The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLVI-4/W6-2021
https://doi.org/10.5194/isprs-archives-XLVI-4-W6-2021-329-2021
https://doi.org/10.5194/isprs-archives-XLVI-4-W6-2021-329-2021
18 Nov 2021
 | 18 Nov 2021

SUPPLEMENTING SATELLITE IMAGERY WITH SOCIAL MEDIA DATA FOR REMOTE RECONNAISSANCE: A CASE STUDY OF THE 2020 TAAL VOLCANO ERUPTION

A. L. F. Yute, E. J. G. Merin, C. J. S. Sarmiento, and E. E. Elazagui

Keywords: social sensing, remote sensing, post-disaster reconnaissance, Named Entity Recognition, Support Vector Machine, Diwata-2, volcanic ash

Abstract. Social sensing and satellite imagery are named as the top emerging data sources for disaster management. There is a wealth of data, both in quantity and quality that can be extracted from social media platforms such as Twitter, given that the content published by users is generally in real-time and includes a geotag or toponym. To reduce costs, risks, and time, performing reconnaissance using remote sources of information is highly suggested. This study explores how social media data can be used to supplement satellite imagery in post-disaster remote reconnaissance using the January 2020 Taal Volcano Eruption in the Philippines. Tweets about the volcanic eruption were scraped, and ashfall-affected locations mentioned in tweet content were extracted using Named Entity Recognition (NER). To visualize the progression of the tweeted locations, dot density maps and hotspot maps were generated. Additionally, a potential ashfall extent map was generated from processed DIWATA-2 satellite imagery using Support Vector Machine (SVM) classification. An intersection of both dot density map and ashfall extent map was performed for comparative analysis of both data. Validation was carried out by matching the ashfall-affected locations with ground reports from local government offices and news reports. The use of social media data complements satellite image classification in the detection of disaster damage for a quick and cost-efficient remote reconnaissance. This information can be utilized by rescue teams for faster emergency response and relief operations during and after a disaster.