holograph
-
Posts
813 -
Joined
-
Last visited
Posts posted by holograph
-
-
The exercise by Holograph is a good idea, but there is something wrong with the coordinates for MY3374.
If you Inverse between each pair of the 4 points in one datum and compare to the distances in the other datum, you find reasonable agreement between all the pairs that do not involve MY3374 (0.066 m, 0.161 m, and 0.037 m), and orders of magnitude larger difference on the ones involving it (19.8 m, 26.6 m, 26.4 m).
Looking at it another way, if you ignore the fact that they are different ellipsoids, and Inverse between the US1901 and NAD83 values for each point, 3 of them shift about 41 or 42 meters at 99 degree azimuth, but that point shifts 33 meters at 139 degrees.
I modified the least squares calculation to use MY0001 instead of MY3374, and now the result is within about 2 feet of the NAD83 location.
So the "bad" control point MY3374 did affect the calculation to the tune of 16 feet of error, as you would expect, but not as much as if it were the unlucky choice of the other methods, which may have resulted in 136 feet of error.
-
Well, as I said above, my test was Klemmer's method vs Papa-Bear's method. The results were nearly identical to each other (31 inches Latitude and less than one inch Longitude). My thinking is that two different methods yielding essentially the same result pretty much proves both methods. I'll test it again in the process of doing the actual conversions for my trip, and advise. I suppose I might have just gotten lucky on one test.
Klemmer's and Papa-Bear's methods were essentially the same method, just differing in the detail of when the average was taken. They are both assuming a simple linear shift in datum -- no stretching, warping, or rotation. The least squares method makes no such assumption -- its assumption is that the measured angles between stations should change as little as possible, and it is able to adjust for rotation, stretching, and warping.
In the case of the CA-NV boundary, since all the stations were part of the same survey project and adjustment, the method that Klemmer and Papa-Bear are using should work as well or better than anything else. The CA-NV boundary has the advantage of a dense neighborhood of stations that have coordinates in both datums and were part of the same network, so it is easy to find nearby control points to use and it is not likely that the network was stretched or warped significantly.
It is possible/likely that the orientation of the datums varied slightly, which would introduce a systematic error in azimuth, but it would be easy to check that by comparing the azimuths between control stations. In other words, a difference in datum orientation would mean a difference in the assumed direction of true north, and should be accounted for when using the Forward routine to compute the location of the missing station.
That could be another slight advantage of the least squares method -- it only uses differences in azimuth rather than absolute azimuths -- so it won't be affected by differences in datum orientation.
-
If you Inverse between each pair of the 4 points in one datum and compare to the distances in the other datum, you find reasonable agreement between all the pairs that do not involve MY3374 (0.066 m, 0.161 m, and 0.037 m), and orders of magnitude larger difference on the ones involving it (19.8 m, 26.6 m, 26.4 m).
Looking at it another way, if you ignore the fact that they are different ellipsoids, and Inverse between the US1901 and NAD83 values for each point, 3 of them shift about 41 or 42 meters at 99 degree azimuth, but that point shifts 33 meters at 139 degrees.
You may be right that MY3374 is suspect, but it is also true that it is not unique in that aspect. I have a small dataset of 430 stations in the same neighborhood with both US 1901 coordinates and NAD83 coordinates. The shift in latitude averages about 7.1 meters southward, with a standard deviation of 4.4 meters, and a shift in longitude of 43 meters eastward, with a standard deviation of 7.4 meters. The result is that the azimuth of the change vector varies considerably across the dataset. The least squares method tends to smooth out that variation.
I didn't single out MY3374, other than filtering out obvious outliers (probably due to misprints in the source document for the US 1901 coordinates). The way I chose the control was by selecting a few first- and second-order stations surrounding the test point. I didn't pay attention to whether the control points had a uniform change, because past experience indicated to me that the changes are not consistent, and that was the motivation for using least squares.
edit: corrected eastward/westward average
-
Just for the heck of it, I used a complicated method using least squares to estimate the location of a station, using surrounding control points whose positions are known in both the old datum and NAD83. The working principle is that the station was located using triangulation, so by transforming the assumed geometry of the network in the old datum to the current geometry in NAD83, we can estimate the station position in NAD83.
It's kind of Klemmer's and Papa Bear's methods on steroids.
