GPS Precision and Accuracy 

Definitions

Unlike any other radio-telemetry technique, the accuracy and precision of GPS data can be quantified statistically. Accuracy is defined as closeness to true location while precision is defined as the proportion of locations within a pre-defined quantile (See Flash Figure below). The ideal scenario is to have all locations accurate and precise while complex analytical techniques are required to overcome the errors in the other scenarios.

Figure 1 - Graphical representation of precision and accuracy. The intersection of both axes represents true location and the axes are measured in meters.

Setting Precision and Accuracy on BlueSky Telemetry™ Collars

The accuracy of a location and the probability of acquiring a location in BlueSky Telemetry™ GPS collars can be set by the user.   The user can decide whether an horizontal accuracy of < 3m, through to < 40m is required while the probability of acquiring a location in different terrains or landscapes can be programmed into the collar.  Obviously there is a cost: the higher the required accuracy the greater time the processor remains active and the greater the power demands.  The user therefore has to balance improved accuracy with increased power demands. No other collar on the market gives the user such flexibility.

For example, in an agricultural feedlot experiment, the required accuracy may be very high in order to separate out the use of different grazing patch's and an accuracy better than 3m can be set.  In such a scenario, the GPS processor will remain active until the estimated accuracy (determined by the processor itself) is less than the chosen accuracy during setup.  Only then will the location be saved into memory as a "Valid" location.  The opposite scenario may prevail in studies of ranging behaviour of tundra caribou where the accuracy of the maps per se may be no better than 50m.  An horizontal accuracy setting of <40m will therefore be more appropriate and again the GPS processor will remain active until the estimated accuracy (determined by the processor itself) is less than the chosen accuracy during setup.  Only then will the location be saved into memory as a "Valid" location.  

The probability of the collar acquiring a location can also be set by the user.  For example in a wooded landscape where the view to the sky and consequently satellites may be restricted, BlueSky Telemetry™ collars can be set to calculate a location when data from only three satellites are available.  By contrast in an open landscape with full visibility, the collar can be programmed to calculate a location when a minimum of five or more satellites are visible.  The greater the number of satellites used to acquire a location, the greater the accuracy.  However, in the wooded landscape, there is a trade off between accuracy set above and actually obtaining a location which is a decision that the user has to make based on the objectives of their study.

Further Reading

Regardless of data presented by GPS receiver and GPS collar manufacturers, all GPS receivers must be benchmarked to test for instrument error and errors specific to each site before deployment on study animals. For a full description of methodologies, analysis and discussion of techniques that can be used to measure GPS accuracy and precision, please see the papers by Moen et al. 1996; Moen, Pastor, & Cohen 1997; Moen, Pastor, & Cohen 1998; Rempel, Rodgers, & Abraham 1995; Rodgers, Rempel, & Abraham 1996 and Hulbert & French 2001. These papers and many others can be viewed in the database of all radio telemetry papers published since 1980.

View references