How accurate can a smartphone GPS be?

Is a smartphone now good enough to replace a dedicated GPS receiver in the field?

These two questions have come up numerous times this past semester. Saratoga Springs is in the process of working on a tree survey and the GIS Center has been doing some minor supporting work. Since most people don’t own a dedicated GPS unit, tree surveyors were curious to know if a smartphone would be sufficient. This piqued our interest, as it could make our work easier if students could use their phones in the North Woods or around campus to collect location data.
To answer this we conducted a simple experiment. To begin, we used two Garmin GPSmap60CSx, a Verizon iPhone4S, and an AT&T iPhone 4S. Walking around a loop on campus, we recorded waypoints at known locations that were visible in satellite imagery and aerial photos. We loaded these waypoints into Arc Map, registered the actual locations, and then measured the relative error of the GPS units and the iPhones. Here are the waypoints, overlaid on the imagery layer of campus. The canonical location is in blue, the GPS units are in red, and the iPhones are in green.

The short analysis is that iPhones are still worse than a dedicated GPS, but not that much worse. For general positions, an iPhone is perfectly sufficient if you are fine with single-digit meter errors. If you want to be within a meter or two, a GPS is the best bet

The following table contains the distance from the canonical location to the GPS units and the iPhones.

  Site #     iPhone 4S (AT&T)     iPhone 4S (Verizon)     GPS Unit 1     GPS Unit 2  
    1     18.28     14.07     0.24     2.50
    2     6.06     5.80     0.83     0.81
    3     6.22     9.72     2.71     3.17
    4     2.74     5.43     3.05     3.31
    5     6.11     5.19     2.45     0.75
    6     6.44     8.78     4.43     2.84
    7     10.75     5.19     3.15     1.00
    8     13.88     8.40     2.06     0.84
    9     10.64     13.28     3.31     3.51
Avg. w/out 1       7.86     8.43     2.88     1.69

To get an average that represented the majority of the sample, we removed the first point from the table. This data point was collected right after moving outside from indoors, and my best guess is that the location information that the iPhones received was from WiFi triangulation and did not have information from more reliable sources(cell towers and GPS signals). Removing the outlier resulted in an average that better represented the other eight results.
I believe that we have reached a point where smartphone location is good enough for projects that are “campus-sized”. When points are separated by large distances, the position provided by iPhones is accurate enough to discriminate between locations. If precision is paramount or points are close together a GPS unit is still required to get significant results.