|
| 基于多要素空间聚类的南方红黄壤区中低产田分类与障碍类型识别——以海南省澄迈县为例 |
| Classification and obstacle-type identification of medium-low yield farmland in the southern red and yellow soil region based on multi-factor spatial clustering：a case study of Chengmai County，Hainan Province |
| Received:September 12, 2025 |
| DOI：10.13254/j.jare.2025.0920 |
| 中文关键词: 中低产田，空间聚类，耕地分类，障碍类型，澄迈县 |
| 英文关键词: medium-low yield farmland, spatial clustering, arable land classification, obstacle type, Chengmai County |
| 基金项目:国家重点研发计划项目（2023YFD1900101）；四川省重点研发计划项目（24ZDYF1583）；海南省研究生创新科研项目（RC2500003501） |
| Author Name | Affiliation | E-mail | | Zhang Ziyi | College of Tropical Agriculture and Forestry, Hainan University, Haikou 570100, China | | | Hu Yueming | College of Tropical Agriculture and Forestry, Hainan University, Haikou 570100, China
South China Academy of Natural Resources Science and Technology, Guangzhou 510000, China | | | Yang Hao | South China Academy of Natural Resources Science and Technology, Guangzhou 510000, China
College of International Tourism and Public Management, Hainan University, Haikou 570100, China | | | Wang Lu | College of International Tourism and Public Management, Hainan University, Haikou 570100, China | wlu@hainanu.edu.cn | | Liu Yunan | College of Information and Communication Engineering, Hainan University, Haikou 570100, China | | | Zhao Xiaoyang | College of Tropical Agriculture and Forestry, Hainan University, Haikou 570100, China | | | Zhang Dongming | Hainan Provincial Key Laboratory of Arable Land Conservation, Haikou 570100, China | |
|
| Hits: 362 |
| Download times: 55 |
| 中文摘要: |
| 为探索适用于南方红黄壤区的中低产田精细化分类方法，提升其识别与障碍诊断的科学性与可操作性，本研究以海南省澄迈县为研究区，构建了由土壤pH、土壤有机质含量、地形坡度和耕地利用类型组成的多属性耕地分类指标体系。采用K-means聚类算法（K-means）、高斯混合模型（GMM）、自组织映射网络（SOM）和基于密度的空间聚类算法（DBSCAN）4种空间聚类方法对研究区耕地进行分类，并通过聚类质量指数（CQI）、轮廓系数（SC）、戴维斯-布尔丁指数（DB）和卡林斯基-哈拉巴斯指数（CH）等内部效度指标对不同算法的分类效果进行综合评价。结果表明，K-means算法在综合聚类表现相对较优，SC为-0.103，CH为187.451，可为后续中低产田分类与障碍类型识别提供基础。基于K-means聚类结果，将研究区耕地划分为3个一级类、8个二级类和17个三级类。在此基础上，结合耕地质量等级约束及《全国中低产田类型划分与改良技术规范》（NY/T 310—1996）的判别规则，识别出中低产田的两类主导障碍类型，即坡地梯改型和瘠薄培肥型。研究表明，本研究构建的要素构建-空间聚类-等级约束-障碍归因-空间叠加方法能够有效揭示县域耕地内部属性差异及主导障碍，为中低产田精细化分类和障碍识别提供可操作方法，并可作为分类-诊断-改良技术路径的基础。 |
| 英文摘要: |
| To explore a refined classification method for medium-and low-yield farmland in the red and yellow soil regions of southern China and to improve the scientific reliability and practical applicability of farmland identification and obstacle diagnosis, Chengmai County, Hainan Province, was selected as the study area. A multi-attribute farmland classification indicator system was constructed, incorporating soil pH, soil organic matter content, terrain slope, and arable land use type. Four spatial clustering methods：K-means clustering（K-means）, gaussian mixture model（GMM）, self-organizing map（SOM）, and density-based spatial clustering of applications with noise（DBSCAN）were applied to classify arable land in the study area, and the classification performance of different algorithms was comprehensively evaluated using internal validity indices, including the clustering quality index（CQI）, silhouette coefficient（SC）, daviesbouldin index（DB）, and calinski-harabasz index（CH）. The results showed that the K-means algorithm exhibited relatively better overall clustering performance, with an SC of -0.103 and a CH value of 187.451, providing a basis for subsequent classification and identification of obstacle types in medium-low-yield farmland. Based on the K-means clustering results, arable land in the study area was classified into three primary classes, eight secondary classes, and seventeen tertiary classes. Furthermore, by integrating cultivated land quality grade constraints and the classification criteria of the National Technical Specification for Classification and Improvement of Medium-and LowYield Farmland（NY / T 310—1996）, two dominant obstacle types of medium-and low-yield farmland were identified：the nutrientdeficient improvement type and the sloping land terrace improvement type. The findings indicate that the proposed factor constructionspatial clustering-grade constraint-obstacle attribution-spatial overlay workflow effectively reveals intra-county farmland attribute differences and dominant obstacles, and it provides an operable approach for refined classification and obstacle diagnosis of medium-and low-yield farmland. |
| HTML
View Full Text
View/Add Comment Download reader |
| Close |
|
|
|
