Background

Water salinity has long been deemed an influential factor in the flocculation of fine sediment (e.g., Odd, 1988). Fundamental to this understanding is the small mass of the fine sediment particles (which allows them to aggregate when close enough to each other) and the fact that the particles are often negatively charged due to the isomorphous substitution and preferential adsorption in their crystalline structure (Partheniades, 2009) or adsorption of low molecular weight (LMW) organic compounds (Tipping & Cooke, 1982;Tipping & Higgins, 1982). The negative charge causes a repulsion force that prevents the particles from getting close enough to aggregate. Increased salinity can promote flocculation by shrinking the electric double layer (EDL) that surrounds the charged clay minerals (Gregory & O'Melia, 1989). The EDL consists of a Stern layer that contains ions with a charge opposite to the clay mineral surface charge and a diffuse layer that contains free ions with a higher counterion concentration. When the ion concentration of water increases, both the Stern layer and the diffuse layer shrink, making it easier for particles to get close enough to each other for van der Waals forces to dominate and allow particles to aggregate and form flocs (Gregory & O'Melia, 1989). An extensive line of research has provided evidence for the growth-enhancing effects of salinity on the flocculation of cohesive sediment both in laboratory experiments and in field surveys.

Laboratory studies have generally reported that increased ion concentration of water leads to a growth in floc size and an associated increase in mud settling velocity. Edzwald et al. (1974) was one of the earliest attempts to quantify the effects of ion concentration in the water on suspensions of clay minerals such as kaolinite, illite, and montmorillonite. Based on their experiments, they reported enhanced aggregation for all clay suspensions with increased salinity. Further research has shown increase in floc size (e.g., Abolfazli & Strom, 2022;Guo et al., 2021;Mietta et al., 2009;Mikeš & Manning, 2010) and settling velocity (e.g., Li et al., 2021;Portela et al., 2013). While added salt is generally acknowledged to increase the size and settling velocity of clay-based flocs in the laboratory, the effects of salinity increases have also been found to have a threshold past which further increases in salt concentration have little effect on floc properties. For instance, S = 15 ppt was reported by Mietta et al. (2009), S = 10 ppt by Mikeš and Manning (2010), and S = 7 ppt by Guo et al. (2021) to be the concentration threshold for salinity effects.

It is difficult to isolate the effects of increased salinity on flocs in natural settings since other flocculation drivers such as turbulence, suspended sediment concentration, and the organic fraction of suspended sediment tend to co-vary along with salinity. Nevertheless, field studies have also generally described a positive, though less pronounced, relationship between salinity and floc size. In the Yangtze estuary, the largest flocs were observed in high slack water when salinity peaked and shear was lowest, while smaller flocs were found during flood and ebb tides when turbulence was stronger (Guo et al., 2017); whether the peak in floc sizes was due to changes in salinity or turbulence or was a result of larger material being advected into the sampling region is unknown. In the Ems Estuary, the largest flocs were observed in the seaward direction, where the suspended sediment concentration

Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 3 of 16

was not necessarily at its maximum (van Leussen, 2011). The presence of a halocline was suggested as the main driver for vertical variation in floc size in the Pearl River estuary by Y. Zhang et al. (2020). However, the authors later argued that, although increased salinity is important in enhanced fluctuation, the density stratification due to saltwater intrusion played a stronger role in modulation of the floc size distribution in the Pearl River estuary (Y. Zhang et al., 2021).

Another driver of cohesive sediment flocculation is organic matter (OM). OM is known to be a significant factor in aggregation dynamics of sediments and has been suggested to affect flocculation in freshwater (Droppo et al., 1997), estuarine (Eisma, 1986), and coastal (Fettweis et al., 2022) environments. In fact, Eisma (1986) proposed that OM was the first-order influence on floc size and that increases in salt concentration with an estuary have a minor to negligible effect. And, the laboratory experiments of Mietta et al. (2009) showed that flocs formed from mud devoid of OM were smaller compared to those formed from mud with natural OM. Due to their structural complexity, and relatively larger size compared to clay crystals, OM can modify the characteristics and behavior of clay particles in a myriad of ways. Microorganisms or a part of their structure can be bound in and become a part of floc structure (Droppo, 2001;Droppo et al., 1997). They can change the surface charge of clay particles (Beckett & Le, 1990), act as a binder for clay particles due to their complex structure (Labille et al., 2005;Theng, 2012), or grow on the particles (Shen et al., 2019;Tang & Maggi, 2018). The OM present in freshwater can have terrestrial or autochthonous origin and thereby vary seasonally (Lee et al., 2019). Additionally, despite the historical assumption that cohesive sediment is unflocculated when transported in fluvial systems, an increasing body of observational (e.g., Droppo et al., 1997;Fox et al., 2014;MacDonald & Mullarney, 2015;Osborn et al., 2021), experimental (e.g., Abolfazli & Strom, 2022), and analytical (e.g., Lamb et al., 2020;Nghiem et al., 2022) evidence support the ubiquitous presence of flocs in rivers and streams. Without the presence of measurable salinity in most cases, the explanation for the presence of flocs in freshwater systems has been the presence of OM serving as a binder for the inorganic and organic components of the flocs.

