OpenStreetMap

How can I get accurate locations in a forest ?

Posted by adaviel on 18 October 2014 in English (English).

I’m trying to improve the accuracy of the mapped features in Smoke Bluffs Park in Squamish. This is a very popular rock-climbing area with many documented and bolted routes, so I’d like to be able to use a GPS to find a particular climb. Although the bluffs themselves are mostly exposed, most of the approach trails to top and bottom are in forest, and don’t show up on the common satellite photo sites. I was using 3 different GPS sets here - a Garmin GLO bluetooth unit talking to a Nokia N810 tablet, an Asus ME173X tablet with builtin GPS, and a Nokia E71 phone with builtin GPS. I had assumed the GLO (GPS+GLONASS) would be most accurate - it is sold as a GPS, tracks more satellites and has a faster update rate. But when I checked the tracks recorded on October 11th, there are places where “going” and “returning” traces diverge, walking the same trail. I returned on October 13th to check some features, with 3 GPS sets running, and got 3 different traces. In some cases these are far enough apart that I’m no longer sure which trail I was on at the time - the trails are in some cases quite close together. At one point I tried leaving the GLO at a fixed location, and walking twice around a loop with the Asus in order to later subtract the traces and get DGPS accuracy, but this was not very successful - the corrected trace was little better than the original. I’m not sure if I would get better results from two identical units, e.g. two GLOs.

Has anyone any suggestions ? Would a second GLO used for DGPS post-process be better ? screenshot

Location: Northridge, Squamish, Squamish-Lillooet Regional District, British Columbia, V8B 0X5, Canada

Comment from Rps333 on 18 October 2014 at 20:31

Try Stava Slide. It takes millions of track to determine the mean. Lots of fun to use as well.

Your lucky tons of tracks in that area. http://labs.strava.com/routing-errors/#250/14/-123.12528/49.74941

Comment from aseerel4c26 on 18 October 2014 at 23:41

it may also help to filter for HDOP accuracy to only record/use the good values. And set a high sampling rate of e.g. 1 second.

Comment from SimonPoole on 19 October 2014 at 08:14

Getting good tracks in high foilage areas is difficult to very difficult and even professional grade equiment will have problems. The previous comments have already pointed I number of things out, a couple of more from me:

  • avoid wet weather (just makes the problem worse)
  • depending on the type of forest and where you live, surveying the tracks outside of summer/spring (aka winter, late fall, early spring) will help
  • using a ballon(s) may help

Comment from alexz3 on 19 October 2014 at 10:33

hi from Vancouver.

I mountain biked in Squamish a couple of times.

Large trees will make your maps fuzzy. From a radio perspective, the trees will absorb and diffuse the signal, in this case, averaging multiple data points will help a bit. Ie, for Jack’s trail, you can use the center of mass of all the tracks that show up in the tracks display.

When you’re near a cliff or in an urban canyon, signals will reflect off of a surface, so sometimes your gps will will cause a gps track to diverge on the second tracing of a path. in the divergent case, you have to discard one of the tracks instead of ‘averaging’ it. Usually, the track which is ‘into’ the mountain will be the incorrect track. Visually, you can picture the mountain as being a mirrored surface. Your gps will see some satellite signals directly, and some signals will be reflected.

The higher end GPS units will have a good antenna to use the first signal it sees, and treat subsequent signals as an echo, or reflected signal.

Comment from alexz3 on 19 October 2014 at 10:49

also, look into the bc-mosaic image set for bc. It has some better imagery than the Bing images. http://www.openstreetmap.org/edit#map=17/49.70559/-123.13594

in some cases, you may be better off using walking papers http://www.walking-papers.org/ for mapping instead of a gps.

Garmin suggests going to the same spot at different times, so you have a different satellite constellation for each measurement.

http://garmin.blogs.com/softwareupdates/2009/04/waypoint-averaging.html#.VEOXJlS9-6k

Comment from SK53 on 19 October 2014 at 12:54

This is a great little project. Only stopped at Squamish for lunch & a beer so don’t really know what the forest trails are like.

I think as others have said you are pushing the limits of what consumer-grade GPS might deliver. Here’s a local open space where I regularly get 4 m accuracy on my etrex display (also used by the local geo-specialists to test their fancy pro-GPSes), you can see the variability of the tracks ((1) from OSM; (2) from Strava)

NUJC Playing Field with GPS traces

NUJC Playing Field with Strava Heatmap

I think part of what you should be doing as the next steps are detailed micro-mapping of features along the trails. I don’t know exactly what things will be useful, but anything which helps refine precise location and most importantly gets the relative topology of everything correct. In your situation I’d probably try to grab pictures along the trails: Mapillary provides a good way to review such image sequences.