The result is mixed: you can get lucky with the simpler methods, and get a good estimate, but you can also get unlucky and get a bad estimate. The least squares method tends to even things out. Here is a sample using 3 control points and a test point, with coordinates in the old US Standard Datum of 1901, as well as NAD83:
Control points:
MY2568 US 1901 - N42.641119 W71.106908, NAD83 - N42.641056 W71.106410
MY4858 US 1901 - N42.611644 W70.730322, NAD83 - N42.611579 W70.729815
MY3374 US 1901 - N42.342606 W71.133944, NAD83 - N42.342380 W71.133682
Test point:
MY5139 US 1901 - N42.502850 W70.965125, NAD83 - 42 30 10.04107(N) 070 57 52.65249(W)
The least squares method estimated MY5139's position to be 42 30 9.89 (N) 70 57 52.79 (W) NAD83
Using a single vector from MY2568, MY5139's position was 42 30 10.03 (N) 70 57 52.66 (W)
Using a single vector from MY4858, MY5139's position was 42 30 10.26 (N) 70 57 54.45 (W)
Using a single vector from MY3374, MY5139's position was 42 30 10.26 (N) 70 57 54.45 (W)
Those estimates have errors of 18 feet, 1 foot, 136 feet, 136 feet, respectively.
So if you chose to use the vector method from MY2568, you would be lucky and get within 1 foot, but if you chose MY4858 or MY3374 (or averaged all three), you could be off by a hundred feet.
The least squares method is far too complicated for me to describe here, but I may try to encode it in a spreadsheet at some time. Most of it is already in a spreadsheet that I developed to test it, but it would take considerable work to put it in an easily-used form. I only did it because I've been curious about ways to convert old datums, and have wanted to try some kind of geometry-based least squares for a long time.
-
Maybe Aich's metal?
Here is a description from an 1888 metallurgy book: The Metallic Alloys: A Practical Guide for the Manufacture of all Kinds of Alloys, Amalgams, and Solders, Used by Metal-Workers; Together with their Chemical and Physical Properties and their Application in the Arts and the Industries; with an Appendix on the Coloring of Alloys. (They really knew how to write book titles in the old days!)
Aich's metal. -- This alloy, named after its inventor, consists of a brass to which a considerable degree of tenacity has been imparted by an addition of iron. It is especially adapted for purposes where the use of a hard, and at the same time, tenacious metal is required.According to analyses of various kinds of this metal, it shows, like other alloys, considerable variations in the quantity of the metals used in its preparation. Even the content of iron to which the hardening effect must be ascribed may vary within wide limits without the tenacity, which is the principal property of this alloy, being modified to a considerable extent .The best alloy, which can be called an Aich's metal, is composed of copper 60 parts, zinc 38.2, iron 1.8. The content of iron must be limited to from 0.4 to 3.0 per cent. Another Aich's metal showing excellent properties is composed of copper 60.2 parts, zinc 38.2, iron 1.6.The chief property of Aich's metal is its hardness, which is claimed to be not inferior to that of certain kinds of steel.It has a beautiful golden yellow colorand is said to oxidize with difficulty, which makes it of great value for articles exposed to the action of air and water. -
The January statistics are available on the statistics page.
Because I updated the statistics on January 9, there were no significant changes to the NGS counts between then and now. I did not update the maps or the counts by county.
The counts for the Geocaching.com logs were updated.
-
Mornin' all,
Anyone out there know if Wyoming has set their's yet?
Cheers,
JCS
In Dec. 2005, I contacted Richard Muscio, President of the PLSW, and he stated that to his knowledge, no center of population monument had been set in Wyoming.
-
The revised December statistics are available on the statistics page. The maps and counts by county have been updated also.
Since it is still relatively near the beginning of the month, I decided to process the updates, which are mostly from December and earlier. The most recent recovery report added to the datasheets was from January 6, 2010. Of course, this will mean that the upcoming update in February will show lower counts.
-
Not quite agreed. I've recovered many triangulation station, many with no reference marks (but some with "witness marks" such as tree blazes) or whose reference marks were not entered into the NGS database. Ditto no box store. Ditto third order. Practically all the US-Canadian boundary monuments are third order and no box score. But yes they were occupied triangulation stations (or traverse stations) and are not intersection stations.
Bottom line - you gotta dig deep. There is no simple formula.
Not only that, but the resurrected PD0674 Grand Manan station of the Eastern Oblique Arc appears on maps as an intersection station, but it is First Order!
So the rule is: there are no rules.
-
...
We have found 3 fire lookout towers, each with a disk in the center underneath. We are having trouble figuring out what exactly a “Found” would consist of because we found the four tower foundation supports and the disk but the tower structure is gone.
For example, we found the disk and the 4 foundation supports for RD1845. I have copied the text below. In the 1954 description. It says the tower was destroyed but the station was found in good condition? Huh?
