GPSr Accuracy: Understanding Variance?

[+pgrig]

I have had several situations recently in which I was trying to establish the likely former position of a destroyed station by using my GPSr to "find" its coordinates and then place a stake at that point. Since setting my GPSr to "go to" the station's position seems to put it into an endless process of tiny variations, I have taken to simply trying to move the receiver around on the ground until its reading "just about matches" the station's reported coordinates, in deg/min/sec (to 2 places).

 

So my question is: How do I interpret the linear error represented by .01 sec. to .03 sec of latitude or longitude in measuring a station's position? Or put another way, how do I figure the likely linear (or circular) error of my reading, expressed in feet?

 

Also, can someone help me with a few sources that I can use to read up on (and begin to understand) what I am talking about here?

 

[i use a Delorme PN-20.]

[+PFF]

The system you describe is much better than using the "pointer" at close range. Congratulations for discovering this!

 

Here's a technique which makes it even easier to find a specific spot:

 

Before leaving home, convert the coordinates into Decimal Degrees. I use the conversion feature at

http://www.gpsvisualizer.com/ and I start with the DDMMSS.SSSSSS reading

on the NGS data sheet. (But it also works with DDMM.MMM.)

 

Out in the field, use the UNITS function on your GPSr to change the reading to DD.DDDDD.

 

Now, when you watch the screen on your GPS unit, you will notice that the reading changes with A VERY SMALL AMOUNT of horizontal movement.

 

This can be a hassle if you have to change the settings at each benchmark. But it really helps "zoom in" on a mark!

 

A work-around to changing settings constantly is to download the marks into your GPSr in advance. Then you can run all day in whatever coordinate mode you wish. A bonus is that in my downloading program (MAPSOURCE), I can create a printout of all the marks, giving the coordinates in Decimal Degrees. For reference, I staple this on the back of the map which MAPSOURCE creates.

 

Keep in mind that the accuracy of your GPSr did not suddenly improve. It's just that now you can see one to two decimal points more than what is normally displayed; hence, you see the readings before being rounded off.

 

This is an "advanced" technique, in a sense. It is a notch above the way we generally operate. But again, it can make a difference when searching for a difficult mark, because moving only a few inches changes the reading. Even in the DDMMM mode, it is much more accurate than using the "pointer"!

 

-Paul-

[+pgrig]

Thank you, Paul!

 

This sounds great. I will try it just as soon as our "Miami Monsoon" weather lets up and I can get out again. Since I usually hunt marks that have not been reported in 20 years or more, I encounter many cases in which the "blah, blah, blah structure/ledge/tree" is now a smoothly graded lawn.

 

I have to confess that I don't use Waymarking for routes I work out in advance. I just arrive at the site and generally work only from the Description. The GPSr is usually a last resort, right before the metal detector. I may have to figure out how to really use this GPSr gadget...

 

And oh yes...could you please respond to the initial question? What do a few hundreths of a second in lat/long represent in linear feet of difference in position around here (Mass.)?

 

By the way, if you can understand this entry (MY2767), you'll see why this question of GPSr precision has started bugging me...

Edited August 12, 2008 by pgrig

[NGS Surveyor]

To determine the number of feet in one second of latitude and one second of longitude in MA, I used NGS' on-line program at: http://www.ngs.noaa.gov/cgi-bin/spc_getpc.prl which converts lat. and long. to state plane coordinates. First I entered approximate values of lat. and long. for the center of MA and ran the conversion. Then I added one second to the latitude and ran the conversion again. Then I subtracted the two northerly values and got 101.22 feet. I then did the same for long. (subtracting the easterly values) and got 75.28 feet. (On second of longitude is of course less because the meridians are converging toward the Poles. You can quickly obtain an approximate value for one second of longitude by multiplying the cosine of the latitude times 100, in this case it comes out to about 74 feet.)

 

My old rule-of-thumb was one second of latitude is approx. 100 feet, so that very close to true in MA. So, 0.1 seconds of lat. is about 10 feet and 0.01 seconds of lat. is about 1 foot.

 

For longitude (at the latitude of MA), 0.1 seconds is about 7.5 feet and 0.01 seconds is about 0.75 feet.

 

Hope this helps,

 

GeorgeL

NGS

[+Black Dog Trackers]

What do a few hundreths of a second in lat/long represent in linear feet of difference in position around here (Mass.)?

Use this program to find out this kind of information. Start the program, then select the "Distance" tab. Then, in the Point1 and Point2 windows, put the same set of coordinates. Here is an example N 42 26.804 W 071 03.925

In the Distance Units dropdown menu, select Feet. Change the Point2 coordinates by 0.002 minutes or whatever you want to check. Click Go.

[+pgrig]

Thank you, GeorgeL and BDT! So a couple of hundredths of a sec. in lat or long is roughly a couple of feet of measurement error on the ground. That's reassuring for some of my past efforts at locating likely detroyed old marks. I will try to repeat your calculations and work an example through FizzyCalc. It's great to have folks like you looking over our shoulders!

 

-Paul

[+jwahl]

I generally use a rough approximation mentioned above. 1 second is 'about' 100 feet and 0.01 sec. is 'about' 1 foot. But also remember that the typical recreational grade GPS receiver even with WAAS may only be good to 5-10 feet under ideal conditions. If there is much in the way of structures or tree canopy, this degrades considerably. Averages help a little.

 

You can do a test yourself by taking a number of positions at a given point at different times and even on different days, and see how much they vary.

