Physical, geochemical, and clay mineralogical properties of unstable soil slopes in the Cameron Highlands

2.1 Site description

The Cameron Highlands is a high topographic region in the Pahang State, Malaysia, situated in the western part of the Peninsula. The study area is within the northwest and southwest part of the Cameron Highlands (101°21′–101°26′E and 4°24′–4°36′N) and covers an area of 205.4 km² (see Figure 1). The four principal localities, namely, Ulu Telom, Brinchang, Ringlet, and Tanah Rata, overlain by the sampling points and drainage lines are represented in Figure 2. The elevation above sea level of the highlands ranges between 300 and 2,060 m [20]. Only ∼15% of the area is flat and near the main river. The river sources are near the water divide separating Pahang from Perak, and on a general perspective, their flow is easterly, with two principal rivers, namely, Sungai Telom (northern part) and Sungai Bertam (central part). The upper part of Sungai borders the Kelantan and Perak States. Sungai Telom and its tributaries drain part of the plateau around Kampung Raja. Sungai Bertam, which drains a substantial portion of the plateau, emanates from the hills close to the Kea farm area and Gunung Brinchang, flowing southernly [21].

Figure 1

Relative position of (a) Peninsular Malaysia, (b) Cameron Highland, and (c) Study area.

Figure 2

Study area overlain by sampling point classification and drainage pattern.

The study region is the hilly terrain of the Cameron Highlands (see Figure 3). More than 60% of the highlands have a slope angle of +20° [22]. The annual rainfall is high, averagely ranging from 2,500 to 3,000 mm [23]. The two pronounced wet seasons are from September to December and from February to May. The peaks are from March to May and from November to December. The single day recorded rainfall high ranges from 87 to 100 mm. During these times, many streams and rivers in the Cameron Highlands could overflow, flooding the surrounding areas, and landslides, such as debris flow, can occur along the river valleys [24]. The average daily temperature is 24°C during the day and 14°C during the night. More than 500 soil series exist in Malaysia, most of which are residual and coastal alluvial [25]. The Cameron Highlands soils are primarily derived from the local parent rocks [26]; however, with time, natural and artificial processes, such as erosion, landslides, farming, climate, and road construction, have influenced the soils. The studied soil falls under sandy loam and sandy clay loam classification according to the USDA standard. Only ∼4% of the entire Cameron Highlands is settlements, 8% is used for various agricultural purposes, 86% is forest, and 2% is reserved for recreation and other endeavors. Furthermore, tourism, agriculture, and hydroelectric power generation are the principal income sources for this area [27]. The two principal catchment areas of Hulu Telom (north) and Bertam Valley (south) provide the required water resources needed for recreation, irrigation, hydroelectric power generation, and domestic use [22]. Intense agricultural practices coupled with the rapid urbanization of this area continue to negatively affect the environment, resulting in land degradation, erosion, landslides, and flooding.

Figure 3

Elevation map of the study area.

2.2 Soil sampling and categorization

A total of 20 soil samples were taken from slope surfaces of varying landslide conditions using a hand auger at a depth of 50 cm from the surface. The geographic coordinate of each sampling point was recorded using a handheld portable global positioning system. To avoid contamination, the sampler (hand auger) was cleaned with water before and after each sampling. Also, direct contact with the soil was avoided. The soil was directly collected into a plastic bag, labeled, and stored in a cool box; it was then transported to the laboratory for analysis. The soil types of the 20 sampled slopes cluster around sandy loam and sandy clay loam section of USDA textural triangle. Samples were primarily grouped based on the observable landslide condition of the slopes on the field. Thirteen samples were from slopes that had experienced landslides of varying severities and classified as landslide-affected slopes (LAS). Seven samples were from locations where no landslides had occurred and grouped as nonlandslide-affected slopes (NAS). The categorization was further evaluated using hierarchical cluster analysis (HCA) and principal component analysis (PCA) based on the physicochemical characteristics of the soils.

