VNE is a cyberGIS-based visual analytics tool that allows users to (1) delineate neighborhoods based on their selection of variables describing socioeconomic and demographic profiles and (2) explore which neighborhoods are susceptible to the impacts of disasters based on specific socioeconomic and demographic characteristics. Firefox or Google Chrome is the recommended web browser for reaping the best performance of VNE.

• Getting Started
• Input parameter description
• Example Result Visualization
• A Case Study
• Interactive Features of the Visual Interface (Tutorial Videos are included)
• Data
• Related Resources
• Contributors
• References
• License

To use it in CyberGISX, follow steps below:

1. If you do not have a CyerGISX account, create a CyberGISX account with your institution's email address or Google email address at https://cybergisxhub.cigi.illinois.edu
2. Once you log in CyberGISX, go to https://cybergisxhub.cigi.illinois.edu/notebook/vulnerable-neighborhood-explorer, and click the button "Open with CyberGISX". Wait for 3 to 5 seconds. All related source codes will be automatically copied, and the main notebook will be opened. Then, click the “play button” at the top to run.

A video tutorial on using VNE in CyberGISX

Note: The video is currently muted by default. Please unmute the video before playing to hear the audio.

Example visaulizations are available in the folders below:

• VNE_Chicago
  
• VNE_Chicago2
  
• VNE_Chicago_kmeans_C5
  
• VNE_Chicago_kmeans_C5
  
• New_York_kmeans_C5
  
• Phoenix_kmeans_C5
  
• Miami_kmeans_C5
  
• US_kmeans_C5
  
• Chicago_extended_kmeans_C5
  
• Illinois_kmeans_C5
  

Exploring Neighborhood-level Social Vulnerability to COVID-19 in Chicago.

Two images (Fig 1 and 2) below show the result visualization of VNE, which allow users to explore socioeconomic and demographic disparity in COVID-19 outbreaks as well as vulnerable neighborhoods and their socioeconomic and demographic characteristics. Two input data were used:

• COVID-19 confirmed and test cases at the zip code level in Chicago. They were downloaded from the website of the Illinois Department of Public Health. The data used reflects the duration of the COVID-19 outbreak until July 11th, 2020, when the data was downloaded.
• American Community Survey (ACS) 5-year estimates from 2014 to 2018. From ACS, 18 variables representing different socioeconomic and demographic statuses were collected.

Below is a detailed description of each chart and map that can be visualized using the Vulnerable Neighborhood Explorer (VNE).

• In Fig1B, users can select disaster-related data via a dropdown menu, choosing between total infection rates or rates categorized by race/ethnicity (e.g., White, Hispanic) at Zip Codes outlined in white. Fig1A displays the density distribution of the selected data, providing geographic context

• Fig1C presents neighborhood delineations obtained through k-means clustering of the eighteen American Community Survey variables at the census Zip Code level (as labeled on the x-axis in Fig 2A). Each neighborhood typically comprises multiple Zip Codes.

• Fig1D shows the proportion of Zip Codes in each cluster, serving as a legend. Clusters include "White Rich Owner," "White Hispanic Aging Suburban," "Asian Elite Renter," "Hispanic Laborer," and "Black Poor." Users can name each cluster by analyzing the Z-score means of the variables using the heatmap and bar charts in Fig 2A and 2B. This can be updated in Config.js by clicking the third link generated by VNE. For labeling instructions, see [Han et al., 2023].

• Fig1 E-G display column charts of the total infection rate and the infection rates among White and Hispanic populations within each neighborhood, respectively.

• Fig1 H-CJ complement these with box plots showing the distribution of these rates across the five clusters.

• Fig 2A and 2B illustrate the mean z-scores of each input variable, highlighting the distinguishing characteristics of the clusters in Fig1C. Z-scores indicate how far each variable's value is from the mean.

  • Fig 2A compares the mean z-scores of the eighteen variables across the five neighborhoods.
  • Fig 2B shows the sorted mean z-score for a selected neighborhood which can be chosen using the dropdown box at the top of this chart. In this case, five horizontal bar charts representing the mean z-score of each of the eighteen variables can be created on the interface of the interactive map.

Figure 1: The initial part of the Vulnerable Neighborhood Explorer (VNE) visualization.

Figure 2: The subsequent part of the Vulnerable Neighborhood Explorer (VNE) visualization.

VNE is a geovisual analytics tool that utilizes a Coordinated Multiple Views (CMV) approach, allowing users to interact with several interconnected visual representations simultaneously. This enhances the analysis of complex data relationships and facilitates the extraction of meaningful insights. Here is a video showcasing the visual features of the Vulnerable Neighborhood Explorer (VNE).

Note: The video is currently muted by default. Please unmute the video before playing to hear the audio.