 

I find the most accurate way to navigate to a known point is to use the direction indications from a few points nearby (50 feet of so) and intersect them. Or if you are on the move and in the clear the so called compass screen.

 

Thank you, GeorgeL and BDT! So a couple of hundredths of a sec. in lat or long is roughly a couple of feet of measurement error on the ground. That's reassuring for some of my past efforts at locating likely detroyed old marks. I will try to repeat your calculations and work an example through FizzyCalc. It's great to have folks like you looking over our shoulders!

 

-Paul

[+Klemmer]

pqriq:

 

The previous responders probably aren't familiar with the Delorme PN-20 features, which I also use. A couple pointers that might help you:

 

1) It has an excellent averaging function, which is very handy. After saving a waypoint, you can even go back and re-average, or continue averaging. Look for the "AVE" function on the save waypoint or edit waypoint screens. Experiment with it, if you haven't.

 

2) Paul's method of going to decimal degrees works for him (and others) because they get more precision (accuracy?) that way, since typically the commonly used decimal minutes format only goes to three decimals (or similar / worse issue with decimal seconds). The PN-20 goes to four digits in decimal minutes (about 0.6 feet or slightly less per 4th decimal place). I haven't explored how many decimals it goes to in other formats, since I tend to stick with decimal minutes on it anyway. Yes, that 0.6 feet or so is way more precision than the GPSr has accuracy [get the terms right?], but when you are searching (or averaging) it gives you good insight into what is going on. You can more easily see it "steady up" or wander (in heavy tree cover, etc). Very handy.

 

3) With the PN-20 mapping features, we have another huge advantage: We can load USGS 7.5 minute Topographic "Quad" maps onto the GPSr. If you haven't bought them, think about it. You can get most whole states for $50 for PN-20 owners, or small sections from the $100 free downloads included with PN-20 purchase. What advantage? Most (80 - 90% or so out here in CA?) NGS benchmarks are shown on the USGS Quads. The error that is typical in scaled marks was (I believe) mostly introduced in the paper process of looking at marks on maps and measuring (scaling) their location. With the PN-20, loaded with the hunt area USGS Quads, you can often put the cursor on the "X" for the benchmark on map, and go right to it in many cases. You still have GPSr accuracy to be concerned about, and all the other usual issues (marks missing, buried, etc), but it often works VERY well. Delorme registration (alignment) of the maps to their GPSr position is never perfect, but for me, it has always been better than the GPSr accuracy. Some may point out the USGS quads are often out of date. True. But with a couple button pushes, we can switch to the Delorme Topo USA maps, which (in my experience) are very much up to date, and are being maintained.

 

Good luck!

P.S. Delorme has a very helpful & friendly forum, if you haven't been there. Link from their website.

 

Klemmer

Edited August 12, 2008 by Klemmer & TeddyBearMama

[+GEO*Trailblazer 1]

I started using this method when I first started and still use it today.

 

I had a geodetic grid for my area and calculated it out originally with the GPS unit.

I started with 90 00.000 and 36 00.000.

 

90 00.000 to 91 00.000 = 56.025 Mi. divided by 60 = 0.93375 Mi.

36 00.000 to 37 00.000 = 68.953 Mi. divided by 60 = 1.14921 Mi.

 

36 00.000 to 36 00.010 = 60.673 feet.

36 00.000 to 36 00.001 = 6.673 feet.

36 00.000 to 36 00.0001 = .673 feet. (GPS Trackmaker has 5 decimal places.)

 

90 00.000 to 90 00.010 = 49.302 feet.

90 00.000 to 90 00.001 = 4.93 feet.

90 00.000 to 90 00.0001 = .493 feet.

 

Edited

my math...duh!!

This can also be done by putting a local grid for your area in the GPS and seeing the distances of the same.

Edited August 12, 2008 by GEO*Trailblazer 1

[Bill93]

As others have pointed out, the waypoint averaging function is very useful. For best results, I would get to the approximate location, plant a wooden stake or unique rock, and average a waypoint over that stake for as long as I could stand, no less than a half minute and maybe minutes. Then I would figure out how far North/South and East/West my desired point was (maybe using the Measure Distance function on the GPSr map) and tape from the first stake to the new location (hopefully less than 10 ft).

 

If you go back hours or days later, then average a new value for comparison. Not only is more averaging always better, but the average of separate short sessions is better than the same number of minutes in one session due to the chance to average out slower changes in the atmospheric conditions that affect the satellite signals.

 

I'm jealous of your display resolution. I have the typical 5 place degrees (best), 3 place minutes, or 1 place seconds. The rounding or truncation to even dd.ddddd degrades the results when signal quality is extremely good. I also don't have the "continue average" function, which is great for the multiple-session work.

[Wintertime]

If you go back hours or days later, then average a new value for comparison. Not only is more averaging always better, but the average of separate short sessions is better than the same number of minutes in one session due to the chance to average out slower changes in the atmospheric conditions that affect the satellite signals.

And also because the GPS constellation will be more different over multiple days than over a few minutes or hours on the same day, right?

 

Patty

[Bill93]

That too. The change in constellation helps randomize both local reflections and the atmospheric paths, resulting in better averaging. A longer average in one session may indicate good accuracy but still have some bias that didn't change over those minutes, whereas a later session will usually have a different bias.

[+GEO*Trailblazer 1]

If you remember we did a EDM talk once as well.

 

I went to both Staions here in SW Missouri and got readings I liked for my handheld and it gave me great confidence.

EDM Calibration