Flutterby™! : A Map to the Past 2009-12-06 20:25:20.11869+00

A Map to the Past

2009-12-06 20:25:20.11869+00 by petronius 11 comments

Although my habit is to use Mapquest when I need directions, I have also used Google Maps on Occasion and it seemd to work OK. However, the Chicago Tribune reports some odd anomalies in their mapping system, such as listing the location of Herzl Community College on Douglas Avenue. The only problem is that Herzl is now called Malcom X college and moved to Congress Parkway 50 years ago. There are other problems, like transposing roads to the wrong county and depicting neighborhood nicknames that haven't been used in a century. And this is in just one city. Any strange geography out where you live?

concept14: You were (are) entirely correct. No one makes maps or produce data that's more than just good enough for their needs. Geographic data is very, very expensive to collect and create. My friend, Peter Batty, has given a keynote presentation where he compares the free USGS data to the very expensive Ordnance Survey MasterMap. The USGS data is not meant to be particularly comprehensive and cover all grounds. Utility companies, for instance, map features at much higher detail than the USGS ever will. The OS MasterMap tries to be a dataset useful for these highly detailed purposes.

JT: What you mention is the beauty of OpenStreetMap. Local knowledge can be encoded in the dataset because anyone can edit. A major issue with geographic information is that it is an abstraction, typically applied by people who lack true local knowledge. Google's new map allows for easier change requests by users. GOOGs hope is that people will note errors, ultimately making their data better. What they've hit on is that it doesn't take very many people flagging errors to make a better map for everyone. I'd personally like to see the USGS adopt a similar strategy.

In Gulfport, MS there is a road that changes names a few times. The locals know where these changes are, however on the map, it's just listed as Highway 49.

From the beginning, Hwy 90, and through downtown, it is 25th Avenue. After reaching about 17th St it turns into R.B. Meadows Blvd. It stays that way until it reaches Pass Road. Pass Road is also known as 25th Street, which makes the intersection of Pass Road and Hwy 49 known as "25th and 25th" due to 25th Street crossing 25th Avenue. It begins north again and becomes known as another street which I can't remember the name of between Pass Road and 28th Street, where it again becomes R.B. Meadows Blvd until 34th Street. It then becomes Highway 49 again.

If you're looking at mapquest for a location, it's not that hard to find anything, but if you talk to a local for directions to something... it could get a little confusing.

Thanks for the update, Ebradway. I should clarify that when I referred to the USGS producing data that was just good enough for their needs, I was referring to the digital mapping from the 1980s that was one of the original foundations of TIGER.

Let me elaborate a little on our process at USGS. This is for The National Map - a US spatial data infrastructure made up of eight data layers - maps aren't what they used to be... Unlike the old topo sheets, the purpose of The National Map extends beyond science and the environment. It's designed to be a uniform set of spatial data for national security and disaster response. We still have the most complete hydrography network dataset of pretty much any country - but we are also working to provide a similar dataset for transportation, structures and other layers. In addition to the online National Map Viewer (and standard RESTful APIs behind the viewer for direct access to the data), the US Topo provides stable map products that can be printed or used with a PDF reader.

Our transportation layer is based partly on Census TIGER data which we clean up and partly on data provided to us by our partners at State and Local GIS departments. The data is normalized to a consistent scale within a specified error bounds. You would think that Census would take the data back from us since we cleaned it up, but that's not the case. If you grab data directly from Census, you don't get our changes. I'm not entirely sure how Census originally created their data, but I bet it's a continued refinement of a base TIGER dataset, which means it predates GPS.

Our data is available under something like a public domain license. It's not exactly public domain but I think the main difference is that terrorists or enemies of the state are not licensed to use it. This license is only slightly less promiscuous than public domain (which terrorists can use) and infinitely more promiscuous than any GPL, Creative Commons or other license. The USGS used to charge a fee to put the data on some kind of media - but now it's all available online for free. I do think the gateway sites like Geospatial One Stop put limits on how much you can grab in a day but the vector data is relatively small in file size. You'd only really be limited if you were grabbing aerial imagery or LandSat data.

Because our data is so freely available, everyone grabs it to use as a starting point. I'm pretty sure if you dug deep enough into TeleAtlas and NavTeq, you'd find USGS data. Right now, Google Maps in the USA is based on USGS data combined with what they've collected with their StreetView vehicles and user corrections.

