Apple made a really rich Mobility Trends dataset available. I simply could not resist to download it and plot the data different than the time series plots.

I do like heatmaps, I should add, I do love datashader, as it turned out, the amount of data did not make it necessary to use it, classic bokeh worked just fine.

Obvisously, the data needed some massaging, I decided to break the dataset up into one pandas data frame for countries, and one for the cities. We are looking for “situational sensors” of what is going on around the world, so this should help us understand the level of granularity we may want and the insights we could get.

Mobility by Country

The x axes show the countries in the dataset, the y axis is the time. The colour code illustrates the percentage mobility serach requests have dropped. I am using blue-red to be colorblind-safe.

There are a few countries that stand out, in particular, Hong Kong and Macao, where mobility dropped very early compared to the other nations. South Korea shows a steady decline in mobility search queries, so does Singapore. I cannot, at this point, iunderstand the Saudi Arabia trend, but that seems to be a weekly pattern.

It is fairly obvious that all countries had severe declines in mobility around April 10…April 17. This actually surprised me as I am in week 5 (I think) of home office and have not been using Apple or Google maps to plan a trip for weeks, it seems.

Mobility by City

These data actually do look different. It is different granularity, but it really appears as if citizens of major towns really only stopped planning trips around the April 14..15 line. Japanese cities seem to remain resilinent to the current crisus. Seoul got affected very early, so did Milan and Rome.

And then there is Carneval, obiously in Rio, around Feb 22, but also in Cologne. Finally, I had been under the impression that the MWC got cancelled in Barcelona, right now I do not understand where that increase in mobility searches late February/early March comes from.

The Takeaway

We use the University of Oxford CORONAVIRUS GOVERNMENT RESPONSE TRACKER a lot, this is a wonderfully curated dataset. Mobility data, at high aggregation rates, can positively be used as sensors to understand when the public actually adopted lockdown measures, and to what extent.

Disclaimer: This information can be used for educational and research use. The author is not a health care professional and it is not recommended to use the views in this document for any healthcare related decision making.