The study was carried out in the Mandi geographical area, which is situated in Himachal Pradesh, India’s mid-Himalayan region (see Figure 1). Our research specifically concentrated on the region along the route that connects Mandi and Kamand; the coordinates of this area are from 76°55′54′E to 76°59′42′E and from 31°42′25°N to 31°46′26°N. The research area was situated between 850 and 1,250 m above sea level. The frequent landslides in this area were the driving force behind our decision to research in Mandi (Kahlon et al., 2014). These landslide occurrences have sparked worries about the possible impact of precipitation, especially in a region where 1,380 mm of precipitation falls on average annually (Kahlon et al., 2014). We determined that 55 places along the 25-km route in Mandi, Himachal Pradesh, were landslide-prone because they had recently had landslides. Two different kinds of data were gathered from the landslide areas for our investigation. There are two types of soil features: one is based on TK, and the other is geotechnical. Below are the specifics of both data collection methodologies.

Twelve experts from a reputable university, aged 21–33 years (mean = 29, SD = 2.8), participated in this study. All experts provided informed consent before their participation. Experts included a mix of geotechnical engineers, environmental scientists, and individuals with expertise in traditional soil classification methods such as TK. The experts held either master’s or doctoral degrees and specialized in civil engineering, geotechnical engineering, and electronics and communication engineering. Seven experts had extensive research experience in geotechnical engineering and slope-stability analysis, whereas the remaining five had field experience in landslide risk assessment, soil mechanics, geomorphological analysis, and traditional soil classification techniques. Experts were selected based on the following criteria: (1) a minimum master’s degree in geotechnical engineering, civil engineering, or other engineering disciplines; (2) research experience in landslide risk assessment and soil characterization; and (3) familiarity with traditional knowledge systems. Their combined academic and field experience facilitated the integration of conventional geotechnical assessments with TK-based soil classification, enabling a comprehensive analysis of soil-environment interactions. The experts conducted field assessments at 55 randomly selected landslide-prone sites and graded soils based on TK principles. The twelve experts evaluated distinct soil properties, including texture, color, and scent, and systematically recorded their observations using a structured questionnaire. To ensure consistency and minimize subjective bias, the assessments were conducted independently by each expert, and inter-rater reliability was analyzed to confirm agreement in soil classifications. This structured approach allowed for direct comparisons between conventional geotechnical features and traditional soil attributes, bridging the gap between conventional engineering assessments and TK soil categorization. Their collective expertise in both traditional and conventional soil characterization methods brought valuable perspectives to the study, allowing for a more comprehensive evaluation of soil properties and their implications for landslide susceptibility. By integrating traditional knowledge with conventional geotechnical assessments, this study highlighted the potential of TK as a complementary tool for rapid soil assessment in landslide-prone areas.

The soil samples from 55 locations were collected at a maximum depth of 0.5 m along sloping surfaces for geotechnical analysis. Each sample included in situ moisture content, geographical coordinates, and recent landslide locations (Vogel et al., 2023). Soils were extracted by using a stainless-steel sampler (10 cm diameter and 13 cm in height) between 0.3 and 0.5 m depths and with the help of a trowel for rocky strata, weighing approximately 40 kg for each sample. The samples were analyzed in the laboratory for 16 geotechnical features such as in situ bulk density, in situ water content (wn), plasticity index, liquid and plasticity limits, saturation permeability, and more (ASTM Committee D-18 on Soil and Rock, 2009; American Society for Testing and Materials. Committee D18 on Soil and Rock, 2004; Nhantumbo and Cambule, 2006; Nguyen et al., 2022). In addition, the standard Proctor test was conducted to obtain optimum moisture content and dry density. Relative compaction was estimated by comparing in situ and laboratory dry density. The in situ density was used to compute the porosity. The geotechnical variables measured include gravel (%), sand (%), clay (%), silt (%), liquid limit, plastic limit, density, fine content (%), natural water content (%), specific gravity, and porosity (refer to Table 1).

