GPS tracklog inconsistencies

[Peakbagr]

For those who use Nat'l Geographic on your computers to plan routes and then upload tracklogs - Do you sometimes see that the tracklog of where you hiked does not correspond with where you knew you'd been, like right on clearly known summit or high point? This happens to me occasionally where I know I've prowled over the wooded highpoint but the tracklog on the computer indicates I was a hiking just below the top.
I've thought this might be an error on some USGS maps that the NG mapping software copied. I use a good GPS unit, 60CSx. I know that it doesn't have commercial accuracy, but its accuracy level is near 12' to 16' and I'm guessing this is not the issue.

Not a big deal for me, but a cartographer friend using her own GIS is seeing some of the same things. I'm tried to see if this shows up with other frequent GPS users from time to time.

 

[spider solo]

Though I cant say on the Nat'l Geographic aspect of the question.
I would say it has to do with how frequently the Gps makes it's track..
I think the wooded summit being the key .

If it takes a reading every 10 minutes and it misses a couple of readings an easy 20 minutes has passed without it really knowing where you are.

Perhaps that's not the case for you but it is a common GPS situation.

For example mine (76csx) often shows me going overland when I have actually paddled around a point of land. It's just that I went around the land or obstruction before it took it's next reading and shows me going right through it.

 

[BorealChickadee]

I've had the same issue this summer several times with a new 60csx. Walking along a peninsula, both out and back the track had me clearly in the lake. It was laughable but at the same time raised the question "why".

Spider...to this techno challenged human, if decreasing the time between points would help how would one go about it?

 

[RoySwkr]

Spider...to this techno challenged human, if decreasing the time between points would help how would one go about it?

My sister used a Trimble to make GPS maps for the Forest Service, one of the big issues was switchbacks.

You could set the sampling interval on the GPS menu but too often meant the storage would fill up fast, so she would stand at the end of each switchback for the sampling interval (30 sec?) to make sure it got a point there (the unit would store only a limited number of named waypoints). Then of course you had to make sure the mapping software didn't average out those points.

As to the original question, I doubt that any hiking maps produced claim to be accurate to 16 feet even if the GPS is. There is a well-known map producer who has gotten people lost in at least 2 states because of trails in the wrong place or misnamed, but hikers love them because they're waterproof

 

[MichaelJ]

Check your datums. The software may expect the track to be in one (NAD83, for example) but your track could be in another. Such mismatches can cause a consistent several feet (or more) off.

 

[spider solo]

I've had the same issue this summer several times with a new 60csx. Walking along a peninsula, both out and back the track had me clearly in the lake. It was laughable but at the same time raised the question "why".

Spider...to this techno challenged human, if decreasing the time between points would help how would one go about it?

Yes the reverse is true for me I have paddled many a time on solid ground.

For me to change the intervals I go into "track log" (then it offers 4 options one of which is
"setup" where I go to change how I want it to record.
You can set it for Auto, time or distance
I choose Auto
Below that box is a box called, Interval
with choices of
normal... more time... less time

(I don't actually see how many minutes those choices equal but it might tell me somewhere else) .


I have mine set for Auto
normal

(I go easy on it so my batteries last longer)

Hope that is of some help. I think yours will offer much the same choices.
Good Luck

 

[DougPaul]

it misses a couple of readings an easy 20 minutes has passed without it really knowing where you are.

If you have the track recording set this sparsely, you deserve what you get...

A typical consumer GPS computes your position once per second (independent of the tracklog settings). Depending on the settings, you can record at longer intervals. When I am recording data for bakeoff style repeatability experiments, I set the interval to 5 seconds when walking (~20 feet at walking speeds).

A sparse track recording interval will not affect the accuracy of the recorded points, just the interpolated track (usually a straight line) in between the recorded points. (The GPS produces one location per second, the recording interval simply influences how many and which locations are saved in the tracklog.)

Auto normal
(I go easy on it so my batteries last longer)

I also set my recording interval to auto, normal for normal use. On a recent hike where I was walking a bit faster than 2mph, it recorded a point at an average interval of 47 ft = 15 seconds.

I doubt that the track recording interval has any effect on battery life.

Doug

 

[spider solo]

Interesting I'm really surprised that it tracks so frequently.

When having the Spot device track it is once every 10 minutes, so when people want to see where I am it could be 9 minutes ago or 1 minute or so ago depending when they are checking.

Back with the GPS
So, with the Auto setting do we know if we are checking time or distance.
if distance how far apart.

