Open mraross opened 3 years ago
The ITN Canford polygon used to stretch out much farther in all directions. The TNRD boundaries were redefined a couple of years ago at the request of EComm and other agencies to be based more on community polygons. The current ITN Canford polygon (which covers an unincorporated area) stops at parcel boundaries just west of the BCGNIS Canford point.
I don't know if the Canford polygon is "correct" or not, nor if the BCGNIS point represents an accurate location for Canford. However I don't understand the title of this ticket since the first two autocomplete returns from Canford are: Canford, BC [DRA locality treated as a DRA locality] Canford in Lower Nicola [BCGNIS locality]
Working with civic number cases, one also gets results consistent with the current definitions of the DRA locality polygons: 2757 Merritt-Spences Bridge Hwy 8, Canford, BC 1600 Merritt-Spences Bridge Hwy 8, Lower Nicola, BC
It would be useful to determine if ITN and/or BCGNIS needs to be tweaked to make them more spatially consistent with each other.
Can the relative locality generation script be changed to ignore any DRA locality that only appears once in BCGNIS? This eliminates the need for accurate placement of BCGNIS names relative to the DRA locality coverage map.
The locality JSON file geocoder input file contains the processed combination of DRA and BCGNIS localities. DRA features that do not have a matching/containing BCGNIS name/point will have a locality type set to unknown. In the file, these will be the entries with LOCALITY_TYPE_ID = -1. This could be useful in identifying particular data review cases such as Canford.
Summary of issue: The Canford locality point in the BCGNIS is in the DRA locality polygon "Lower Nicola" which results in the duplicate values Canford (DRA locality point) and Canford in Lower Nicola (BCGNIS locality point)
There a larger list of BCGNIS locality points that don't spatially overlap with a same-named ITN Locality polygon. I extracted the BCGNIS points and ITN polygons where the names are the same, then did a spatial overlay. Not taking into account the 25m "fudge factor" or simplifying the names, I found a total of 67. A few examples:
I think it is worth pointing out that in some cases these clearly refer to the same actual locality, eg. Canford, Dewdney, and Spatsum, while others actual refer to different localities, eg. Summit Lake and some other examples I am familiar with:
We could increase the fudge factor even more, maybe a few kilometers even? But also we might want to consider getting GeoBC to adjust the locality polygons (assuming the GNIS locations are the more accurate of the two).
Also, as I recall, the 25m fudge factor did resolve a lot of those cases (enough to warrant adding it).
I did an update to my analysis limiting the results to only BCGNIS localities that are within 2000m of a same-named ITN polygon. There are about 40 of these BCGNIS points. Some manual review is probably needed before contacting BCGNIS or GeoBC/ITN. But there is more than just this 1 locality that needs to be looked at. I recommend we close this issue, and work on the bigger list. For now, I've created a Jira ticket wit my findings LOC-5238
There is only one BCGNIS locality named Canford but Canford appears in geocoder as both Canford and Canford in Lower Nicola.