See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/374229558
Survey Verification of In-Woods GNSS Accuracies for a Wide Variety of
Receivers
Presentation · September 2023
DOI: 10.13140/RG.2.2.29645.82406
CITATIONS READS
0 45
2 authors, including:
Jacob Strunk
US Forest Service
33 PUBLICATIONS 663 CITATIONS
SEE PROFILE
All content following this page was uploaded by Jacob Strunk on 27 September 2023.
The user has requested enhancement of the downloaded file.
Survey Verification of In-Woods GNSS
Accuracies for a Wide Variety of Receivers
2023 Forest Inventory Workshop
Sept 13, 2023
Kalama, WA
Jacob Strunk Thomas Broch
GNSS, or commonly, GPS
1) Global Navigation Satellite System (GNSS)
2) GNSS: the generic term for satellite-based
positioning
3) Real-time approximately 20m accuracy
*GPS is the American constellation of satellites
Convenient, Portable, Cheap, Everywhere
Amazing Accuracy*
*Possible, but Not Guaranteed
The Mythology of GNSS
1) It is hard to measure GNSS accuracy
2) Therefore people “feel” their equipment is
good, because
A. Vendor told me so
B. I spent a lot of money
C. The software says it’s good
D. Someone I trust told me it’s good
E. I already own it
F. I have been using it for a long time
G. Looks bad if I bought some garbage
In-Woods GNSS is Difficult
1) Bunch of “stuff” in the way
2) Can absorb / reflect signals
3) Advertised accuracies not relevant
Forestry is Evidence Based …
1) How good is new gadget?
A. Measure with gadget (e.g., Haglof ultrasonic DME)
B. Measure with known device (tape)
C. Compare
i. Known distance
ii. Gadget
2) Or laser range finder
3) Or prism plots
4) Or GNSS …
How Good is GNSS?
1) Take positions with Geodetic Survey (total station)
A. “True”
B. ≈2 inches error
How Good is GNSS?
1) Take positions with Geodetic Survey (total station)
A. “True”
B. ≈2 inches error
2) Take position with GNSS receiver
3) Compare True versus GNSS
A. Compute error in feet
How Good is GNSS
1) Conceptually easy!!
2) Practically, very difficult
A. Requires survey traverse in the forest …
B. Collect a bunch of GNSS data
C. Analyze correctly
Our Study
1) Forestry Sciences Lab in Olympia, WA
A. Survey Traverse put in by WA DNR
B. ~ 20 survey points
C. 150’ – 180’ closed canopy forest
2) Test variety of budget receivers
A. Autonomous positions
B. Short occupations
3) A few better receivers
A. Short / long occupations
B. Autonomous / post-processed / RTK
Receiver L1 L2 L5 Constellations RTK/PPK cost
Emlid RS2 x x GPS, GLO, Gal, Beid yes $2,600
EOS Arrow 100 x GPS, GLO, Gal, Beid RTX possible $3,000
Garmin 64s x GPS, GLO no $300
Garmin 65 x x GPS, GLO, Gal no $350
Garmin 65s x x GPS, GLO, Gal no $400
Garmin GPSMAP 60CSx x GPS no $200
Google Pixel 4 x x GPS, GLO, Gal, Beid no $800
iPhone 11 x GPS, GLO, Gal no $500
Javad Triumph 2 x x GPS, GLO, Gal, Beid yes $3,500
Juniper Geode GNS2 x GPS, GLO, Gal, Beid no $2,500
Motorola Edge+ 2022 x GPS, GLO, Gal, Beid RTX possible $850
Samsung Galaxy S20 FE x GPS, GLO, Gal, Beid no $600
Samsung Galaxy S22 Ultra x x GPS, GLO, Gal, Beid no $1,200
Trimble Geo 7x x x GPS, GLO, Gal, Beid yes $9,000
Trimble R1 x GPS, GLO, Gal, Beid RTX possible $1,500
Point Averaging
More is better
Single point may have large errors
Outlier detection
Averaging reduces magnitude of
errors
CLT – with caveats …
How to interpret:
GNSS errors for individual survey
points…
Worse
How to interpret: Bad
Results – Short Occupations
True Versus GNSS (GPS)
First Result:
Mostly bad!
30 – 60 seconds:
1. 13 ft average
error
2. 80 ft errors!!
3. 20+ ft errors
common
First Result:
Mostly bad!
30 – 60 seconds:
1. Most average
errors are 10 ft
or worse
Second Result:
Some are REALLY awful!
Garmin GPSMAP 60CSx
Average error: 24 ft
Third Result:
Some were OK ?
Arrow: 6 – 12 ft
Javad: 4 – 9 ft
Garmin 65: 6 - 9 ft
Fourth Result:
Open better than FOREST OPEN
forest
Forest: 13 ft
Open: 8 ft
Aside: Differential
correction
1) GNSS on known positions (survey
monuments)
A. Get GNSS errors
B. Errors -> Correct rovers
Differential
correction
Differential Correction:
“meh”
“aut” = autonomous
“proc” = differential correction
In 30-second occupations, only a
couple “corrections” …
Dedicate base may work better
Results – Long Occupations
15 minutes of data
True (Survey) versus GNSS (GPS)
New data (not in previous figures)
Short: 30 – 60 Seconds Long: 15 minutes
15 mins data:
Much better!!
• Medians better than 10’
• Javad receivers: high-
performance (3.28 ft HRMSE)
• Others: not high-performance
“aut” = autonomous
“proc” = differential correction
Conclusions
1) Collect point averages!
A. Not single points
2) Short occupations imprecise
A. 12 foot or worse typically
B. 30 feet or worse possible
3) High Performance (3.28 feet – 1 meter HRMSE)
A. Need: Survey grade receivers
B. Need: Long occupations (15 min +)
C. Need: Tripod mounted
D. Differential correction?
Emlid RS2 / Trimble 7x not “High performance”
4)Avoid offsets
A. Human error worse than GPS error
B. Correct azimuth (declination)
C.From versus to
5) RTK, RTX not suited for dense canopy*
A.Use DGNSS/ DGPS instead
*Limited testing here, but other experiences corroborate
*makes sense given that it is RTK ≈ differential correction in real
time without a pc
END
View publication stats
