Lopa & Lingo building footprints

Lopa & Lingo building footprints is an import of building outlines in the Ituri province of the Democratic Republic of Congo (DRC).

Update: This project was initially started as an import. After comments on the OpenStreetMap import mailing list it was however decided to try to retrieve the original imagery. Ultimately limited access to the imagery was received and the mapping was done based on this image. Much as RapiD, the automatically generated data by Z_GIS was used as help to speed up the mapping. The mapping was done in December 2020, for all edits the changeset comment #LopaLingobuildings was used.

Goals

Add building footprint data in the Lopa & Lingo area in Ituri province in DRC. Having building footprint data on OpenStreetMap (OSM) will provide additional detail for humanitarian organisations such as Médecins Sans Frontières (MSF) and others working in the area.

Schedule

The planning of the import started in March 2020, we would like to have the import completed in July 2020.

Import Data

Background

The two datasets are Esri shapefiles (.shp) created by the Department of Geoinformatcics - Z_GIS, University of Salsburg for MSF.

• The data for Lingo is based on a WorldView-3 satellite image acquired on 01 January 2020 with 50 cm spatial resolution, provided by EUSI (European Space Imaging).
• The data for Lopa is based on a WorldView-2 satellite image acquired on 15 January 2020 with 50 cm spatial resolution, provided by EUSI (European Space Imaging).

For both datasets buildings have been semi-automatically extracted using object-based image analysis and visual image interpretation. Building geometry from automated analysis has been regularised for homogeneous results. Automated extracted buildings were validated by visual image interpretation, whereby also missing buildings were added where needed. In the end, the complete dataset was again validated by an independent interpreter.

Data source site: The datasets are not online available. This import is set up by MSF staff who have access to the internal database.
Data license: Permission to share was given in a private email and states: "The provision of building footprint data (as created by our partner ZGIS) from the MSF Master database is fully in line with the data sharing dynamic that MSF wants to promote. In order to ensure the widest possible dissemination of this data, I am pleased to inform you that after running through the MSF internal geodata sharing policy, MSF agreed to share those non-sensitive data for OpenStreetMap under ODC-ODBL license. This under the condition that MSF and ZGIS will be mentioned on the OpenStreetMap wiki page as contributors to OpenStreetMap."

MSF approval

OSM attribution: Z_GIS & Médecins Sans Frontières
ODbL Compliance verified: Yes

OSM Data Files

Find here the source data files that we want to import. The link is temporary and will be removed once the import is finished.

Import Type

This is a one-time import, but the workflow of the import is designed to be launched every time we have a building footprint dataset created  by Z_GIS. Since the dataset is relatively small, we will use JOSM to upload the data onto OSM.

Data Preparation

Data Reduction & Simplification

The two datasets together contain 29 620 building footprints with a simple geometry - circular or rectangular. After data alignment, we will remove the buildings from the dataset that overlap with buildings that already exist on OSM.

Tagging Plans

Most of the attributes of the original dataset are irrelevant, only the type of building is considered as relevant because it contains information about the material of the roof. To determine these tags we used the tags as described on the following OSM wiki page: https://wiki.openstreetmap.org/wiki/Key:roof:material

In addition all features will be tagged with building=yes.

Find here a list with all original attributes and their corresponding translation into the OSM tagging schema:

Lingo

Z_GIS attributes	Z_GIS attributes meaning	OSM tag	Comments
area_sqm=*	Size of the building in squared meters.	/	Irrelevant
type= very_large_metal large_metal medium_metal small_metal very_small_metal organic plastic facility other_building	Information about the type of the building.	roof:material= metal metal metal metal metal palm_leaves plastic / /	There will be no tags given to the original attributes type=facility and type=other_building, since the material of the roof is unknown for these features.
sensor=WV3	Satellite that took the imagery that was used to retrieve the buildings.	source:geometry=WV3 AND source:roof:material=WV3	Will be added as changeset comment and as a tag to the object like suggested on the OSM DRC email list.
country=COD	ISO3 country code of Democratic Republic of Congo, the country where the buildings are.	/	Irrelevant
location=Lingo	Name of the area where the buildings are.	/	Irrelevant
acq_date=2020-01-01	Date satellite imagery was acquired.	source:geometry:date=2020-01-01 AND source:roof:material:date=2020-01-01	Will be added as changeset comment and as a tag to the object like suggested on the OSM DRC email list.
		building=yes	New tag that will be added