To account for the influence of drivers such as OM and salinity on mud transport, one must have a mathematical model for floc size or settling velocity that can be integrated into a larger hydrodynamic and sediment transport framework (e.g., Sherwood et al., 2018;Tarpley et al., 2019;Teeter, 2000;Verney et al., 2011;Winterwerp, 1998). Recently, there has been a push to include the impact of OM on floc population models (e.g., Shen et al., 2019). However, models that account for alterations in the floc sizes or settling velocity brought on by changes in salinity are more limited, perhaps due to the lack of data suitable for model development. The two models that include a salinity effect are the approaches by Delft3D (Deltras, 2021) and Horemans et al. (2020). In Delft3D, the salinity effect is modeled directly on the settling velocity of the mud (i.e., floc size is not modeled). The Delft3D salinity-driven model assumes that settling velocity, w s , approaches a maximum value as salinity increases and approaches a value past which increases in salt no longer influence floc size and settling velocity, S max . Horemans et al. (2020) (hereafter, H20) takes a similar approach in that salinity effects are limited to a range of values below a cutoff or threshold salinity. But, rather than mapping these effects directly to the mud settling velocity, they propose a floc aggregation efficiency parameter, usually taken as a constant in floc size models, as a function of salinity. The equation for the efficiency parameter they developed was based on three data points from Edzwald et al. (1974) (hereafter referred to as E74), and using it allowed them to predict changes in floc size, and hence settling velocity, due to changes in estuarine salinity.

Study Purpose

The studies discussed above point to the significance of both ion concentration of the water and OM in the flocculation of cohesive sediments. Either can likely independently influence flocculation, and it is unclear if the independent effects are linearly additive or if there are interactions between the two that override the other or provide a joint effect. Answering this question is particularly important for modeling of mud in natural environments because both cations and OM are always present in some form. In addition, while changes in salinity are thought to drive changes in flocculation rate and equilibrium size, efforts to parameterize and model its behavior in natural environments are scarce and not well tested or calibrated.

In this study, our first goal is to test the hypothesis that the positive correlation between increased aggregation rate and equilibrium floc size and increased salinity is more pronounced when organic matter is present. That is, we expect the effect of salinity change on mud flocs to be positively modulated by organic matter. The rationale for this hypothesis is that the presence of both salt and OM affects flocculation by changing the electric charges that

Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 4 of 16

are present on the ions, clay particles, and OM molecules and the alterations they make to the inter-particle interactions. This suggests that there exists a potential interaction between the different components of flocculation, that is, sediment particles, ions, and OM. The specific questions we aim to answer in this study are: (a) is there a fundamental difference in the way natural mud flocs and those devoid of OM respond to an increase in salinity; and (b) is there a particular range of salt levels past which increases in salt in the suspension of either type of sediment no longer influence the flocculation properties? A broader question of consideration associated with our study is whether or not salt has any effect on a mud suspension if organic binders are already present and causing flocculation in freshwater rivers before they enter estuarine environments. The second goal of our study is to use data from the experiments to develop a method for including the impacts of changes in salinity within a Winterwerp (1998) type formulation for dynamic or equilibrium median floc size.

Materials and Methods

Approach Overview

To test our hypothesis, we measure the size distribution, equilibrium median size, and aggregation rate of mud flocs in a laboratory mixing chamber for three sets of experiments using OM-free kaolinite clay, natural sediment containing OM, and that same natural sediment treated to remove its organic content. Salinity and turbulent shear rate are varied across each of the three sets of experiments (Table 1). Data are collected with a floc camera to measure the size distribution of the particles in suspension and an optical backscatter sensor (OBS) is used to measure the turbidity. Comparisons between experiments with and without OM and across increasing salinities are used to test the hypothesis and answer the specific research questions. The turbulent shear rate during the experiments proceeds through step-downs in mixing rate to mimic the weakened turbulence conditions in a river plume, delta, or estuary. Below we present a detailed description of the mixing chamber, the camera system, the salt and sediment used in the experiments, and the experimental procedure.

