As you can easily imagine, different station-measurement-pairs will produce different predictions for a questioned point. Therefore, the final quality at this point is determined by averaging over these different predictions, depending on the distance of the measuring point who made them. In this process, a function is needed that will return the weight of a measurement in this average, depending on its distance. For this purpose, an exponential function was chosen.
This leaves you with another parameter to set, which is called "DistanceCoefficient". The higher this value, the softer the exponential function will fall, thus causing measurements that are farther away to have an effect on the averaging. The lower the value, the steeper the function falls, thus only very close measurements have an effect on the questioned point (in extreme, only the closest measurement has an effect, producing rather funny-looking splits in the image).
A recommended setting in a well-calibrated map is 
(
).