-------- Original Message --------
Subject: Re: Phoenix street network data
Date: Mon, 16 Feb 2009 19:17:17 -0600
From: Chieko Maene <[log in to unmask]>
Reply-To: [log in to unmask]
To: [log in to unmask]
References: <[log in to unmask]>
Hi everyone,
I finally had a chance to create geocodable street data using TIGER2008
- I had to. I know my patrons would ask me about it. Yes, TIGER2008 has
been improved greatly, especially geometry-wise (more lines, spatially
accurate) - how wonderful.
Though, as Edward wrote, I realized it isn't easy to use TIGER2008 for
geocoding. I thought it's just a matter of joining associated tables
(EDGES/FEATNAMES/FACES) and merge fields in ArcGIS, but I was wrong. The
main problem for me was the one-to-many relationship between EDGES and
FEATNAMES. (In fact, EDGES's Prime Key TLID isn't unique - i.e. there
are quite a few duplicate TLIDs in EDGES, looks like it's because the
edges are between two counties - so, it's actually many-to-many
relationship.) Since ArcGIS cannot handle one-to-many outer join (i.e.
ArcGIS will grab only the first matching row information, not all
matching rows), you would need another software that you are comfortable
to use, statistical software or DBMS such as access or foxpro (my
choice), so that you won't lose useful information/rows from FEATNAMES
table.
It worked for me in the end but I spent a whole weekend to do this. I
think it's better if we can geocode addressed using relational
geodatabase in ArcGIS, not using single-table shapefiles (i.e. streets)
so that we can keep table relationships without combining to create a
large and un-normalized single table (i.e. shapefiles.) I don't know if
I can do that without ArcSDE though, and I would love to know how to do
it. Or, I can just wait until ESRI releases a new version of StreetMap
N.America dataset based on TIGER 2008..
Or, as Edward pointed out, we can use third-party geocoding tools. I
think the USC website is wonderful. I don't know if the USC geocoder is
based on TIGER 2008, though.
I was a fan of batchgeocoder but later, I found some inaccurate results.
I also like to be able to select candidates for ambiguous addresses..
kind of..
One more thing: at the end, I wanted to know how good my result was - I
tested my final TIGER2008 Chicago area roads data to geocode Chicago
area addresses, and then compare the result with that of the ESRI 2008
StreetMap N. America streets dataset (from ESRI Data & Maps 9.3,
TeleAtlas base. *Not* a premium version.) To be fair, I clipped ESRI
2008 StreetMap streets using the same 7 counties boundaries. Then I
geocoded the addresses under same conditions (i.e. same method: US
Streets with City, State and ZIP, same thresholds, no alias tables,
etc.) The results: Unmatched cases(%) without interactive matching,
TIGER2008: 5,574 (23%) vs. ESRI2008: 5,650 (23%), total N was 24,158.
Happily, TIGER2008 did slightly better than ESRI 2008 StreetMap, though
the difference was minuscule.. (The reason for the high unmatched cases
is because my addresses contained out-of-the-extent addresses, such as
outside state addresses.)
Just in case, here is what I did to create the Chicago metropolitan area
(7 counties) geocodable streets. Please feel free to correct me if I am
doing wrong.
Also, if anybody wants the final data, you can download the Chicago
metropolitan area TIGER2008 street file from here (no metadata..)
http://www.library.northwestern.edu/map/TIGER2008_NE7CO_ROADS.zip (64MB)
[Note: I didn't bother to add a extra address-ranges table (ADDR) since
dealing with just three tables was enough for me. Well, I guess FACES is
optional too, but it's good to include additional left/right side
polygon attributes for those who like to aggregate data based on
boundaries, i.e. tracts, blocks.]
(1) In ArcGIS : append 6 counties EDGES shapefiles to the main county
(cook county.) Select road edges only ([roadflg]='Y') and also get rid
of miscellaneous road types we don't want to use, i.e. alleys and misc.
trails ((MID([MMTFCC],1,3)='S17'OR (MID([MMTFCC],1,3)='S18') and then
finally add a new identifier field, FID2 (copy value from FID), to keep
a unique ID key for each polyline.
(2) In DBMS software: append all 7 counties' FEATNAMES tables. Do the
same for the 7 counties' FACES tables. Create a query, [EDGES left outer
join FEATNAMES on EDGES.TLID=FEATNAMES.TLID]. From FEATNAMES, you will
get parsed road information (dir, pretype, name, surfix, type) with
variations of street names. (i.e. "E. Main St" in Cook County is also
known as "W County Line Rd", "Main St", "Lake-Cook Rd", etc. We want all
these variations/rows for better geocoding results.) After that, I got
left and right sides face/polygon information by joining EDGE-FEATNAMES
table and FACES, twice (left side, and then right side) - this part is
straightforward since it's all one-to-one relationship.
(3) Optional - In DBMS software: You can further continue to query
(join) to get textual place and state abbreviation names, i.e.
"Chicago", and "IL". Get the information from other TIGER tables (i.e.
COUNTY or PLACE.) Also, you may want to rename fields so that ArcGIS can
recognize geocodable fields (i.e. instead of LFROMADD, change it to
L_F_ADD).
