density of a floc is given by:

The density of the solid fraction ρ _s (represented by black pixels) is at this stage unknown. For mineral sediment flocs it can be expected to be close to 2600 kg m ⁻³ and for organic rich particles, expected to be much lower (1016-1600 kg m ⁻³ ) (Deng et al., 2019; Chassagne et al., 2021; Deng et al., 2021). Note that the assumption that the particle’s properties do not change along the direction indicated by c can lead to erroneous estimations.

The water samples were taken at 0.6 mab depth and were analysed using the LabSFLOC2 camera system (Laboratory Spectral Flocculation Characteristics, version 2) that recorded the settling velocity of particles from a pipetted amount of sample. The particle size, shape and effective density were estimated from the settling velocity (Manning et al., 2007b; Manning et al., 2010; Manning et al., 2017).

LabSFLOC2 was developed by Manning (Manning and Dyer, 2002; Benson and Manning, 2013; Manning et al., 2017), and it utilizes a high magnification 2.0 MP Grasshopper monochrome digital video camera (Manning and Dyer, 2002) to observe individual flocs as they settle in a 350 mm high by 100 mm square Perspex settling column. The video camera, positioned nominally 75 mm above the base of the column, views all particles in the center of the column that pass within a 1 mm depth of field, 45 mm from the Sill TZM 1560 Telecentric (maximum pixel distortion of 0.6%), 0.66 (1:1.5) magnification, F4, macro lens fitted behind a 5 mm thick glass face-plate.

During sampling, a modified pipette is used to carefully extract a floc sub-sample from the Van Dorn chamber and is filled to produce a fluid head of 50 mm, which results in a video image control sample volume nominally of 400 mm ³ (1 mm image depth and 8 mm nominal video image width, with a nominal 50 mm high suspension extracted with a modified pipette). This controlled volume permits in principle LabSFLOC2 calculated floc mass to be compared and calibrated directly to ambient SSC. In the present study, this was not performed.

The pipette sub-sample is immediately transferred to the LabSFLOC2 settling chamber, whereby the aperture of the pipette was brought into contact with the LabSFLOC2 settling column water surface and permitted the flocs to pass from the vertically held pipette to the chamber and settle solely under gravity, i.e. naturally and unassisted. Thus, the flocs allowed to pass into the settling column were naturally segregated as they fell by the process of differential settling; i.e. the fastest falling aggregates would be observed first.

Settling flocs are viewed as silhouettes (to reduce image smearing) resulting from a CCS LDL-TP-43/35-BL, 43 x 35 mm, homogeneous blue (470 nm) back-illumination LED panel, located at the rear of the settling column. The video images are streamed in real time as AVI files to a laptop PC via a FireWire-B PCI card interface. The digital floc images are captured at a frame rate of 25 Hz (one frame is 0.04 s), at a resolution of 1600 × 1200 pixels, with an individual pixel nominally representing 5 μ m (confirmed by independent calibration). The AVI files are not Codec compressed, so they could be analysed with MATLAB software routines. During post-processing, the HR Wallingford Ltd DigiFloc software - version 1.0 (Benson and Manning, 2013; Manning et al., 2017), is then used to semi-automatically process the digital recordings to obtain floc size and settling velocity spectra.

The sphere-equivalent floc diameter, D , was calculated from the measured major and minor axis of each observed two-dimensional floc by (Manning and Schoellhamer, 2013):

Each floc settling velocity was determined by measuring the vertical distance that the floc travels between a sequence of frames. Stokes Law then enables an estimation of individual floc effective density (Manning and Schoellhamer, 2013),

where W _s is the settling velocity, μ is the water viscosity and g is the gravitational acceleration.

During the survey day, 26 samples were analyzed using the LabSFLOC2 camera system.

A video camera mounted under the Van Dorn sampler was lowered at each sampling time, to take videos of the particles at the sample location. The camera was an Imaging source GigE color camera, I/O Aptina CMOS MT9P031, with a maximum resolution of 2592 × 1944. The pixel size is 2μm×2μm . The videos taken with the underwater camera were used for qualitative analysis of the flocs and not for quantitative size analysis.

The experiments were conducted with grab samples taken in the neighborhood of the survey (at the mooring indicated on the map and defined as PoR sediment.The samples were kept in a dark refrigerated room for two weeks before being used. The samples were sieved so as to remove particles coarser than 125 μm.

The microalgae Skeletonema costatum was chosen in the study, as being one of the most predominant algae species in the coastal area (Rijstenbil, 1987). The algae was obtained from the hatchery Roem van Yerseke B.V. The concentration of the algae stock suspension was about 5×10⁶ cells/mL and was kept in a dark environment to avoid photosynthesis and limit changes in concentration. The algae stock was refreshed every two weeks.

