Notices by Eldan Goldenberg (eldang@mapstodon.space)

I spent a couple of hours helping a friend learn how to use QGIS. She got her project done, figuring most things out for herself. Today I got back to my office after a little time away, and found this waiting for me with a thank you card.

[original graphic from https://github.com/souperneon/null-island ; artist's page about it at https://www.ektad.com/projects/null-island ]

I just want to rave for a minute about this world map of the spread of humans: https://vis.social/@infobeautiful/113527888543173510 (h/t @trochee )

First, it's a great use of the "Dymaxion" projection, which I don't often see in the wild, and rarer still without the distracting triangular cutouts of the ocean. I usually care too much about the oceans to want to use this one, but it's a good fit for the story being told here. (1/2)

But perhaps more importantly: it's a reminder of how comparatively recent human expansion into most of Europe was. I'm more familiar with the arguments over the peopling of the Americas, which if anything may have happened earlier than shown here. But in that whole argument, I tend to forget that the whole range of estimates has people continuously present where I live now (NW North America) much longer than where I grew up (Britain).

@dymaxion ...which keeps forgetting to clean its glasses?

I have a new favourite geospatial open data bug: US Federal Register 220519-0117 "Change to County-Equivalents in the State of Connecticut".

Sooo... a thing I'm working on for a client is to show various sorts of demographic data at the US Census Tract level, to serve as context layers for the main thing we're building. Most of them come from the 2020 Census, and are straightforward enough to handle: one set of geometries with consistent IDs that all the stats tables can be joined by. (1/?)

There are a couple that are drawn from the American Community Survey, which uses 5-year averages to make up for its comparatively smaller sample size in each year. Since the latest 5-year window started before the 2020 Census, they have to be mapped as 2010 Census Tracts. So far so good.

I noticed that the entire state of Connecticut was missing from one of those layers, so I had to investigate. Sure enough, it's in the data table, so what's going wrong here? (2/?)

I learned that there are two distinct sets of FIPS codes (if you don't know what a FIPS code is, all that matters here is that it should be unique and consistent). One of our data sources was consistent with the Tract geometries, and had Tract names like "Census Tract 1001, Fairfield County, Connecticut". The other had different codes, and names like "Census Tract 1001; Greater Bridgeport Planning Region; Connecticut"

Argh. (3/?)

Once I found that Federal Register announcement I could piece it together. From 1960 until 2022, the US Census Bureau was using outdated labels for subdivisions within Connecticut. Starting in 2022, they now use the up to date planning region names, which have been issued new FIPS codes because they don't have a perfect 1:1 correspondence with the old counties. The problem for me is that while the 2018-2022 ACS 5-year estimates have to use 2010 Tract geometries, they use the 2022 codes. (4/?)

@mastohost @jeremy That's reasonable! You have to keep your own workload manageable. And you're providing a great service here, thank you for it.

@jeremy would you be willing to install this patch? https://queer.party/@maffsie/109367901406518202

This may be partly because I'm used to Mastodon 3.x behaviour as @eldang@weirder.earth, but I keep being frustrated by copying the URL for someone's profile or toots and getting the captive replica of it within mapstodon instead of their actual page.