Global Journal of Human-Social Science, B: Geography, Environmental Science and Disaster Management, Volume 22 Issue 3

Table 6 presents the results of the modeling of the soil-water characteristic curves. The characteristic curves of the studied soils present high suction values and air intake points, since the more weathered the soil, the higher the presented suction values are (Boszczowski and Ligocki, 2012). The shape of the curves suggests a bimodal behavior, indicating that both micropores and macropores control water inlet and outlet flows (Feuerharmel et al., 2006). The high suction values obtained are also similar to those obtained by Marinho and Stuermer (2000) when they studied a mature residual soil of Gnaisse (45% clay), compacted in normal and modified Proctor energies, obtaining air intake values ranging between 1000 and 2000 kPa and residual suction of 15,000kPa. b) Cluster Analysis The dendrogram created by processing the codes in Python resulted in the hierarchical structure presented in Figure 3. Seeking to understand how the algorithm effectively grouped the data, a top-down dendrogram analysis was carried out, i.e., noting which geotechnical variable or variables may have been used as a soil divider in each Cluster. For this, three cut-off points were defined in the dendrogram, as highlighted in the figure, and the clusters created were identified. It is emphasized that the clusterization obtained went through a much more complex process than is described here, since more than fifty geotechnical attributes were considered. Figure 3: Analysis of the dendrogram of the hierarchical clustering of the soils of this research Initially, in the first cut-off point, the algorithm divided the soils into two large groups (A and B), which may have been formed considering the results of the compaction test (optimum moisture and maximum dry apparent specific mass), since, as can be seen in the graph of Figure 4, there is a clear division of these groups through the compaction curves. Thus, it is observed that group A is composed of soils with higher values of optimum moisture (Wo > 16%) and lower values of ρ dmax (< 1,8 g/cm³) and group B of soils with Wo < 16% and ρ dmax >1,8 g/cm³), It was also noted that group A is composed entirely of USCS's ML class soils that take into account LL and PI. Volume XXII Issue III Version I 18 ( ) Global Journal of Human Social Science - Year 2022 © 2022 Global Journals B Clustering of Fine-Grained Tropical Soils using Data Science Tools Applied to their Geotechnical Properties