Distribution and development characteristics of loess seismic landslides induced by the 1622 M 7 earthquake in Northern Guyuan, Northwest China

Coseismic loess landslides induced by historical strong earthquakes represent the predominant seismogeological hazard across the Loess Plateau, and a thorough inventory of their distribution and kinematic behavior is fundamental to regional seismic hazard evaluation and to the mechanistic understanding of loess seismic disasters. Although well-studied events such as the 1654 Tianshui M 8 and 1920 Haiyuan M 8½ earthquakes have received systematic investigation, the coseismic loess landslides triggered by the 1622M 7 earthquake have long been confined to brief entries in historical catalogs, with no clear understanding of their spatial distribution, controlling factors, or kinematic characteristics. To address this gap, we combined high-resolution remote sensing interpretation with field verification to identify 322 coseismic loess landslides and compile the first comprehensive inventory for this event. We analyzed the relationships between landslide distribution and seismic intensity, fault distance, seismic wave propagation direction, and loess ridge orientation, and we summarized the topographic conditions and geometric parameters governing landslide development. The results show that: (1) Landslides mainly occur in seismic intensity zones VIII and IX and form a dense NNW-trending belt along the southern segment of the Yantongshan Fault, concentrated on the fault hanging wall with sliding directions nearly orthogonal to both fault strike and loess ridge orientation; their distribution is jointly controlled by ground-motion intensity, fault geometry, topography, and seismic wave propagation direction. (2) Source slopes exhibit a typical loess–mudstone dual structure with a three-segment vertical profile; landslides preferentially develop on loess slopes 30–50 m high with original angles of 20–30° and free-face angles of 40–50°. (3) Landslides are dominated by small-to-medium sizes and show typical peeling-type failure along the loess–mudstone interface, with landslide angles mostly 10–20° and weathered scarp angles of 24–30°. (4) The dominant runout distance is 30–60 m, and the equivalent friction coefficient (μ) follows a normal distribution peaking at 0.25–0.30, indicating moderate-to-low friction and medium-to-short runout behavior; μ displays significant negative power-law correlations with volume (V) and runout distance (LR), revealing that larger landslides have lower friction and longer travel distances. (5) The inventory confirms that the southern Yantongshan Fault is a Holocene active fault, and the coseismic rupture length estimated from landslide distribution is 40–50 km, yielding a magnitude consistent with historical records. These findings fill a major gap in the study of historical earthquake-induced loess landslides in southern Ningxia and provide important theoretical and practical value for improving the historical loess seismic landslide database of the Loess Plateau, assessing fault activity, and analyzing regional seismic hazard.