[Talk-us-massachusetts] Hello, and some tagging questions!

[Greg Troxel]

Aria Stewart via Talk-us-massachusetts
<talk-us-massachusetts at openstreetmap.org> writes:

> Hey! I’m Aria, working on mapping a bunch in Salem.

Welcome to the list.

> I’ve got questions about the best way to approach some situations:

There is a lot here, so a few generalities first:

  We map what exists and then rendering happens, so we should not be
  particularly concerned about how things *look*, and should not enter
  objects with the intent of parks being green.  You didn't hint at this
  but it's a common and big issue.

  There are two spearate concepts, which don't appear as strongly in OSM
  docs as they should:

    land use: what humans use land for.  This tends to line up with
    property boundaries, as in "this parcel is a park".  Although it
    could be "this part of this larger parcel is a park and the other
    part is something else".  It is ok and normal to have a polygon that
    matches the property boundary and to put land use tags on it.

    land cover: are there trees, grass, rocks, sand, pavement, etc.
    This more or less by definition does NOT line up with property
    boundaries although human activity may do one thing on one side of
    the line.  We have evolved a rough consensus on this list that land
    cover tags should not be added to polygons that align with property
    boundaries.  The reason is that it's almost always wrong, and
    excludes trees outside the property, or things that aren't trees
    inside, etc.

  There have been imports and they all have their baggage, although I
  think more benefit than harm.
  
  There are also relations which have outer and inner ways when you want
  to express "this area is grass except for this sub-part which isn't."

> Parks: Salem has three notably different styles of map objects for parks. 
> - Lappin Park
> (https://www.openstreetmap.org/edit#map=20/42.52141/-70.89585) has an
> outer park area, and inner grass and footpath objects.

Assuming that's a "city park", meaning human-controlled landscape for
enjoyment (vs a nature reserve), which seems obvious given location and
imagery, then leisure=park seems right.  That should be on the parcel
boundary, more or less.  Having objects for paths makes sense.  footway
implies a presumption for walking but allowing bicycles and horses.  So
you may want to add access tags if those aren't allowed.

This park was imported from the MassGIS Open Space layer so probably
needs cleaning up.  The boundary should be some blend of matching the
parcel boundary (the actual boundary, and L3 parcels is a good guess
absent better evidence), but perhaps extending to the demarcation
between sidewalk and what feels like park.  I say this last point
because the legal road layout and the in-use road layout are often a bit
different (this is about how the world is, not OSM).

Modulo clenaing up the park object, this seems ok.

> Lafayette Park
> (https://www.openstreetmap.org/edit#map=20/42.51753/-70.89327) is
> similar, though the ‘park’ area goes right up to the centerline of the
> street. I’m never sure how precise to be vs how space filling.

Legally, the park goes to the edge of the parcel, and after that it is
road.  Practically there is a transition from park grass to sidewalk or
a fence.  The park boundary should be some blend of the two.  Definitely
not road centerline as the road is not park.

> - Forest River Park
> (https://www.openstreetmap.org/edit#map=17/42.50658/-70.88447) is a
> mess of objects that don’t seem to quite line up with my GPS, and seem
> too distinct — should the green spaces be separate from the park? Or
> should the parking be on TOP of the park? how about the wetlands? It
> all overlaps, it’s all part of the park (except maybe Pioneer Village,
> which is at least fenced off separately, though it’s accessed through
> the park)

A park can contain wetlands and paved areas for parking.  That is not
logically inconsistent.  The leisure=park tag should be on a closed way
matching the property boundary.   Then landuse for grass,
amenity=parking, natural=wetland, etc.   If you draw a big grass polygon
you may want to make it a relation and exclude the parking lot as a
strategy, but the parking lot probably has road so probably you don't
need to do that.

I don't know about Pionoeer Village.  You want to ask "does the
government consider this to be 'in the park'" and look on L3 parcels to
see if it is under a separate deed.  do park rules apply?  Am guessing
it is separate.

> - Palmer cove park
> (https://www.openstreetmap.org/edit#map=17/42.51452/-70.88898) has
> some of the same amenities as Forest River Park, but they’re unmarked
> — is that valuable to go trace out? How should they be layered and
> tagged?

Yes, it's useful.   Just keep landuse on polygons aligned with ownership
and new objects for landcover.

> Textured surface for crossings: is that the textured corners? Or the
> whole crossing that’s textured? We’ve got a few textured and raised
> crossings in town, but mostly textured corners and painted markings,
> and some untextured corners and unmarked crossings. What’s the best
> practice here?

I think it's about the edge bumps per ADA.

> The MassGIS aerial imagery and the Bing maps imagery are offset from
> the buildings on the map in town pretty pervasively. What do you
> trust? how do you deal with sidewalks that cross buildings? Is there a
> strategy that doesn’t involve moving things?

I'll try to write more later, but generally I trust the 2021 MassGIS
Aerials.  I do not trust or distrust Bing.  Buildings are from MassGIS
from LIDAR and it's surprising they are far off.  I would expecte
meter-level agreement.   Can you send an example?

(My basic advice is tread lightly for now in fixing this and don't do
anythign large scale until we can discuss more.  THese are hard things
to fix.)

Greg