Generating Geolocated Synthetic Population to Assess Travel Need to Access Opioid Treatment Centers


  • Aiden Pape, Computer Science, Middlebury College

Faculty Mentor(s)

  • Shangjia Dong, Civil and Environmental Engineering, University of Delaware


A synthetic population dataset is used as the foundation for predicting and testing theĀ  societal impact of public policy and climate disasters. In an effort to produce highly granular housing-household data, this paper presents a method to generate a Two-Layer Geolocated Synthetic Population. We use Synthetic Reconstruction techniques including Entropy Minimization and Truncation Replication Sampling to generate the individuals along with the households. The methodology uses Census Data from the US Census Bureau including aggregated data and PUMS micro-data to create the synthetic population. Each household is geolocated to the housing units and the structures identified using the county parcel data and structure data from the (NSI) National Structure Inventory. We use Sussex County, Delaware as a case study area which is highly prone to coastal flooding. A high resolution spatial data with individuals and households is generated with heterogeneous characteristics such as age, sex, earnings, marital status, race, ethnicity, building type, number of persons in the household, tenure status, and household income. Moreover, we map the households probability to visit the Opiod Treatment Center based on the demographic attributes of the individuals. The results provide a framework to identify the vulnerable communities for targeted support during natural disasters such as flooding and show the benefit provided by integrating synthetic populations and housing unit structures.