2.3 Soil parameters determination

The grain size distribution was determined by mechanically sieving dry samples and hydrometer analysis following the ASTM D422-63 2007 standard method. Each sample was sieved using test sieves of BS 410, and the grain mass from each sieve was estimated. Hydrometer analysis was performed on samples retained in the pan using the principle of sedimentation of soil grains in water, in which the sample was added to the solution of dispersion agent (sodium hexametaphosphate powder) and distilled water. Sample readings were taken at intervals and the grain diameter was calculated using Stoke’s law. The actual hydrometer reading, corrected hydrometer reading, and temperature were noted, and grain size distributions were estimated.

The moisture content (MC) was analyzed on wet samples using the gravimetric method following the British standard method (BS 1377 1990). The organic matter content (OMC) was determined using the loss on the ignition (LOI) method [28]. The oven-dried samples at 105°C for 24 h were heated in a furnace (LT 1200–6 Furnace) at 550°C for 4 h. The weight of each sample was recorded before and after heating, and the OMC was subsequently determined using equation (1):

where M ₁ is the weight of the crucible, M ₂ is the weight of the crucible plus the sample, and M ₃ is the weight of the crucible plus the ash.

The pH of each soil sample was measured using an Accumet AB150 pH meter following the British standard method (BS 1377). Soil samples were dried, disaggregated, and sieved through a 2 mm mesh and placed in tubes, and distilled water was added with a soil/water ratio of 1:2.5. Tubes were subsequently shaken in an orbital shaker for 1 h before settling for 24 h. The pH meter was calibrated with a buffer solution at pH 4, 7, and 10, and the pH of each sample was measured by dipping the pH probe into the tube above the soil level. CEC was determined using the methylene blue (MB) spot test method [29]. Samples were sieved through a 69-µm mesh, and 15 mL distilled water was added to 5 g of soil sample; then, 0.5 mL MB solution was incrementally added. After each increment, the solution was stirred, and a drop from the mixture was placed on a filter paper until a permanent blue halo formed around the dark blue droplet on the filter paper. The CEC was calculated using equation (2):

where V is the volume of MB used, N is the normality of MB, and W is the sample weight.

Concentrations of anions were measured using a Metrohm ® 861 advanced compact ion chromatographer (IC). Standard solutions formulated from certified multiple elements with varying concentrations (standard 1, 2, and 3) were used for calibration. The extraction process was conducted on the dried soil, and 2 mL of extract from each sample was attached to the filter using a syringe and plugged into the IC for the analysis. The result was obtained from IC Net 2.3 software installed on a system connected to the IC.

The clay mineralogy of soil samples was determined using the X-ray diffraction (XRD) technique, following the United State Geological Survey method (USGS 01-041), using PANanalytical EMPYREAN DY1032 machine made in Netherland. Nine representative samples were selected: seven from LAS and two from NAS. The samples were prepared by placing them in test tubes below 5 mL, and three drops of Calgon were added, after which distilled water was added until 30 mL. The whole setup was placed inside ultrasonic cleaner for 10 min, brought out and shaken for 1 min, and then allowed to settle for 30 min. After the samples preparation, each sample was subjected to four treatments: (1) air dried/normal, (2) glycolation, (3) heating at 350°C, and (4) heating at 550°C. Normal involves allowing the prepared sample solutions to cool on a glass slide for 15 h. Glycolation requires the treatment of samples with ethyl glycol to expand the swelling clays such as smectite. Prepared samples were placed in a container containing glycol, covered, and placed in the oven for 12 h at 69°C. The third and the fourth treatment is the heating of the prepared samples in the furnance at 350 and 550°C for 30 min and 1.5 h, respectively, to reveal changes in the crystal structure. After each treatment (Normal – 550°C), the setup was placed in the X-ray machine for 12 min/sample at an angle of 20–30°, and the mineralogy was captured into the software (Expert high score v.3.0) attached to the machine. All X-ray diffractograms (Normal – 550°C) of each sample were compared to see any changes, which will lead to the determination of the list of the available clay minerals.

