This is part 2 of a 4-part series describing our Oldest GeoKid’s grade 5 science project from a few years back. Part 1 contains useful background information for those not very familiar with how handheld GPS receivers work. Please keep in mind that this project was carried out by a 10 year-old, with a little help from the parents. He used our Garmin MAP60CSx to mark all the coordinates recorded as observations for his experiment.
After completing his background research, it was time to design an experiment to determine which variable most affected the accuracy of our GPSr’s calculation of its location. For his experiment, Oldest GeoKid planned to compare our GPSr’s displayed location – calculated under different variable conditions – to a precisely known location. He selected three Alberta Survey Control Markers as the “truth” locations against which he could compare his results. The coordinates for the survey markers are contained in an Alberta government database and are determined by survey-grade GPS receivers, accurate to within centimeters. These locations are also “findable” locations for one of our all-time favourite geocaches: GC43F3 – Brass Cap Cache.
One of the locations was selected because it is close to our house, which would make it relatively easy for Oldest GeoKid to get out to gather observations under various weather and atmospheric conditions. The other two were selected because they are affected by significant multi-path error and poor satellite geometry.
Once his locations were selected, Oldest GeoKid developed a matrix to record two things: the condition of each variable under consideration and the coordinates calculated by the GPSr. He had a few challenges the first time he went out to record some observations. Here’s an extract from his project report:

“My Dad went with me on December 3 to take the first set of readings at BCP211. We had some problems – we ran out of memory on the GPSr. It was really cold and the car battery died, so we had to get a boost before we could go home.
“I found that the data collection sheet I first set up didn’t work very well, so changed it for any other days’ readings.
“What went OK was I took about 20 readings for each set of conditions. I made note of the conditions in effect for each set of readings. It was sunny in the morning, when DOP was low and it was cloudy in the afternoon, when DOP was high. The highest DOP was around 3 and the lowest DOP was around 2.
“The GPS readings were downloaded into the computer through a USB connection. Once the readings were in the mapping software, they were copied into an Excel spreadsheet for analysis.
“For each set of readings, Excel was used to calculate an average reading. The average readings were compared to see if there were any strong patterns. I couldn’t find any patterns in the results.”
Here are some of his comments regarding subsequent observations:

“Other sets of readings at BCP211 were collected on Dec 7, Dec 10, Dec 27 and Dec 31. DOP ranged from a low 1.8 to a high of 3.6. Weather conditions ranged from sunny to cloudy to a blizzard. There was no rain during the last half of November or the month of December, so I didn’t get any readings in the rain. In total, I have 1400 readings from this location.

“All the readings at BCP175 were taken on January 9 in cloudy weather conditions. DOP ranged from a high of 3.5 to a low of 2.0. A total of 320 readings were taken at this location.
“The readings at BCP082 were all taken on January 12 in sunny weather conditions. DOP ranged from a high of 3.3 to a low of 2.0. A total of 240 readings were taken at this location.”
DATA ANALYSIS:
Analyzing almost 2000 data points took considerable time, as well as parental assistance. Here’s an overview of the process; if you’d like more detail please leave a comment and we’ll do our best to respond.
- For each observation, Excel was used to calculate the variance, in meters, between the observed latitude and longitude and the “known” latitude and longitude (i.e. the published location for the Alberta Survey Control Marker). These calculations took into consideration the fact that the farther north you go, the smaller the distance covered by each degree of longitude.
- The data sort feature in Excel was used to group observations taken under certain variable states and then Excel’s graphing features were used to help identify patterns.
- Recorded coordinates were grouped into the following categories:
- those within 2.5 meters (8 feet) of the true coordinates;
- those between 2.5 and 5.0 meters (8 to 16 feet) of the true coordinates; and
- those more than 5 meters (16 feet) from the true coordinates.
SUMMARY OF RESULTS:
- At BCP211 20% of the marked coordinates were > 5 meters away from the true location while 34% of the marked coordinates were within 2.5 meters of the true location.
- At BCP175 76% of the marked coordinates were more than 5 meters away from the true location.
- At BCP082, all the marked coordinates were more than 5 meters away from the true location.
Since BCP175 and BCP082 are both located in downtown Calgary, near many tall buildings, this immediately showed that multi-path error can lead to very inaccurate coordinates calculations by handheld GPSrs.
The rest of the findings from his analysis of nearly 2000 data points will be shared in part 3. Watch for it in the next few days.
Pingback: Handheld GPS Receivers: Results From Experiment to Assess Which Variable Most Affects Coordinates Accuracy | Out and About with the GeoKs