Introduction

Tidal estuaries are complex systems often including temporally and spatially variable processes of sediment dynamics and morphodynamics (Dyer 1995; Green et al. 1997; Friedrichs et al. 1998). With climate change, associated sea level rise, and expected increase in storm intensity and frequency, the preservation, sustainability and resilience of coastal ecosystems, including tidal estuaries, is threatened (Wolanski and Chappell 1996; Morris et al. 2002; Orton et al. 2012). Coastal ecosystems are often further stressed by human impacts through recreational and industrial usage (Lesourd et al. 2001; Kirwan and Megonigal 2013). Detailed investigation of surficial sediment characteristics is crucial for evaluating sediment dynamics, monitoring the health of an ecosystem, and planning coastal engineering actions. Geotechnical sediment characteristics can provide important information for such investigations. In situ sediment strength, sediment density, and pore pressure behavior are directly associated with sediment erodibility, and consolidation when being deposited (e.g., Lintern et al. 2005; Foster et al. 2006; Grabowski et al. 2011; Stark et al. 2011). However, few geotechnical field datasets are available from tidal estuaries to date, and the ones available often lack spatial coverage and/or a detailed resolution of the seabed surface (Missiaen et al. 2008; Stark and Wever 2009; Carling et al. 2009). Therefore, a still unsolved research question is how do geotechnical properties, such as bulk density, sediment strength and pore pressure behavior, correspond to local sediment grain size distribution, transport history, hydrodynamic forcing conditions, and morphology in a tidal estuary?

Two working hypotheses were pursued in this field study. First, estimates of sediment strength from a portable free fall penetrometer (Stegmann et al. 2006; Stoll et al. 2007; Seifert et al. 2008; Stark and Wever 2009; Mulukutla et al. 2011; Stark et al. 2011, 2014; Stephan et al. 2012; Morton et al. 2014) reflect the available sediment types, and local morphology in tidal channels. Second, dilative or contractive soil behavior can be inferred from pore pressure measurements of the same device. The following specific research objectives were addressed in this field study: (1) correlation of sediment composition to the in situ sediment strength of bed surface sediments; (2) assessment of the impact of bathymetric features (bottom slope) on sediment strength; (3) correlation of the pore pressure behavior during and after high velocity (~4.75 m/s) impact to sediment type and soil mechanical behavior.

A field study using the BlueDrop penetrometer was conducted on 27 September 2014 in the Great Bay Estuary, New Hampshire, USA (Fig. 1), which includes 227 penetrometer deployments along four transects (Fig. 1) with varying water depths and bathymetric features. The penetrometer dataset is complemented by sediment samples and cores from the same areas.

Fig. 1
figure 1

Survey locations in the Little Bay and Great Bay regions of the Great Bay Estuary. Transect 1 starts in Broad Cove and extends north into the estuary channel. Transect 2 (subdivided into transects 2-1, 2-2 and 2-3) covers a large area of shallow water depth from the west end of the Scammel Bridge to north of transect 1. Transect 3 traverses Little Bay in an east–west direction north of Adams Point, and transect 4 crosses the opening to Great Bay in a northeast to southwest orientation

Physical setting

The Great Bay Estuary, located in southeastern New Hampshire, USA, is comprised of Little Bay, Great Bay, and the tidal sectors of five rivers: the Bellamy, Oyster, Lamprey, Squamscott, and Winnicut. The estuary is connected to the Gulf of Maine via the Lower Piscataqua River, which connects to Little Bay near Dover Point under the General Sullivan Bridge (Fig. 1). Like much of New England, the area has been heavily influenced by former glacial processes (Goldthwait et al. 1951). Due to the estuary’s proximity to the coast, it has been subjected to reversing cycles of marine transgression and regression as the continental crust and sea level responded to the growing or shrinking polar ice sheets (Ward 1995). Although the timing of events is not fully understood, a river valley was created by a combination of ice plucking, subglacial erosion, and fluvial processes following glacier retreat. The sea level gradually rose after the last ice age, flooding the valley, and resulting in the current water level ca. 2,000 years ago (Goldthwait et al. 1951; Ward 1995).

The glacial history has resulted in a highly variable bathymetry characterized by steep channels surrounded by shallow mud flats. Channel depths in Little Bay and Great Bay range from less than 3 m to ~24 m (Bilgili et al. 2005). The hydraulic conditions are tidal dominated as salt water is delivered into the estuary by the Lower Piscataqua River exchanges approximately 40% of the volume of Great Bay (Ertürk et al. 2002; Bilgili et al. 2005). In comparison, fresh water input from the five rivers constitutes less than 2% of the tidal prism (Swift and Brown 1983; Ertürk et al. 2002; Bilgili et al. 2005). The tides are dominated by the semidiurnal M2 constituent, and typically range from 2 to 4 m between spring and neap cycles, exposing up to 50% of the mud flats at low tide (Bilgili et al. 2005). The tide to depth ratio is approx. 0.18, and maximum current velocities in the main Little Bay channel are greater than 2 m/s, but can be locally stronger or weaker depending upon cross-sectional area (Swift and Brown 1983; Ertürk et al. 2002; Bilgili et al. 2005). In general, the estuary can be considered a shallow, tidal dominated, well-mixed estuary with sub-critical flow regime (Swift and Brown 1983; Bilgili et al. 2005; Wengrove et al. 2015).