Lopa

Z_GIS attributes	Z_GIS attributes meaning	OSM tag	Comments
area_sqm	Size of the building in squared meters.	/	Irrelevant
type= large_sized_metal medium_sized_metal small_sized_metal very_small_sized_metal organic plastic facility other_building	Information about the type of the building.	roof:material= metal metal metal metal palm_leaves plastic / /	There will be no tags given to the original attributes type=facility and type=other_building, since the material of the roof is unknown for these features.
sensor=WV2	Satellite that took the imagery that was used to retrieve the buildings.	source:geometry=WV2 AND source:roof:material=WV2	Will be added as changeset comment and as a tag to the object like suggested on the OSM DRC email list.
county=COD	ISO3 country code of Democratic Republic of Congo, the country where the buildings are.	/	Irrelevant
location=Lopa	Name of the area where the buildings are.	/	Irrelevant
		source:geometry:date=2020-01-15 AND source:roof:material:date=2020-01-15	Were not attributes in the original dataset, but will be added as tags since they were suggested on the OSM DRC email list. They will be added as well as a changeset comment.
		building=yes	New tag that will be added

Changeset Tags

We plan to use the following changeset tags

• comment=#LopaLingobuildingimport #Ituri #RDCongo #Z_GIS #MSF #MissingMaps
• source=WV3 (for Lingo) OR source=WV2 (for Lopa)
• source:date=2020-01-01 (for the Lingo dataset) OR source:date=2020-01-15 (for the Lopa dataset)

Data Transformation

The datasets are created in a different projection as the OSM database and based on a different image than currently is used for mapping on OSM. This means that there is a slight misalignment between OSM and the datasets that we will have to correct.

An automatic realignment of the data to the existing data and available imagery has not proved possible. We'll thus have to proceed with a manual correction of the misalignment. Splitting up the dataset in chunks of 500 buildings will allow us to corrected manually.

Data Transformation Results

The data transformation will be done in ArcGIS/Qgis and the results can be found here. The link is temporary and will be removed once the import is finished.

Data Merge Workflow

Team Approach

The import will be undertaken by MSF staff and experienced OSM mappers from the Missing Maps volunteer community, using a specific import OSM user account: https://www.openstreetmap.org/user/MSF_imports.

References

This import is referenced in the Import Catalogue and on the DRC import page. The local OSM community in DRC will be contacted and asked for input on the import plans.

Workflow

The upload will be done through JOSM. We will follow the next steps:

1. Data preparation in ArcGIS/Qgis:
   • Unnecessary attributes are deleted and converted to fit the OSM tagging scheme
   • Buildings that are overlapping with buildings on OSM are removed from the dataset.
   • Split the shapefile in chunks of 500 buildings
   • save the shapefiles as .osm files
2. Open the .osm file into JOSM (open data plugin needs to be installed)
3. Data validation phase 1:
   • Align the data with Maxar Premium Imagery (this is the most recent imagery available)
   • Run the JOSM validator & use the Missing Maps Map Paint style to correct all existing errors
4. Download OSM data in the same layer
5. Data validation phase 2:
   • Run the JOSM validator & use the Missing Maps Map Paint style to correct all existing errors
     • First correct for crossing buildings and highways - most issues can be solved by correcting the road.
     • Next correct for overlapping buildings - this should not be possible because they should be filtered out already, but if this is the case please remove the building that you want to import.
   • In addition go over all buildings to verify manually and correct all other errors
6. The validated data is uploaded to OSM with the above described account and changeset tags.

Conflation

As there is not much data in OSM in the area, this does not cause a lot of problems. Without any data transformation yet only 284 buildings overlap with existing OSM buildings. So we will simply remove all buildings from our dataset that are overlapping with buildings already in OSM. The remaining issues will be corrected manually with help of the JOSM/Validator and the Missing Maps JOSM Style.

See also

The email to the OSM DRC mailing list was sent on 2020-07-03 and can be found in the archives of the mailing list: https://lists.openstreetmap.org/pipermail/talk-cd/2020-July/thread.html.

The email to the import mailing list was sent on 2020-07-08 and can be found in the archives of the mailinglist: https://lists.openstreetmap.org/pipermail/imports/2020-July/006298.html