2.4 Statistical analyses

The data were carefully prepared, and outliers were winsorized. The data normality was ensured using the Kolmogorov–Smirnov test via OriginPro 2020. The classification of samples based on soil properties across slopes of different landslide conditions was evaluated by HCA and PCA using OriginPro 2020 and Statistica 13, respectively. In HCA analysis, Ward’s cluster method and Euclidean distance type were applied. Also, anion properties were excluded in HCA because of fewer variations across the slopes. However, all tested properties were used in the PCA. PCA also aids in visualizing the multivariate dataset as a system of coordinates in a high-dimensional data space. Furthermore, the correlation between variables was visually summarized using PCA. A one-way analysis of variance (ANOVA) and Pearson’s correlation analysis were performed using OriginPro 2020. ANOVA was used to evaluate the significance of the difference in soil properties between the LAS and NAS, which was conducted using Fisher’s least significant difference at a significance level of p = 0.05. The significance of the association among the soil properties (independent variables) and the relationship between these properties and LS (dependent variables) was determined using Pearson’s correlation analysis.

3.1 Field observation and soil properties of the study area

Two principal landslide types were observed in the field, namely, shallow and deep, both of which were characterized by the slide surface’s thickness. Shallow landslides are those with a slide surface of less than 4 m deep, while deep landslides are those with slide surfaces of above 4 m deep. Most sampled slopes are not highly vegetated and fall within the V–VI weathering grade. The weathering grade categories are those in which the rock is completely weathered with or without the original mass still in place. Field slope measurement showed that the degree of the slope of the entire sampled locations ranges from 30 to 80° with an average of 59.45°. The degree of the LAS and NAS ranges from 42° to 80° (mean = 63.69°) and 30–70° (mean = 51.57°), respectively. Laboratory analysis (See Table 1) revealed that the soil grain size distribution of this area is characterized by high sand content (57.13–80.08%, mean = 70.99%), low silt (2.22–23.11%, mean = 10.82%), and low clay (4.62–27.87%, mean = 18.19%). According to the USDA soil classification system, the gradation test of the soil reveals that the soil is dominated by sandy loam and sandy clay loam soil (Figure 4). The MC ranges from 13.41 to 30.62%, with an average of 20.2%. Other soil properties in the study area were OMC (1–5.31%), pH (4.05–5.21), and CEC (1.71–4.00 cmolc kg⁻¹). The prominent anions included SO 4 2 − , F⁻, Cl⁻, and NO 3 − , with average concentrations of 253.41, 23.06, 128.72, and 1134.68 mg kg⁻¹, respectively. The clay mineralogy obtained from XRD results showed an abundance of kaolinite and illite with traces of chlorite (see Table 2).

Figure 4

Soil type classification according to USDA standard.

3.2 Classification of soil sample based on field observation, HCA, and PCA

HCA (see Figure 5) and PCA (see Figure 6) classifications of soils across slopes of different landslide conditions coincide with field observations (see Table 3). Samples from slopes with no landslides (less susceptible) and those from landslide-affected hills (high susceptible) independently clustered on both HCA and PCA. While soils from less-susceptible slopes were grouped, landslide-affected hill samples clustered into two. The two variations of failed slope samples slightly coincide with landslide types (shallow and deep) based on the field observation (see Table 3). For convenience, samples were grouped into NAS and LAS. In this study, NAS samples are low-susceptible slopes and NAS are high-susceptible slopes.

Figure 5

Hierarchical cluster analysis using ward’s cluster method and Euclidean distance type (All tested parameters were used except anions and clay mineralogy). Observations 1–20 represent the soil samples.

Figure 6

Principal component analysis (all tested properties included except clay mineralogy). NAS depict non-landslide affected samples while LAS indicates landslide affected samples.

3.3 Comparison between soil properties of NAS and LAS

The one-way ANOVA results (see Table 4) revealed a significant difference (p < 0.05) between NAS and LAS samples regarding sand, clay, OMC, and CEC. The box plot (see Figure 7) shows differences between these four properties across the LS. The soil grain size distribution of NAS samples is characterized by low sand (67.09%), low silt (9.39%), and high clay content (23.52%). Furthermore, this class exhibits relatively high MC (21.06%), high OMC (3.52%), low pH (4.63%), and high CEC (3.42 cmolc kg⁻¹). However, the differences in the silt, MC, and pH are statistically insignificant across the LS. The anion concentrations in this group have relatively low sulfate (213.94 mg kg⁻¹), low fluoride (12.49 mg kg⁻¹), low chloride (56.88 mg kg⁻¹), and high nitrate (1179.89 mg kg⁻¹). Although anions vary across LS, this variation is statistically insignificant.

