Abstract: Terrace has been one of the most important cropland resources in the mountainous areas. But the terrace and cropland abandonment has been more significant issue in the process of marginalization with the advancement of urbanization in recent years. This study aims to reveal the spatial differentiation and influence mechanism of terrace abandonment for the sustainable use of terraces in mountainous areas. According to the perspective of cropland marginalization, an indicator system of distributional characteristics was constructed to explore the development of the terraces abandonment-prone areas. At the same time, the average nearest-neighbor index and the independent samples t-test were adopted to reveal the aggregation and neighborhood effect of terraces abandonment, respectively. Finally, the impact mechanism of terrace abandonment was obtained using a binary logistic regression from the ‘cost-benefit’ perspective. The results show that: 1) Terraces were the most vulnerable to abandonment if the greater slope, difference in height from settlements, the distance from settlements, elevation and distance from the nearest road, except for the less distance from the terrace to the forested area. 2) The terraces after abandonment shared the aggregation characteristics. The degree of aggregation was lower on the abandoned terraces than that on the ploughed terraces. The reason was that the cultivated terraces were required to centralize, in order to save as much labor cost as possible. The process of abandonment usually started with the most marginal terraces, which were more susceptible to the wildlife encroachment. 3) Abandoned terraces shared the stronger neighborhood effect, compared with the abandoned sloping cropland. More influence was from the surrounding abandoned cropland. The impact of abandoned cropland on terrace abandonment generally occurred within a range of 100m. The abandoned terraces were tended to decrease and level off, when the range exceeded 100m. The strongest impact of abandoned cropland on terrace abandonment occurred in the 10m range, where the abandonment rate of terraces as Swinney high as 60%. 4) At the plot scale, whether terraces were abandoned or not was depended on the endowment of terraced plots. Specifically, the variables, such as topographic conditions, the distance from settlements, and shape of the plot were positively correlated with the terrace abandonment, whereas, the quality of terraces and size of the plot were negatively correlated. The terrace abandonment was attributed to the urbanization and industrialization, where the higher labor costs were caused to abandon less productive terraces. The endowment of a single terrace was dominated the costs and benefits of farming, and then to make a decision on whether to abandon the terrace or not. At the same time, the neighborhood effect of abandoned terraces was formed the negative feedback, further contributing to the spread of terrace abandonment. In the future, the introduction, research and development and upgrading of suitable machinery for farming should be emphasized in the management of terraces. At the same time, it is very necessary to combine social capital for the ‘production + culture and tourism’ terrace complex, in order to improve the management system of terraces. The regional differences should also be considered for the decision-making on the sustainable use of terraces.