We performed a Pearson’s correlation analysis to establish the strength of relationships between traditional soil characteristics of color, texture, smell, and taste and geotechnical properties of moisture content, density, and gravel content. Correlations were considered statistically significant using a two-tailed test of significance at p < 0.05. Strong correlations, r ≥ 0.5, were expected to confirm how well the traditional and conventional soil categorizations agreed. For example, we predicted that brown soil would correlate positively with density and moisture content, indicating compactness, while white soils, which are loosely structured, would negatively correlate with gravel content.

To further interpret the contribution of soil to slope instability, we used established literature thresholds and mean values observed in our dataset to qualitatively categorize the landslide susceptibility of each geotechnical feature as low, moderate, and high.

Soils with high porosity of more than 60% have lower particle packing and higher void ratios; hence, they are more susceptible to increase in water infiltration and pore pressure accumulation, which is a critical precursor to slope failure under rainfall conditions. With a mean porosity of 66.6%, these soils thus fall in the category of having high landslide susceptibility (Rahardjo et al., 2007; Sidle and Ochiai, 2006).

Fines content close to 32% represents a transitional soil system of silt and clay that retains water but without the cohesion attributed to high-clay content soil. Consequently, such soils are moderately susceptible to instability, especially under wet-dry cycles and gravitational loading (Thevanayagam, 1998).

A plastic limit of ∼16% would indicate that the soil changes from semi-solid to plastic behavior at a moderate level of moisture. These soils have a tendency to show reduced shear strength under increasing saturation and, therefore, moderate susceptibility (Van Westen et al., 2008).

The liquid limit of 24.4% falls within the range typically associated with soil softening under rainfall and seismic triggers, thus classified as high susceptibility. Clays with LL >20% have been linked to reduced stability during extreme hydrological events (Varnes, 1958).

Plasticity index values of 8%–12% represent transition soils that are neither excessively plastic nor brittle. Accordingly, they are moderately susceptible to deformation by the action of applied external stressors (Skempton, 1953).

An average saturated water content of 12.5% indicates that these soils are capable of developing high moisture retention, with an associated gain in weight and loss of matric suction, which may well increase the potential for slope failure under sustained rainfall. It is considered to be of high susceptibility (Fredlund and Rahardjo, 1993).

While optimum moisture content at 10.3% also indicates moderate risk, it is favorable for compaction in dry conditions, but saturation beyond this point can destabilize the soil matrix and increase landslide risk (Das and Sobhan, 2006).

A sand content over 50% improves permeability normally and enables good drainage, hence a minimal development of pore pressure. Since the mean sand content is 52.4%, these soils are to be classified as having low susceptibility (Bishop, 1955).

Saturated unit weight of 15.9 kN/m³ indicates moderately dense soils, which, when saturated, may develop considerable driving forces downslope and justify a moderate risk classification (Das and Sobhan, 2006).

A specific gravity of 2.6 corresponds to a stable silicate mineralogy with coarse fractions, which indicates structural resistance and effective drainage. This corresponds to low susceptibility according to (Das and Sobhan, 2006).

Gravel content contributes to macro-porosity and facilitates drainage, lowering water retention and pore pressure accumulation, with a mean of 14.8%. Consequently, these soils are rated as low susceptibility (Macciotta and Hendry, 2021). Particle size parameters: D10 = 0.4 mm and D60 = 2.3 mm indicate that the well-graded soil allows for good compaction, reducing the risks of internal erosion, hence supporting low susceptibility. On the other hand, D30 = 1.6 mm and D50 = 2.2 mm indicate intermediate grading, which may show some sensitivity under stress and hence are rated as moderate susceptibility according to USDA Soil Survey (Garc and Frankenstein, 2015). Hence, the susceptibility classification in Table 1 integrates these parameter thresholds and their known geomaterial behaviors to qualitatively indicate the relative potential for landslide initiation in the study area. This context strengthens the engineering geological relevance of the dataset by relating lab-measured soil properties to field-observed landslide patterns in a Himalayan terrain dominated by colluvial and weathered metamorphic formations.

