0

I have a historical addresses dataset in form:

|Name   |Description  | 

Where "Name" is a single field with house number, street name, city and country. I have digitised two georeferenced historical maps of the location to cover the period and account for street name or geometry changes and for the addition of or removal of streets compared to present day.

This gives me point data for individual addresses and attributed polylines with geometry for the streets in the city:

AddressName: "59 Great Strand Street, Nulltown, GISland"  
DigitisedMapName: "Great Strand Street, Nulltown, GISland"

I now wish to join these up matching based on the street name, giving the street geometry to the address data and plotting the address a certain % along the street based on the house number.

I have tried the ArcMap 10.2.2 address locator but get 0% matched leaving only a manual selection (not feasible given the number of records).

Is it possible to complete this task in PostgreSQL instead to strip the mapped geometry from that table and give it to the AddressName record? Or are the wheels going to fall off in the joining accuracy?

  • Do you have access to FME? Example imgsblog.wordpress.com/2015/07/15/… – Mapperz May 19 '16 at 15:47
  • Hi, yeah I got the free student trial version of FME. I don't have really any experience of using it so the blog post you have provided doesn't really make sense to me, unfortunately. I have just tried the feature merge using the spatial data and the non-spatial address list as readers, set what they should merge based on but I don't know what I am outputting to? Do I define an empty csv to write to? – Paws May 19 '16 at 17:25
  • Add a Feature Writer for csv blog.safe.com/2016/01/fmeevangelist143 – Mapperz May 19 '16 at 20:30
0

Given your 0% match combined with your mention of digitized historic maps (big props to you for putting in the work required to represent change over time!!) it's likely that you need to link some additional attribute data to your roads before any sort of geocoding will work.

every road segment, for example, in the u.s. census bureau's road network is packed with info that aids geocoding. for example the attributes LFROMADD | LTOADD | RFROMADD | RTOADD provide the starting/stopping street numbers for both sides of the segment. a spatial/attribute join to add attributes to your historic maps might be more trouble than it's worth, so the best plan might be...

geocode your addresses using a modern data set chock-full of attribute goodness and visualize the resulting points over your old maps. you'll have to do some manual editing to correct for new/moved/abandoned roads, but the vast majority of addresses remain stationary over time. should work.

  • Hi there, thanks for your advice. I have already trialled an online geocoder and there are, of course, a good number that match. There are loads that don't, unfortunately, because in this case there has been substantial change over time due to slum clearance and political unrest. The fact that the records go back to 1834 doesn't help! In the "description" of the street name I have details of what other streets are connected though ultimately I think you're right - the address locator I have tried to define isn't firing properly. – Paws May 19 '16 at 17:29
  • I have used a combination of a modern geocoder with the gazetteer I created using the historical maps. It really came down to incongruities in the data with spelling mistakes or spaces throwing the address locator off. The historical addresses are from a scanned ledger so need a significant amount of attention! – Paws May 21 '16 at 10:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.