[Talk-us] natural=* and landuse=* multipolygons at the urban interface

Steve Friedl steve at unixwiz.net
Tue Aug 15 13:37:55 UTC 2017

The challenge with the Scrub from Hell is that it’s mostly one huge area (role=outer in 8 segments) and 40 or something role=inner that provide “cutouts” for things other than natural=scrub;  this could be a lake or it could be a city or whatever. This is the largest single object I’ve ever had to work with in OSM.

I guess I’ll start looking for natural boundaries to split the big area into smaller ones which will make them more manageable (and separate). 


Some have suggested just removing the scrub entirely,  but this is going to make the Santa Ana mountains look lousy; it’s almost entirely scrub, and that’s where I hike.




From: Nathan Mixter [mailto:nmixter at gmail.com] 
Sent: Monday, August 14, 2017 2:12 AM
To: OSM Volunteer stevea <steveaOSM at softworkers.com>
Cc: talk-us <talk-us at openstreetmap.org>; David Kewley <david.t.kewley at gmail.com>; Rihards <richlv at nakts.net>; Steve Friedl <steve at unixwiz.net>
Subject: Re: [Talk-us] natural=* and landuse=* multipolygons at the urban interface


Steve is indeed correct. ID isn't really designed as an editor of complex relations since it is browser based. In JOSM, you can select the relation and all the members 
and easily see how the elements interact. That being said, it is fine to simplify the FMMP relationships by splitting them into smaller parts like you did or even to delete 
the relationships altogether. Especially in Orange County, it might be easier just to delete it. 

The FMMP data are designed primarily to focus on farms. The original designation of "farm" has now been changed to "farmland". Unfortunately it doesn't break it 
down further to separate things like orchards and vineyards. The data are usually pretty good in separating farmland from everything else. But the everything else is where
it gets vague. The FMMP data are not designed to break down landcover designations like grassland, scrub, meadows or woods.
In that sense, it really isn't a landcover import.

Even with true landcover imports like Corine, the distinction between what is classified a certain way is often arbitrary compared with what you would expect to find by 
looking at imagery.

Often the FMMP landcovers and landuses will be grouped together as non farm land. Not too helpful for OSM purposes. If there is no actual farmland like in the area 
around Orange County it makes sense just to delete it entirely since nothing is gained if the area covered is too broad. 

Development in Orange County, California pushes into areas currently
covered by polygons (often large multipolygons) tagged as natural=scrub,
landuse=meadow, or landuse=[farm|farmland]. These were part of the FMMP
import http://wiki.openstreetmap.org/wiki/California_Farms.
Mostly I try to leave those large multipologons alone, because I don't feel
confident I can handle them properly, and because I'm using iD (due to
using a Chromebook), where relation handling is rudimentary.
But I'd like to update the urban-wildland boundary, where new suburban
developments are pushing into former wildland, farmland, or (historical?)
"grazing land". See for example the new development (with 2017 imagery
recently added to Bing) at
Editing these huge multipolygons, and reviewing others' edits to them,
becomes very cumbersome, at least to me. It seems to me probably sensible
and reasonable at the urban edge to split off small parts of these
multipolygons, e.g. at roads, to make the smaller bits easier to edit and
review in the context of the expanding urban edge.
As one test / demonstration edit (
http://www.openstreetmap.org/changeset/51090963), I carved off a bit of
natural=scrub from a large outer role of a multipolygon, into its own
polygon. I manually added new boundary way segments, stitched them together
into the existing ways, copied tags, and made the split-off piece its own
polygon, independent of its original parent multipolygon. I did the split
at an existing highway=residential object (Golden Ridge Lane).
I know, I should find a way to use JOSM, which I expect makes this much
easier. :)
Meanwhile, does this seem a reasonable approach to making the urban
interface a bit more manageable in the future? I.e. splitting off parts of
large multipolygons (so long as they don't have names or other unique
identifiers that matter, just generic tags things like natural=scrub), to
make future editing easier?
I know for the above example of a new residential area, I could make a
landuse=residential island, and make it an inner role in the surrounding
landuse=meadow multipolygon. But at some point as the urban sprawl expands,
it seems to me it makes more sense to stop pretending the area is dominated
by the natural features, and make it clear it's dominated by e.g.
landuse=residential, with possibly interspersed natural features like scrub.
What would the group suggest?
Is my test edit reasonable, or should it be reverted?
P.S. As an aside (not my main point today), the FMMP-based distinction in
this area between scrub and meadow seems awfully arbitrary. I could be
mistaken, but I don't believe the "meadow" is actually used today for
grazing nor feed harvesting, and in the aerial photography, it appears
indistinguishable from the adjacent "scrub". It appears (and I'm nearly
certain from driving by) that there's both substantial grass and
substantial woody plant cover, in similar ratios in both "meadow" and
I don't believe there's any current agricultural use of that land, at least
not near where I'm giving examples today. There might be some
large-acreage, semi-wildland grazing or feed harvesting activity remaining
in Orange County, but I've not noticed any.
As documented in the FMMP wiki page
http://wiki.openstreetmap.org/wiki/California_Farms, the FMMP designation
"Grazing Land" was mapped to landuse=meadow.
But the FMMP designation of "Grazing Land" explicitly does not mean that
there *is* grazing activity there, just that it is "...land on which the
existing vegetation, whether grown naturally or through management, is
suitable for grazing or browsing of livestock." (See for example
http://www.conservation.ca.gov/dlrp/fmmp/Documents/soil_criteria.pdf.) So
wildlands that will never again see livestock, or harvesting for livestock
feed, can still be designated Grazing Land by FMMP. Those areas map better
to natural=grassland or natural=scrub, I think.
So landuse=meadow seems less useful than natural=scrub or natural=grassland
for many of these areas. Even though this is a secondary point today, I'd
welcome comments on this as well.



