高分解能衛星データを用いた森林タイプ判別 : 佐倉市の里山地域を例として
Classification of forest types using high resolution satellite data : A case of Satoyama landscape in Sakura City

Nami HASEGAWA, Yoshinobu HOSHINO, Keitarou HARA, Noritoshi KAMAGATA, Katsuhiro NAKAO
2013 Shokusei Gakkaishi  
∫ず α ψy ∫ ゐeterocycla were detected as the dominant spe − cies representing the 8 forest types . The size ofaplot was l 6 m × 16mand corresponded to l6pixels ofthe IKONOS data. We developed classi 盒cation tree models fbr each fbrest type by using 5 explanatQry variables as fbllows : Normalized Difference -「 egetation Index ( NDVI ) , mean ofthe red band, standard deviations ofthe 。 ,、, − i。 fr 、,ed band"ed b、nd ・nd g ・・en b ・ nd , E・ ch m ・ d。 1・h・ w ・ d ・ g ・ ・d p ・・f ・rm ・nce ( ・・ea und ・曲 e
more » ... ce ( ・・ea und ・曲 e cu 隅 [ AUC ユ> 0. 8) , except that by M / aponicbls type , Each model involved a characteristic combination ofthe ex − planatory variables . The standard deviations of the near − in仕ared band and red band wore adoptcd as the ex − planatory variables in some models such as that fbr P1 竃 γ 〃ostachys bambusoides type . Thus , textural property was also shown as usefUl fbr the identification offbrest types when using the high − resolution satelhte data ; 72 % ・fth・ pl ・t・ w ・・e ・identifi ・d ・・ rr・ctly. This re ・ult i・ di・at ・・ th・t high − re ・・ 1・ti・ n ・at・ llit ・ d・ta c ・ ・ ld b・u ・ed f・・
doi:10.15031/vegsci.30.25 fatcat:gbqyeb557rdrjndakagaapvu3q