The sediments in the estuary owe their origins to the glacial history of the region, and range from coarse sands to fine-grained mud. A detailed sediment map for the region can be found in Wengrove et al. (2015). Coarse sediments are derived from glacial till and glacial outwash, while fine sediments come from the Presumpscot Formation, a cohesive deposit of silty clay with variable amounts of sand formed from rock or glacial flour being deposited in a marine environment at the edge of the receding glaciers (Ward 1995). In general, the surficial sediments are related to water depth with fine-grained sediments located in shallow water, and coarse sediments in deeper water. More than 50% of surficial sediments contain more than 50% mud based on the Shepard classification system (Shepard 1954; Wengrove et al. 2015). The depth dependency of sediments can be related to the hydrodynamics where stronger flows through the channels have eroded and/or prevented fine-grained material from accumulating. Deeper sections of the channels have little unconsolidated material, and are mostly exposed bedrock (Ward 1995; Wengrove et al. 2015).

Materials and methods

BlueDrop free fall penetrometer

The BlueDrop penetrometer follows the conceptual idea of the Nimrod penetrometer (Stark et al. 2011). Both probes are designed to be approx. fluid-dynamically shaped, and to be suitable for deployment in areas of energetic hydrodynamics and active sediment dynamics (Stark et al. 2011, 2014). The probe falls vertically through the water column and penetrates the uppermost layers of the seafloor to distances ranging from 5–300 cm, depending on the sediment type. The BlueDrop is 63.1 cm long with a diameter of 8.75 cm and mass of 7.7 kg when equipped with a conical (~60°) tip.

The BlueDrop penetrometer houses five vertical microelectromechanical systems (MEMS) accelerometers of different measurement range and accuracy (measurement ranges from ~1.7 to 250g, with g being gravitational acceleration), a dual-axis MEMS accelerometer (±55g) to measure inclination, a pressure sensor located behind the tip (defined as u2 location in standard Cone Penetration Testing; Lunne 2012), a data logger sampling at 2 kHz, and a power supply. The characteristics of the impacted sediments are deduced from analysis of the deceleration and pressure profiles over the penetration depth. The point of impact, and thus, the water–seabed interface, was defined as the point of time when the penetrometer is decelerated from its free fall velocity after accounting for deceleration due to the increase of rope drag (Stoll 2004; Aubeny and Shi 2006; Stark and Wever 2009; Stephan et al. 2012). The accuracy was maximized by using the MEMS accelerometer with the highest resolution (range ~1.7g) for identification of the point of impact (Dorvinen et al. 2015). Penetrometer velocity and penetration depth are derived by single and double integration of the impact induced deceleration, respectively. The data logging rate of 2 kHz and impact velocities up to ~6 m/s result in a vertical resolution of better than 1 cm.

Estimating an equivalent of quasi-static bearing capacity (qsbc)

The extent of penetration and deceleration experienced by an object penetrating into soil are functions of the sediment resistance force, F sr, allowing the assumptions that the penetrometer is falling freely, and that the increase of rope drag over the length of the sediment penetration is negligible due to the limited penetration depth. Following that, F sr can be calculated using the deceleration, dec, measured during impact and penetration into sediment:

$$ {F}_{\mathrm{sr}}={m}_{\mathrm{b}}\ast \mathrm{d}\mathrm{e}\mathrm{c} $$
(1)

where m b is the mass of the penetrometer in seawater (Stoll 2004; Aubeny and Shi 2006; Stoll et al. 2007; Stark et al. 2011; Stephan et al. 2012, 2015). Different mechanisms contribute to the sediment resistance force: the shearing resistance force of the sediment F S, the soil buoyancy F B, and the skin friction F SF (Aubeny and Shi 2006; Chow and Airey 2013; Morton et al. 2014, 2016). Assessments of soil buoyancy for any given probe, penetration depths, and soils ranging from coarse sand to soft mud lead to a potential contribution of 1–15%. A contribution of 15% would occur if the penetrometer would reach a penetration depth of 100 cm into dense clay, a more theoretical scenario than realistic. Thus, a contribution of soil buoyancy of only 1–5% is expected. The contribution of skin friction is also negligible based on the short penetrometer length with maximum diameter, and on results by DeJong et al. (2000) and Frost and DeJong (2005). After neglecting inertia, F B and F SF, it can be assumed that F sr~F S, and the equivalent ultimate dynamic sediment bearing capacity, q ud, can then be estimated using the sediment resisting force:

$$ {q}_{\mathrm{ud}}=\frac{F_{\mathrm{sr}}}{A} $$
(2)

where A is the area of the layer subjected to loading.

Sediment resistance force and bearing capacity depend on shearing velocity, here being expressed by the penetration velocity, v. The initial impact velocity varies with water depth dependent on the required free fall depth to achieve terminal velocity and rope drag. During penetration into the seabed, v decreases from the initial impact velocity to zero. Therefore, a strain rate correction must be performed (Dayal et al. 1975; Steiner et al. 2013). Based on the encouraging results when applied to the correction of high velocity impact penetrometers, the approach proposed by Dayal et al. (1975) is used herein as with many other studies (Stegmann et al. 2006; Stoll et al. 2007; Stark et al. 2012a; Steiner et al. 2013; Stephan et al. 2015), in which the strain rate correction factor, f, is calculated as

$$ f=1+ K \log \left(\frac{v}{v_0}\right) $$
(3)

where K is a dimensionless factor assumed in the range 1 to 1.5 following results for high speed impact velocities (Stoll et al. 2007; Stark et al. 2011, 2012a; Stephan et al. 2015), and the industry standard penetration rate of v 0=0.02 m/s is used for the reference velocity (Lunne et al. 1997). The strain rate corrected bearing capacity, qsbc, simulating a quasi-static penetration (penetration at 0.02 m/s), can then be estimated by

$$ \mathrm{qsbc}=\frac{q_{\mathrm{ud}}}{f} $$
(4)

It should be noted that the optimization of the strain rate correction for high velocity impact penetrometers of varying shapes is still the subject of ongoing research.

