The development of methods for dividing a region into sub-regions that are spatially contiguous and as homogeneous as possible

本研究首次提出了一种基于等值线的区域化方法，并在计算机环境下实现了该方法和其他区域化方法。然后，我们将包括基于等值线的方法在内的几种区域化方法应用于大量的数据集，并从子区域的同质性角度对这些区域化方法进行了比较。我们发现，在空间单元数很少的情况下，基于等值线的方法的平均性能优于其他方法。我们还发现，当数据集中存在较强的空间自相关性以及属性变量之间的相关性很弱时，基于等值线的方法的性能可能会逊于其他方法。虽然基于等值线的方法在某些条件下可能表现不佳，但我们证明，如果空间单元的数量足够大，基于等值线的方法并不会表现不佳。总体而言，基于等值线的区域化方法的性能似乎令人满意。然而，执行该方法的算法还有待进一步开发。我们希望在未来解决这个技术问题。

This study first proposed an isopleth-based regionalization method and implemented the method and other regionalization methods in a computer environment. We then applied several regionalization methods including the isopleth-based method to a numerous number of datasets and make comparisons between the regionalization methods in terms of the homogeneity of sub-regions. We revealed that the isopleth-based method on average outperforms other methods when the number of spatial units is very small. We also revealed that the isopleth-based method may underperform other methods when there is strong spatial autocorrelation in a dataset and also when the correlation between attribute variables is very weak. Although the isopleth-based method may underperform under certain conditions, we however demonstrated that the isopleth-based method does not underperform if the number of spatial units is large enough. Overall, it appears that the performance of the isopleth-based regionalization method is satisfactory. However, the algorithm for executing the method should be developed further. We hope to tackle the technical issue in the future.