Can somebody explain that? Isn’t the center of the tower the “station”?
Before I log RD1845 I am trying to decide if I should call it destroyed because the tower structure is gone or call it found because we found the disk?
...
It appears that RD1845 was an intersection station, but that the geodetic position was transferred to the disk that the Forest Service plumbed underneath. So the geodetic position was preserved, but the tower was destroyed. If you found the Forest Service disk, you found the position, and can record its condition as poor, found, or destoyed (if the disk is now destroyed). The station (position) is still an intersection station, though, because the position was originally determined by intersecting the top of the tower.
It is of course possible that the original survey crew did do observations from the tower, in which case it may not be an intersection station. Only the field notes could tell you for sure.
-
If the datasheet does not specifically mention in the description that the station is an intersection station (or was intersected) is there a way to positively identify if the station is an Intersection station?
I see datasheets with adjusted coordinates and no mention of setting type and figure that is a good indicator of an intersection station but I suspect that more clues are needed to absolutely identify an Intersection station?
Absolutely? An intersection station is a horizontal control point that was not occupied with a theodolite or GPS, and the published datasheet doesn't tell you that, so you can never know absolutely unless you have access to the original observation notes. There are a couple of reliable clues, though.
- Intersection stations will be 3rd order horizontal control.
- Intersection stations will be established with "classical" methods.
- Intersection stations will be artifacts whose purpose is not primarily geodetic. i.e. they will be buildings or structures that predate the first observation.
- Intersection stations will be points that would not be "occupied", i.e. if you can't envision a person putting a theodolite at the point, there's a good chance it's an intersection station.
- If you see "the point observed was..." in the description.
- If you have a map of the triangulation network, and the lines meeting at the point are dashed or dotted.
Even mountain peaks might be intersection stations if no one bothered to hike up to the top with heavy equipment and take sightings on other stations of the network.
- Intersection stations will be 3rd order horizontal control.
-
I've updated the statistics page, but there were only 2 additional GEOCAC NGS recovery reports, both of which were backdated recoveries for April 2008. The geocaching statistics were updated as usual.
-
...
Here are the two National Geodetic Survey (NGS) - previously Coast & Geodetic Survey (C&GS) resources that some of use use to research these older (and very interesting) projects:
C&GS Annual Reports (1837 - 1965)
C&GS Special Publications (1898 - 1969)
There are some indexes & bibliographies on those pages that are searchable (on-line or in Acrobat) that make finding things a little easier. At a quick glance, I did see a number of surveys in Louisiana, but didn't see the completion of the arc. I didn't check Arkansas.
...
Special publication 187 is the document for Arkansas. There was a series of publications from 1885 to 1925 that documented the triangulation on the U.S. Standard datum, but the series of publications was never completed for all states, and Arkansas was not included. Then another series stated in 1930 through 1941 on the NAD27 datum, but again didn't cover all the states, but it at least included Arkansas.
-
Thanks Bill and DR. I looked at the 1940 London map. Might make it a lot easier to find the benchmark location. Also it will be pretty cool to go around the county now and look at where old things used to be.
Thanks!
I've done a lot of work with the historical Topos from the late 1800's and early 1900's, and I have to warn you about a few things. First, the spatial accuracy is generally horrible by today's standards. Second, the datum is probably the old US Standard, which is approximately the NAD27, but for Ohio, it might be the Lake Survey. The datum alone accounts for large errors unless you adjust for it -- meaning that you have to reproject the map into a modern datum such as NAD83 or WGS84. Finally, the USGS seemed to be not very careful about the location of railroads on those old maps. So you have to allow for possible discrepancies of hundreds of yards, if not more.
Having said that, I've found the old maps to be useful for locating old rail lines and even old roads that have long disappeared. Using the old maps as a guide, it is often possible to locate the faint traces of the features on modern aerial surveys (e.g. Google Earth, Bing Maps) which provide accurate modern coordinates.
Another resource that is available in some states is the old Agriculture Adjustment Administration aerial photography from the 1930's. It's amazing how much stuff is clearly visible in those old photos, when it is totally obscured in modern images. The problem with Google Earth and Bing maps is that they like to have nice thick green forests in the images, which obscures the ground features. In the 1930's, many of the forests were logged and thinned, so ground features were much more visible.
If you can't find the AAA photos, you could try the USGS Digital Ortho Quads. Those photos were generally taken without foliage cover, so features are usually clearer, and they are georeferenced to modern datums. DOQs used to be available at Terraserver, but they seem to keep moving.