OpenStreetMap also started with USGS and TIGER data in the US. Originally they started with TIGER because the OSM people in the US didn't realize the USGS data was a little better. OSM is terribly sensitive to copyright issues and even tell their users not to use Google StreetView to gather names off of street signs. There's definitely no TeleAtlas or NavTeq data in OSM. TeleAtlas and NavTeq are at the opposite ends of the licensing spectrum from the USGS. They delineate data use right down to the size of the screen on which you view it - car navigation, cell phone, etc. Google split from TeleAtlas to allow them to make their cell phone-based navigation application for Android. The TomTom app for iPhone costs $95 when most other apps are under $10. The differential in price is the licensing and data costs.

I worked with the Tiger data when it first came out in the 1980s, and it was obvious that just because all those decimal digits were there, they weren't necessarily reliable. It was clear that the metro and non-metro data had different provenances.

The underlying issue is that all these geographic data compilers are serving different needs and produce data that is good enough for their purposes. The USGS only produced 1:24,000 maps where the countours and watercourses were more important than the roads and buildings. If the Census Bureau puts your house in the right block that is good enough for them. The commercial firms annotated the base data with one way streets and classifying the major highways from the cowpaths and so on. I think Grass Valley has one or two no-left-turn intersections in town, but of course nobody needed to send a crew out into the mountains to look for any out there and thus the original data never gets reviewed.

Nowadays targeting urban consumers by neighborhood seems to be important in the marketing industry (see e.g. http://www.maponics.com/Neighb...es/neighborhood_boundaries.html; disclosure: my employer is a customer of theirs), and the obsolete Chicago neighborhood names may have come from some project at Google to attempt to fill this need.

Eric, on the alignment issue, this was an old farm road, it may have even been possible that it used to go up over the knoll that the vector data had it on at one point. I was just amazed that the data that road data was derived from had to be many decades old, for some reason I'd thought that they'd started with TIGER data sets, and I figured that was done by direct GPS acquisition.

If they're pulling information off of old USGS quads that explains why there's century old stuff floating around modern digital databases, but that's an impressive undertaking in itself.

The Open Street Map project starts with the same TeleAtlas data (I believe), but allows you to submit fixes/changes/edits yourself - kind of like a map-wiki.

The combining of various databases (databasi?) would explain some of the Chicago issues. We have many named neighborhoods with very specific boundaries that were autonomous towns up until the era immediately before WW1. The little suburbs joined the city in order to link up to the elaborate water system we built. Some old, less formal names could have stayed on some old lists. We also completely reworked the street numbering system and modified many street names in 1914, a system still in use.

Here's the thing, though. If they had simply asked somebody in Chicago we could have told them that Dundee Road is in the next county, McCormickville no longer exists, and Herzl College got caught up in the social change of the mid-century. Who does fact check at Google?

As Markd states, GOOG maps are no longer based on licensed TeleAtlas data. It's an amalgamation of various datasets Google could drum up. A large chunk comes from the USGS (disclaimer, that's where I work, but we don't make money so there's no conflicts). A surprising amount of the USGS data is sourced from color-separated scans of the old 1:24,000 topo quads which fell out of production in 1992. Many of the topo quads hadn't been updated since the 60s, especially in remote areas. To further complicate the problem, those remote areas were not surveyed by local residents, so the names used on the maps were not necessarily the name in use.

But you have to consider the effort of mapping the entire USA at 1:24,000. There were over 55,000 separate quad sheets made - all by hand - every contour, every stream, etc.)

As for roads not aligning with imagery, it's probably due to the fact that the vector data (roads) is at a different scale from the raster data (imagery). The trend is to show the highest resolution imagery possible (like the .3m imagery from the USGS 133 Most Populated Places imagery) but then simply overlay the vector data. 1:24,000 scale vector data typically has uncertainty on the order of 30m. So the roads could be as much as 100 pixels off from the imagery while still actually being accurate.

Google's using different map data (which they own). Should make it a lot easier to get changes made (took close to 5 years to get my street address in western PA to point to the right place on the map). There's a 'report a problem' link - if you find something amiss, please report a problem, even if it's just a one-liner.

(ObDisclaimer, I work for the GOOG, but not in the geo department. I was giddy when my street address finally got fixed. Even being on the inside, I couldn't get it corrected.)

I know I've mentioned that the last time we were up near Grass Valley we were blindly following the TomTom and ended up on a road that petered out into nothingness. So we pulled up Google Maps and Google Earth on the iPhone, which, indeed, thought that that road existed as well, but showed it probably 150' to one side of where the satellite imagery (and the GPS) thought we were.

Haven't looked much deeper, but we on the west coast also have a relatively recent history. Old is 150 years, and much of our populating is from the booms post WWII, so if it's on the map it probably hasn't changed names much.