For fun, I have been playing with making population density visualization with three.js and the WorldPop dataset for global population at high resolution.

Here are some places that I've lived in recently:

Bay Area
Taiwan
Tokyo

Perhaps I will finish this viewer and post it in the future, and write a blog about it, but for now it's pretty jank and just a few hours of me and LLMs.

The WorldPop dataset issue

I picked the WorldPop dataset for this, as I was interested in animating population vs time. It turns out that doing so is not all that interesting. But, it led me to something else.

It looks like the methods that WorldPop is using to project population out to 2030 don't quite give sensible results. See the below animation, we are looking at southern California, with Las Vegas off in the distance.

Los Angeles San Diego Las Vegas

As the year counts past 2025, we see something quite strange. Irvine population absolutely explodes, far exceeding downtown LA. And somewhere in south los vegas the same happens. Something must be off in how the forward population projections are done.

From the release statement:

The previous ‘Global1’ 2000-2020 data exhibited a phenomenon in some locations that we called “the donut effect”, whereby population density was observed to decrease in city centres and increase in neighbouring rural areas between annual gridded estimates. The phenomenon was negated in the updated Global2 timeseries modelling using a new approach. Whereas the Global1 2000-2020 data were modelled using a RF approach that disaggregated the whole population for each year in the annual timeseries, the new approach for the Global2 timeseries disaggregated the difference in population between the target year and preceding year on top of the preceding year’s gridded dataset (e.g. for the 2020 gridded dataset, the population difference between 2019 and 2020 was modelled “on top” of the 2019 gridded dataset).

Perhaps, this artifact comes from using differences in population over areas that have had a lot of recent population growth, but in such a way that clearly will not hold in the same trend over the next 5 years.