Lead Presenter Biography

Krishnan Viswanathan leads the data analytics practice at WRA. He has experience using a vast array of data sources to understand passenger and freight travel behavior.

Co-Authors

Presentation Description

The destructive potential of hurricanes requires that officials make plans for evacuating vulnerable populations when large storms approach well in advance. In order to develop effective evacuation plans, planners need a thorough understanding how the population in different areas respond to hurricane threats and evacuation orders. The traditional approach to collecting this information involved conducting surveys, which are time-consuming, costly, and increasingly suffer from low response rates. Further, the data collected relies on accurate recollection of response behavior by respondents – sometimes many years after the hurricane event. In this regard, passively collected location-based services (LBS) data derived from mobile devices has the potential to address several of these shortcomings by (1) offering larger sample sizes many orders of magnitude at a fraction of the cost; (2) capturing multiple days of travel around the hurricane event for a more complete picture of the before, during, and after periods; and (3) providing spatial and temporal granularity allowing for more robust evacuation patterns – including day of evacuation, evacuation location, and post-evacuation behavior.

In this study, for Florida Division of Emergency Management, LBS data were used to analyze the evacuation travel behavior around three major hurricanes in Florida between 2016 and 2018. Several scalable algorithms were developed to translate raw device pings to activity stays and trips, and then further sequenced and processed to generate a series of behavioral metrics. These include evacuation rates (percentage of people who leave their home locations to evacuate), evacuation county (where in Florida or out of state did people evacuate to), type of refuge (public shelters vs hotels/motels vs others), response curves (cumulative curves representing evacuation times for different segments of the population), and clearance times (time to clear the impacted regions). Another aspect of the analysis was to compare evacuation behavior of residents in site-built home locations vs mobile homes and residents of different risk zones, since the choice of home type and location has equity considerations (lower income individuals tend to stay in mobile homes, lack transportation options, and very vulnerable to hurricane impacts).

The results from this data-driven analysis were validated against a hurricane evacuation survey done by Florida DOT. These metrics were subsequently used to update inputs to the Transportation Interface for Modeling (TIME), Florida’s hurricane evacuation model. The purpose of TIME is to help decision-makers better prepare their communities to undertake evacuations in a safe, orderly, and timely manner. 

Presentation File