Abstract:
Abstract: Land leveling accounts for the major part of project quantity and investment of land consolidation. Rational planning and implementation of land leveling project have important practical significance to reducing the project investment. Partitioning of leveling area is beneficial to reducing the project quantity and facilitating project organization. However, the existing empirical partitioning methods are mainly based on farmland irrigation and drainage facilities, and put little concerns on the project quantity and land consolidation efficiency. A partitioning method of land leveling area based on spatial clustering was proposed and tested in this paper. Firstly, the farmland was divided into a regular grid and the grid cells were used as clustering units. The grid cell size followed the present DEM (digital elevation model) resolution, and was 5m × 5m in this paper. Secondly, clustering variables were selected to ensure that the natural and agricultural infrastructure conditions in one partition are as consistent as possible. These clustering variables included elevation, relative position to the road, land ownership, and spatial coordinates. Thirdly, the spatial clustering was achieved by two-step cluster algorithm to adapt to different clustering variables. Result schemes were evaluated through three quantitative indexes: project quantity, elevation range and shape index. The comparison of empirical and spatial clustering partition schemes both with 9 partitions shows that: the scheme using clustering method save 24% project quantity and 11% elevation range; the shape index of the scheme using empirical method was 11% lower than the scheme using clustering method. The empirical method balances the influence of various factors and its partitions get more regular shape, but it lacks the consideration of project quantity. In the clustering method, the project quantity and elevation range can be significantly reduced due to the consideration of elevation. Although the factors of road and land ownership are also considered in the clustering method, they only have limited impacts on the shape of partition boundaries. Various partition schemes using spatial clustering method were obtained by adjusting the number of clusters. The output of each partition scheme included partition number, partition distribution and evaluation indexes. The comparison of spatial clustering partition schemes with different numbers of partitions shows that with the increasing of number of partitions, the project quantity gradually reduced, the elevation range increased; the shape index decreased, the partition patches became smaller and regularized, and the complexity of partition border were reduced. The comprehensive evaluation index was calculated by weighted average of these three indicators, and the three corresponding weights were 0.47, 0.37 and 0.16 that were calculated by Analytic Hierarchy Process in this paper. The comprehensive index declines rapidly at first then slowly increase with number of partitions. When number of partitions is 7, the minimum value of comprehensive index appears, which means the partition scheme has the best performance overall. To sum up, using spatial clustering method to partitioning the leveling area can effectively reduce the project quantity and improve the flatness after land consolidation, which are beneficial to reducing investment and can satisfy the future needs of agricultural production. However, the clustering method performs less rational than empirical method in aspect of the regularity and distribution patterns. Therefore, in practice, the clustering scheme could be used as the preliminary foundation, and then the scheme can be adjusted depending on experts experience to give full play to advantages of both methods.