基于MaxEnt模型与聚类算法的良渚古城遗址景观分布研究  

Spatial distribution of landscape heritage in Liangzhu ancient city based on MaxEnt model and clustering algorithm

作　　者：余孙哲 寿松涛 张亚平 杨国福 聂文彬 徐斌[1] YU Sunzhe;SHOU Songtao;ZHANG Yaping;YANG Guofu;NIE Wenbin;XU Bin(College of Landscape and Architecture,Zhejiang A&F University,Hangzhou 311300;Fuyang Department of Agriculture and Rural Affairs,Hangzhou 311400;College of Art and Archaeology,Hangzhou City University,Hangzhou 310015)

机构地区：[1]浙江农林大学风景园林与建筑学院,杭州311300 [2]杭州市富阳区乡村振兴服务中心,杭州311400 [3]浙大城市学院艺术与考古学院,杭州310015

出　　处：《安徽农业大学学报》2024年第1期67-73,共7页Journal of Anhui Agricultural University

基　　金：浙江省重点研发计划(2019C02023)资助。

摘　　要：为探索良渚古城遗址景观空间分布特征,预测潜在遗址景观分布概率,研究通过Geodatabase平台对已发现的135个遗址点构建景观空间数据库,利用SAGA-GIS、主成分分析(PCA)采集筛选17个相关变量,通过K-means聚类法识别划分遗址群,最后运用MaxEnt模型对各遗址群建立预测模型,进行刀切法检验变量,并绘制出各遗址景观群潜在的分布概率图。结果表明:1)良渚古城遗址景观可依据14大筛选变量划分为4类,并根据聚类结果与模型变量响应曲线,提取出傍山依水、远山近水、居山取水和低丘止水4种分布模式;2)其中影响度最大的4个主导变量分别是坡度(61.8%)、风暴露指数(34.8%)、地形开阔度(21.7%)和地形湿度指数(41.1%);3)良渚古城遗址景观潜在的分布区域为大遮山南部平原及大雄山西北部湿地,4类遗址景观高概率分布的区域面积分别为11.56、33.76、61.34和76.87 km^(2)。In order to explore the distribution characteristics of the Liangzhu ancient city site landscape,predict the potential distribution probability of the site landscape,and study the construction of a landscape spatial database for 135 discovered site points through the Geodatabase platform,we utilized SAGA-GIS and principal component analysis(PCA)to collect and select 17 relevant variables,and employed the K-means clustering method to identify and categorize site clusters.Finally,we applied the MaxEnt model to establish predictive models for each site cluster.We conducted the knife-cutting method to test variables and produced potential distribution probability maps for each site landscape cluster.The results indicated that:1)The landscape of the Liangzhu ancient city site can be cate-gorized into four types based on 14 selected variables.According to the clustering results and the response curves of model variables,four distribution patterns were extracted:‘near mountains and water’,‘distant mountains and near water’,‘residing on mountains and obtaining water’,and‘low hills and stopping water’.2)The four dominant varia-bles with the greatest impact were slope(61.8%),storm exposure index(34.8%),terrain openness(21.7%)and ter-rain humidity index(41.1%).3)The potential distribution areas of the Liangzhu ancient city site landscape were the southern plain of Dahe Mountain and the northwest wetland of Daxiong Mountain.The high probability distribution areas for the four types of site landscapes were 11.56,33.7661.34 and 76.87 km2,respectively.

关 键 词：遗址景观 MaxEnt模型 空间分布 K-MEANS聚类 良渚古城 

分 类 号：TU986[建筑科学—城市规划与设计]