Soil Classification and Site Variability Analysis Based on CPT—A Case Study in the Yellow River Subaquatic Delta, China

Abstract

The Yellow River Delta is located at the junction of the Yellow River and Bohai. The impact function from the river and the dynamics of the ocean tides make the soil composition and distribution in this area substantially complicated. In order to test the distribution and variation of the soil layers in the Yellow River Delta, the soil layers in the test area were classified and the variation was calculated using the cone penetration test (CPT). The following conclusions were drawn: (1) the soil in the measured area is mainly composed of sensitive fine-grained soil, accounting for about 70% of all soil types, and the content of sensitive fine-grained soil in the far-sea position is higher than that in the offshore position in the direction perpendicular to the coastline. (2) It has a high vertical variability index (VVI) at the near-shore location, above 45%, and a low vertical variability at the far coast, generally below 20%. (3) The horizontal variability index (HVI) changes significantly near the coast, and it remains below 45% in the test area.

1. Introduction

A large number of cables and pipelines are distributed on the bottom of the Yellow River Delta, which is an important area for offshore oil and gas production [1]. Many international researchers have paid full attention to the investigations of the geological disasters in the Yellow River Delta [2,3,4,5], finding that the study area is characterized by complex topography and geological disasters. The shoreline boundary on either side of the estuary moves towards the sea at the average speed of 100 m/month [6]. The engineering characteristics of the sensitive fine-grained soil are relatively complicated, and may produce fluid mud to cause potential marine geological disasters [7,8]. The wave actions will directly lead to changes in the seabed topography at the mouth of the Yellow River, making the construction of the area, the engineering survey of the soil, and the prediction of geological disasters on the seabed difficult [9,10]. The Yellow River Delta is located at the intersection of the Yellow River and Bohai, which will cause sediment accumulation, and the effect of waves will cause the erosion of silt and sediment [11]. Changes in the erosion and deposition of the Yellow River Delta will seriously affect the development of ports [12]. Besides this, the deposition of the seabed results in frequent instability accidents, making it challenging to develop offshore projects on a large scale [13]. Therefore, the geological conditions in the Yellow River Delta are relatively complex, with an uneven distribution of soil layers, diverse soil composition, the influence of ocean tide dynamics, river erosion and accumulation on the coast, etc. Various conditions make the assessment of the strata of the Yellow River Delta region and exploration work very complicated. As a result, it is very important to investigate the variability of the entire test area. The variability is composed of the value and change of horizontal and vertical variability.

The cone penetration test (CPT) and its enhanced version (piezocone, CPTu) have seen extensive applications in many kinds of soils due to their continuity and repeatability [14]. Compared with traditional laboratory tests, CPT—with the advantages of fast, continuous, repeatable and reliable measurements—was often the preferred method for in-situ marine geological investigation, and abundant experience has been gained in the interpretation of CPT results for soil types [15], meaning that CPT is used in engineering geological monitoring widely. Due to the effect of ocean dynamics, sedimentary minerals will be transported and accumulated again [16], and the mineral composition of surface sediments in the modern Yellow River Delta is rather complicated [17]. These factors made the stratigraphic variability of the delta area relatively large. The soil classification information obtained by CPT can provide essential and useful information to geotechnical engineers [18,19,20]. The proposed soil classification charts by Robertson [21] have been popular and widely used [22,23]. Using the data obtained by the cone penetration test, it is convenient to investigate the changes of the soil layer and calculate the properties and parameters of various soil bodies. It can meet the simple operation and protect the nature of the original soil. Using the measured data to comprehensively investigate the variability values can draw more accurate results.

Due to the complex composition of the seabed soil in the Yellow River Delta, the study of the quantity and composition of the soil layers cannot provide the characteristics of the changes in the soil layers in detail. Therefore, it is necessary to continue to quantify the changes and calculate the changes in variability. Around the location of the area promoted by the Yellow River Delta, we established several points that can be used to investigate the horizontal and vertical changes, that is, the changes in the direction of the vertical coast of the Yellow River Delta near the coast and parallel to the coast.

Then, we used the variability calculation method summarized by previous scholars to investigate the characteristics of the soil variability in this area. Finally, we preliminarily summarized the change characteristics of the soil layer. Combined with the advantages of the cone penetration test, the composition of the soil layers in the Yellow River Delta region can be explored and the variation of the seabed soil can be quantified. The classification of the soil layers and the analysis of the site variability provide support for the Yellow River Delta underwater engineering construction, and make a preliminary exploration of the composition and variation of the underwater seabed soil.