(4) In DBMS software: find duplicate FID2 (unique identifier for EDGE
shapefile polylines), and separate the duplicate sets by the number of
duplicate (i.e. group1: duplicate 1, group2: duplicate 2, group3:
duplicate 3, etc.. I found up to 9 duplicate name groups as a result of
EDGES-FEATNAMES one-to-many joining.. I am not sure if this part makes
sense to anybody..)
(5) In ArcGIS: join EDGES polylines from (1) and the first group of the
EDGE-FEATNAMES-FACES joined table from (4) using a common field,
EDGES.FID and EDGES-FEATNAMES-FACES.FID2. Export as EDGES_1. Remove the
previous join in EDGES polylines and join EDGES and the second group of
EDGE-FEATNAMES-FACE (duplicate group 2) and then export only the joined
rows (i.e. only rows with matching join fields) as EDGES_2. Repeat this
for all duplicate group tables. Lastly, merge all EDGES groups (EDGES_1,
EDGES_2, EDGES_3, etc.) by appending others to one.
(6) Optional: I added a spatial index in ArcCatalog for faster rendering
in ArcGIS. TIGER files don't come with these index files, .sbn & .sbx. I
didn't add any attribute indexes since "create a geocode locator" will
create indexes for geocoding.
Sincerely,
Chieko
--
Chieko Maene, MS, MLIS
Maps & State Documents Librarian
Government and Geographic Information and Data Services
University Library
Northwestern University
1970 Campus Drive
Evanston, IL 60208-2300
Phone: (847) 467-3679
Fax: (847) 491-7603
[log in to unmask]
http://www.library.northwestern.edu/map/
http://geospatial.edublogs.org/
Maps-L Moderator wrote:
> -------- Original Message --------
> Subject: RE: Phoenix street network data
> Date: Sat, 14 Feb 2009 12:28:52 -0800
> From: Edward Sullivan <[log in to unmask]>
> To: <[log in to unmask]>, <[log in to unmask]>
> References: A<[log in to unmask]>
>
>
>
> This is a bit "outside the box", but rather than going to the trouble of
> locating, standardizing, and performing quality checks on 'free' road
> network and address layers, I've lately taken to using the Batch Geocode
> site to do many of my smaller (< 5,000 addresses) geocoding tasks -
>
> http://www.batchgeocode.com/
>
> I believe the Yahoo! Geocoding API this 'mashup' page utilizes is using
> the very good GDT/TeleAtlas road network base maps, so the quality of
> the results are typically as good or better than what I can accomplish
> with free data and desktop GIS geocoders.
>
> There are more advanced batch geocoding services available to academic
> users without charge from the USC GIS Research Laboratory:
>
> https://webgis.usc.edu/
>
>
>
> Edward A. Sullivan, III
> Senior Technical Associate
> Economic & Planning Systems, Inc.
> 2501 9th Street, Suite 200, Berkeley, CA, 94710-2515
> Voice: 510-841-9190 FAX: 510-841-9208
> Email: [log in to unmask]
> Web site: www.epsys.com
>
> Due to the potential that information exchanged by electronic media can
> deteriorate, be damaged, lost or modified, intentionally or otherwise,
> use of this electronic data by anyone other than Economic & Planning
> Systems, Inc. shall be at the sole risk of such user and without
> liability or legal exposure to Economic & Planning Systems, Inc.
>
> The recipient is responsible for verifying the accuracy of data against
> governing hard copy documentation. If there is a discrepancy between
> the hard copy and the electronic copy, the hard copy will govern.
>
> Recipient assumes all risks in the changing or modification of data and
> revisions or updating of hard copy documents.
>
> -----Original Message-----
> From: Maps, Air Photo & Geospatial Systems Forum
> [mailto:[log in to unmask]] On Behalf Of Maps-L Moderator
> Sent: Friday, February 13, 2009 11:43 AM
> To: [log in to unmask]
> Subject: Phoenix street network data
>
> -------- Original Message --------
> Date: Fri, 13 Feb 2009 11:38:48 -0800
> From: andrew nicholson <[log in to unmask]>
> To: <[log in to unmask]>, <[log in to unmask]>, tanya
> <[log in to unmask]>
>
>
>
>
> Hi everyone,
>
> My apologies for the cross posting.
>
> We are looking for a good quality road network for Phoenix Arizona to
> allow students to geocode too. We have downloaded the Tigerline files,
> but they are not very good for this purpose. Can anyone provide us with
> suggestions for a source or two for high quality and free Phoenix street
> network data?
>
> thanks for your help,
> Andrew
>
>
> --
>
> Andrew J.P. Nicholson
>
> GIS/Data Librarian
>
> Hazel McCallion Academic Learning Centre
>
> Room 360A
>
> University of Toronto Mississauga
>
> 3359 Mississauga Road North
>
> Mississauga, Ontario
>
> CANADA
>
> L5L 1C6
>
> Phone:(905)828-3886
>
> Fax:(905)569-4320
>
> Email: [log in to unmask]
> <mailto:[log in to unmask]>
>
> Web: http://www.utm.utoronto.ca/library/
>
>
>
>
> ------------------------------------------------------------------------
> Windows Live(tm): Keep your life in sync. See how it works.
> <http://windowslive.com/howitworks?ocid=TXT_TAGLM_WL_t1_allup_howitworks
> _022009>
|