Hydro-disintegration of loess–paleosol sequences and its implications for microscale landform evolution

• Four disintegration modes dictate loess micro-landforms: knolls, pillars, stepped terraces. • Sandy/silty loess > 80 % mass loss vs. clayey loess depth-dependent dormancy below 10 m. • Fissures & clay cementation control collapse-driven vs. particulate erosion mechanisms. The disintegration dynamics of loess–paleosol sequences (LPS) constitute a critical precursor to soil erosion and serve as a determinant in landform evolution across the Loess Plateau of China (LPC). The spatial heterogeneity of LPS distribution, coupled with the alternating stratigraphy of loess and paleosol units within individual gully slopes, engenders differential erosion patterns with distinct magnitudes and mechanisms. This study deciphers the lithostratigraphic controls on disintegration dynamics through comparative analysis of three lithofacies: sandy (SL), silty (SIL), and clayey loess (CL). Vertical profiling revealed stark contrasts: SL and SIL exhibit high disintegration potential (>80 % mass loss), correlating with intensive erosion, while CL displays depth-dependent behavior—moderate activity (>40 %) in upper layers (<10 m) transitioning to dormancy in deeper strata. Microstructural characterization identified four failure modes: Type I (particulate dispersion) dominates homogeneous sandy loess, enabling gradual lateral erosion that shapes dome-shaped knolls; Types II–III (bulk collapse), prevalent in fissured silty or clayey loess, drive vertical pipe development, forming walls and pillars; Type IV (surface spalling) in dense CL preserves stepped terraces. Disintegration mechanisms are governed by textural–structural synergies: SL’s loose, isotropic fabric favors particle-by-particle detachment, whereas SIL/CL’s vertical fissures and clay cementation dictate collapse-driven erosion. These modality-specific processes create a tripartite landscape evolution model: SL domains regress through lateral erosion, SIL and upper CL zones evolve via episodic collapse, and stable CL strata resist degradation. The findings establish microstructure–erosion linkages critical for predicting landscape responses to hydrological forcing, offering mechanistic baselines for erosion modeling and paleo-landform reconstruction.