Sustainable agriculture can rely mainly on four dimensions at present, including the resource adaptability, economic effectiveness, ecological sustainability, and social development viability. However, it is a high demand to revert the non-grain cultivated (NGC) lands in the mountainous regions of southern China. Existing agricultural production can be disrupted to trigger the severe secondary land abandonment, particularly after the plantation destruction and reconversion. It is urgently needed to implement the contextually adaptive strategies under complex circumstances. Temporal assessments can also be incorporated for the spatially differentiated restoration. Such measures should be systematically organized, according to the diagnostic evaluations of the local socio-ecological conditions and production system resilience. This study aims to systematically analyze the criteria of the cultivated land identification in the Big Food View from the food system perspectives. The land types were identified after NGC reversion. Taking Jianyang District as the study area, the interpretation of the high-resolution remote sensing was employed to map the NGC patterns and distribution. A multi-dimensional diagnostic index system was developed to assess the cultivated land use potential. The multi-dimensional coupling coordination was also found in the NGC utilization. Finally, a self-organizing map network was applied to identify the functional constraints and then prioritize restoration sequences. The results showed that: (1) The types of primary crops included the forest land (60.08%), tea plantations (19.45%), citrus orchards (9.86%), pond aquaculture (6.27%), grape cultivation (4.15%), and dragon fruit (minimal) in the non-grain cultivated plots. These plots were predominantly distributed in the low-elevation hills (100–400 m) with the sunny slopes (6–20°), indicating the strong topographic adaptability. (2) A preliminary level was found in the multi-dimensional coupling coordination degree (MCD) of the cultivated land use, with an average MCD of 0.652. The largest proportion was observed in the primary coordination zones (46.99%), which was mainly distributed in Huangkeng, Mashan, and Xushi towns. Marginal coordination zones (35.29%) were concentrated in the east and northwest, while the high-quality and good coordination zones (14.96%) were primarily located in river valley lowlands. Intermediate coordination zones (2.66%) were clustered in southern Jianyang. While the near-dysfunctional zones (0.10%) were largely influenced by the terrain and ecological constraints. (3) The single cropland was covered 1133.68 ha in the functional deficiencies, where the social development was the most prominent limitation; The areas with two and three weaknesses were measured as 3 097.52 and 1 915.27 ha, respectively, with the resource adaptability and social development as the dominant combinations. According to the potential and deficiencies, the restoration zones were classified as the priority (30.52%), sub-priority (42.87%), medium-term (26.50%), and long-term (0.1%). In conclusion, the differentiated and precise restoration should be adopted in response to the functional limitations. Particularly, there are the dual constraints of resource adaptability and social development. Economic and ecological objectives can be balanced to effectively promote the sustainable utilization of cropland. These findings can offer a scientific basis for the targeted re-cultivation and sustainable management of non-grain cropland in southern mountainous. The theoretical and practical significance can also provide to advance the multidimensional and coordinated agriculture.

This research aims to elucidate the spatiotemporal evolution trends of cropland conversion to forest or fruit production in the southern mountainous regions, with a specific focus on Fujian Province, China. A comprehensive approach was adopted over the study spanning five-time points from 2000 to 2020. A raster dataset comprised the land-use data, natural environmental variables, locational conditions, and economic policy factors. Subsequently, the cropland conversion to forest or fruit production was identified through spatial overlay analysis. The optimal parameter was utilized to geographically weight the detector model. The terrain gradient classification and spatial autocorrelation analysis horizontally and vertically determined the spatiotemporal features of cropland "conversion to forest or fruit production". Furthermore, the indicators were selected from three dimensions – natural environment, locational conditions, and economic policies – to detect the key factors and interactive influences on the significant cropland conversion to forest or fruit production. The research findings indicated the following key results: 1) Cropland conversion to forest or fruit production exhibited an overall trend of "steady development followed by a sudden increase." The fourth time period witnessed a significant surge, with the converted area reaching 98 670.24 hectares. The primary regions of conversion were gradually shifted from east to west, with an increasing number of counties with conversion rates exceeding 1%. There was a reduction in the large-area patches, as the number of patches increased. 2) Horizontally, the degree of conversion shared the transition from "higher in the east, lower in the west" to "higher in the west, lower in the east." Moran's I increased from 0.396 to 0.672, indicating a clear spatial clustering of the conversion behavior. 3) Vertically, the cropland conversion to forest or fruit production exhibited a distinct topographical gradient, primarily occurring at elevations between 0 and 1 000 m, slopes of 0°-2° and 10°-20°, and terrain position indices ranging from 0.1 to 0.8, particularly in regions with the shaded and semi-shaded slopes. 4) Multiple factors were dominated in the cropland conversion to forest or fruit production. At the single-factor impact level, the economic policy demonstrated a higher explanatory power, compared with the natural and locational factors. The impact of each factor varied in the different periods, but the elevation, distance to residential areas, and GDP consistently emerged as the dominant factors with substantial explanatory power. At the multi-factor interaction level, there was mutual reinforcement among factors, with the more pronounced interactions between different types of factors within the same category. Notably, the interaction between locational factors and other elements exhibited a particularly significant enhancement. The cropland conversion to forest or fruit production complicatedly evolved from the combined influences of natural endowment, locational conditions, and economic policies. The differentiated management should be adopted in the control measures for the cropland conversion, according to the baseline conditions of cropland and the diverse driving mechanisms involved. Additionally, the cropland occupation and replenishment can be balanced to promote the spatial displacement between plain areas with forest or fruit production and hilly cropland. The research findings can offer valuable scientific insights for the coordinated development of cropland production and diversified agriculture in the southern mountainous regions.