Mixing Chamber and Data Acquisition Setup

The experiments were carried out in a 13 L mixing chamber (27.5 × 27.5 × 25 cm) equipped with an overhead stirrer motor connected to a paddle that allows the mixing rate and hence turbulent shear rate (G) to be adjusted within the chamber. G is a measure of the dissipation rate of turbulent kinetic energy, ϵ, and is defined as

, where ν is the kinematic viscosity of water. G is a widely used parameter in modeling the flocculation process because turbulence drives both aggregation (due to increasing the likelihood of particle collision) and breakup of flocs (due to increased mechanical stress exerted on flocs) (e.g., Lee et al., 2011;Winterwerp, 1998). G was estimated using a relationship proposed by Logan (2012) based on tank and paddle geometry and paddle speed:

where C D is the drag coefficient, A and R are the paddle area and radius, respectively, s is the paddle speed, ν is the fluid kinematic viscosity of water, and V T is the volume of water in the chamber. Our tank and paddle setup inevitably results in inhomogeneous turbulence with higher shear near the paddle and lower shear moving out away from the paddle. The G we report and link to the overall size distribution of flocs and particles in the tank via Equation 1 is a tank or volume-averaged G value. Therefore, in these experiments, we take the resulting size distribution of particles observed in the tank to be the integrated outcome of the various shears experienced by the particles as they are advected throughout the well-mixed tank. We use this approach partly for pragmatic reasons (it is difficult to create truly homogenous turbulence in a vessel in which the same set of particles can be tracked), and because no spatial gradients in suspended sediment concentration or suspended particle size have been observed within the tank. We speculate that floc size distributions measured in our tank at a particular G may be slightly smaller than those in a homogeneous turbulence field all else being equal because of the faster time scales associated with floc breakup in zones of higher shear.

Experiment set Sediment type Salinities tested [ppt] 1 Kaolinite clay 0, 2, 10 2 Natural mud (natural OM) 0, 0.5, 1, 2, 3, 5, 10 3 Treated mud (devoid of OM) 0, 2, 10

Table 1 Experiments Were Conducted Using Three Types of Sediment at Various Salinity Levels Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 5 of 16

The mixing tank was also equipped with a camera system that consisted of a waterproof LED light source placed inside the tank and a camera placed outside of the tank. The camera records images of particles passing through the slit created by the light source and the wall of the mixing chamber (Figure 1). The system is capable of capturing images of flocs in the size range of approximately 5-1,500 μm. Details on the mixing chamber and the camera system can be found in Tran and Strom (2017). The images captured by the camera were then organized using a Python script that was based on the procedure by Keyvani and Strom (2013) and processed in ImageJ to identify the particles present in each image. The particle, or floc, sizes were then extracted using the measured projected area of each particle as

, where d f is the floc diameter or size, and A f is the measured area of each floc. Associated floc volumes were calculated as:

= 3 ∕6 , where V f is the floc volume. The identified particles or flocs were then sorted into log sized bins to produce floc size distributions based on floc volume and size statistics such as d 50 (the median floc size by volume). Defining d 50 by volume eliminates the need to assume a fractal dimension for flocs. Additionally, obtaining actual and real-time images of flocs with the camera system enabled us to visually compare the flocs formed in different conditions in addition to measuring their size (compared to the particle size and volume distribution estimation methods that are based on the principles of laser diffraction).

Salt

Previous experimental flocculation studies have used different types of salt to represent sea salt; for example, salts used include table salt (sodium chloride) (e.g., Nasser & James, 2006), commercial aquarium salts (e.g., Tan et al., 2014), and a mixture of multiple salts (e.g., Edzwald et al., 1974). While all of these salt mixtures increase the ion concentration and salinity of water, it has been shown that in addition to the concentration of salt (i.e., ion concentration), the type of salt can be quite important in setting the flocculation behavior of mud (e.g., Abolfazli & Strom, 2022). Divalent and polyvalent cations (compared to monovalent ions), for instance, are more efficient in facilitating the bonds between the sediment particles mostly due to cation bridging effects (Lai et al., 2018;Theng, 2012). To recreate saline water as close as possible to seawater, we used an American Society for Testing and Materials (ASTM) grade sea salt substitute (Lake Products Company LLC, Florissant, MO, USA) in our experiments. This salt contains nine other constituents in addition to sodium chloride including magnesium chloride and sodium sulfate. Deionized (DI) water was used as the background water in all experiments to eliminate the effects of the ions that are already present in tap water and their potential variations over the period of the study.

