Publishing Legacy Street Centerlines for Public-Safety Dispatch
A city open-data team plans to publish a 1998 street centerline file with 4,800 segments for a public-safety dispatch app. The metadata states positional accuracy of ±12 m, but staff later conflated the file to 2018 orthophotos; 7% of sample addresses geocode to the wrong block. The file has no lineage for the conflation, its horizontal datum is recorded as 'unknown,' and a vendor added 300 segments with no attribution. The dispatch app will use the centerlines to locate 911 calls. The team has 10 working days, one GIS analyst, and a vendor offering a 'certified' version for $15,000. The police department wants the file live before a city event. Staff propose publishing as-is with a disclaimer that the data are 'for reference only.'
1. Assess the fitness for use of this centerline file for public-safety dispatch. What evidence matters most?
Read the worked answer
Fitness-for-use assessment: The file is not fit for dispatch as described. Dispatch is a life-safety use that depends on address-to-location accuracy, current street network, and known lineage. The controlling evidence is not the stated ±12 m alone; it is the combination of an unknown horizontal datum, missing conflation lineage, unattributed vendor segments, and a 7% wrong-block geocoding rate. A 7% error rate on sampled addresses means roughly 1 in 14 calls could be routed to the wrong block if the centerline is the locating source.
An unknown datum prevents reliable transformation and integration with current address points, emergency service zones, and GPS. Without conflation lineage, the team cannot tell which segments were moved, how they were moved, or whether the vendor additions are in the same datum and geometry. The ±12 m figure may be inherited from the 1998 source and may not describe the conflated file. A disclaimer does not cure a life-safety accuracy problem; it only shifts notice.
Conclusion: The file may be acceptable as a general reference layer, but not as the authoritative dispatch locator. The city should either obtain a version with complete metadata and demonstrated address accuracy, or restrict the file to non-dispatch uses while retaining the current dispatch source.
2. Design a 10-day QA and metadata plan with one analyst. What would you sample, measure, and document?
Read the worked answer
10-day QA plan:
- Day 1-2: Recover control information. Locate the original 1998 source metadata, the conflation project file, and the vendor delivery record. If the datum cannot be recovered, treat the file as unreferenced and do not publish it as authoritative.
- Day 2-4: Stratified sample. Select 150 segments: 50 from the original 1998 set, 50 from the conflated set, and 50 vendor-added segments. Include downtown, suburban, and recent-growth areas. Compare each segment to 2018 orthophotos and current address points. Measure offset and record direction, block-face match, and attribute completeness.
- Day 4-6: Address geocoding test. Geocode 200 known address points across the city. Report match rate, wrong-block rate, and positional error. Compare against the current dispatch source. Flag segments with no name, no address range, or duplicate names.
- Day 6-8: Metadata reconstruction. Document source, dates, datum and epoch if known, conflation method, vendor scope, positional accuracy by stratum, attribute completeness, and known gaps. Do not copy the ±12 m statement into the new metadata unless the sample supports it.
- Day 8-10: Decision memo. Recommend publish, publish with restrictions, or reject. Include the sample table and a map of failures. If the vendor certified version has a complete metadata package and passes the address test, compare cost and schedule.
With one analyst, this is tight but feasible if sampling is stratified and automated geocoding is used.
3. What governance decision should the team make about publishing before the city event? Justify the decision.
Read the worked answer
Governance decision: Do not publish this file as the dispatch locator before the event. Publish it, if at all, as a clearly labeled reference layer in a separate open-data channel, with a known datum and a documented accuracy statement, and keep it out of the dispatch application. Notify the police department and the city's GIS data steward in writing of the 7% wrong-block finding and the unknown datum.
The justification is proportionality of risk: a public event increases call volume and makes location errors more consequential. A disclaimer does not provide the operational control that dispatch needs. The team should use the existing dispatch source for the event, then complete QA and either accept the vendor's certified version if its metadata and test results meet the city's standard, or commission a new conflation with lineage. If the vendor version is purchased, require a sample-based accuracy report, datum statement, and attribution for added segments before acceptance. Record the decision, the responsible data owner, the review date, and the conditions for later promotion to dispatch. This is a data-governance decision, not only a technical one: the data owner must accept residual risk explicitly.