Modeling Social Interactions Between Households for Evacuation Behaviors in the Devastated Areas



Modeling Social Interactions Between Households for Evacuation Behaviors in the Devastated Areas

Authors

Junji Urata, Graduatedschool Of Civil Engineering , The University Of Tokyo, Eiji Hato, Graduatedschool Of Civil Engineering , The University Of Tokyo

Description

This paper focuses on residents’ mutual assistance during evacuations and aims to model the social interactions between households for evacuation behaviors. The main objectives of this paper are to propose the choice model of the residents’ pairs of the social interactions behaviors and to derive the choice set from the spatiotemporal and social groups.
The framework of the proposed model derives from the network GEV model and has two nest. The first nest is the households’ groups and the second nest is the residents. The groups are partitioned by group detection algorithms for social networks and the spatial networks The utilities of the households’ pair is defined by the other-regarding preferences. The model evaluates the difference of the behavioral constraints under the disasters as the factors of the other-regarding preferences and the dynamic change of the parameters with the disaster situation. The model is calibrated by the real social interactions between households in the devastated areas. By the survey, households’ activities were assembled by the interviews.
The choice set formations are governed by the distance of pairs, the difference of their damages and the difference of their behavioral constraints. The current study assumes that the small neighborhood groups and the other-regarding preference influence mutual behaviors. The processes of the group detection using the road networks enable to simulate the evacuation behaviors according to households’ location at that time.

Abstract

1. Objective
In natural disasters, for example a typhoon, a tsunami and a volcanic disaster, most people need someone’s help to get information and to evacuate. This paper focuses on residents’ mutual assistance during evacuations and aims to model the social interactions between households for evacuation behaviors. The main objectives of this paper are to propose the choice model of the residents’ pairs of the social interactions behaviors and to derive the choice set from the spatiotemporal and social groups.

2. Methodology and Data
The framework of the proposed model derives from the network GEV model and has two nest. The first nest is the households’ groups and the second nest is the residents. The groups are partitioned by group detection algorithms for social networks and the spatial networks and the whereabouts of the residents at that time are used in the group detection. The groups are used for sampling the alternative households’ pairs if the pairs of the different groups are chosen depending on their feature. The utilities of the households’ pair is defined by the other-regarding preferences and the distributions of the random error terms have the correlations between the households’ pairs if they have the same households. The model evaluates the difference of the behavioral constraints under the disasters as the factors of the other-regarding preferences and the dynamic change of the parameters with the disaster situation. The model is calibrated by the real social interactions between households in the devastated areas.
The evacuation activities data was gathered in the city of Niihama after the heavy rain and mudslide disasters in 2004. On the survey, households’ activities were assembled by the interviews. A purpose of the survey is to know the households’ awareness of danger, the risk management activities and the cooperation activities along the time series.

3. Results
The network analysis of the social interactions shows that the road networks influence the pairing of the social interactions. The households’ networks are partitioned by the road network and their habitation. The result of calibration indicates that distances affect the choices of the pair but the heavy damaged households are assisted many people despite their distances. The choice of the pairs from the heavy damaged households’ groups and other group is low influenced of their utility because the gaps of the pair utilities are less by the homogeneity parameter μ. The allocation parameter in residents’ nest of the assisted residents being over 1/2 indicates that people who are assisted get more utilities.

4. Implications
The model is applied to the social interactions pair choices in the natural disasters and provides the insights regarding its underlying determinants. The choice set formations are governed by the distance of pairs, the difference of their damages and the difference of their behavioral constraints. The current study assumes that the small neighborhood groups and the other-regarding preference influence mutual behaviors. The processes of the group detection using the road networks enable to simulate the evacuation behaviors according to households’ location at that time.

Publisher

Association for European Transport