EARTHQUAKE RISK LEVEL ANALYSIS ON JAVA ISLAND USING SPACE-TIME GAUSSIAN PROCESS REGRESSION (GPR)

• Main Author: Wisnugraha Sugiyarto, Aditya
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/66752
• Institution: Institut Teknologi Bandung

Indonesia has more than 400 volcanoes and 130 of them are active volcanoes and some of them are located on the seabed. Indonesia is also the meeting point of 2 series of active volcanoes (the ring of fire) and also Indonesia has dozens of active faults. There are large faults in Indonesia such as the Great Sumatran Fault which stretches for 1900 km and the convergent flat boundary of the Banda Sea which creates more seismic activity. This causes Indonesia to be a country that often experiences earthquakes. Therefore, there is a need for a special study regarding the handling of natural disasters, especially in the case of earthquakes, which we cannot deny if we look at the geological conditions of Indonesia. Moreover, Indonesia has a population of around 270 million people which makes it the fourth most populous country in the world. Of the entire territory of Indonesia which consists of about 17 thousand islands, there are several islands that have the most dense population. The most populous island is Java, where more than half (65%) of Indonesia's population lives, so full attention needs to be given to handling to minimize the risk of an earthquake occurring. By minimizing the risk of an earthquake occurring, we can save other factors that would be affected, such as economic factors that are key in a country. There are many ways that can be done to minimize the risk of earthquakes, such as giving special treatment to areas that still have a high level of earthquake risk. The treatment can be in the form of distribution of different allocations of funds depending on the level of disaster risk in the area. Therefore, this study focuses on analyzing the level of disaster risk in Java. The results show that modeling with the space-time Gaussian Process Regression (GPR) method has good results for predicting the level of disaster risk based on spatial and time-series components, namely MAPE for train data and test data are 9.51% and 8.13% respectively with using the RBF kernel. This result is better than other kernels, and with this model it can be used to predict the mapping that will occur in the next few months by entering several values for the predictor variables (year, month, latitude, longitude, depth, and province). So with this prediction of the level of risk, the government can use it to determine how much funds are allocated in each province.