Hierarchical Cluster Analysis of Dense GPS Data and Examination of the Nature of the Clusters Associated With Regional Tectonics in Taiwan

Abstract

Collision of the Luzon Volcanic Arc is the primary causative factor behind the creation of Taiwan Island. Additionally, the indention of the Peikang basement highs in the Taiwan Strait and the back arc spreading of the Okinawa Trough are responsible for the characteristic tectonics in Taiwan. Identification of active tectonic boundaries is important for understanding neotectonics in Taiwan. We analyzed Global Positioning System (GPS) horizontal velocity data in Taiwan with a hierarchical cluster analysis method. We also developed a statistical scheme introducing information entropy to quantitatively assess the fuzziness of the obtained cluster boundaries, which enables us to distinguish a trustworthy discontinuity. The hierarchical representation, which is expressed by a dendrogram, suggests the relative importance of tectonic sources in this region. The major clusters, which appear at higher hierarchy, are likely to represent major tectonic features such as the Longitudinal Valley fault and lateral extrusion in southwest Taiwan. Those at lower hierarchy suggest velocity discontinuities associated with local geological structures such as active faults. For example, boundaries in east offshore Taiwan and southern Taiwan are identified without using geological information by this analysis. In addition to the identification of these clear tectonic boundaries, our statistical cluster assessment shows the existence of fuzzy cluster boundaries along the Coastal Plain and the Western Foothills, reflecting continuum deformation in west Taiwan.

Cluster Analysis of Dense GNSS Velocity Field Reveals Characteristics Associated With Regional Tectonics in New Zealand

“…In order to identify an important tectonic structure, the stability of cluster allocations have to be quantified. To this end, we follow the cluster stability assessment method in Takahashi et al (2019). In this section, we introduce and demonstrate the method by using an artificial data set.…”

“…Following Takahashi et al. (2019), we define the stability of a station with a sum of the information entropy, $${S}_{i}$$, as follows. $$$\begin{array}{c}{S}_{i}=\sum\limits _{j=1}^{N}{S}_{ij}\end{array}$$$ where $$$\begin{array}{c}{S}_{ij}=-{p}_{ij}\,{\mathrm{log}}_{2}{p}_{ij}-\left(1-{p}_{ij}\right){\mathrm{log}}_{2}\left(1-{p}_{ij}\right)\end{array}$$$ and $$$\begin{array}{c}{p}_{ij}=\frac{{R}_{ij}}{M}\end{array}$$$ …”

“…To this end, we follow the cluster stability assessment method in Takahashi et al. (2019). In this section, we introduce and demonstrate the method by using an artificial data set.…”

Self‐Potential Ambient Noise and Spectral Relationship With Urbanization, Seismicity, and Strain Rate Revealed via the Taiwan Geoelectric Monitoring Network

“…Until 2019, the GPSLAB (http://gps.earth.sinica.edu.tw) has been managed more than 500 GPS stations operating in Taiwan. The GPS data were employed to perform numerous studies regarding crustal deformations and plate motions in Taiwan (Takahashi et al, ; Yu et al, , ; Yu & Kuo, ).…”

Cluster Analysis of Marine Terraces and Quantitative Seismotectonic Interpretation of the Boso Peninsula, Central Japan

“…The K‐means is considered as an unsupervised machine learning method. The cluster analysis has recently been utilized in various geophysical fields, such as in geodesy (Shimpson et al, 2012; Takahashi et al, 2019), structural geology (Eymold & Jordan, 2019), and seismology (Perol et al, 2018). We used a class of the K‐means clustering called a model‐based clustering, which assumes, as prior information, a specific structure in the spatial distribution of the point cloud data set.…”