Detection of River Change in Modeling Flood Vulnerability using Support Vector Machine (SVM) Methods in Tallo River Makassar City

Abstract

The transformation of river morphology and the rising frequency of flooding in urban environments have emerged as increasingly concerning environmental challenges, particularly in Makassar City. The Tallo River, one of the primary waterways traversing the city, exhibits notable dynamic changes driven by both natural processes. In the contemporary era, flooding stands as one of the most recurrent natural disasters, occurring unpredictably and posing serious risks, especially in major metropolitan areas. Such events frequently disrupt daily activities, leading to traffic congestion and obstructing ground transportation. Residential zones situated near riverbanks are particularly vulnerable to its impacts. Moreover, climate change exacerbates these conditions by contributing to increasing environmental unpredictability and need through a monitoring. The purpose of this research is to analyze river morphology changes and assess flood susceptibility in the Tallo River, Makassar City, using Support Vector Machine (SVM) classification methods. Approximately, there are 20% of the area experienced significant changes during 2018 in Tallo River. As water discharge continues to increase, the volume of water mass also rises accordingly. To detect the spatial distribution of flood vulnerability along the Tallo River, which flows through Makassar City, this study utilizes Land Use and Land Cover (LULC) data from 2017 and 2024. These datasets were classified using the Random Forest model, achieving accuracies of 0.89 and 0.87, respectively values that meet the standards for land use change accuracy. Flood vulnerability is also influenced by low elevation values, particularly areas below 0 meters, which are classified as wetland zones. In the Tallo River area, which is part of the Jeneberang Watershed, the dominant class is moderate flood vulnerability, covering approximately 138.48 hectares. Remote sensing technology combined with machine learning approaches especially supervised classification techniques widely used for both binary and multivariate classification tasks, demonstrating high accuracy in detecting and classifying flood vulnerability.