Figure 7

Box and whisker plot indicating differences in the parameters across slope classification. NAS depict non-landslide affected samples while LAS indicates landslide affected samples.

The grain size distribution of LAS samples is similar to the NAS soils but inversely proportional to each other. Sand is higher in LAS (73.86%), clay is lower (15.32%), and silt is not significantly different (11.59%). Also, LAS exhibits lower MC (19.74%), lower OMC (2.39%), higher pH (4.71%), and lower CEC (2.68 cmolc kg⁻¹). Regarding anions, LAS possesses higher fluoride (28.75 mg kg⁻¹), higher chloride (152.24 mg kg⁻¹), slightly lower nitrate (1110.32 mg kg⁻¹), and higher sulfate concentrations (274.67 mg kg⁻¹). Bromide and nitrite concentrations are negligible because, of 20 samples, they were detected only in 2 and 1 samples, respectively.

Pearson’s correlation analysis (See Table 5) shows that LS was moderately and negatively correlated (p < 0.05) with clay, OMC, and CEC, but correlated positively with sand. However, the MC, pH, and anion concentration do not have any linear relationship with LS. Sand has a strong negative association with clay content and CEC and, to a lesser degree, with OMC (p < 0.02). Also, OMC exhibited a moderate positive correlation with MC and CEC (p < 0.05).

4.1 Relationship between LS, physical-geochemical parameters, anion concentrations, and clay mineralogy

The Cameron Highlands in Malaysia are susceptible to landslides because of the nature of the soil, rapid urbanization, and frequent rainfall. However, a distinction in soil properties was noticed between the LAS, which is highly susceptible, and the NAS, which is less susceptible. Sand, clay, OMC, and CEC are critical because of their statistical association with LS. Soil texture is a physical property that determines the soil slope strength. Soils with a high sand percentage with no sufficient clay-sized soil to bind the solution ions and fill the void have larger free pore spaces and weak bonds. When this feature (high sand) exists in the upper part of the slope, it becomes vulnerable to landslides (especially if the deeper part of such slope is more clayey). Thus, slopes with such soil fail more easily when there is a sufficient trigger mechanism, such as rainfall. Moreover, the electrostatic charges on the clay surface attract and bind the solution ions, increasing cohesion between the soil particles and, consequently, increasing slope stability [30,31]. Thus, when the clay content in a soil slope is very small (especially in the first two horizons) as is the case in this study, there would not be enough negative charge to bind the solution ions. Hence, the soil body would be easily slipped with continuous heavy rain because of the lose bond and over saturation.

Othman et al. [32] suggested that a soil slope with less than 15% clay content is very prone to landslide. However, soil with higher clay in the deeper layers may increase landslide because of the lower infiltration rendering the less clayey upper (sandy) part vulnerable to failure after saturation. However, at the surface level, high clay concentration (depending on the type of the clay) has a potential of reducing slope failure because of a strong bond and under saturation. Also, sand allows higher water infiltration that could apparently reduce slope failure if the deeper part is not highly clayey. The landslide in the study area usually occurs during the monsoon period when there is intense continuous rainfall. As the slope becomes saturated due to prolonged rainfall, water cannot sink in anymore. The water would become movable with a high force within the slope due to the additional water from the rainfall, which would lead to slope failure. The positive association between sand and LS and the negative relationship between clay and LS reflect the link between soil texture and LS. This finding correlates well with the works of refs. [13,32].

Clay was negatively associated with sulfate, chloride, and nitrate. However, these anions appear not to directly influence LS in the study area; therefore, there could be other factors that influence their concentrations or the correlation may be random. However, SO 4 2 − is a known landslide hazard factor as it increases the swelling properties of the soil, weakening the cohesive force of attraction, thus leading to reduction in the compressive strength [33]. For example, ref. [34] found a positive correlation between SO 4 2 − concentrations in streams and the ratio of the landslide area to the catchment area, which signified a high probability of landslide occurrence in areas with high SO 4 2 − concentrations in water. Note that the hazardous effects of SO 4 2 − in the alteration of physical and engineering characteristics of soil depend on both the SO 4 2 − concentration and the chemical form and period of interaction [33]. The inverse relationship of sulfate with clay content may be an indication to the mechanism of the soil strength response to sulfate and clay concentrations [31,33]. It implies that soil slopes with low clay and high sulfate at the upper horizon have a higher tendency of failure.

