Composition of simulation data for large-scale disaster estimation

Abstract

When a large-scale natural disaster occurs, it is necessary to quickly collect damage information so that disaster-relief operations and wide-area support in accordance with the scale of the natural disaster can be initiated. Previously, we proposed a fast spatio-temporal similarity search method (called the STSim method) that searches a database storing many scenarios of disaster simulation data represented by time-series grid data for scenarios similar to insufficient observed data sent from sensors. When the STSim method is naively applied for estimating disasters occurring at multiple locations, e.g., fire spreading after a large-scale earthquake, it must prepare a huge number of combinations consisting of scenarios that represent disasters at multiple locations. This paper presents a combination method of simulation data in order to apply the STSim method for estimating disasters occurring at multiple locations. This proposed method stores scenarios, each of which represents a disaster occurring at a single location, to a database; thus, reducing the number of scenarios. After a disaster occurs, it extracts and composes scenarios similar to observed data, resulting in efficient disaster estimation in any situation. We conducted performance evaluations under the assumption that an earthquake occurs below the Tokyo metropolitan region and estimating the spread of fire in the initial response. These results of the processing time for estimating the spread of fire show that the processing time is within 10 minutes, which is practical.