Travel demand behavioral model after disaster earthquake in Tehran Iran

3.1. Earthquake Scenario

According to JICA, Ray fault is the strongest and north of Tehran fault comes after that. In the event of activation of these faults, earthquakes of magnitude 7 and 5.5 on the Richter scale will respectively occur in the south of Tehran and they have been considered as the earthquake scenarios in this study. Since the time of an earthquake (day or night, working or non-working day) has also a substantial effect on the post-earthquake behavior of the individuals and where they would go after the earthquake occurrence, the hypothetical earthquake scenario should also consider time of earthquake occurrence in order to model the travel demand [5]. Therefore, 4 earthquake scenarios including an earthquake that occurs on the day (day earthquake), an earthquake that occurs at the night (night earthquake), a moderate earthquake and a powerful earthquake have been considered in this study. It is assumed that day earthquake occurs in the morning of a working day that all employees are at work and the schools are open. Considering this situation, the most critical state will happen to the traffic in the transport network. Another earthquake is assumed to occur at midnight while people are asleep in their homes (Such as the Bam earthquake, 2003). In addition, the damages caused by the moderate and the powerful earthquakes are estimated [1]. 1) Specifications of the earthquake of magnitude 5.5 (moderate): minor damage (up 10%) to the typical buildings made with bricks and metal or the buildings over the age of 30 (type B). Damages (up to 40%) to weak and badly designed brick buildings (type C) including cracks and falling plaster walls and ceiling. There would not be a severe damage to the lifelines and the phone would be usable [1]. 2) Specifications of an earthquake of magnitude 7 (powerful): well-designed structures (type A) would be bent and their location on the Foundation would be displaced. Type (c) structures would destroy and type (B) buildings would be heavily damaged (up to 85%) or completely destroyed. The streets with the widths less than 6 meters would be closed. Some roadway bridges would be damaged. Telephone, water, gas and electricity would be cut off. There would be the probability of fire and explosion at gas stations after the earthquake [1].

3.2. Data Acquisition

The survey was conducted in such a way that the individuals were interviewed face to face due to the requirement to explain the situation to them. In this regard, 364 individuals were asked to fill in the questionnaire designed with the stated preference (SP) methods. In the first part of the questionnaire, the individuals were asked questions about the type of their building (structure of the building and its age) in order to specify the extent of damage against the moderate and powerful earthquakes. In the second part, explaining earthquake scenarios and damages caused by them, the individuals were asked about their first post-earthquake trip purposes. The trips made after the day earthquake are defined with 4 purposes as follows: A. Relief and rescue: trips on relief/rescue as well as trips to confirm the safety of the family members, like looking forwards children at schools or other family members out of home. B. Return-to-home: trips from work to home. C. Immediate evacuation: trips in order to leave the place where the individual at the time of the earthquake is and to go to a safe place (i.e. the nearest park or outdoor space to work/home appointed by the disaster management organization).

D. No travel: individuals who do not make a trip. The post-earthquake trips for the night earthquake are defined with 3 purposes as follows: A. Inquiry on safety: checking up offices. Since there is a possibility of looting shops and offices after the earthquake, many individuals prefer to make a trip to their shops or work and collect their valuable property. B. Immediate evacuation: trips in order to leave the place where the individual at the time of the earthquake is and to go to a safe place (i.e. the nearest park or outdoor space to work/home or somewhere out of town). C. No travel: staying at home. In the third part of the questionnaire, the individuals were asked about their personal characteristics such as age, gender, marital status, having children, education and past experiences of the disaster, and their economic characteristics such as size of their family, the number of their private vehicles, the vehicle used to go to work under normal conditions and house/apartment ownership. The data were collected according to a survey of people working in zone 10 of Tehran in which 52% of streets have a width less than 6 meters and wouldn't be accessible after the earthquake occurrence [1]. According to the JICA studies, this zone may sustain the most severe damages in terms of human casualties and destruction of buildings. Therefore, it is considered as the study area in this research. Since the way in which individuals make decisions when they are at work was concerned, the interviews were conducted while they were at work. The individuals considered in this study were employees, teachers and shopkeepers (all of whom owned shops).