The low OMC, CEC, and pH discovered here were aided by quick laterization, leading to the declination and, sometimes, the disappearance of primary minerals, which eventually contribute to the acidity and low organic materials of the soil [35]. High OMC in the soil contributes to the reduction in its acidity [36,37]. However, short-term changes in the soil OMC alone have a little effect on its pH because of the soil buffering capacity. Acidic soils are at risk of Ca, Mg, and K depletion. They exhibit a lower solubility rate of Mo and P compounds and a high rate of colloidal dispersion, lowering the concentrations of Mo, P, and Ca in acidic soils and, consequently, reducing the binding effect within the soil [38]. Soil with a high CEC can retain more cations (such as Ca, Mg, Na, K, and Al) to the exchange areas of OMC and clay, improving soil stability through the binding effect that results from the attraction between solution ions via electrostatic surface charges of clay in the soil [39]. Statistical analysis in this research shows that OMC and CEC were negatively interrelated with the sand content but positively with each other. The natural relationship between these parameters supports the statistical results, as elaborated on in the previous sentences. Thus, soil slopes with high OMC, high CEC, low sand, and high pH would be relatively stable.

The abundance of kaolinite and illite in the study area indicates the weathering of primary minerals of muscovite and feldspar that form the bedrock [40,41], implying that the rock is in the last stage of weathering, weakening the slope strength [12]. Illite clay originates from the weathering of rocks rich in K and Al, such as mica and K-feldspar. Similarly, kaolinite formation is related to the weathering of aluminosilicate minerals or hydrothermal alteration [42]. In the case of weathering as a form of origin of kaolinite, orthoclase feldspar is usually the weathered minerals. During the weathering process, silicon is replaced by aluminum and cations, such as Na, Ca, K, and Mg, and Fe is leached [43]. The lower cation concentrations, mentioned earlier, occasioned by the formation of kaolinite make the soil slope vulnerable to failure [32].

4.2 Mitigation measures

The physical and chemical properties of soil slopes in the Cameron Highlands are among the factors that cause frequent landslide occurrences. The causative factors identified must be improved to increase the mitigation of this hazard. Vegetation is critical in slope stability. A good root system architecture could increase the shear strength of weak soil, reducing the influence of rain and increasing the cohesiveness of the soil, thus preventing it from falling. Unfortunately, from our field observation, most slopes where samples were taken in the Cameron Highlands are not highly vegetated, exposing them to further risk of failure. Also, most slopes visited fall within the V–VI weathering grade according to the International Association for engineering geology and the environment classification; therefore, the rocks are in their last stage of weathering, exposing them to more risk of failure. Authorities should investigate and plant vegetation types using root system architecture that could stabilize the soil slopes based on their physical, geochemical, and mineralogical properties. Also, low-cost chemical alteration of the soil slope toward improving their strength will be an effective mitigation measure. If the identified causative factors in this study are improved, it will further assist in mitigating the hazard.

4.3 Future research directions

Future researchers may increase the number of samples and cover more natural slopes in Cameron Highlands to further confirm the claims in this research. Another interesting study in this area is the investigation of the vegetation types and plant root system architecture that could stabilize the soil slopes based on their physical, mineralogical, and geochemical properties. The influence of heavy metals, cations, and soil color variables of the study area on LS is another needed research. Also, low-cost chemical alteration of the soil slope toward improving their strength will be an exciting topic. Furthermore, the identification of factors that influences the soil properties (that reduces slope strength) reported in this article will be an interesting research. Apart from the Cameron Highlands, other soil slopes with similar characteristics that have undergone shallow and deep-seated landslides in Malaysia should be studied regarding physical, geochemical, and mineralogical properties. The results should be compared to ascertain the most contributing parameter(s) in the country. Also, there are very few soil mineralogical studies of Cameron highland. This is also a gap in the research that needs attention.