Sediments

Kaolinite clay, natural mud, and natural mud devoid of OM were all used in the experiments. The first set of runs was conducted using kaolinite clay (ActiveMinerals, Maryland, USA). Kaolinite has less total negative charge compared to other common minerals found in freshwater sediment (e.g., montmorillonite and illite) due to smaller isomorphous substitution that takes place in its tetrahedral or octahedral sheets. Because of this, and also due to the presence of positive charges on its edges and negative charges on its surfaces, kaolinite can create flocs even in DI water (Partheniades, 2009). Natural sediments typically consist of various inorganic and organic components. The base sediment used in the second and third sets of experiments was fine bed sediment collected from Stroubles Creek, which is a stream in southwest Virginia, USA. The sediment was wet sieved to include only the material in the clay and silt size range (<62.5 μm). During the study, the natural mud was kept in a dark fridge at 4°C. The organic matter content of the mud was measured as 11.7% based on loss on ignition. The sediment used for the third set of experiments was the same as the one used for the second set except it was treated with sodium hypochlorite following Siregar et al. (2005) to remove OM content. Experiments with kaolinite clay, which is devoid of OM, enabled us to draw a comparison between the naturally OM-free flocs and the flocs formed from the treated fine sediment. Figure 2 shows the disaggregated size distribution of the kaolinite clay and bed sediment used in the experiments. All flocculation Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 6 of 16 experiments were conducted at a sediment concentration of 100 mgL -1 . Pre-weighed dry kaolinite clay was used for the first set of experiments.

For the second and third sets, we used wet sediments, the OBS, and a calibration between NTU and the mass concentration to set the experiments to the 100 mgL -1 . The calibration curve was produced by filtering, drying, and weighing disaggregated particles. For both the treated and untreated natural sediments, 55 NTU was found to produce a concentration of 100 mgL -1 .

Experimental Procedure

In each experiment, the tank was first filled with 13 L of DI water, followed by the addition of salt (or no salt in the DI water experiments). In all three sets of experiments, the dry or wet sediment was added to 40 mL of DI water, and the mixture was sonicated for 15 min to break up any aggregates that may be present. Once the sonicated sediment was added to the tank, the experiments commenced with a 1-hr period of low turbulent shear (G = 35 s -1 ) where flocs were allowed to grow from a sonicated state. We referred to this first hour as the initial growth (IG) phase. Flocs were then broken up during a high shear (HS) phase at G = 550 s -1 over a period of 15 min to allow the process of flocculation to start from a more natural state (from turbulence-generated and not sonicated particles). After the HS phase, the suspension was put under five periods of descending turbulent shear rates (i.e., G = 95, 70, 50, 35, and 20 s -1 , respectively) to mimic the range of shear a river could experience as it makes its way to the sea. For the kaolinite clay, each period lasted 90 min. For the natural sediment, these periods were 150-min long as we had previously observed that natural sediment required more time compared to kaolinite clay to reach equilibrium in size after each step-down phase.

Results

Overview

The floc size population was extracted from the images that were captured at a frequency of 1 Hz and grouped into 1-min samples. The size distribution of each minute was then used to calculate the characteristic floc size statistics such as the d 50 (median floc size by volume) as a function of time (Figure 3a). The d 50 time series also provides the rate of change of the characteristic floc size as a function of G and S. Generally, flocs rapidly grew as soon as the sonicated sediment was added to the mixing chamber (Figure 3b). Flocs that form during this stage were then broken down during the high shear phase. From this point, they then increased in size again with each step-down in shear (Figure 3b). d 50 was found to reach an equilibrium size faster at higher turbulence levels, and kaolinite reached equilibrium for a given shear rate faster than the natural mud. At the lowest shear rate (G = 20 s -1 ) using natural mud, some of the flocs grew large enough to settle out of suspension. This led Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 7 of 16 to a negative slope in the d 50 time series in roughly the second half of the G = 20 s -1 stage (Figure 3). We present the results by sediment type in the following sections.

