Why Bother Directly Estimating Systematic Error?

If system (3) produces track-free maps, the skeptic might wonder, why even bother directly estimating the systematic error in the data? First, as mentioned earlier, system (3) leads to a loss of resolution in the final map. Second, in many cases we may have prior information about the distribution and magnitude of the systematic error, which we could then include as an ``inverse model covariance'' (regularization operator) in system (4).

By directly estimating and subtracting systematic errors, we have more faith that the final map is an accurate representation of the true quantity. While the authors of the previously mentioned papers on Galilee and Madagascar were more interested in resolving the topographical features than the value of the underlying field, in many applications, the field itself is most important. Furthermore, when the systematic error is explainable by physical or other phenomena, we want to have control in its estimation.