County Accuracy

[holograph]

A week or so ago, there was a brief discussion about how accurate the county names on the datasheets are. I was curious about how many datasheets might be in error. I've created a spreadsheet that contains more than 3,000 PIDs that appear to be filed with the wrong county. The spreadsheet is limited to those PIDs that are 1 km or more from the border of the county named in their datasheets. I also restricted it to the lower 48 states, since Alaska is known to be problematic.

 

The record discrepancy is DM0128 which appears to be 756 km from the county it was filed under!

 

If any of you are curious and want to check our some of the PIDs in your area, you can download the spreadsheet (120 kb zip file) here.

 

Caveat: I recommend that before you use this spreadsheet to submit corrections to Cheryl, you should independently compare the locations of the PIDs to the counties they were filed under. How you do that is up to you.

 

Some of the discepancies may be due to county realignments, but most are probably due to errors when they were submitted. I used the county boundaries from the 2004 Tiger/Line Census bureau file (second edition). That dataset is likely more up to date than many or most other online maps, especially the ones based on USGS quads.

 

And yes, this is how I entertained myself when we were getting drenched by a week of rain here in the Northeast.

[+rogbarn]

OK, I'm impressed! Let me know what software and/or process you used to do that?! I'm guessing it is a pretty definative list, right?

 

I will reiterate that there are many reasons why this situation can occur and each one needs to be researched before asking for an update. As I mentioned in my other post, I map the coordinates and then try to determine if the description matches the mapped coordinates. If they do, then it is probably a simple county error. While a simple county error is the most common, other problems I have run into include having a description that is mixed up with another datasheet (often with the same or similar designation or, less frequently, with a nearby PID) and a typo in the coordinates (off by a degree or some other obvious amount). The confused descriptions can be difficult to figure out but I have figured out some of them before sending off my findings to the NGS. Other times, I just mentioned that it isn't right and let them try to figure it out. Most of those are still outstanding issues, I've never seen an update come thru. I'm guessing that field research would be needed to clear those up.

[holograph]

The process:

 

1. Extract the political boundaries of each county from the 2004se Tiger/Line files. (The political boundaries incorporate the 3-mile limit offshore, the boundary lines within lakes, rivers, and bays, and other political boundary positions.) Software: custom Java.

 

2. Extract the lat,lon of each of the 750,000+ pids. Encode the datasheet's state and county as a FIPS code. Software: custom Perl.

 

3. Determine which counties contained which PIDs based on the county boundaries and the locations of the PIDs. Software: GIS.

 

4. Compare the PID's datasheet state and county to the county that contained the PID's position. Create a set of PIDs that appear to be located in counties other than the one listed in the datasheet. Software: GIS.

 

5. For each PID in that set, compute its distance to the boundary of the county that is listed in its datasheet. If the distance exceeds about 1 km, then list the PID. Software: GIS.

 

The 1 km criteria was abitrary, it was just a way to filter out the numerous cases of PIDs that are slightly outside the border of the county they are filed under. The filtering eliminates bogus discrepancies that arise from scaled coordinates and county boundary imprecision.

 

I guess the list is definitive as far as it goes, unless I made some programming blunders somewhere, which is why everyone should do their own checking. Code is never perfect, it just gets less imperfect with more testing and refinement.

[Bill93]

I've investigated AA3903, AA3904, and AA3905 in Iowa, submitted an email, and seeing no update then also logged the situation by web entry.

 

The listed county was right, and the coordinates were wrong. The USGS Quad name on the data sheet matches those wrong coordinates. The elevations are wrong enough no one would be fooled into using them.

 

As I believe it to have happened, 3 benchmarks were to be destroyed by road construction. Someone went out and reset them and submitted the data. Unfortunately, those marks were not in the NGS data base The office found three marks with the same designations, '5', '8', and '12' in the NGS files, in an area about 30 miles away where road construction might also have been moving things, and entered the reset data by offsetting the data for those wrong PIDs.

 

Now you've given me a few more to check out when I travel through the right areas.

[+BuckBrooke]

Yikes! Again, how cool! It looks like I've gone after at least a dozen of these already, nearby Albuquerque. More research is needed on my part. This quest also sounds like it will mesh well with finding PIDs in the counties GEOCAC hasn't been seen yet.