A structured questionnaire was developed based on the principles of traditional soil classification to gather expert evaluations and quantifiable data (see Table 2). As shown in Table 2, the questionnaire assessed soil properties, including soil color (e.g., red, white, yellow), smell (e.g., earthy and musty), texture (e.g., stickiness, smoothness), and taste (e.g., sweetness and saltiness), using a 5-point Likert scale (+2 = significant agreement, +1 = agreement, 0 = neutrality, −1 = disagreement, −2 = significant disagreement) for standardized expert responses (Alabi and Jelili, 2023; South et al., 2022; Kumar and Brijesh, 2023). The 32 questions included in the questionnaire were carefully selected following a comprehensive review of the traditional texts and contemporary literature on soil properties. Experts in geotechnical engineering and environmental science evaluated the relevance of each question to ensure alignment with both the traditional and conventional soil classification methods. For more details on the questionnaire, refer to the Supplementary Material.

Before the main data collection, a calibration session was conducted with all the participating experts to standardize the interpretation of qualitative descriptors, such as color, smell, taste, and texture. Experts were trained on how to assign ratings using the five-point Likert scale (−2 to +2) and were shown representative soil samples (black, white, and brown) to achieve a consensus on interpretation. This session ensured that the qualitative evaluations were harmonized across experts. To validate the questionnaire, two geotechnical experts independently reviewed the 32 questions for clarity, relevance, and comprehensiveness before full implementation. Each question was rated for validity and consistency using a 5-point Likert scale, and the inter-rater reliability was assessed using Cronbach’s alpha, which was computed as 0.86, indicating a high level of agreement among experts. Additionally, Cronbach’s alpha for the full dataset of 660 responses across 32 items was 0.77, further confirming the reliability of the questionnaire. This validation process ensured that the questionnaire effectively captured the intended soil characteristics while maintaining consistency in the expert assessments.

Experts administered the questionnaire at randomly selected landslide-prone sites to evaluate soil fertility, moisture content, water retention, drainage capacity, stickiness, hardness, and disturbance response. The effects of vegetation cover on soil stability were also analyzed. Soil color was noted to assess both organic and mineral components, providing insight into soil composition. These expert evaluations, guided by structured criteria, ensure consistency across assessments and enable objective comparisons between soil samples.

The structured questionnaire provided a standardized framework for evaluating soil characteristics, enabling comparisons across diverse landslide-prone sites, while minimizing variability in expert assessments. Given the challenges of conducting direct measurements in remote locations, the questionnaire facilitates an efficient and field-adaptable method for soil classification. This approach also incorporates diverse disciplinary perspectives, as experts from geotechnical, and civil engineering backgrounds apply a consistent evaluation framework to the same soil samples.

All statistical analyses were performed using SPSS software (IBM Corp, 2016). To determine the suitability of the data for factor analysis, the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett’s Test of Sphericity were employed (Gupta and Shukla, 2018). The KMO measure assesses the proportion of variance among variables that might be a common variance, indicating the appropriateness of factor analysis. A KMO value above 0.5 was considered acceptable for proceeding with Exploratory factor analysis (EFA) (Rahardjo et al., 2007). Bartlett’s Test of Sphericity tests the null hypothesis that the correlation matrix is an identity matrix, which would indicate that variables are unrelated and unsuitable for structure detection. A significant result (p < 0.05) from Bartlett’s test suggests that there are adequate relationships among variables for factor analysis (Gupta and Shukla, 2018). All traditional knowledge (TK) descriptors were transformed into numeric values using a five-point Likert scale (−2 to +2), representing the ordered intensity of each sensory attribute. These ordinal scores were treated as quasi-interval data, suitable for multivariate analysis. Significant correlations (r ≥ 0.3, p < 0.05) confirmed that the Likert scale transformations captured meaningful gradations in soil properties. The suitability of the transformed three datasets for factor analysis was further verified using KMO (>0.7) and Bartlett’s test (p < 0.01).

To minimize inter-rater bias among experts with different disciplinary and educational backgrounds, individual questionnaire scores were recorded independently and later aggregated using the mean Likert value for each descriptor. This approach reduces individual influence while preserving the collective trend. Before data collection, a calibration session was held with all experts to standardize the rating criteria and align the interpretation of qualitative descriptors, functioning similarly to the Delphi-style consensus process.