Pore pressure measurements

Pore pressures in a sediment during undrained loading increase for a contractive sediment such as normally consolidated clays or loose sands, and decrease for dilative sediments such as over-consolidated clays or dense sands (e.g., Lambe and Whitman 1969). Furthermore, pore pressures in sandy sediments, regardless of density, can be expected to equilibrate to hydrostatic pressures shortly after loading has ceased, while clayey sediments have a much slower response due to their very different hydraulic conductivities (e.g., Lambe and Whitman 1969). Pore pressure responses to dynamic penetration tests have been investigated by numerous authors and are discussed below, including the possibility of partial drainage (Lee and Elsworth 2004; Seifert et al. 2008; Houlsby and Cassidy 2011). The pore pressure response during high impact velocity penetration and for the penetrometer at rest are investigated here.

Sediment sampling and analysis

Surficial sediment samples (78 total) were obtained from a Ponar grab sampler deployed from a small boat. The device collects sediment to a depth of about 10–15 cm. Sediments obtained along transects 1 to 3 were analyzed for grain size distribution with a Malvern Mastersizer laser particle size analyzer capable of detecting particle sizes from 0.01 μm to 2 mm at a constant logarithmic interval of 0.1354. Size fractions were classified according to Shepard (1954), where greater than 75% sand (>62 μm) is denoted as sand, between 50% and 75% sand as muddy sand, between 25% and 50% sand as sandy mud, and less than 25% sand as mud. The median (d 50) grain diameter was obtained for each sample.

In October 2014, 12 sediment cores were obtained by scuba divers along transect 4. Scuba divers carefully inserted and extracted 10.8 cm diameter plastic core barrels from the sediment at prescribed locations. The top 2 cm of each core was extracted with a hydraulic press, corresponding to a sample volume of 183.22 cm3 with estimated error of 5% (based on ±1 mm height error in extracting the volume), and stored in plastic bags in a refrigerated room at 4 °C. A 2 cm3 subsample was removed from each sample to be used for sediment grain size analysis. The core top samples were oven-dried at 60–100 °C for at least 24 h, cooled to room temperature in a desiccator, and reweighed. The difference in mass of the samples before and after drying was assumed to be entirely due to the removal of water. A correction was made to the mass of the sediment to account for the mass of salts in the pore water that remains in samples after drying (Dadey et al. 1992). The water temperature and salinity at the time of sample collection was used to estimate a water density ρ w=1.02 g/cm3. The fractional volume of water to sediment is the porosity, n, determined from ρ w (to estimate water volume) and the density of quartz (ρ g=2.65 g/cm3; to estimate sediment volume). Assuming the pore fluid density is the same as ρ w, dry bulk density, ρ d, can be calculated:

$$ {\rho}_{\mathrm{d}}={\rho}_{\mathrm{w}}\left(1- n\right) $$
(5)

This method does not depend on the sample volume, an unfavorably rough measurement, and instead relies on porosity determined from the fractional loss of water, a relative and more precise value than volume (Dadey et al. 1992).

Water depth during penetrometer surveying

Water depths at the time of penetrometer deployments were estimated using an onboard Garmin acoustic echo sounder with 0.1 m digital output resolution. The location of the 200 kHz transducer was about 20 cm below the waterline on the rear transom of the vessel, approx. 2–3 m horizontally from the position where the drop was made. Acoustic derived depths were determined assuming a sound speed of 1,500 m/s, this being a good approximation for the seawater temperature (15 °C) and salinity (30 psu) during the deployment. The vessel (a 14 foot Carolina Skiff) bobbed vertically in the water in response to (small) waves and personnel moving on the vessel. The bottom is determined by threshold detection algorithms internal to the sonar. The best estimate is that the sonar-derived water depths are slightly (approx. 10–20 cm) below the true value owing primarily to the transducer location and precise waterline in a moving vessel.

Results

The BlueDrop penetrometer was deployed a total of 227 times at 116 distinct locations along the four transects (Fig. 1). Transect 2 was divided into the three sub-transects 2-1, 2-2, and 2-3. At most locations, the penetrometer was deployed twice to test reproducibility of the results. Provided the penetration profiles were similar, average values were used for the classification of sediment strength at each location (83 of 116 locations, 71.6%). For duplicate deployments with dissimilar results, the results have been reported separately (11 of 116 locations, 9.5%). As there is some spatial separation between duplicate deployments (usually in the order of single meters depending on vessel drift and water depth), this can be an indicator for spatial inhomogeneity of the seabed surface. Twenty deployments (out of 227, 8.8%) were not included in any of the analyses due to impact on rock or other disturbances in the data (e.g., tangled rope).

Sediment strength and composition

The results from a single deployment showing the deceleration, velocity, and estimated qsbc profile through penetration are shown in Fig. 2. The maximum average value of qsbc reached over the penetration, qsbcmax, was determined for each deployment. For example, qsbcmax was 9.8±1.4 kPa at a sediment depth of 43 cm during deployment 3 of transect 1 (Fig. 2).

