Object-based Classification of Land Impacts under Typhoon Pablo in Compostela Valley, Mindanao

摘要

Immediate land impact analysis during and right after the occurrence of any catastrophic disaster provides crucial function for producing timely information to stimulate proper disaster reduction measures. In this area of interest, remote sensing has thus been widely applied and rapidly developed in terms of its high spatial resolution and multiple spectrogram. However given the concern of cost of high spatial resolution image, and complex socio-ecological system in the Southeast Asia, estimating direct land impacts of typhoon related hazards from interpreting delicate landscapes at fine scale is still challenging. Interpreting hurts of agricultural land that also refers to crops and trees uprooted or broken by windstorm and/or debris/mud flow is of criticality. This study presents an Object-Based Image Analytical (OBIA) method using SPOT5 satellite imageries (10m resolution) to analyze land impacts in Compostela Valley, Mindanao in the strike of typhoon Pablo in December 2012. Various object-based features are integrated including spectra, texture and geometry of 4 spectral bands to assess especially the hurts of cultivated fields and the windthrow effects. The accuracy assessment shows that the accurate classification of debris/mud flow reaches 88.9% and of windthrow attains 83.3%. The result reinforces that distinguishing windthrow from debris/mud flow or flood is possible. This analytical method highlights the importance to enable a fast regional-scale assessment of typhoon damages and especially cost-effective option for inaccessible areas.