基于自组织神经网络的生态敏感性分区--以北京市房山区为例

Ecological sensitivity division based on SOM-a case study of Fangshan district in Beijing

采用自组织神经网络(SOM)模型,以北京市房山区为例,对其进行生态敏感性分区.分别以土壤侵蚀、地表水环境、地下水环境和生境为生态敏感性因子,作为SOM模型的4个二维输入矩阵,通过多次迭代学习和自组织聚类,使结果在4维(4个生态因子)生态敏感性空间内最大限度地逼近房山区生态特征分布.结果表明,房山区西北部山区的敏感性最高;东部平原是地表、地下水非常丰富的地区,属中度敏感;二者之间的丘陵浅山区敏感性相对较弱.

The ecological sensitivity division based on SOM（Self-Organizing Feature Maps） model was introduced and validated through the case study of Fangshan district in Beijing. Soil erosion, surface water, ground water and habitat, in shape of 2 dimension matrix, were identified as the SOM model input. Through iteration algorithm and self-organization clustering, the model output approached in maximum the pattern of Fangshan ecological sensitivity in 4-dimensional space. In general, its northwestern forested area was the most sensitive area, and its eastern plain, with plentiful surface water and ground water, belonged to medium ecological sensitive region. However, the region between these two parts was in relatively low sensitivity.