2. In-Situ Survey

In order to determine accurate geological conditions and soil classification in the north of the Yellow River Subaquatic Delta, ten stations were arranged, of which one is perpendicular to the coastline, two others are parallel to it, as shown in Figure 1. The independently-developed Submarine Engineering Geological Environment Investigation System (SEGEIS) carried out a one-month CPT survey—as shown in Figure 2.

Figure 1. The stations and survey lines.

Figure 2. Underwater CPT system, developed independently.

The surface layer of the area consisted of yellow-brown silt with a thickness of 4.7 m and a medium density, of which parts were interspersed in the silty clay. The standard test shot number was 12–31, and the median diameter of the topsoil particles was about 0.03 mm, of which the saturation and density were 27.4–32.3%, and 19.4–20.1 kN/m³, respectively. The in-situ investigation system of the seabed engineering geological environment had many functions, such as in-situ observation, in-situ detection, static penetration, penetration, the extraction of sediment static sampling tubes, seabed environmental lighting, video shooting, and photography. The underwater CPT equipment, developed independently, was an in-situ investigation system for the subsea engineering geological environment. A high-quality pressure strain sensor was used to measure the cone tip resistance (range 0–50 MPa) and the sidewall friction resistance (range 0–1 MPa). The pore water pressure was measured by a piezoelectric element (range 0–2 MPa). The tilt angle was measured by an accelerometer (inclination angle in both the X and Y directions, range 0–15°). Finally, the sampling and penetration methods were static and hydraulic. The parameters of the underwater cone sounding test equipment are shown in Table 1.

Table 1. The main parameters of the independently-developed underwater CPT system.

Specifications	Parameters
Water depth of working	100 m
Probe parameters	Φ36 mm
Sampling tube parameters	Φ75 mm/Φ110 mm (drive pipe)
Penetration rate	1.2 m/min
Penetration force	2.5 t
Hydraulic pressure	10 MPa (Penetration)/12 MPa (Extraction)
Cone tip resistance range	50 Mpa
Side friction force	1 Mpa
Cone tip cross-sectional area	10 cm2
Sample diameter	75 mm
Sampling depth	5 m

3. Soil Classification

Soil classification charts have been used widely to assess soil behavior based on CPT results [22,23]. The soil classification chart [21] is divided into twelve zones in order to represent the different grain size distributions, as well as the different soil states or behaviors coded by numbers. For example, Zone 1 indicates a soil type of sensitive fine grains, while the CPT data points of classical stiff fine-grained soil will fall in Zone 11, according to Figure 3, as given by Robertson [21].

Figure 3. Soil behavior type chart.

Fine-grained soil is widely distributed in the Yellow River Delta region, and some sandy soil with larger particle sizes also exists. As shown in Figure 4, there is an increase by 10 MPa sharply at depths of 1.5 m and 2 m, etc. The continuous accumulation of pore water pressure under the action of tides and waves in the sediments of the Yellow River Delta results in the continuous accumulation of the strength of the surface soil [24]. Typically, there is high cone resistance in sands but low resistance in clays, while the friction ratio is contrary, according to Robertson’s research [25].

Figure 4. The in-situ CPTu data.

The pore water pressure is distributed in 150 kPa–200 kPa, and the different sites vary significantly in depth. The pore water pressure will drop steeply at a certain depth. As such, the excess pore pressure shows a decreasing trend near 150 kPa–200 kPa in 0.5 m, at 3.4 m depth below from the surface for S9 and 2.1 m depth below for S4. At 4.5 m, it drops by about 140 kPa in S5. There is a high negative pore pressure in the case of high-density fine sands and dilative silts according Cai et al. [23]. Affected by the over-consolidation state, the pore water pressure may also become small or even negative [26].

