[MZZ54-P02] Emergency flood detection using multiple information: Integrated analysis of natural hazard monitoring and social media data
Keywords:flood, emergency detection, effective rainfall, Twitter
Nowadays the extreme weather events occur more frequently due to climate change. In October of 2019, the eastern japan was hit by a large and high-speed typhoon, Hagibis (no.19). This unprecedented typhoon caused the evacuation of 237 thousand people, the casualties of over 300 people, and the house damage of over 94,000 households throughout the affected area. Flood is one of the most devastating natural disasters thus providing effective early warning is critical to reducing disaster impacts as preparing for such extreme weather events. Current warning against rainfall-induced disasters is issued based on the rainfall threshold, which varies depending on the disaster type. For example, the amount of 1.5 hours-effective rainfall and the river water level have been used as relevant criteria for flood hazards. However, warning based on only natural hazard monitoring is not enough to detect areas subject to disaster. Integrated analysis natural hazard monitoring and social media data could improve the warning system to enhance awareness for disaster managers and citizens about the emergency event. We analyzed time-series data including rainfall, effective rainfall, river water level monitoring data, and Twitter data relating to disaster events during 5 days from 11 to 15 October, focusing on the most affected area around the watershed of Abukuma-river and Chikuma river in Japan. The analysis using 47,520 tweet samples revealed that spatial and temporal resolution and extent of these monitoring data influence to detect the emerging change and spatial deviation relate to the flood events. Thus, the design of monitoring system should meet the criteria for emergency detection.