PREDICTION OF EARTHQUAKES MAGNITUDE USING GENERALIZED SPACE ̉̉ TIME AUTOREGRESSIVE MODELS (GSTAR)
Indonesia is an archipelago that located in the very complex and active tectonic area. This condition leads Indonesia into the world's highest seismic potential area. One of Indonesia region that frequently hit by earthquakes are Banda Sea region. In this research, it is predicted the strength...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/23543 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Indonesia is an archipelago that located in the very complex and active tectonic area. This condition leads Indonesia into the world's highest seismic potential area. One of Indonesia region that frequently hit by earthquakes are Banda Sea region. In this research, it is predicted the strength of the earthquake magnitude using Generalized Space Time Autoregressive (GSTAR) with assume that the spatial location are heterogeneous. The dependence and influence among observed location is pesented by weight matrix. In general, there are several weight matrix are used, but in this research the other weight matrix is defined. The weight matrix are determined by considering the location area at subduction zones, that are called the uniform-subduction weight matrix and the binarysubduction weight matrix. The subduction zone is a zone at the boundary between the plates which are convergent. The location of the research is arranged in the grids. Ten grid is selected to represent ten spatial location. Since there are observation with zero values then it is defined some data sets for modeling. First,the modeling involve all seismic data includes the zero values. Second, the data set consist of all seismic observation by replacing all the zero values with the random numbers between ( , which and is mean and the standard deviation of a magnitude below 4 SR. Third, all seismic observation below 3 SR are replaced by random numbers between 3 SR and 3.5 SR. Fourth, seismic observation which under 3.5 SR are replaced by random numbers between 3.5 SR and 4 SR. For prediction, it is used three scenarios. The best model obtained by model which used subduction weight matrix. Based on Invers of Autokovariace Matrix (IacM), it is concluded that Data Set 1 is not stasionary, so that GSTAR model is not appropriate. The better model obtained for Data Set 2 and Data Set 3 is GSTAR (1; 3) by using binary-subduction weight matrix, the appropriate model to predict Data Set 4 is GSTAR(1; 2) by using uniform-subduction weight matrix . |
|---|