The Particle Size Distribution of samples was obtained by Static Light Scattering (SLS) using a Malvern Mastersizer 2000 which provides a full PSD every 30 s. The size range of the SLS device is 0.01-10000 μ m in 100 logarithmic-spaced bins. The samples are stirred in a 1L-jar at a constant rate of 35 rpm (Shear rate = 30 s ⁻¹ ), and pumped through pipes to and from the SLS device at a constant rate of 10 rotation per minute (rpm) which is about 12 s ⁻¹ . The set-up is the same as the one used by Ibanez et al. (Ibanez Sanz). The shear rates used are usually higher than those observed in the field (which are of the order of 0−10 s ⁻¹ ), especially at neap. However, these rates were determined in preliminary experiments to ensure a minimal breakage of the aggregated fractions and avoid settling in the pipes.

When a steady-state PSD was reached for a given algae concentration, the concentration of algae was adjusted to a higher content. The percentage of algae content was increased in steps to get concentrations of 10, 20 and 30 volume %. A separate measurement was done for a concentration of 5 volume %. Each percentage corresponds to the volume taken from the microalgae stock suspension divided by the total volume in the jar. At steady-state, some sub-samples have been pipetted and studied with the LabSFLOC2 camera system that was mounted in the laboratory. The description of the LabSFLOC2 camera system is given in Section 2.14. In order to obtain a good signal to noise ratio by SLS we used sediment and algae concentrations that are 100 times more concentrated than the ones observed in-situ during our one-day survey. The time scale of aggregation is therefore not be representative of in-situ conditions but an indicator whether flocculation is (or not) taking place.

From the underwater camera records and the LISST-HOLO pictures, different types of particles are identified, as shown in Figure 2. One type corresponds to (mineral) sediment-rich flocs, having an effective density close to 1600 kg m ⁻³ with settling velocity of 1 - 10 mm/s. These particles have compact form and dark color (high density). An other particle type corresponds to organic matter-rich particles having an effective density between 16 - 160 kg m ⁻³ with a settling velocity of 0.1 - 1 mm/s. These particles display an open or anisotropic structure and have a lighter color (low density). They are likely composed of inorganic and organic particles such as Transparent Extracellular Polymeric (TEP) substances (Jarvis et al., 2005; Spencer et al., 2010). For some of these flocs, the elongated algae structure is still recognizable, although the algae is coated with (in)organic debris and can be in some state of coiling. The third type corresponds to bare algae, or algae bound to very little amount of sediment, which have a very low density of about 16 kg m ⁻³ with low settling velocity, mostly around 0.1 - 0.5 mm/s. The microalgae Skeletonema costatum, for instance, is known to form elongated strings (Karp-Boss et al., 2007; Many et al., 2019). Despite having such different settling velocities and densities, all these particles are composed of different amounts of organic and inorganic matter. As detailed in (Chassagne et al., 2021), flocs composed in part of organic and inorganic particles are defined as Class 2 particles, whereas pure organic particles (like microalgaes) are defined as belonging to Class 3. Class 1 particles are defined as pure mineral clay particles, and none of the particles observed using video techniques were belonging to this class. As the survey day was very calm, only very small clay particles (<1 μ m) are expected to be in suspension.

The LabSFLOC2 camera cannot measure sizes below 10 μ m, a range that is accessible to the LISST-HOLO. This implies that in Figure 3 no particle below 10 μ m are recorded using the LabSFLOC2. As it was a calm day, not many particles were recorded, and a large spread in settling velocity was recorded the whole day. On the other hand, with the LabSFLOC2 technique, more large particles (> 100 μ m) were observed than with LISST-HOLO. This can partially be due to the fact that differential settling is occurring while particles are left to settle after being pipetted into the settling column. This would also lead to the cluster of fast-settling particles observed at 10:40 for example (also seen at 12:30) in Figure 3. Another possibility for these fast-settling particles is the coiling of flocs under shear that occurs continuously and lead to flocs as illustrated in Figure 2, at the bottom left. In Figure 4 a large spread in density for a given floc size is observed (left panel) - this spread is related to the spread in settling velocity for a given floc size, as shown in Figure 3. This spread was also observed in many studies, see for instance (Manning et al., 2007a; Manning and Schoellhamer, 2013). The spread in density for a given size can be attributed to two factors: first, the composition of a floc (the relative amount of mineral and organic matter within a floc) and second, a conversation artifact due to the shape of a floc. For highly anisotropic particles such as algae, the equivalent diameter will be estimated by:

where the AR the aspect ratio. For a 1 mm long algae, with AR=10 , this gives an equivalent diameter of about 100 μ m, which is the size that many flocs of aspect ratio close to 1 also have. Plotting the density as a function of the particle major axis can therefore help to better differentiate the different types of particles as illustrated in Figure 4, right panel where some elongated particles are now visible in the LISST-HOLO plot.

