A substantial area of abandoned sloping farmland on the Loess Plateau is undergoing natural vegetation succession, which significantly influences runoff and soil loss. However, the responses of soil erosion processes to vegetation and soil property changes driven by natural abandonment remain unclear due to the absence of long-term continuous observations. To address this, an eight-year continuous study was conducted to monitor rainfall, vegetation, soil properties, runoff, and also soil loss across six slope gradients (5 degrees-30 degrees). The first two years (2016 and 2017) represented tilled bare land, while the subsequent six years (2018-2023) represented abandoned land. The results showed that the runoff depth (RD) and soil loss rate (SLR) exhibited a decreasing trend as the duration of abandonment increased. Both RD and SLR initially increased and then decreased with slope gradient, with a critical slope angle identified at 25 degrees. SLR increased linearly with RD from 2016 to 2023, and the slope of their fitted equation was higher before abandonment but fluctuated downward with the increase in abandonment years. The runoff coefficient and SLR significantly decreased with increasing plant density, litter biomass, moss crust cover, saturated hydraulic conductivity, and total porosity (p < 0.05). Additionally, threshold-dependent responses were observed in the relationship between rainfall and runoff. When the comprehensive rainfall index (CRI)-constructed using rainfall depth, duration, and intensity-exceeded 0.23, a significant correlation emerged between rainfall and runoff. Below this threshold, vegetation and soil properties exerted stronger influences, leading to a disordered rainfall-runoff response. Vegetation restoration also complicated the response of soil loss to runoff. A significant relationship between runoff and soil loss was only observed when runoff exceeded 0.46 mm; below this threshold, the relationship remained irregular. These findings from long-term field observations provide a scientific basis for improving soil erosion prediction models and support the development of sustainable and high-quality soil and water conservation measures on the Loess Plateau.