TY - JOUR
T1 - Classification of habitats by “Classification Tree”
T2 - A case study on riparian habitats of birds. Ecol. Civil Eng. 5(2), 189-201, 2003
AU - Katoh, Kazuhiro
AU - Ichinose, Tomohiro
AU - Takahashi, Toshimori
N1 - Copyright:
Copyright 2017 Elsevier B.V., All rights reserved.
PY - 2003
Y1 - 2003
N2 - Multivariate analysis of ecological data has been applied to environmental analysis and evaluation. Recently, some ecologists reported that classification and regression trees (CART) are ideally suited for the analysis of complex ecological data. We use classification tree to analyze the relationship between avian species composition and habitat conditions from 37 study plots located in a riparian area of the Tama-gawa River, Tokyo. The data were comprised of census data of birds and vegetation structural information. First, the study plots were classified by TWINSPAN based on the avian species composition. Then, we tried to recover the grouping of the study plots by classification tree or canonical discriminant analysis using the vegetation structural information so that we could find the relationship between avian species composition and vegetation structure. Classification tree analysis performed almost as well as canonical discriminant analysis. Classification tree models explain variation of a single response variable (here, avian fauna type) by repeatedly splitting the study plots into more homogeneous groups, using combinations of explanatory variables (here, vegetation structural parameters). This structure is simple, suitable for dealing with high-order interactions, so that classification tree can give easily interpretable results. Finally, based on the comparison of classification tree and canonical discriminant analysis, we concluded that classification tree was more suitable than discriminant analysis for landscape evaluation and planning.
AB - Multivariate analysis of ecological data has been applied to environmental analysis and evaluation. Recently, some ecologists reported that classification and regression trees (CART) are ideally suited for the analysis of complex ecological data. We use classification tree to analyze the relationship between avian species composition and habitat conditions from 37 study plots located in a riparian area of the Tama-gawa River, Tokyo. The data were comprised of census data of birds and vegetation structural information. First, the study plots were classified by TWINSPAN based on the avian species composition. Then, we tried to recover the grouping of the study plots by classification tree or canonical discriminant analysis using the vegetation structural information so that we could find the relationship between avian species composition and vegetation structure. Classification tree analysis performed almost as well as canonical discriminant analysis. Classification tree models explain variation of a single response variable (here, avian fauna type) by repeatedly splitting the study plots into more homogeneous groups, using combinations of explanatory variables (here, vegetation structural parameters). This structure is simple, suitable for dealing with high-order interactions, so that classification tree can give easily interpretable results. Finally, based on the comparison of classification tree and canonical discriminant analysis, we concluded that classification tree was more suitable than discriminant analysis for landscape evaluation and planning.
KW - classification trees
KW - discriminant analysis
KW - landscape planning
KW - species composition
KW - vegetation structure
UR - http://www.scopus.com/inward/record.url?scp=84999467476&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84999467476&partnerID=8YFLogxK
U2 - 10.3825/ece.5.189
DO - 10.3825/ece.5.189
M3 - Article
AN - SCOPUS:84999467476
VL - 5
SP - 189
EP - 201
JO - Ecology and Civil Engineering
JF - Ecology and Civil Engineering
SN - 1344-3755
IS - 2
ER -