Fig. 2
figure 2

Deceleration, velocity, and estimated qsbc for deployment 3 on transect 1. qsbc data are a possible range (blue shading) following K=1 to 1.5. qsbc at penetration depth <4 cm is biased by A→0, and is removed from analysis here. The decrease of qsbc with the halt of the penetrometer is a result from using the deceleration for the calculation. Thus, values of qsbc associated with ν→0 must be rejected. Here, this accounts for penetration depths >45 cm. The maximum quasi-static bearing capacity encountered at this site for a penetration depth of 45 cm qsbcmax is 9.8±1.4 kPa

The sediments along transect 1 (Fig. 3) were classified as muddy sand and sandy mud up to approx. 100 m along the transect from location 1, and the qsbcmax values did not exceed 11 kPa. With an increase in water depth (≥3 m), the sediments were classified as sand, and the values of qsbcmax increased slightly to 11–34 kPa. In the area of transects 2-1 to 2-3, the sediment classified as predominately sand with some muddy sand beyond 1,000 m west of location 24 on transect 2-3. For the short transects 2-1 and 2-2, sand content varied little with 85–95% and 90–100%, respectively. The associated median grain sizes ranged around 200 and 200–350 μm, respectively. Sand content and median grain sizes along transect 2-3 are presented in Fig. 4. The average qsbcmax values in this area (26.3 kPa) were higher than along transect 1 (12.4 kPa), particularly between 600 and 1,000 m along transect 2-3 (45.4 kPa). Only eight locations along transect 2-1 were characterized by soft sediments with qsbcmax <20 kPa. The sediment varied along transect 3 (Fig. 5) from sandy mud (distance 0–120 m) to muddy sand (distance 120–260 m) to sand (distance 260–1,080 m), and back to muddy sand (distance 1,080–1,400 m) and sandy mud (distance 1,400–1,600 m). These variations are reflected in the qsbcmax (Fig. 5). Specifically, a peak qsbcmax of 55.4 kPa was associated with the location of the peak sand content (100%), and values of qsbcmax <20 kPa were consistently associated to sand contents of less than 75%.

Fig. 3
figure 3

Transect 1: bathymetric map including deployment location numbers (left), qsbcmax and pore pressure response type (upper right), as well as sand content and median grain size from sediment samples (central and lower right). Distance is measured along the transect from location 1 offshore

Fig. 4
figure 4

Transect 2: bathymetric map including deployment location numbers, qsbcmax and pore pressure response type. Distance is measured along each transect from the respective starting points: location 1 for transect 2-1, location 13 for transect 2-2, and location 24 for transect 2-3. Sand content and median grain size are also presented for transect 2-3

Fig. 5
figure 5

Transect 3: bathymetric map including deployment location numbers (upper panel), qsbcmax, pore pressure response type, as well as sand content and median grain size from sediment samples. Distance is measured along the transect from location 1

The results indicate a relationship between qsbcmax and sediment composition. Using measurements where the BlueDrop deployment and sediment sample locations were within a distance of 20 m, sediments with a sand content greater than 40% and smaller than 70% yielded a qsbcmax of ~10 kPa (Fig. 6). Samples with a sand content of more than 85% showed significantly more scatter with qsbcmax ranging from 10–55 kPa. Comparing qsbcmax to the median grain size d 50, a power function qsbc ≈ 0.99d 50 0.57 can be derived, yielding a coefficient of determination R 2=0.59 with most scatter occurring for d 50 >150 μm (Fig. 7).

Fig. 6
figure 6

qsbc vs. sand content of deployments and samples with a distance of less than 20 m from each other

Fig. 7
figure 7

qsbc vs. d 50 of deployments and samples with a distance of less than 20 m from each other

The sediment cores obtained along transect 4 were analyzed for bulk density characteristics. An elevated qsbcmax was measured at a transect distance of 1,200–1,350 m where a higher dry bulk density of the sediment was found (Fig. 8). Sediment samples collected between 520 and 630 m were denser than 1.0 g/cm3, and were located in a small channel (Fig. 8) generally associated with sandier sediments. Thus, a higher dry bulk density may be expected, but it is unclear why the values of qsbcmax did not reflect this. No sediment samples were obtained at a transect distance of 0–535 and 950–1,200 m. A direct comparison between the dry bulk density and the values of qsbcmax along transect 4 is presented in Fig. 9, considering only locations with a distance of less than 20 m from each other.

Fig. 8
figure 8

Transect 4: bathymetric map including deployment location numbers (upper panel), qsbcmax, pore pressure response type, as well as dry bulk density from sediment samples. Distance is measured along the transect from location 1

Fig. 9
figure 9

qsbc vs. dry bulk density of deployments and samples along transect 4 with a distance of less than 20 m from each other

Recorded pressure signatures

Three distinct profiles of pressure recordings after impact were identified: types A, B and C. Exemplary results are presented in Figs. 10, 11 and 12, respectively. For the type A deployment, an initial increase in pressure during penetration dissipated quickly after the penetrometer came to a stop. While resting in the sediment, a very gradual increase in pressure was observed (Fig. 10). The qsbc estimates and deceleration identified a top layer of very low sediment strength (sediment depth 0–12 cm; qsbc <0.5 kPa) over another weak sediment layer (sediment depth 12–30 cm; qsbc ~2 kPa), before penetrating into a stiffer substratum (qsbc ~10 kPa; Fig. 10). In the first 22 cm of penetration through the two weak top layers, the recorded pressures p were smaller than the hydrostatic pressure projection p h (Fig. 10). As the penetrometer moved into the stiffer stratum, a clear increase of p to higher than p h was observed. The development of p<p h in the upper 10–20 cm of the sediment was a common characteristic of type A sediments. Calculating the ambient hydrostatic pressure from the penetrometer measurement upon contact with sediment with consideration of the Bernoulli effect yielded p b=38.3 kPa, representing only a difference of ~28 cm in water depth between the penetrometer measurement and the echo sounder (an error accounted for by uncertainty in the water depth derived from the sonar). It should be emphasized that the measured pressure did not increase rapidly and immediately with a decrease in velocity associated with the probe coming to rest (Fig. 10), as the Bernoulli effect would suggest. In the following, p s denotes the pressure corresponding to the static water depth determined from the onboard echo sounder.

