*Yuei-An Liou1、Kim-Anh Nguyen1、Trong Hoang Vo1
(1.National Central University)
キーワード:Urban greenspace, vulnerability, typhoon, Google Earth Engine, Sentinel-2 MSI
The need of accurate and up-to-date spatial data for decision-makers in public and private administrations is steadily increased. People concern about the management of urban greenspace for smart city and disasters, especially with the of enhanced adaption under the context of climate change with possible consequence of more frequent weather extremes. Here, we present urban greenspace (UGS) extraction in Taiwan derived from the Sentinel-2 MSI imagery. The UGS is classified with support of Google Earth Engine (GEE) using classification and regression trees for Machine Learning (CART) and then converted into GIS environment in ArcGIS 10.5 for further analysis. A total of 360 field survey points were used for training and validation. Land cover is classified into four classes, including 1) dense tree cover, 2) tree cover, 3) water body, and 4) miscellaneous. Later, the UGS was further extracted and classified into roadside tree, park, and shrub or grassland. Overall accuracy of classification is 88.6 % with mean Kappa coefficient of 0.84. The extracted UGS was combined with other indicators to generate the vulnerability map to typhoons. The UGS vulnerability map was classified and ranked into five levels (very low, low, medium, high, and very high) to distinguish the risk patterns and levels of urban greenspace in Taiwan to typhoons. Results further indicate the applicability of Sentinel-2 MSI data as an effective dataset to study urban greenspace at city scale. This study provides an approach to monitor and manage UGS by using freely accessible satellite imagery dataset and open-source platform GEE. Results are useful for decision makers to mitigate the impacts and damages associated with typhoons and enhance the adaptive capacity.