Different sediment samples were collected with a Van Veen grab at two locations in lake Markermeer: the southwest (SW) and the northeast (NE) sites (see Figure 1). The upper layer of the lake bed consists of a thin (circa 0.1 m) layer of soft sediment, of which only the few upper millimeters are oxic (Van Duin, 1992). The sediment composition of this soft sediment layer is detailed in Table 1, labeled as sample NE2B-70%. Underneath this soft sediment layer, a thick layer of Holocene deposits (clay/silt mixtures, peat, or sand) is present (Rijkswaterstaat, 1995). The material from this layer tested in this study is referred to as clayey silt and was sampled at the southwest sampling site (Table 1). Markermeer sediment mineralogy is dominated by illite and smectite at all sampling sites, while kaolinite (<11%) and chlorite (<6%) are present in small amounts (see Barciela–Rial, 2019). After sampling, all samples were transported to the laboratory and stored in dark conditions at 4°C.

For the incremental loading (IL) oedometer test and the fall cone test, sample NE2B-70% (NE site) was used as the “mother” sample (where the suffix indicates the percentage sand content). From this material, subsamples were prepared in the laboratory with 0, 10, 20, and 40% sand content with respect to the total dry mass (M _sand /M _dry;total ): samples NE2B-0%, NE2B-10%, NE2B-20, and NE2B-40%, respectively. These subsamples were obtained by separating the fines and sand fractions by sieving and recombining afterward. The value of the % sand contents in mass is almost identical to their correspondent % sand in volume because of the similar particle densities measured for the sand and fine fractions. The water content with respect to the fine fraction was kept constant for all subsamples. For the fall cone test, a sample from the deeper clayey silt material SW1B was also tested.

Additionally, a subsample from the 40% sand sample from the NE (sample NE2B-40%), tested with IL and fall cone tests, and sample SW1B, tested with the fall cone test, were oxidized to mimic the behavior of the Markermeer sediment after being in contact with atmospheric conditions for a long time. As natural oxidation is slow, the oxidation process was accelerated in the laboratory. This was done by treatment with hydrogen peroxide (H₂O₂), following an adapted procedure from the British Standards (British Standards Institute, 1990a) including two drying–rewetting cycles. First, the samples were dried in an oven at a constant temperature of 50°C until there was no further mass loss. Then, subsamples of 100 g were rewet by adding demineralized water and oxidized according to the standard. After centrifuging to remove the oxidant, the material was dried again at 50°C and rewet with filtered Markermeer water. pH was measured before and after oxidation by immersing a pH electrode in the samples, which showed that the treatment did not change the pH of the sediment. The labels of the treated samples are identified with the identifier T: samples SW1BT and NE2B-40%-T. The origin and composition of all the samples for the IL and fall cone tests are described in Table 1.

Atterberg limits and undrained shear strength were determined according to ISO (2016) and ISO/TS (2004b), respectively. For the liquid limit (LL), 80 g/30^° tip angle fall cone tests were performed. The plastic limit (PL) was determined using the rolling thread test. The total organic matter (TOM) content was obtained by loss on ignition (LOI), according to the European Standard (EN, 2012), and the amount of total organic carbon (TOC) and its reactivity were determined using the Rock Eval machine (Behar et al., 2001). The particle density was measured using a gas pycnometer (ISO/TS, 2004c), and the particle size was determined using a hydrometer and by dry sieving according to the British Standards Institute (1990b). Sample characteristics are presented in Table 1. In this table, sand particles are defined as those with size >63 μm, silt as the particles between 2 and 63 μm, and clay particles as those smaller than 2 μm.

The undrained shear strength c _u was obtained with the fall cone test following the ISO/TS (2004b) standard for each sample prepared with a range of different water contents.

IL tests were undertaken according to ISO/TS (2004a) to determine the compression and consolidation properties. With an IL test, a cylindrical soil sample is enclosed in a stiff steel ring in between two porous stones and incrementally loaded through the upper porous stone. The typical test interpretation is based on the Terzaghi consolidation theory (Terzaghi, 1923), which assumes small deformations. For a comparison with the more general large strain consolidation theory (Gibson et al., 1967), check the supplementary materials. As the initial water contents of the thick dredged sediment samples studied were higher than soil samples traditionally studied with this device, the loading protocol was adapted according to the initial water content of the sample to ensure small deformations to enable typical test interpretation to be undertaken. For most of the samples, the initial load was 0.6 kPa (Table 3). The IL experiments were performed in a climate-controlled room with a constant temperature of 10°C. The bulk and dry densities of the test specimens were determined by weighing before the test and after drying for 24 h in an oven at 105°C.

