Event-based mapping and spatial pattern analysis of landslides in parts of central Vietnam

Systematic inventorying after large triggering events is essential for understanding the spatio-temporal dynamics of landslides and enabling advanced predictive analysis. The landslide inventories triggered by typhoon Ketsana, tropical storm Podul, and typhoonMolave in the highlands of central Vietnam are presented to better understand the spatial interaction (i.e., clustering, dispersion, or independence) among these landslide events using a global clustering indicator Ripley’s K. A total of 8,744, 915, and 10,257 landslides were mapped and attributed to typhoons Ketsana, Podul, and Molave, respectively, utilizing high-resolution RapidEye (5m/px) and PlanetScope (3m/px) satellite imagery. The inhomogeneous Ripley’s K and cross K-functions were employed for intra-inventory and inter-inventory landslide distribution analyses to assess the clustering, dispersion, or randomness of landslide occurrences at various distances. The findings reveal pronounced clustering among landslides triggered by typhoons Ketsana and Molave. In contrast, tropical storm Podul caused landslide clustering up to 22.5 km, after which a strong aversion was observed. Interactions between landslides from the Ketsana and Molave inventories exhibited clustering within the 0–8.1 km range and repulsion beyond 8.6 km. Conversely, landslides from the Ketsana and Podul events displayed repulsion at all distances, indicating the absence of clustering. Subsequently, a Spatial Kolmogorov-Smirnov test was conducted on seven continuous covariates—elevation, topographic slope, topographic aspect, Topographic Position Index, drainage density, annual average rainfall, and distance from mapped geologic faults. A Chi-square test was applied to lithology as a categorical covariate. These analyses aimed to evaluate the spatial independence of landslide distribution concerning these covariates. A statistically significant p-value in these tests indicated a non-random spatial association between landslide occurrences and the covariates, signifying spatial dependency among them.