[HOT] Tracing tagged buildings
travis.driessen at pdx.edu
Mon May 4 17:48:12 UTC 2015
Hey Klaus, Robert, et al,
My name is Travis Driessen and I am a Urban Planning master's student
studying smart growth here in Portland, Oregon. Klaus, Robert Pierre, you
have some great points.
I want to weight in on the nodes vs. polygon in terms of housing/building
mapping. I think Klaus has some great points and in terms of logistical
operations, speed, and the optimal use of the data it would seem the
following order would be the most useful for emergency responders (and
later for long-term planning). I am very new to HOT so it is possible that
some issues that are being raised by newbies may have been dealt with by
the HOT community, but I think it is useful none the less to be considered.
1) Land Use Polygons: For emergency responders entering new areas, it would
seem just first knowing the general area of the city/village in terms of
Residential, Commercial, Industrial, Agricultural, (other categories) would
be first priority.
2) Density: It seemed from the discussion that housing/building shapes
where being used to determine density and areas where people live. This can
be done with point density analysis. A centroid of polygon is taken anyway.
All points can then be spatially joined to the land use polygons and you
will have values for priority rescue ops. Similarly in transportation
network analysis we just use land use parcels centroids which then get
snapped to street intersections.
3) Speed: The amount of time making a point file compared to the amount of
time to draw a polygon is minutes to seconds. Allowing the OSM community to
dramatically map priority areas and help determine strategic locations for
rescue ops based on the most people to be attended to in the hours and days
following a disaster.
4) Aerial Imagery: The quality of aerial imagery did not allow for
polygons to be correctly shaped. It would seem that, while people are
making points from the existing data, drones could be sent out for
reconnaissance of quality aerials to support future waves of improving the
5) Iterations: The data will never be perfect, but it can always be
improved. Downloading a point file data set after there is better quality
aerial data and then drawing polygons if need be. I guess I need to read
more up on why polygons are eventually needed (do they help emergency
responders figure out where to enter a building?), but from an urban
planning perspective in terms of where roads need to be built, where water
and sanitation infrastructure should be built, electricity, etc. this all
can be done from simple point nodes.
On Mon, May 4, 2015 at 10:06 AM, Klaus Hartl <k127 at gmx.de> wrote:
> Hello Robert,
> thank you for your response!
> Regarding your second remark, which is quite the unemotional and pragmatic
> evaluation of my notes that I was hoping to receive, I see that it makes
> sense to change my workflow.
> I won’t map any further buildings as nodes then.
> Since other mappers could face the very same decisions, please let me
> point out how I came to my odd decision to map buildings as nodes:
> Whether or not we call a mapper experienced, I don’t see experience as to
> know tagging rules by heart. Since these could change over the years, just
> like visualization rules do, it does matter how those rules are
> recapitulated in case of need. In my case, like I did, I read the *schema
> specification for the key building*, and nothing more since
> attributing *a node is not denoted invalid* there*:*
> *Note about using this tag on nodes : although buildings are better
> represented with their footprints (a closed way or a multipolygon
> relation), OSM is working by iteration and some areas in the world don't
> have good aerial imagery or public datasets offering building footprints.
> Therefore, buildings on nodes should be tolerated until better sources are
> And that’s where I see the odd and thus a risk of this (anti)pattern to
> Maybe we could adjust or refine either the specs or our judgement on
> applying these specs in order to arrange this procedure more even.
> Is there any opinion on that?
> *Klaus / k127*
> p.s.: I just took a look at the *Building Tools* Plugin for JOSM,
> which kind of supersedes my two-pass contribution approach by providing a
> neat two-and-a-half-click action for creating a perfect, orthogonal
> building shape.
>  http://wiki.openstreetmap.org/wiki/Key:building
>  http://wiki.openstreetmap.org/wiki/JOSM/Plugins/BuildingsTools
> Am 04.05.2015 um 14:11 schrieb Robert Banick <rbanick at gmail.com>:
> Hi Klaus,
> *First of all,* thanks for providing such a measured response to a not
> very measured message. I’m sorry you got such a rude message in the first
> place and want to assure you that it doesn’t reflect HOT’s attitude, both
> stated by the organization and unstated within the community, towards
> errors by new contributors. Everyone has to start somewhere and errors are
> *Secondly*, I do have to agree with the point of the message. The fact
> is your iterative work process doesn’t fit with the contribution-validation
> process HOT has set up to make it easy for everyone to work together.
> There’s no graceful way in the technical tools or HOT’s workflow to reflect
> that buildings-as-nodes are a transitional step by you towards perfect
> data. Thus it creates the potential for others to waste time “correcting”
> what seems like a mistake.