Particle density estimations are given in Figure 4 for the samples measured at 10:40 by LabSFLOC2 and LISST-HOLO. The data for that time is representative for all the three main type of particles found during the survey day. The density of the flocs recorded using the LabSFLOC2 was evaluated using the method described in Section 2.1.4. The density of the flocs recorded using the LISST-HOLO was evaluated using the method described in Section 2.1.3. This method requires the solid fraction density ρ _s as input, see eq.(3). From Figure 4, it can be seen that for particles having a major axis D _major <300 μm, the data found by LISST-HOLO and by LabsFLOC match when ρ _s is taken to be the one for mineral clay (which is done for the data plotted in the left panel as well as the one indicated as ρ _s = 2600 kg m ⁻³ in the right panel), whereas for particles having a major axis D _major >300 μ m, the data found by LISST-HOLO and by LabsFLOC match when ρ _s is taken to be close to the density of organic material (which was taken to be 1016 kg m ⁻³ ). This is done in the right panel (ρ _s = 1016 kg m ⁻³ ). This implies that a black/white scale to estimate the solidity as proposed in Section 2.1.3 is not appropriate to estimate densities and that a proper greyscale analysis should be performed, whereby the greyscale would be calibrated with the inorganic/organic composition of a floc. This can not be done at present, as this analysis requires a proper calibration of the greyscale colours based on more extended research.

The fact that density of flocs decreases with size has been reported by numerous studies (Gibbs, 1985: Al Ani et al., 1991; Vaezi et al., 2011; Mehta et al., 2014; Ramirez Mendoza). This density/size dependence is usually modelled using the relation

where D is the particle diameter evaluated according to eq.(4) and D ₀ is a reference size, usually taken to be the one of a primary particle. The parameter D _f found from fitting the data is often refered to as a fractal dimension. As we now have seen, even though eq.(7) usually models well the data, no information can be gained from D _f as in the situations where organic matter plays a role in flocculation, ρ _s will be a decreasing function of D.

As the size of the particles did not change significantly during the whole day, the PSD data from the LISST 100X, the LISST-HOLO and the LabSFLOC2 camera system were averaged over the entire day for each instrument. This enabled having a statistically significant data set, as the amount of particles detected was low during the day. Figure 5 shows the day average particle size distributions estimated by the three techniques. Although the LabSFLOC2 camera data are not directly comparable with the LISST data due to differences between the two techniques (Mikkelsen et al., 2005), the overall mean size of the particles obtained from the LabSFLOC2 camera correlates with the mean particle size obtained from the LISST data and fluctuates between 150 and 220 μ m.

The clay-silt fraction (< 10 μ m) can only be assessed by LISST 100X and is out of the LISST-HOLO and LabSFLOC2 system range.

The averaged PSD measured by LISST 100X is trimodal with peaks in particle sizes at 300, 200 and 70 μ m. It was shown by Karp-Boss et al. that biological species cause trimodal distributions in LISST 100 data due to their very anisotropic shape (Karp-Boss et al., 2007). The PSD measured by LISST-HOLO displays two marked peaks which overlap the 200 μ m and 300 μ m LISST 100X size peaks. Graham et al. showed that the PSD produced by LISST-HOLO usually indeed overlap with the PSD of LISST 100X in the size range (20-500 μ m) (Graham et al., 2012). The LabSFLOC2 data on the other hand displays one wide peak that tends to be bimodal, overlapping mainly the 200 μ m peak and slightly the 300 μ m peak. Graham and Nimmo Smith (2010), who compared the LISST-HOLO PSD with the Malvern Hydro 2000G PSD, have found that the LISST-HOLO PSD showed multiple modes as compared to the monomodal PSD produced by the Malvern Hydro 2000G, which were attributed to overlapping large particles in the field of view of the holographic system. In our case, the presence of highly anisotropic particles leads to same kind of effect. The PSD found using the LabSFLOC2 technique is based on the particle size estimation given by eq.(4). Both minor and major axis are visually determined and therefore no error is made when identifying the particles.

Agrawal et al. inverted the data found by LISST 100X using different kernel matrix to produce particle size distribution for non-spherical particles (Agrawal et al., 2008). However their method does not apply to very anisotropic particles such as bare algae and flocculated algae as are present in our observations. This is why a combination of holographic pictures and/or other video or camera recording, are necessary to interpret the LISST 100X PSD as was also concluded by Graham et al. (2012).