The samples tested had 0, 10, 20, 40, and 70% sand, and the tests are thus referred to as I-NE2B-0, I-NE2B-10, I-NE2B-20, I-NE2B-40, and III-NE2B-70%, where the first identifier (I, II, or III) indicates differences in the initial ratio of the water content and LL. Two additional IL tests were performed to investigate the different compression behaviors of the treated (oxidized) samples with 40% sand content with different initial water contents: test II-NE2B-40%-T and test III-NE2B-40%-T. Test II-NE2B-40%-T was started at a water content of 1.6 times the LL, and test III-NE2B-40%-T at 1.2 LL. This corresponds with a water content of the fines fraction of 0.9 and 0.7 [M _w /M _s,fines ], where M _s,fines is the dry mass of the fraction <63 μm (Table 2). The 1.6 LL water content test was selected because it was the maximum water content of the sample below which no self-weight consolidation occurred (Barciela–Rial, 2019). The reduction in the water content until 1.6 LL (sample II-NE2B-40%-T) or below (sample III-NE2B-40%-T) was done by atmospheric drying.

For the second additional test (III-NE2B-40%-T) and for the 70% sand content test (III-NE2B-70%), a different loading program with higher initial stresses was applied. This was necessary because of the higher stiffness of these samples. The three different loading plans (for the different initial fines water contents I, II, and III) are described in Table 3.

Figure 2 shows the undrained shear strength (c _u ) results. Herein, c _u is plotted against the non-corrected ϕ s (i.e., ϕ s f + ϕ s s a ) and ϕ mod . They follow a linear relationship on the log–log scale (i.e., power law). For each sample, a decrease in undrained shear strength with a larger water content (lower ϕ s or ϕ mod ) is observed. Plotting against ϕ s does not allow to identifying the TFC nor determining patterns of similar behavior. However, plotting against ϕ mod shows that the samples containing 0, 10, 20, and 40% sand overlap because they have the same OM composition (same mother sample). Thus, the TFC has not yet been reached at 40% sand. This is not the case for the sample with 70% sand content, implying that a TFC somewhere between 40 and 70% sand content. Therefore, the mechanical behavior of sample NE2B-70% is governed by a granular skeleton. Other authors found TFCs at similar sand content. For example, Thevanayagam (1998) found that the clay matrix governed the undrained shear strength behavior below 70% (mass) sand content, while Kumar and Wood (1999) found a TFC between 30 and 40% (mass) fines contents.

Sample SW1B exhibits lower initial ϕ mod than the samples originating from the other sampling sites because of its larger amount of OM and subsequent larger ability to bind water (different sampling sites). This sample is therefore located at the left in Figure 2. On the other hand, treated samples SW1B-T and NEB-40%-T moved to the right because of their loss in ability to bind water given the lower lability (reactivity) of the remaining OM, reflected in a lower PI. It is remarkable that the slopes of the curves remain almost identical, suggesting a similar internal structure of the clay matrix.

Figure 3 shows the traditional log(effective stress)—void ratio (e) space and shows the effective stress against the corrected volumetric concentration ϕ mod obtained from the IL tests. All the non-treated samples with sand contents up to 40% (i.e., tests I-NE2B-0%, I-NE2B-10%, I-NE2B-20%, and I-NE2B-40%) show virtually the same power law relation with overlapping lines in the log–log plot (right panel). This indicates that the TFC occurs for sand contents above 40% but below 70%. This agrees with the findings for the undrained shear strength.

The treated samples (tests II-NE2B-40%-T, III-NE2B-40%-T) also follow a power law relationship. The σ _eff–ϕ _mod correlations from these treated samples have a steeper slope, implying a larger increase in effective stress for an increase in ϕ mod with respect to the non-treated samples (i.e., larger n _f , see Eq. 2 and the Supplementary Material). This indicates that the floc structure became denser as a result of the treatment, and primary particles may touch each other. This is in accordance with the findings of Barciela–Rial (2019), who found the same change of behavior after oxidation treatment for samples tested upon the constant rate of strain (CRS) consolidation tests for other Markermeer sediment samples.

The σ _eff–ϕ _mod correlation of sample of test III-NE2B-70% does not overlap with the other non-treated samples. This is because the fines content is below the TFC, and therefore, a granular skeleton is present. Remarkably, it still follows a power law, while, for instance, Schultze and Moussa (1961) presented exponential relationships for sand compressibility.

The compression index C _c calculated according to the geotechnical standard (i.e., applying Eq. 4 with e, instead of e mod ) is presented in Figure 4A,C as a function of the effective stress and corrected solid content ϕ mod . Overall, C _c decreases with both increases in effective stress and solid contents but becomes nearly constant during the last phases of the tests. C _c appeared to be invariant with sand contents up to 10%. Possibly the sand particles are too diluted in the fines matrix to affect the compression, although an experimental error may have affected the results as well. For sand contents of 20% or more, C _c decreases with sand content. This is caused by the incompressibility of sand particles as has frequently been reported in the literature (e.g., Monkul and Ozden, 2007; Watabe et al., 2011; Simpson and Evans, 2016). However, when correcting C _c for the amount of sand (C _c,mod calculated by using e mod in Eq. 4; Figure 4B,D), the C _c,mod for samples with 0–40% sand overlap, indicating that the TFC has not yet been reached with sand contents up to 40% (see Figure 4).

