Racine County Buildings and Address Import
Goals
Add all missing building footprints for Racine County, and add accurate address information to both new and existing buildings.
Schedule
Importing began in November and will continue until the project is complete, tentatively aiming for May 2019 or earlier.
Import Data
Background
Microsoft US Building Footprints
Data source site: https://github.com/Microsoft/USBuildingFootprints
Data license: https://opendatacommons.org/licenses/odbl/
ODbL Compliance verified: yes
Wisconsin V4 Parcel Database
Data source site: https://www.sco.wisc.edu/parcels/data/ or http://geodata.wisc.edu/opengeoportal/
Data license: Per Department of Administration's Peter Herreid (email received on 29-Oct-2018:
The statewide parcel data is in the public domain and there are no restrictions on its use. The statewide parcel database is considered to be a public record under Wisconsin’s Public Records Law (s. 19.31-39, Wis. Stats).
ODbL Compliance verified: yes
OSM Data Files
Data files are all located at https://sites.google.com/site/racinebuildingimport/, though not shown on the home page. If you search for "osm", you'll see them.
Import Type
This is a one-time import which will not involve any automation. Data will be uploaded to OSM using JOSM.
Data Preparation
Data Reduction & Simplification
Parcel data has many, many fields which are unimportant. Table will be filtered first by "IMPROVED" field, which indicates presence of a building. After that, only about 8 fields are related to site address; others will be dropped.
Tagging Plans
Parcel data splits address across several fields. These will be combined into single fields equivalent to "addr:housenumber" and "addr:street". Parcel data includes "ZIP" and "PLACE_NAME" fields, which are already consistent with "addr:postcode" and "addr:city". All footprints will be assigned "building"="yes".
Changeset Tags
Per email from Peter Herreid at Wisconsin DOA, no specific attribution is required. Will be including "source=Wisconsin Land Information Program" on changesets.
Data Transformation
Footprints will be spatially joined to parcel layer using "wholly within" predicate, to avoid misidentifying footprints which straddle parcel lines. Duplicates will be removed from footprints layer, and all unmatched records from the parcel layer will be exported to an address POI layer. The footprints layer, while good, often misses small buildings or buildings under vegetation, and those have to be added manually,. Having the address tags in a nearby POI will make copying faster.
For "straddler" buildings, many are just barely over a parcel line. In these instances, if >75% of the footprint lies within a single parcel, that parcel's data is added to the footprint. These footprints are given a "fixme" tag to flag for additional review, just in case. In cases where the footprint is more evenly split between parcels, these are also flagged for review, but address tags are not copied from POIs.
For extant buildings in OSM lacking detailed address tags, Merge address points is used to bring in data from parcel layer.
Data Transformation Results
*.osm files for project are, on the whole, pretty good. Something must have gone wrong during the conversion, but the JOSM validator has auto-fixes that address duplicate nodes and white spaces in address fields.
Data Merge Workflow
Team Approach
Project will be managed through the US Tasking Manager.
Workflow
- Choose task in TM3.
- Open *.osm file.
- Compare footprints against satellite imagery, correcting and adding as needed.
- Copy address tags from POIs to footprints as needed.
Conflation
During the Data Transformation process, footprints which intersect with extant buildings in OSM will be removed from the import dataset. Address POIs will still be retained to update OSM buildings whenever possible. Additionally, each section of the import will be compared against extant buildings immediately prior to uploading, to ensure that building additions made prior to Data Transformation are not missed.
QA
The quality of the imported data will be reviewed per-feature, owing to the inconsistent nature of the footprint dataset.
See also
The email to the Imports mailing list was sent on 2018-10-20 and can be found WHERE EXACTLY.