> I can understand how this system would work really well when you’re
> managing a task or area by yourself. But HOT tasks are done with others and
> the system is designed so that we build on one another’s work. Also
> consider that no responding agencies are looking for buildings as nodes and
> hence your transitional data adds no value until entered as an area.
> *Finally*, a gentle reminder to experienced: if you encounter systematic
> errors from users, however seemingly basic or disastrous, please give them
> the benefit of the doubt with a *nice* message. Inviting new users to
> contribute guarantees mistakes, but it creates a lot more value than harm:
> we have to accept the mistakes as part of the process. If I was a new user
> and read the message above I guarantee I would be scared off mapping — and
> that means HOT just lost a potential longtime contributor.
> Sent from Mailbox <https://www.dropbox.com/mailbox>
> On Mon, May 4, 2015 at 11:14 AM, Klaus Hartl <k127 at gmx.de> wrote:
>> Hi HOT,
>> with this message I’m not particularly answering the previous one rather
>> than intending to jump on that topic due to some misunderstandings I got
>> notified by concerned users via private message (which I’ll post here), on
>> which a little clarification is needed. If the following issues are
>> clarified elsewhere, I’d like to thank you for that notice in advance and
>> excuse any double posting.
>> Some OSM mapper wrote to me:
>> Hi I'd like to let you know that the Task #658 (
>> http://tasks.hotosm.org/project/1018#task/658) is a complete mess thanks
>> to you and a few other users. Why don't you read the instructions before
>> starting to work on the map? You've entered thousand of buildings as nodes,
>> when instruction states that buildings has to be entered as polygons and
>> now someone is going to waste precious time in order to correct your
>> errors. I hope this was your only mistake. I'm not going to waste any more
>> time by writing to you; please, read carefully the instructions BEFORE any
>> Have a nice one Regards
>> I’d like to post my reasons to this list so that
>> - it can be validated by all and
>> - further misunderstandings can hopefully be avoided
>> Hello […],
>> Thank you for your friendly notice and for honoring OSM volunteer work.
>> You're definitely right proposing to trace buildings as areas.
>> Hence, I am fully aware that I created information what you might
>> consider a lack of quality.
>> However, the reasons for registering buildings like I did are these:
>> 1. I was working on an HOT task (in case you don't know, please see
>> ). Buildings are not part of the main objective of this task, which is
>> rather to find flat spaces suitable for potential helicoper landings among
>> 2. Regarding my paradigm of contributing to OSM data more
>> generally, I tend to improve data quality in several iterations, this means
>> to break up the task into various pieces (which of course have to be
>> consistent), if it isn't justifiable to solve the task as an one-off (cf.
>> 1.). The first iteration in the given case would be to register buildings
>> as quickly as possible. Technically spoken, in JOSM, I copy one building
>> node and then per instance point the mouse cursor on the right spot while
>> pasting. You're right when you call this far from perfect, but it's
>> something me or others can start from later. And regarding the schema ,
>> attributing a single node looks fairly valid to me. Tracing the building as
>> an area would therefore be part of the next iteration step since some exact
>> adjustment is required per object, which renders the effort many times
>> If you've got any remarks or further questions, please don't hesitate
>> to state them.
>> Cheers and happy mapping
>> *Klaus / k127*
>> -  http://tasks.hotosm.org/project/1026#task/114
>> -  http://wiki.openstreetmap.org/wiki/Key:building
>> Klaus / k127
>> Am 04.05.2015 um 01:50 schrieb Brad Neuhauser <brad.neuhauser at gmail.com>:
>> On Sunday, May 3, 2015, Dan S <danstowell+osm at gmail.com> wrote:
>>> Hi -
>>> 2015-05-03 22:03 GMT+01:00 Phil Allford <pallford at gmail.com>:
>>> > 1. Should I delete the single node tag for a house when I trace a
>>> > JOSM warns of object within object... I left the original tags.
>>> Yes, delete it - it's important not to lose any extra tags that might
>>> be there, so make sure of that (but in many cases it's just
>>> building=yes or whatever).
>>> Advanced JOSM users like to merge the old node into one of the new
>>> building's nodes, moving the tags from node to way, so that the
>>> object's history is "connected". Don't feel obliged to do that if it's
>>> Probably most new users aren't using Potlatch, but for anyone that is,
>> you can convert from node to area and keep all tags by selecting the node,
>> then shift-clicking where you want another corner to be.
>> HOT mailing list
>> HOT at openstreetmap.org
> HOT mailing list
> HOT at openstreetmap.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the HOT