| Author |
Message |
![[Post New]](/gforum/templates/default/images/icon_minipost_new.gif) 16/02/2011 05:32:19
|
Silent3
Joined: 15/02/2011 03:51:23
Messages: 8
Offline
|
Why are there a bunch of cities in Louisiana and Tennessee with a timezone of America/Indiana/Knox? In Florida with America/Kentucky/Monticello? In Texas with America/North_Dakota/New_Salem?
I started out trying to determine timezones using nothing but latitude/longitude as a way to compensate for the lack of timezone info in the postal code database. I obtained shapefiles for timezones from http://efele.net/maps/tz/world/ to use for this process.
Then I decided to see what would happen if I looked up timezones for cities in the allCountries.zip file, which already have timezones, to see if the shapefile results agreed. For this test I limited myself to American cities since using the shapefiles is a painfully slow way to determine a time zone. (You wouldn't notice much of a processing delay one city at a time, but when working hundreds of thousands of locations, you might as well go to bed and check your results in the morning.)
Out of 195783 cities (and other geographic locations) 19256 cities produced different results, nearly 10% disagreement between allCountries.zip and the shapefiles.
As far as I can tell so far the differing timezones are at least timezones with the same current UT offset and DST rules. To the extent that Olson timezones relate to the history of timekeeping in a given location, however, the results I'm getting from the shapefiles seem to make more sense.
None of the differences are mere differences in alias names (like America/New_York and US/Eastern being different names for the exact same zone).
Has anyone here worked with the shapefiles from efele.net? If so, what do you think of their accuracy?
By the way, even with the difficulties I'm having, I want to say that I appreciate the work that must go on behind the scenes at geonames.org to process and provide the data that has been made available here.
|
|
|
 |
![[Post New]](/gforum/templates/default/images/icon_minipost_new.gif) 16/02/2011 06:47:42
|
Silent3
Joined: 15/02/2011 03:51:23
Messages: 8
Offline
|
I've just noticed that the vast majority of disagreements on timezones were for places with a feature code/class of "A.ADMD".
There were only 254 disagreements for "P.PPL" items, such as Akiachak, AK (ID#7262936, 60.91227, -161.40071), where GeoNames says America/Nome, and the shapefiles say America/Anchorage.
|
|
|
 |
![[Post New]](/gforum/templates/default/images/icon_minipost_new.gif) 19/02/2011 16:50:45
|
Silent3
Joined: 15/02/2011 03:51:23
Messages: 8
Offline
|
Anyone? Anyone? Bueller?
|
|
|
 |
|
|