UCGIS-CyberGIS Center at UIUC Award for Advancing Reproducible Geospatial Research

 Two of the six groups were selected by their peers, mentors, and research staff for recognition this year.

 

Group 6 Workflow Summer School 2019The Context Makes the Difference: Reproducibility and Replicability in Measures of Spatial Accessibility

A food desert is defined as “a region lacking access to healthy foods as well as a range of other nutritious options”, but calculating it is challenging because of the following reasons: available maps are very general or cover large aggregations of data and its existing visualization is outdated and static. To overcome the problems, we investigated “What impact do individual user preferences have on accessible food service areas?” We have developed a tool for calculating a food desert based on personal preference. The function takes a transportation mode (walk, bike, drive) and personal context (travel time threshold, and store rating) as input and generates a customized food desert map. The implementation includes data retrieval from Google Map API and Open Street Map, as well as a network analysis with NetworkX, GeoPandas, and Jupyter notebook. The general structure of models can be reproducible and replicable at will. By opening and parameterizing the model for individual input, not only can the same model be replicated in different places, different parameters can be tuned for individual access. Thanks to Jupyter notebooks and CyberGIS infrastructure, we have demonstrated collaborative problem sharing, exploration, and investigation. Using diverse libraries of spatial analysis, our approach exemplifies the cross-discipline nature of shared scientific discovery using CyberGIS.

Group 6 Summer School 2019Group Members (L-R): Xuan Zhang (Univ of Georgia), David Lafferty (UIUC), Coline Dony (AAG; Group Mentor), Forrest Bowlick (Univ of Mass at Amherst), Yuqin Jiang (Univ of South Carolina), Jinwoo Park (Texas A&M), and Alexandra Timmons (UIUC).

 

 

 

 

 


 A Reproducible and Replicable Spatially Explicit Agent-Based Model Using CyberGIS-Jupyter: A Case Study in Queen Anne Neighborhood, Seattle, WA

Influenza is a contagious respiratory illness caused in humans by four species of Orthomyxoviridae influenzavirus. More than 900,000 people were hospitalized and over 80,000 people died from the flu in the 2017-2018 season (CDC, 2018). Although well-known epidemic models can capture various spatiotemporal phenomena of disease spreading, it is challenging to simulate the dynamic process and understand the complex contact network of individuals and environments in a geographic context. One approach to this challenge is the Spatially Explicit Agent-Based Model (SEABM), which this project proposed to reproduce and replicate in the Queen Anne neighborhood in Seattle.

The main objectives of this project were to conceptualize influenza transmission, simulate the spreading process, and explore the spatiotemporal patterns of the outbreaks. In doing so, we first obtained and processed geographic datasets containing households/workplaces/schools from the City of Seattle Open Data Hub, population distribution (i.e., age, sex, race, and income) and household sizes from the American Community Survey 2013-2017, and 2018-2019 weekly influenza data from the Seattle & King County Department of Public Health. We then synthesized the population in different environments and implemented the SEABM model with adjustable settings of initial parameters, such as the exposure, reproduction, and infection rate. Finally, we visualized the contact network and the number of infection cases over the course of the outbreak, animated the dynamic process of influenza transmission, and produced heat maps to study the socioeconomic patterns of influenza outbreaks.

The results of our simulation showed consistency with the 2018-2019 weekly influenza outbreak reports and successfully captured the dynamic process of influenza transmission. Our framework, methodology, and code could be easily adaptable to reproduce and replicate this in other cities or areas with the help of CyberGIS-Jupyter. Future work will explore the geovisualization of large contact networks over the course of an outbreak, improvement of model parameters in broader contexts, and coordination with observational reports.