[OSM-talk] Improvements to Massachusetts Data

Christopher Schmidt crschmidt at metacarta.com
Thu Jan 8 01:06:05 GMT 2009


After some urging by TimBL, I put some more work into uploading some new
data in Massachusetts. There are two new imports of import:

 * Buildings Layer, http://www.mass.gov/mgis/lidarbuildingfp2d.htm

This layer contains building outlines for all of Metro Boston. There are
no attributes attached to the data. (However, for state-run agencies
like fire stations and the like, the node is usually in the center of
the way, generating a reasonable look.)

Cambridgeport, my home neighborhood:


MGH, original home of the "MUMPS" programming language that powers


Fenway Park, home of the Boston Red Sox:


The tool used to upload this data was polyshp2osm, a new-style upload
script that has support for namespaced source data attributes, and
(unlike my previous shape uploading tool) is nicely documented, and
checked into OSM SVN at: 


This tool supports polygons-with-holes via the 'multipolygon' relation,
which is useful for buildings like Camridge Friends School:


When uploading, unfortunately, I did not sufficiently check all data,
and of the 300,000+ buildings imported, 9273 were duplicated, exactly,
in the source data. These uploads made it into the upload, but I have
since deleted them, so there should not be many of these left. I
apologize for the temporary inconvenience.  


The real reason I started working on this project with Tim, however, was
to get the green space in Massachusetts updated. This took a bit more
effort, because there is a large quantity of data attached to each
greenspace feature, as described on:


The end result is that Massachusetts is much more 'green' than it used
to be:


When uploading, a choice was made to use leisure=recreation_ground, when
what should probably have been used was landuse=recreation_ground, so
many areas that were unspecified "recreation areas" -- which range from
parks to state forests -- are currently not green on the map. r13029
adds this rendering to the Mapnik stylesheet, with Steve8's support, and
so the next time there is a style update, many of these 'recreation'
areas should start showing up. 

However, many recreation_ground areas in Massachusetts really have
better possible tags -- many of them are, for example, actually parks.
Some of these changes can be automated: I'm currently working on a
conversion tool to read all leisure=recreation_ground tags that I
created and convert them to leisure=park if the item has 'Park' in its
name. Another common name which will get renamed is places with "Golf
Course" In the name.

(I'm also interested if anyone has a suggested tag for country clubs? I
didn't see anything relevant in the wiki or tagwatch.)

Each of these uploads had each of its 50 files manually opened in JOSM,
and compared against existing data. Conflicts were merged by hand;
generally deleting the data from the new upload, rather than from the
existing data, thugh in some cases significantly more accurate places
were uploaed in place.

I apologize if I have missed any merges in this way; this was a
many-many-hours task, so I may have missed some.

The next big task is to get water uploaded: probably using:


I haven't done much planning on this yet, but I will be doing it at:


I want to comment that I have been extremely pleased with the results of
both Osmarender and Mapnik at styling these polygons: I find them to be
quite nice in both cases. So good work to everyone involved in both
cases. (And probably to some Germans for adding enough buildings to make
building rendering important ;))

Best Regards,
Christopher Schmidt

More information about the talk mailing list