Non-grain cultivated land is distributed in the areas with a concentrated population and frequent mining activities in recent years, particularly with the continuous development of urban expansion and industrialization. The scale effect of non-grain cultivated land can be one of the most important geographic phenomena in spatial heterogeneity. The optimal strength of related factors can also be conducive to the sustainable production of cultivated land in the mineral grain composite areas. In this study, the aggregation and dispersion of non-grain cultivated land were considered in the different geographic scale units from the scale effect and spatial contiguity. The study area was taken as the largest copper mine and grain composite area in Jiangxi Province of China. Nine factors were selected considering the scale of 500-6100m block, watershed, and administrative village. The spatial measurement model was used to analyze the type of non-grain cultivated land in the mineral grain composite area. Firstly, the land survey data was used to quantify the non-grain cultivated land in 2020. Global Moran's I index was used to test the multi-scale spatial autocorrelation of three types of non-grain cultivated land in the mineral grain composite area. The block scale was classified into four scale levels; Then, the traditional least squares method (LSM), spatial lag, and spatial error models were used to carry out the binary simulation regression of non-food cultivated land and the related factors. A comparison was made on the strength of the influencing factors at different scales. The results show that: (1) three forms of non-grain cultivated land were distributed mainly in the central and western regions, as well as the riparian zones with less distribution in the mountainous areas. The overall distribution of non-grain cultivated land showed strong spatial autocorrelation and strong spatial heterogeneity, with significant spatial spillover effects at most scales; (2) The outstanding scale effects were found in the spatial heterogeneity of non-grain cultivated land in the mineral grain composite area. Four zones were divided, according to the actual situation. (3) There was an inconsistent significance of related factors of non-grain cultivated land in different scales. The significant rate was 90.2% in the regression of non-grain cultivated land on the small scale, with an average value of the R2 0.236, which was better than other scales. The scale of the catchment area was better than that of administrative villages; (4) There was also a clear scale effect in the strength of related factors in the non-grain cultivated land, indicating the most significant in the small-scale level. the average regression coefficient of the main factors was greater than 0.25, and the significant rate reached 92%. Therefore, the cultivated non-grain cultivated land in the mineral grain composite area was affected by multi-scale, multi-factors, and multi-dimensions. The influence of related factors at different scales was the optimal choice to solve the cultivated non-grain cultivation. It is necessary to consider the spatial dependence and scale effect in the existence of non-grain arable land in the future comprehensive land space improvement in the mineral grain composite area. The key factors should also be optimized for the land space, according to local conditions.