Spatial Pattern Analysis and Spatial Regression of Geological Disaster Risk in Indonesia

Abstrak

Indonesia is a country with a high level of vulnerability to geological disasters such as earthquakes, landslides, and volcanic eruptions due to its position in the meeting zone of three active tectonic plates. This study aims to map the spatial pattern of geological disasters Local Indicators of Spatial Association (LISA) and analyze the effect of spatial variables on the number of victims using a spatial regression analysis approach. Data were obtained from BNPB and analyzed using the OLS method and three spatial regression models, namely the Spatial Autoregressive Model (SAR), Spatial Error Model (SEM), and General Spatial Model (GSM). The LISA results show that several provinces are included in the high-high and spatial outlier categories that are prone to disasters. Regression analysis shows that for earthquakes, the SEM model is the best model with a pseudo-R² of 0.888 and the lowest AIC. For landslides, the GSM model gives the best results (pseudo-R² = 0.7860 and AIC = 78.389), while for volcanic eruptions, the spatial model does not show significant spatial effects, so the OLS model is sufficient to represent the relationship between variables. The variable number of events proved to have the most significant influence on the number of victims across all types of disasters. This finding emphasizes the importance of a spatial approach in risk mapping and geological disaster mitigation planning in Indonesia.