[Tagging] Watershed or Drainage Basin relation draft proposal

Christoph Hormann osm at imagico.de
Thu Sep 13 11:22:04 UTC 2018

On Thursday 13 September 2018, Joseph Eisenberg wrote:
> Christoph,
> So you believe the ridges are verifiable (and the network of
> waterways, I assume), but potentially parts of the watershed would
> not be verifiable because eg. terrain is too flat? 

There are many reasons why a the watershed structure can be practically 
non-verifiable on the ground.  Relatively flat topography is only one 
of them.  I would estimate that at least about 40 percent of the Earth 
land surfaces have a drainage structure that is practically 
non-verifiable (in the sense that independent estimates from several 
mappers would not converge to the same geometry).  The percentage would 
be even higher if you want it to be verifiable through remote mapping.

> I was thinking 
> that in fairly flat areas it is still possbile to see which way water
> flows in drainage channels, and it's often possible to find the
> highest line throught he terrain when surveying.

Nothing speaks against mapping physically observable features like 
drainage channels but mapping additional abstract geometries derived 
from them in OSM makes little sense IMO.

Note in flat areas with artificial drainage channels the actual drainage 
structure can be extremely complex.

> Would these examples be verifiable?
> [...]

Not having first hand knowledge of these cases means i can't really 
tell.  A waterway is a geometry directly physically observable on the 
ground.  The watershed divide is an abstract geometry OTOH. You see a 
ridge and *assume* that water from one side of the ridge flows into a 
certain drainage system and on the other side it flows into a different 
one.  But you don't observe this.  You might have a simple case where 
this seems obvious but the fundamental problem is still there.  In 
humid climate areas you can make the mapping more reliable by first 
diligently mapping the waterway network in the whole area but then what 
is the point in mapping the watershed divides in addition?

Also note in priciple watersheds form an infinitely deep hierarchy of 
geometries.  To be able to practically map these you would have to 
define a discretization system in this hierarchy which would be an 
arbitrary up-front decision with no basis in the physical world.

> I value your opinion Christoph, because I hoped this relation might
> encourage more complete mapping of ridges, watersheds and drainage
> basins, thus making it easier to render good maps, eg
> http://www.imagico.de/map/water_generalize_en.php

If you want to facilitate better maps w.r.t. hydrography the best way 
would be to put all the time and energy you might put into mapping 
watersheds into mapping the waterway network. Priorities should be:

* correct connectivity
* correct distinction onto artificial and natural
* correct flow direction
* completeness and unifomity in detail of mapping

Where these conditions are fulfilled processing the waterbody data for 
accurate maps is orders of magnitude easier than elsewhere.  Compared 
to that the would-be gain of having watershed geometries available in 
addition would be relatively small.

Christoph Hormann

More information about the Tagging mailing list