3

Picture of centroids off coast and between islandsI am new to QGIS. Imported the census population weighted centroids shape file retrieved from NHGIS as a vector layer into QGIS and they are nearly always where I would expect them but a sprinkling of them are just off the coastline or between islands. This doesn't seem to allow for correct overlay. It automatically imports with a different CRS but I exported it as another layer with the project CRS and it looks exactly the same. What is going on? Is it supposed to be like that?

Here is the link to the census latitude and longitude data. I'm using the population weighted centroids for census block groups.

ClaireR
  • 31
  • 3
  • Maybe there's some population on some islands ? – Kasper Apr 07 '23 at 04:59
  • Yes, on the islands for sure... But in between the islands? I am overlaying boundaries and not sure what to do about that. But also sometimes, the points are off the coast. I added the latitude and longitude from the census website and it is the same, so it seems this is just how the data is. – ClaireR Apr 07 '23 at 05:28
  • 3
    Could you please share a link of the data that produces the problem you mention? I don't know if a picture is necessary but it tells more. – Kadir Şahbaz Apr 07 '23 at 07:37
  • If you have a mainlad + island or like 3 islands, it is completely reasobable that the (witghted) centroid might also fall outside of the land-polygons, depending on the geometry. Nothing special, this is inherent to the concept of centroid. See: https://gis.stackexchange.com/a/415575/88814 and https://gis.stackexchange.com/q/43384/88814 – Babel Apr 07 '23 at 09:29
  • See also: https://gis.stackexchange.com/a/236300/88814 – Babel Apr 07 '23 at 09:36

2 Answers2

4

Using the Identify tool on one of those offshore points shows that the POPULATION value is 0.

enter image description here

Ideally, you'd overlay a map of the census blocks to see if the data makes sense. You can download census block maps from https://www.census.gov/geographies/reference-maps/2020/geo/2020-census-block-maps.html (unfortunately in PDF format - there may be better sites).

As you can see from the maps, there are census blocks located off the coast - possibly to capture boats in territorial waters.

The points located off the coast all appear to have a POPULATION value of 0 (values of 0 in yellow). This appears to be perfectly valid given the census block locations. I imagine these are geographic centroids given there is no population to weight them.

enter image description here

Tom Brennan
  • 4,787
  • 6
  • 26
  • Thank you, good to know I'm not losing population there. Try as I might, I could not click on the dots to get information on them. I'm not sure what I was doing wrong. I could figure out the assigned county by filtering them. NHGIS provides the shape files modified: "Most substantially, we erase coastal water areas to produce polygons that terminate at the U.S. coasts and Great Lakes shores" but they apparently use the centroids as given by the census bureau. Not sure the value of allowing them to go offshore. – ClaireR Apr 08 '23 at 05:03
  • 1
    Ctrl+Shift+I gets you the Identify Tool, and you need to have the Identify Results Panel enabled (View->Panels->Identify Results) – Tom Brennan Apr 08 '23 at 07:21
  • Get shapefiles for census blocks for Washington state here: https://ofm.wa.gov/washington-data-research/population-demographics/gis-data/census-geographic-files – Babel Apr 08 '23 at 10:18
2

It is completely normal to have census block that are uninhabited, see: Nobody lives here: The nearly 5 million Census Blocks with zero population

You should make yourself comfortable about how the data (weigthed centroids) were modelled based on raw data. I downloaded census blocks data for Washington state from here: https://ofm.wa.gov/washington-data-research/population-demographics/gis-data/census-geographic-files

There is e.g. the centroid at +48.237194, -123.919819 and overlay of census tracts shows that it belongs to TractCE 990110, county Clallam - and neighboring polygons belong to the same tract/county:

enter image description here

Now dissolve and merge all polygons with the same TractCE and Countyname: use aggregate tool) or simply select them withTRACTCE= 990100 and COUNTYNAME='Clallam'. You see that you get a weirdly shaped polygon, completely inside water (see next screenshot). The centroid is indeed in the middle of it:

enter image description here

Babel
  • 71,072
  • 14
  • 78
  • 208
  • Thank you, Babel. Your example of how the weighted centroids are developed is just geographical, correct? Not population weighted centroids, right? – ClaireR Apr 08 '23 at 13:08
  • 1
    When a census tract has 0 inhabitants, centroids can only be calculated geographically - no possibility for "weighted" – Babel Apr 08 '23 at 14:06