Element 84 Logo

COVID-19 Cases Over Time with Kepler.gl


This blog post discusses how to quickly visualize data, like that of COVID-19. We are not medical experts and the visualizations have not been peer reviewed. We recommend visiting data sources, like the CDC, for data on COVID-19.

Data visualizations have been crucial to communicating the severity of the COVID-19 pandemic. Problems like this benefit from visualizations that have spatial, temporal, and magnitude axes. Many of the visualizations I’ve seen omit one of these components. For example, on some sites, maps display current counts of cases but don’t show trends over time. Bar charts can show changes across time but you lose the spatial component.

One solution is an animated map which brings together the three dimensions:

  • Location on the map represents physical location
  • Color, size of icons, or other indicators can show magnitude
  • Animating changes shows variations across time.

This video shows COVID-19 data displayed within Kepler.gl, an open source geospatial analysis tool.

COVID-19 spread across the US

View the interactive map

  • A point represents a single county reporting cases of COVID-19.
  • The color of the point and its radius shows the number of cases in that county.
  • A date filter adds a slider that allows seeing the data at different points in time.

Kepler.gl is tool developed and open sourced by Uber. It runs completely in the browser without additional software. It’s easy to select and load a dataset and then configure how you want to visualize the data without programming.

Limitations and Future Improvements

Since this is county-based data the map should show the county boundary. It was easier for me to set it up as points with varying radiuses. When you zoom out the points overlap and are a bit crowded. One improvement would be collapsing spatial boundaries as you zoom out. For example, as you zoom out, display state boundaries and counts instead of county boundaries.

Kepler.gl tries to be field aware and does make good guesses about data formats. It can still be a bit finicky with data formats. For example, the date slider only works with dates that have a time component. I had to append “T00:00:00Z” to the date field to enable this feature.

Data Sources

NY Times provided the source data for this visualization. Source

I added fields for totals per county and latitude and longitude. Download that data here: https://covid19.labs.element84.com/us-counties-with-lat-lon.csv

Jason Gilman

Principal Software Engineer