Regarding the effect of the oxidation treatment, the treated samples have smaller C _c values because of a lower ability to bind water (Barciela-Rial et al., 2020); thus, less pore water is expelled upon loading. The C _c –effective stress plot also shows that when using the void ratio e, instead of the corrected void ratio e mod to calculate C _c (Figure 4A), the compressibility behavior of the treated sample with 40% sand (III-NE2B-40%-T) and the natural sample with 70% sand (III-NE2B-70%) is the same for loading scheme III (starting at higher initial stress). For the test with the treated sample at higher initial water contents (II-NE2B-40%T), the compressibility values were slightly larger.

The treated sample III-NE2B-40%-T and the sandy sample III-NE2B-70% exhibit a different tendency; the compression index increases with the effective stress. This is likely to be because both samples behave as silty sand. The increase in compressibility with effective stress for sandy sediment has been shown by multiple authors. For instance, Mesri and Vardhanabhuti (2009) showed an increase in C _c with effective stress for dense quartz sands with a maximum C _c value below 0.01 for stresses of 100 kPa. It is therefore concluded that the oxidation treatment for the material with 40% sand content changed the TFC.

Figure 4E shows C _c,mod as obtained from Eq. 6 with the fitting parameters determined from results presented in Figure 3. For sand contents of 40% or smaller, the C _c,mod curves overlap. The values obtained also agree with the ones calculated with e mod and Eq. 4, that is, Figure 4D. This implies that Eq. 6 may be used to determine C _c,mod as a function of ϕ mod , directly with the material parameters determined from standard settling experiments (e.g., Merckelbach and Kranenburg, 2004b).

Figure 5 presents a schematic of the mechanical behavior of the cohesive sediment with organic material. Herein, a possible collapse of flocs upon loading is not addressed. In this schematic, the different behavior found for Markermeer samples are classified into four (sub)categories: (1.1) fines dominated with OM (no sand), (1.2) fines dominated with OM and some sand (passive sand particles), 2) fines dominated with oxidized OM and some sand (passive sand particles), 3) sand dominated with OM and some fines (passive fines), and 4) sand dominated with oxidized OM and some fines (passive fines). The transition from case 1.2 to case 3 and that from case 2 to case 4 determine the TFC.

For sand contents of 40% and lower, the behavior of the Markermeer sediment is dominated by the fines. However, for the sediment sample with 70% sand, the sand particles form a skeleton which dominates its behavior. The same sand-dominated behavior is observed for the treated samples. This means that the material parameters have changed. However, the constitutive model still appears to be applicable for these samples as the σ _eff – ϕ mod curves still follow the power law (e.g., Figure 3). The fact that oxidation of the OM changed the behavior of a sample with 40% sand from fines-dominated to sand-dominated suggests that the physical particle interactions are affected by the presence of OM and by the type of the organic matter present (fresh organic matter or oxidized organic matter). Thus, the type of OM is also relevant, as discussed by Barciela–Rial et al. (2020) who showed that the mechanical behavior upon drying of the Markermeer sediment is correlated stronger with the reactivity of OM than with the total organic matter content.

The activity plot (Skempton, 1953) reflects the relation between the plasticity index (PI) as a function of the clay content, yielding the sediment’s activity (the slope of the curve) and the critical clay content above which the sediment depicts cohesive behavior. Active sediments undergo considerable changes of volume when wetted (swelling) or dried (shrinkage), for example, Mitchell and Soga (2005). Figure 6 presents the activity plot for the samples used for the IL-tests with an activity A = 1.8 and critical clay content of 6%. The removal of the OM (treated sample) shifts the activity plot to the right, at the same slope (activity). However, the critical clay content increases substantially to 17%, which is in agreement with the aforementioned analyses on the impact of treatment on the mechanical behavior of the sediment.

The sediment studied in the present work has much larger initial water content than usually studied in classical soil mechanics yet lower than those at which the Merckelbach and Kranenburg model has traditionally been applied. The results showed that at these relative low water contents, the constitutive equations of Merckelbach and Kranenburg work well. The mechanical response to external stresses can be described with these equations. This is in line with the findings of, for example, Chassagne (2019); Barciela–Rial (2019), who showed that the fractal theory could reproduce the evolution in time of consolidation for natural sediment mixtures also at large initial concentrations of sediment. A higher fractal dimension applies than traditionally available in the literature (between 1.8 and 2.2 for flocs, e.g., Merckelbach, 2000), which makes sense as a higher fractal dimension is an indication of a more compacted clay. However, the fact that the samples obey power law relationships does not prove that they are fractal, and further research would be needed to quantify this, which is out of the scope of this study.

When building with the dredged sediment, the initial water content and sand content may vary considerably, and which will also affect the consolidation behavior. The practical question is how to cover various sand or water contents with as few tests as possible. The present study shows that trends in behavior can be identified from σ _eff–ϕ _mod and c _u –ϕ _mod relations. At sand contents below the TFC, the effect of sand can be accounted for through a correction of the sediment volumetric concentration, and all σ _eff–ϕ _mod and c _u –ϕ _mod curves collapse (overlap). However, the behavior of the treated samples NE2B-40%-T and SW1B-T did not overlap with the natural samples and exhibited a steeper increase with the volumetric fraction. This is consistent with the increase in the critical clay content after treatment, as discussed before.