Virus-free.  <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link> www.avast.com 


On Sun, Aug 13, 2017 at 9:29 PM, OSM Volunteer stevea <steveaOSM at softworkers.com <mailto:steveaOSM at softworkers.com> > wrote:

David, I would contact Nathan Mixter directly (in OSM, nmixter, import account Eureka gold) and ask him what he thinks, as he is (largely speaking) the original importer of these (and many other, very large) imports, many of which, unfortunately generated consternation or reversion.  You might ask him what his plans are to "upkeep" the data he has imported.

Nathan is a friend of mine I met in OSM (on a personal and "let's go hiking/camping/backpacking together" level) and I have helped him on both improving the Santa Cruz County (my home) and Monterey County (next door to both of us) landuse imports that he initiated.  Together, we did the single-county FMMP import of Monterey County (only, I didn't help with other counties) over many months (instead of the days Nathan thought it might take) as I wanted to convey the care, vetting, quality assurance and teamwork that such an endeavor truly requires to get it right (or much closer to right, as I still think Monterey County's landuse from this import is "pretty good," if I say so myself).  I/we documented what we did if you click around the links in our wiki, already introduced in this thread.

In short, these landuse polygons are indeed very large, unwieldy or virtually "just kill me now" highly difficult to edit using iD (PLEASE use JOSM to edit complex polygons like these!).  I declare that they aren't anything "sacred," especially as new human urban development simply outdates more and more edges of these data as obsolete.  Subtle differences between scrub and meadow, while I admire your diligence in determining "what is best" for a given area, are not hard-and-firm.  I'd characterize these FMMP imports as "2010-12 data, roughly applied to OSM to avoid large blank areas in California" (except Monterey County, were I was very careful to apply the lipstick carefully so there was no piggy ugliness about it).  So, should these FMMP import (multi)polygons need to be changed, edited, modernized and especially trimmed down to more manageable size, please, get a read from Nathan if you can, then take the controls of JOSM firmly in your hands and go for it!  Especially as those bulldozers build those suburbs.

Nathan, you might please chime in either on-list or via email to this distro; thank you.  If you wish, I additionally invite anybody to contact me off-list to ask about this topic should you care to know further details, though Nathan is the primary importer of these data.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20170815/ca2ca3b8/attachment-0001.html>

More information about the Talk-us mailing list