-
The November statistics are available on the statistics page. The maps and counts by county have been updated also.
The new recovery reports seem to be a few stragglers from the beginning of the month. The most recent recovery report added to the datasheets was from November 3.
-
Thanks, Dave. I watched the whole thing and found it interesting and informative. I also learned how to correctly pronounce some acronyms and words that I have been mentally mispronouncing.
-
WOW, Arizona is getting close to "done". I thought about making a day trip to fill in some of western NY, but instead planned a trip to Florida for the shuttle launch.
Arizona benefits from having large counties and plentiful benchmarks. I tried once to do a coloring scheme based on % of benchmarks recovered, but that merely handicaps a different set of counties that have large numbers of unrecoverable benchmarks, and gives small counties an advantage.
The current scheme is at least easy to understand. Other schemes might require pages of mathematics to explain.
-
The October statistics have been updated at the statistics page. The maps and the totals by county have been updated also.
There were 1,360 GEOCAC recovery reports added to the NGS datasheets. The most recent recovery report was dated October 27, so it looks like the statistics are up to date.
-
Darn! Hudson County, NJ is so small, my one find in Jersey Cty doesn't show on the map!
I need to work on larger counties!Ta da!
-
The September statistics are available on the statistics page. The counts by county and the maps have been updated, also.
There were about 350 recovery reports added to the datasheets. It looks like the most recent recovery report was dated September 3, so many of your September recoveries haven't shown up.
Kayakbird was busy in North Dakota and Minnesota, though!
-
The statistics page has been updated with the August statistics. The counts by county and maps have been updated, also.
edit: corrected image links
-
holograph, I'm sorry about the two letter initials. I'll start going to JMA then when I do my next batch of benchmarks. I did a few a couple of days ago that are JA, they were all in Franklin County. But then I noticed that someone else is using JMA, so tell you what, you tell me which is easier for you and I'll do it.
As for the one in Maryland, that was a trip I was on, and happened to be near it, so I snagged it to say I got a BM and a cache in Maryland. I doubt that I'll be near there anytime soon. I would however be in the Nelson County area of Kentucky more often than that (in-laws house) so that's a probable area for me.
You may as well continue with 2 letters then. If there are two people in the same area using the same 3-letter initials, I won't be able to keep them separate. If one of you uses 2-letters, and the other uses 3-letters, then I (actually my programs) at least know they are different people.
-
Awesome. and I see where my work is...I'm that little sliver of green in western Arkansas.
Question, can you see what the Benchmark that is initialed JA is? It might be mine, and I was just curious.
It looks like it was HV7850. It will show up in next month's statistics.
As a rule, when people enter the NGS reports with two-letter initials, I don't automatically assume that it is the same person that has previously entered reports under those initials. I have a secondary lookup table that I maintain by state and optionally county, to resolve two-letter initials. Since that was the only mark recovered in Maryland under the initials "JA", it wasn't automatically assigned to you. Since it doesn't look as though you get into Maryland very often, I've narrowed the match so that Maryland "JA" recoveries only in Montgomery county will be assigned to you.
When people enter reports with three-letter initials, I generally assume that it is the same person throughout the entire country. There are only a few exceptions, for instance Camper1 and Gnkihog both use the same "JLH" initials, and I have to separate them by state.
When a new set of initials shows up, I try to find a corresponding Geocaching.com log entry to find out who it is. After that, the process automatically assumes that three-letter initials are always the same person. The only way I can learn that multiple people are reporting under the same three-letter initials is when someone complains that their statistics are wrong.
Basically, it is a lot easier for me if people use three-letter initials, because there is less manual processing.
-
The July statistics are available on the statistics page. The county totals and the maps have been updated, also.
1,817 datasheets had new GEOCAC recovery reports this month. Most of the updates were for reports from May 14 through July 28, with a smattering of earlier reports.
From the map, it looks like the people out West have been busy. The Easterners, especially in the northeast, may have been discouraged by the unusually cool and wet spring and summer this year. It also looks like someone burned a trail of recoveries from Texas to Montana!
Kayakbird captured three new northernmost stations in the 48 contiguous states: the 8th, 9th, and 10th most northern stations of the lower 48.
I need to work on larger counties!
Very Old Datum Conversion to WGS84 Datum
in Benchmarking
Posted
Bill93 and Billwallace, I posted a description of the process at this page.
As I said, it is too complex to do routinely without having some software to do it automatically. I was just trying a proof-of-concept, and may or may not try to make it easy to do with an Excel spreadsheet or comparable program.