Fig. 10
figure 10

Type A pressure response (transect 1, location 12) showing recorded pressure p versus time and hydrostatic pressure as measured by the vessel-mounted acoustic echo sounder p s (upper panel), deceleration, velocity and qsbc versus penetration depth (lower left panel), and p and the projected increase due to hydrostatic pressure p h versus depth. Blue shading Measurements in water. Dark brown vertical line Time of initial impact. Brown shading Measurements in sediment. Black vertical line in upper panel Time at which BlueDrop came to rest. Blue arrow and text box Hydrostatic pressure based on the penetrometer measurement when considering pressure correction regarding velocity based on the Bernoulli effect p b at first contact with sediment

Fig. 11
figure 11

Type B pressure response (transect 2, location 34) showing recorded pressure p versus time, hydrostatic pressure as measured by the vessel-mounted acoustic echo sounder p s and pressure at impact measured by the penetrometer considering the Bernoulli effect p b (upper panel), deceleration, velocity and qsbc versus penetration depth (lower left panel), and p and the calculated increase due to hydrostatic pressure p h versus depth, following the format of Fig. 8

Fig. 12
figure 12

Type C pressure response (transect 3, location 9) showing recorded pressure p versus time, hydrostatic pressure as measured by the vessel-mounted acoustic echo sounder p s and pressure at impact measured by the penetrometer considering the Bernoulli effect p b (upper panel), deceleration, velocity and qsbc versus penetration depth (lower left panel), and p and the calculated increase due to hydrostatic pressure p h versus depth, following the format of Fig. 8

Type B profiles are characterized by a consistent and more rapid increase in pressure toward p s and p b after the penetrometer stopped (Fig. 11). An initial decrease in pressure was recorded upon embedment of the pressure sensor inlets in the exemplary profile of type B. However, some type B profiles display increasing pressures upon impact. The qsbc estimates indicate low initial sediment strength (qsbc <2 kPa) in the first ~12 cm of penetration overlying a stiffer intermediate stratum (sediment depth 12–25 cm; qsbc ~4 kPa), before an even stiffer stratum was encountered (sediment depth 25–40 cm; qsbc ~17 kPa; Fig. 11). Recorded pore pressure increased significantly beyond p h upon penetration into the intermediate and stiffer layers (Fig. 11). The difference between p s and p b represented a difference of ~44 cm in water depth, a value slightly larger than the potential uncertainty of the echo sounder (~20 cm), and may be related to small-scale bathymetric features.

In the case of type C sediments (Fig. 12), the pressure increased to an approx. constant value shortly (~0.1 s) after the penetrometer came to rest. This value was slightly larger than p s and slightly smaller than p b. The qsbc estimates indicated low initial sediment strength (qsbc <1 kPa) in the first ~6 cm of penetration followed by an abrupt increase to ~55 kPa at a depth of ~11 cm (Fig. 12). Upon sensor embedment, an immediate increase in pressure was recorded (>p h), followed by a rapid drop (<p h) when the penetrometer came to a stop, before the pressure increased rapidly during rest to approx. p S<p<p b (Fig. 12). The immediate response of the recorded pressures upon initial penetration, and upon stopping, was a characteristic of type C sediments.

Grouping the results was based on the criteria of the pressure profiles correlated to sediment strength characteristics (Table 1). In general, water depth, maximum deceleration, and qsbcmax increased from type A to type C, while the penetration depths decreased accordingly. Some deployments (37.9%) showed a clear delay in the pressure response. Such behavior can be explained with a clogged filter ring that is protecting the pressure transducer from direct soil pressure, and contamination of the canals, or more likely, is associated with air bubbles in the filter ring. Such data are not included in Table 1. The water depths reported in Table 1, and discussed in the remainder of this paper, were determined from the vessel-mounted acoustic echo sounder. However, using the pressure measured by the penetrometer upon impact and correcting for the Bernoulli effect led to good agreement (differences less than ~0.25 m) between the water depth estimated from the penetrometer measurement and the echo sounder.

Table 1 Comparison of deployments based on response type

Discussion

Sediment strength (cf. qsbc) and composition, morphology

With a higher sand content, an increase in friction angle is expected, leading to greater sediment strength (Terzaghi 1951). The increase of portable free fall penetrometer deceleration with a transition in sediment composition from fine-grained sediments to sand has also been reported by Stoll et al. (2007), Stark and Wever (2009), and Mulukutla et al. (2011) who classified sediments directly based on the deceleration. Many factors may contribute to the observed scatter within the sand samples (Fig. 6), including the separation distance, possible temporal changes of the sediment between BlueDrop deployments and sediment sample retrieval, different sand grain sizes, differences between the depth of the sample and sediment depth where the qsbc reached a maximum, a sloped seabed surface, and effects of the failure mechanism. The influence of gradation, seabed slope, and failure mechanism are discussed below. It is expected that the correlation would improve with more temporally and spatially co-located samples, suggesting active sediment transport and redistribution processes.