Thought for sure when leaving a "bread crumb" trail you could set it to a quarter mile if you wanted.

I was thinking with the time setting the Gps does indeed check the satellites frequently much like you say, but does not actually record a track point anywhere near that frequently.


I'll be "checking the books" when I get a chance but that was my understanding.

The more tracking you use the more battery you consume...oh now I need to read the manual again.....

 

[DougPaul]

The GPS fundamentally estimates your position in cartesian (xyz) cordinates and then converts it to a lat-lon-altitude format all without regard to the topography. (So yes, it can place your boat on land, your car on the water, and you below lake/sea level.) Only then does it plot the position on a map (which can have its own errors) and the position is presented errors and all rather than guessing at where you "should" be. If this position is illogical based upon additional info that you have, then so be it. Get used to it.

Typical use of consumer GPSes can be far less accurate than the rated specs. The 90% chance of the indicated horizontal position being within 10 meters of your true horizontal position spec assumes a clear skyview, no trees, no multipath, etc*. How many of us use external antennas up on a pole or on top of our heads and have a full level horizon in all directions? Furthermore maps have errors which can sometimes be large in some areas. The map features (eg man-made objects, landforms, and land-water boundaries) can also change much faster than maps can be updated.

* A surveyor will check the predicted satellite constellation (and satellite health) and wait for good condition to make his measurements. He will also stay at one place and average to improve accuracy. (Current best horizontal accuracy is <1 cm for 24hrs of averaging using multi-frequency carrier phase and post-processing. Your consumer GPS can't get anywhere near this accuracy under any conditions.)

FWIW, I have a track recorded during a 9 mi kayak trip on the Sudbury and Concord rivers--the 60CSx GPS was in a below-deck dry-bag between my knees. I see no place where it obviously puts me "on land" when plotted on an NG 24K topo map. When I display the track on a Garmin Topo 2008 100K scale map, it shows some land incursions. (I don't have a 24K scale Garmin-compatible map for the area.)

Check your datums.

Some GPSes/mapping software is sensitive to the datum settings at the time of the data transfer. Other GPSes/software assumes that the datum of the data is always WGS84 and is not sensitive to the settings at the time of transfer. (FWIW, GPSes work in WGS84 internally and only convert for input or output, such as to the display.)

Doug

 

[psmart]

Interesting I'm really surprised that it tracks so frequently. When having the Spot device track it is once every 10 minutes, so when people want to see where I am it could be 9 minutes ago or 1 minute or so ago depending when they are checking

I believe the Spot reporting interval is kept fairly long to save satellite bandwidth, and perhaps to extend battery life. But the GPS receiver in the Spot is probably capable of getting a fix more often - you just can't see it. This is one of the limitations of the Spot as compared to a conventional GPS. They're designed for different purposes: Overall course recording vs. interactive navigation.

 

[DougPaul]

When having the Spot device track it is once every 10 minutes, so when people want to see where I am it could be 9 minutes ago or 1 minute or so ago depending when they are checking.

The transmitted track points could be of typical GPS accuracy and anything in between is interpolation (ie guesses). GPS locations also include time which is presumably transmitted by Spot.

Transmitting the location would consume large amounts of power. As psmart suggested, a long transmit interval will conserve batteries and satellite bandwidth.

Back with the GPS
So, with the Auto setting do we know if we are checking time or distance.
if distance how far apart.

Thought for sure when leaving a "bread crumb" trail you could set it to a quarter mile if you wanted.

I suspect the auto settings use a combination of time, distance, and possibly track shape. (You don't need as many points to accurately represent a constant speed on a straight line as you do for varying speed and/or a curving track.)

I was thinking with the time setting the Gps does indeed check the satellites frequently much like you say, but does not actually record a track point anywhere near that frequently.

A typical GPS computes the position once per second independently of the other settings except for the battery saving mode. If you don't compute your position frequently enough, you will lose satellite lock. (The trip computer is based upon this once-per-second data.) Garmin gives you three options for track recording: by time, by distance, or auto (some combination as noted above).

The more tracking you use the more battery you consume...oh now I need to read the manual again.....

Track recording consumes very little power--it is just a write in non-volatile memory. Receiving the signals, computing the location, and updating the display consume far more power.

Doug

 

[spider solo]

I've never been more quoted in all my life.
Kinda makes you feel all warm and fuzzy

Yup, I did find it in the manual
time..in seconds
distance...in feet
auto ....map source decides which points to drop.

To Peakbaggers original question: the answer would be map inaccuracy mixed with GPS inaccuracy.
But not dropped points because of tree cover if my revised understanding is correct?

