<p>Evaluating rainfall-induced landslides is crucial for mitigating potential damage to human life and property. Modeling the uneven destruction caused by rainfall and understanding the underlying mechanisms are both challenging and essential. This study proposed an improved multi-field coupling discrete element model for dynamically analyzing the full process of rainfall-induced landslides and their influencing factors, validated through comparison with experimental results. In this model, elements were assigned calibrated water content parameters that directly affect their strength. A dynamic seepage network model was developed based on the inter-elements contact relationships. The results shown that landslides triggered by heavy rainfall on sandy slopes are typically shallow. Stress concentration first occurred at the slope toe, where rainfall-induced degradation initiates collapse, gradually propagating across the slope. A sensitivity analysis was conducted on the reduction coefficients of cohesion and internal friction angle in the saturated state. The results shown that an excessively high reduction coefficient for cohesion leads to shallow landslides, while an excessively high reduction coefficient for internal friction angle results in deep landslides. In most fine-grained soils, cohesion is more significantly controlled by saturation than by the angle of internal friction, which explains why most rainfall-induced landslides are shallow. The presence of weak layers and joints can control the depth of the sliding surface. This study provided new insights into modeling rainfall-induced landslides.</p>