Interpolating datasets with trends: A modified median polish approach

This article investigates an alternative method to deal with data sets in the presence of trends. Median polish kriging (MPK) was introduced as an alternative solution to universal kriging or intrinsic random functions of order k (IRF-k) for estimation in the presence of trends. The maps obtained using the original MPK algorithm show banding artefacts which do not appear in the reference data set. A modified version of MPK was introduced to attempt to remove the banding artefacts. The results confirm the improvement in quality of estimate using the modified version of MPK (called MPKm), which takes into account the problems of clustered samples and boundary effect associated with the re-addition of the trend along bands. The variation introduced in the median polish algorithm proved to be satisfactory in eliminating the artefacts.