Cultivated land fragmentation is one of the key factors to hinder the high-quality agriculture. It is of great significance to construct a parcel-scale measurement system of the cultivated land fragmentation. Differentiated governance strategies are often required to promote the cultivated land fragmentation. This study aims to measure and classify the farmland parcel fragmentation in the hilly and mountainous areas. An example was then taken in the Hekou Town, Yunan County, Guangdong Province, China. A theoretical review was presented on the concept and connotation of the cultivated land fragmentation. Natural, tenure and cultivation of the parcels were then incorporated to combine with the questionnaire surveys and the decision-making on the high-standard farmland. A parcel-scale fragmentation measurement index system was constructed using the “natural endowment degree, tenure dispersion degree, and cultivation accessibility degree”. The three-dimensional composite index of parcel fragmentation was then calculated using TOPSIS model. The local spatial autocorrelation was utilized to explore the spatial aggregation relationship. Each administrative village was evaluated for the balance and differences in the variation of the mean and coefficients. Finally, the three-dimensional Rubik's cube model was used to form the guidance program for the classified consolidation. The results showed that: 1) In the dimension of natural endowment degree, the comprehensive evaluation level of the parcels in the study area was ranged from 0.219 to 0.716, where the majority was in the the lower and higher levels. In terms of the spatial differentiation, there was the relatively balance in the variation of the mean level and coefficient after comprehensive evaluation on each administrative village. Parcels with the superior natural endowment degree were formed the contiguous clusters in space. While the rest parcels exhibited the smaller clusters or lower levels of the aggregation, due to the low natural endowment degree. In the dimension of tenure dispersion degree, the comprehensive evaluation level of the parcels was ranged from 0 to 0.902, where the majority was in the lower and low grades. Parcels with the different comprehensive evaluation grades exhibited the relatively homogeneous spatial distribution. In the dimension of cultivation accessibility degree, the comprehensive evaluation level of the parcels was ranged from 0.026 to 0.974, with the majority in the lower and higher grades. Higher and high-grade parcels were tended to cluster spatially, due to the denser, higher-grade road networks and greater parcel adjacency. Other parcels, however, were largely constrained by natural conditions, leading to the high levels of the parcel adjacency. 2) In terms of the spatial correlation, the natural endowment degree was dominated by the high-high and low-low aggregation. The tenure dispersion degree was dominated by the low-low aggregation. And the cultivation accessibility degree was dominated by the high-high and low-low aggregation. 3) Three-dimensional Rubik's cube model was combined with the spatial combination and control constraints. The parcels were classified into the high-quality conservation, endowment enhancement, tenure adjustment, facility improvement, comprehensive consolidation, and prohibited consolidation. Differentiated consolidation strategies were proposed after optimization. The finding can also provide a strong reference to construct the parcel consolidation of the high-standard farmland in the hilly and mountainous areas.

The optimal spatial layout of farmland is of great significance in improving the production capacity. The agricultural and geographical patterns can also be better matched for the national food security. However, the previous researches focus mainly on the spatial layout at the scale of specific administrative districts. It is still lacking in the replacement of farmland with the restored land. Taking Fengkai County in Guangdong Province as an example, this study aims to follow the micro adjustment on "transferring out existing farmland and transferring in restored land". Firstly, the spatial superposition and comprehensive evaluation methods were used to identify the farmland that transferred out with the fragmented, unstable and low production potential. Secondly, the screening utility function was constructed based on the three objectives of the maximization of agricultural production suitability, the minimization of the importance of ecosystem service functions, and the maximization of layout compactness of the restored land. An optimal model was also constructed to adjust the spatial layout of farmland, according to the two aspects of transferring in and out. Taking the balance of the number of transfers in and out as a constraint, the spatial layout of farmland was optimized at different administrative scales. The results showed that: 1) A total of 1,689.89 hm² of farmland was transferred out of the study area, accounting for 7.62% of the total area of farmland. The overall distribution of farmland was characterized by "global fragmentation and local aggregation". 2) The average suitability of agricultural production and integrated ecosystem services in the restored land were 67.04 and 40.52, respectively, indicating the superior resource endowment in agricultural and ecological production. 3) There were significant differences in farmland that transferred in at administrative scales. The average suitability of agricultural production increased by 2.08 in the farmland that transferred in, whereas, the mean value of comprehensive ecosystem services decreased by 0.67 in the intra-county balance scenario, compared with the intra-town one. The screening of farmland that transferred in was more effective at a wider spatial scale. 4) The quality, stability, and contiguity of optimal farmlands were improved in both scenarios. The optimization was much more outstanding in the intra-county balance scenario, compared with the intra-town one. There were also significant impacts on the administrative scales. An optimal expansion in the screening scale of restored land can be expected to improve the overall quality of farmland and reduce the probability of screening ecologically important land. There were two contributions of this study. Firstly, an optimal model was constructed on the spatial layout with better applicability. The dynamic adjustment was realized to optimize the current farmland and the restored land. Secondly, there were some differences in the spatial layout of farmland at the scale of administrative regions. The administrative regions can be considered to formulate the policies for farmland protection. The finding can provide a strong reference for the decision-making on the optimization of spatial layout of restored land and farmland.