Kaolinite

In the DI water experiment (S = 0 ppt), the sonicated kaolinite clay readily flocculated once it was added to the mixing chamber (Figure 4), reaching a d 50 of about 40 μm within the 1-hr initial growth phase. Flocs were found to be weak enough to notably break up during the high shear (HS) phase of the experiment but ultimately grew back up to about 45 μm in diameter at G = 20 s -1 by the end of the experiment (Figure 5).

The next two kaolinite experiments were conducted with added salt (Table 1). The presence of salt in the water caused the flocculation to be inhibited relative to the DI water experiment, leading to a decrease in the equilibrium floc size; in the experiment at S = 2 ppt, floc size increased to only 20 μm during the initial growth phase, and the d 50 did not exceed 20 μm even at G = 20 s -1 . Although flocs were slightly larger in the experiment at S = 10 ppt compared to those at S = 2 ppt, they were still smaller compared to those in the DI water experiment at all turbulence levels (Figure 5).

The flocs' response to reduced turbulence during each step-down phase occurred quickly. New equilibrium sizes were reached at each new shear level within 20 min of the change in shear. To quantify their growth rate, we calculated Δd 50 /Δt, which is the change of d 50 over a specified duration of time, Δt (15 min for the kaolinite experiments), immediately after G was reduced at each step-down phase. The response was stronger in the DI water experiment, in which d 50 increased at a rate of 0.2-0.4 μm min -1 at G > 35 s -1 and more than 0.5 μm min -1 at G = 20 s -1 (Figure 5). In comparison, in the salt experiments, d 50 increased by only a few μm at each step-down with Δd 50 /Δt never exceeding 0.2 μm min -1 at any turbulence level.

Bed Sediment With Natural OM

Similar to the kaolinite experiments, we first examined the flocculation behavior of the natural bed sediment without the addition of salt using DI water (S = 0 ppt). In the absence of added salt, two distinct groups of suspended particles were observed across all shear rates. The larger of these two groups consisted of flocs in the >200 μm size range. The number of particles and total sediment volume in this larger group was, however, strongly outnumbered by particles in the second group, that is, the smaller un-or less-flocculated particles (Figure 6a).

Six salt experiments were conducted at salinities ranging from S = 0.5-10 ppt (Table 1). A key takeaway from these experiments is that it only took a very small increase in salinity to strongly enhanced the flocculation of natural mud such that at S = 0.5 ppt, the number of flocs were noticeably greater at all of the turbulence levels compared to the DI water experiments (Figure 6b).

The increase in d 50 in response to increased salinity can be traced back to the aggregation of unaggregated particles and smaller flocs mostly in the 25-75 μm size range. That is, as salinity increased, the unaggregated particles became more and more integrated into the floc structure. The loss of fine material is evident in the images as the Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 8 of 16

background of small dispersed particles in DI water reduces with salinity (Figure 6). The change in floc population in response to increased salinity is also evident in the particle size distribution (PSD) (Figure 7). While some relatively large flocs with the size of >200 μm did form in DI water, the fraction of discrete unaggregated particles was notably larger than those in the experiment at S = 0.5 ppt. The PSD moved toward larger particle sizes as salinity increased, leading to a larger d 50 .

The equilibrium floc size (d 50 ) increased notably from the DI water experiment to S = 0.5 ppt, and then to 1, 2, and 3 ppt. These enhancing effects of increased salinity, however, were found to be most evident at S ≤ 3 ppt, and a further increase in salinity from 3 to 5 ppt and then to 10 ppt did not result in notable increases in d 50 (Figure 8). The stronger effects of increased salinity at lower salinity levels were observed at all shear levels. Salinity affected not only the equilibrium floc sizes but also the growth rate of flocs in response to each reduction in turbulent shear rate (Figure 8b). While in DI water the equilibrium floc size increased at a very slow pace (or decreased as some larger flocs with diameters of roughly >250 μm settled out of the suspension), the presence of salt at the level of only 0.5 ppt in water increased the flocculation rate by as much as 2 μm min -1 at G = 20 s -1 . Similar to the equilibrium floc size, the enhancing effects of salinity on flocculation rate had a threshold, and the strongest effects were found at S ≤ 5 ppt.

