GPS Accuracy: Part Two

This is a follow up to my previous post on GPS accuracy. Basically, I returned to the same location --the wading pool in the Roosevelt Baths Quad at Saratoga Spa State Park-- and recorded additional sets of GPS fixes using the Garmin GPSMap 64s and the Garmin eTrex. Image One is a screen capture from Google Earth showing the results. The green markers represent the fixes recorded using the 64s and the red markers were recorded using the eTrex. The red dots represent the actual locations.

Image One: Screen capture from Google Earth. The green markers represent locations recorded by the Garmin GPSMap 64s and the red markers are from the Garmin eTrex Venture HC. The red dots show the points where I stood to record the locations.

My intent, as described in detail in the previous post, is to understand the level of accuracy that can be obtained using inexpensive (less than $300) GPS devices and to determine the factors that affect that accuracy. From that perspective, the average accuracy of the fixes recorded on this visit to the Spa State Park site are similar to those obtained on the previous visit. Slightly better overall, but not substantially different.

The majority of the fixes recorded by both units where within ten feet of the actual and both recorded some fixes within five feet of the actual. Using the criteria discussed in the previous post these are excellent results. My methodology is to record a fix with each unit, walk to the next location and repeat the process. Starting at the upper left I proceeded to the lower left, lower right, and upper right corners of the pool. The green marker labeled 001 was the first fix recorded using the 64s and the red marker labeled 001 was obtained a few seconds later using the eTrex. I made three circuits so there are three markers for each unit at each corner.

On the third circuit I used the location averaging feature available on both devices. When using averaging the GPS device grabs a sequence of fixes at the location and averages those fixes to produce a single fix that should be more accurate. I have not seen a description of the algorithm these devices use to do the averaging, but an educated guess is that the units temporarily store fixes obtained using different sub-sets of the full set of satellites from which the unit is receiving signals. Comparing these trial fixes would allow the unit to drop signals that don't fit well with the others. See the previous post for a discussion of conditions that can reduce GPS accuracy.

To make it easier to see how these results compare I grabbed an additional screen capture showing a closer view of the lower left corner of the pool (the second stop on the circuit). And this brings me right back to the title of the first post; GPS accuracy is not a simple matter. As you can see, the distance from the Green marker 010 to the actual is 1.3 feet as measured using the Google Earth measurement tool. That fix, obtained on the 3rd circuit of the pool using the averaging feature of the 64s, was the most accurate fix obtained by either unit. But the 64s also recorded one of the worst fixes of the day at that same location. The green 002 marker, recorded by the 64s on the first circuit of the pool, is 16 feet away from the actual. That's not terrible but it is the 2nd worst result for either device on any of the three circuits. Given that the second time around the 64s recorded a fix about five feet away from the actual (green 006) it was probably the case that the 002 marker incorporated a signal from a satellite with corrections pending, and that that through off the result.

In summary, the accuracy you obtain using inexpensive GPS devices will almost always be better than the 30 foot level that the manufacturers claim for these devices. And it can be much much better with best case results under five feet. But that level of accuracy requires an understanding of how these devices work and what factors reduce accuracy. To obtain these 'best case' results you should turn your device on several minutes before recording the first fix -this allows your GPS device to download the satellite correction data it needs to make proper use of each satellite. Secondly, use the averaging feature of your GPS if it has one. Or, take several readings at the same location a few seconds apart and do the averaging yourself.

Most important, be skeptical. It may be that the one foot accuracy I got on this go round is a fluke. After all, I didn't see that on my first visit to the same site. Leaving me to wonder, what will the next visit show?

Image Two: Screen capture from Google Earth. The 1.3 feet distance from the actual location to the green marker 010 (third circuit) was most accurate fix by either unit. Confouding that, the green 002 marker, also recorded by the 64s, was one of the worst.

GPS Accuracy: Not a simple Matter

GPS accuracy is a tricky topic to discuss. There are a number of factors that limit GPS accuracy and the 'accuracy' of any GPS device varies almost continuously while in use. Even so, a few general rules do apply and I'm going to discuss what I've learned here.

You probably have GPS in your smartphone (and possibly your camera or other devices) but I'm going to focus on hand-held 'recreational' GPS units. This type of GPS receiver can be purchased for under $100 or you might spend as much as $400 (more than that and you are looking at a different class of device). I'm currently working with a Garmin GPSMap 64s that I bought for $230. Over the last 15 years I've purchased three similar devices:

• The GPSMap 64s - purchased in Jan. 2016, referred to throughout as the '64s'
• A Garmin eTrex Venture HC. Six or seven years old, the 'eTrex'. Around $130
• A Magellan eXplorist 210. Fifteen years old. I no longer use this device.

GPS receivers like these are typically described as being accurate to 10 meters (30 feet). But that is the kind of safe generalization that manufactures make to avoid legal disputes. Under favorable conditions these units common fix locations within 10 to 15 of the actual location. Under ideal conditions these units can provide coordinates that are within a few feet of the exact location. If an even higher level of accuracy is needed you can purchase GPS receivers that come with very sensitive external antennas and that make use of ground-based correction signals and post processing of the raw GPS data. At the highest end these unit achieve sub-centimeter accuracy (better than half an inch). These units currently cost thousands of dollars.

But how do you judge the accuracy of locations saved using your GPS receiver? One way is to upload the 'marks' or 'waypoints' captured by your GPS device to a computer for display on top of a map or geographically referenced imagery. I do this using Google Earth. To judge accuracy at the level of a few feet you'll need to record locations at a place that you can identify precisely on a map. The place that I'm using for these tests is seen in Image One. This satellite view from Google Earth shows a complex of buildings located in the Saratoga Spa State Park near my home. The square in the center of the image is a wading pool (it was empty when this satellite image was made). Image Two shows the wading pool with the map zoomed in for a closer look. In Image Two I added a marker place precisely on one corner of the wading pool. The properties dialog then shows the geographic coordinates for that location.