3.3. Descriptive Statistics

Table 1 gives an overview of all Descriptive statistics. 77% of people in the statistical sample own a car, but only half of them go to work with their cars. About 20% of people in the statistical sample have had past experience of a powerful earthquake and none of them selected the choice "no travel" in the questionnaire. Table 1. The statistical characteristics of the collected data Variable

Gender (male=1, female=0) Age Marital status (single=0, married=1) Education (High school diploma=0, Diploma=1, BS/BA/BE =2, MS/MA/ME=3, Ph.D./Doctor=4) Having a child who goes to school (Yes=1, No=0) Family size Job (employee)=0, self-employment=1) Car ownership (Yes=1, No=0) Car ownership at work (Yes=1, No=0) House/apartment ownership (Yes=1, No=0) Past experience of a powerful earthquake (Yes=1, No=0) stability of the building (Good (groups A and B) = 1, weak (group C) = 0) the time interval from work to home (min) the time interval from work to school (min)

3.4. Logit Model

In this study, the multinomial logit model was used to analyze the acquired data. This model is capable of designating the share of each trip purpose and also factors affecting the selection of any of the trip purposes. The general form of the multinomial logit model is given in equation (1) [15]. P𝑛 (𝑖) =

Uin: The utility of alternative i for the person n, Ujn: The utility of alternative j for the person n, Pn(i): The probability of choosing i for the person n from C possible alternatives. In order to build the discrete choice logit models, the maximum likelihood method shown in equation (2) is used. 𝐿𝐿(𝛽) = ∑𝑛 ∑𝑖𝜖𝐶 (𝑦𝑖𝑛 𝑙𝑛(𝑃𝑛 (𝑖)))

LL (β): Log-likelihood function after building the model; yin=1, if the individual n chooses i; yin=0, if the individual n doesn't choose i [15]. The goodness of model fit measure (ρ2 statistic) is used to validate the model that is calculated as follows: 𝜌2 = 1 −

LL(0): Log-likelihood function assuming all parameters set to zero. LL(β): Log-likelihood function after building the model; LL(c): The value of the log-likelihood functions for every constant term.

5.1. Day Earthquake

Statistical analysis of the trip purposes, based on the magnitude of the earthquake, showed that after occurring the powerful earthquake, 23% of the individuals will make trips on relief/rescue, 68% of them will return to home and the others will make trips to the safe places. After the moderate earthquake, 10% of the individuals will make trips on relief/rescue, 24% of them will return to home, 12% will make trips to the safe places and the others will stay where they are. Table 3 gives an overview of the results of the multinomial logit model for the postearthquake trip selection. The dependent variable in the model is the trip purpose after an earthquake on a working day. In order to perform the analysis and build the model, in addition to the previously defined variables, different composition variables were used, some of which improved the model. The ρ2 statistic for the model was obtained 0.425 that is a fairly good indicator for the discrete choice models. The model has also predicted the trip purposes in 67% of the observations correctly. The parents*gender combined variable is obtained negative, indicating that mothers are more willing to provide relief which confirms the results obtained by [13]. The results also showed that parents are more likely to provide relief and it corresponds with the results obtained by [11]. If the earthquake occurred is powerful, people will be more willing to choose a relief trip due to a sense of danger. Comparing the coefficient of the variable for the earthquake magnitude in relief/rescue and return-to-home utility functions, it is found that after powerful earthquake, the individuals are more willing to provide relief and then return to home. In such situation, they wouldn't tend to have an immediate evacuation or stay where they are, immediately after the earthquake. Since the individuals who have a stable house/apartment wouldn't feel a need to check up their homes, they are more likely to provide relief. However, the individuals who own a house/apartment prefer to

return to home and check up their property and valuable things. The negative coefficient of travel time variable indicates that the longer the travel time, the weaker the individuals' tendency to provide relief, which confirms the result obtained by [13]. The coefficient of the variable is small due to the travel time in minutes in this model and the time interval being in the range of 3 to 120 minutes (large numbers). The results indicate that as the family size is greater, the individuals are more willing to return to home and not to go to a safe place first, after the earthquake. Men who are over the age 50 are more willing to go to a safe place than the women of the same age, after the earthquake. This demonstrates the impact of gender on trip purposes and points to the fact that the gender variable affects the post-earthquake trip purposes. Since individuals make decisions with respect to the sense of danger as well as their information on the earthquake, the education variable has no effect on decision making. According to the conducted interviews, only a small number of the individuals (6 individual) had received a high education (MSc and Ph.D.) in the field of earthquake and they were the only ones who had developed a plan for their family to come together in a predetermined location after the earthquake. The location was easily accessible and close to all family members. Since the number of these people was small, it has not affected the model. The variable for the previous experience of a powerful earthquake has not been significant in the model. This could be due to the small number of respondents who have experienced a powerful earthquake. However, none of them have chosen to make no trip. Since the individuals believe that using cars in the critical situation is impossible due to the damages to the streets, having a private car at work had also no impact on the behavior of the individuals. Table 3. Results of the multinomial logit model of the trip selection for the day earthquake Trip purposes