The EFA was executed using Principal Component Analysis (PCA) for factor extraction, followed by varimax rotation to achieve a clearer factor structure (Dziuban and Shirkey, 1974). EFA was executed using Principal Component Analysis (PCA) for factor extraction, followed by varimax rotation to achieve a clearer factor structure. The stability of the extracted factors was evaluated by comparing the loading patterns obtained from the traditional, geotechnical, and combined datasets. The factors consistently aligned with the same soil categories (fine-grained, coarse-grained, and earthy), confirming the stable structural relationships. The internal consistency of each factor was verified using Cronbach’s alpha, which ranged from 0.75 to 0.86. Given the moderate sample size, confirmatory factor analysis (CFA) was deferred to future studies with larger datasets. The EFA was chosen for its ability to identify latent constructs and relationships within the dataset, which included both traditional and geotechnical soil properties (Gupta and Shukla, 2018). This method is particularly suitable for datasets in which multiple variables are expected to be grouped into distinct underlying factors, such as fine-grained, coarse-grained, and earthy soil categories (Gupta and Shukla, 2018). The traditional feature-based dataset consists of 660 data points derived from expert evaluations across 55 landslide sites (with twelev experts assessing each site). The geotechnical features dataset, on the other hand, comprised 55 laboratory analyzed soil samples collected from these same sites. In addition, the sample size met the recommended thresholds for EFA, ensuring the reliability of the extracted factors. By uncovering the underlying structure of soil properties, EFA allows for the development of a robust soil categorization framework that effectively integrates traditional and geotechnical characteristics.

The eigenvalues represent the amount of variance explained by each factor (Shrestha, 2021). Following the Kaiser Criterion, factors with eigenvalues greater than 1.0 were considered significant, as these factors explain more variance than an individual variable (Yong and Pearce, 2013). However, the ultimate threshold for retention was the cumulative variance explained by these factors. For each dataset, only the factors that collectively explained greater than 55% of the total variance were retained. Scree plots were also used to visually confirm the number of retained factors (Howard, 2016). This combined approach ensures that the retained factors meaningfully contribute to the overall explanatory power of the analysis, while minimizing noise.

Next, one-way ANOVA was used to explore the potential impacts of soil factors identified by EFA and to investigate the differences in factor scores among the demographic and expert groups, focusing on age and expertise level (education level and discipline) (Belkhiri and Narany, 2015).

The results revealed several significant correlations, highlighting the interdependencies among soil color, composition, water retention, and geotechnical properties. Brown soil showed a significant positive correlation with rocks and debris (r (658) = 0.112, p < 0.05) and density (r (658) = 0.351, p < 0.01), suggesting that denser soils with rock fragments tended to exhibit brown coloration, which may indicate higher compaction levels. In contrast, white soils were positively correlated with silt content (r (658) = 0.267, p < 0.01) and negatively correlated with gravel content (r (658) = −0.202, p < 0.01), indicating that soils with high silt proportions and fewer coarse particles appear lighter in color. Soil water retention exhibited a strong positive correlation with earthy soil smell (r (658) = 0.483, p < 0.01) and moisture content (r (658) = 0.617, p < 0.01) but was negatively correlated with sand content (r (658) = −0.417, p < 0.01). These results suggest that soils with high organic content and finer particles tend to retain more water, whereas sandy soils drain moisture more rapidly. The specific gravity was significantly correlated with cohesion (r (658) = 0.521, p < 0.01), supporting the notion that denser soils tend to have stronger particle bonding and higher shear resistance. Stickiness was positively correlated with silt content (r (658) = 0.352, p < 0.01), suggesting that fine-grained soils exhibited greater plasticity and cohesion, reinforcing their vulnerability to landslides under wet conditions. Additionally, silt content was strongly correlated with the plasticity index (r (658) = 0.239, p < 0.01), reinforcing the importance of fine-grained soils in determining soil deformation characteristics. Soil drainage exhibited a strong positive correlation with soil density (r (658) = 0.432, p < 0.01), indicating that denser soils may exhibit better compaction and structural stability, and reduce permeability. Smoothness was significantly positively correlated with the sweet taste of soil (r (658) = 0.137, p < 0.01), which aligns with the presence of certain mineral compositions possibly linked to organic-rich soils. Additionally, vegetation cover was negatively correlated with soil thickness (r (658) = −0.276, p < 0.01), implying that thicker soil layers may have lower vegetation cover owing to potential instability or poor nutrient retention. Soil clumping was positively associated with the average internal friction (r (658) = 0.256, p < 0.01), indicating that well-aggregated soils may resist shear forces more effectively.

