Spatial Regression Modeling and Distribution of Submarine Landslides in the Negros–Sulu Trench System

Large submarine landslides are a global concern as they can trigger tsunamis with no clear precursors. While geological characterization of submarine landslides remains a challenge to many areas worldwide, the availability of global bathymetric datasets and spatial analysis tools has led to progress in mapping these submarine geomorphological features. Morphological and statistical analyses of submarine landslides and their attributes enable the identification of regions susceptible to large submarine failures and covariates that are good predictors of submarine landslide volume. This study identifies significant clusters of large submarine landslides mapped (n=1214) in the Negros–Sulu Trench System by testing the spatial dependence of volume using global Moran’s I and Getis-Ord (Gi*) statistic. This study further explores a spatial model that best elucidates the distribution of submarine landslide volume. Global Moran’s I suggests significant positive spatial autocorrelation, while Gi* statistic reveals local clustering of large-volume submarine landslides, where the densest clustering occurs offshore of southern Panay Island. Among the 18 spatial regression models, the (1) univariate spatial Durbin, (2) nested, and (3) spatial Durbin error with the maximum slope as the predictor have the lowest Akaike information criterion (AIC) of 2056.1, 2057.0, and 2057.8, respectively. The spatial regression models also revealed that mean depth is a poor predictor of submarine landslide volume. Log likelihood-ratio test suggests a simpler option of the nested model. The spatial Durbin error model better represents the underlying local heterogeneities such as sediment flux and subduction processes in triggering submarine landslides than the global spillover effects of the spatial Durbin model. Furthermore, this study highlights the dominant role of slope and tectonic processes that induce oversteepening, triggering large submarine landslides that may induce damaging tsunamis. The identified offshore areas with significant clustering of large submarine landslides are valuable information for offshore geophysical surveys and tsunami hazard assessment in the region.