報告書番号

V05027

タイトル（和文）

生態系アセスメントのスコーピング支援ツールの開発－地理情報を用いた火力・原子力発電所立地における動物相の推定－

タイトル（英文）

Development of a support tool for scoping phase of ecological impact assessment -Prediction of terrestrial fauna on the location of thermal and atomic power plants from geographical information-

概要 （図表や脚注は「報告書全文」に掲載しております）

発電所の環境アセスメントでは，スコーピング段階で事前に既存情報を整理し，対象地域の生態系の概略を把握することが必要である．これまで，既存資料に乏しい動物については，不十分な情報から生息種を推察せざるを得なかった．本研究では，過去の火力・原子力発電所のアセスメントにおける動物調査データと，GISデータを解析し，地理情報から動物相を推定する統計モデルを作成した．動物相のタイプはTWINSPANにより５つのタイプに区分され，北海道と沖縄で各々特徴的なタイプ，本州，四国，九州では3タイプに分けられた．さらに，本州，九州，四国で見られる3タイプは，ロジスティック回帰分析の結果，土地利用から計算される生息環境要因によって統計的に判別できることが明らかになった．本研究で得られた統計モデルは，スコーピング段階での生態系の概略把握に必要な動物の情報を迅速かつ簡便に提供するツールである．

概要 （英文）

To assess the ecological impact before construction of electric power plants, species lists of ecosystem components are necessary. Topography and vegetation information are easily obtained, because thematic maps of whole area of Japan already exist. However, occurrence data of animal species are hardly obtained except for field survey, especially on the location of power plants. In this study, we tried to construct the predictive models of ground vertebrate fauna by analyzing past environmental assessment data and geographical database. Based on the collected data of mammal, reptile, and amphibian, TWINSPAN classification was conducted. Five types of ground vertebrate fauna were divided as A type distributed on Hokkaido, B, C, and D types distributed on Honshu, Shikoku, Kyusyu, and E type distributed on Ryukyu islands. Species richness of faunal types except for Hokkaido and Ryukyu types were significantly decreased in response to environmental changes from vegetated hilly area to artificial lowland area in order of B, C and D. Environmental differences between B and C types were statistically significant to all environmental variables, whereas differences between C and D types were statistically insignificant to some variables except for land use. The results of statistical tests show that land use strongly affected faunal components, thus we constructed predictive models of faunal types by logistic regression. Two models were obtained to discriminate B from C・ D types, and C from D types. Selected variables of the former model were forest rate, paddy field rate, waste land rate, and distance form the area occupied above 50% of forest. Selected variables of the latter model were distance form the area occupied above 50% of forests and that of paddy fields. Discriminate accuracies of the former and the latter models were 84.5% and 84.4 % respectively. By using this model, species list of a developing area are easily estimated, and scoping processes of ecosystem impact assessment will be efficiently implemented.

報告者

キーワード