Stark et al. (2012a, 2012b) documented that variations in sediment strength in sandy sediment can directly be associated with differences in grain size (ranging from coarse to fine sand), as well as in density. Figures 6, 7 and 9 suggest that grain size may represent the governing property for the in situ sediment strength in fine-grained sediments ranging between approx. 50≤d 50≤100 μm. It should be mentioned that samples with d 50 ranging from 100 to 150 μm were not abundant. For sandy sediments, other properties seem to have a significant impact on qsbc in addition to the sand content and median grain size.

Impact into a noticeably inclined seabed surface can reduce the bearing capacity significantly, generally attributed to less sediment contributing to shear resistance downslope (Terzaghi 1951; Meyerhof 1953). However, the slopes in this study (mostly up to 5–10%) seemed to impact the results only in very few cases of very steep slopes. Transects 2-1 and 2-2 contained mostly sandy sediments, but also crossed a noticeable bathymetric gradient, approaching a channel trench from the shallow flats (Fig. 4). Deployment locations 10–14 are on the steepest sections of these transects; however, qsbc is only noticeably low for locations 10–11 (Fig. 4).

A change in failure mechanism with sediment depth can also affect the qsbc. In the case of a general bearing capacity failure, it is assumed that the failure surface extends to the ground surface, the sediment is incompressible, the sediment displaces as a rigid mass, and thus, the bearing capacity is a function of the soil strength (Coduto 2001). This type of behavior may be expected during initial penetration resulting from little overburden stress resisting the displacement of sediment. In contrast, a punching failure assumes the failure surfaces do not extend to the ground surface, the failure surfaces are poorly defined, and the bearing capacity is a function of both the soil strength and compressibility (Coduto 2001). This type of failure may be expected at deeper sediment depths, or in very loose sediment top layers. The depth and degree to which the different failure mechanisms occur is likely a function of the sediment characteristics (e.g., density, gradation, over-consolidation ratio) and penetration velocity, and is currently subject to research for high velocity impacts. In the present case, qsbcmax values represent an in situ assessment of the sediments’ ultimate load carrying ability, inclusive of the aforementioned factors such as inclination and compressibility. Any attempt to calculate shear strengths from qsbc presented here would require assumptions whose validity may be difficult (or impossible) to confirm in the field. No attempt was made to derive the sediment shear strength. Laboratory experiments in a large modified cone penetration test calibration chamber are currently in preparation to investigate this issue under controlled laboratory conditions.

Regarding the scatter in qsbcmax for the sandy sediments, differences in sand density that may be associated to most recent sediment relocation processes may indeed lead to variations in the in situ friction angle and shearing resistance, as well as differences in the failure mechanism. This is particularly evident when considering that penetration depths of 15–30 cm were often reached in the sandy sediment, indicating overall rather loose sands (Stark et al. 2012a). Additionally, Figs. 8 and 9 document a relation between bulk density and qsbcmax with an increase of qsbcmax with dry bulk density in accord with geotechnical theory.

High impact velocity pressure records for sediment characterization

Pore pressure measurements have been a standard procedure in cone penetration testing (CPTU) for many years (Lunne et al. 1997; Lunne 2012). More recently, they have also been performed using both rapidly winch lowered and free falling CPTU (Stegmann et al. 2006; Seifert et al. 2008). Stegmann et al. (2006) presented results obtained using an industrial CPTU tip at impact velocities ranging from 0.3–6 m/s, which showed significant changes in pore pressure response during penetration and after the probe came to rest. Those authors emphasized the potential of using pore pressure recordings of free fall devices for sediment characterization. However, many of the physical and soil mechanical processes leading to the widely varying pore pressure responses under high velocities are still unclear.

Measurement performance

In this study, a significant amount of pore pressure recordings (~38% of the deployments) were rejected from further analysis, as the pore pressure recording appeared delayed and damped. One likely explanation is that negative excess pore pressures may have developed while penetrating a desiccated layer of surficial sediments, causing desaturation of the porous filter ring (Sandven 2010). After each initial impact, negative excess pressures would promote desaturation, and shallow water depths would not provide enough ambient pressure to re-saturate the filter ring (Sandven 2010). This issue impacted 91% of the measurements at water depths shallower than 5 m, and only 15% of the measurements at greater water depths. Large areas of the surveyed tidal flats are exposed at low tide, which could lead to over-consolidated surficial sediments by dessication that may dilate upon shearing. In future surveys, this problem may be mitigated by using mineral oil for the saturation of the pressure inlet and filter ring system instead of water (Sandven 2010).