 

[DougPaul]

Watch out--here come some more quotes...

I've never been more quoted in all my life.
Kinda makes you feel all warm and fuzzy

Yup, I did find it in the manual
time..in seconds
distance...in feet
auto ....map source decides which points to drop.

To Peakbaggers original question: the answer would be map inaccuracy mixed with GPS inaccuracy.

Correct.

But not dropped points because of tree cover if my revised understanding is correct?

Points are not dropped because of tree cover. The GPS can lose lock due to loss of the satellite signals (perhaps due to tree cover) and simply has no location points to save (or drop). This creates a break in the track which will generally break the tracklog into separate segments*. (These are marked in a GPX tracklog with the <trkseg> </trkseg> delimiters. A track with no breaks will be recorded as a single segment.)

* The GPS manufacturer can also choose to do something else...

Some display programs may concatenate all segments of a track so you cannot differentiate a break from sparsely saved points.


One way to think of the track saving process is: The GPS creates one location point per second (call it the "full tracklog") and then saves some subset in the tracklog (according to the interval setting). Each individual tracklog point is still as accurate as it was when it was in the full tracklog--the dropped points do not affect the saved points in any way**. So you should check only the saved points for accuracy, not the interpolation between the points (which is typically straight lines).

** Note that the tracklog points include time as well as lat, lon, and elev. Here is a sample one segment, two point tracklog in GPX format. By convention, the datum is WGS84:
<gpx>
<trk>
<name>sample track</name>
<trkseg>
<trkpt lat="42.420040565" lon="-71.360199759">
<ele>9.269897</ele>
<time>2010-06-15T16:34:57Z</time>
</trkpt>
<trkpt lat="42.420050874" lon="-71.360201184">
<ele>57.816528</ele>
<time>2010-06-15T16:35:00Z</time>
</trkpt>
</trkseg>
</trk>
</gpx>
(You can have multiple trksegs in a trk.)


FWIW, the question about how come the GPS puts my boat on land, my car in the water, me under water, the summit away from the summit mark on the map, etc has been asked and answered many times...

Doug

 

[Paradox]

It seems to me that a better indication of GPS accuracy is reproducability. For a typical "out and back hike" my 60CSx can put my "out" track as much as 40-50 feet away from my "back" track. If I check my Accuracy of GPS reading on my unit it will typically give me a figure about half of that value (12 - 30 feet). This discrepancy does not concern me.

 

[DougPaul]

It seems to me that a better indication of GPS accuracy is reproducability. For a typical "out and back hike" my 60CSx can put my "out" track as much as 40-50 feet away from my "back" track. If I check my Accuracy of GPS reading on my unit it will typically give me a figure about half of that value (12 - 30 feet). This discrepancy does not concern me.

Reproducability is necessary, but not sufficient. For instance, if the GPS always said that you were at the North Pole, the reproducability would be very good, but the accuracy would be terrible (unless, of course, you actually reside at the North Pole... ).

The only way to truly measure the accuracy of a GPS is to use it to measure the location of a surveyed point. (Its location must be known to much higher accuracy than the GPS accuracy.) You have to take a measurement every 30 sec or so for ~30 days to average across satellite locations and ionospheric conditions. This gives you ~86K measurements to estimate the average error and standard deviation and/or estimate a distribution. (The accuracy is usually stated as x% probability of being within y meters of the true location. If the distribution is assumed to be a 2D Gaussian, then the standard deviation can be converted to an x% prob for y meters with a bit of simple math.)

That said, it is much easier to measure the reproducability than the accuracy. One way is to keep the GPS stationary, take a bunch of measurements spread out over time, and compute the average and standard deviation. (I have several 12 hr tracks of this form.)

Comparing incoming track to the outgoing track from an out-and-back trip is also a good method--you may recall that it is what we did on the GPS bakeoff. (http://www.vftt.org/forums/showthread.php?t=14406 http://www.vftt.org/forums/showthread.php?t=15025) Since the goal of the bakeoffs was to compare GPSes, we recorded simultaneous tracks of a hike to remove the effects of satellite location, ionosphere, and skyview from the comparison. I have used this same paridigm in some more recent comparisons.


The indicated accuracy (actually EPE=estimated position error) given by the GPS makes assumptions which are often (usually?) not met. The probabalistic error will not be better than the EPE, but is often (usually?) worse. Thus the number is not rigorously believable. Smaller EPE is better than larger, but it is no guarantee of better accuracy.

Doug