The data for the locations of the 10 stations are arranged in the chart [21]. There is a clear distribution of 10 measuring points in area 1, and there no points in areas 11 and 12, revealing preliminarily that the soil composition in the area is mainly sensitive fine-grained soil, as shown in Figure 3. 50% of the points are in zone 1, which is mainly related to the sediment accumulation in the Yellow River Delta, according to the relevant research [27]. The sediment of the Yellow River is mainly from the Loess Plateau, and the sensitive fine-grained soil is mainly in the depth of 40 cm below the surface, while the deep layer is dominated by clay [28] where there is an obvious effect from the wave actions [27]. The method of soil layer classification in this area is shown in Table 2. By referring to the classification results in Figure 3 and Table 2, the soil layer results are obtained in the test area. As shown in Figure 5, there is very fine sand on the delta front, which is consistent with previous studies [4]. Sedimentation in the sedimentary area is slow when it is near the land in the west, and steep when it is far away from the land in the east. According to previous research [29], from the coast to the deep sea, the sediment composition shows a coarse-fine-coarse variation.

Table 2. Non-normalized CPT soil behavior type (SBT) chart [21].

SBT Zone Robertson et al. (1986)	Proposed Common SBT Description
1	Sensitive fine-grained
2	Clay—organic soil
3	Clays: clay to silty clay
4 & 5	Silt mixtures: clayey silt & silty clay
6 & 7	Sand mixtures: silty sand to sandy silt
8	Sands: clean sands to silty sands
9 & 10	Dense sand to gravelly sand
12	Stiff sand to clayey sand
11	Stiff fine-grained

Figure 5. The soil layers profile.

4. Site Variability Analysis Based on CPT

By calculating the site variability, the geological conditions of the measured area can be evaluated, and the civil engineering construction of the site can be guided. There are three main parts about the analysis on the site variability of the static penetration test. The first part is the calculation of the vertical variability, including the intra-layer variation index (VVI_IL), log vertical variability index (VVI_log), and cone resistance vertical variability index (VVI_qc). The second part is the calculation of the horizontal variability. The final part is a comprehensive analysis of the obtained data of the horizontal variability and vertical variability, as well as the variability of the site considering the geology and marine factors of the measured area.

When processing the CPT data, all of the abnormal data caused by instrument error, improper operation, and special geological phenomena should be correctly identified and interpreted reasonably.

4.1. Vertical Variability Index (VVI)

First, the depth and q_c values measured by CPT are detrended. It is necessary to fit the data with functions. The polynomial function is used to fit the q_c data and depth for the data detrending processing in this paper. The polynomial function used is denoted as f(x). After the fitting of the function is completed, the function is detrended, and the function used is as shown in the following Equation (1):

$$\omega =\frac{y-f\left(x\right)}{f\left(x\right)}$$

(1)

where f(x) is the q_c value obtained after the polynomial function processing, y is the q_c value in the original data, and ω is the q_c value after the detrending. The relationship between the q_c data and the depth after the detrend processing of S8 is shown as an example in Figure 6.

Figure 6. CPT detrended data with the fitted trendline.

The value after detrending ranges from −1 to 1. The crossing point and crossing distance are shown in Figure 6; the scale of the fluctuation-normalized coefficient of coefficient variation (COV) is defined as SNC, which will indicate the variety of the soil layers. Furthermore, the area of one side of the trendline will indicate the value of the scale of fluctuation (SF), which represents the variety of the soil layers. Although the cone penetration test method suffers a measurement error [30] when the soil layer thickness is small, the influence of the layer thickness on the variability is not significant. Therefore, the error of the cone penetration test in the case of a minimal-thickness soil layer is ignored in this paper.

4.2. Intra-Layer Variation Index (VVI_IL)

By considering that the intra-layer variability index is affected by COV and SF, a new parameter—SNC—is defined. The variation index of intra-layer size has a significant connection with the each-layer soil thickness and the number of soil layers in the entire soil section. First of all, the data obtained from the CPT is used in calculating each soil COV with soil layer maps and different soil type layer classification. Then, the relationship between the data after the detrending process with crossing points is established. The SF value used in this paper has been introduced [31], which relates to the model and the soil type used in the calculation. According to the research, many differences vary from 0.07 m to 4 m in marine or clay soil types regarding the SF value [32]. Finally, the SNC value is found with the formula proposed [33]. Figure 7 clearly illustrates the calculation process.

Figure 7. The SNC calculation method (adopted by Salgado et al. 2019) [33].

Here, n_c is the number of the crossing point, ${\mathsf{\Delta }}_i$ is the crossing distance, and the value of L_r is 1. ACF means the Autocorrelation Function, which is used to determine the distance of the correlation, which changes with the function type [34]. The twice-positive area formed by the fitting function and the trend line is regarded as the SF value in this paper. The main factor affecting the variability in the soil layer is q_c; thus, only the SNC of q_c needs calculating [33]. The SNC results are shown in Table 3.