The Bernoulli effect

The Bernoulli effect clearly impacts the pressure recording during free fall through the water column. Correction for the Bernoulli effect generally gave an estimate of water depth in agreement to the water depth measured using the echo sounder, accounting for uncertainties of the echo sounder measurements, and potential impacts of small-scale bathymetric features (discussed above). It should be mentioned that no high accuracy bathymetric measurements were made during the penetrometer survey. Nonetheless, further investigations in a tow tank are planned to derive a full calibration. For the present field study, it can be concluded that the Bernoulli effect influences the pressure measurement during fall through the water column, and that the pressure measurement should be corrected for this. A Bernoulli correction of the pressure during penetration into the sediment, however, appeared inappropriate based on the presented measurements. It rather seems that the penetrometer is entering sediment enclosure at an ambient pressure that is affected by the Bernoulli effect. Particularly, the fact that the pore pressure decreases further when the penetrometer comes to a stop, and then slowly approaches the hydrostatic pressure (Figs. 11 and 12), suggests that a Bernoulli correction cannot be directly applied to the pressure during penetration. More research is needed here, and measurements under controlled laboratory conditions are planned. For the following considerations, it is assumed that the penetrometer enters the sediment with a lowered ambient pressure due to the Bernoulli effect, and that the pressure reading does not equilibrate instantaneously with a decrease in velocity due to full embedment into sediment. As a consequence, the uncorrected pressure p is correlated to its extrapolated hydrostatic pressure profile p h based on penetration depth, as well as the actual hydrostatic pressure p s~p b, allowing the rejection of a correction for the Bernoulli effect for the penetration into sediment. This assumption has also been applied in other studies (e.g., Stegmann et al. 2006; Seifert et al. 2008).

Drained versus undrained conditions

Excess pore pressures and an undrained pore pressure response due to changes in stress are commonly expected in fine-grained soils due to their low permeability, but not in sandy sediments with an overall higher hydraulic conductivity. Thus, the observation of positive excess pore pressures (>p h) in sandy sediments must be associated with the high impact and penetration velocities (~4.75 m/s). In the case of free fall penetrometers, the rate of penetration and loading is constantly decreasing with the loss of kinetic energy. Therefore, depending on sediment type and penetration velocities, the loading conditions for dynamic penetrometers may be initially undrained, and start to transition to a drained failure (pressures remain at the hydrostatic level) by the end of penetration (Houlsby and Cassidy 2011; Steiner et al. 2013). During the transition from undrained to drained conditions, the sediment may be considered partially drained. In this condition, there may be some movement of the pore fluid, but the movement is still restricted by the sediment’s hydraulic conductivity during loading, leading to an excess pore pressure (Duncan et al. 2014). This could have an important consequence on the interpretation of sediment strength from dynamic penetrometers as the state of drainage during penetration may be unknown, and shear strengths of sediments (or soils) are typically only determined for the fully drained or undrained conditions (Duncan et al. 2014).

Numerous authors have proposed the use of a “non-dimensional velocity” to determine the degree of drainage during penetration (Randolph 2004; Randolph and Hope 2004; Chung et al. 2006; Kim et al. 2008; Lehane et al. 2009; Robertson 2010; Steiner et al. 2013). The non-dimensional velocity is given by:

$$ V=\frac{vD}{c_{\mathrm{v}}} $$
(6)

where V is the dynamic non-dimensional velocity, v the penetration velocity, D the penetrometer diameter, and c v the coefficient of consolidation of the sediment. Steiner et al. (2013) reported that values of V ≥30 and <0.3 correspond to fully undrained and drained conditions, respectively. For the assessment of the drainage conditions, they suggested that the highest value of c v should be used in the soil profile (meaning the sediment within the penetration depth that drains the fastest).

Using a velocity of 4 m/s to represent most of the penetration distance in this study (e.g., 35 cm of 45 cm penetration in the example shown in Fig. 3), fully undrained conditions exist for c v <1.17*10–2 m2/s, and fully drained conditions exist for c v ≥1.17 m2/s. Using a velocity of 1 m/s to represent the last few centimeters of penetration, c v <2.92*10–3 m2/s and c v ≥2.92*10–1 m2/s yield fully undrained and drained conditions, respectively. Lambe and Whitman (1969) reported typical c v=5*10–7 m2/s for soils with a liquid limit of 30 undergoing undisturbed compression, representing fully undrained for the above conditions. Additionally, Duncan et al. (2014) reported c v for clays to be in the range of 2.8*10–8 to 2.8*10–6 m2/s, which also correspond to fully undrained conditions during penetration. Duncan et al. (2014) documented c v=2.8*10–5 m2/s for silts (~100 times those of clays), and c v=2.8*10–3 m2/s for sand and gravels (~100 times those of silts), resulting in fully undrained conditions for silts, and primarily undrained conditions for sand with the possibility of partially drained conditions as the penetrometer slows.

Sediment characterization

Knowing that the penetration process was (highly) likely undrained, the pressure response can be used to investigate the different sediment types. Positive excess pore pressures or supra-hydrostatic conditions indicate contractive sediment layers, and thus, likely represent normally consolidated clays to loose sands. Sub-hydrostatic conditions (here being considered p<p h) correspond to dilative materials. These sediment layers may represent over-consolidated clay or dilative silty sands.

The development of sub-hydrostatic pressures (p<p h) in the upper ~10–20 cm of penetration was a characteristic of type A and type B deployments (at least 41.7% of type A; 40.7% of type B; only 10.3% of type C). The presence of sub-hydrostatic pressures in the early stages of penetration, and their greater association with type A results, can be explained by a layer of desiccated or over-consolidated clay at the upper regions of sediment in the tidal flats. This assertion is supported by the differentiation of sediment type by the relative rate of the pressure response after penetration (Table 1). The increasing rate of pressure response from type A to type C sediments reflects increasing permeability, or hydraulic conductivity of the sediments. The implication is that type A results correspond to fine-grained clayey sediments, type C results correspond to coarse-grained sandy sediments, and type B results correspond to an intermediate or mixed sediment. This notion is corroborated by the sediment samples shown in Figs. 3, 4, 5 and 8. Type A results occurred most often in areas with less than 75% sand and small median grain sizes d 50. Type C profiles occurred exclusively in areas with more than 75% sand, whereas type B profiles were observed in areas with variable amounts of sand, but values of d 50 <300 μm (Figs. 3, 4, 5, 8).

