Averaging To Get A More Accurate Position?

[+Bob Blaylock]

  A few days ago, a friend of mine was telling me about a project in which he was involved.  He was trying to measure a change of elevation, over a period of three years, with that change expected to be on the order of a few millimeters.

 

  Several measurements were taken over a period of several hours, on each of several days, using a GPS receiver.  The process was to be repeated again in three years.

 

  I gather this was done using a regular, consumer-grade GPS receiver.  I don't know if that receiver was WAAS-enabled or not, but at best, it could only have been accurate to a radius of several feet.  I guess the theory was that if a large number of measurements were taken over the course of several days, and averaged, they could come up with a much more accurate measurement than what would be expected of a single measurement.

 

  There were also a few various measures taken to avoid some sources of interference.  For example, the GPS was placed on an aluminum disk, on the end of a pole raised above the ground, supposedly to avoid reflected multipath signals, and workers were forbidden to park vehicles near the GPS.

 

  It does seem believable to me that using such methods, you could get indeed get a more accurate measurement; but it does not seem believable to me that you could this way get a measurement as accurate as this project seems to require.  (Remember, they are looking for an elevation change on the order of a few millimeters.)

 

  What do you think?

[Photobuff]

No way with consumer stuff. I've averaged for days using SA Watch, and I'm confident one can get within a foot for horizontal, and maybe a couple feet for elevation, maybe a bit better, but not mm. The geodetic class instruments can do it, but I believe it takes them several days of data collection, and a lot of post processing. I believe it can also be done using a remote station referenced to a known benchmark, if you have the equipment to do it, and a nearby benchmark. They're wasting their time.

[toddm]

Very interesting, were they trying to evaluate isostatic rebound or a orogeny event of some kind?

 

I know this is done a lot for those types of evaluations but as Photobuff said pretty expensive equipment is used for it and days worth of data consolidated.

 

It would also be interesting to evaluate different algorithms and see how they fare with long term data in a consumer unit, but I believe it would be hard to get anything in the mm range. At some point it becomes exponentially hard to reduce the error factor even a small amount.

 

When you really start to evaluate all the variables in such a project, gravity and magnetic anomalies, original sources of error in benchmarks, climatic conditons etc. the measuring units error factor, etc. it's amazing they can get within mm's with any equipment. It would be interesting to see if even some of the top professional gear could really give repeated results of finding the same spot within mm's from year to year.

[+Deneye]

One of my favorites..."You can't get there from here."

 

In this case though, "It can't be done".

 

The North American plate would have shifted by a millimetre or 2 in the time period given, thereby causing a dislocation of the physical point on Earth and the virtual point of location (lat-lon) which means that even if the device could pick up the exact same virtual spot, it would be measuring the physical one differently, with the possibility that the original physical spot hadn't of change except by moving sideways.

 

Plus what the others above have said

[+GOT GPS?]

I think the best way to measure land, like near faults, and volcanoes, is to use lasers, and tilt sensors, to measure the rise, and fall of the land, and the change of slope of the land.

 

Sub-meter is hard to do in the first place, using expensive equipment, and the satellites themselves, are not perfect, including the change in position and angle of the antennas on the satellites, now if the antennas on the satellites were at the exact center of the mass of the satellite, that would help a little.

 

-

Geoff

[GeoidPS]

It comes down to statistical analysis. The standard deviation should be 90% confidence level.

[+Sputnik 57]

My bet is that if GPS was involved at all, they had to be using Trimble or other high end equipment, and comparing elevation differences between a fixed elevation ground unit and the unit at the site. Even so, I'm not sure sub-millimeter is possible, but I've never gotten to work with such fancy toys.

[+Elde]

A few days ago, a friend of mine was telling me about a project in which he was involved. He was trying to measure a change of elevation, over a period of three years, with that change expected to be on the order of a few millimeters.

On top of everything else mentioned in this thread - a standard GPS display probably doesn't have enough decimal places to display to that accuracy.

[+GOT GPS?]

Would be in interesting experiment to do all this measuring, then raise up the GPS a couple inches, then try it again, to see, if you see a difference.

 

The Tidal wave that happened over a year ago was one event that changed the earth in an instant, and maybe that could have been measured with GPS units, as a noticeable change.

 

-

Geoff

[+Alan2]

A few days ago, a friend of mine was telling me about a project in which he was involved. He was trying to measure a change of elevation, over a period of three years, with that change expected to be on the order of a few millimeters.

On top of everything else mentioned in this thread - a standard GPS display probably doesn't have enough decimal places to display to that accuracy.

 

eLDE'S RIGHT. eVEN IF THE gps WAS EXTREMELY ACURATE, THE "PRECISION" LIMITS THE DATA REPORTED. iF YOU SET THE COORDIANTE TO utm FOR EXAMPLE, THE BEST YOU CAN READ IS + OR - 1 METER. uSING MINUUTES DECIMAL, YOU'RE OFF A FEW FEET. oops sorry about the caps.

[Photobuff]

Actually, I think you can do far better than the display precision. By averaging enough numbers, the percentage one way or another determines the next digits, potentially way beyond what's displayed. For best results you can do what SA Watch does, which is weight the data depending on the quality of the data. Eventually it comes down to systematic errors and signal to noise. Someone who understands the statistics should be able to calculate how long one would have to average for a desired precision, given a certain quality and resolution of data. Still, I don't believe mm precision is possible with consumer GPS units due to a plethora of unknown and uncontrolled error sources, that probably won't average out in any reasonable time frame.

[toddm]

Just for giggles I left my 76csx on overnight averaging a waypoint, and in the morning it was still reading an accuracy of around 3' and it was off at least 8' vertically.