講演詳細

経営部門[Forest Management]

日付	2023年3月27日
開始時刻	14:45
会場名	Room 3
講演番号	D21
発表題目	16-year of forest dynamics in Hokkaido: an accurate and on-time forest change detection with Google Earth Engine. 16-year of forest dynamics in Hokkaido: an accurate and on-time forest change detection with Google Earth Engine.
要旨本文	Complex land cover dynamics come from extreme seasonal variations of mixed evergreen and deciduous forests with agricultural land use. Mixed landscape hinders accurately detecting forest change events. Therefore, we aimed to detect monthly forest change more accurately to provide better estimates for carbon-neutral community activities. Google Earth Engine cloud-processing with Continuous Change Detection and Classification time-series algorithm efficiently separated drastic land-cover changes from Japanese forests' seasonality. We stacked 16-year spectral information from Landsat images to select input features by random forest algorithm. Then, we trained randomized classification using JAXA-ALOS forest map as reference and validated forest area differentiated from agricultural field using annual digital height models from airborne laser scanning data. Our approach provides more reliable estimates to monitor Hokkaido's forest with an overall accuracy of 92% for annual mapping.
著者氏名	○Camila Marques1,2 ・ Akira Kato1 ・ Mari Kohri2 ・ Tomoko Furuta2 ・ Hiroshi Sasakawa2 ・ Masahiko Kanamori2
著者所属	1Graduate School of Horticulture, Chiba University ・ 2Forest Information Group, Japan Forest Technology Association
キーワード	forest change, time-series analysis, Google Earth Engine, Continuous Change Detection and Classification
Key word	forest change, time-series analysis, Google Earth Engine, Continuous Change Detection and Classification