Using the questionnaire, twelve experts assessed each of the 55 locations, generating a total of 660 responses. An EFA was performed on these 660 responses, and the results are presented in Table 3 (Mvududu and Sink, 2013). Before conducting EFA, the suitability of the data was assessed. The KMO measure of sampling adequacy was 0.897, which confirmed that the sample size was appropriate. Bartlett’s Test of Sphericity was significant (χ² = 2,137.6, p < 0.01), suggesting that the dataset exhibited sufficient intercorrelations for factor extraction. PCA was employed for factor extraction, initially identifying four factors. However, factors that explained less than 55% of the total variance were excluded, resulting in three retained factors. KVN rotation was applied to optimize factor separation. Figure 2 shows the scree plot, where the eigenvalues dropped sharply after the third factor, confirming the retention of the three dominant factors. The final model retained three factors, which explained 65.17% of the total variance. The eigenvalues were 4.8, 4.7, and 3.8, corresponding to fine-grained soil (Factor 1), coarse-grained soil (Factor 2), and earthy soil (Factor 3), respectively, as detailed in Table 3. The factor loadings ranged from 0.504 to 0.838, while communalities varied between 0.294 and 0.926 (Table 4), indicating strong associations among the extracted variables. To interpret the retained factors, factor loadings were analyzed to identify the dominant soil characteristics. Factor 1, representing fine-grained soil, showed high loadings for variables such as clay content, black color, and water retention, which aligns with traditional classification of moisture-retentive and fertile soils. Factor 2, associated with coarse-grained soils, included high loadings for sand content, gravel, and coarseness, confirming the alignment with well-drained, low-cohesion traditional soil descriptions. Factor 3, representing earthy soils, exhibited high loadings for brown color and earthy smell, reinforcing its classification as an organic-rich soil with moderate cohesion.

Fine-grained soils included clay, silt, black color, high water retention, moisture content, high fertility, and a musty smell (Table 5). These attributes align with TK descriptions of fertile and moisture-retentive soils that support plant growth, but may experience stability issues under saturation. Coarse-grained soils comprised sand, gravel, rocks, debris, yellow and white colors, coarseness, easy crumbling, stickiness, healthy plants, and high-water drainage (Table 3). The presence of yellow and white coloration in conjunction with coarse particles suggests that these soils are well drained and prone to surface erosion rather than deep-seated movement. Earthy soils are characterized by a brown color, earthy smell, and moderate water retention. The significant correlation between earthy smell and water retention (r (658) = 0.483, p < 0.01) applies specifically to earthy soils, reinforcing their classification as moderately moisture-retentive with respect to organic content (Table 5). This classification aligns with TK description of ordinary soils that exhibit moderate stability but require external reinforcement for sustained fertility and structure. These findings validate the alignment between traditional soil classifications and conventional geotechnical assessments, demonstrating that qualitative attributes, such as color, smell, and texture, are strongly associated with measurable soil properties.

The geotechnical characteristics of the soil samples were analyzed using EFA on 55 laboratory test results to identify the significant underlying factors (Mvududu and Sink, 2013). The extraction results are summarized in Table 6. PCA was employed for factor extraction, with a KMO measure of sampling adequacy of 0.506, confirming that the sample size was acceptable for the factor analysis. Bartlett’s Test of Sphericity was significant (χ² = 462.3, p < 0.01), indicating that the correlation matrix was not an identity matrix and that factor analysis was appropriate. Factors that explained greater than 55% of the total variance were retained. The EFA of the geotechnical dataset identified three factors that explained a cumulative variance of 68.32%. KVN rotation was applied to improve factor interpretability (Table 6). Figure 3 illustrates the scree plot, which demonstrates a significant decline in eigenvalues after the third factor, thus supporting the three-factor solution. The three extracted factors were classified as fine-grained, coarse-grained, and earthy, based on the dominant geotechnical properties loading onto each factor. A factor loading threshold of 0.5 was applied, ensuring that each item loaded uniquely onto one factor without cross-loading (Table 7). The factor loadings ranged from 0.573 to 0.943, with communality values between 0.415 and 0.848, indicating a strong variable association with the respective factors. The interpretation of these factors was guided by factor loadings, with higher loading values indicating stronger contributions from specific geotechnical properties. Factor 1 (fine-grained soil) included cohesion, plasticity index, and specific gravity, confirming its classification as highly cohesive soil with high water retention. Factor 2 (coarse-grained soil) had high loadings for sand content and friction, reinforcing its characterization as a well-drained, granular soil. Factor 3 (earthy soil) was associated with water content and moderate cohesion, aligning with the traditional descriptions of soils with organic content and moderate stability.

