Post-Disaster Recovery Assessment Through Land Cover And Land Use Change Analysis Using Remote Sensing.pdf

Post-disaster recovery is a complex phenomenon that includes physical reconstruction, economic and social recovery, and is the least understood phase of the disaster cycle. Disasters introduce changes to the affected land and subsequent post-disaster activities through land cover and land use change (LCLUC). LCLU information extracted from satellite imagery is widely used in the RS disciplines, but its value in the recovery context has not been explored.

A study was conducted in Tacloban city, Philippines, which was devastated by super typhoon Haiyan in 2013. The available data consisted of 3 WV2 images from 8 months before, right after, and 4 years after the typhoon. A methodology was developed based on a generic, action-oriented, forward-looking conceptual framework, which included transition patterns to characterize different recovery statuses.

Support vector machine (SVM) was employed for classification purposes, and a detailed comparison of the performance of linear- and RBF-based SVM relying on various settings of hand-crafted features was conducted. The best combination of SVM with image features (SVM+GLCM+NDVI2, SVM+LBP+NDVI2; LC and LU tasks respectively) was applied in 3 time-span images to produce LCLU maps. The result showed that well-designed hand-crafted features could show competitive performance in a complex task involving classes from simple and small to abstract and big regarding complexity and size, respectively.

The LCLU maps were stacked and, based on the developed CF, different TPs from the stacked LC and LU maps were extracted. The final products were LC- and LU-based recovery maps, which were further up-scaled to a region level. It was found that the characteristic of the post-Haiyan recovery in Tacloban city can be explained through the LCLUC information. Results of this study showed that 168 ha of the area had positively recovered by the time of the most recent image, while 69 ha showed negative recovery in both LC- and LU-based recovery maps. Positive recovery was mainly related to the recovery projects and was in part effective to build back the damaged area and build impervious surfaces back better.