Table 3. The SNC.

Station	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10
SNC (%)	4	9	34	15	13	22	3	11	89	85

The results are quite different, where S9 and S10 approach 90%; S3 and S6 are near 30%; S2, S5, and S8 are about 10%; and S1 and S7 are just about 3%. The soil stratification, SNC data, and the variability index of the intra-layer size are related to the thickness of the soil layer, the number of normalized q_c crossing points in each layer, and the COV of each layer of the variation coefficient. Therefore, the SNC at S9 and S10 is bigger than those of the other stations. Furthermore, the thickness of a single layer and the COV is large, resulting in a large SNC. There is a prime feature within S4 and S5, given that both of them have a dominant thickness; even the thickness of S5 is over 4 m in a single layer between S4 and S5. However, there is clay or silty clay type in both S4 and S5, making the VVI_IL in S4 and S5 small. All in all, the thickness of a single layer has a significant effect on the SNC.

4.3. Log Vertical Variability Index (VVI_log)

The logarithmic vertical variability index mainly calculates the diversity factors (DF) within a layer. Firstly, the soil layer is classified according to Table 4. The numbers of the sand, clay, and mixed soil obtained are respectively recorded as N_S, N_C, and N_MS. Then, the proportions of the D_S, D_C, and D_MS—of which the sum is 1—are calculated. The D* shown in Figure 8 represents the distance between the soil of any content of the three, and the three of them are 1/3, which is the degree of variation of the three. Theoretically, it fluctuates between 0 and 1, such that the DF ranges from 0 to 100. The lower the DF, the higher the dominance of the soil; therefore, the vertical variability will be smaller. Eventually, the final VVI_log value is obtained. The main steps in the DF and VVI_log are in Figure 9.

Table 4. Soil group classification for the modified Robertson chart used to group soils for the calculation of dominance factor D_i and diversity factor DF.

Soil Group	Soil Types
Sand	Gravelly sand to sand (very loose to very silty sand)
	Clean sand to silty sand (very loose to very dense)
Clay	Organic clay
	Sensitive fine grained
Mixed soil	Silty sand to sand silty
	Clayey silty to silty clay
	Clay to silty clay

Figure 8. D* schematic diagram (modified from Salgado et al., 2019) [33].

Figure 9. The VVI_log calculation method (modified from Salgado et al., 2019) [33].

The NDL is the number of layers per unit length. In this test, various values of NDL were calculated, with a range of 1 to 5.7, revealing that there are complex and diverse soil types and layers in the Yellow River Delta. The calculated VVI_log value is normalized with the log₁₀(x) function in order to keep the value obtained in the range of 0 to 1. As illustrated in Table 5, S1, S5, S8, S9, and S10 are below 90%, indicating that there is a dominant soil layer, or that the number of soil layers is small. Taking S4 as an example, there is a large proportion of sensitive fine-grained soil, but the number of soil layers at this point is 14, which is exceptionally numerous. Therefore, the VVI_log value of S4 is relatively large, resulting in a large vertical variability, which means that the value of VVI_log can intuitively show the influence of the number of soil layers and the thickness variation of the different soil layers.

Table 5. The VVI_log.

Station	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10
VVIlog (%)	82	94	96	99	79	98	99	83	77	46

4.4. Cone Resistance Vertical Variability Index (VVI_qc)

VVI_qc is a COV of q_c, and the standardized COV of q_c as the VVI_qc value of the soil layer, because the VVI_qc value is related to the soil layer profile and the composition of the soil type. We can calculate the COV from the obtained qc value, and we should normalize the result if it is between 0–1. In a soil layer comprising a single soil, especially a single clay or silty clay layer, the VVI_qc value is below 10%, such as for S5, S8, S9 and S10, as listed in Table 6. On the contrary, if the profile consists of multiple soil types, the number of soil layers is plentiful, and the physical characteristics of the soil types are quite different, then the VVI_qc is relatively large, just as for S1, S2, S4, and S7. Comparing S8 with S7, the number of layers is similar, and the main components of the soil are clay and silty clay, but the number of soil layers in S7 is five times as many as for S8, so the VVI_qc of S7 is higher than that of S8.

Table 6. The VVI_qc.