The early pressure response (before the pressure sensor is fully embedded) that was sometimes observed may reflect strong deformation of the sediment due to high energy impacts. For weak, low density sediments, a high speed impact may initiate a pressure wave pushing material and water away from the probe in an explosive manner. Evidence of this effect may be seen in the upper portions of the pressure response, where the pressure begins to deviate from the hydrostatic line before the pore pressure sensor inlets enter the sediment (Fig. 12). The larger sub-hydrostatic pressures that begin after approx. 5 cm of penetration are attributed to the dilation of desiccated material (as described above), but the initial fluctuations may be a result of material being ejected from the impact site, accounting for the effects occurring prior to the sensor entering the seabed. For higher density materials such as sand, the impact energy may not be great enough to completely eject the sediment from the impact site, and the material would bulge around the penetrometer in a general bearing capacity failure. The increase in pressure prior to the sensor entering the seabed (Fig. 12, lower right panel) could indicate a sediment bulging around the penetrometer. While this interpretation is consistent with physical and soil mechanical concepts, and such sediment deformations have been documented around penetrometers before, this cannot be proven using the available dataset. Measurements with an attached camera are planned to investigate this issue further.

Sediment surface deformations and high energy impacts would likely cause a significant amount of disturbance of sediment properties. For example, materials initially dilative may become more compressible after being remolded, which would also influence their permeability. Furthermore, the impact and penetration process may cause some level of sediment mixing as the displacements occur, and high gradients from induced excess pore pressures may lead to void redistribution as the pore pressures dissipate. Therefore, the drop in pressure observed when the penetrometer comes to rest (Figs. 10, 11, 12) may be associated with temporary gaps or channels between the sediment and penetrometer, allowing for the rapid dissipation of induced pore pressures, effects of suction, or further velocity effects related to differences in the probe’s velocity and the surrounding material and fluid.

Differences in the rate of p approaching p s, when the penetrometer has completed its penetration and is at rest at the respective sediment penetration depth, are clearly related to the sediment types (Table 1 and Figs. 10, 11, 12), and are consistent with the expectations of permeability. The pressure in type A deployments approached hydrostatic pressure according to the water depth very slowly (~0.5 kPa/s; Fig. 10), consistent with expected low permeability in fine-grained sediments on the tidal flats. In type C deployments, p~p s within less than 0.1 s (~147 kPa/s), consistent with the fact that most of these sediments are sandy (Fig. 12). Type B deployments show a wider range, in accord with more mixed sediments. In the example shown in Fig. 11, p is approaching p s with ~4 kPa/s.

Conclusions

An estimate of quasi-static bearing capacity (qsbcmax) derived from deceleration records using a portable free fall penetrometer was related to sand content and median grain size (Fig. 13). Sediments with a sand content ranging between 40–70% were consistently soft, while sediment strength varied significantly for sediments with a sand content above 85% (Fig. 6). The correlation within the group of sandy sediments was further improved by using d 50 (Fig. 7), but a significant amount of variability in sediment strength was still associated to d 50 >150 μm. This variability is at least partially associated to differences in bulk density (Fig. 9), which likely resulted from sediment transport history. A reduction of qsbcmax was observed along steep channel slopes, consistent with geotechnical theory; however, overall slopes of 5–10% seemed to have a minor impact on the results (Fig. 5). These findings are essential information for the improvement of the understanding of the characteristics of tidal estuary surface sediments, and thus, prediction of long-term sediment dynamics, or the impact of engineering actions.

Fig. 13
figure 13

Conceptual sketch of findings related to the impact of particle size distribution, bulk density and morphology of tidal estuary sediments on sediment strength, and resistance against the penetrometer

The relative rate at which recorded pore pressures approached hydrostatic pressure after penetration correlated well to sediment type and strength, and can be used to classify the sediment. While this has been widely accepted for standard offshore cone penetration testing, it has rarely been documented for high impact, portable free fall penetrometers, particularly in that standard dissipation of positive excess pore pressure over time was not observed, but rather pressure equilibrated from sub-hydrostatic values to hydrostatic. Additional features in different phases of the penetrometer’s penetration into the sediment were identified in the pressure signatures that were related to differences in sediment type, and first hypotheses were presented to explain this behavior. This included that pore pressure changes during penetration suggested additional insight into the nature of the material (dilative versus contractive), supporting working hypothesis 2 (Fig. 14). Furthermore, pressure deviations occurring prior to the sensor entering the bed may provide evidence of ejected low density mud, or a bulging failure in sand. A correction of the pressure for the Bernoulli effect during free fall through the water column seemed crucial to estimate hydrostatic pressure correctly (Fig. 14). However, the results indicated that the same approach is unlikely directly applicable to the penetration process due to the sediment–penetrometer interaction.

Fig. 14
figure 14

Conceptual sketch of findings related to pore pressure measurements. A decrease in measured pressure during fall through water can be associated to the Bernoulli effect. Results indicate that sediment surface behavior is likely reflected in the pressure measurements just upon impact. Dilative versus contractive sediment behavior is reflected in the pore pressure results as well as hydraulic conductivity, being dependent on grain size and density. The role of the Bernoulli effect during penetration is indicated but still unclear