Marginalization of cultivated land, resulting from rapid urbanization, exists as an important form of land use change, and represents a new research direction in land-use and land-cover change (LUCC). This article proposes a classification of such marginalization on the basis of elasticity of input and income, categorizing marginalization of cultivated land as either policy-induced (PIM), nature-induced (NIM) or economy-induced (EIM) marginalization. These classifications are further explored as either positive or negative marginalization, depending on whether the land is transformed from or into cultivated land. This innovative framework is applied to analyses of marginalization in Lianjiang County, located in southeastern coastal China. This research analyzes characteristics of spatial-temporal evolution of categories of marginalization of cultivated land using 3D kernel density methods. Significant findings point to spatial-temporal processes and driving forces of marginalization, including: (1) Concentrations of positive (P-PIM) and negative (N-PIM) policy-induced marginalization both occur and agglomerate in separate spaces, with the former mainly in the southeastern portion of the county and the latter in the northwest. (2) By contrast, patterns of positive (P-NIM) and negative (N-NIM) nature-induced marginalization complement each other in space - N-NIM tends to be more discrete in areas with P-NIM aggregations, and vice versa. (3) Finally, areas with aggregations of positive (P-EIM) and negative (N-EIM) economy-induced marginalization overlap. The research suggests that relevant land use policies should be formulated in response to these characteristics of cultivated land marginalization so as to address marginalization of cultivated land, especially as associated with rapid urbanization.