Station	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10
VVIqc (%)	58	66	29	48	10	33	42	3	6	3

Eventually, the vertical variability index can be calculated by Equation (2), and the VVI value is listed in Table 7.

$$VVI=0.2VV{I}_{IL}+0.2VV{I}_{\log }+0.6VV{I}_{qc}$$

(2)

Table 7. The VVI.

Station	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10
VVI (%)	52	60	43	52	10	24	46	21	37	28

The VVI is related to the soil layer content, number, and thickness. When the composition of the soil layer is single within three layers, the vertical variability is relatively single, which makes the value of VVI small, i.e., under about 40%. The vertical variability is various; meanwhile, when a single soil layer is thicker where there are five or more layers in one meter, and the number of soil layers is larger in one soil profile where the totally different layer numbers exceed ten, this will also cause more considerable vertical variability, and the value of VVI will even reach 50%. Therefore, the number of layers has a significant effect on the vertical variability.

4.5. Horizontal Variability Index (HVI)

When calculating the coefficient of the horizontal variation index, the size of the cone tip resistance and the depth of the sounding need to be considered as two factors affecting the horizontal variability, because they have a deficient level of variability to some extent if the changes in the two groups of qc data have a high correlation.

First, the influence factors of q_c should be indicated. The differences between q_c and the average of q_c are calculated. The f₀ is defined by dividing the previous maximum value q_c with the absolute q_c value |q_c|. Secondly, the correlation coefficient ρ_xy between q_c and the depth is calculated under the consideration of the depth effect. The parameters of f₁ and f₂ are figured with the Equations (3) and (4). The eventual HVI value can be calculated using Equation (6). The main calculation process is shown in Figure 10, where ρ_xy is the correlation coefficient, and s is the vertical distance between the two CPT data points [33].

$$f_1=\frac{{\rho }_{xy}+1}{2}$$

(3)

$$f_2=1-e^{\frac{s}{4}}$$

(4)

$$f=[0.8(1-f_0)+0.2f_1]f_2$$

(5)

$$HVI=1-\frac{{\displaystyle \sum _1^{n}f}}{n}$$

(6)

Figure 10. The HVI calculation method (modified from Salgado et al., 2019) [33].

The result of HVI is shown in Table 8. The horizontal variability index is related to the content of hard clay. According to the previous research [33], the high content of hard clay or silt will make the horizontal variability index large; otherwise, the horizontal variability index will be small. Besides this, due to changes in the carrying capacity of the Yellow River and the carrying capacity when entering the sea, the particle size of the Yellow River Delta varies from the estuary to the sea. Therefore, the horizontal variability has a major change with the direction of the vertical coastline [35]. In terms of the soil stratification in the overall soil profile, the content of hard clay in S6 is higher. Therefore, the horizontal coefficient of variation of S6 is the largest, while the rest are medium.

Table 8. The HVI.

Station	S1	S2	S3	S4	S5	S6	S7	S8	S9	S10
HVI	44	45	38	59	56	66	45	43	42	35

6. Conclusions

In-situ tests were carried out on the Yellow River Delta using independently-designed submarine CPT equipment. The classification of the soil layers and the analysis of the site variability were performed based on the obtained data. The main conclusions are as follows:

(1)

Sensitive fine-grained soil is widely distributed in the Yellow River Delta. The classification results revealed that 70% of the soil layers in the test area are mainly sensitive fine-grained soil, and the soil composition type is distributed variously. There is an obvious change in the direction perpendicular to the coastline where the content of sensitive fine-grained soil in the near-coast is less than that in the far-coast.

(2)

Vertical to the coast, the water depth changes complexly and is obviously affected by hydrodynamic forces. The vertical variability index varies obviously in the direction perpendicular to the coastline. The vertical variability in the offshore areas is basically maintained in the range of more than 45%, which is much larger than the value in distant sea areas, which is less than 20%.

(3)

The soil layer changes are complex because of the shallow water depth and the obvious effect of the wave load near the shore. The diversification of the horizontal variability mainly occurs near the coast, and gradually decreases from the northwest direction to the southeast direction. There is no significant change of the horizontal variability distribution in the measured area, and its value is maintained at a moderate level of about 45% or less.

(4)

Based on the previously-available geological composition of the Yellow River Delta, the vertical variability and soil distribution are significantly different in the disturbing delta front and the flat delta front. In the distribution delta front, the soil composition is diverse, with high vertical variability basically exceeding 40%; in the flat delta front, the soil composition is homogeneous, with low variability, and is below 20%.