Machine Learning for Mapping the Distribution of Clay Minerals in Sidoarjo Mud Volcano (LUSI) Using Support Vector Machine (SVM) Algorithm

Abstract

Abstract Sidoarjo Mud Volcano (LUSI) is a natural phenomenon that has led to the emergence of volcanic mud eruptions in the city of Sidoarjo, East Java Province, with a generated volume of volcanic mud of approximately 80.000 m3 per day. However, only about 60.000 m3 can be managed daily, resulting in an excess of volcanic mud that accumulates and overflows into the surrounding area from the center of the mud eruption. This volcanic mud include high-temperature gas and fluid, which pose significant obstacles and limitations for conducting research at the location. Remote sensing methods, which involve measuring the electromagnetic spectrum energy emitted by an object without direct physical contact in the field, have been employed as a solution to overcome the limitations of access. The Advanced Spaceborne Thermal Emission and Reflection (ASTER) satellite’s image data will be utilized, specifically the Short-Wave Infrared (SWIR) subsystem consisting of bands 4 to 9. This subsystem is selected due to its high reflectance sensitivity to clay minerals, allowing for easy recognition and identification of the imagery. SWIR data will be extracted using band ratio math, which involves sensitive band combinations for specific minerals, followed by normalization to minimize atmospheric effects and provide optimal and clay-sensitive outcomes. The resulting SWIR extraction data, amounting to 58.164 data points, will be subjected to the machine learning method of the Support Vector Machine (SVM) algorithm to maximize classification accuracy. The dataset will be divided into 40% training data and 60% testing data. The training data will establish the foundational model for the SVM algorithm, which will subsequently be utilized by the Support Vector Classifier (SVC). The SVC will employ the Radial Basis Function (RBF) kernel with a gamma parameter value of 10 and a C parameter value of 750. Based on these decisions, the entire dataset can be classified using the established parameters, resulting outcome in the form of a distribution map of clay minerals in Mount Lumpur Sidoarjo (LUSI). The presence of clay minerals, including Kaolinite, Chlorite, Illite, and Smectite, will be uniformly and predominantly distributed in the northern direction from the center of the volcanic mud eruption site.