statewide address suggester

[gregbunce]

What's the idea?

• To create a statewide address suggester. Essentially, this tool/service would allow a user to click on a map and the service would return an approximate street address for that location. Currently, we offer this service in Atlas, however, It's limited to the areas where we have road centerline data. This tool would extend the geographic extent of the service, making it statewide.

Who would this benefit?

• Primarily addressing authorities (county engineers) throughout the state and possibly 911 folks. Many in those groups are not using GIS. This service, then, would extend the capabilities of GIS, to many non-GIS professionals. Currently, some counties in the state are using paper and pencil to perform these functions.  This tool would greatly enhance their existing workflows.  

Is funding needed?

• If this idea gains enough traction, I feel as though this work is eligible for a grant or some level of one-time, state funding.  

Estimated level of work?

• 60 to 80 hours.

What are some of the known steps to make this happen?

1. Build a pseudo addressing grid for each of the 231 addressing systems in the state. This would look similar to a grid of road centerlines but evenly spaced and with buffered address ranges on each segment (ie: 100 to 199) corresponding with the estimated distance from the center of the address system. 
2. Create a web service (or extent the existing one) to perform the reverse geocoding on this new pseudo grid. 
3. Create a web map interface, or extend the functionality of Atlas, where users can utilize this new functionality.  

Some technical thoughts:

• Under the hood, this tool would perform a reverse geocode on (1.)the existing road centerlines and then if an existing road segment was not found (which would be the case in most rural areas in the state) then, (2.) it would rely on the estimations in the new pseudo addressing grids

What are some of the known limitations or challenges?

• There are 231 address grids in the state, but apparently, there is not a standard distance for a city block across the state.  That said, we'd need to do some upfront work to determine what a city block is for each of the address systems.  Erik suggested we can use the existing road data to do some analysis, which could roughly determine this distance. 

Resources

• Existing VBA code that builds addressing grids
• A few existing address grids that David Buell created can be found scattered about in this directory (aka: search for 'address grids').

[rkelson]

Are the existing grids, that were used for the original addressing that David has in drive, going to be utilized somehow?

[gregbunce]

I suppose we could take a look at those and see if they have value. I would guess since David created them, then they're pretty darn accurate and have been adjusted to reflect what's on the ground. I saw a handful of counties, but not a ton. definitely worth looking at.

that's my two cents. curious to hear from others.

[gregbunce]

I added a link to David's SpillmanBackups folder to the idea resources above, where the grids are scattered about.

[rkelson]

Yeah I don't believe David created them as I recall making them originally for the counties I was working on. I think he got them from the original county stewards. I only recall doing the addressing in Kane and Daggett.

[gregbunce]

Yeah, definitely share any info, back stories, or data that you have on this. It will help shape the discussion and/or next steps.

[gregbunce]

Status/Progress Update:
The GIS Coordinator, (Ken Richards) from Iron County asked if we could convert the VBA code to python. Having this code in a workable language has value to them. Erik and I told him we would have that done by the end of June.

/remind me on 06/01/2021

[eneemann]

I created an initial python script to build out an address grid, which can be found in my Python-Tools repo at build_address_gridlines.py. Everything is in NAD 1983 UTM 12N (EPSG: 26912), so the units need to be in meters. The script generates an approximate address grid from the following user inputs:

• origin x coordinate
• origin y coordinate
• number of blocks to build in each direction
• average east-west block distance
• average north-south block distance
• grid name
• output directory

The output looks like a lattice of gridlines, each containing a label with its approximate street name (e.g., 700 E).

It does a decent job based on some initial tests in Salt Lake City and Provo (see images below). We knew this would be tricky due to inherent inconsistencies and real-world variations, but when accounting for possible rectangular grids with different x/y block distances, I think this is a good first step.

Weaknesses

• If block distance varies in a city, this grid won't line up with the variations very well
• The output seems slightly rotated in the counter-clockwise direction, which becomes noticeable far from the grid's origin. This might be an artifact of the UTM 12N CRS.

Salt Lake City

• not bad for about 6 miles out from the grid's origin

Provo

• east-west oriented gridlines are doing pretty well 48 blocks north of the grid's origin
• north-south oriented gridlines appear shifted a bit counter-clockwise (1 block west) this far from the origin

[mheagin]

You want to share it with Cydne to test or run a test on Elsinore in Sevier Co first?

[eneemann]

@mheagin Is Elsinore one of the areas they need? I'll probably share with some folks next week, they'll need to be able to edit and run the script with Python 3 from an ArcGIS Pro install. Otherwise, I could build a grid for someone and send them the final product. Iron county was recently asking about it, so I'll reach out to them.

[mheagin]

@eneemann That is the big annexation that is going to go through next week, so she will be bombarded with address requests. Might be best to create a grid and just send it to her, not sure of here coding skills. She trained under Cynthia

[eneemann]

Okay, I'll get one going for her.

[gregbunce]

I've created a google drive folder/directory in the agrc_public_share where we can compile all the known, existing address grids from across the state. I've added Sanpete's grid that I obtained from their GIS dept as well as any other grids I've found in random places in the agrc/ugrc google drive.

please add any address grids to this folder that you are aware of. Thanks!