Treated Bed Sediment With OM Removed

Mud floc size and structure were clearly different when the treated sediment was used in the flocculation experiments. Interestingly, in the experiment with no added salt (DI water), the treated bed sediment devoid of OM did not form any visible flocs, and the suspended particles retained their sonicated dispersed form (Figure 9a).

It was only in the presence of salt that treated sediment started to aggregate and form flocs. In the experiment at S = 2 ppt, small flocs with the size of <20 μm formed immediately after the sonicated sediment was added to the mixing tank. However, d 50 of the flocs never exceeded 30 μm during any of the shear step-downs (Figure 10). Similar behavior was observed in the experiment with S = 10 ppt. Although the size and growth rate of flocs were slightly greater compared to those in the experiment with S = 2 ppt, their d 50 hardly reached 35 μm even at G = 20 s -1 . 21699291, 2023, 7, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JC019255 by Virginia Tech, Wiley Online Library on [20/05/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 9 of 16

Discussion

Organic Matter and Flocculation of Natural Mud

Gums such as xanthan gum, guar gum, and chitosan are often used in laboratory experiments looking to study the role of organic material on flocculation (e.g., Furukawa et al., 2014;G. Zhang et al., 2013;Zeichner et al., 2021). In this study, we used an alternative approach by removing OM from natural sediment rather than adding gum to the suspension.

Our experiments revealed a stark difference between flocs formed from natural unaltered bed sediment (which contains OM) and flocs that form from suspensions of the same sediment after it has been treated to remove the organics (OM-free). Figure 12 visually highlights this distinction. In the images, salinity, shear rate, and base inorganic sediment components are identical in panels (a) and (b). The only item that is different is the OM content. Untreated bed sediment formed larger flocs, where most mud particles were incorporated into the floc structure. The incorporation of discrete particles caused the background to be clearer, resulting in more defined and darker flocs in the images due to their higher contrast with the clearer background. In comparison, flocs were smaller in the treated sediment experiment. The difference we observed in floc size is greater than that reported by Mietta et al. (2009), where the increase in floc size from OM-devoid mud to OM-containing mud was limited to <30 μm at G = 35 s -1 . The flocs formed from the OM-free mud were also more transparent and fragile than the OM-containing flocs. These differences between the sediment that contained OM and that devoid of OM were present at all salinity and turbulence levels that were tested.

The strong flocculation-enhancing role of the OM that is naturally present in freshwater sediment has been attributed to the structure and weight of OM molecules. Organic biopolymers, such as the extracellular polymeric substances (EPSs), contain numerous charged or uncharged groups in their structure, all contributing to the forces between the organic and inorganic components of the mud (Lai et al., 2018). The uncharged groups of biopolymers can interact with the negatively charged surface of clay minerals through the van der Waals forces. Hydrogen bonds can occur between the basal hydroxyl surface of silicates such as kaolinite and polar groups of biopolymers (Theng, 2012). Cationic groups of the polymers can further interact with clay mineral surfaces through electrostatic forces, while the anionic groups can attach to negatively charged surfaces of clay minerals through polyvalent cations acting as a bridge between the two (Philippe & Schaumann, 2014). All of these interactions can lead to a complex looped structure of OM-sediment bonds in the floc structure (Figure 11). The role of OM in flocculation is not limited to forming the general size and shape of flocs as a whole. The presence of OM can also alter the nature of the primary particles, that is, those particles that are the base-level building blocks of flocs. At this level, sediment particles and OM can firmly bind, thereby increasing the size of these building blocks (Fall et al., 2021). When OM is absent from the sediment, flocs form by absorption of differently charged sections of mineral structure (such as positively charged edges and negatively charged faces of kaolinite) (Partheniades, 2009) or due to polyvalent cation bridging between ABOLFAZLI AND STROM 10.1029/2022JC019255 10 of 16

two negatively charged clay particles (Mietta et al., 2009) or shrinking of the EDL and the resulting net attractive van der Waals forces (Spielman, 1978). In all of these cases, the structure of the flocs is different from the flocs that contain OM (Figure 12).

Salinity and Flocculation

Salinity has historically been considered a driving factor in the flocculation of mud deposition in estuarine zones (e.g., Odd, 1988). Our results support the idea that the presence of salt increases both the rate of aggregation and the equilibrium size of mud flocs formed from fluvial bed sediment. While this general trend or principle is in line with historic understanding, the experiments presented here offer three more refined points of consideration to this general trend and observation regarding salt and its role in the flocculation of natural mud.

