Talk:Import/Past Problems

Chessboard forests in the U.S.

This analysis badly misunderstands the nature of the imported USFS data:

Data imported from http://data.fs.usda.gov/geodata/edw/edw_resources/shp/S_USA.BasicOwnership.zip is grossly wrong. Forests have a chessboard appearance probably derived from an attempt at very low-resolution half-toning. That is, there are regular squares of forest where none exist on the ground. The squares are one mile wide. For a good example, see the area NE of Grass Valley, California, U.S.A. This data ought to be deleted. Sample way: http://www.openstreetmap.org/way/370128450#map=15/39.3965/-120.4460 .

The file downloaded from the USFS represents land ownership, not vegetation cover, and the half-toning hypothesis completely misses the mark. Historically (mostly in the 1800s) it was common practice in the American West for the federal government to distribute land title to state or private entities in exactly the 1-mile-square checkerboard pattern seen in that file. As frontier territory was annexed to the U.S., the "unoccupied" land (apologies to the prior occupants, especially Native Americans) was deemed to be federally owned by default. The federal government would then distribute the ownership (in some areas) for various purposes, such as providing revenue-producing land to support local schools, or incentivizing private construction of transcontinental railways. This distribution created the checkerboard pattern seen in the import, including within boundaries (e.g. national forests) that would otherwise suggest uniform federal ownership. But this checkerboard ownership usually does not produce corresponding vegetation cover patterns that would be visible in aerial photos.

It is a fair philosophical question whether OSM wants to represent the distinction between federal/non-federal land title: I can see good arguments both ways, and the ultimate solution may be situational. BUT the USFS data is not "grossly wrong", and the analysis quoted above draws the wrong lessons. The issue isn't about data quality, but about potential conflicts between the purposes and goals of OSM vs. the agencies that provide data to us.

The prior analysis is far enough off the mark that I will delete it, in hopes that someone will provide a better discussion of the import issues in the future. --Werewombat (talk) 04:06, 8 December 2017 (UTC)

Clarification of chessboard forests

I accept that my original analysis was mistaken, and have corrected it; it needed fixing, not deleting. The data is misleading, and I have pointed out that if it is to represent vegetation or land use, it should be edited. I don't agree that it's a 'philosophical question whether OSM wants to represent the distinction between federal/non-federal land title'; it's actually a question of tagging policy and the semantics of the map. There are quite a few ways to tag forests, and it's a difficult area. In this case it's hard to see how using the raw land ownership boundary is helpful. --Graham Asher (talk) 09:31, 8 December 2017 (UTC)

Your fix seems perfect to me. Good job - I couldn't have done as well. Werewombat (talk) 19:59, 12 December 2017 (UTC)