On January 23, 2024, a magnitude 7.1 earthquake struck Wushi County in the southern Tianshan Mountains, Xinjiang Province, China. This earthquake caused numerous landslides, both on terraces and non-terraces. However, current research lacks a comprehensive examination of the effects of river terraces, as a distinct geomorphic feature, on coseismic landslides. Moreover, variations in sampling strategies introduce uncertainty into the modelling results of coseismic landslide susceptibility assessments. Therefore, this study provides a comprehensive analysis of the spatial distribution, characteristics, controlling factors, and susceptibility assessment of terrace and non-terrace landslides across a 2894 km² affected area. We developed the first coseismic landslide inventory via extensive remote sensing analysis and field surveys. The results revealed that the Wushi event triggered at least 4009 coseismic landslides, covering a total area of 15.68 km², approximately one-third of which occurred on river terraces. Statistically, terrace and non-terrace landslides are prone to occur in specific zones of topographic, geologic, hydrologic, and seismic factors, which have relatively high landslide area proportions, landslide area densities (LADs), and landslide number densities (LNDs). These factors have the most significant control over terrace landslides, with higher LAD and LND values than non-terrace landslides do, which is called the terrace effect. Terrace landslides preferentially occurred in areas with intermediate elevations, hillslopes with steep gradients, low-elevation depressions, and high peak ground acceleration (PGA) values. Under the two different sampling strategies, the landslide susceptibility assessments performed using the Bayes network (BN), decision table naive Bayes (DTNB) and radial basis function classifier (RBFC) models demonstrate that source-based machine learning (ML) models achieve superior predictive performance compared with mass-based models. Notably, river terrace areas exhibit both higher proportions of high/very high susceptibility zones and greater landslide concentrations within these zones than the entire study region does. The tectonic geomorphic conditions of the Wushi earthquake-induced landslides are macroscopically controlled by regional uplift and river erosion, whereas earthquakes provide direct triggers for landslide occurrence. The clustering and size of terrace landslides are controlled by the terrace effect, especially the effect of morphological terraces. This study increases the understanding of the spatial distribution of earthquake-induced terrace landslides in the Tianshan Mountains, offering a scientific foundation for disaster prevention and mitigation in comparable regions.

Terrace landslides,Tianshan Mountains,Ms. 7.1 Wushi earthquake,Effect of morphological terraces