First, it appears to us that the primary role of salt is to make it easier for unaggregated small particles to flocculate with existing flocs or each other. Our experiments did not show a simple relationship whereby flocs do not form in freshwater and then do form when salt is added. For the untreated natural bed sediment, we still observed the formation of flocs even without the addition of any salt, that is, flocs were present in the pure DI water experiment (Figure 6a). However, a notable fraction of the suspended matter did remain in a discrete, unaggregated state, and the presence of flocs and unaggregated material resulted in a bimodal PSD (Figure 7). We expect that the existence of flocs with a diameter of >200 μm was most likely driven by OM, particularly macromolecules that contain multiple charges that act as a binder for sediment particles because identical experiments with the treated bed sediment (OM removed) in DI water did not produce any flocs (Figure 9). The bimodal size distribution of the suspended material for the S = 0 ppt experiment was likely the reason for the drop in equilibrium d 50 at G = 20 s -1 due to settling of a fraction of the largest particles (Figure 8a), which also resulted in negligible growth rate at this shear rate (Figure 8b). Considering the role of salt on the PSD then, the primary role of the salt in the natural sediment case was to change the PSD from a bimodal (unflocculated and flocculated) distribution to one that is more uni-modal (flocculated); or to change the fraction of material in suspension that resides in larger flocs.

Second, the PSD for the natural bed sediment with OM had a stronger response to increases in salt than the PSD for the bed sediment without OM. That is, the presence of OM did not override or remove a flocculation response to changes in salinity. Instead, it enhanced them. Without OM present, floc sizes increased in response to an increase in salinity. But the change in size, as represented by d 50 , was limited to a change of approximately 10-15 μm from S = 0 ppt to S = 10 ppt (Figure 10); translated to settling velocity of approximately 0.03 mms -1 difference. In contrast, flocs formed with the untreated bed sediment experienced a change in d 50 of 50-150+ μm with the addition of salt depending on shear rate (Figure 8); a change of almost 0.5 mms -1 in terms of settling velocity for G = 35 s -1 . We suspect that increasing salinity had less of an effect on the treated sediment because when OM is removed, the electrically charged points of contact that can cause Journal of Geophysical Research: Oceans ABOLFAZLI AND STROM 10.1029/2022JC019255 12 of 16

is present, even when the overall conductivity of the water is still low relative to marine or estuarine standards. We, therefore, expect that flocs in rivers are governed not just by the organic matter, but also by other ions, salt or otherwise, that may be present. As a result, one can imagine a case where the organic matter and background ions, that is, non-saltwater-intrusion-related ions, are sufficient to promote the flocculation of a large fraction of the mud load in freshwater systems. Additionally, our findings suggest the effects of salinity on modulating the fraction of mud flocculated or the overall floc PSD to be limited to the head of a well-mixed estuary or to the fresh-saltwater interface in a vertically stratified salt wedge system.

As a final note, given the complex interaction between clay minerals, OM, and ions, we speculate that the order in which OM and ions are added to the clay affects the flocculation process. In the present study, we added salt to mud that already contained OM to mimic the increase in ion concentration as freshwater mud experiences higher levels of salinity as it approaches estuaries and oceans or experiences a spike in ion concentration due to deicing salt runoff (Abolfazli & Strom, 2022).

How Applicable Are These Findings to Coastal Mud?

Our experiments were conducted using natural mud gathered from a local stream in the Valley and Ridge province of Virginia, that is, Stroubles Creek. Stroubles Creek is a tributary to the New River, which in turn is a tributary to the Mississippi River. Sediment found in the creek can ultimately make its way to the Gulf Coast region, and spikes in salinity do occur in the creek following runoff of roadway deicing salts during winter (Abolfazli & Strom, 2022;Lakoba et al., 2021). Therefore exploring the role of organic matter and salinity on the flocculation of Stroubles Creek mud specifically has utility for understanding its transport dynamics. Nevertheless, it is reasonable to question how broadly applicable the experimental results are to other muds-and in particular, muds that are more proximally located to coastal zones where changes in salinity are more eminent.

