WikiProject Bhutan/Thimphu Thromde data import
The goal is to import around 10,000 buildings from one side, and more than 7,000 housenumber nodes and hundreds of still unmapped streets on the other, all from Thimphu Thromde, in Thimphu area, Bhutan.
- Preparation, discussion: To be started on Saturday 2nd July 2016.
- Import: Expected to start as soon as community has solved any issues.
The original data consists of three different sets:
- Street geometries.
- Building address numbers.
- Building footprints.
Many highways are already mapped in the area.
The street geometries (865 in total) can help adding those highways still unmapped.
Many of them come with the name. Those names can be added to those highways already mapped that are still unnamed, and will be added of course to those streets not yet in the OSM database.
Moreover, they come with a type field (9 different values) that can be useful for the users to tell what kind of highway they are. They will, anyway, check those values against aerial imagery before adding info.
Some buildings need manual realignment against Bing, that appears to be very well aligned according to available GPS traces. Mapbox will be used too, as it appears to be more up-to-date than Bing.
The original building dataset consists of 13,626 ways and 30 multipolygon relations of building footprints. There are around 1,500 building inside building and 2,500 crossing ways. More about that on data reduction and simplification further down.
The building dataset contains several field attributes, none of them useful for import.
ODbL Compliance verified: YES
This import consists of two independent imports: a) a building footprints import, that will be done manually by the community using the Tasking Manager and following a strict workflow, and b) a highways and street numbers import, that will be done by a small group of experienced mappers, also manually but not through the Tasking Manager.
Data Reduction & Simplification
The data is originally in shapefile format. You can download the original data from here.
For the buildings dataset, we have:
- Made an alignment with Bing. Bing imagery appears to be well aligned according to available public GPS traces. In any case, users will have to realign some footprints as a first step when importing them.
- Fixed 7 duplicated relations.
- Deleted all multipolygon building relations except one.
- Fixed 3 duplicated nodes.
- Fixed 8 self intersecting ways and 1 unclosed building.
- 3 buildings converted to relations (buildings with empty spaces inside).
After all that, we got an .osm file composed of 13,610 buildings as simple closed ways and 4 buildings composed of an outer and inner polygons. Of those buildings, we have:
- Building inside building: 1,531
- Crossing buildings: 2,579
Al of them means there are 2 different polygon for the same building. For each of those conflicts, users will chose one of each 2 polygons (the one that appears to be the most correct) and ignore (delete) the other.
As for the street geometries, they have been simplified with the JOSM "simplify ways" tool (SHIFT+Y), going from 27,217 nodes down to 6,222.
For the streets and street numbers datasets we will keep them in two different datasets. The street numbers dataset will be further divided in subsets of 400-500 nodes that will be assessed and imported separately, along with the streets from the streets file that qualify for import (not yet present in OSM and geometry of good quality).
You can download the three modified data sets in .osm format here.
We will use the following changeset tags:
- comment=Bhutan data Import
To the comment=* tag, you should add particular comments to the changeset.
Taken the data reduced and simplified, we added the tags described in tagging.
Data Merge Workflow
As described, this import will be done in two independent parts:
- Buildings: Manually by the OSM community, through one Tasking Manager project.
- Streets and house numbers: Manually, by a small group of experienced mappers.
You have a workflow wiki for the import of the building footprints here.
You have a workflow wiki for the import of the house numbers and named street geometries here.
In case of any trouble, JOSM reverter will be used.
As a general rule, all info that is already present will be respected. We will only replace one street when the geometry of the OSM one is clearly poorer than the import dataset counterpart. The same will apply for building footprints.
We won't replace any name that is already in OSM, unless we have on the ground info to do that.