There are several modes that allow the location of the Base Station to be set or surveyed, and the way these are used determines the type of positional accuracy and stability that is achieved. The amount of drift that is seen in the position of a roving receiver using the Base Station corrections is relative to the DGPS mode that is being used. For example, if RTCM 40 cm 95 % CEP is being used, then the rover position will drift no more than 40 cm from its relative position 95 % of the time.
Likewise for RTK 2 cm 95 % CEP, this will not drift more than 2 cm from its relative position 95 % of the time. The important word to note in this statement is relative. The Base Station provides drift corrections to the rover based on the positional information that has been set; so if the position that has been set in the Base Station is offset from the actual absolute position on earth then this offset will also be present in the rover position.
There are two positional plots below to illustrate this point below. The graph on the left shows a drive around a marked circle at different times of the day using 20 cm 95 % CEP DGPS corrections from a Base Station with a 24 hour averaged position that has been stored. You can see that the position of each circuit is in close proximity and that the positional error is due only to the 20 cm drift in relative accuracy.
The plot on the right shows the same drive logged immediately after each of the first plots, but with the Base Station position set using ‘SET TO CURRENT’ before each circuit. While the relative positional data for each individual circuit is stable to a 20 cm radius, a positional shift of nearly 2 m can be seen between the green and red plots. So in this case where a geographical feature is being referenced and data is being collected during different test sessions (even without the Base Station antenna being moved), this ‘SET TO CURRENT’ method is not suitable.
However, it should be noted that if the ‘SET TO CURRENT’ location had been stored and reused for each measurement; the results would have looked almost identical to those of the averaged position. This makes ‘SET TO CURRENT’ a perfectly valid method for comparing data with relative positioning from different test sessions. The difference would have shown when comparing the absolute geographical positions of the two sets of data.
The ‘SET TO CURRENT’ data would have had a positional offset as compared to the ‘SET TO AVERAGE’ data, and this is the important thing to keep in mind as it can cause problems if forgotten. For example, If the extents of a circuit were mapped not using the same location corrections as those being used during testing, the lines taken by the vehicle and the extents of the circuit could be misaligned by several metres.
The same effect of offsetting would be seen if an averaged position was used but the Base Station GPS antenna was not placed in the same position as the original average was taken. Any error in locating the antenna would be translated into the correction at the rover. It should also be noted that any error in a manually set location as compared to the actual position would result in the same type of error in the correction.
A similar effect to the right hand plot would be produced if stored location was used but without correctly locating the GPS antenna in the same position for each use. This is why it is important to mark the physical location carefully if you intend to reuse it.