Image One: Building complex in Saratoga Spa State Park. The square shape in the center of the image is a wading pool. For scale, the wading pool is 93 feet on a side. Image captured from Google Earth.

Image Two: The wading pool again, this time zoomed in for a closer view. A marker was added and the properties dialog shows the geographic coordinates of the marker (in the Decimal Degrees format). Coordinates captured using GPS can be compared with a known location such as the corner of the pool to judge accuracy.  Image captured from Google Earth.

As mentioned, using GPS to record coordinates that are less than 5 feet from the actual location typically requires the use of some type of correction technology. Correction is needed because various factors, some of which vary locally, interact in ways that reduce GPS accuracy. GPS satellites reside in geo-synchronous orbits --the satellites orbit in a constant position relative to the surface of the earth-- and the accuracy of GPS hinges on knowing the precise orbital position of each satellite. But those orbits are constantly perturbed by outside forces and regular correction is needed to compensate for this. This type of correction is built into the system but it is not instantaneous. At any given point in time some satellites are not exactly where the system thinks they are and this reduces accuracy. Which is one reason why more is better when it comes to the number of satellites your GPS receiver can use. Signals from additional satellites can be used to average out some types of errors in the system.

To get really accurate locations your device needs to make use of signals from ground-based stations. These stations broadcast signals that your GPS receiver can incorporate into its' calculations and these stations don't move. The location of the ground based stations is known with great precision. Some consumer GPS receivers, including the Garmin GPSMap 64s, can use the WAAS ground correction network but I'm not clear on when this is used and the effect that it has on the accuracy of locations recorded using the 64s, The underlying problem with ground-based correction is that you need to be within the range and coverage area of a transmitter. Currently this is common in urban areas and uncommon away from cities. [Note: I've really over-simplified this, follow the link above for more detailed information]

Really good accuracy also requires the use of a highly sensitive antenna and this is probably the most significant factor in limiting best case accuracy in low cost GPS receivers. This is especially true when using GPS in the dense forests of the Adirondacks. Your GPS receiver must receive signals from at least four satellites (three provides a rudimentary fix) and these signals can be blocked by a dense tree canopy as well as by mountains or buildings. Intriguingly, the GPSMap 64s has a port that allows for the use of an external antenna. I have not yet begun to explore what that might mean in terms of the best accuracy level this receiver can achieve. 

But, without further ado, let's look at some comparisons. I recently did an informal comparison of the accuracy of locations recorded using a Garmin GPSMap 64s compared results obtained using the older Garmin eTrex. I marked the same locations using the GPS in my smartphone to provide an additional set of comparisons. Image 3 shows the results.

Image Three:

I recorded the location of each of the four corners of the wading pool shown in the images above using each device. Since the corners have indents I marked the exact locations with the red circles to show where I took the measurements. Locations recorded using the 64s are represented by the green markers (note that the marker points are the at the recorded location). The locations recorded using the eTrex are represented by the purple markers. And the locations recorded using my phone are represented by the pink markers. As you can see, the 64s and eTrex achieved similar levels of accuracy. The eTrex was actually the winner; the mark closest to the physical location was recorded by the eTrex and that mark is about five feet away from the map image location. However, the 64s was more consistent and that also could be used to decide which device is "more accurate." This was a bad day for the phone. I usually get decent results using the GPS in my phone but on this day at that place it was all over the map. One mark, labeled Phone-3, seen on the left side in the image, was off by almost 200 feet. It should have been around the corner of the wading pool at lower right.

The extra mark at the upper right, labeled "E4b" was recorded by the eTrex and I included it to support my contention that the 64s was more consistent. I actually recorded two marks on each device; standing in the same spot. These marks were generally very close to each other but in this one case the first mark recorded by the eTrex was 25 feet away. Inconsistent.

A few conclusions can be drawn from this test. For starters, while the place and conditions at the time I recorded the data were favorable, the results are not particularly impressive. Both of the Garmin receivers marked the locations at points within ten feet of the actual, but only a couple were with five feet. I plan to return at a different time of day to repeat the test to see how much the results vary. Different levels of accuracy are to be expected because at different times your receiver will obtain GPS signals from different satellites. That the eTrex unit produced the closest marks could actually be an accident. I've been comparing the fixes obtained using both devices informally for over a month and in nearly every case the 64s has performed better. As the title of this post says, GPS accuracy is not a simple matter.

Image Four

Recording fixed locations under controlled conditions is interesting, but my main use for these devices is to record tracks while hiking in the Adirondacks. Image four shows a portion of a track recorded by both devices on a recent outing. The light blue line represents the track recorded by the 64s and the purple line is from the eTrex. The 64s track is more accurate and more precise. This comparison is interesting because it removes bias from the comparison (More on that in a future post).

My contention that the the tracks recorded by the 64s are more accurate comes from knowing the details of this outing. In the lower part of the image, the four distinct lines should overlap almost perfectly because that portion of the route follows a path that I walked along on the inbound portion of the hike and then again, in the opposite direction, on my return. My actual route might have varied by a few feet in places but those lines should basically overlap. In places there could be differences of a few feet between the inbound and outbound tracks but the two tracks should nearly overlap.

Measuring the gap in Google Earth I find that the it averages about 15 feet for the track recorded by the 64s and by 40 feet for the track recorded by the eTrex. In this context the 64s was significantly more accurate.