| Popis: |
Background: Petrographical analysis is a visual method to describe the mineral properties of a rock sample, related to its composition and classification. The mineral composition is mostly determined using a comparator, thus it is merely a quantitative method. This paper uses multivariable statistics as tools to support the visual classification. Materials and method: We took 89 thin sections from five locations, based on secondary data, gathered from five final projects from Department of Geology ITB: Mt. Lamongan-Probolinggo (LAM), Mt. Banyuresmi-Bogor (BAN), Mt. Kromong-Palimanan (KRM), Mt. Sangkur- Garut (SAN), and Mt. Wangi-Purworejo (WAN). The locations is Quaternary volcanic (LAM, KRM, SAN, and WAN) and Tertiary (BAN), consists of 85 igneous rock samples, three pyroclastics, and one sedimentary rock sample. The mineral percentage data was then analyzed using Principal Component Analysis (PCA) and Cluster Analysis (CA) with R. Our data frame size was 89 rows x 24 columns. Results and discussions: Our preliminary results show a separation between Quartenary and Tertiary samples on PCA and CA plots. We identified three clusters: cluster 1 BAN; cluster 2 LAM, KRM, SAN, WAN, and BAN, and cluster 3 LAM, KRM, SAN, WAN, and BAN. Based on our PCA plot, cluster 1 shows a strong influence of: chlorite, calcite, and quartz; cluster 2: piroxene, plagioklas, olivine, glass and cluster 3: olivine, piroxene, opaque, glass, and feldspar. Following the results we see that there are samples from BAN detached from the rest. We believe this is due to an anomalous volcanic process on that spot. On the other hand we see the similarity between samples from different locations. Based on those analyses, we perceive that multivariable statistics can support visual petrographic analysis by reading the data structure, selecting stronger variables and extracting clusters. |
