Evaluation and consolidation of cultivated land fragmentation based on integration of function zoning and multi-cluster algorithms

Abstract: Taking the cultivated land in the Xinbei district of Changzhou City in Jiangsu Province of China as the research object, an evaluation model was constructed for the patch-scale arable land fragmentation using the integrated functional zoning and regional clustering. Taking the cultivated land patch as the basic unit, some indicators were firstly selected, including the patch area, contiguous degree, and shape index. A spatial analysis was conducted to calculate the indicators, such as aggregation in ArcGIS 10.6. Secondly, the area weighting was used to expand the patch index to the area with the administrative village as the unit, where the Getis-Ord Gi* was further used to identify the cold and hot areas of each index for the degree of regional fragmentation. Thirdly, the research area was divided into functional zones using two-step clustering, where the administrative villages with neighboring geographical locations and similar fragmentation attributes were clustered into one zone. Finally, the Python-based Sklearn library was selected to implement the K-means clustering with the silhouette measure. The silhouette measure was introduced to determine the optimal number of clusters and the best clustering. The clustering data was then used to evaluate the degree of fragmentation of the cultivated land. The results showed that: 1) There were the insignificant area, continuous regular areas, and discrete complex zone, according to the clustering data of functional zoning. 2) The cultivated land patches were classified into three categories: Category 1, the number of patches was 17 332, the average area of patches was too small, the degree of continuity was low, the area of patch was concentrated in 0-10 000 m2, the area accounted for 21.98%, and the contiguous degree was concentrated in 1-4, mainly distributed in the central area; Category 2, the number of patches was 4 535, the complex and irregular shape of patches, the area accounts for 9.65%, the shape index was concentrated in 1.5-2.5, evenly distributed in the whole area; Category 3, the number of patches was 4 091, the patches were concentrated and regular, where the area accounted for 68.37%, the contiguous degree were concentrated in 5-10, and the shape index was concentrated in 1-1.5, mainly distributed in the peripheral area. Some consolidation suggestions were proposed, according to the fine fragmentation attributes of different categories of cultivated land. A fragmentation model of patch-scale arable land was constructed using the integrated multi-clusters and functional zoning. The index was applied for the landscape pattern indicator to the vector patch scale. There was no need to assign the weight to various indicators during clustering. The regional fragmentation degree was utilized for the specific fragmentation pattern within the region. The model can quickly, intelligently, and low-costly evaluate the fragmented cultivated land patches, which is conducive to further planning and consolidation. The model can be expected to fast, intelligently, and low-costly evaluate the finely divided farming patches for further planning and improvement. This finding can provide a strong reference to improve the evaluation of cultivated land fragmentation.