Return-to-home Immediate evacuation and going to a safe place No travel LL(0) LL(c) LL(β) ρ2 ρ2c Accuracy of the estimation (%)

Variable coefficient Gender * having children -3.1041 Marital status* having children 5.9301 stability of the house 1.4921 Travel time -0.0264 Magnitude of the earthquake 4.2188 Constant coefficient 3.5857 Magnitude of the earthquake 3.4206 Family size 0.3524 Constant coefficient 6.0758 Family size -0.4643 House/apartment ownership -0.7725 Age over 50* Gender 2.4314 Constant coefficient 5.3853 -463.02086 -415.0472 -266.3161 0.425 0.358

P-value 0.0000 0.0000 0.0114 0.0256 0.0000 0.0002 0.0000 0.0219 0.0000 0.0657 0.0669 0.0002 0.0000

5.2. Earthquake Occurring at the Night

Statistical analysis of the trip purposes, based on the magnitude of the earthquake, indicated that after the powerful earthquake, 10% of the individuals will make trips to check up offices and 88% will make trips to the safe places. After the moderate earthquake, 46% of the individuals will have an immediate evacuation and 52% prefer to stay at their homes. According to the acquired 607

data, more than half of the individuals whose first trip after the earthquake are to have an immediate evacuation, prefer to seek refuge in a safe place out of the city. Therefore, a large number of the individuals tend to leave the city after the earthquake which may lead to a heavy traffic on the roads leading to the city outlets. Table 3 indicates the results of the multinomial logit model. The dependent variable in the model is the trip purpose after the earthquake at the night. The fitness index (ρ2 statistic) of the model was obtained 0.572 that is a fairly good indicator for the discrete choice models. The model has also predicted the trip purposes in 73% of the observations correctly. Table 4. Results of the multinomial logit model of the trip purpose for the night earthquake Trip purposes Inquiry on safety

No travel LL(0) LL(c) LL(β) ρ2 ρ2c Accuracy of the estimation (%)

Variable coefficient Gender * Job 3.5338 Gender * Age over 50 0.0501 Gender * travel time -0.0478 Earthquake magnitude 6.1756 Constant coefficient 4.8685 Gender * Job -1.0014 Earthquake magnitude 4.5517 Constant coefficient 3.0351 Family size 0.3870 Stability of the building 0.7329 -336.9357 -255.3770 -157.2732 0.572 0.389

P-value 0.0001 -1.6912 0.0851 0.0000 0.0002 0.0365 0.0000 0.0076 0.0184 0.0508

The model results show that men who own a shop are more likely to check up their shops and properties after the earthquake while the individuals who have a government job (employees), as well as men over the age 50, prefer to go to a safe place. The negative coefficient of travel time*gender variable for inquiry on safety trips represents that if the work is far from the home, the individuals show less tendency to chuck up offices and prefer to make a trip on relief/rescue as well as trips to confirm the safety of relatives and acquaintances or have an immediate evacuation. The variable for the gender being significant for the inquiry on safety choice (in the questionnaire) is due to the importance of job and property for men owing to their greater sense of responsibility toward financing the family. At the event of a powerful earthquake, the individuals would be more inclined to check up the shops and there would be also a great desire to have an immediate evacuation. The individuals who are confident of stability of their buildings prefer to stay at their homes due to less sense of danger. However, since the coefficient obtained for the magnitude of the earthquake is high, it is clear that staying at home will be only chosen by the individuals after a moderate earthquake. The modeling results indicate that as the size of the family is greater, the individuals are more likely to stay at home with their families. The variable for the previous experience of a powerful earthquake was also evaluated in the model that slightly improved the model, but the level of significance of this variable was not acceptable. Therefore, it is not included in the model. This could be due to the fact that a few individuals interviewed had previous experience of a powerful earthquake. Providing a more comprehensive statistical sample, this variable may be significant and influential in the individuals' traffic behavior.