Fine-grained soils were characterized by cohesiveness, specific gravity, and plastic limits (Table 8). The correlation between specific gravity and cohesion (r (658) = 0.591, p < 0.01) reinforces the geotechnical properties of fine-grained soils, highlighting their compact nature and resistance to structural failure under dry conditions. However, their high plasticity index and low permeability render them prone to reduced shear strength under saturation conditions. Coarse-grained soils comprised silt and the plasticity index, indicating their role in the soil deformability and drainage properties. The strong correlation between silt content and plasticity index (r (658) = 0.573, p < 0.01) suggests that fine material within coarse-grained soils contributes to their cohesion and water retention capacity, thereby affecting slope stability. Earthy soils exhibit sand, friction, and water content, confirming their role in soil drainage and surface erosion. These findings confirm that fine-grained soils, owing to their high cohesion and plasticity, require targeted stabilization strategies to mitigate landslide risks. Similarly, coarse-grained soils retain silt, which contributes to instability under fluctuating moisture conditions, despite their high drainage capacity.

The EFA was conducted on 660 questionnaire responses and 55 laboratory test results to identify the shared structural components across the traditional soil classifications and geotechnical properties (Mvududu and Sink, 2013). The extraction results including eigenvalues are presented in Table 9. The KMO measure of sampling adequacy was 0.788, confirming that the combined dataset was suitable for the factor analysis. Bartlett’s Test of Sphericity was significant (χ² = 1685.4, p < 0.01), ensuring that inter-variable correlations justified factor extraction. PCA was applied for factor extraction, initially identifying four factors. However, based on the scree plot and total variance contribution, factors explaining less than 55% of the total variance were excluded. Figure 4 illustrates the scree plot, which shows a sharp eigenvalue drop beyond the third factor, justifying the retention of the three dominant factors. The final EFA model retained three factors that collectively explained 71.46% of the total variance (Table 9). The factor-loading analysis followed the same procedure as in the previous sections, with items loaded onto a given factor if they exhibited a factor loading ≥0.5 and did not cross-load onto other factors (Table 10). The retained factors were classified as fine-grained, coarse-grained, and earthy soil types, aligned with both the traditional and geotechnical soil categorizations. The combined EFA dataset demonstrated a strong alignment between the traditional attributes and conventional geotechnical properties. For instance, factor loadings for soil color and moisture retention in the traditional dataset were strongly associated with geotechnical properties, such as water content and cohesion, confirming the validity of integrating traditional and quantitative methods. The correlation between earthy smell and water retention (r (658) = 0.332, p < 0.01) further supports the reliability of qualitative assessments for landslide-prone soil classification.

