文献类型：期刊文献

英文题名：Reconstructing historical forest spatial patterns based on CA-AdaBoost-ANN model in northern Guangzhou, China

作者：Zhan, Xin[1] Yu, Shixiao[1,5] Li, Yide[2] Zhou, Zhang[2] Cao, Honglin[3] Tang, Guangda[4]

第一作者：Zhan, Xin

通信作者：Yu, SX[1]

机构：[1]Sun Yat sen Univ, Sch Life Sci, Dept Ecol, State Key Lab Biocontrol, Guangzhou 510275, Peoples R China;[2]Chinese Acad Forestry, Res Inst Trop Forestry, Guangzhou 510520, Peoples R China;[3]Chinese Acad Sci, South China Bot Garden, Guangzhou 510650, Peoples R China;[4]South China Agr Univ, Coll Forestry & Landscape Architecture, Guangzhou 510642, Peoples R China;[5]Sun Yat sen Univ, Sch Life Sci, State Key Lab Biocontrol, Guangzhou 510008, Peoples R China

年份：2024

卷号：242

外文期刊名：LANDSCAPE AND URBAN PLANNING

收录：;Scopus(收录号：2-s2.0-85177773016);WOS:【SSCI(收录号:WOS:001165609200001)，SCI-EXPANDED(收录号:WOS:001165609200001)】；

基金：We are grateful to many colleagues who assist in the field survey. This research was funded by the National Natural Science Foundation of China (Grant 31770513, 31872701) , and Key Program of Bureau of Forestry and Landscaping of Guangzhou Municipality (SLYKX2016-06) .

语种：英文

外文关键词：CA-AdaBoost-ANN model; Forest landscape; Historical forest spatial patterns; Remote sensing; Species distribution

摘要：Influenced by natural and man-made factors-especially urbanization-regional forest landscapes and structures are in a dynamic process of constant change. It is of great significance to reconstruct the historical pattern of forest landscapes and construct maps of forest landscapes for long time series. Based on the investigation of Fengshui and carbon sequestration forests in northern Guangzhou city, this study combined with an artificial neural network (ANN) improved using the AdaBoost algorithm to create a cellular automaton (CA) to reconstruct species compositions and spatial distributions of historical forests. The model had the best effect at a 30 m spatial scale. At 30 m spatial scale, compared with the actual forest community's spatial distribution, the overall accuracy of forest community distribution reconstructed by our retrospective model in 2000 was 89.17 %, the Lee Sallee index was 0.7972, and the landscape similarity index was 84.62 %. In 1990, the overall accuracy was 86.78 %, the Lee Sallee index was 0.7604, and the landscape similarity index was 80.84 %. This study provides an effective method for the reconstruction of forest vegetation patterns and the prediction of future scenarios.