As one step toward more broadly testing the response of different muds to changes in salinity both with and without natural organic material, we conducted experiments similar to those outlined in the methods and results for the Stroubles Creek mud using mud obtained from the bed of the main channel of the Mississippi River near Venice, LA. The mud was collected as part of a larger project in January 2021 using a Shipek grab sampler. Tests were run on the treated (OM removed) and natural (no removal of OM) muds in DI water and DI water with enough salt added to bring the salinity to 2 ppt. Images of the suspension are shown in Figure 13. Images and floc size measurements indicate that the general behavior concerning salts and organic matter for the Mississippi River mud was the same as that of the Stroubles Creek mud. The unaltered mud in DI water produced a bimodal distribution with a few larger flocs and many smaller aggregates and unflocculated particles (Figure 13a). Adding salt resulted in a much higher degree of flocculation, a more unimodal distribution, and an overall larger median size (Figure 13b). Without OM (i.e., for the treated case) no large flocs formed in DI water at S = 0 ppt (Figure 13c). The addition of salt led to a more flocculated state, but floc sizes were smaller than those produced at the same shear and salinity level but with the presence of OM (Figure 13d). The primary Our back-calculated ′ increased with S (the values obtained are represented by boxplots in Figure 14). Following Horemans et al. (2020), we averaged all calculated ′ for each salinity and then fit a hyperbolic tangent function through the average. The result of this process yielded:

where ′ 0 = 0.05 denotes ′ at S = 0 ppt Equation 6 has a total residual sum of squares (RSS) of 0.064 and a R 2 = 0.73 when using the average ′ for each S.

Overall, our data suggest a similar functional shape in the relationship between ′ and S as that Horemans et al. ( 2020) and the Delft3D models. However, our data show a slight increase in the response of ′ to increasing S and one that occurs at a lower salinity. For example, the change in ′ with S in our case all occurs before a change in ′ is predicted by the Horemans et al. ( 2020) model (Figure 14).

While our aim was to provide a functionality between ′ and S, the data did show that ′ was also affected by G.

In the analysis, we present here we have grouped all of the variations with G at a single salinity value (reflective in the boxplots in Figure 14) into an average. However, we did observe a systematic increase in ′ with a decrease in G. It should also be noted that it is possible that the equilibrium floc sizes are underestimated at the lowest shear rate in our experiments (i.e., G = 20 s -1 ) due to the settling of a small fraction of the flocs. The model presented in Equation 6 is based on the results of our experiments. How generic the equation is remains unknown. We expect that the form of the equation will capture the general behavior of mud flocs in the presence of increasing salinity and that scaling the equation by the zero-salinity aggregation efficiency will aid in making it extendable to other suspensions of very sediment and organic content. However, it is possible that the constant coefficients (i.e., 0.4, 1.2, and 0.55) may need to be adjusted to capture a given mud's behavior most accurately.

With the natural mud used in our experiment, most of the change in floc size and growth rate occurred at salinities below 5 ppt with the positive effects of salinity on enhancing ′ being evident even at salinities as low as 0.5 ppt. While the model predicts a reversible response with S, one would expect the physical process to not be perfectly reversible. Additional data are needed with other natural sediment mixtures to better understand the level of generality that our data and Equation 6 reflect.

Conclusion

Our experiments reveal a strong interaction between organic matter and salinity in driving the flocculation dynamic of natural mud. Moving from DI water to salinities of 10 ppt increased the equilibrium floc size and growth rate of flocs formed in natural sediment both with and without natural organic matter. Without organic matter, flocs were limited in size to 20-30 μm even at 10 ppt. With natural OM present, a limited number of large flocs formed even in pure DI water. However, the fraction of material flocculated remained low, resulting in a bimodal suspended particle size distribution. Adding just a small amount of salt to bring salinity to 0.5 ppt notably increased the fraction of flocculated material and changed the particle size distribution from bimodal to unimodal. Increases in salt up to 3-5 ppt decreased the number of unflocculated particles further and increased floc sizes such that the average reached between 100 and 200 μm depending on the shear level. Increases in salinity past this level led to only marginal increases in floc size for a given shear rate. The data suggest that both organic material and ions associated with river or estuarine water are likely present if high degrees of flocculation, with flocs on the order of 100 μm and larger, are observed. The data also suggest that very low levels of salinity (e.g., 0.5 ppt) are needed to enhance floc size. It is therefore likely that large salinity-driven changes to the floc size distribution in nature are limited to the head of a well-mixed estuary or to the fresh-saltwater interface in a vertically stratified salt wedge system in rivers with low background freshwater ion concentration. Change in floc growth and equilibrium size brought on by salinity changes in the presence of natural mud and OM can be captured with a salinity-dependent aggregation efficiency parameter in a Winterwerp (1998) type formulation.