Errors that Can Affect Tracking Device Tracking Data

The following are some of the main errors that can potentially affect data acquired from GPS sensors (points 1 to 5), and that can be classified as GPS location bias, i.e. due to a malfunctioning of the GPS sensor that generates locations with low accuracy (points 6 to 9):

1. Missing records. This means that no information (not even the acquisition time) has been received from the sensor, although it was planned by the acquisition schedule.

2. Records with missing coordinates. In this case, there is a electronic tracking devices failure probably due to bad GPS coverage or canopy closure. In this case, the information on acquisition time is still valid, even if no coordinates are provided. This corresponds to ‘fix rate’ error.

3. Multiple records with the same acquisition time. This has no physical meaning and is a clear error. The main problem here is to decide which record (if any) is correct.

4. Records that contain different values when acquired using different data transfer procedures (e.g. direct download from the sensor through a cable vs. data transmission through the GSM network).

5. Records erroneously attributed to an animal because of inexact deployment
information. This case is frequent and is usually due to an imprecise definition
of the deployment time range of the sensor on the animal. A typical result is
locations in the scientist’s office followed by a trajectory along the road to the
point of capture.

6. Records located outside the study area. In this case, coordinates are incorrect
(probably due to malfunctioning of the GPS sensor) and outliers appear very far from the other (valid) locations. This is a special case of impossible movements where the erroneous location is detected even with a simple visual exploration. This can be considered an extreme case of location bias, in terms of accuracy.

7. Records located in impossible places. This might include (depending on species) sea, lakes or otherwise inaccessible places. Again, the error can be attributed to GPS sensor bias.

8. Records that imply impossible movements (e.g. very long displacements, requiring movement at a speed impossible for the species). In this case, some assumptions on the movement model must be made (e.g. maximum speed).

9. Records that imply improbable movements. In this case, although the movement is physically possible according to the threshold defined, the likelihood of the movement is so low that it raises serious doubts about its reliability. Once the location is tagged as suspicious, analysts can decide whether it should be considered in specific analyses.

GPS sensors usually record other ancillary information that can vary according to vendors and models. Detection of errors in the acquisition of these attributes is not treated here. Examples are the number of satellites used to estimate the position, the dilution of precision (DOP), the temperatures as measured by the sensor associated with the GPS and the altitude estimated by the GPS. Temperature is measured close to the body of the animal, while altitude is not measured on the geoid but as the distance from the centre of the earth: thus in both cases the measure is affected by large errors.

More GPS tracking solutions at http://www.jimilab.com/ .