[GIS] have gaps in OSM data

openstreetmap

I just starting to work with raw-OSM data, so maybe my question is a very basic one.

When I download OSM data from Geofabrik I never get the same coverage of the polygons as in the rendered raster OSM data – at least for many areas.

In the screenshot below you see the rendered OSM data as background and the polygons of the downloaded raw-OSM data above. For some areas I get the polygons (the light-red colored area, transparent to see the rendered OSM data below), but you see e.g. the forest in the middle of the sceenshot and the farmland in the lower part which are missing in the raw-OSM data. Even the boundary between forest and agricultural area is not available (e.g. as Polyline in the .osm.pbf file).

Screenshot: Rendered OSM data (background) and raw-OSM data (only Polygons)

Being used to work with shapefiles, I would assume that all areal features like land use are provided as polygons. But this is not the case obviously. The alternative -which I do not beliefe to be true – would be that the raw data have gaps in it.

Can someone explain what I am doing wrong?

Thanks

some technical details: I tried different data sets (meaning: different areas) and data formats (shapefiles, osm.pbf) but always have the same issue. I also tried to download OSM data via the OpenStreetMap plugin in QGIS.

Best Answer

If you want to have the same look as the background tiles, you have to use the same toolchain.

That requires to have a Postgis database filled with osm2pgsql.

The Openstreetmap plugin of QGIS has some limitations in the correct creation of multipolygons (although much better than the old plugin had some yers ago), and the ogr2ogr data provider used by "Add vector layer" has too.

The shapefiles from geofabrik are only a subset of the OSM database, with attention to roads. It might be that some kind of strange forest multipolygons are dropped there.

See this example for relation 3541951 missing in geofabrik shapefile detmold-regbez-latest.shp.zip:

enter image description here

Tagged as landuse=forest, but not in the landuse or natural layer.

Introduction

This will likely require significant amount of manually work to detect and remove the duplicated data. When you're detecting and resolving the duplicate data; you'll want both sources to be in the same geo format: shapefile, PostGIS DBs, or as OSM data.

Workflow

The following workflow is based on having both sources of data as OSM before merging and resolving duplicate data.

There's a couple options to convert the data into OSM:

Convert the shapefile data into OSM however you'd like. Versions of ogr2ogr released in 2013 or later (version 1.10 or later, IIRC) can also convert SHP to OSM. There's also ogr2osm as you had noted: there's a couple different versions of ogr2osm, no matter which one you use - I prefer pnorman's, it's the most up-to-date. No matter what, make sure the translation files are compatible with the version of ogr2osm that you're using (for the sake of simplicity, the ones that I've linked to should be compatible with the version of ogr2osm). See here as examples for translation files that are compatible with pnorman's ogr2osm.

Ensure the translation file is complete with of all the information that you want in your shapefile. The translation file will convert your Types and attributes of the shapefile into what OSM calls Tags, which consist of Keys and Values.

1a. run ogr2osm.

Open josm, download the conflation plugin
Your gov data is now a osm file. Open josm, File > open Your data is there as a layer.
If you already have the OSM data locally stored on your computer, open it in josm, it will also open as a new layer.
Merging these two sources of data together and resolving the duplicate data is known as conflation. Run the conflation plugin and resolve all of the conflicts.

If JOSM runs out of memory (e.g., when using large files), separate the types of attributes and complete this workflow multiple times, each with a different kind of data (e.g. boundaries and land uses; highways; buildings), and then finally merge the osm files together using osmium or another tool.

B. JOSM can also read shapefiles although SHP support isn't perfect and this method assumes the shapefile can be loaded entirely into memory...

Start JOSM.
Open the shapefile (e.g., filename.shp).
Select all.
In JOSM, Edit the attributes and properties that were imported from the SHP, and change them so each attribute corresponding an OSM tag.
Save as OSM format.
Continue from A4 and conflate

Import as OSM

Import the OpenStreetMap data into the system as follows:

Change to the directory containing OpenStreetMap (OSM) files converted using JOSM.
Execute the following commands in the database:

    CREATE EXTENSION hstore;
    osm2pgsql -j -W \
              -d osm filename.osm

The -j option is key as it instructions osm2pgsql to import the tags into an hstore column, this preserving the underlying data structure and will import all tags into the database.

Create Mapnik Layer

To have the data appear on the map, add a layer and a style for that layer. This can be as simple as the following:

Edit mapnik-stylesheets/osm.xml.
Insert the following XML code before the closing </Map> tag...

...

<Layer name="prefix_zone" status="on" srs="&osm2pgsql_projection;">
  <StyleName>zones</StyleName>
  <Datasource>
    <Parameter name="table">
    (select way from prefix_line order by tags desc, z_order) as zones
    </Parameter>
    &datasource-settings;
  </Datasource>
</Layer>

Create Mapnik Style

Continuing from the previous section:

Find the last </Style> tag (around line 3350).
Insert the following XML code before the &layer-shapefiles; directive:

...

<Style name="zones">
  <Rule>
    &maxscale_zoom1;
    &minscale_zoom19;
    <LineSymbolizer stroke="#0065BD" stroke-width="2.5" />
  </Rule>
</Style>

Roadmatcher

roadmatcher is another tool that might be helpful

Best Answer

Related Solutions

[GIS] Large Open Street Map OSM file to Shapefile. Is there any hope for 32-bit FME

[GIS] Conflate (merge) private shapefile data with OSM data