Comment from Vincent de Phily on 20 October 2014 at 10:36

Just do your surveying in two steps : the first in summer while carrying a flame-thrower, the second a few days later while carrying your GPS. Problem solved :p

Comment from Nakaner on 20 October 2014 at 11:58

If you want to do DGPS, you need to recievers which output raw data. One receiver is mounted at a position with known, accurate coordinates, the second receiver is the receiver you carry with you. If you want to know your position in real time, you have to have a data connection between the to receivers (e.g. via Internet over cell phone networks) and the basis has to send its data via a NTRIP server. That’s why post-processing is easier.

Comment from Alan Trick on 20 October 2014 at 22:30

If I remember correctly, It’s mostly deciduous trees in the Smoke Bluffs, so SimonPoole’s suggestion of going in the winter is probably good. December-February is probably your window for minimal foliage.

On fairly labor intensive solution would be to use a rangefinder and calculate positions based on existing known positions.

Comment from adaviel on 21 October 2014 at 17:49

Thanks for all the comments. This is annoying me now; I thought it would be simple - just record some tracks and plot them. But usually I’m on a motorbike on a logging road, and 100m accuracy is often good enough.

Strava if I understand it relies on many different track uploads, and treats them all equally. Not so helpful here without a lot of volunteers. I can’t get around here on a bicycle, let alone a motorcycle in reasonable time - a couple of trails have sections with a fixed rope… (Well, OK, there is an adjacent area with a trials motorcycle course, but those guys are crazy and I don’t have one right now to map that).

I went and ordered a second GLO to play with anyway, before reading all the comments. I had been to a Trimble technical presentation talking about post-processing for DGPS, but that was years ago when SA was still in force. I’m not sure how to get raw data, or even log satellite data, from the GLO. I have maemo-mapper on my N810, and an Android tablet with Viewranger, but those just log regular GPX tracks with no HDOP. Anyone know any good Android bluetooth apps ?

Yes, I took some photos, but it’s hard for me to compare. I have the Squamish Select climber’s bible, which has lots of photos, so I need to go back, with the book, and positively identify the climbs while collecting and averaging GPS data. Hoping it doesn’t rain. The tops of some of the bluffs have a fairly good view of the sky.

The forest cover is I think coniferous, away from the wider trails which have some deciduous trees. I don’t suppose anyone would appreciate an “accidental” forest fire.

I’m not sure walking papers would be so helpful. The existing map is partly my fuzzy tracks away from the residential area, and there aren’t any good landmarks. The satellite photos show mostly trees.

Rangefinder - don’t have one (well, only an indoor one good for 50 feet). A survey crew would have some trouble - there isn’t line-of-sight anywhere and the trails aren’t straight.

Comment from MapMakinMeyers on 22 October 2014 at 02:32

Strava doesn’t share data back with the larger community; reaching out to them is a joke…

If I were trying to get a accurate GPS trail through heavy foliage I would hike/ bike/ walk/ drive the route several times and average out the data; take a few geotagged pics. Line up with some imagery from EarthExplorer (if in US) - and make it yourself. Sometimes emailing county gis departments or park departments for baseline data doesnt hurt.

Cheers!

Comment from dieterdreist on 24 October 2014 at 15:53

IMHO of your set of GPS-enabled devices the only device to use is the GLO, forget about the phone and the tablet. Additionally to the good suggestions like avoiding wet foilage and possibly foilage at all (winter), setting the recording rate to 1 second, I suggest to be careful where you put the receiver. Put it as high as possible and don’t create body shadow on it, e.g. put it on top of your hat. You should definitely not put it into any kind of pocket or bag.

Comment from Alan Bragg on 30 October 2014 at 20:06

Dieterdreist’s wear is a good idea. I’ve been trying to carry my iphone running the Galileo APP on my hat without much success. The Glo is much lighter and I could easily velcro it to my hat. The Glo update rate is 10 Hz and it’s not possible to set the recording rate to 1 second.

Comment from Warin61 on 17 December 2014 at 08:11

As you have found .. the GPS does not perform well in these situations. So others GPS will also have similar problems. This is when mapping become actually more important. The map should represent what is on the ground so that a person there can gather clues as to what track to take. So I’d not be too worried by ‘accuracy’ in location but much more concerned with ‘accurate’ representation of the features so it can be used as a map without GPS.

The GPS reported location should include something on the ‘error’ .. use that as a guide as to how far out the recorded tracks could be. The mapping you put into OSM should take the GPS tracks as a guide ..


My smart phone GPS tracks .. well it does the Glonass thing too .. but it is not as good in tree cover as my much older non Glonass GPS .. its an obsolete Garmin 60Cx … umm 7 years old. The smart phone looses satellites where the 60Cx does not. If you have an older dedicated GPS that way work better than the newer stuff.


Login to leave a comment