• Overview & Cross-Filtering for Maps
• Local Visualization for Distribution & Proportion Charts
• Disaster Impact Across Neighborhoods: Column Chart & Box Plots
• Tukey's Honest Significant Difference (HSD) Test
• Exploring Neighborhood Characteristics Using Z-Score Means
• Change Subjects
• Advanced Options & Fixed Axes for Direct Comparisons
• Where Results Are Saved & Sharing Your Visualizations

A key feature of CMV is cross-filtering, where actions in one view—such as selecting a region in a chart—automatically update related views. For instance, consider three visualizations: a map showing COVID-19 infection rates by Zip Code (Fig 3A), a map displaying different neighborhood types (Fig 3B), and a proportion chart of Zip Code areas per neighborhood (Fig 3C). When a user hovers over a neighborhood category in the proportion chart, both maps update to reflect the selected area. Hovering over "C0 White Rich Owner" highlights the data in this neighborhood on both maps, showing a generally low infection rate (Fig 3D, E and F). Conversely, hovering over "C4 Black Poor" highlights that neighborhood, indicating a higher infection rate (Fig 3G, H and I). This interactive cross-filtering enables users to compare infection rates across neighborhoods, focus on specific areas to identify localized trends, and understand how different neighborhoods relate to infection rates. In summary, VNE's CMV and cross-filtering capabilities allow for dynamic exploration and analysis of disaster-related data, such as COVID-19 infection rates, across different neighborhoods.

Figure 3: Partial visualization of the VNE output. Panels (D) and (E) show the map view when the C0: White Rich Owner neighborhood is selected in the proportion chart (F). Panels (G) and (H) show the map view when the C4: Black Poor neighborhood is selected in the proportion chart (I).

Another key feature of VNE is brushing, where highlighting a specific data point in one visualization, simultaneously highlighting the corresponding regions across all linked maps. On the right in Fig 4, the box plot depicts the distribution of COVID-19 infection rates across five neighborhoods, with each dot representing the infection rate of a specific Zip Code area. Users can lasso-select these dots (Zip Codes) on the chart, triggering the corresponding Zip Code areas to be highlighted on the two maps. In this example, the box plot focuses on the Hispanic laborer neighborhood, revealing a higher distribution of infection rates compared to other neighborhoods. By selecting areas above the 75th percentile within this neighborhood, the highlighted Zip Code areas represent the top 25% of infection rates in the Hispanic laborer neighborhood.

Figure 4: The result visualization is created by setting only the box plots to 1 and leaving the other chart options unselected.

VNE integrates synchronized map views with cross-filtering capabilities to enhance data exploration. When a user pans or zooms one map, all linked maps automatically adjust to display the same geographic area. This synchronization enables simultaneous analysis of COVID-19 infection rates and neighborhood types within focused regions. Additionally, zooming into a sub-region on the maps triggers the distribution chart to dynamically display the infection rate distribution within the current map boundaries. Furthermore, the chart illustrated in Fig 5D updates in real-time to show the proportion of polygons within each neighborhood. Fig 5 presents maps of a region that has been zoomed in from Fig1 A-D. Please note that the distribution charts (Fig 5A) and the proportion chart (Fig 5D ) appear differently in these figures, reflecting the views before and after zooming. Specifically, C2: Asian Elite Rent is not displayed in Fig 5D because the zoomed-in map in Fig 5C no longer includes this neighborhood once zoomed from the view shown in Fig 1C.

Figure 5: Partial visualization of the VNE output. (A) illustrates the distribution of the variable selected in (B), where (B) represents the COVID-19 infection rate. (C) and (D) depict neighborhood types. These images are screen captures taken from the visualization featured in the accompanying video.

• American Community Survey (ACS) 5-year estimates from 2014 to 2018
• COVID-19 Zip Level Tests and Cases

• CyberGISX
• Leaflet
• PlotlyJS
• D3
• GEOSNAP

Su Yeon Han¹, Joon-Seok Kim², Jooyoung Yoo³, Jeon-Young Kang⁴, Alexander Michels⁵, Fangzheng Lyu⁵, Furqan Baig⁵, Jinwoo Park⁵, Shaowen Wang⁵

¹ Geography and Environmental Studies, Texas State University, San Marcos, TX, USA
² Computer Science, Emory University
³Spatial Sciences Institute, University of Southern California
⁴ Department of Geography, Kyung Hee University, South Korea
⁵CyberGIS Center for Advanced Digital and Spatial Studies, University of Illinois at Urbana-Champaign, Urbana, IL, USA

Han, S. Y., Kang, J. Y., Lyu, F., Baig, F., Park, J., Smilovsky, D., & Wang, S. (2023). A cyberGIS approach to exploring neighborhood‐level social vulnerability for disaster risk management. Transactions in GIS, 27(7), 1942-1958.

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

If you have questions, please contact Dr. Su Yeon Han at su.han@txstate.edu at Geography and Environmental Studies, Texas State University