The overall objective of this RAPID proposal is to use the Protective Action Decision Model (PADM) as a guide to collecting empirical data on people's behavioral responses to the Palu earthquake and tsunami. Specifically, this project has collected a uniquely valuable dataset to respond the following seven major research objectives (RO): RO 1: Collected data on the amount of time it took officials to decide to issue tsunami evacuation warnings; RO 2: Collected data on tsunami warning sources, channels, messages and warning dissemination times. This will include data on people's ability to recognize earthquake shaking as an environmental cue to tsunami onset; RO 3: Collected data on people's evacuation participation rates (vs. sheltering in-place), preparation times, and departure times; RO 4: Collected data on people's evacuation logistics (e.g., route choices, destination/accommodations choices, evacuation durations, and evacuation costs); RO 5: Assessed the extent to which earthquake/tsunami brochures were distributed or workshops were held before the event and if they predict shorter evacuation preparation and departure times; RO 6: Assessed the ability of physical, social, and household contexts; social and environmental cues; socially-transmitted warnings; demographic characteristics; prior experience; and cultural background to predict warning receipt, risk perception/personalization, evacuation decisions, and evacuation departure times; and RO 7:  Assess the ability of demographic data to predict people's evacuation logistics.

This project achieved three interdisciplinary objectives that link social science warning research, transportation engineering, and casualty modeling.  First, it extended warning research, which has focused substantially on hurricanes, to tsunamis--thus providing more information on warning response in rapid onset events.  Second, it seeked to replicate and extend the findings from the one major earthquake/tsunami, the 2009 American Samoa. Third, it omplemented the recent study of the Uttarakhand India flash flood. 

The data on warning dissemination and household preparation times will significantly reduce the uncertainty associated with current estimates of tsunami evacuation departure times that are used in evacuation transportation models. More accurate tsunami ETEs will provide a sound scientific basis for estimating casualty rates under existing terrain and evacuation route system conditions. In turn, this baseline condition can be used to examine the potential effectiveness of hazard education programs that increase recognition of earthquake shaking as an environmental cue to tsunami onset, thus eliminating authorities' decision times and reducing warning dissemination time distributions. 

This study used the Protective Action Decision Model (PADM) as a guide to collect about 488 household residents' responses to the 2018 Palu Indonesia M7.5 earthquake and tsunami. The results show that (1) The disparity between the widespread availability of motor vehicles and the transportation mode actually used to evacuate (foot) requires an explanation. Of course, not all respondents had access to motorized vehicles with which to evacuate. Moreover, the research team's interviews and observations indicated the majority of the motor vehicles are motorcycles. Thus, even people with motorcycles might have evacuated by foot because 1) motorcycles were unable to evacuate all family members, 2) motorcycles might have taken more time to prepare to evacuate than evacuating by foot, 3) damage to the road network by liquefaction and collapsed buildings might have impeded vehicular travel. Whatever the explanation is, this result is consistent with the 2004 Thailand tsunami in which 75% of the respondents evacuated by foot and the evacuation expectations survey in Kamakura City Japan, in which 71% of the respondents expected to evacuate by foot. However, these results are quire different from the 53% of the respondents in the American Samoa tsunami who evacuated by car and the 75% of those in the evacuation expectations survey in the CSZ that expected to evacuate by car. Consequently, further research is needed to identify the determinants of tsunami evacuation model choice.  (2) Although 43% of the respondents had a motorized vehicle (car, truck or motorcycle) available to them, 67% of them evacuated by foot, 29% in their own motor vehicles, and 4% carpooled in a friend's or neighbor's vehicle. For their evacuation destinations, 31% respondents went to other villages but 69\% stayed within their own village. For their evacuation accommodations, most respondents went to a public location such as a park (37%), but a few went to the home of a relative (6%), a mosque (5%), an official shelter (4%), the home of a friend (2%), or another location (47%). (3) 65% of them left within 5 minutes, 76% within 10 minutes, 84% within 15 minutes, and 95% within 30 minutes. Only 4% took more than 40 minutes to leave.

This study also compared the response behaviors from this study with the past earthquake and tsunami events such as the 2009 American Samoa (M8.0), the 2011 Christchurch (M6.3), the 2011 Hitachi (M9.0). The purpose of this comparison is to identify what are the commonalities and differences. 

Last Modified: 07/09/2021
Modified by: Haizhong Wang