地球信息科学学报 ›› 2012, Vol. 14 ›› Issue (6): 800-806.doi: 10.3724/SP.J.1047.2012.00800

• 空间分析综合应用 • 上一篇    下一篇

基于SOFM网络的景观功能分类——以北京及周边地区为例

冯喆1, 吴健生2, 高阳1, 彭建1, 宗敏丽2, 王政2   

  1. 1. 北京大学城市与环境学院 地表过程与模拟教育部重点实验室, 北京 100871;
    2. 北京大学深圳研究生院 城市规划与设计学院 城市人居环境科学与技术重点实验室, 深圳 518055
  • 收稿日期:2012-11-03 修回日期:2012-12-05 出版日期:2012-12-25 发布日期:2012-12-25
  • 通讯作者: 吴健生(1965-),男,湖南人,副教授,研究方向为景观生态学与GIS。E-mail:wujs@szpku.edu.cn E-mail:wujs@szpku.edu.cn
  • 作者简介:冯喆(1984-),男,北京人,博士研究生,研究方向为景观生态学。E-mail:sucreal@126.com
  • 基金资助:

    国家自然科学基金重点项目(41130534)资助。

Classification of Landscape Functions Using SOFM Neural Network: A Case Study from Beijing and Its Peripheral Area

FENG Zhe1, WU Jiansheng2, GAO Yang1, PENG Jian1, ZONG Minli2, WANG Zheng2   

  1. 1. College of Urban and Environmental Sciences, Peking University; Key Laboratory for Earth Surface Processes of the Ministry of Education, Beijing 100871, China;
    2. The Key Laboratory for Environmental and Urban Sciences, Shenzhen Graduate School, School of Urban Planning & Design, Peking University, Shenzhen 518055, China
  • Received:2012-11-03 Revised:2012-12-05 Online:2012-12-25 Published:2012-12-25

摘要:

景观多功能性是景观生态学研究的热点领域,需要一种既能体现景观多功能整体性,又能表征各功能间独立性的表达方法。本文以北京及其周边地区为研究区,以500m栅格为最小评价单元,使用空间化的统计数据表征物质生产功能,使用植被生物量与土壤含碳量之和表征碳汇功能,使用潜在水土流失量与实际水土流失量的差值表征土壤保持功能,使用生态系统服务功能的评估结果表征生境维持功能,使用人口空间化数据表征居住功能。在计算5种景观功能强度后,通过自组织特征映射模型将土地栅格进行聚类分析。研究结果表明:景观功能强度具有空间异质性。景观功能强度可分为以农地为优势景观,以物质生产为主要功能的农业功能区域;以农地和城市用地为优势景观,以居住和碳汇为主要功能的城市功能区域;以林草地为优势景观,以土壤保持和生境维持为主要功能的生态功能区域;以及优势景观不明显,各项功能均衡发展的过渡功能区域4类。该分类方法既可较好地表达多功能景观的功能分异和空间分异,又能为其研究土地利用和生态管理实践提供理论依据。

关键词: 北京及周边地区, 聚类分析, 景观多功能性, SOFM网络

Abstract:

Landscape multifunction is a hotspot in the fields of landscape ecology. In order to explore a method which can reflect both integrity and independence of landscape multifunction, this research focuses on the clustering of landscape functions, taking Beijing and its peripheral area, China, as the study area. Five landscape function intensities, material production, carbon storage, soil retention, habitat conservation, and population support, are calculated using a variety of ecological models and indices in a grid map. Then, based on the results of landscape multi-function calculation, the study area are clustered through self-organizing feature map model. The quantitative results show that different regions turned out to have different and relative unique effects on the regional priority functions. Beijing and its peripheral area can be divided into four landscape function regions: agricultural region, whose dominant function is material production; urban region, whose dominant functions are population support and carbon storage; ecological region, whose dominant functions are soil retention and habitat conservation; and transition region, which does not have dominant functions, but reflects the interaction between human and nature. The validation of the results also shows that the presented SOFM neural network model is an effective and appropriate method for cluster analysis. Clustering results based on the SOFM model exhibit significant regional heterogeneity, with notable regional differences in the four clustering types within the research area. This spatial comprehensive dataset, combined with the independence from mechanistic ecological assumptions of the SOFM network approach provides a unique opportunity to validate and assess modeling efforts. The dominant landscape functions influencing regional development differ from one area to another. Furthermore, characteristics of the landscape indices and functions vary with region. Despite its limitations and uncertainty, the application of the presented method on clustering landscapes function using the SOFM model organization in connection with high performance computers is encouraged as a very interesting and important goal for future studies. The approaches to achieve sustainable regional development were illustrated and their importance highlighted for policy makers and stakeholders.

Key words: Beijing and its peripheral area, cluster analysis, SOFM neural network, landscape multifunction