The composition of each extracted factor is listed in Table 11. Fine-grained soil (Factor 1) accounted for the highest variance (24.86%), followed by coarse-grained soil (18.55%), and earthy soil (14.32%). Fine-grained soil exhibits properties such as water retention, stickiness, smoothness, hardness, soil clumping tendency, and silt content. A significant correlation was observed between stickiness and silt content (r (658) = 0.450, p < 0.01), thereby reinforcing the role of fine particles in soil cohesion and permeability. This supports the traditional classification of fertile moisture-retentive soils, which exhibit high plasticity under saturation, making them susceptible to landslides in wet conditions. Earthy soil includes density, earthy smell, musty smell, plastic limit, and high moisture content. A significant correlation was found between the earthy smell and plastic limit (r (658) = 0.312, p < 0.01), suggesting that organic-rich soils retain moderate plasticity, water content, and balance stability with permeability. This aligns with traditional description of transitional soils that exhibit mixed characteristics of moisture retention and structural variability under wet-dry cycles. Unlike fine-grained and earthy soils, no statistically significant correlations were observed for coarse-grained soils. This suggests that coarse-grained soils, which include components such as sand, gravel, and rocks, behave more independently of finer geotechnical and traditional soil attributes. Their properties are largely governed by the granular structure rather than cohesive interactions, reinforcing their high permeability and low susceptibility to plasticity-related instabilities. These findings demonstrate that transitional qualitative soil classifications closely align with measurable geotechnical parameters in fine-grained and earthy soils, reinforcing the importance of integrating traditional ecological knowledge into conventional landslide risk assessments.

A one-way ANOVA was conducted to examine the impact of the demographic variables (age, discipline, and education level) on the three soil types (fine-grained, coarse-grained, and earthy soils) (see Table 12).

Age did not significantly affect soil classification across fine-grained (F (2, 657) = 0.761, p = 0.51, η ² = 0.01), coarse-grained (F (2, 657) = 0.761, p = 0.41, η ² = 0.01), or earthy soils (F (2,657) = 0.500, p = 0.61, η ² = 0.01). This suggests that soil categorization was not influenced by age differences among the experts, which is consistent with previous findings that expertise-based assessments remain consistent across age groups.

Discipline (civil engineering (CE), computer science engineering (CSE), and electronics and communication engineering (ECE)) significantly influenced fine-grained soil categorization (F (2, 657) = 2.9, p < 0.05, η ² = 0.01). Tukey’s post hoc test (see Table 10) indicated that CE participants assigned significantly higher scores to fine-grained soils than ECE participants (CE (M = 16.1, SD = 2.3) > ECE (M = 14.3, SD = 2.1), p < 0.05). However, no significant differences were observed between CE and CSE (p = 0.56) or between CSE and ECE (p = 0.15). These findings highlight that CE participants, trained in geotechnical principles, prioritized fine-grained soils more than ECE participants, who may have placed less emphasis on soil cohesive properties. Similarly, discipline significantly influenced the classification of earthy soils (F (2, 657) = 3.9, p < 0.05, η ² = 0.03). ECE participants rated earthy soils significantly higher than CE (ECE (M = 9.2, SD = 0.6) > CE (M = 7.3, SD = 1.3), p < 0.05) and CSE (ECE (M = 9.2, SD = 0.5) > CSE (M = 7.2, SD = 1.4), p < 0.05), whereas CE and CSE did not differ significantly (p = 0.99). These results suggest that ECE experts may have been more inclined to consider traditional qualitative attributes, such as smell and texture, leading to a stronger emphasis on earthy soil classifications. For coarse-grained soil, no significant effect of discipline was found (F (2, 657) = 1.567, p = 0.25, η ² = 0.01), indicating that disciplinary training did not influence classification. This may be attributed to the straightforward nature of coarse-grained soils, which exhibit clear geotechnical properties, such as high permeability and low cohesion, reducing the likelihood of subjective interpretation.

Education level (Ph.D., postgraduate (PG), or graduate (G)) did not significantly influence the classification of fine-grained soils (F (2, 657) = 1.662, p = 0.23, η ² = 0.01) or coarse-grained soils (F (2, 657) = 0.267, p = 0.92, η ² = 0.01), suggesting that expertise across these levels resulted in similar assessments for these soil types. However, education level significantly affected the earthy soil classification (F (2, 657) = 2.892, p < 0.05, η ² = 0.04). Post-hoc analysis (Table 10) showed that PG participants rated earthy soils significantly higher than both G (PG (M = 9.1, SD = 0.8) > G (M = 8.1, SD = 1.3), p < 0.05) and Ph.D. participants (PG (M = 9.3, SD = 0.9) > Ph.D. (M = 7.5, SD = 1.3), p < 0.05). No significant difference was found between PhD and G participants (p = 0.89). This suggests that PG participants who are at an intermediate level of specialization may have been more open to integrating both traditional and geotechnical attributes in their assessment of earthy soils.