 

Hmm, a number of the ones listed in Socorro county, NM, that should be in Sierra county, such as DS0509 are in White Sands Missile Range. Those are going to be tough to check/log.

Edited October 18, 2005 by BuckBrooke

[+rogbarn]

A great list and great info on how it was obtained. It makes me want to learn a bunch more programming languages.

 

I note one possible problem (I only have one example, so it could easily be something else). EB1383 is not on your list. It coordinates as listed on the datasheet are 33 11 13. (N) 078 04 20. (W) which places it in the Atlantic off the coast of SC/NC by roughly 50 miles. It might be a coordinate typo, I estimated its coordinates according to its description as 34-11-34.8 (N) 78-04-16.7 (W). Is it possible that the datasheet coordinates place it outside all counties and therefore got dropped from your list?

[holograph]

Yes, I dropped all PIDs that weren't located inside a county or within its offshore 3-mile limit. Most of them are in Canada, Mexico, or in the Gulf of Mexico. I did that on purpose, since there is no way other than checking by hand to tell if their county names are unrealistic.

 

There were about 300 such "offshore" marks that aren't in the list I published (excluding those in the territories and Hawaii). EB1383 was one of the ones off the Atlantic coast as well as EX0312, BR1004, and FW1605, and a couple dozen others off of Florida, New York, and Maine.

[Bill93]

OP0948 is an example of another way that things can be mixed up and why it is so hard to get a program to figure them out.

 

The data sheet says Kossuth Co, Iowa, the description seems to place it in Emmet Co, Iowa, and the coordinates are very close on the Minnesota side of the Emmet Co IA / Martin Co. MN line.

 

It showed up in Holograph's search because of the distance in Iowa from the named county, but his program listed it as MN because of the almost negligible distance across the state line.

[+PFF]

Holograph wrote:

Caveat: I recommend that before you use this spreadsheet to submit corrections to Cheryl, you should independently compare the locations of the PIDs to the counties they were filed under. How you do that is up to you.

 

Amen! Sometimes the "error" is intentional. See my comments in the Warren County NC vs. Mecklenburg County VA thread for a specific example of a mark placed in one state and indexed in another.

 

I'm uneasy about the continued interest in "correcting" the NGS database. Most GEOCACHING.COM participants are laypeople. We're working with information submitted by professionals for specific purposes. Something may look obvious to us. But as Paul Harvey says, always consider that you may not know "rest of the story". [see Coast Guard Life Jackets, below.]

 

Never submit a "correction" or "destruction" request unless you have personally been to the mark. And accept NGS's decision if they let it stand "as is". They have the Big Picture.

 

-Paul-

 

While working aboard a U.S. Coast Guard cutter, I noticed that the life jackets were not stamped USCG APPROVED. Now, THAT didn't make sense! So I asked the Captain why the rest of the boating world had to use approved life jackets, but the USCG didn't follow the same rules.

 

"Remember the detailed safety instructions and numerous drills you went through before coming aboard?" the Captain asked. "Our life jackets are the best available units. They will keep your head out of the water, even if you are unconscious. And they will float for a LONG time. But they are very complex, and we could never expect the boating public to master their use. Therefore, the units approved for consumers are easy to slip in to. The USCG only will approve devices that the public-- including the very young and seniors with less-than-nible fingers--can handle."

Edited October 23, 2005 by PFF

[ArtMan]

I'm going to post here mainly to bump up this thread and again highlight Holograph's excellent work.

 

Over the past couple of years, I've notified NGS of numerous PIDs in the wrong county. Some seem to be one-off goofs. Others are cases where an agency of one county sets a mark across the boundary in another jurisdiction, but the mark is listed in the county of the setting agency, not its actual location.

 

There are a lot of errors in Virginia, which has many independent cities, i.e., municipalities that are not part of a county. As a result, marks located in the city of Falls Church may be listed as being in surrounding Fairfax County, and vice versa.

 

Redrawing borders also results in errors. A hundred or so marks listed in Chesterfield County, Virginia, are actually in the City of Richmond as a result of Richmond annexing part of the adjacent county in 1970 (?).

 

These discrepancies have in the past been reported to Cheryl Malone at NGS, but she has retired, and we are awaiting word on how they should be reported going forward.

 

-ArtMan-